Phosphorus magnetic resonance spectroscopy and imaging (³¹P-MRS/MRSI) as a window to brain and muscle metabolism: A review of the methods

Abstract

Phosphorus magnetic resonance spectroscopy and spectroscopic imaging (³¹P-MRS/MRSI) non-invasively provide very important information regarding energy metabolism as they can detect high-energy metabolites and membrane phospholipids in vivo. They have repeatedly proven their utility in the study of healthy and disease conditions in skeletal muscle and brain tissue, as many disorders are related to imbalances in bioenergetics processes. However, they are not often used in a clinic setting, as there are technical challenges imposed by the low sensitivity and low concentration of metabolites leading to low signal to noise ratio (SNR), coarse spatial resolution and very long acquisition times. This paper presents an overview of the main techniques used for the acquisition, data reconstruction and processing of ³¹P-MRS/MRSI experiments with emphasis in methodological aspects, as well as some of their main applications in the study of skeletal muscle and brain tissue. Also, recent advances in the development of accelerated methods for acquisition of ³¹P-MR data are discussed.

Introduction

Phosphorus magnetic resonance spectroscopy (³¹P-MRS) and phosphorus magnetic resonance spectroscopic imaging (³¹P-MRSI) offer a very unique window to non-invasively look at tissue metabolism in vivo. These approaches are capable of tracking high energy metabolites and membrane phospholipids, all of which are involved in cellular energetic processes [1]. Thus phosphorus spectroscopic methods have been attractive across the scientific community in order to study energy metabolism in healthy and pathophysiological conditions. Furthermore, its noninvasive nature makes it an excellent candidate to perform repeated measurements and track disease progression or treatment response. In addition to the information contained in static spectra, these methods also allow evaluation of oxidative energy production by the mitochondria through an exercise challenge and subsequent recovery in skeletal muscle [2].

Although these methods have been around since the beginnings of MRI in the 1970’s, the current technological developments have been spearheaded by the ultra high magnetic field community. This includes improved radio-frequency (RF) coil designs, new data sampling methods and better reconstruction approaches all of which provide significant improvements to the technology making them more attractive to clinicians and researchers. This work provides an overview of the current status of ³¹P-MRS/³¹P-MRSI with focus on the technological developments and the main contributions in the study of skeletal muscle and brain metabolism.

The phosphorus nucleus is visible to nuclear magnetic resonance (NMR) due to its natural spin of ½. This isotope has a 100% natural abundance and albeit its relatively low sensitivity of 6.7% when compared to proton (¹H), it has encountered relevance when used for in vivo applications due to its presence in important high energy and phospholipid metabolites [1,3].

The chemical shift range where in vivo phosphorus compounds resonate spans around 30 ppm in contrast to the narrow 5 ppm window for ¹H MR spectra, additionally the amount of relevant metabolites is considerably less. Within the ³¹P-MRS spectrum, the fundamental high-energy metabolite phosphocreatine (PCr) is the dominant signal (in brain and skeletal muscle), thus it is often used as a reference and assigned the value of 0 ppm. PCr serves as an immediately available energy reserve ready to restore the expenditure of adenosine tri-phosphate (ATP) in energy demanding processes. ATP contains three phosphate groups; α, β and γ, that give rise to three independent signals in the spectrum at −7.56 ppm, −16.15 ppm and −2.53 ppm respectively (noting the intracellular magnesium (Mg²⁺) concentration will vary the resonant frequency of the β group). The interaction of the phosphorus nuclei with other nearby spins results in a homonuclear J-coupling effect, causing a line splitting ATP signals into doublets for γ-ATP and α-ATP, and a triplet for β-ATP. The ratios of peak heights are based on binomial expansion, hence doublets are 1:1 and the triplet 1:2:1.

Moving to the positive side of the spectrum, a relatively small singlet signal at approximately 4.85 ppm corresponds to inorganic phosphate (Pi), a degradation product of energy metabolism. This peak shifts in frequency according to intracellular pH (discussed below). Next to Pi, the cell membrane precursors phosphomonoesters (PMEs) phosphocholine (PC) and phosphoetanolamine (PE) are observable at around 6.2 ppm and 6.7 ppm, often times measured as a single PME signal at magnetic field strength B₀ ≤3 T. Similarly, phosphodiester (PDE) degradation products of phospholipid metabolism, glycerophosphocholine (GPC) and glycerophosphoethanolamine (GPE) are observed at 2.95 ppm and 3.50 ppm respectively. Depending on the signal to noise (SNR), other important phosphorus containing metabolites such as nicotinamide adenine dinucleotide (NAD) in its oxidized and reduced form (NAD+ and NADH, respectively) and uridine diphosphate glucose (UDPG) are also visible, specially at ultra-high field (i.e. 7 T).

Aside from these metabolic signals, other important information can be deduced from the analysis of the ³¹P spectrum. First, the chemical environment where phosphorus-containing metabolites reside may change with different physiological or pathological conditions. Of main relevance, the chemical shift of Pi is dependent on the intracellular pH [6], while the resonance of PCr remains constant, the resonance of Pi shifts with pH. Thus it is possible to calculate the pH using the difference in chemical shifts from PCr and Pi (δ) through the so-called modified Henderson-Hasselbach equation as follows:

$$pH(i)=p{K}_A+\log \left[\frac{(\delta -{\delta }_{HA})}{{\delta }_A-\delta }\right]$$

Here, $p{K}_A=6.75$ is the dissociation constant of Pi, ${\delta }_{HA}=3.27$ and ${\delta }_A=5.63$ are the chemical shifts of the protonated and non-protonated forms of Pi, respectively. As the main signal from Pi comes from cytoplasm (sarcoplasm in muscle tissue), it is the intracellular pH (pH(i)) that is measured. Second, the concentration of adenosine di-phosphate (ADP) under physiological conditions is too low to be detected using ³¹P-MRS, however it can be determined from the concentrations of creatine (Cr), ATP and PCr due to the chemical equilibrium in the creatine-kinase (CK) reaction [7], thus:

$$C_{ADP}=\frac{[Cr]\times [ATP]}{[PCr]\times [H^{+}]\times [K_{CK}]}$$

Where $K_{CK}=1.66\times {10}^9$ is the equilibrium constant of the CK reaction, assuming that PCr accounts for 85% of total Cr [4,5]. Finally, the cellular concentration of free Mg²⁺, a physiologically important divalent cation, influences the chemical shift of phosphorus compounds as the Mg complexes of ADP and ATP act as substrates for ATPases and kinases. Thus, the concentration of free Mg²⁺ can be deducted from the chemical shifts of β and α-ATP or PCr and β-ATP [8,9]. Fig. 1 shows examples of ³¹P-MRS spectra acquired at 3 T whereas Table 1 summarizes the metabolites, chemical shifts and their in vivo concentrations in skeletal muscle and brain tissue [4,5].

Prior to metabolite quantification, free induction decay (FID) curves (i.e. time domain signals) acquired from a ³¹P-MRS experiment are preprocessed. This step typically includes zero and first order phase corrections, zero filling to increase spectral resolution and spectral apodization, the latter applied mainly for filtering, especially when SNR is poor.

SNR is perhaps the most important characteristic in obtaining reliable metabolite quantification. If adequate, a common procedure for performing spectral fitting is to use Advanced Method for Accurate, Robust and Efficient Spectral Fitting (AMARES) [10], a time domain algorithm that incorporates prior knowledge and offers flexibility of the fitting parameters. The AMARES routine is available in the jMRUI software package [11] or as part of the MATLAB based OXSA toolbox [12]. Also, the popular MRS fitting package LCModel has been successfully used in the quantification of ³¹P spectra [13].

Accurate quantification of ³¹P metabolites in vivo is a challenging task as factors such as field inhomogeneity, relaxation times and coil sensitivities may influence the measurements. Thus, metabolite ratios are often used when reporting results from MRS experiments, especially in clinical literature. However, ratios are incapable of providing information about metabolic changes in pathological conditions, as they may be ambiguous. In order to properly assess them, metabolite concentrations need to be reported. Such quantification requires a calibration or reference compound (C_ref) of known concentration and generally, one of two methods is used to obtain the absolute values: in the first one, an external solution with a known concentration is placed outside of the object of study but within the coil sensitivity region. This requires the chemical standard to resonate outside of the spectral metabolite regions to avoid overlap and subsequent difficulty in differentiation. Furthermore, that approach is made more complicated by variation in B₁+ field and B₀ homogeneity, with respect to the sampled tissue volume. One example of such a compound is hexachlorocyclo-triphosphagene (HCCTP), which resonates at 22.9 ppm, well outside the range of physiological metabolites [14,15]. In the second approach, a metabolite naturally found in the tissue with established concentration is selected as the reference. From a quantitative review of experiments conducted in skeletal muscle [5], the concentration of ATP ($[ATP]\approx 8.2mmol/L$ cell water) is appropriate as an internal reference for quantification of healthy muscle metabolites. In the case of brain tissue, a reference value of $\text{3 }mmol/L$ for γ-ATP is often used as internal reference [4,16]. Thus, the concentration of the metabolite (C_met) of interest is determined as follows:

$$C_{met}=[C_{ref}]\cdot \frac{S_{met}}{S_{ref}}\cdot C_{MR}$$

Where S_ref and S_met are the signals detected from the reference and metabolite, respectively, and C_MR is a correction factor accounting for differences in relaxation time, position relative to the coil, magnetic susceptibility or in general any other difference between the reference and the metabolite of interest.

A straightforward way to understand the main requirements for ³¹P-MRS is by comparing it with ¹H as all clinical and research MR systems clearly use the latter as their primary focus. Some disadvantages can be deduced from this comparison such as reduced sensitivity, lower concentration in tissue, longer T₁ and shorter T₂ relaxation times of ³¹P compounds. On the other hand, some advantages are also present. First, the significantly larger spectral dispersion of phosphorus metabolites leads to a better separation of signals. This is especially important when measuring PE, PC and GPC as they all are choline-containing metabolites that cannot be separated by ¹H spectroscopy. Also, there are no fat or water dominating signals in phosphorus spectra, meaning no frequency selective suppression techniques are required to null these dominating signals. A further difference lies in the relaxation mechanisms that govern each nucleus, whereas for ¹H the T₁ relaxation is dominated by the magnetic dipole-dipole interaction, the relaxation of ³¹P metabolites is strongly influenced by the chemical shift anisotropy mechanism, which gets stronger as the magnetic field increases [17,18].

The lower gyromagnetic ratio and consequently lower Larmor frequency of phosphorus compounds is the main difference resulting in differing technical aspects, when comparing to proton scans, as all components in clinical scanners are tuned/optimized to the later. Thus, in order to use those same hardware systems, additional components such as a broadband transmitter/receiver and RF-coils are required. Most organ specific ³¹P-MRS studies are performed using single loop surface coils to take advantage of their high sensitivity, however their strong variability in B₁+ flip angle due to the inhomogeneous excitation profile can compromise the results for spectroscopic imaging applications. As a result, complex phased arrays and whole volume coverage coils with high sensitivity and homogeneous B₁+ excitation have been developed [[19], [20], [21], [22], [23]].

A technological development that has played a key role in improving the spectral quality of ³¹P-MRS experiments is the use of ultra-high field MR systems (i.e. B₀ ≥ 7 T) [18,24]. The main features include an SNR boost of roughly double as well as better spectral resolution and the possibility to gain more SNR per unit of time due to the shortening of T₁ relaxation times as a consequence of the above mentioned chemical shift anisotropy mechanism [18]. This gain in SNR can be traded for a higher resolution or faster acquisition times. Unfortunately, the availability of such systems for human scanning is still very limited.

Further improvement of signal in phosphorus MRS experiments is possible through the following mechanisms. First, analogous to the homonuclear coupling, there is another coupling interaction between the ³¹P and ¹H nuclei that causes line broadening of the in vivo spectra, especially at the PME and PDE resonances. Applying RF irradiation at the proton frequency during the ³¹P signal acquisition reduces this heteronuclear coupling. This process effectively decouples the interaction and results in narrower spectral lines, making possible the separation of signals such as PC and PE as well as GPC and GPE at low field strength (i.e. B₀ ≤ 3 T) [25]. Older decoupling approaches used continuous RF transmission at the ¹H frequency which escalated the specific absorption rate (SAR). Later, however, clever combinations of RF cycled pulses, such as WALTZ-4 and WALTZ-16, and multiplets of these called supercycles, reduced SAR whole maintaining efficient proton decoupling [26]. Second, the utilization of the nuclear Overhauser effect (NOE) can also enhance the signal of interest. By applying proton RF irradiation during the ³¹P inter-pulse delay, the sensitivity for the detection of the later has been reported to improve up to 80% in skeletal muscle at low field [27] and 44% in brain at ultra high field [28] In theory the maximal possible NOE is 0.5(^1Hγ/^31Pγ) = 0.5(42.576 MHz/T / 17.23 MHz/T) = 124%. Here, it is important to mention that a drawback of the above-mentioned techniques is the significant increment in the specific absorption rate (SAR) due to the ¹H-RF irradiation. Also, it is difficult to predict the maximal NOE one will get in practice, which adds to complexity in absolute quantitation. Hence, other polarization transfer methods have been explored to enhance the sensitivity of some ³¹P-MRS signals [[29], [30], [31], [32]].