Fructose-1,6-bisphosphate (FBP) sometimes provides substantial cerebral protection during hypoxia or ischemia. 31P/1H nuclear magnetic resonance spectroscopy of cerebrocortical slices was used to study the effects of FBP on hypoxia-induced metabolic changes. In addition, 13C-labeled glucose was administered and 13C nuclear magnetic resonance spectroscopy was used to search for FBP-induced modulations in glycolysis and the pentose-phosphate pathway.
In each experiment, 80 slices (350 microm) obtained from ten 7-day-old Sprague-Dawley rat litter mates were placed together in a 20-mm nuclear magnetic resonance tube, perfused, and subjected to 30 min of hypoxia (PO2 < 3 mmHg). Nine experiments were performed, with n = 3 in each of three groups: (1) no treatment with FBP; (2) 60 min of prehypoxia treatment with FBP (2 mM); and (3) 60 min of posthypoxia treatment with FBP (2 mM). 31P/1H Interleaved nuclear magnetic resonance spectra at 4.7 T provided average adenosine triphosphate, intracellular pH, and lactate. Cresyl violet stains of random slices taken at predetermined time points were studied histologically. Some experiments had [2-13C]glucose in the perfusate. Slices from these studies were frozen for perchloric acid extraction of intracellular metabolites and studied with high-resolution 13C nuclear magnetic resonance spectroscopy at 11.75 T.
With no pretreatment with FBP, hypoxia caused an approximately 50% loss of adenosine triphosphate, an approximately 700% increase in lactate, and a decrease in intracellular pH to approximately 6.4. Pretreatment with FBP resulted in no detectable loss of adenosine triphosphate, no increase in lactate, and minimal morphologic changes but did not alter decreases in intracellular pH. 13C Nuclear magnetic resonance spectra of extracted metabolites showed that pretreatment caused accumulation of [1-13C]fructose-6-phosphate, an early pentose-phosphate pathway metabolite. Posthypoxic treatment with FBP had no effects compared with no treatment.
During severe hypoxia, pretreatment with FBP completely preserves adenosine triphosphate and almost completely preserves cell morphology but does not alter hypoxia-induced decreases in intracellular pH. Pretreatment also substantially augments the flux of glucose into the pentose-phosphate pathway.
PREVIOUS in vivo and in vitro animal studies found that pretreatment with fructose-1,6-bisphosphate (FBP) provides protection of hemodynamic and vital organs during shock, hypoxia, and ischemia. [1–5]Subsequent studies that focused on cerebral hypoxia and ischemia found that pretreatment with FBP reduced central nervous system injury. [4–9]When FBP is cerebroprotective during oxygen deprivation, adenosine triphosphate (ATP) is preserved and levels of extracellular glutamate are reduced. Such findings have inspired hope that FBP ultimately might serve a role or provide valuable mechanistic insight in the prophylaxis and therapy for human cerebral ischemia. Certain other studies, however, found that FBP fails to provide protection during cerebral ischemia. [11–13]More significantly, there is no comprehensive explanation of the mechanisms by which FBP provides protection. Some have suggested an empirical conclusion-that there is lack of central nervous system protection if:(1) FBP is given after ischemia, or (2) ischemia is “too severe,” or (3) the administered FBP dose is lower than a threshold value, typically 500 mg/kg. [4,11–13]We note two exceptions regarding pretreatment. Kuluz et al. found that postischemic administration of FBP reduced infarct size in a rat middle cerebral occlusion model, and Sola et al. demonstrated that it reduced central nervous system injury in neonatal rats after hypoxia. In both studies, there was a threshold for central nervous system protection.
Because we thought that noninvasive nuclear magnetic resonance (NMR) spectroscopy of respiring brain slices might demonstrate fundamental metabolic differences when FBP is protective, we used noninvasive31P/sup 1 H NMR spectroscopy to study the effects of pretreatment with FBP on hypoxia-induced changes in intracellular ATP, intracellular pH (pHi), and lactate. It has been postulated that FBP might sustain levels of ATP by entering cells and acting as a metabolic substrate that normally requires ATP expenditures. In glycolysis, two ATP molecules are expended in converting glucose to FBP: one by hexokinase, which phosphorylates glucose on cell entry, and another by phosphofructokinase, which phosphorylates fructose-6-phosphate (F6P). Intracellular entry of FBP would spare these ATP costs. FBP might also stimulate the pentose-phosphate pathway (PPP; also called the phosphogluconate pathway), which is known to be present in nerve endings, where it is activated during stresses that require local repair or reduction of oxidative toxins. We therefore also explored the feasibility, of using13C NMR spectroscopy to detect FBP-induced modulations of glycolytic and PPP metabolites after administering13C-labeled glucose. We also performed histologic examinations of 20-micro meter thick Nissl-stained sections cut from slices that had been removed at predetermined times in the protocol.
Materials and Methods
Obtaining slices of high metabolic integrity requires minimizing injury from two sources: postdecapitation ischemia and mechanical trauma during brain removal and slice cutting. To combat the former, we used isoflurane/oxygen anesthesia and mild hypothermia before decapitation. [15–18]To combat the latter we worked quickly, taking only 30 sec from decapitation to placement of four slices in oxygenated medium, and we used a technique established earlier in which only brain surface slices are obtained, thereby assuring that only one side of each slice was traumatized during cutting. Electrophysiologists who use brain slice preparations for microelectrode investigations of single synapses have long known that slice neurons are edematous, metabolically impaired, and electrically dysfunctional in the immediate postdecapitation period but that after 2 h of optimum care, morphology, metabolism, and synaptic function return to normal. [20–22]We also have shown that our31P/sup 1 H NMR spectra from brain slices are identical to in vivo NMR spectra, except for increases in amplitude of the inorganic phosphate (Pi) and lactate peaks. [15–18]The increases in Piand lactate are consistent with the notion of an [nearly =] 50-micro-thick injury layer at the cut surface. We recently confirmed this estimate for the size of the injury layer in c-fos and heat shock protein 70 in situ hybridization studies of slices, in which the size and state of the injury layer was made quantifiable by immunohistochemical stains sensitive to apoptosis and deoxyribonucleic acid damage. Messenger ribonucleic acid for c-fos and heat shock protein 70, are, respectively, markers of neuronal insult (intracellular calcium excess) and injury (intracellular protein denaturation). The [nearly =] 50-micro region that we identified as the mechanically traumatized injury layer was further confirmed as such by showing that decreases in postdecapitation neuronal damage occurred only in that region as a response to immediate postdecapitation treatment with dizocilpine (MK-801). (Neurons in cell culture previously were shown to be protected by dizocilpine from mechanical trauma, i.e., from mechanical streaking of the culture plate. ) We also have observed that perfused slices undergo mild edema immediately after decapitation, that histologic morphology returns to normal within 2 h, and that messenger ribonucleic acid expression of heat shock protein 70 does not occur in slices if hypoxia is avoided. 
Our protocol for obtaining cerebrocortical slices (350 micro thick) was approved by the Committee on Animal Research at University of California, San Francisco. Details of animal protocols,31P/sup 1 H NMR experimental protocols, and artificial cerebrospinal fluid composition its perfusion system have been described previously. [15–18]In each experiment, 80 live cerebrocortical slices (350-micro meter thick and [nearly =] 3.2 g total wet weight) were obtained from 20 7-day-old Sprague-Dawley rats and perfused in a glass NMR tube 20 mm in diameter. Interleaved31P/sup 1 H NMR spectroscopy studies were performed at 4.7 T. Experiments began 120 min after metabolic recovery (defined as the return of the phosphocreatine:ATP ratio to [nearly =] 1.5 after decapitation ischemia. [15–18]Five-minute NMR acquisitions resulted in spectra from which we determined ATP, phosphocreatine, pH sub i and extracellular pH (pHo), and the ratio of intracellular lactate to N-acetyl-aspartate (NAA). In this system, NAA levels in ex vivo1H NMR spectra remain constant for [nearly =] 4 h if the slices are killed and perfusion is continued. In previous studies, we demonstrated that our slices are robust and that31P/sup 1 H NMR spectra each remain constant for > 16 h if hypoxia is not induced, i.e., if slices are perfused only with oxygenated media.
Nine31P/sup 1 H NMR spectroscopy studies of hypoxic slice ensembles were performed. There were three complete studies (n = 3 in each of three groups):(1) no treatment, (2) pretreatment with FBP, and (3) post-treatment with FBP. The FBP concentration, 2 mM, was selected from earlier in vitro studies that found FBP to be protective. [4–6,9]FBP was obtained as D-fructose-1,6-bisphosphate trisodium salt octahydrate ([nearly =] 99% pure; Fluka Chemical Corporation, Ronkonkoma, NY). Artificial cerebrospinal fluid osmolarity, measured with a Wescor vapor pressure osmometer (Logan, UT), was not significantly changed by the addition of 2 mM FBP (304.8 +/- 2.8 vs. 308.2 +/- 1.6 milliosmoles per liter [mOsM], for artificial cerebrospinal fluid alone vs. artificial cerebrospinal fluid + 2 mM FBP). For the FBP pre-treatment group, slices were initially perfused for 60 min with oxygenated, artificial cerebrospinal fluid containing 2 mM FBP. For the no treatment group, no FBP was administered before or after hypoxia. After the 60-min perfusion, 30 min of hypoxia was induced in two steps:(1) 10-min perfusion of artificial cerebrospinal fluid that was made hypoxic via equilibration with 5% CO2/95% N2(with or without 2 mM FBP); and (2) subsequently cessation of perfusate flow for 20 min. The total hypoxic episode was 30 min in each case. Artificial cerebrospinal fluid PO2was always < 3 mmHg, as measured with a polarographic oxygen electrode (Cole-Parmer Model 9071 DO2meter equipped with a “Clark”-type polarographic oxygen electrode, UK). During the 30-min hypoxia episode, a flow of 100% N2was maintained in the gas region of the tissue perfusion chamber (above the fluid). After hypoxia, a 60-min recovery period was accomplished by perfusing the slices with artificial cerebrospinal fluid that had been equilibrated with 95% O2/5% CO2. This artificial cerebrospinal contained no FBP, corresponding to the no treatment and FBP pretreatment groups, whereas the post-hypoxic artificial cerebrospinal fluid contained 2 mM FBP, corresponding to the FBP posttreatment group. Slices were removed for fixation at five different predetermined times in the protocol for sectioning, cresyl violet staining, and histologic examinations. Removal times corresponded to postdecapitation, end of decapitation recovery (or beginning of NMR studies), beginning of hypoxia, end of hypoxia, and end of recovery.
Six high-resolution13C NMR spectroscopy studies of metabolite extracts were performed at 11.75 T in two separate series of hypoxia experiments; n = 3 in each of the two groups (with and without pretreatment with FBP). In each group, 5 mM [2-sup 13 C]glucose (Cambridge Isotope Laboratories, Andover, MA) was administered during the 60 min before hypoxia; otherwise, the protocol was exactly the same as in31P/sup 1 H ex vivo NMR studies. The metabolite extraction procedure is described in the section on13C NMR methods.
Slices taken from the perfusion chamber at different predetermined times were transferred to tubes containing 10% formalin in phosphate-buffered 0.9% saline solution (pH [nearly =] 7). After 24–72 h, the tissues were washed, dehydrated, and embedded in paraffin. Adjacent 10-micro meter-thick sections were cut in the plane of the slices, stained with cresyl violet (Nissl stain), and examined with phase contrast light microscopy (Zeiss Axioskop Model 20, Thornwood, NY). Representative fields of 50–140 adjacent cells, corresponding to areas of [nearly =] 104micro meter2, were photographed after counting shrunken and dead neurons.
Determination of Free, Ionized Ca sup 2+ in the Perfusate
Because of the possibility that FBP might chelate calcium, nine sets of six measurements were made of free Ca2+ in separate experiments. Each measurement was from a 100-micro liter aliquot of medium withdrawn from the bottom of the NMR tube (slice chamber). The nine sets were for three different experimental conditions: n = 3 for FBP-treated slices; n = 3 for untreated slices; and n = 3 for perfusate alone (no slices). The six measurements in each set corresponded to different sampling times:(1) before perfusion with 2 mM FBP;(2) after 30 min of pretreatment with 2 mM FBP (i.e., after half of the pretreatment);(3) after 60 min of pretreatment with 2 mM FBP (i.e., end of pretreatment);(4) after 30 min of hypoxia;(5) after 60 min of reperfusion recovery; and (6) after 120 min of reperfusion recovery. Recordings were obtained using a Ca2+-selective ion electrode (Orion model 93–20, Boston, MA) and a single junction reference electrode (Orion model 90–01) that connected to an Orion model 250A pH/temperature/mV meter (0.1-mV resolution, +/- 0.2 mV accuracy). Before the measurements were taken, a linear Ca2+ calibration (-19.4 -+ 20.2 mV; r2= 0.984) was prepared by measuring voltage differences for Ca2+ standards (0.09–2.9 mM). Readings were taken immediately after aliquot collection. Perfusate concentrations of Ca2+ were determined from the millivolt readings using the aforementioned calibration curve.
Nuclear Magnetic Resonance Methods
Two different types of NMR spectroscopy experiments were performed in this study: in vivo and high-resolution. The basic NMR phenomenon is at the heart of both methods, as is the basic technique: pulsed fourier transform NMR spectroscopy. A brief review of relevant, basic NMR principles is given in Appendix 1, which provides a foundation for Appendix 2, in which specific details are presented regarding NMR parameters in this study.
Data are reported as mean +/- SD. Average relative NMR metabolite values for different time points were compared first with a repeated-measures analysis of variance to determine if relative metabolite values appeared unchanged throughout time, i.e., consistent with the null hypothesis. If the null hypothesis was rejected for a particular metabolite, data for that metabolite at each of the aforementioned time intervals were then subjected to a multiple-comparisons test. For this we used Bonferroni-corrected t tests, as in our previous studies. [15–18]
(Figure 1) shows the time dependence of31P/sup 1 H NMR results. The most striking result in Figure 1is that during hypoxia, pretreatment with FBP results in almost total preservation of ATP and almost no increase in lactate. In the group pretreated with FBP, ATP levels were unchanged throughout the entire period of study (null hypothesis accepted, P > 0.63). Figure 1also reveals that without pretreatment with FBP, hypoxia causes ATP and phosphocreatine to rapidly decrease to levels not detectable with NMR spectroscopy (i.e., less than [nearly =] 0.5 mM). During hypoxia, in the group pretreated with FBP, phosphocreatine levels decreased more slowly than in the no treatment group (P = 0.0001, n = 3), although they became undetectable for both by the end of the hypoxia period (Figure 1(B)).
In slices not pretreated with FBP and in untreated slices, ATP returned to [nearly =] 50% of that in controls (P < 0.0001, n = 3 for both), and phosphocreatine returned to [nearly =] 80% of that in controls (P < 0.0001, n = 3 for both) after 60 min of hyperoxic recovery (Figure 1(A and B)). In all groups, pHidecreased during hypoxia to [nearly =] 6.4 from 7.1 (P < 0.0027, n = 3). Similarly, in all groups, pHireturned to control levels during hyperoxic recovery (Figure 1(C)). The rate of pHidecrease and recovery was the same in all groups. NAA signal amplitudes were never seen to change; however, the lactate:NAA ratio increased from [nearly =] 1.0 to [nearly =] 7.0 in slices not treated with FBP (Figure 1(D)) and in slices posttreated with FBP. In contrast, pretreatment with FBP eliminated any increase in lactate (Figure 1(D)). Statistically, the lactate:NAA ratio was unchanged throughout the protocol (null hypothesis accepted, P > 0.63). (In NMR spectra from cerebrocortical slices, there is a small background lactate signal even in the control group. This is commonly attributed to the [nearly =] 50-micro-thick injury layer at the edge where slices were cut. [20,23])
(Figure 2) shows representative cresyl violet-stained 20-micro-thick transverse sections from slices taken after 60 min of reperfusion. Shrunken nuclei and cell swelling can be found in sections taken from slices not pretreated with FBP. Qualitatively, average slice thickness after 60 min, as seen during light microscopy, was also noticeably greater in untreated and posttreated groups and was consistent with edema. Histologic features in sections from FBP-pretreated slices were not different from those in controls.
As can be seen in Figure 3(A), dynamic range limitations prevented our one-pulse, proton-decoupled13C NMR ex vivo spectra from showing glycolytic and tricarboxylic acid TCA cycle metabolites that might have incorporated the C2 label. The relatively high concentration of labeled glucose (5 mM), compared with micromolar concentrations of metabolic intermediates, requires sophisticated NMR pulsing methods to suppress the large glucose peaks. The dynamic range problem is smaller in13C NMR spectra of perchloric acid (PCA) extracts (Figure 3(B and C)), because the perfusate is not present when obtaining extract spectra and the NMR signals are narrower in extracts. Some contribution from the perfusate might persist in extract spectra, however, because interstitial extracellular fluid is present in slices that are removed and frozen for the PCA extraction process.
Despite problems of dynamic range, Figure 3(B and C) display a striking metabolic feature. A significant signal can be seen at 63.75 ppm in extract spectra of slices taken at the end of hypoxia but only in slices pretreated with FBP. This particular resonance peak is associated with the alpha, beta-[1-sup 13 C] carbon of F6P, shown schematically in Figure 4, and its existence indicates that labeled glucose entered the PPP. [25–27]The [1-sup 13 C]F6P signal was not seen when [2-sup 13 C]glucose was metabolized aerobically, by untreated slices or by FBP-treated slices. Figure 4shows that when [2-sup 13 C]glucose gets metabolized to F6P via glycolysis, the label remains in the C2 position. When [2-sup 13 C]glucose enters the PPP, however, the first step is decarboxylation at C1 of glucose. This causes the13C label, formerly at C2, to become a label at C1. At the end of the 30-min hypoxia period, for slices pretreated with FBP (n = 3), the average ratio of the [1-sup 13 C]F6P signal to that for the beta-anomer of [2-sup 13 C]glucose was 0.09 +/- 0.01. This corresponds to a [1-sup 13 C]F6P concentration of [nearly =] 0.47 +/- 0.03 mM. No detectable increases in lactate were seen in13C NMR spectra of PCA-extracted metabolites.
Measurements of free, ionized calcium concentrations in the perfusate are given in Table 1. Addition of FBP to perfusate obtained from the Cell Culture Facility at University of California, San Francisco caused an [nearly =] 17% decrease in the calcium concentration of the input perfusate. Aliquots of artificial cerebrospinal fluid taken from the NMR tube after no-flow hypoxia (with all slices still in place, before reflow was started), however, showed that hypoxia led to [nearly =] 50–60% decreases in calcium in the surrounding perfusate. As can be seen in Table 1, calcium concentrations in aliquots taken at the end of hypoxia were identical for the FBP pretreatment and no treatment groups.
As in earlier studies that found FBP to be protective, [1–9]acute energy failure in cerebrocortical tissue during rapid-onset hypoxia was substantially ameliorated by FBP pretreatment, as was the rise in tissue lactate concentrations. FBP-induced reductions in lactate production during hypoxia have been observed previously. [4–6,28,29]It is possible that FBP enters cells, causes feedback inhibition of phosphofructokinase, and retards or arrests glycolysis. It is nevertheless unusual that, in slices pretreated with FBP, pHidecreased during hypoxia without an increase in lactate. Where are the hydrogen ions coming from if not from lactate? It is probable that most hydrogen ions come from ATP hydrolysis and that those coming from lactate are a small percentage of the total hydrogen ion pool. This would be consistent with the findings of Hida et al., who showed that intracellular acidosis is not always coupled to increases in intracellular lactate. The simultaneous preservation of ATP and increase in hydrogen ions in our severely hypoxic brain slices remain unexplained, however.
Our13C NMR extract spectra from slices pretreated with FBP provided results that are consistent with our ex vivo1H NMR results. In both studies, there were no lactate increases after hypoxia. Our13C NMR extract spectra results for untreated slices differ, however, from our ex vivo1H NMR results. The administration of [2 sup 13 C] glucose did not result in hypoxic production of [1,213C]lactate and [2,3-sup 13 C]lactate in slices not treated with FBP. In preliminary ex vivo13C NMR studies of naturally abundant13C, we noted small but detectable control levels of naturally abundant13C-lactate (peak no. 6, Figure 3(A)). Naturally abundant13C-lactate increased with hypoxia in a manner consistent with our1H studies (data not shown). One explanation is that we did not provide sufficient time for migration of the13C label. In principle, perfusion with [2-sup 13 C]glucose should result in13C NMR spectra of PCA extracts showing formation of [1,2-sup 13 C]lactate and [2,3-sup 13 C]lactate, with the resonance positions for C1, C2, and C3 being 183, 69, and 20 ppm, respectively. [24–26]Our slices were exposed to13C glucose for only 60 min, however. Studies by others suggest that this might not have been enough time for sufficient quantities of the label to advance to lactate through glycolysis. It is possible that longer exposure to13C-labeled glucose might result in sufficient13C-labeled lactate. Long exposures involve more extensive migrations of the13C-label, however, which sometimes causes difficulty in the interpretation of measurements. Experiments with13C-labeled FBP might be helpful, but the same experimental challenges remain that were present in our13C-labeled glucose studies-overcoming the dynamic range issue and having enough time for sufficient label migration through metabolic pathways.
Because diverse organ systems appear to benefit from pretreatment with FBP during oxygen deprivation, it has been postulated that the primary action of FBP is metabolic and that during oxygen deprivation FBP improves ATP supply versus demand. [4–6]A possible mechanism by which FBP sustains ATP levels is by acting as a metabolic substrate that stimulates the PPP, causing increased formation of triose phosphates. Bader-Goffer et al. previously pointed out that in13C NMR studies C1-labeled glucose was useful for avoiding the PPP, whereas C2-labeled glucose was useful for probing it. Similarly, Ben-Yoseph et al. showed that the use of isotopically substituted D-[1,6-sup 13 C;6,6-sup 2 H2]glucose can be used to measure relative activities of PPP when used in conjunction with microdialysis. It has been demonstrated in vitro that FBP activates the nonoxidative arm of the PPP, which interconverts various sugars without generating NADPH from nicotinamide-adenine dinucleotide phosphate and converts glucose to triose phosphates with no expenditure of ATP. Triose phosphates from the PPP must still go through the TCA cycle, however, ff full ATP production is to occur. Furthermore, during hypoxia, one would expect an increase in triose phosphates to lead to an increase in lactate; therefore, an FBP-induced increase in PPP-related F6P does not entirely explain FBP-induced ATP conservation. Additionally, the PPP is more active in immature brain, and the results, which were obtained in tissue from 7-day-old rats, might be unique to the neonate.
Alternative hypotheses of FBP protection have included calcium chelation, free-radical scavenging, and protease inhibition. [33,35–37]Hassinen et al., suggested that FBP protection in perfused hearts came primarily from extracellular calcium chelation. They also found redox preservation, however, which must have resulted from other mechanisms. In two previous studies using intracellular fluorescent calcium indicators on astrocytes and brain slices, FBP was found to prevent hypoxia-induced increases in extracellular glutamate and intracellular Ca2+, and hypoxia-induced depletion of ATP. [7,10]Although we found that the addition of FBP to the normal perfusate caused an [nearly =] 17% decrease in Ca2+, free calcium in the perfusate at the end of no-flow hypoxia was unchanged by pretretament with FBP. Such Ca2+ measurements are extracellular, however, and changes in intracellular Ca2+ concentrations, which are regulated by mitochondria, need only occur in the nanomolar range to cause deleterious cascades. Our Ca2+ measurements do not make it possible to assert or exclude that FBP protection is explained by reduced calcium uptake.
A protective effect of FBP during hypoxia was exhibited in respiring neonatal brain slices, demonstrating that in vivo physiologic mechanisms, such as increased cerebral blood flow, are unlikely to wholly account for in vivo results. Intracellular mechanisms that permit FBP to be protective during oxygen deprivation remain unknown. It seems likely, however, that improved NMR spectroscopy techniques would be very useful for elucidating metabolic changes during studies of this type.
Appendix 1: Brief Review of Basic Issues in Ex Vivo and High-Resolution NMR Spectroscopy
For both types of spectroscopy, conditions for observing the NMR phenomenon are established in steps. [38–40]First, one applies a strong, constant magnetic field, typically 2–20 T in strength. (For comparison, the magnitude of the earth's magnetic field is [nearly =] 50 micro T.) The magnetic field causes all magnetic dipoles (objects behaving like compass needles) to precess about the direction of the applied field, a phenomenon called Larmor precession, at a frequency (called the Larmor frequency) that is the product of three factors: a universal constant, the strength of the magnetic field, and the strength of the magnetic moment. Nuclear magnetic moments are typically [nearly =] 1,000 times smaller than the magnetic moment of the electron or magnetic moments of atoms and molecules. At 1 T, Larmor frequencies for sup 1 H,31p, and13C are, respectively, 42.576, 17.236, and 10.705 MHz. At 4.7 and 11.75 T Larmor frequencies for1H are, respectively, 200 and 500 MHz.
The Larmor frequencies calculated from a knowledge of the applied field apply only to isolated dipole moments, i.e., isolated1H,31P, and13C nuclei, floating alone in space, experiencing only the applied magnetic field. Nuclei are not alone, however; they are surrounded by atomic electrons, the rest of the parent molecule, and numerous other molecules. The static magnetic fields used in NMR spectrometers are uniform to approximately 1 part in 10 . Magnetic fields generated within a molecule by orbiting electrons, however, add to the applied magnetic field, changing the total magnetic field and final precession frequency at a particular molecular position, typically by several ppm. Consequently, within a single molecule, different electronic environments can be experienced by chemically different nuclei of the same type. Such is the case for the three31P nuclei in ATP. Each has a different Larmor frequency (by [nearly =] 5–14 ppm) that is clearly distinguishable with NMR spectroscopy.
In the second step of NMR spectroscopy, a rapidly rotating magnetic field is applied to the sample for a short period, typically tens to hundreds of microseconds, by a radiofrequency excitation coil. Magnetic moment transitions occur if and only if the frequency of the applied radiofrequency field is the same as the nuclear precession frequency. This phenomenon is known as nuclear magnetic resonance. The magnetic moment transitions absorb energy from the radiofrequency magnetic field, and this is sensed by a detection coil. Absorption signals occur at resonance frequencies, i.e., at radiofrequencies corresponding to the different precession frequencies of chemically distinct nuclei, for all molecules in the sample. Thus, NMR spectroscopy is absorption spectroscopy, and the shift in resonance frequencies caused by different chemical environments is called the chemical shift.
The radiofrequency detection coil is often physically the same as the radiofrequency excitation coil. (Circuitry to the coil, however, must then be changed rapidly from excitation circuitry to detection circuitry.) Although excitation by a radiofrequency pulse is typically centered at the Larmor frequency of a particular nucleus, there are band widths to the excitation (and detection), and these are made wide enough to include all resonance frequencies of interest. For sup 31 P spectroscopy of biologic tissue, sharp absorption signals can be detected at separate frequencies (over a range of [nearly =] 30 ppm), e.g., for phosphoenthanolamine, sugar phosphates (e.g., FBP, F6P, G6P), Pi, phosphodiesters, phosphocreatine, and the alpha, beta, and gamma31P nuclei in ATP.
We have mentioned that different molecular positions for particular nuclei are often uniquely identifiable. Such is the case for Pi, which exists primarily as one of two species at physiologic pH: H2PO4sup -(acid)[left and right arrow] H sup + HPO4sup --(base), with pKa[nearly =] 6.75. The chemical shift of the sup 31 P resonance in H2PO4is sup -- is [nearly =] 2 ppm different from the chemical shift of the31P resonance in HPO4. The31P nucleus experiences a different magnetic field in the two Pispecies because of the absence or presence of a single proton. For a particular P1molecule, protonation and deprotonation occur so rapidly that the31P nucleus experiences both species many times during the period of NMR resonance detection (milliseconds). The sup 31 P nucleus of Piin this situation is said to be in rapid exchange between the two species, and the measured resonance frequency for the31P nucleus turns out to be the weighted average of the resonance frequencies for the two Pispecies. The weights, which are the ratios of the two phosphate species to total Pi, uniquely relate to the chemical shift measured for Pi. Theoretically, the ratio of the two phosphate species also relates uniquely to pHivia the Henderson-Hasselbalch equation. Thus, pHiand the measured P sub i chemical shift are linked, as explained in Appendix 2, in which an explicit formula is given.
After a radiofrequency pulse is applied to a sample, one detects the NMR signal for a fixed period, during which it decays exponentially. The measured signal is called a free induction decay (FID). Many FIDs are summed and averaged to eliminate contributions from random fluctuations (noise). After an average FID has been obtained, a fourier transform is performed to obtain the NMR spectrum, a plot of NMR signal amplitude versus frequency. (The process is analogous to that used for power analysis of electroencephalographic spectra-in which electroencephalographic amplitudes are recorded for time epochs and fourier transformation is used to get amplitudes vs. frequency.) Averaging FIDs is an important method of increasing signal-to-noise ratios in NMR spectra. Typically, it takes a few minutes to accumulate the necessary number of FIDs, and one waits >or= to 1 s after a radiofrequency pulse before providing the next one. A long interpulse delay gives nuclei time to relax after radiofrequency excitation. Because resonance frequencies are different among nuclei, one can perform a radiofrequency excitation and FID collection for31P (or sup 1 H) during the time that1H (or31P) nuclei are relaxing if one has appropriate electronic equipment and is able to perform adequate computer tracking of the data. It is thus possible to obtain NMR spectra for two or more nuclei during the same 5-min period required to obtain a spectrum for only one. This is known as interleaved spectroscopy. The previous section is general and limited; omitted are technical descriptions of phase, a quantity that describes the frequency coherence of detected signals, and quadrature detection, a method that facilitates the correct assignment of frequencies relative to a reference frequency. Interested readers are referred to comprehensive texts. [38–40]
In high-resolution NMR spectroscopy, nuclei have parent molecules that are freely mobile in dilute solutions. We used high-resolution NMR spectroscopy to study metabolites extracted from brain slices. We obtained the metabolites by rapidly freezing the slices in liquid nitrogen, pulverizing the frozen tissue to a powder, and dissolving it in cold perchloric acid solution to preserve metabolites and inactivate enzymes. In brain slice studies, it is very convenient to obtain and freeze the tissue quickly. In addition, higher concentrations of compounds of interest can be made after lyophilizing solutions with extracts. With high-resolution NMR spectroscopy, one destroys the cells but detects all atomic nuclei because all molecules are freely mobile in solution.
The situation for in vivo NMR spectroscopy is different. In vivo, molecules are often bound or partitioned in macromolecular structures. As texts explain, large macromolecules and small ones bound to them move and tumble more slowly, causing broader NMR signals from reporting nuclei. The NMR signal intensity, or area under a signal's curve, is proportional to the concentration of the signal (assuming that FIDs are acquired with long interpulse delays that lead to full relaxation). For a fixed concentration, i.e., for a fixed NMR signal intensity, the height of an NMR signal diminishes as it widens. Macromolecular interactions can thus cause the height of a signal (amplitude) to become so small that it is no larger than the signal amplitude for random noise. In such cases, the signal is no longer detectable with NMR spectroscopy. With in vivo NMR spectroscopy, cells are not destroyed, but line broadening causes fewer nuclei to be detectable by NMR, and it causes some detectable signals to overlap with others.
Appendix 2: Details of Methods We Used for Ex Vivo and High-Resolution Nuclear Magnetic Resonance Spectroscopy
Interleaved31P and1H spectra were obtained on a Nalorac Quest Model 4400 (Martinez, CA) 4.7-T NMR instrument, operating at 81 and 200 MHz, respectively, using methods described previously. The 20-mm NMR tube containing 80 slices was positioned inside a custom-made probe. The probe had a 24-mm (outer diameter) Helmholtz coil on a Pyralux flexible composite (Du Pont Electronics, Towanda, PA) that was tuned to1H. An inner four-turn 23-mm (outer diameter) solenoidal coil was tuned to31P. Each acquisition consisted of 2,048 complex data points for both31P and1H. Time-sharing was such that it took 0.84 s for a single31P acquisition and 1.27 s for a single1H acquisition. Individual spectra for31P and sup 1 H were generated after 120 acquisitions in quadrature phase detection mode. The spectral width was +/- 4,000 Hz. Nuclear magnetic resonance signal intensities for each metabolite were determined by numerical integration of optimal computer fits to corresponding NMR resonance peaks (Nalorac Quest Model 4400 Curve Fitting Program; and MacFID Curve Fitting Program, Tecmag Inc., Bellaire, TX).
sup 31 P Magnetic excitation was typically via one radiofrequentTpulse of [nearly =] 30 micro s, causing a tip angle of [nearly =] 45 [degree sign]. The broad31P phospholipid hump was subtracted using a convolution difference method having a 500-Hz exponential filter.31P Metabolite concentrations were measured relative to corresponding signal peak areas in the control run (time = 0 min) and expressed as [Peak area]time=+/[Peak area]time= 0. Relative ATP levels were determined from the beta-ATP peak at 16.3 ppm. It is apparent from prior T1measurements by others that pulse conditions do not allow observation of fully relaxed signals. For example, the following T1values have long been known for 4.7 T :[nearly =] 3.5 s for phosphocreatine;[nearly =] 3.8 s for phosphomonoesters (PME);[nearly =] 2.6 s for Pi; and [nearly =] 1.3 s for beta-ATP. Acquisition conditions were the same for all experiments, however. If one assumes, as others do, that T1values do not change, then changes in NMR signal intensity in any metabolite are directly proportional to changes in the concentration of that metabolite. Therefore, no correction for saturation is needed when measuring relative changes of a particular brain slice metabolite. Intracellular pH was calculated from the Pi-phosphocreatine chemical shift difference, sigma, using an equation that was tested by several groups and ultimately refined : pHi= 6.75 + log ([sigma - 3.27]/[5.69 - sigma]).
Spin-echo1H NMR spectroscopy experiments were initiated with a 100-ms low-power presaturation pulse centered on the water resonance. A 136-ms spin-echo delay (TE) for refocusing pulses permitted substantial discrimination against more rapidly relaxing lipid signals near the 1.32-ppm lactate peak.1H Magnetic excitation pulses required a 50-watt amplifier and typically occurred via a radiofrequency pulse of [nearly =] 196 micro s duration, causing a tip angle of 90 [degree sign]. A Lorentzian-to-Gaussian transformation for each1H spectrum was performed by multiplying by a 12-Hz exponential and a 7-Hz Gaussian.
Ex vivo proton-decoupled13C NMR spectra of perfused brain slices were also obtained at 4.7 T during special studies in which there was administration of either [2-sup 13 C] glucose or normal glucose. (Proper determination of low-concentration metabolite changes after [2-sup 13 C]glucose administration requires a knowledge of background signal intensities occurring from naturally abundant13C [nearly =] 1.1% of all carbon nuclei].) Ex vivo13C NMR studies were performed using a custom-built probe tuned to 50.32 MHz. Ten minutes was required to obtain one spectrum. Each 10-min acquisition consisted of 4,096 complex data points, obtained in quadrature phase detection mode. The spectral width was +/- 12,000 Hz.13C Magnetic excitation was typically from one radiofrequency pulse of [nearly =] 70 micro s, causing a tip angle of 45 [degree sign]. An exponential filter of 12–15 Hz was used to reduce noise in the spectra. A WALTZ-16 scheme was used during acquisitions for broadband proton decoupling. Studies of naturally abundant13C metabolites required the addition of three 10-min acquisitions.
High-resolution sup 13 C NMR Spectroscopy of Extracted Metabolites
Additional separate studies with13C-enriched glucose were conducted for NMR studies of extracted metabolites. In these, all 80 brain tissue slices (3.2 g wet weight) were removed just before hypoxia or at the end of the 30-min hypoxia period and were immediately frozen in liquid nitrogen. After perchloric acid extraction, proton-decoupled13C NMR spectra were obtained with an 11.75-T GN500 instrument (Brucker Instruments, Billerica, MA) operating at 126 MHz. In the extraction procedure, frozen slices were ground into a fine powder in a cryo-cooled mortar and then dissolved in a 1.0 N HCIO4solution containing 20 mM edetic acid, using a 1:2 (weight:volume) ratio. Separation of cell debris from perchloric acid extracts was accomplished at 4 [degree sign] Celsius by 10 min of centrifugation at 15,000 rpm (Sorvall Refrigerated Centrifuge, IL). The H of the perchloric acid extract was adjusted to [nearly =] 7 using 1 N KOH. Solutions were subsequently frozen in liquid nitrogen and lyophilized using a Labconco freeze-dryer (Fischer Scientific, Fremont, Newton, CT). Immediately before high-resolution NMR studies, the freeze-dried perchloric acid extracts were dissolved in 0.5 ml deuterium oxide (D2O) and transferred to a 5-mm (outer diameter) NMR tube.
Data acquisitions for high resolution13C experiments were accumulated in 4-h blocks. Typically, four to six blocks were summed to obtain improved signal-to-noise ratios. Each high-resolution NMR acquisition consisted of 4,096 complex data points, obtained in quadrature phase detection mode. The spectral width was 30,000 Hz.13C Magnetic excitation was typically one radiofrequency pulse of [nearly =] 20 micro s in duration, causing a tip angle of 45 [degree sign]. As in ex vivo NMR studies, an exponential filter of 3–7 Hz was used to reduce spectral noise, and a WALTZ-16 scheme was used for broadband proton decoupling. Chemical shifts were referenced to resonance peaks for [beta-C2]glucose at 74.1 ppm. [25–27]