In vivo mouse myocardial ³¹P MR spectroscopy using 3D ISIS: technical considerations and biochemical validations

Abstract

Phosphorous-31 magnetic resonance spectroscopy (³¹P-MRS) provides a unique noninvasive window into myocardial energy homeostasis. Mouse models of cardiac disease are widely used in preclinical studies, but application of ³¹P-MRS in the in vivo mouse heart has been limited. The small-sized, fast-beating mouse heart imposes challenges regarding localized signal acquisition devoid of contamination with signal originating from surrounding tissues. Here, we report the implementation and validation of 3D Image Selected In vivo Spectroscopy (ISIS) for localized ³¹P-MRS of the in vivo mouse heart at 9.4 T.

Cardiac ³¹P MR spectra were acquired in vivo in healthy mice (n = 9) and in transverse aortic constricted (TAC) mice (n = 8) using respiratory-gated, cardiac-triggered 3D ISIS. Localization and potential signal contamination were assessed with ³¹P-MRS experiments in the anterior myocardial wall, liver, skeletal muscle and blood. For healthy hearts, results were validated against ex vivo biochemical assays. Effects of isoflurane anesthesia were assessed by measuring in vivo hemodynamics and blood gasses.

The myocardial energy status, assessed via the phosphocreatine (PCr) to ATP ratio, was ~25% lower in TAC mice compared to controls (0.76 ± 0.13 vs. 1.00 ± 0.15; P < 0.01). Localization with 1D ISIS resulted in two-fold higher PCr/ATP ratios than measured with 3D ISIS, due to high PCr levels of chest skeletal muscle that contaminate the 1D ISIS measurements. Ex vivo determinations of the myocardial PCr/ATP ratio (0.94 ± 0.24; n = 8) confirmed in vivo observations in control mice. Heart rate (497 ± 76 beats/min), mean arterial pressure (90 ± 3.3 mmHg), and blood oxygen saturation (96.2 ± 0.6 %) during the experimental conditions of in vivo ³¹P-MRS were within the normal physiological range.

Our results show that respiratory-gated, cardiac-triggered 3D ISIS allows for noninvasive assessments of in vivo mouse myocardial energy homeostasis with ³¹P-MRS under physiological conditions.

Introduction

Mouse models are widely used in preclinical studies on the pathogenesis of cardiomyopathies. Disturbed myocardial energy metabolism has been identified as an important contributor to the development of cardiomyopathy (1). Assessment of the myocardial energy status is therefore instrumental in characterizing disease progression or treatment response. The high-energy phosphates adenosine 5’-triphosphate (ATP) and phosphocreatine (PCr) are essential for providing energy for cellular processes such as sarcomere contraction in cardiomyocytes, and for energy transport and buffering. The inherent instability of high-energy phosphates compromises accurate assessment of the myocardial energy status with biochemical techniques, which require disruptive or terminal biopsies, precluding longitudinal in vivo investigations. Many magnetic resonance imaging (MRI) and, to a lesser extent, spectroscopy (MRS) methods have been introduced to study the in vivo mouse heart noninvasively (2–4). These methods allow for longitudinal studies of disease progression and assessments of the effects of therapeutic strategies.

Phosphorous-31 MRS (³¹P-MRS) is the only method that provides a noninvasive window into in vivo high-energy phosphates (1). Localized signal acquisition is essential to restrict the obtained spectrum to the heart, excluding signal from nearby liver tissue or chest skeletal muscle. Localization methods for ³¹P-MRS include single-voxel as well as chemical shift imaging (CSI) approaches. ³¹P-CSI allows for an assessment of regional differences in myocardial energy status (5), but is susceptible to intervoxel signal contamination due to Fourier bleeding (6). In contrast, single-voxel localization with 3D Image Selected In vivo Spectroscopy (ISIS) (7) leads to a better defined voxel shape (8), but voxel size is typically much larger compared to CSI, and commonly includes the whole left ventricle (LV) (9).

Localized ³¹P-MRS of the in vivo mouse heart is very challenging due the small organ size (~100 mg), the high heart rate (~500 beats/min), and the intrinsically low sensitivity of ³¹P-MRS. Methods for in vivo 1D and 2D ³¹P-CSI of the mouse heart were initially demonstrated in healthy mice (10) and in a transgenic mouse model for cardiomyopathy (11). These experiments were performed at a constant TR, which is essential for a reliable signal quantification. None of these methods used cardiac triggering or respiratory gating to account for physiological motion of the tissue of interest. Together with the effects of Fourier bleeding in CSI methods, tissue displacement could lead to contamination of the spectrum with signal from the liver, blood, and/or chest skeletal muscle.

One early study describes the application of cardiac-triggered 3D ISIS for in vivo ³¹P-MRS of the mouse heart (9). While cardiac triggering was used in those experiments to synchronize the acquisitions with the cardiac cycle, no measures were taken to ensure a constant TR. When applying ISIS under partially saturated conditions (i.e., if TR < 5 × T₁), variations in TR can lead to signal contamination (12,13) as well as modulations of signal amplitudes due to T₁-dependent partial saturation effects. Those early experiments were performed at 2.35 T and required a lengthy experimental time of nearly 3 hours (9). At higher field strengths, scan time can possibly be reduced to acceptable values while maintaining sufficient SNR for signal quantification.

Here, we report the implementation of 3D ISIS for single-voxel localized ³¹P-MRS of the in vivo mouse heart at 9.4 T. Because 3D ISIS is a multi-shot localization method and hence particularly sensitive to motion artifacts, we employed both respiratory gating and cardiac triggering whilst maintaining a steady state of magnetization with dummy excitations during respiratory gates to ensure a constant TR (14). Results were validated against ex vivo biochemical assays of myocardial PCr and ATP concentrations. Hemodynamics and blood gasses were measured to assess potential effects of isoflurane anesthesia on the cardiovascular physiology. To demonstrate suitability for cardiac applications, the method was applied to a well-characterized mouse model of heart failure (15,16).

Methods

Animals

All procedures were approved by the Animal Ethics Committee of Maastricht University (Maastricht, The Netherlands). Male C57BL/6 mice (n = 8; body weight = 26.2 ± 2.6 g) underwent transverse aortic constriction (TAC) surgery as described previously (17). Anesthesia was induced using 2-3% isoflurane in 0.4 L/min flow of 1:1 O₂:medical air, after which mice were intubated for mechanical ventilation. Analgesia was provided using buprenorphine (0.1 mg/kg subcutaneous). An incision was made above the first intercostal space to gain access to the aortic arch. The aorta was tied off together with a 27G needle between the innominate artery and the left common carotid artery with a 6-0 silk suture. Subsequent needle removal left an aortic stenosis, inducing LV pressure-overload. The chest was closed and the intubation tube was removed to allow full recovery. Seven weeks after surgery, MR data were acquired as described below. Healthy mice (n = 9; body weight = 24.4 ± 2.0 g) served as controls. Following the MR measurements, anesthetized mice were sacrificed by exsanguination. Blood was collected in EDTA tubes for ex vivo analysis with ³¹P-MRS.

MR protocol

Mice were anesthetized with 2-3% isoflurane in 0.4 L/min flow of medical air and positioned prone in a purpose-built support cradle above a custom-built, actively decoupled, two-turn ³¹P surface coil (Ø 15 mm) for signal reception. Anesthesia was maintained with 1.2-1.6% isoflurane in a continuous flow of 0.4 L/min medical air. The front paws were taped onto gold-coated ECG electrodes integrated in the anesthesia mask. A respiratory balloon was positioned underneath the lower abdomen. Vital signs were monitored and used for MR gating and triggering by the SA Monitoring and Gating System 1025 (Small Animal Instruments, Stony Brook, NY, USA). Mouse body temperature was maintained using a heating pad with integrated warm water flow, and monitored with an external abdominal fiber optic probe. The setup was inserted into a horizontal-bore 9.4 T magnet (Magnex Scientific, Oxon, UK), interfaced to a Bruker Avance III console (Bruker Biospin MRI, Ettlingen, Germany), and controlled by the ParaVision 5.0 software package (Bruker Biospin). The system was equipped with a 740 mT/m gradient set, and a volume coil (Ø 54 mm) composed of a quadrature ¹H birdcage resonator and a linear ³¹P birdcage resonator (RAPID Biomedical, Rimpar, Germany), used for ¹H MRI and shimming, and for radiofrequency transmission for ³¹P-MRS, respectively.

Scout ¹H MR images were acquired to confirm positioning of the heart within the sensitive area of the ³¹P surface coil. A segmented, prospectively cardiac-triggered, respiratory-gated fast low-angle shot sequence was used to acquire cine ¹H MR image series of 16-18 frames per cardiac cycle. Four 1-mm LV short-axis slices were complemented with 4- and 2-chamber long-axis views, and used for quantification of LV function and morphology as well as for anatomical reference during 3D ISIS voxel planning for localized ³¹P-MRS. Imaging parameters were: field of view 30 × 30 mm²; matrix = 128 × 128; TE = 1.8 ms; TR = 7 ms; flip angle = 15°; number of averages (NA) = 4. Total acquisition time was ≈ 20 minutes.

Subsequently, an 11 × 11 × 11 mm³ voxel in the sensitive area of the surface coil was shimmed manually by minimizing the ¹H₂O line width, acquired with a respiratory-gated, cardiac-triggered point resolved spectroscopy sequence (18). Calibration of the ³¹P sinc excitation pulse (pulse length = 1.2 ms; bandwidth = 32.0 ppm) was performed by varying pulse power to achieve maximal signal from a spherical phantom (Ø 5 mm; 15 M phosphoric acid) positioned underneath the ³¹P surface coil. After removal of the phantom, unlocalized pulse-acquire ³¹P MR spectra were obtained from a subset of animals (n = 5 per group) to assess metabolite T₁ values using conventional saturation-recovery experiments. Parameters were: 1.2 ms 90° sinc excitation pulse; bandwidth = 32.0 ppm; γ-ATP on resonance; TR = 500, 1000, 2000, 4000, 6000, and 15000 ms; NA = 1024 - 32.

Next, a respiratory-gated, cardiac-triggered 3D ISIS sequence was used for localized cardiac ³¹P-MRS. Dependent on heart size, a 3D ISIS voxel (TAC: 326 ± 43 μL vs. control: 175 ± 8.8 μL) was positioned to enclose the end-diastolic LV, carefully excluding the liver and chest skeletal muscle (Figure 1A-B). 3D ISIS parameters were: TR ≈ 2 seconds; NA = 768 (96 3D ISIS cycles) preceded by 1 dummy cycle; 6.25 ms 180° adiabatic hyperbolic secant inversion pulses (bandwidth = 37.5 ppm); 1.2 ms 90° sinc excitation pulse (bandwidth = 32.0 ppm); γ-ATP on resonance; 2048 acquisition points. Although the excitation pulse was calibrated, the potential effect of spatial contamination by so-called ‘T₁ smearing’ due to inhomogeneous excitation pulses (19) was further minimized by choosing the least-optimal inversion order in the left-right orientation. This avoided contamination of the spectra with signal from chest skeletal muscle (anterior-posterior orientation) or the liver (head-feet orientation). Triggering was timed at ECG R-wave upslope detection. Respiratory gating causes fluctuations in effective TR, leading to variations in longitudinal magnetization between subsequent acquisitions. If longitudinal magnetization is not equal at the start of all eight acquisitions within one 3D ISIS cycle, cancellation of unwanted signals in the addition/subtraction scheme is incomplete. Thus, when measuring at TR < 5 × T₁, a constant TR is required to minimize signal contamination. Moreover, measuring at constant TR prevents complex modulations of signal amplitudes that hamper corrections for partial saturation effects. Therefore, we performed unlocalized dummy excitation pulses during respiratory gates to achieve an essentially constant TR of ≈ 2 seconds. Acquisition time was < 40 minutes. In a subset of healthy mice (n = 6), 3D ISIS was also performed at TR ≈ 15 seconds; NA = 192 (24 cycles). These measurements were used to verify the partial saturation correction factor obtained with unlocalized saturation-recovery experiments.

Figure 1.

End-diastolic left ventricular (LV) MR images obtained from a control mouse (A) and a mouse with a transverse aortic constriction (TAC) (B). The constriction is indicated by the arrow. Dilated hypertrophic cardiomyopathy is evidenced by increased LV wall thickness and LV cavity volume in the TAC mouse. Rectangles indicate the voxels selected for localized ³¹P-MRS with 3D ISIS. Panels C and D display ³¹P MR spectra acquired in vivo with 3D ISIS in a healthy mouse heart and a TAC heart, respectively. Myocardial PCr/γ-ATP, corrected for partial saturation, was lower in TAC mice (n = 8) compared to healthy controls (n = 9) (E). Data are expressed as mean ± SD. **, P < 0.01. α-, β-, γ-ATP, α-, β-, and γ-phosphate groups in ATP; 2,3-DPG, 2,3-diphosphoglycerate; PCr, phosphocreatine; P_i, inorganic phosphate.

To compare 3D ISIS localization with 1D localization, we acquired spectra using 1D ISIS of the anterior myocardial wall in a slice essentially parallel to the surface coil (Figure 2A) in a subset of mice (n = 6). Acquisition parameters were kept similar to the 3D ISIS approach, except for NA = 384 (192 1D ISIS cycles) and slice thickness = 1 mm. Furthermore, to obtain spectra solely from the myocardium and to rule out any contamination with signal from LV cavity blood, 3D ISIS was performed with a small voxel (1 × 3 × 3 mm³) positioned within the anterior myocardial wall (Figure 2B). Acquisition time was ≈ 2.5 hours with NA = 3072 (384 cycles).

Figure 2.

Comparisons of 1D ISIS and 3D ISIS localization for cardiac ³¹P-MRS. Geometrical localization for (A) 1D ISIS in the anterior myocardial wall (slice thickness = 1 mm), indicated in transversal and sagittal reference images. Circle and ellipses indicate the position of the ³¹P surface coil. 3D ISIS voxel localization in (B) the anterior myocardial wall (1 × 3 × 3 mm³) and (C) enclosing the whole LV (6 × 6 × 6 mm³). The corresponding ³¹P MR spectra from (D) 1D ISIS (NA = 384), (E) 3D ISIS in the myocardium (NA = 3072), and (F) 3D ISIS of the whole LV (NA = 384). All spectra were acquired in the same mouse at TR = 2 seconds. (G) PCr/γ-ATP ratio assessed via 1D ISIS (as in A) and 3D ISIS of the whole LV (as in C). **, P < 0.01 (two-sided paired t-test, n = 6). α-, β-, γ-ATP, α-, β-, and γ-phosphate groups in ATP; 2,3-DPG, 2,3-diphosphoglycerate; PCr, phosphocreatine; P_i, inorganic phosphate. SA: short axis, 4-ch LA: 4-chamber long-axis, 2-ch LA: 2-chamber long-axis.

Additionally, spectra from liver and hind limb skeletal muscle (n = 3) were acquired in vivo with respiratory-gated and cardiac-triggered 3D ISIS (Figure 3) to evaluate the performance of 3D ISIS localization with respect to in vivo ³¹P-MRS of these tissues as reported in the literature (20,21).

Figure 3.

Voxel positioning for 3D ISIS in (A) the liver (5 × 5 × 5 mm³) and in (B) hind limb skeletal muscle (4 × 4 × 4 mm³) in the mouse, and the corresponding ³¹P MR spectra for (C) the liver (NA = 768) and (D) skeletal muscle (NA = 384). Note the prominent resonance of PCr detected in skeletal muscle, whereas it is absent in liver tissue. α-, β-, γ-ATP, α-, β-, and γ-phosphate groups in ATP; PCr, phosphocreatine; P_i, inorganic phosphate.

Finally, to investigate the contribution of blood metabolites to the cardiac ³¹P MR spectra acquired in vivo, spectra of fresh blood were measured ex vivo using a pulse-acquire sequence. A vial (n = 6) with ~1 mL of blood in EDTA was positioned just over the surface coil and maintained at 37°C by a heating pad. Parameters were: 1.2 ms 90° sinc excitation pulse; γ-ATP on resonance; TR = 2000 ms; NA = 512.

Image analysis

LV cavity and myocardial wall volumes were quantified by semiautomatic segmentation of the cine images (Pie Medical Imaging, Maastricht, The Netherlands) as described previously (22), yielding LV end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), ejection fraction (EF), and LV myocardial mass.

³¹P-MRS data analysis

Fitting of the metabolite signals to Lorentzian line shapes was performed in the time domain using AMARES in jMRUI (23). The PCr resonance at 0.00 ppm was used as an internal chemical shift reference. The ATP resonances at −2.48 ppm (γ; doublet), −7.52 ppm (α; doublet) and −16.26 (β; triplet) were fitted with equal amplitudes and line widths within each multiplet, and a J-coupling constant of 17 Hz. The γ-ATP line widths (LW_γ-ATP) were constrained relative to the PCr line width (LW_PCr) according to an empirically determined relation: LW_γ-ATP = LW_PCr + 14.85 Hz (n = 63; r = 0.78; P < 0.001). Blood 2,3-diphosphoglycerate (2,3-DPG) resonances obscured the inorganic phosphate (P_i) resonance. Therefore, these signals were fitted with two peaks: one for 2,3-DPG_{5.4 ppm} and P_i at 5.4 ppm, and one for 2,3-DPG_{6.3 ppm} at 6.3 ppm.

Saturation-recovery curves of PCr, γ-ATP and α-ATP were fitted by a mono-exponential function to estimate the corresponding longitudinal relaxation rate constants R₁. Mean R₁ values were used to determine metabolite T₁ values via R₁ = 1/T₁.

The in vivo myocardial energy status was expressed as the PCr/γ-ATP ratio, corrected for partial saturation.

Because blood contains ATP, but no PCr (24), signal contamination from blood in in vivo cardiac ³¹P MR spectra may lead to underestimation of the myocardial PCr/γ-ATP ratio. Therefore, we evaluated whether the PCr/γ-ATP ratio in the current work could be affected by blood contamination. The [γ-ATP/2,3-DPG_{6.3 ppm}]_blood ratio in spectra acquired in fresh blood was determined as a measure for ATP content in blood. The contribution of resonances from blood metabolites to in vivo spectra was assessed by quantifying the [2,3-DPG_{6.3 ppm}/γ-ATP]_LV ratio in the cardiac 3D ISIS spectra. Next, the relative contribution of signal from ATP in the blood to the ATP signal obtained with in vivo 3D ISIS was calculated as: [γ-ATP/2,3-DPG_{6.3 ppm}]_blood × [2,3-DPG_{6.3 ppm}/γ-ATP]_LV × 100%.

Validations

To test our in vivo ³¹P-MRS method against conventional biochemical assays in ex vivo tissue samples, validation measurements were conducted in a separate cohort of healthy mice (n = 4). Following the acquisition of 3D ISIS-localized myocardial ³¹P MR spectra as described above, the heart was immediately snap frozen upon thoracotomy. Myocardial high-energy phosphate concentrations were determined spectrophotometrically as described previously (25).

To monitor the effects of anesthesia on hemodynamics and blood gasses, healthy mice (n = 5) underwent cannulation of the carotid artery after induction of isoflurane anesthesia. Mean arterial pressure (MAP) and heart rate were monitored using a heparinized saline-filled catheter connected to a blood pressure transducer (Baxter TruWave PX600F, Edwards, Irvine, CA, USA), and recorded using LabVIEW 5.1 (National Instruments, Austin, TX, USA) as described previously (26). Each experiment was continued for 1.5 hours, and mimicked the experimental conditions of the in vivo ³¹P-MRS acquisitions in terms of mouse body temperature, heart and respiratory rates, and anesthesia maintenance as described above. After 1.5 hours, a 100-μL arterial blood sample was obtained via the cannula and subsequently analyzed using a blood gas analyzer (RAPIDPoint 500, Siemens Healthcare, Erlangen, Germany). Immediately thereafter, the anesthetized mice were sacrificed by snap freezing the heart upon thoracotomy for ex vivo spectrophotometric assay of PCr and ATP concentrations (25).

Statistical analyses

Data are reported as mean ± SD. The statistical significance of differences was analyzed using two-sided paired or unpaired t-tests, as appropriate. The level of significance was set at P < 0.05.

Results

MRI: LV hypertrophy in TAC mice

We assessed in vivo LV morphology and function from cine MR images to confirm the hypertrophic phenotype and impaired cardiac performance in TAC mice (Table 1, Figure 1A-B). LV mass was 95% higher in TAC mice compared to healthy mice (P < 0.001), indicating LV hypertrophy in TAC mice. Concomitantly, EDV (P < 0.001) and ESV (P < 0.001) were higher in TAC mice compared to controls. This translated in a lower SV (−33%, P < 0.001) and EF (−65%, P < 0.001) in TAC mice. Combined, these data illustrate the development of dilated hypertrophic cardiomyopathy with severe systolic dysfunction after 7 weeks of LV pressure-overload in TAC mice.

Table 1.

LV morphology and functional parameters obtained with MRI in control mice and TAC mice.

	Control (n = 9)	TAC (n = 8)	P
LV mass (mg)	90.0 ± 14.9	176 ± 19.1	***
LV mass/body weight (mg/g)	3.7 ± 0.6	6.0 ± 0.9	***
Heart rate (beats/min)	495 ± 50	539 ± 31	NS
End-diastolic volume (μL)	63.7 ± 10.8	123 ± 26.1	***
End-systolic volume (μL)	20.8 ± 5.4	94.1 ± 24.8	***
Stroke volume (μL)	43.0 ± 7.3	28.7 ± 6.6	***
Ejection fraction (%)	67.6 ± 5.6	23.9 ± 5.2	***

Data are expressed as mean ± SD. Effect of TAC (two-sided unpaired t-test):

^***P < 0.001; NS, not significant.

³¹P-MRS: in vivo myocardial energy status

Typical ³¹P MR spectra acquired with 3D ISIS in a healthy mouse heart and a TAC heart are shown in Figure 1C-D. Resonances of PCr, and ATP are indicated. Inorganic phosphate (P_i, ~5 ppm) was obscured by 2,3-DPG arising from blood. Spectral line width for PCr was 0.37 ± 0.15 ppm.

Conventional pulse-acquire ³¹P MR saturation-recovery experiments of the mouse chest were performed in order to estimate the high-energy phosphate metabolite T₁ relaxation times at 9.4 T. T₁ values did not differ between control mice and TAC mice, and were 2.54 ± 0.41 s for PCr, 1.45 ± 0.25 s for γ-ATP, and 1.09 ± 0.31 s for α-ATP. Given a TR of 2 seconds, this resulted in a partial saturation correction factor for PCr/γ-ATP of 1.37 in 3D ISIS experiments. In healthy mice (n = 6), 3D ISIS was also performed under fully relaxed conditions. These localized acquisitions yielded a partial saturation correction factor for myocardial PCr/γ-ATP of 1.38 ± 0.28 for spectra acquired at TR = 2 seconds, which is in good agreement with the value derived from the unlocalized pulse-acquire saturation-recovery experiments. Myocardial PCr/γ-ATP, corrected for partial saturation, was lower in TAC mice compared to healthy controls (0.76 ± 0.13 vs. 1.00 ± 0.15; P < 0.01, Figure 1E), which is indicative of a compromised myocardial energy homeostasis in TAC mice.

Localization of a slice containing anterior myocardial wall with 1D ISIS (Figure 2A) systematically resulted in higher PCr/γ-ATP ratios than those obtained with 3D ISIS of the whole LV (2.12 ± 0.61 vs. 1.08 ± 0.25; P < 0.01, Figure 2G). Skeletal muscle tissue surrounding the anterior myocardial wall within the sensitive area of the surface coil (Figure 2A) likely contributed to the higher PCr/γ-ATP ratio observed with 1D ISIS. This was corroborated by spectra (Figure 2E) obtained with a very small 9-μL 3D ISIS voxel positioned in the anterior myocardial wall (Figure 2B), which qualitatively confirm the PCr/γ-ATP ratio of ~1 for healthy myocardium.

³¹P-MRS of liver, skeletal muscle, and blood

Using the respiratory-gated and cardiac-triggered 3D ISIS approach, we acquired ³¹P MR spectra from in vivo mouse liver and hind limb skeletal muscle. Absence of PCr in liver tissue was confirmed (Figure 3C), illustrating that essentially no signal from PCr in adjacent skeletal muscle contaminated the spectra that were obtained from the liver with 3D ISIS. Spectra obtained from hind limb skeletal muscle yielded a PCr/γ-ATP ratio of 3.59 ± 0.58 (Figure 3D), which is typical for healthy mouse skeletal muscle in resting conditions (21,27).

In spectra obtained from fresh blood (Figure 4), the blood ATP content was estimated via the [γ-ATP/2,3-DPG_{6.3 ppm}]_blood ratio, which was 0.22 ± 0.14. Contribution of signal from metabolites in the blood to the cardiac spectra, estimated via [2,3-DPG_{6.3 ppm}/γ-ATP]_LV, was similar for controls and TAC mice, and was 0.18 ± 0.10. The relative contribution of signal from ATP in the blood to the ATP signal in cardiac 3D ISIS spectra was therefore approximately 4%. These results show that contamination of the 3D ISIS spectra with signal from ATP in LV blood is marginal, and only minimally affects the myocardial PCr/γ-ATP ratio.

Figure 4.

³¹P MR spectrum obtained from fresh blood. 2,3-DPG, 2,3-diphosphoglycerate; PDE, phosphodiesters; α-, β-, γ-ATP, α-, β-, and γ-phosphate groups in ATP.

Validations

The in vivo PCr/γ-ATP ratio measured with 3D ISIS-localized ³¹P-MRS in healthy control mice (1.00 ± 0.19; n = 19) matched conventional ex vivo spectrophotometric assays in myocardial tissue (0.94 ± 0.24; n = 8; Figure 5A). In a subset of healthy mice, both in vivo and ex vivo measurements were performed. Spectrophotometric assays in myocardial tissue mirrored in vivo ³¹P-MRS measurements of myocardial PCr/γ-ATP ratios in the same hearts (1.00 ± 0.32 vs. 0.94 ± 0.14; n = 4; Figure 5B). Concentrations of PCr and ATP in ex vivo myocardial tissue samples (n = 8) were 18.3 ± 5.7 mmol/kg dry weight and 19.3 ± 2.5 mmol/kg dry weight, respectively.

Figure 5.

Comparison (A) of in vivo 3D ISIS-localized ³¹P-MRS measurements (n = 19) and conventional ex vivo spectrophotometric assays (n = 8) of the myocardial PCr/ATP ratio in healthy mice. Data are expressed as mean ± SD. In a subset of healthy mice (n = 4), both ³¹P-MRS and biochemical assays were performed. The difference in PCr/ATP ratio between the two methods is plotted against their mean per mouse in a Bland-Altman plot (B). Dashed lines indicate the limits of agreement (mean ± 2 SD) around the mean (solid line).

Mouse hemodynamics and blood gasses were measured after 1.5 hours of anesthesia to assess the potential effects of the isoflurane anesthetic regimen on the cardiovascular physiology. Throughout these experiments, which mimicked the experimental conditions of the ³¹P-MRS procedure, MAP (90 ± 3.3 mmHg) and heart rate (497 ± 76 beats/min) remained well within the physiological range for healthy C57BL/6 mice (28). Moreover, after 1.5 hours of anesthesia, blood oxygen saturation and other blood gas parameters were normal (Table 2). These data illustrate that the experimental conditions of the in vivo MR protocol have no major influence on mouse hemodynamics or blood gasses.

Table 2.

Hemodynamics and blood gasses measured in control mice after 1.5 hours of anesthesia (1.2-1.6% isoflurane in a continuous flow of 0.4 L/min medical air).

	C57BL/6 (n = 5)
Heart rate (beats/min)	497 ± 76
Mean arterial pressure (mmHg)	90 ± 3.3
pH	7.41 ± 0.03
pCO2 (mmHg)	31.6 ± 3.5
pO2 (mmHg)	110.7 ± 4.4
Total hemoglobin (mM)	9.2 ± 0.2
O2 saturation (%)	96.2 ± 0.6

Data are expressed as mean ± SD.

Discussion

We describe a noninvasive method to study the in vivo myocardial energy status in healthy and diseased mice using 3D ISIS-localized ³¹P-MRS. Acquisitions were respiratory-gated and cardiac-triggered, whilst maintaining steady-state conditions using dummy excitations to ensure accurate localization. We validated our method against conventional biochemical assays of PCr and ATP concentrations in ex vivo myocardial tissue collected immediately after acquiring the in vivo ³¹P MR spectra. Using a ¹H/³¹P MR setup, cardiac ³¹P-MRS with 3D ISIS was combined with cine ¹H MRI to assess morphological, functional, and metabolic parameters in a single experimental session of less than 2 hours. With this approach, we identified a reduced myocardial energy status, evidenced by a ~25% lower PCr/γ-ATP ratio, accompanied by hypertrophic growth and severely impaired myocardial function in a surgical mouse model of heart failure. These pathophysiological changes are consistent with previous studies (16,29).

The myocardial PCr/ATP ratio in healthy mice was ~1, measured both in vivo with ³¹P-MRS and ex vivo with spectrophotometric assays in the same hearts. This value is on the lower range of the normal PCr/ATP values reported in literature for humans (overall mean: 1.72 ± 0.26) (30). In humans, heart rate and cardiac work can increase up to three-fold during dobutamine stress (31) or exercise (32). In healthy mice, dobutamine infusion induced a heart rate increase of only ~39% (27), whereas running exercise increased mouse heart rates by only 40% to 90% (33,34). Indeed, it has been suggested that mice may have a lower cardiac energy reserve (35), given the narrower dynamic range of heart rates and cardiac work in mice compared to men.

Notably, the myocardial PCr/ATP ratio for healthy mice in the current work was lower than PCr/ATP values obtained in other in vivo ³¹P-MRS mouse studies (9–11). With measurements of physiological hemodynamics and blood gasses during 1.5 hours of anesthesia (28,36), we have ruled out adverse effects of the experimental conditions on mouse cardiovascular physiology that may cause a decreased PCr/ATP ratio. Instead, many technical aspects of localized ³¹P-MRS acquisition and quantification could contribute to the variability in the MRS-derived PCr/ATP values reported in literature (30), and should be taken into consideration when comparing between different laboratories (37,38). The three main causes for apparent discrepancies in literature reports are differences in 1) correction for partial saturation, 2) contamination of the spectra by signal from the liver or skeletal muscle tissue, and 3) contaminating blood in the LV cavity (30). Below, each of these issues are addressed for the current study.

Partial saturation correction

For the correction of partial saturation effects, we used metabolite T₁ values as determined by unlocalized pulse-acquire saturation-recovery experiments. Because the T₁ values in chest skeletal muscle could be different from those in cardiac muscle (39), we validated the correction factor in healthy mice by acquiring 3D ISIS-localized spectra from the LV myocardium under fully relaxed conditions. Indeed, the partial saturation correction factor derived from unlocalized saturation-recovery experiments was in agreement with measurements localized to the heart. This observation validates the assumption (40) that in the healthy mouse the T₁(PCr)/T₁(γ-ATP) ratio is essentially the same for chest skeletal muscle and myocardium at 9.4 T (11).

Minimizing signal contamination

Because 3D ISIS requires multiple acquisitions for signal localization, the method is particularly sensitive to motion artifacts and consequential contamination from tissues surrounding the heart such as liver and chest skeletal muscle. Additional contamination (‘T₁-smearing’) can be introduced by differences in longitudinal magnetization between subsequent acquisitions within one ISIS cycle due to imperfect flip angle of the excitation pulse combined with a TR < 5 × T₁ (12,13,19). Similar effects occur when TR is not constant between acquisitions, while measuring at TR < 5 × T₁. Previous studies in rodents (9) and in humans (41–43) using 3D ISIS for cardiac applications did not measure at constant TR for a steady state of magnetization, nor at fully relaxed conditions. The 3D ISIS sequence presented here uses a respiratory-gated and cardiac-triggered timing strategy to ensure localized inversion of the LV signal at identical cardiac and respiratory phases for all acquisitions, in combination with dummy excitations during respiratory gates to maintain a constant TR. In addition, by using the γ-ATP signal to determine the PCr/ATP ratio, we minimized the chemical shift displacement between PCr and ATP. PCr signal was absent in 3D ISIS-localized liver spectra, demonstrating that skeletal muscle located between the surface coil and the voxel of interest did not contaminate the spectra and that signal acquisition was effectively restricted to the liver. Indeed, localized ³¹P-MRS of the liver (20) will also benefit from the strategy proposed here (44).

Moreover, we demonstrated the effect of potential contamination of the spectra with signal from chest skeletal muscle by comparing 3D ISIS of the whole LV with 1D ISIS of a slice containing the anterior myocardial wall, similar to 1D CSI approaches reported previously (10,45). Localization within the selected slice is realized by surface coil positioning (Figure 2A), which may introduce signal from the surrounding tissue to the resultant spectrum. The PCr/γ-ATP ratio obtained with 1D ISIS was consistent with values observed with 1D CSI (10), but two-fold higher than the myocardial PCr/γ-ATP ratio measured with 3D ISIS.

Blood contamination

Blood in the LV cavity contains ATP, but no PCr. If signal from ATP in the blood contributes to the signal acquired from the voxel of interest, the myocardial PCr/ATP ratio may be underestimated (46). A correction factor can be applied to account for signal contribution from ATP in the blood to the spectra, which depends on both the amount of ATP in the blood and the amount of blood contributing to the spectrum. We showed that the amount of ATP signal from fresh blood, in terms of [γ-ATP/2,3-DPG_{6.3 ppm}]_blood, was approximately 20%, which is consistent with previous reports of ³¹P-MRS studies of blood in humans (24) and mice (9,47). The contribution of signal from metabolites in blood to the in vivo 3D ISIS spectra of the whole LV, estimated via [2,3-DPG_{6.3 ppm}/γ-ATP]_LV, was less than 20%, and was not different between healthy and TAC mice. The high velocity of flowing blood during acquisition may attenuate the peaks from blood metabolites in spectra obtained in vivo (48). Based on these observations, we estimated that the contribution of signal from ATP in the blood to the ATP signal in in vivo 3D ISIS spectra was only 4%, and concluded that correcting for blood contamination was not necessary in this study. Moreover, spectra obtained with 3D ISIS selecting a small voxel confined to the anterior myocardial wall were similar to those obtained from the whole LV, confirming minimal contribution of signal from blood to the latter.

Study limitations

With our current approach of determining the PCr/ATP ratio, it is not possible to detect similar decreases in both PCr and ATP concentrations. Absolute quantification of metabolite concentration will therefore be more sensitive to changes in cardiac energy metabolism. Indeed, it has been shown that in pressure-overload mouse hearts, not only myocardial PCr levels drop, but also ATP concentrations decrease, illustrating the added value of absolute quantification over the PCr/ATP ratio (45,49). Nonetheless, the PCr/ATP ratio has been used to identify perturbations in myocardial energy homeostasis in various mouse models of disease (16,50,51). Furthermore, a reduced PCr/ATP ratio has been shown to be an important indicator of disease severity (16).

A drawback of using single-voxel localized ³¹P-MRS of the entire LV is that myocardial energy status cannot be assessed at a regional scale, which is of importance in investigations of myocardial ischemia. In vivo 2D ³¹P-CSI of the mouse heart (11) would allow for mapping of myocardial energetics to assess differences between infarcted and remote regions in mouse models of myocardial ischemic insult. Nonetheless, because many preclinical animal studies focus on pathologies that have a global effect on the heart, such as aortic stenosis, diabetes, and inborn errors of metabolism, the current 3D ISIS approach for localized ³¹P-MRS can be a valuable addition to the toolbox of mouse cardiac MR methods (4).

Even at the high magnetic field strength of 9.4 T, we were not able to unambiguously detect and quantify myocardial P_i, because the signal was obscured by signal from 2,3-DPG in the blood. Potentially, reliable detection of P_i would provide additional metabolic insights, as decreasing PCr levels may be mirrored by increasing P_i levels. Additionally, the chemical shift of P_i could potentially be used to quantify in vivo myocardial pH. Notably, even in the human heart P_i is often undetectable, suggesting that myocardial P_i may only be partially MR visible (30).

In conclusion, the present work describes a noninvasive approach to assess myocardial energy status in the in vivo mouse using single-voxel 3D ISIS-localized ³¹P-MRS, which was validated by conventional biochemical assays in ex vivo myocardial tissue samples from the same hearts. The method encompasses a respiratory-gated, cardiac-triggered 3D ISIS sequence with dummy excitations during respiratory gates, to ensure a well-defined localization. This method identified differences in high-energy phosphate metabolism between the healthy mouse heart and a widely used model for heart failure, the TAC mouse. Importantly, results were obtained under physiological conditions, which are difficult or impossible to achieve with disruptive methods such as open-thorax protocols or isolated perfused-heart setups. Furthermore, the 3D ISIS-localized spectra can be obtained within 40 minutes, leaving room for measurements of cardiac function with MRI during the same experimental session. We anticipate that localized ³¹P-MRS will provide valuable contributions to preclinical investigations of cardiac disease progression and therapeutic intervention efficacy.

List of Abbreviations

α-, β-, γ-ATP

α-, β-, and γ-phosphate groups in adenosine 5’-triphosphate

CSI

chemical shift imaging

2,3-DPG

2,3-diphosphoglycerate

ECG

electrocardiogram

EDV

end-diastolic volume

EF

ejection fraction

ESV

end-systolic volume

ISIS

image selected in vivo spectroscopy

LV

left ventricle

LW

line width

MAP

mean arterial pressure

MRI

magnetic resonance imaging

NA

number of averages

PCr

phosphocreatine

PDE

phosphodiesters

P_i

inorganic phosphate

³¹P-MRS

phosphorous-31 magnetic resonance spectroscopy

SNR

signal-to-noise ratio

SV

stroke volume

TAC

transverse aortic constriction