Chemical Exchange Rotation Transfer imaging of Phosphocreatine in Muscle

Abstract

In chemical exchange saturation transfer (CEST) imaging, the signal at 2.6 ppm from the water resonance in muscle has been assigned to phosphocreatine (PCr). However, this signal has limited specificity for PCr since the signal is also sensitive to exchange with protein and macromolecular protons when using some conventional quantification methods, and will vary with changes in the water longitudinal relaxation rate. Correcting for these effects while maintaining reasonable acquisition times is challenging. As an alternative approach to overcome these problems, here we evaluate chemical exchange rotation transfer (CERT) imaging of PCr in muscle at 9.4T. Specifically, the CERT metric AREX_double,cpw at 2.6 ppm was measured in solutions containing the main muscle metabolites, in tissue homogenates with controlled PCr content, and in vivo in rat leg muscles. PCr dominates CERT metrics around 2.6 ppm (though with non-trivial confounding baseline contributions), indicating CERT is well-suited to PCr specific imaging, and has the added benefit of requiring a relatively small number of acquisitions.

INTRODUCTION

Phosphocreatine (PCr) and creatine (Cr) are two major metabolites of the creatine kinase (CK) reaction that play vital roles in muscle energetics (1,2). PCr serves as a mobile energy reserve. During the CK reaction, these two molecules convert:

$$\text{Cr}+\text{ATP}\leftrightarrow \text{ADP}+\text{PCr}+{\mathrm{H}}^{+}$$ (1)

Measurements of both PCr and Cr are essential for evaluating the CK reaction and for assessing metabolism in muscle. While proton magnetic resonance spectroscopy (¹H MRS) has been used to measure total Cr, it cannot differentiate between PCr and Cr because they resonate at almost the same frequency (3). Thus proton spectroscopy has a limited ability to measure energy metabolism. Phosphorus-31 (³¹P) MRS has been used to measure [PCr], but suffers from low spatial resolution and sensitivity (4).

Chemical exchange saturation transfer (CEST) is a contrast mechanism which applies long, frequency selective RF pulses to saturate protons in exchanging pools and detect subsequent changes in water signals. Due to the accumulated effects of the exchange, CEST provides an indirect measurement of solutes with significantly enhanced sensitivity compared to direct detection via conventional magnetic resonance spectroscopy (MRS). In addition, although Cr and PCr have similar resonance frequency offsets at around 3 ppm from tetramethylsilane (TMS) in MRS spectra (which measure non-exchanging carbon-bound protons), they show quite different resonance frequency offsets in CEST spectra (which indirectly measure, and hence have the resonance frequencies of, exchanging amine protons), with Cr at around 2 ppm and PCr at both 2 ppm and 2.6 ppm from the water line. Thus the peak at 2.6 ppm potentially provides a way to distinguish signal contributions from these two molecules. Previously, CEST has been used for high-sensitivity Cr imaging in muscle (5-7). More recently, CEST and CEST variants have been used for PCr imaging by us (8) and others (9).

There are multiple approaches for CEST imaging of Cr and PCr. In some previous CEST creatine imaging studies, a magnetization transfer ratio with asymmetric analysis (MTR_asym) has been used to quantify CEST signals (5,6). Although this method requires only three sampling points and is very simple, the results are influenced by semi-solid magnetization transfer (MT) asymmetry effects and relayed nuclear overhauser enhancement (rNOE) effects originating from the other side of the water peak. In addition, asymmetry approaches typically require significant additional data acquisitions to correct for B₀ inhomogeneities (10). An alternative multiple-pool Lorentzian fit has been used to quantify the CEST signals from Cr and PCr (7,9), but its application is limited by the long acquisition time necessary to obtain the entire Z-spectrum. In this work, we apply our previously reported CEST variant, chemical exchange rotation transfer (CERT) imaging (11-13), to quantify CEST signals from PCr around 2.6 ppm in muscle. In addition, since components of the CEST and CERT signals do not add linearly, the direct subtraction of label and reference signals employed in previous methods can only remove 0^th-order effects, but not higher order effects (14). Hence, we combine CERT with an inverse analysis method based on the previously published apparent exchange-dependent relaxation (AREX) approach (15) to further increase the specificity of PCr imaging. Our goal is to examine the specificity and sensitivity of this CERT measure of PCr when using as few as three data points.

MATERIALS AND METHODS

CEST methods and metrics

The conventional CEST metric MTR_asym is the difference of CEST signals acquired at the solute resonance and the opposite side of the water line:

$${\text{MTR}}_{\text{asym}}=\frac{S_{+}-S_{-}}{S_0}$$ (2)

where (−) represents the offset of the exchanging species and (+) is the offset on the opposite side of the water peak. S₀ is the signal acquired in the non-irradiated control case.

CERT is a CEST variant that detects exchanging solutes via their selective rotation by RF pulses. Most commonly CERT signals are acquired using trains of pulses with irradiation flip angles (θ) of π and 2π, while keeping the RF frequency offset and average irradiation power (B_{avg power}) constant (11-13,16). Eq. (3) shows the definition of a particular CERT metric which uses a constant pulse width (cpw):

$${\text{MTR}}_{\text{double},\text{cpw}}={\phantom{\mid }\frac{S_{-}({\theta }_2=2\pi ,d{c}_2)-S_{-}({\theta }_1=\pi ,d{c}_1)}{S_0}\mid }_{B_{\text{avg power}}}$$ (3)

The duty cycle (dc) is varied to maintain constant pulse width (pw). Here dc is set to be 0.18 for 2π irradiation and 0.72 for π irradiation so that the pulse widths for the two irradiations are both 13 ms with an average irradiation power of 1 μT.

To further increase specificity, we combined the CERT and AREX approaches to define a new metric:

$${\text{AREX}}_{\text{double},\text{cpw}}=R_{1obs}{\phantom{\mid }(\frac{S_0}{S_{-}({\theta }_1=\pi ,d{c}_1)}-\frac{S_0}{S_{-}({\theta }_2,d{c}_2)})\mid }_{B_{\text{avgpower}}}$$ (4)

where R_1obs (=1 / T_1obs) is the observed water longitudinal relaxation rate. By taking the difference of inverted signals, the AREX approach corrects for non-linear contributions to the measured signal (17). Further, by scaling the metric by R_1obs, AREX accounts for the effect of water relaxation rate changes on the measured signal after irradiating at the solute resonance. Finally, direct water and macromolecular saturation effects are cancelled by using the same average irradiation power and offset when applying pulse trains of π pulses or train 2π pulses.

In vitro validation

To study the ability of the CERT approach to quantify [PCr], 5 solutions with metabolites at the high end of their physiological concentrations (20 mM Adenosine triphosphate ATP, 50 mM creatine Cre, 50 mM phosphocreatine PCr, 300 mM Glycogen, and 50 mM Lactate) were prepared in 1× (phosphate buffered saline) PBS and titrated to pH of 7.0. Ex vivo tissue samples were prepared to evaluate the ability of CERT to quantify the PCr. Specifically, tissue homogenates were prepared by removing whole rat brains from freshly sacrificed rats (18,19). The intact brain tissue was weighed and washed quickly in buffer composed of 70 mM HEPES with 70mM KCl. After addition of 4 times CK buffer (w/w) (45 μl D₂O + 22.5 μl HEPES + 22.5 μl KCl + 5 μl MgCl₂ + 5 μl N-Acetyl-L-Cysteine (NAC)), the tissues were homogenized, and centrifuged. The supernatants were divided into two samples. Sample #1 was used as a control; PCr powder was added to sample #2 to reach 10 mM.

Animal experiments

Five healthy rats were immobilized and anesthetized with a 2%/98% isoflurane/oxygen mixture. Respiration was monitored and a constant rectal temperature of 37°C was maintained throughout the experiments using a warm-air feedback system (SA Instruments, Stony Brook, NY). All procedures were approved by the Animal Care and Usage Committee of Vanderbilt University.

MRI

All measurements on animals and samples were performed on a Varian DirectDrive™ horizontal 9.4T magnet with a 38-mm Litz RF coil (Doty Scientific Inc. Columbia, SC) at 37°C. Continuous wave (CW) and pulsed Z-spectra were acquired with RF offsets from −2000 Hz to 2000 Hz (−5 ppm to 5 ppm at 9.4 T) with an interval of 50 Hz (0.125 ppm at 9.4 T). A control scan was performed to acquire S₀ by setting the RF offset to 100,000 Hz. For Z-spectra studies on solutions and tissue samples, the first points of each free induction decay (FID) were used for data analysis. For Z-spectra from in vivo rat imaging, a single-shot SE-EPI sequence was employed using a FOV of 30 × 30 mm², a slice thickness of 2 mm, a matrix of 64 × 64, a receiver bandwidth of 250 kHz, and one acquisition. B_{avg power} (the square root of the mean of the square of the field amplitude) was 1 μT for all irradiations, based on our previous experience isolating the effects of slower-exchanging protons from those of intermediate and fast-exchanging protons in Cr and proteins, along with direct and MT effects (12,20). The total irradiation duration was 5s for animal experiments and 8s for sample experiments performed before acquisition with a 2s delay time before the next acquisition. CW CEST signals were acquired with a long rectangular pulse. CERT signals were acquired using a pulsed-CEST sequence with a series of Gaussian RF irradiation pulses followed by acquisition (11-13). R_1obs for animal experiments was obtained using a selective inversion recovery (SIR) sequence (21). R_1obs for samples was obtained using an inversion recovery fast spin-echo (FSE) sequence.

RESULTS

In order to determine the sensitivity and, more importantly, the specificity of CERT approaches to imaging PCr, we acquired data from chemical phantoms, tissue homogenates with controlled PCr content, and in vivo rat leg muscle. These samples provide a means for examining the signal specificity under increasing complexity, from 1) simple known molecular content, to 2) complex molecular content but with no large background macromolecular contributions and with controlled PCr content, and finally to 3) the full in vivo case.

Fig. 1 shows the experimental MTR_double,cpw and AREX_double,cpw spectra from solutions containing the main metabolites in muscle. MTR_double,cpw and AREX_double,cpw at 2.6 ppm are dominated by contributions from PCr, while MTR_double,cpw and AREX_double,cpw at 2 ppm have sizable contributions from Cr, PCr, and ATP. The results in vivo will depend on the corresponding metabolic concentrations and may not match our phantoms, which were taken on the high end of physiological ranges.

Fig. 1:

Measured MTR_double,cpw and AREX_double,cpw spectra from solutions containing the main metabolites in muscle. T_1obs are around 4 s for all samples.

To further study the ability of CEST and CERT to detect PCr, we performed CEST experiments on tissue homogenate and tissue homogenate + PCr. Results are shown in Fig. 2. We used rat brain as the source for the homogenate, rather than muscle, to start with very low concentrations of PCr as our baseline condition (22). After adding PCr to the homogenate, the CEST effect at 2.6 ppm increased, confirming the PCr basis for the CEST peak at 2.6 ppm. Also note that although MTR_asym shows a negative sloping baseline in Fig. 2c, MTR_double,cpw in Fig. 2d and AREX_double,cpw in Fig. 2e show reduced and flatter baseline, suggesting that CERT is effectively more specific, or at least has a simpler relationship to PCr than the conventional asymmetric analysis.

Fig. 2:

CW Z-spectra (a), CERT Z-spectra (b), MTR_asym spectra using CW CEST data (c), MTR_{double, cpw} spectra (d), and AREX_{double, cpw} spectra (e) for sample 1 homogenate, and sample 2 homogenate + PCr. Arrows indicate the CEST signals from PCr at around 2.6ppm. T_1obs are 3.7 s and 3.5 s for sample 1 and sample 2, respectively. Note that the 2π spectra in figure (b) serve as baselines with no rotation-exchange effects in the CERT approach.

Fig. 3 shows the experimental CW and pulsed Z-spectra, along with the resulting MTR_asym, MTR_double,cpw, and AREX_double,cpw spectra from five healthy rat legs. Note the negative MTR_asym values at farther offsets from water, indicating possible confounding contributions from asymmetric MT or nuclear Overhauser enhancement (NOE) effects originating from the other side of the water peak. These effects cause a sloping negative baseline signal, making interpretation difficult. In contrast, the MTR_double,cpw and AREX_double,cpw metrics specifically quantify the CEST signals from PCr without large unwanted baseline distortions, though there is still the possibility of non-specific signal contributions at both 2 and 2.6 ppm. Fig. 4 shows corresponding maps of anatomy, and AREX_double,cpw maps at 2 ppm and 2.6 ppm.

Fig. 3:

Experimental CW and CERT_cpw Z-spectra (a), MTR_asym spectrum (b), MTR_double,cpw spectrum (c), and AREX_double,cpw spectrum (d) from rat legs (N=5). Arrows indicate the CEST signals from PCr at around 2.6ppm. T_1obs is 2 ± 0.07 s.

Fig. 4:

Maps of tissue anatomy (a), AREX_{double, cpw} at 2 ppm (b), and AREX_{double, cpw} at 2.6 ppm (c) from a representative healthy rat leg. ROI was drawn from region where shimming is good (the root mean square deviation of B₀ field map was < 60 Hz). Anatomy image was from the control scan in the CEST experiment, and was cropped to match the PCr image and to avoid an aliasing artifact at the edge of the image.

DISCUSSION

MTR_asym has been widely used to isolate the CEST signal from direct water saturation (DS) and semi-solid MT effects. While MTR_asym has the benefit of being a simple metric for quantification, it is sensitive to confounding influences from upfield rNOE effects and/or semi-solid macromolecular asymmetry, creating a negative sloping baseline, that be seen in the current work (figure 2c). In this case, the contribution of PCr to the homogenate + PCr MTR_asym values would be difficult to isolate without comparison to the values in homogenate. Such a tissue-comparison approach has been applied in the CEST literature (e.g. (23)) due to the difficulty of separating the targeted signal from the negative sloping baseline. Note, however, that the need for such comparisons and the relative contributions of confounding signals depends strongly on the pulse sequence parameters. For example, one approach to Cr imaging uses MTR_asym with higher-power and shorter duration irradiations than the current work in order to use exchange-rate filtering to increase specificity (6).

The CERT results (Figs. 2d, 2e, 3c and 3d) show positive MTR_double,cpw or AREX_double,cpw signals at 2.6 ppm and appear to have a baseline signal flatter and closer to zero, though this assessment is somewhat arbitrary and inexact. A benefit of this diminished and simpler baseline is the viability of acquiring data only at the offset of interest, in this case 2.6 ppm, because fitting the baseline dependence on frequency is not necessary, or at least less necessary. (A separate R_1obs measure is still necessary to calculate the AREX metric.) In contrast, MTR_asym or raw Z-spectra analysis typically uses data acquired at many frequency offsets to correct for baseline effects. For example, a three-point method, which uses the average of two nearby signals as reference, has been previously used to quantify amides at 3.5 ppm in brain (24,25). However, this method is unlikely to accurately quantify the CEST signal at 2.6 ppm because it is close to the nearby amides at 3.5 ppm and amines at 2 ppm, and we know of no attempts to quantify PCr using such an approach. There has been a recent study that used a multiple pool Lorentizian fitting of the Z-spectrum to image PCr (9). Note that this study fit the data using 78 frequency offsets and did so at 11.7 T, the very high field facilitating fitting offset dependent models. In contrast, the AREX_{double, cpw} metric uses a single offset and no signal modeling.

Proteins may also produce significant CEST signals from amides, amines, and hydroxyls (18,19,26,27). Our previous studies on bovine serum albumin (BSA) and egg white albumin (EWA) indicate that the guanidino amine of proteins may have significant contributions to the CEST signal at 2 ppm, but no significant contributions at around 2.6 ppm (13,18). In addition, although fast exchanging amines have broad peaks which may influence the CEST signals at 2.6 ppm, their CEST effects can be mostly removed in CERT due to its exchange filter effect (11,13). This suggests that proteins in muscle make negligible contributions to CERT PCr imaging. This paper thus shows that CERT may offer a viable approach to PCr imaging and a means to study the CK reaction. However, some aromatic and amide protons between 1.5 ppm and 4 ppm, which have slow exchange rates, may affect the CERT signal at 2.6 ppm. This issue of spectral selectivity is worse at clinical field strengths, where distinct solute peaks are difficult, though not impossible (28), to discern. On the plus side, CERT methods avoid contributions from the rNOE at −3.5 ppm, which is on the other side of the water peak and can thus confound CEST metrics based on asymmetry, such as MTR_asym.

While we have emphasized the benefits of a small and flat baseline using the CERT approach, this baseline is not zero. Results in phantoms can give a very rough estimate of the size of this confounding factor in vivo. In Fig. 1b, the AREX_double,cpw signal at 2.6 ppm from PCr is 4.9%s⁻¹, suggesting 0.1% s⁻¹ per 1mM PCr. In Fig. 2e, the subtraction of AREX_double,cpw signal at 2.6 ppm between the homogenate and homogenate+Pcr is 0.7%s⁻¹, suggesting 0.07%s⁻¹ per 1mM PCr. Note that the two calibrations from a simple sample and tissue homogenate are comparable. Using 0.07%s⁻¹ per 1mM PCr to calibrate the rat leg results at 2.6 ppm in Fig. 3d gives 52.4 mM PCr in rat legs. Typical in vivo PCr mM concentrations are in the mid-thirties, suggesting that a confounding baseline of roughly 1/3 the signal value at 2.6 ppm, which seems visually reasonable looking at Fig. 3d. A more precise estimation of the baseline, or a more specific measure, requires non-trivial quantitative tissue modelling and parameter fitting. This would likely require longer data acquisitions and is beyond the scope of this work, which is focused on the utility of a rapidly-acquired model-free metric.

There are two notes of caution in overinterpreting the homogenate results. First, we used tissue homogenates from rat brain to evaluate the CEST signal in muscle due to the difficulty making homogenates from muscle fibers. Although the tissue homogenates in muscle and brain are different, both of them have rNOE and asymmetric MT effects and can be used to evaluate the ability of CERT to remove the influence from the resulting slope in the baseline that confounds conventional asymmetric analysis. Second, any assessment of the sensitivity of CEST or CERT metrics to changes in solute concentrations is possibly confounded by similar dependencies on changes in exchange rates, which are related to variations in tissue pH and/or temperature. Quantitative fittings can avoid this co-dependency, but require assumptions about tissue models.

The AREX_{double, cpw} metric can be acquired in as little as 3 points, though this does not include the required measurement of R_1obs or correct for any possible B₀ and B₁ inhomogeneities. However, in situations where R_1obs variations are negligible, R_1obs normalization may not be essential. We previously showed little B₀ and B₁ dependence when using a similar CERT metric applied to the amide resonance at 3.5ppm (11), but it is still an open question for the current PCr study. Supporting Figure S1 show MTR_asym, MTR_{double, cpw}, and AREX_{double, cpw} spectra from the shimmed area and the non-shimmed area. Much bigger standard deviations from the non-shimmed area than that from the shimmed area were found in the spectrum of MTR_asym, but not MTR_{double, cpw}, and AREX_{double, cpw}. In summary, while the proposed CERT approach shows promise in imaging PCr, especially in producing a relatively flat and small confounding baseline signal, there are still several unresolved issues, including robustness (under field inhomogeneities and tissue variations), separation of concentration and exchange effects, and viability under clinical field strengths and acquisition times. The current study demonstrates PCr sensitivity in phantoms and healthy muscle, but we have not shown spatial correlation vs gold standard or independence from tissue state under pathologies. Given its low signal-to-noise, such validations using ³¹P will be challenging.

CONCLUSION

AREX_double,cpw at 2.6 ppm can produce reasonably specific PCr imaging using three acquisitons, though with non-negligible baseline signal contributions and with separate measurement of R_1obs.

Supplementary Material

Supplementary material

Supporting Figure S1. MTR_asym spectrum (a), MTR_double,cpw spectrum (b), and AREX_double,cpw spectrum (c) from shimmed area (blue) and non-shimmed area (red) on rat legs (N=5). Arrows indicate the CEST signals from PCr at around 2.6ppm. Note that the standard deviations, especially those close to water peak, are relatively bigger in the non-shimmed area than that in the shimmed area.

Abbreviations used

CEST

chemical exchange saturation transfer

CERT

chemical exchange rotation transfer

MT

magnetization transfer

AREX

apparent exchange-dependent relaxation

cpw

constant pulse width

pw

pulse width

dc

duty cycle

CW

Continuous wave

B_{avg power}

average irradiation power

MTR_{double, cpw}

CERT metric using a constant pulse width

AREX_{double, cpw}

CERT metric using AREX and a constant pulse width

MTR_asym

magnetization transfer ratio with asymmetric analysis

¹H MRS

proton magnetic resonance spectroscopy

³¹P MRS

phosphorus magnetic resonance spectroscopy

FID

Free induction decay

SIR

Selective inversion recovery

FSE

Fast spin-echo

PCr

phosphocreatine

Cr

creatine

CK

creatine kinas

TMS

Tetramethylsilane

ATP

Adenosine triphosphate