Cardiac Magnetic Resonance Spectroscopy

Introduction

Cardiovascular magnetic resonance (CMR) imaging uses the ¹ H nucleus in water (H ₂ O) and fat (CH ₂ and CH ₃ groups) molecules as its only signal source, and therefore offers little insight into the biochemical state of cardiac tissue. In contrast, MR spectroscopy (MRS) of the heart allows the study of many other nuclei. It is the only available method for the noninvasive assessment of cardiac metabolism without needing the application of external radioactive tracers. Information on the major nuclei of interest for the metabolic study of cardiac tissue by MRS is given in Table 9.1 , including ¹ H, ¹³ C, ²³ Na, and ³¹ P. Although, theoretically, many clinical questions can be answered with cardiac MRS, the main reason why MRS has not yet fulfilled its potential in clinical cardiology is related to the fundamental physical limitations of the method. MRS is often used to study nuclei other than ¹ H, which have a much lower intrinsic MR sensitivity than ¹ H. Furthermore, metabolite concentrations in vivo are typically in the mM (millimolar) range, which is several orders of magnitude lower than for water (approximately 80 M) or fat (often >1 M). Therefore the temporal and spatial resolution of MRS has so far remained far behind that of CMR imaging. Successful applications of MRS are those in which the unique metabolic insights of MRS are more important than obtaining high spatial or temporal resolution.

TABLE 9.1

Elements (Nuclei) Most Frequently Studied by Magnetic Resonance Spectroscopy

Nucleus	Natural Abundance	Relative MR Sensitivity	Myocardial Tissue Concentrations
¹ H	99.98%	100%	H ₂ O 110 M; up to ~90 mM (CH ₃ - ¹ H of creatine)
¹³ C	1.1%	1.6%	Labeled compounds, several mM
²³ Na	100%	9.3%	10 mM (intracellular); 140 mM (extracellular)
³¹ P	100%	6.6%	Up to ~18 mM (PCr)

MR , Magnetic resonance; PCr , phosphocreatine.

Although this chapter focuses on clinical cardiac MRS, some explanation of the experimental principles of MRS is important even for the clinical reader: MRS has been a widespread method in experimental cardiology, ever since the first ³¹ P-MR spectrum from an isolated heart was obtained by Radda's group in the 1970s, and since then experimental MRS studies have offered numerous fundamental insights into cardiac metabolism. Furthermore, only with an understanding of the major implications of experimental MRS studies are we able to fully appreciate the potential of the method and extrapolate to clinical cardiac MRS applications that should become feasible in the future, once the technical challenges presently limiting the clinical utility of MRS have been overcome. For those interested in greater detail on experimental MRS and methodological background, comprehensive reviews of the subject are available elsewhere. Complementary clinical reviews are also available.

Physical Principles

The basic principles of MRS (see the reference list for textbooks on the general physical principles of CMR and MRS ) are best explained from the most extensively studied nucleus, ³¹ P, and from the most widely used animal model, the isolated buffer-perfused rodent heart. These principles apply to MRS of all nuclei. An MR spectrometer consists of a high-field superconducting magnet (currently up to 23.5 T, where tesla [T] is the unit of magnetic field strength) with a bore size ranging between ~5 cm and ~1 m. The nucleus-specific probe head with the radiofrequency (RF) coils, which are used for MR excitation and signal reception, is seated within the magnet bore. The magnet is interfaced with a computer, a pulsed magnetic field gradient system, an RF transmitter, and RF receivers. MRS requires a much higher magnetic field homogeneity than CMR; therefore, before any MRS experiment, the magnetic field must first be homogenized with shim gradients. Spin excitation is achieved by transmitting a radiofrequency pulse through the RF coils into the subject. The resulting MR signal, known as the free induction decay (FID), is then received from the subject by RF coils and recorded by the scanner. The signal (or FID) oscillates with time but within an overall exponentially decaying “envelope.” It typically reaches negligible levels after tens of milliseconds. An MR spectrum is then computed from the FID (by applying the discrete Fourier transform formula ). The MR spectrum describes the signal intensity as a function of resonance frequency. Because of the low sensitivity of MRS, in practice, many FIDs are averaged to obtain MR spectra with an adequate “signal-to-noise ratio” (SNR; i.e., the height of a peak in the spectrum after “matched filtering” divided by the standard deviation of baseline noise ). The required number of averages depends on the concentration of the metabolite under investigation, its MR relaxation times (T1, T2, T2*), the “filling factor” (the mass of the heart relative to the coil size), the natural abundance of the nuclear isotope studied, its relative MR sensitivity (see Table 9.1 ), the pulse angle, and the pulse repetition time (TR). For a perfused rat heart experiment at ≥7 T, 100 to 200 FIDs are typically acquired. To quantify metabolite concentrations from MR spectra, it is vital to correct for partial saturation, which depends on the selected pulse angles and TR. This is because the maximum MR signal is only obtained when nuclei are excited from a fully relaxed spin state, that is, when a time of at least 5 × T1 has passed since a previous excitation (e.g., T1 of phosphocreatine [PCr] at 7 T is ~3 s requiring TR of ≥15 s, at 1.5 T the PCr T1 ~4.4 s requiring TR of ≥22 s); acquisition of fully relaxed MR spectra therefore requires long TRs, leading to prohibitively long acquisition times. In practice, we take advantage of the fact that the initial part of the FID contains most of the signal. Use of shorter TRs therefore yields spectra with higher SNR for a given acquisition time, but some of the signal is lost because of saturation. Such spectra are termed “partially saturated.” The extent of saturation also varies for different ³¹ P-resonances, because the T1s of ³¹ P-metabolites such as PCr and adenosine triphosphate (ATP) are significantly different (T1 of PCr is 2–3 × longer than T1 of ATP). Therefore for metabolite quantification from partially saturated spectra, the spectral peak areas are determined (e.g., fitted using the Advanced Method for Accurate, Robust, and Efficient Spectral Fitting [AMARES] or LCModel ), and then correction factors are applied to obtain the metabolite concentration. By comparing fully relaxed and saturated spectra, these factors can be determined for each metabolite. In practice, TRs and pulse flip angles for MRS are chosen to yield acceptable measurement times at an ~20% to 50% degree of saturation. Larger degrees of saturation, resulting from the use of extremely short TRs, make quantification of spectra unreliable because of uncertainties in the metabolite T1 values and the dynamic chemistry of these metabolites.

A ³¹ P-MR spectrum from an isolated, beating rat heart, obtained in 5 minutes at 7 T with a TR of 1.93 seconds and a pulse angle of 45 degrees, is shown in Fig. 9.1 . A typical cardiac ³¹ P-spectrum shows six resonances, corresponding to the three (γ, α, β) ³¹ P atoms of ATP (the resonance at the right shoulder of the α-ATP peak represents the ³¹ P atom in nicotinamide adenine dinucleotide [NAD ⁺ ]), PCr, inorganic phosphate (Pi), and phosphomonoesters (PME), mostly representing adenosine monophosphate (AMP) and glycolytic intermediates. The reason why only a small number of ³¹ P resonances is detectable, in spite of many more ³¹ P-containing metabolites being present in the heart, is that, for ³¹ P nuclei to be visible, they must be present above a certain concentration threshold of ~600 µM and free to tumble in solution. Largely immobilized metabolites such as plasma membrane phospholipids do not yield a quantifiable MR signal because of their very short T2 values; instead, these metabolites contribute to the broad “baseline hump” of ³¹ P spectra, which is particularly pronounced in the brain. Because of the phenomenon termed “chemical shift,” different metabolites resonate at distinct frequencies, allowing their discrimination from each other. Chemical shift is given in units of parts per million (ppm) relative to the resonance frequency of a reference compound (e.g., phosphoric acid H ₃ PO ₄ ). ³¹ P-metabolite resonances extend over a range of ~30 ppm. The physical basis of chemical shift is that nuclei at different positions in the molecule have different spatial distributions of electrons around them, which interact differently with the scanner's magnetic field. Fitting of spectral peak areas (e.g., with AMARES) and saturation correction gives metabolite concentrations in arbitrary units. Often metabolite ratios, such as the PCr/ATP ratio, are then calculated—implicitly assuming that the concentration of the metabolite in the denominator (ATP here) is well regulated enough to act as an endogenous reference. Alternatively, using an exogenous reference phantom, absolute metabolite concentrations can be computed. Frequently, phenylphosphonic acid (PPA) is used for this purpose because this generates an additional peak in the spectrum, which does not overlap with the cardiac ³¹ P resonances. The most significant advantage of MRS over destructive methods such as traditional biochemical assays, where the tissue has to be frozen and extracted, is that the MRS measurement is noninvasive. Therefore, spectra can be acquired sequentially, and the response of energy metabolites to ischemia, hypoxia, or inotropic stimulation can be followed longitudinally in the same tissue. With this approach, each heart can serve as its own control, yielding a more powerful experimental design and substantially reducing the number of required experiments.

FIG. 9.1, 31 P-magnetic resonance spectrum of an isolated, buffer-perfused rat heart obtained in 5 minutes at 7 T. ATP, Adenosine triphosphate; PCr, phosphocreatine; Pi, inorganic phosphate; PME, phosphomonoesters.

Experimental Foundations

³¹ P-MRS

With ³¹ P-MRS, cardiac high-energy phosphate metabolism, that is, the energetic state of the heart, can be monitored noninvasively. ATP is the direct energy source for all energy-consuming reactions in the heart. PCr acts as a temporal energy buffer and as an energy transport molecule in the “creatine kinase energy shuttle” ( Fig. 9.2 ). For this, the high-energy phosphate group is transferred from ATP to creatine in the mitochondria (where ATP is produced by ATP synthase), yielding PCr and adenosine diphosphate (ADP) in a reaction catalyzed by the mitochondrial creatine kinase isoenzyme (CK _mito ). PCr diffuses through the cytoplasm to the site of ATP utilization, the myofibrils, where the back reaction occurs (catalyzed by the myofibrillar-bound MM-creatine kinase isoenzymes; CK _MM ), ATP is reformed and is used for contraction. Free creatine then diffuses back to the mitochondria. The reaction rates at the creatine kinase (CK) isoenzymes complement each other to maintain this steady state. The second important function of PCr and creatine kinase is to control the thermodynamic state of the cell, that is, to maintain free cytosolic ADP at low concentration on exertion, instead of raising ADP and consuming Cr in the mitochondria, thereby up-regulating oxidative phosphorylation. This is a requirement for normal cardiac function, because ADP determines the free energy change of ATP hydrolysis (Δ G ; kJ/mol), a measure of the maximum amount of energy released from ATP hydrolysis (see Neubauer for details on the calculation of free ADP and Δ G from creatine kinase shuttle metabolites). In the normal heart, Δ G is ~−58 kJ/mol. Many intracellular enzymes such as SR-Ca ²⁺ -ATPase and others will not function properly above a threshold value for Δ G of about −52 kJ/mol. CK has also been implicated in minimizing cellular ADP loss, maintaining cellular pH, and activating glycogenolysis and glycolysis.

FIG. 9.2, Creatine kinase energy shuttle. For details, see text. ADP, Adenosine diphosphate; ATP, adenosine triphosphate; CK mito , mitochondrial creatine kinase isoenzyme; CK MM , MM-creatine kinase isoenzyme.

Pi is formed when ATP is hydrolyzed: ATP ⇌ ADP + Pi. Pi increases when ATP utilization exceeds ATP production, for example, during ischemia. Intracellular pH (pH _i ) can also be measured with ³¹ P-MRS, from the chemical shift difference between PCr and Pi, which is pH-sensitive because of shifts in the equilibrium H ₂ PO ₄ ⁻ ⇌ HPO ₄ ² ⁻ + H ⁺ near pH 7.

In principle, the intracellular magnesium ion concentration can also be measured by ³¹ P-MRS from the chemical shift difference between PCr and β-ATP.

Cardiac energy metabolism has been investigated by ³¹ P-MRS under various experimental conditions. For example, the effect of changes in cardiac workload on energetics has been examined. PCr levels do not change with moderate changes in workload, but decline with substantial increases in cardiac work. ATP content remains almost constant with varying workload and during the cardiac cycle, because the creatine kinase equilibrium favors ATP synthesis over PCr synthesis by a factor of ~100. Thus for any situation of myocardial stress, including ischemia, ATP only decreases when PCr levels are substantially depleted. This is the fundamental reason why the PCr/ATP ratio is a highly sensitive indicator of the energetic state of the heart. Changes of myocardial energy metabolism in experimental models of ischemia and reperfusion highlight the potential of ³¹ P-MRS for the detection of ischemia in the human heart. Fig. 9.3 shows an example of ³¹ P-MR spectra during control, ischemia, and reperfusion. After 15 minutes of total, global ischemia, ATP and PCr resonances have disappeared, and almost all the ³¹ P in heart is present as Pi and PME. During reperfusion, PCr and Pi show full, and ATP partial recovery. We showed that, when hearts are pretreated with endothelin-1, a hormone that increases susceptibility to ischemia, recovery of high-energy phosphate metabolism is impaired. By summing up data from several experiments, Clarke et al. demonstrated that the decrease of PCr and the increase of Pi were amongst the very earliest metabolic responses in myocardial ischemia, with significant changes occurring within seconds. Thus if we were successful in measuring energetics in human myocardium with high temporal and spatial resolution, we could directly image parameters that detect myocardial ischemia within seconds after its onset. No other diagnostic approach currently achieves this, although hyperpolarized ¹³ C imaging technology seems likely soon to offer similar insight earlier in the energetic chain.

FIG. 9.3, 31 P-magnetic resonance spectra from an untreated perfused rat heart and a heart treated with endothelin-1 (ET-1) during control, at the end of 15 minutes of ischemia and at the end of reperfusion. For details, see text. ATP, Adenosine triphosphate; PCr, phosphocreatine; Pi, inorganic phosphate.

With the magnetization (saturation) transfer method, the forward rate and velocity of the creatine kinase reaction, a measure of ATP transfer from mitochondria to myofibrils, can be measured in vivo. Creatine kinase reaction velocity correlates with cardiac workload and with recovery of mechanical function after ischemia. In animal models or skeletal muscle, magnetization transfer also allows measurement of the forward rate and velocity of the ATP hydrolysis reaction in the myofibrils as a measure of ATP consumption.

Experimental ³¹ P-MRS studies have substantially contributed to our understanding of the role of energetics in heart failure. Independent of the etiology of heart failure, the failing myocardium shows reduced PCr, unchanged or moderately (by <30%) reduced ATP, unchanged or increased Pi, and substantially reduced creatine kinase reaction velocity and flux. These changes are likely to contribute to the impairment of contractile reserve in failing myocardium, because of the failure to maintain appropriate Δ G values during inotropic stimulation.

¹ H-MRS

Other than ³¹ P, the nuclei with the greatest potential for clinical cardiac MRS are ¹ H and hyperpolarized ¹³ C (of which, more below). Protons have the highest MR sensitivity of all MR-detectable nuclei as well as high natural abundance (see Table 9.1 ). ¹ H is contained in a large number of metabolites, for example, creatine, lactate, carnitine, taurine, and −CH ₃ and −CH ₂ resonances of lipids. Measurement of total creatine by single voxel spectroscopy or navigator-gated echo planar spectroscopic imaging (EPSI) or potentially by creatine chemical exchange saturation transfer (CrCEST) imaging should, in conjunction with ³¹ P-MRS, allow the noninvasive determination of free ADP and Δ G of ATP hydrolysis, as demonstrated by Bottomley and Weiss in dogs and humans. By means of the oxymyoglobin and deoxymyoglobin resonances, tissue deoxygenation can be measured. Technical challenges for ¹ H-MRS include the need for suppression of the strong ¹ H signal from water and the complexity of ¹ H spectra with overlapping resonances, many of which remain to be characterized, and which means that excellent shimming and motion compensation are needed.

¹³ C-MRS

The ¹³ C nucleus has a low natural abundance (~1%), and for a classical ¹³ C-MRS experiment, the heart has to be supplied with ¹³ C-labeled compounds such as, for example, 1- ¹³ C-glucose. Cardiac substrate utilization, citric acid cycle flux, pyruvate dehydrogenase flux or beta-oxidation of fatty acids can then be quantified. Clinical cardiac studies have yet to be reported because of the low sensitivity of ¹³ C-MRS and the requirement for high concentrations of exogenous ¹³ C-labeled precursors.

However, the technique of hyperpolarization can increase the SNR of ¹³ C experiments by a factor of up to 10,000× for a few minutes, until the magnetization decays by T1 relaxation back to thermal equilibrium. This is a very active area of research as new pulse sequences, analysis methods, and injectable hyperpolarized agents are developed to capitalize on this extraordinary but brief burst of enhanced signal-to-noise. Pyruvate is a particularly attractive molecule to hyperpolarize because it plays a pivotal role in substrate uptake before oxidative phosphorylation, enabling fluxes into acetyl acetate, lactate, and bicarbonate to be quantified. These enable experimental assessment of the balance between fats, carbohydrates, and ketone bodies being metabolized by the heart, a balance which may well be disturbed in disease. Hyperpolarization studies in animal models are revisiting many of the systems previously characterized by ³¹ P-MRS to obtain further insight, for example, ischemia-reperfusion models.

²³ Na-CMR

²³ Na-CMR can evaluate changes in total and intracellular and extracellular Na ⁺ during cardiac injury. A cardiac ²³ Na spectrum shows a single peak representing the total Na ⁺ signal. To split the intracellular and extracellular Na ⁺ pools into two resonances, paramagnetic shift reagents, such as [TmDOTP] ⁵⁻ , are added to the perfusate. This method has been used experimentally to examine the mechanisms of intracellular Na ⁺ accumulation in ischemia-reperfusion injury, but ²³ Na-MR shift reagents are not yet available for clinical use. Experimental CMR of total ²³ Na shows that in acute ischemia, the total myocardial ²³ Na CMR signal increases because of the breakdown of ion homeostasis and the formation of both intracellular and extracellular edema (see Kim et al. and Horn et al. ). Furthermore, ²³ Na remains significantly elevated during chronic scar formation post coronary ligation because of the expansion of the extracellular space in scar, and the area of elevated ²³ Na signal correlates closely with histologically determined infarct size. Importantly, ²³ Na content is not elevated in stunned or hibernating myocardium. Thus ²³ Na CMR may allow detection of myocardial viability without the use of external contrast agents. Significant improvements in image quality were made at 3 T, and, recently, high-quality human ²³ Na images and retrogated cine movies have been acquired at 7 T in the human heart. An example of the high spatial and temporal resolution from 7 T is shown in Fig. 9.4 . Furthermore, Umathum et al. have recently presented the first results from a whole-body ²³ Na birdcage coil at 7 T, which can image from the pelvis to the heart in one scan. The outlook for cardiac sodium CMR has recently been reviewed by Bottomley.

FIG. 9.4, 23 Na-cardiovascular magnetic resonance of the human heart. (A) A four-chamber view with field of view (FOV) zoomed to (210 × 273) mm 2 and (B) a short-axis view (FOV = [180 × 150] mm 2 ). The ventricles and the septum are clearly visible in A and B. The left and right atriums are delineated from the ventricles in the four-chamber views. The rib cartilage shows the highest signal intensity because of the high sodium concentration. The temporal resolution is 200 ms, nominal spatial resolution (6 mm) 3 . The images were acquired with the ST acquisition scheme (10,000 projections) and reconstructed with a Hanning filter.

Other Nuclei

The availability of ultra-high field (7 T) CMR scanners is making it possible to perform human in vivo spectroscopy and imaging also of fluorine ( ¹⁹ F), which is not present naturally in the body in mobile form and therefore makes an attractive CMR tracer to study, for example, in drug uptake. Imaging of chlorine ( ³⁵ Cl) and potassium ( ³⁹ K) has also been demonstrated at 7 T, which makes it possible to map cell membrane potential in vivo.

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here

Introduction

Physical Principles

Experimental Foundations

31 P-MRS

1 H-MRS

13 C-MRS

23 Na-CMR