SPECIAL COMMUNICATION
Noninvasive measurements of transmural myocardial metabolites using 3-D ³¹P NMR spectroscopy

Yong K. Cho, Hellmut Merkle, Jianyi Zhang, Nikolaos V. Tsekos, Robert J. Bache, and Kâmil Uurbil

Center for Magnetic Resonance Research and Departments of Medicine and Radiology, University of Minnesota Medical School, Minneapolis, Minnesota 55455

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A completely noninvasive three-dimensional (3-D) static magnetic field magnitude spatially localized ³¹P spectroscopy technique has been developed and applied to study the in vivo canine myocardium at 9.4 T. The technique incorporates both Fourier series windows and selective Fourier transform methods utilizing all three orthogonal gradients for 3-D phase encoding. The number of data acquisitions for each phase-encoding step was weighted according to the Fourier coefficients to define cylindrical voxels. Spatially localized ³¹P spectra can be generated for voxels of desired location within the field of view as a postprocessing step. The quality of localization was first demonstrated by using a three-compartment phantom. The technique was then applied to in vivo canine models and yielded ³¹P cardiac spectra with an excellent signal-to-noise ratio. The in vivo validation experiments, using an implanted 2-phosphoenolpyruvate-containing marker, demonstrated that the technique is capable of measuring at least two transmural layers of left ventricular myocardium representing the subepicardium and subendocardium.

phosphocreatine; ATP; left ventricular hypertrophy three-dimensional; high-energy phosphates

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

THE RELATIONSHIP BETWEEN ENERGY required for contractile work of the heart and oxygen availability can be examined by measuring myocardial high-energy phosphates (HEP) over a range of workloads by using ³¹P NMR spectroscopy. This relationship is expected to be nonuniform across the left ventricular (LV) wall because systolic tension development, and therefore oxygen consumption and ATP utilization rates, are greater in the subendocardium than in the subepicardium of the LV (2, 17-24, 36, 37, 40, 41, 50). Restrictions in coronary artery flow, as encountered during ischemia, also result in transmurally nonuniform perturbations of oxygen supply and HEP metabolism (3, 4, 30-32, 40, 41, 51, 52). Because of these transmural gradients, spectroscopic localization techniques were developed to achieve transmural differentiation of myocardial NMR spectra in instrumented, open-chest preparations where a radio frequency (RF) surface coil was placed in direct contact with the exposed LV surface (1, 13, 15, 16, 34, 35, 43-45). These techniques have demonstrated (31-34, 42-52) that significant differences in HEP content exist across the LV wall of anesthetized open-chest animals even under resting conditions, and that these transmural gradients are perturbed by alterations in workload, ischemia, anesthesia, and LV hypertrophy (LVH).

Whereas these studies in open-chest animals have yielded new insights and information that would not have been available otherwise, generalization of these results to more physiological conditions cannot be assumed. It is well recognized that general anesthesia and surgical trauma can have deleterious consequences on autonomic responses, reflex control of the coronary vascular bed, and metabolic coronary vasoregulation (12, 38, 39). Therefore, the validity of the conclusions derived from the open-chest studies remains to be defined in the intact closed-chest animal. This requires spatially localized spectroscopy studies with sufficient resolution to achieve transmural differentiation in the intact closed-chest animal within reasonable data acquisition times. In this paper, we demonstrate that this is feasible at 9.4 T magnetic field strength, the highest magnetic field so far employed for such medium to large animal cardiac studies. We demonstrate that spectra can be acquired in three-dimensions (3-D) with transmural differentiation; these localized spectra reveal that even in the closed-chest animal there exists a transmural gradient in the phosphocreatine (PCr)-to-ATP ratio (PCr/ATP) across the LV wall under resting conditions and that this gradient is present both in the normal and hypertrophied heart.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

NMR and RF coil. Experiments were performed in a 31-cm horizontal bore, 9.4-T magnet (Magnex) interfaced to a Unity console (Varian). A quadrature RF coil assembly consisting of a coplanar dual loop (28) coil was employed; a single loop coil at the center was used for ³¹P spectroscopy to enhance the RF signal penetration and improve sensitivity. Both coils in the quadrature assembly were tuned to the ³¹P frequency of 162.19 MHz. The coil assembly is illustrated in Fig. 1. The ³¹P coil was slightly curved to allow optimal contact with the chest wall of the animal. A 7-cm diameter single loop circular coil with distributed capacitance tuned to 400.22 MHz was used for ¹H localized shimming and anatomic imaging. The ³¹P quadrature coil and the ¹H surface coil were decoupled from each other on the basis of their geometry so that mutual interference was minimized. Coordinate axes were defined with z parallel to the static magnetic field magnitude (B₀) and xz as the RF coil plane.

View larger version (20K): [in this window] [in a new window]   Fig. 1.   A quadrature 31P radio frequency (RF) coil assembly consisting of a coplanar dual-loop coil (gray lines, 3.5-cm wide and 13-cm long), and, at the center, a 3.5-cm diameter surface coil. Both coils of the assembly are mounted on the surface of a cylindrical former. They were tuned to 162 MHz, the 31P Larmor frequency at 9.4 T. Decoupling was achieved by intrinsic symmetry and its placement to each other. , RF feeding points for the tune-match circuitry. B1, field magnitude.

To compare the performance of ³¹P quadrature coil with a linear coil, the B₁ sensitivity profiles were measured from the quadrature coil and from a comparable size single-turn surface coil. The B₁ sensitivity profiles in the zx-plane through the isocenter are shown in Fig. 2. These sensitivity profiles were obtained by using a homogeneous phantom containing 0.5 M NaPO₄ · H₂O to match the loading of the canine model. Figure 2, A and B, shows data from the linear coil and the quadrature coil, respectively. This comparison indicates that the RF penetration and sensitivity were improved by using the quadrature assembly.

View larger version (30K): [in this window] [in a new window]   Fig. 2.   Comparison of the B1 gradient sensitivity performance profiles between the single-loop coil (A) and the quadrature coil (B). These sensitivity profiles were obtained by using a homogeneous phantom containing 0.5 M of NaPO4 · H2O to match the loading of the canine model. Comparison of the sensitivity profiles indicates that the RF penetration and sensitivity were improved by using the quadrature assembly. Numbers are distance in centimeters along the coil axis that is perpendicular to the plane of the surface coil.

3-D B₀ spatial localization. Spatial localization was achieved with the use of 3-D chemical shift imaging (6) modified to apply a "filter" on the k-space data in the data acquisition stage to optimize data acquisition and simultaneously define the voxel shape. Phase encoding along the three orthogonal axes was on the basis of B₀ gradients. This approach incorporates features of previously described Fourier series windows (FSW) (8-11, 29) and the selective Fourier transform (SFT) methods (5, 25, 26). This approach has been employed previously to define cylindrical voxels at 4.7 T and has produced promising results in preliminary animal experiments (14).

In this study, the 3-D B₀ spatial localization for ³¹P spectroscopy was implemented with a square RF excitation pulse, followed by a half-sine, phase-encoding, B₀ gradient pulse that was incremented in equal steps along all three orthogonal axes. The multipliers: l, m, and n for these increments have values that range from l₀ to l₀, m₀ to m₀, and n₀ to n₀ for unit steps along the x-, y-, and z-axes, respectively. These incremented gradients produce a phase difference of 2 radians along the field of view (FOV) for each axis; spatial coordinates can thus be represented by a phase factor ranging from to +.

The 3-D voxel shape used in this study was a cylinder with its circular dimension in the xz-plane and rectangular dimension in the y-axis. A cylindrical voxel defined by the FSW approach required fewer phase-encoding steps than would be required for a rectangular voxel. The 3-D localization is an extension of the two-dimensional localization that was previously reported (14). The number of data acquisitions for each phase-encoding step was weighted according to the Fourier coefficients

(1)

with the first term representing the coefficients for the circular dimension of the cylinder and the second term representing the coefficients for the rectangular (height) dimension of the cylinder. In B_lmn, l and m are the unit step multipliers for the phase-encoding increments in the circular dimensions and n is the multiplier for the rectangular dimension, J₁ is the Bessel function of the first order, and k =  and w₁ and _c are the angular length and angular radius of the cylinder in radians, respectively. The desired voxel shape at the isocenter of three orthogonal gradients can be described by the window function, W(₁,₂,₃), where ₁, ₂, and ₃ represent the spatially dependent phases along three orthogonal axes. This 3-D voxel can be described as a set of Fourier coefficients in Eq. 1

(2)

where j = . Because convolution of sequences implies multiplication of the corresponding Fourier transforms, the localized spectroscopic signal for this voxel can be generated by the summation of B_lmnS(l,m,n,t) for all values of l, m, and n, where S(l,m,n,t) is the signal acquired after the application of all three orthogonal gradients. Therefore, the spectrum can be generated by

(3)

where p(r,t) represents the time-domain signal corresponding to position r. Because of the limited time of in vivo experiments, not all the significant coefficients can be utilized, so the Fourier series must be terminated. The truncation effect was minimized by terminating the Fourier series at the first zero-crossing point (9). Differences between the actual value of coefficients and the integer number of scans were compensated by multiplying the obtained signals by correction factors. A spectrum from a single voxel was generated with summation with respect to the entire phase-encoding domain. In addition, spectra from an arbitrarily defined spatial location within the FOV can be generated by shifting the voxels to the desired location during postprocessing of the data (15).

The performance of this B₀ localization approach was evaluated by using a ³¹P phantom that consisted of an array of three compartments created by using thin glass plates separated with rubber O rings. The compartments were of equal dimensions of 5-cm diameter and 0.5-cm thickness stacked back to back. The three compartments were filled with different ³¹P solutions, namely 0.5 M tetrasodium pyrophosphate, 1.5 M inorganic phosphate, and 2 M phenylphosphonate.

Animal preparation and positioning. All in vivo animal experimental procedures were approved by the University of Minnesota Institutional Animal Care and Use Committee. The investigation conformed to the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health (NIH publication No. 85-23, Revised 1985).

Eight adult beagles weighing 10-13 kg were anesthetized with pentobarbital sodium (10 mg/kg iv) to allow intubation and control of ventilation with a respirator (Siemens). Anesthesia was then maintained with morphine sulfate (2 mg · kg¹ · h¹ iv). A fiber-optic cable with an infrared sensor (Nonin Medical) was clipped to the ear to monitor the arterial pulse and to trigger the NMR experiments to the cardiac cycle. The animals were placed in the prone position on a coil cradle with the heart directly over the ³¹P coil. In three of the animals, a reference containing phosphoenolpyruvate (PEP) was surgically placed between the chest wall and the anterior LV to allow assessment of the spatial localization technique. These animals were premedicated with acepromazine (10 mg im), anesthetized with pentobarbital sodium (30 mg/kg iv), intubated, and ventilated with a respirator supplemented with oxygen. A left thoracotomy was performed in the fourth intercostal space, and a thin circular phantom (3 cm diameter, 0.4-cm thick, relatively soft plastic bag) containing 0.05 M of PEP was implanted between the chest skeletal muscle and the heart. This arrangement was employed for evaluating the accuracy of spatial localization. PEP was used because it has a unique chemical shift at ~3 parts per million (ppm) higher than PCr and is well resolved from all other signals detected in the ³¹P spectra of cardiac and skeletal muscle. The concentration of PEP was chosen so that its intensity would be comparable to that of PCr in myocardium. After we completed the surgery, the thoracotomy was closed in layers, the chest was evacuated of air, and the animal was allowed to recover for 1-2 wk before the study.

Production of LVH. Four beagle dogs ~8 wk of age were anesthetized with pentobarbital sodium (25-30 mg/kg iv), intubated, and ventilated with a respirator. A right thoracotomy was performed in the third intercostal space, and the ascending aorta, ~1.5 cm above the aortic valve, was mobilized and encircled with a 2.5-mm-wide polyethylene band. While we simultaneously measured LV and distal aortic pressures, we tightened the band until a 20- to 30-mmHg peak systolic pressure gradient was achieved across the narrowing. The chest was then closed, the pneumothorax was evacuated, and the animals were allowed to recover. LVH occurred progressively as the area of aortic constriction remained fixed in the face of normal body growth. At ~6 mo of age when the LV weight-to-body weight ratio had increased by ~50% compared with normal, the animals were returned to the laboratory for study.

3-D ³¹P experiment. Anatomic images were obtained utilizing an ultrafast gradient-recalled echo sequence (turboFLASH) (7) to position the animal so that the center of the anterior LV wall was as close to the isocenter as possible; the relative positions were recorded to be used as a reference for post processing of the data. Magnetic field homogeneity was optimized by using the ¹H water resonance. Localized shimming was performed on an ~1-cm-thick slice of the anterior LV wall that was parallel with the coil plane. Motion artifact was minimized during shimming and subsequent data acquisition by gating the acquisition trigger to the cardiac cycle. In the region of interest, the anterior LV wall, motion due to respiration was found to be minimal; hence, respiratory gating was not employed to reduce the acquisition time. Studies with the PEP containing a "phantom" inserted in the chest cavity also demonstrated that spatial localization did not suffer from this lack of gating to respiration.

³¹P spectroscopy using the 3-D B₀ FSW-SFT technique was implemented with a square RF excitation followed by a half-sine-shaped phase-encoding gradient on all three axes. With the use of the B₁ magnitude measured at the coil center, the RF pulse duration was calibrated to give approximately the Ernst angle at the region of interest (the anterior LV wall) to maximize the signal from this region. 3-D localization experiments were performed by using the two different parameter sets shown in Table 1.

View this table: [in this window] [in a new window]   Table 1.   Parameter sets used in this study

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Phantom experiment. The quality of the 3-D B₀ spatial localization achieved was first evaluated by using the three-compartment phantom described in METHODS. The phantom was attached to a bottle containing saline to mimic the in vivo loading conditions of the RF coil. The phantom was positioned so that its center was aligned with the ³¹P coil center. The RF coils were mounted on a curved platform designed to simulate the canine chest. The phosphorous solution of the first compartment, beginning ~2.5 cm from the coil center, simulates the region of interest within the canine LV wall that is typically ~1.5 to 2 cm from the chest surface in an intact animal.

Figure 3 illustrates a cross-sectional image of the phantom together with the corresponding ³¹P spectra generated by the 3-D B₀ FSW-SFT spatial localization technique. An array of voxels, 2 cm in diameter and 0.5 cm thick, were positioned as indicated in the figure so as to have one voxel cover predominantly one compartment of the phantom. Because each compartment contained a different compound, it was uniquely identified by the chemical shift of that compound. The localized ³¹P spectra illustrate predominantly one peak from each voxel. Although there exists some out-of-voxel contamination from the neighboring voxels, there is virtually no signal contamination between the innermost and outermost voxels.

View larger version (30K): [in this window] [in a new window]   Fig. 3.   Images from the phantom studies together with corresponding 31P spectra generated by the three-dimensional (3-D) B0 Fourier series window (FSW)-selective Fourier transform (SFT) spatial localization technique. A: 1H turboFLASH sequence [repetition time (TR), 6.74 ms; echo time (TE), 3.37 ms; field of view (FOV), 12 × 12 cm/slice, thickness = 0.5 cm] in the xy-plane. An array of voxels was placed centered at the center of each compartment. Locations of the three voxels, one for each phosphorous solution, are indicated by rectangles representing the side view of the cylinders. Each cylindrical voxel was 2 cm in diameter in its circular dimension and 0.5 cm in thickness. B: each spectrum was generated from the corresponding voxel in A. Although there exists some out-of-voxel contamination from neighboring voxels, there is virtually no signal contamination from every other voxel.

In vivo canine experiments. Results of the validation study in one animal with an implanted PEP chamber are illustrated in Fig. 4. PEP was located between the anterior LV wall and the chest wall as seen in the axial image. The voxel locations, from which five localized spectra were extracted, are indicated by rectangles drawn on the image. The spectrum in Fig. 4, voxel a, originated from chest skeletal muscle, and, consistent with this location, PCr/ATP and P_i-to-ATP or P_i-to-PCr ratios were relatively high in this voxel. Voxel b was positioned predominantly over the implanted PEP phantom located between the chest wall and the anterior LV wall. This spectrum shows a large PEP resonance ~3 ppm from PCr with minor contributions from PCr and ATP, which are expected given the partial volume effect that is evident in the image. Spectrum c is from the subepicardium immediately adjacent to the implanted PEP phantom; despite its proximity to the PEP phantom, this voxel shows no contribution from PEP and contains only myocardial phosphorus resonances. Spectrum d is from the voxel located mainly in the subendocardium and mixed with a small portion of blood from LV cavity. This contributes to the decreased signal-to-noise ratio (SNR) in spectrum d compared with spectrum c. Finally, spectrum e is from a voxel that penetrates deeper and is occupied mostly by blood in the LV chamber with a small contribution from interventricular septum.

View larger version (48K): [in this window] [in a new window]   Fig. 4.   Validation of spatial localization by using an implanted 2-phosophoenolpyruvate (PEP) phantom. PEP is located between the anterior left ventricular (LV) wall and the chest wall. A: magnetic resonance image obtained with the use of turboFLASH and spectrum (a) originated from chest wall skeletal muscle as indicated by a Pi resonance and a high phosphocreatine-to-ATP ratio (PCr/ATP). B: set of spatially localized spectra obtained from the respective volume element of the left ventricle. Spectrum b is from the voxel containing the implanted PEP phantom located between the chest wall and the anterior LV wall; this spectrum shows the PEP resonance located at ~3 ppm upfield of PCr. Spectrum c is mostly from the voxel containing subepicardium; spectrum d is mostly from the voxel containing subendocardium mixed with LV cavity blood, which contributes to the decreased signal-to-noise ratio in spectrum d compared with spectrum c, and spectrum e is from a voxel containing mostly blood in the LV chamber with some interventricular septal tissue as indicated by the resonances of 2,3-diphosphoglycerate (2,3-DPG) and small high-energy phosphate (HEP) resonances.

The resonance intensities in these spectra are determined by the spatially nonuniform excitation pulse profile (adiabatic pulses were not used) and the coil sensitivity, which decays with increasing distance from the coil. The excitation pulses were adjusted to yield maximal excitation approximately at the location of the voxel shown in spectrum c in Fig. 4. Consequently, resonance intensities are in general lower in spectra d and e as well as in a. Intensities of spectrum a would have been significantly higher with an adiabatic pulse or a conventional pulse optimized over this voxel; however, given the generally higher phosphorus metabolite content of the skeletal muscle, this was not desirable and was avoided.

In Fig. 5, two different sets of spectra obtained from a normal dog with the use of the parameter set 1 and set 2 are shown with the same vertical scale for comparison. Both sets of spectra were generated from the transmural array of voxels described in the validation study (Fig. 4). Figure 5A was obtained in ~17 min and Fig. 5B in 33 min. The details of parameters used are found in Table 1. As in Fig. 4, voxel a originated from the chest skeletal muscle; spectrum b is from a voxel located between the skeletal muscle and the heart comprising some skeletal muscle and air gap; spectrum c is mostly from the subepicardium; spectrum d is mainly from the subendocardium including some LV chamber; and spectrum e is mostly from blood in the LV chamber as indicated by the resonances of 2,3-diphosphoglycerate (2,3-DPG) as well as some HEP from the interventricular septum. The SNR (PCr/root mean square noise) measured from the subepicardial spectrum was 35 for Fig. 5A and 50 for Fig. 5B. These data demonstrate that ³¹P NMR can be obtained with transmural differentiation of the LV wall into two myocardial layers representing subepicardium and subendocardium with excellent SNR.

View larger version (16K): [in this window] [in a new window]   Fig. 5.   Two different sets of spectra obtained from a normal dog using the parameter set 1 and set 2 are shown with the same vertical scale for comparison. Both spectra sets were generated from the transmurally arrayed voxels. Spectra in A was obtained in ~17 min and B in 33 min. Voxel in spectrum a is principally from chest skeletal muscle; spectrum b is from a voxel located between the skeletal muscle and the heart; spectrum c is mostly from the subepicardium; spectrum d is mainly from the subendocardium including some LV chamber, and spectrum e is mostly from blood in the LV chamber as indicated by the resonances of 2,3-DPG as well as some HEP from interventricular septum included in the voxel.

The PCr-to-ATP ratios (PCr/ATP) measured by using set 1 parameters (Table 1) in eight normal dogs and four dogs with LVH are shown in Table 2. Within groups the difference between the subepicardial and subendocardial PCr/ATP can be evaluated by using paired testing because both were obtained from the same heart. Thus the significance of this difference is not affected by interanimal variability in the average myocardial PCr/ATP. Paired analysis revealed that the subepicardial PCr/ATP was significantly greater than the subendocardial PCr/ATP in both groups of animals (Table 2). Comparison of normal and hypertrophied hearts demonstrated that the mean PCr/ATP was less in LVH than in normal hearts, and this difference achieved statistical significance in the subendocardial voxel (Table 2).

View this table: [in this window] [in a new window]   Table 2.   Myocardial PCr-to-ATP ratio from normal dogs and dogs with LVH

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

3-D B₀ spatial localization. In this report we present ³¹P cardiac spectra with transmural localization in intact animals. As described earlier, 3-D spatial localization was achieved by using three orthogonal gradient-weighted k-space acquisitions for enhanced SNR. Postprocessing of the acquired FIDs enabled accurate placement of an arbitrarily shaped voxel in a desired location within the FOV. The use of a quadrature RF coil improved the B₁ penetration and sensitivity compared with a similar size linear coil. In addition, the high magnetic field, 9.4 T, contributed to increasing the SNR of the spectra. This is evident if one compares 4.7 T spectra presented in previous publications from our group in the open-chest canine preparation with small surface coils placed directly on the heart (32, 34, 42-52). The SNR in these open-chest measurements with RF coils directly in contact with the myocardium are similar for comparable data acquisition times to the closed-chest data presented in this paper.

The total ³¹P data acquisition time was ~17 min for parameter set 1 and 33 min for set 2 (Table 1) with TR equal to 1 s. The 33-min data acquisition time produced spectra with excellent SNR and spatial resolution; however, it is desirable to minimize the total data acquisition time for in vivo experiments. The spectra generated by using a 17-min acquisition time displayed adequate spectral quality with high SNR. A further decrease in the number of phase-encoding steps could further reduce the total data acquisition time, but at the expense of a poorer SNR and reduced spatial resolution.

Transmural distribution of HEP levels in intact animals. The present data demonstrated a statistically significant gradient in PCr/ATP across the LV wall. This is consistent with earlier open-chest studies and suggests that also in the closed-chest animal the greater contractile work performed by the inner myocardial layers is reflected in a lower PCr/ATP and, hence, higher free ADP content, given the fact that the total creatine level is uniform across the LV wall (32, 43, 46). An earlier study suggested that the use of pentobarbital for surgical anesthesia might be responsible for the transmural gradient in PCr/ATP across the LV wall in open-chest animals, because this transmural gradient was less in animals anesthetized with -chloralose (33). The mechanism for the difference in PCr/ATP depending on the choice of anesthesia previously reported in open-chest animals was unclear, but might have been related to impairment of coronary metabolic vasoregulation resulting from general anesthesia and acute surgical trauma in the open-chest state. In the present study, surgical anesthesia was not required during data acquisition. Only a small dose of pentobarbital was used to allow intubation for control of respiration, and thereafter anesthesia was maintained with morphine sulfate. The present finding is in agreement with a previous report in normal human subjects in which a trend toward lower PCr/ATP was found in the subendocardium than in the subepicardium (27). Because the present technique allowed sampling from only two layers across the LV wall, it is possible that selective sampling from a thinner myocardial layer deeper in the subendocardium would show an even lower PCr/ATP.

In these studies, care was exercised to eliminate differential perturbation of the PCr /ATP resonances across the LV wall. All measurements of the ratios were based integrals of the resonances (rather than peak height), so that potential differences in peak height due to magnetic field inhomogeneities across the LV wall are avoided. The methodology could in principle have differential effects on the PCr and ATP resonances as the result of off-resonance effects, thereby altering the PCr/ATP. The off-resonance problem can be especially troublesome if the RF excitation pulse is weak compared with the resonance frequency difference between the resonances in question and the RF pulse. This problem was avoided by placing the RF of the excitation pulse midway between the PCr and ATP resonances, thus assuring identical off-resonance effects irrespective of the spatial location. Even without this precaution, however, off-resonance problems were expected to be minimal for the resonances observed because of the large B₁ values attained (typically 25 kHz in the coil center) versus the frequency difference between the RF pulse and the PCr and ATP resonances (200 Hz at 9.4 T). This is experimentally confirmed in the data by comparing the ATP to ATP resonances in Figs. 5 and 6. Because the RF excitation pulse was located midway between the PCr and ATP resonance, any off-resonance problem would be much more apparent between these two resonances. The essentially constant ratio of ATP to ATP in spectra obtained from different locations demonstrates that off-resonance problems did not interfere with quantitation of the acquired data.

View larger version (19K): [in this window] [in a new window]   Fig. 6.   Comparison of spectra obtained from a normal dog (A) and from a dog with pressure-overload LV hypertrophy (B). Data were acquired using set 1 parameters.

A principal advantage of the closed-chest capability presented in this paper is that the noninvasive and nondestructive nature makes it possible to perform longitudinal studies over a long period of time. For example, myocardial infarction can result in progressive LV dilatation and dysfunction over time. Similarly, LVH can cause decreases of myocardial HEP content and the PCr/ATP, which are related to the severity of hypertrophy (2, 46, 52), but the time course of these abnormalities relative to the development of hypertrophy or heart failure is unknown. In this report, we include ³¹P cardiac spectra from four closed-chest dogs in which pressure-overload LVH was produced by banding the ascending aorta. In a previous study using this experimental model, open-chest ³¹P NMR spectra demonstrated significantly lower PCr/ATP in hypertrophied than in normal hearts. However, hypertrophied hearts have increased sensitivity to the effects of anesthesia and acute surgical trauma, so that the previously observed HEP alterations might have been the result of the experimental preparation. The present data exclude this possibility and demonstrate that the PCr/ATP is also significantly less in LVH than in normal hearts studied in the closed-chest state using minimal levels of anesthesia required for endotracheal intubation and control of respiration. The availability of a noninvasive technique for acquiring spatially localized cardiac ³¹P NMR spectra will allow assessment of the development of HEP abnormalities during the evolution of the myocardial hypertrophic process over time.

In conclusion, 3-D B₀ spatially localized ³¹P NMR spectroscopy at 9.4 T can generate minimally contaminated transmural cardiac spectra from subepicardium and subendocardium with excellent SNR. This technique allows studies previously performed in open-chest animal models to be executed in closed-chest intact animal models. The noninvasive nature of the technique makes it possible to perform longitudinal studies in which the responses of myocardial HEP to cardiac overload can be monitored over time. Moreover, the technique offers the potential for improved characterization of human heart disease using high-field ³¹P NMR spectroscopy.

	ACKNOWLEDGEMENTS

This work was supported by National Institutes of Health Grants HL-33600, HL-50470, HL-61353, HL-21872, HL-58067, HL-58840, and RR-08079 and by a Whitaker Foundation Biomedical Engineering grant. J. Zhang received an American Heart Association Established Investigator Award.

	FOOTNOTES

Address for reprint requests and other correspondence: J. Zhang, Univ. of Minnesota, Cardiovascular Div., Mayo Mail Code 508, 420 Delaware St. SE, Minneapolis, MN 55455 (E-mail: zhang047{at}tc.umn.edu).

The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received 14 March 2000; accepted in final form 25 August 2000.

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

1.   Abduljalil, AM, Rath DP, Hui Z, Aletras AH, McCartney WC, and Robitaille PM. Spatial localization with modified Fourier series windows. Application to the transmural ¹³C-nuclear magnetic resonance analysis of the in vivo myocardium. Invest Radiol 31: 611-618, 1996[Web of Science][Medline].

2.   Bache, RJ, MacHale PA, and Greenfield JC. Transmural myocardial perfusion during restricted coronary inflow in the awake dog. Am J Physiol Heart Circ Physiol 232: H645-H651, 1977.

3.   Bache, RJ, Zhang J, Path G, Merkle H, Hendrich K, From AHL, and Ugurbil K. High-energy phosphate responses to tachycardia and inotropic stimulation in left ventricular hypertrophy. Am J Physiol Heart Circ Physiol 266: H1959-H1970, 1994[Abstract/Free Full Text].

4.   Boerth, RC, Covell JW, Seagren SC, and Pool PE. High-energy phosphate concentrations in dog myocardium during stress. Am J Physiol 216: 1103-1106, 1969.

5.   Brooker, HR, Mareci TH, and Mao J. Selective Fourier transform localization. Magn Reson Med 5: 417-433, 1987[Web of Science][Medline].

6.   Brown, TR, Kincaid BM, and Ugurbil KK. Chemical shift imaging in 3D. Proc Natl Acad Sci USA 79: 3523-3526, 1982[Abstract/Free Full Text].

7.   Chien, D, and Edelman RR. Ultrafast imaging using gradient echoes. Magnes Res 7: 31-56, 1991.

8.   Garwood, M, Robitaille PM, and Ugurbil K. Fourier series windows on and off-resonance using multiple coils and longitudinal modulation. J Magn Reson 75: 244-261, 1987.

9.   Garwood, M, Schleich T, Bendall MR, and Pegg DT. Improved fourier series windows for localization in in vivo NMR spectroscopy. J Magn Reson 65: 510-515, 1985.

10.   Garwood, M, Schleich T, Ross BD, Matson GB, and Winters WD. A modified rotating frame experiment based on a fourier series window function: application to in vivo spatially localized NMR spectroscopy. J Magn Reson 65: 239-251, 1985.

11.   Garwood, M, Ugurbil K, Schleich T, Petein M, Sublett E, From AHL, and Bache R. In vivo spatially localized surface coil NMR spectroscopy utilizing a fourier series window function and two surface coils. J Magn Reson 69: 576-581, 1986.

12.   Hanley, FL, Grattan MT, and Stevens MB. Role of adenosine in coronary autoregulation. Am J Physiol Heart Circ Physiol 249: H558-H566, 1985.

13.   Hendrich, K, Garwood M, and Ugurbil K. Spectroscopic imaging using variable angle excitation from adiabatic plane rotation pulses. J Magn Reson 19: 496-501, 1991.

14.   Hendrich, K, Hu X, Menon RS, Camarata P, Heros R, and Ugurbil K. Spectroscopic imaging of circular voxels with a two-dimensional Fourier series window technique. J Magn Reson 105: 225-232, 1994.

15.   Hendrich, K, Liu H, Merkle H, Zhang J, and Ugurbil K. B₁ voxel shifting of phase-modulated spectroscopic localization techniques. J Magn Reson 97: 486-497, 1992[Web of Science].

16.   Hendrich, K, Merkle H, Weisdorf S, Vine W, Garwood M, and Ugurbil K. Phase modulated rotating frame spectroscopic localization using an adiabatic plane rotation pulse and a single surface coil. J Magn Reson 92: 258-275, 1991[Web of Science].

17.   Hess, DS, and Bache RJ. Transmural distribution of myocardial blood flow during systole in the awake dog. Circ Res 38: 5-15, 1976[Abstract/Free Full Text].

18.   Honig, CR, Kirk ES, and Meyers WW. Transmural distributions of blood flow, oxygen tension and metabolism in myocardium: mechanism and adaptions. In: Coronary Circulational Energetics of the Myocardium, edited by Marechetti E, and Taccardi B.. Basel, Switzerland: Karge, 1967, p. 31-47.

19.   Hood, WP, Thomson WJ, Rackley CE, and Rolett EL. Comparison of calculations of left ventricular wall stress in man from thin-walled and thick-walled ellipsoidal models. Circ Res 24: 575-582, 1969[Abstract/Free Full Text].

20.   Kirk, E, Brooks H, Turbow M, and Sonnenblick EH. Myocardial performance in relation to altered coronary blood flow: heterogeneity of contractility and limitations of ventricular performance. Myocardial Blood Flow in Man; Methods and Significance in Coronary Artery Disease, edited by Maseri A.. Toronto: Minerva Medica, 1972, p. 11-22.

21.   Kirk, ES, and Honig CR. An experimental and theoretical analysis of myocardial tissue pressure. Am J Physiol 207: 361-370, 1964.

22.   Kirk, ES, and Honig CR. Nonuniform distribution of blood flow and gradients of oxygen tension within the heart. Am J Physiol 207: 661-668, 1964.

23.   Lowe, JE, Cummings RG, Adams DH, and Hull-Ryde EA. Evidence that ischemic cell death begins in the subendocardium independent of variations in collateral flow or wall tension. Circulation 68: 190-202, 1983[Abstract/Free Full Text].

24.   Lundsgaard-Hansen, P, Meyer C, and Riedwyl H. Transmural gradients of glycolytic enzyme activities in left ventricular myocardium. I. The normal state. Pflügers Arch 297: 89-106, 1967[Web of Science].

25.   Mareci, TH, and Brooker HR. High-resolution magnetic resonance spectra from a sensitive region defined with pulsed field gradients. J Magn Reson 57: 157-163, 1984.

26.   Mareci, TH, and Brooker HR. Essential considerations for spectral localization using indirect gradient encoding of spatial information. J Magn Reson 92: 229-246, 1991.

27.   Menon, RS, Hendrich K, Hu X, and Ugurbil K. ³¹P NMR spectroscopy of the human heart at 4 T: detection of substantially uncontaminated cardiac spectra and differentiation of subepicardium and subendocardium. Magn Reson Med 26: 368-376, 1992[Web of Science][Medline].

28.   Merkle, H, Wei H, Garwood M, and Ugurbil K. B₁ insensitive heteronuclear adiabatic polarization transfer for signal enhancement. J Magn Reson 99: 480-494, 1992.

29.   Metz, KR, and Briggs RW. Spatial localization of NMR spectra using Fourier series analysis. J Magn Reson 64: 172-176, 1985.

30.   Moir, TW. Subendocardial distribution of coronary blood flow and the effect of antianginal drugs. Circ Res 30: 621-627, 1972[Free Full Text].

31.   Murakami, Y, Zhang Y, Cho YK, Mansoor AM, Chung JK, Chu C, Francis G, Ugurbil K, Bache RJ, From AH, Jerosch-Herold M, Wilke N, and Zhang J. Myocardial oxygenation during high work states in hearts with postinfarction remodeling. Circulation 99: 942-948, 1999[Abstract/Free Full Text].

32.   Path, G, Robitaille PM, Merkle H, Tristani M, Zhang J, Garwood M, From AHL, Bache RJ, and Ugurbil K. The correlation between transmural high energy phosphate levels and myocardial blood flow in the presence of graded coronary stenosis. Circ Res 67: 660-673, 1990[Abstract/Free Full Text].

33.   Rath, DP, Little CM, Zhang H, Jiang Z, Abduljalil AM, Zhu H, Tong X, Brown C, Hamlin RL, and Robitaille PM. Sodium pentobarbital versus alpha-chloralose anesthesia. Experimental production of substantially different slopes in the transmural CP/ATP ratios within the left ventricle of the canine myocardium. Circulation 91: 471-475, 1995[Abstract/Free Full Text].

34.   Robitaille, PM, Lew B, Merkle H, Path G, Sublett E, Hendrich K, Lindstrom P, From AHL, Garwood M, Bache RJ, and Ugurbil K. Transmural high energy phosphate distribution and response to alterations in workload in the normal canine myocardium as studied with spatially localized ³¹P NMR spectroscopy. Magn Reson Med 16: 91-116, 1990[Web of Science][Medline].

35.   Robitaille, PM, Merkle H, Sublett E, Hendrich K, Lew B, Path G, From AHL, Bache RJ, and Ugurbil K. Spectroscopic imaging and spatial localization using adiabatic pulses and applications to detect transmural metabolite distribution in the canine heart. Magn Reson Med 10: 14-37, 1989[Web of Science][Medline].

36.   Spotnitz, HM, Sonnenblick EH, and Spiro D. Relation of ultrastructure to function in the intact heart: sarcomere structure relative to pressure volume curves of intact left ventricles of dog and cat. Circ Res 18: 49-66, 1966[Abstract/Free Full Text].

37.   Tota, B. On the regional metabolism of beef heart ventricles. Acta Physiol Scand 87: 289-295, 1973[Web of Science][Medline].

38.   Vatner, SF, and Braunwald E. Cardiovascular control mechanisms in the conscious state. N Engl J Med 6: 293-297, 1975.

39.   Vatner, SF, and Smith NT. Effects of halothane on left ventricular function and distribution of regional blood flow in dogs and primates. Circ Res 34: 155-167, 1974[Abstract/Free Full Text].

40.   Weiss, HH, and Sinha AK. Regional oxygen saturation of small arteries and veins in the canine myocardium. Circ Res 42: 119-126, 1978[Abstract/Free Full Text].

41.   Weiss, HR, Neubauer JA, Lipp JA, and Sinha AK. Quantitative determination of regional oxygen consumption in the dog heart. Circ Res 42: 394-401, 1978[Abstract/Free Full Text].

42.   Yoshiyama, M, Merkle H, Garwood M, From AHL, Bache RJ, Ugurbil K, and Zhang J. Transmural distribution of glucose uptake in normal and post-ischemic canine myocardium. NMR Biomed 8: 9-18, 1995[Web of Science][Medline].

43.   Zhang, J, Duncker DJ, Xu Y, Zhang Y, Path G, Merkle H, Hendrich K, From AHL, Bache RJ, and Ugurbil K. Transmural bioenergetic responses of normal myocardium to high workstates. Am J Physiol Heart Circ Physiol 268: H1891-H1905, 1995[Abstract/Free Full Text].

44.   Zhang, J, Duncker DJ, Ya X, Zhang Y, Pavek T, Wei H, Merkle H, Ugurbil K, From AHL, and Bache RJ. Effect of left ventricular hypertrophy secondary to chronic pressure overload on transmural myocardial 2-deoxyglucose uptake a ³¹P NMR spectroscopic study. Circulation 92: 1274-1283, 1995[Abstract/Free Full Text].

45.   Zhang, J, Ishibashi Y, Zhang Y, Eijgelshoven MH, Duncker DJ, Merkle H, Bache RJ, Ugurbil K, and From AH. Myocardial bioenergetics during acute hibernation. Am J Physiol Heart Circ Physiol 273: H1452-H1463, 1997[Abstract/Free Full Text].

46.   Zhang, J, Merkle H, Hendrich K, Garwood M, From AHL, Ugurbil K, and Bache RJ. Bioenergetic abnormalities associated with severe left ventricular hypertrophy. J Clin Invest 92: 993-1003, 1993.

47.   Zhang, J, Murakami Y, Zhang Y, Cho YK, Ye Y, Gong G, Bache RJ, Ugurbil K, and From AHL Oxygen delivery does not limit cardiac performance during high work states. Am J Physiol Heart Circ Physiol 277: H50-H57, 1999[Abstract/Free Full Text].

48.   Zhang, J, Path G, Chepuri V, Homans DC, Merkle H, Hendrich K, Ugurbil K, Bache RJ, and From AHL Effects of dobutamine on myocardial blood flow, contractile function, and bioenergetic responses distal to a coronary stenosis. Implications with regard to dobutamine stress testing. Am Heart J 129: 330-342, 1995[Web of Science][Medline].

49.   Zhang, J, Path G, Chepuri V, Xu Y, Yoshiyama M, Bache RJ, From AHL, and Ugurbil K. Response of the myocardial high energy phosphates and wall thickening to prolonged regional hypoperfusion induced by sub-total coronary stenosis. Magn Reson Med 30: 28-37, 1993[Web of Science][Medline].

50.   Zhang, J, Shor L, Yoshiyama M, Merkle H, Garwood M, Homans DC, Bache RJ, Ugurbil K, and From AHL Hyperperfusion and cardioplegia effects on myocardial high energy phosphate distribution and energy expenditure. Am J Physiol Heart Circ Physiol 267: H894-H904, 1994[Abstract/Free Full Text].

51.   Zhang, J, Toher C, Erhard M, Zhang Y, Ugurbil K, Bache RJ, Lange T, and Homans DC. Relationships between myocardial bioenergetic and left ventricular function in hearts with volume-overload hypertrophy. Circulation 96: 334-343, 1997[Abstract/Free Full Text].

52.   Zhang, J, Wilke N, Wang Y, Zhang Y, Wang C, Eijgelshoven MHJ, Cho YK, Murakami Y, Ugurbil K, Bache RJ, and From AHL Functional and bioenergetic consequences of post-infarction left ventricular remodeling in a new porcine model: an MRI and ³¹P MRS study. Circulation 94: 1089-1100, 1996[Abstract/Free Full Text].