Localized rest and stress human cardiac creatine kinase reaction kinetics at 3 T

Changes in the kinetics of the creatine kinase (CK) shuttle are sensitive markers of cardiac energetics but are typically measured at rest and in the prone position. This study aims to measure CK kinetics during pharmacological stress at 3 T, with measurement in the supine position. A shorter “stressed saturation transfer” (StreST) extension to the triple repetition time saturation transfer (TRiST) method is proposed. We assess scanning in a supine position and validate the MR measurement against biopsy assay of CK activity. We report normal ranges of stress CK forward rate (kf CK) for healthy volunteers and obese patients. TRiST measures kf CK in 40 min at 3 T. StreST extends the previously developed TRiST to also make a further kf CK measurement during <20 min of dobutamine stress. We test our TRiST implementation in skeletal muscle and myocardium in both prone and supine positions. We evaluate StreST in the myocardium of six healthy volunteers and 34 obese subjects. We validated MR‐measured kf CK against biopsy assays of CK activity. TRiST kf CK values matched literature values in skeletal muscle (kf CK = 0.25 ± 0.03 s−1 vs 0.27 ± 0.03 s−1) and myocardium when measured in the prone position (0.32 ± 0.15 s−1), but a significant difference was found for TRiST kf CK measured supine (0.24 ± 0.12 s−1). This difference was because of different respiratory‐ and cardiac‐motion‐induced B0 changes in the two positions. Using supine TRiST, cardiac kf CK values for normal‐weight subjects were 0.15 ± 0.09 s−1 at rest and 0.17 ± 0.15 s−1 during stress. For obese subjects, kf CK was 0.16 ± 0.07 s−1 at rest and 0.17 ± 0.10 s−1 during stress. Rest myocardial kf CK and CK activity from LV biopsies of the same subjects correlated (R = 0.43, p = 0.03). We present an independent implementation of TRiST on the Siemens platform using a commercially available coil. Our extended StreST protocol enables cardiac kf CK to be measured during dobutamine‐induced stress in the supine position.


| INTRODUCTION
The rate and flux of the creatine kinase (CK) exchange mechanism have been shown to be sensitive measures of heart failure, 1 which is a prevalent and burdensome disease. 2,3 The rate can be characterised by the pseudo first-order forward rate constant, k f CK . Phosphorus magnetic resonance spectroscopy ( 31 P-MRS) enables noninvasive measurement of myocardial k f CK . 4 In addition to measuring k f CK in myocardium at rest, measuring k f CK during pharmacologically induced stress would enable us to understand the effect of a perturbed CK mechanism in the stressed human heart. 1,5 Schär et al introduced the triple repetition time saturation transfer (TRiST) sequence to measure CK kinetics by 31 P-MRS at 3 T. 6 A TRiST acquisition lasts 40 min (out of complete protocol totalling 84 min); it measures k f CK in a one-dimensional coronal stack of slices covering the heart and chest wall. 7 1D-localised k f CK measurement during inotropic stress was achieved by Weiss et al at 1.5 T using the four-angle saturation transfer (FAST) and the derived "FASTest" method. 1,4 However, FAST and FASTest rely on small-angle adiabatic pulses, which are not achievable at 3 T because of radiofrequency power requirements, power deposition and T 2 relaxation. 8 The 3 T TRiST protocol offers an established measurement technique which can be included in a larger cardiac 1 H MR protocol on the same scanner. While 7 T 31 P-MRS has been used for 3D-localised k f CK measurements, the existing published methods have long duration and the higher field strength restricts the recruitment of subjects who have undergone surgical procedures. 9 An increase in myocardial work can be reliably maintained, without physical exercise (and therefore motion), by administering dobutamine intravenously. However, the duration of intravenous infusion should be kept to a minimum length. Current guidelines prescribe 15 min stress protocols 10 and they recommend patients should be scanned in the supine position, for safety in case of arrhythmia. Measuring k f CK within this timeframe is currently only possible using the Two repetition time saturation transfer (TwiST) method, published during the course of this study. 7 However, although TwiST assumes a fixed phosphocreatine (PCr) intrinsic longitudinal relaxation time (T 1 *), ie in the hypothetical case of PCr not undergoing chemical exchange, it is not yet known whether T 1 * remains constant in clinically relevant groups; eg obese, normal weight, and heart failure.
We therefore propose to perform stress k f CK measurements at 3 T in two steps: first, derive a per subject baseline PCr T 1 * with TRiST; and second, perform two further scans during dobutamine-induced stress to record the stress k f CK in <20 min of dobutamine infusion. This stress saturation-transfer (StreST) protocol yields a stack of 1D-localised k f CK measurements at rest and stress. We describe below the implementation of StreST and validate the underlying TRiST method in skeletal muscle, in human myocardium in the prone position and subsequently in the supine position, where we correlate surgical left ventricular (LV) biopsy-obtained CK activity against preoperative myocardial k f CK in a set of withinpatient paired measurements. We then demonstrate StreST by using it to record normative ranges of rest and stress k f CK in the myocardium of normal volunteers and in older obese and age-matched control cohorts.

| THEORY
CK regenerates adenosine triphosphate (ATP) from PCr according to the equilibrium expression PCr 2− + MgADP − + H + ⇌ Cr + MgATP 2− . CK provides the temporal energy reserve in muscle. TRiST uses steady-state saturation of the terminal (γ) phosphate-group of ATP to measure k f CK , the pseudo first-order rate constant of the CK reaction in the forward (ATP-generating) direction. Since the 31 P nuclei in PCr and γ-ATP are undergoing two-site exchange, continuous saturation of γ-ATP allows the forward exchange rate constant to be determined using using the dual-TR method, 8 and a third to measure M Ctrl PCr (see Figure 1). The intrinsic longitudinal relaxation time T * 1 , can be computed from 11 Therefore, by substituting Equation 2, Equation 1 may be recast in terms of T * 1 rather than T ′ 1 4 : T * 1 is a hypothetical longitudinal relaxation constant for a molecule without chemical exchange. Assuming that T * 1 does not change from one scan to the next (eg in myocardium at rest vs under stress), then an additional measurement of k f CK may be made in two steps: measuring M Ctrl PCr and M ′ PCr . This is a similar assumption to that used in the FASTer/FASTest adaptation of the FAST method. 4 FIGURE 1 (A) Six-step StreST protocol (including TRiST as steps 1-4); (B) 1H localiser with 1D CSI grid overlaid; the coil position is marked in green; orange = skeletal muscle, blue = myocardium, green = most anterior myocardial slice (coil slice distance ≤70 mm); (C) spectra acquired from the four steps of TRiST on a healthy, normal-weight subject; (d) spectra acquired during dobutamine stress as part of the extension StreST measurement. In (C) and (D) signal is normalised to the PCr peak value in step 1

| StreST protocol
Our new StreST protocol comprises a rest measurement of k f CK and PCr T 1 * using TRiST (four steps) and a measurement of k f CK during intravenous pharmacological stress (two additional steps). The six (4 + 2) steps are shown in Table 2. All steps are completed in a single scanning session.
In future studies using the proposed StreST protocol, subjects will be scanned in the supine position, instead of the prone position used in the original TRiST studies. Prone scanning is considered to be unsafe from a cardiac monitoring perspective, especially when scanning patients with heart failure who may be at greater risk of dangerous arrhythmia. One in 400 patients receiving dobutamine experience a life-threatening arrhythmia. 12 3 | METHODS All subjects were recruited in a manner approved by the local research ethics committee. All participants gave written informed consent.

| Hardware, sequence and spectral analysis
All experiments used a 3 T TIM Trio MR scanner (Siemens, Erlangen, Germany). The scanner's body coil was used for 1 H localisation. A 10 cm loop transmit-receive surface coil (Pulse Teq, Chobham, UK) tuned to the phosphorus frequency was used for spectroscopy. The 31 P coil was matched for each subject using an RF sweeper (Morris Instruments Inc., Ottawa, Canada). Coil loading was measured by inversion-recovery on a phenylphosphonic acid/ethanol/chromium acetylacetonate fiducial fixed at the centre of the coil, as previously reported, 13 and expressed as a reference voltage (the RMS RF voltage giving a 1 ms 180°pulse). The 31 P coil was positioned above the apical myocardium of the subject. The positioning was checked using 1 H localiser images of the heart and coil fiducial, and repositioning was carried out based on those images.
TRiST was implemented on the Siemens platform, following the reported description, 6 as follows. The vendor's 1D chemical shift imaging (CSI) sequence was modified to continuously selectively saturate a chosen frequency while waiting to detect an R-wave from an electrocardiogram (ECG) monitor attached to the subject. Once an R-wave was detected saturation was continued for a further trigger delay until diastole, at which point an adiabatic half-passage (AHP) excitation pulse, 1D phase encoding gradients and free-induction-decay readout were applied. Diastole was chosen to minimise myocardial motion.
The selective saturation was provided by a train of amplitude-modulated delay alternating with nutation for tailored excitation (DANTE) pulses. 6 B 1 -insensitive 90 o excitation was provided by frequency-cycled AHP pulses. 8 Spectra were analysed by measuring the amplitude of the phased and apodized PCr peak relative to the baseline. 6 Raw data from TRiST scans on the 3 T Achieva Philips scanner at Johns Hopkins were kindly supplied by Dr Schar and used to validate our fitting approach. k CK f , T ′ 1 and T * 1 were calculated as described in the theory section (Equation 1) and as previously reported. 6 The amount of direct (or spill-over) saturation of PCr by the DANTE pulse (Q) was calculated as the ratio of M Ctrl PCr /M 0, PCr . (Q = 1 in the ideal case when there is no direct saturation, but only saturation via chemical exchange from saturated γ-ATP.) Spectra from cardiac slices were selected using the transverse 1 H localiser and analysed on a per-slice basis.

| Literature values
We surveyed the literature for values of k f CK in human myocardium and skeletal muscle (Table 3). We used the arithmetic means of the literature k f CK values and standard deviations (SD) for both tissues as a reference to validate our results.
We validated our TRiST implementation in the calf muscle of nine healthy volunteers (eight males, 30.6 ± 3.8 years old, 74.1 ± 11.3 kg). The subjects were positioned feet-first-supine in the scanner with the 31 P loop coil under one leg. After 1 H localisation, the 1D-CSI grid was positioned running in the anterior-posterior direction. The protocol followed steps 1-4 in Table 2. Other parameters were: 160 mm field of view (FOV), 16 slices, 3 kHz bandwidth, 512 spectral points, and 200 V AHP transmit voltage (corresponding to 800 W peak power, and~35 μT B 1 + in vivo).
Selective saturation of γ-ATP and control saturation were achieved using 35 V DANTE pulses (corresponding to 24.5 W peak power, and~6 μT B 1 + in vivo).

| Myocardium in prone position
Ten healthy volunteer subjects (six males, 29.6 ± 4.9 years old, 70.7 ± 18.2 kg) were scanned using the newly implemented TRiST protocol (steps 1-4, rest k f CK only) to measure myocardial k f CK . The scans were completed in the prone position as per previously published methods.
CSI acquisition parameters were as follows: 160 mm FOV, 16-step matrix, 3 kHz bandwidth, 512 samples. The CSI grid was positioned perpendicular to a transverse localiser covering the heart, with the CSI delineated dimension aligned coronally (parallel to the band of skeletal muscle lying between the coil and the heart). The AHP transmit voltage was 210 V (ie 882 W peak power), and the amplitude-modulated DANTE voltage was maximised within the constraints of the specific absorption rate (SAR) for the short TR scan (typically to~30 V, ie 18 W peak power). Spectra from cardiac slices were selected using the transverse 1 H localisers for analysis as described above. The data from the most apical slice containing only myocardium and blood (but not skeletal muscle) were also analysed separately. The coil to slice distance was <60 mm for these slices.

| Myocardium in supine position
As detailed above, the full StreST protocol will include administering intravenous dobutamine, for which it is preferred to position the subject supine. To test whether the change of position (prone to supine) affects the initial TRiST measurement in the StreST protocol, we scanned the same 10 subjects as used in the previous section (six males, 29.6 ± 4.9 years old, 70.7 ± 18.2 kg) again. This time, scans were in the supine position, using the newly implemented TRiST protocol (steps 1-4, rest k f CK only). Other acquisition and analysis parameters were identical to the previous section.

| Effect of intra-scan B 0 fluctuation
The potential effects of respiratory and cardiac motion induced B 0 changes on TRiST k CK f values were analysed using Bloch simulations of the full TRiST protocol. A dual-echo CINE gradient echo sequence was used to measure the range of B 0 values present in the un-shimmed apical myocardium of a single subject in different cardiac phases and respiratory states in both supine and prone positions. A sinusoidal frequency sweep with amplitude of 0, 20, 40, 60 and 80 Hz was applied to the Bloch simulation to simulate respiratory motion. The AHP pulse was simulated with three different B þ 1 magnitudes: 12, 23 and 35 μT, and the DANTE saturation pulse was scaled appropriately to simulate the experiment. Simulations were run with 10 000 repetitions, each having a random initial cardiac and respiratory phase. Each independent step of TRiST was simulated and combined to give a measured k f CK . Simulation parameters were taken from Table 1 (heart muscle) in reference 23 with k f CK varied from 0.1-0.5 s −1 . SNR was calculated for PCr and γ-ATP, and the simulation was scaled so the PCr SNR in step 4 was equal to 15.

| Validation of MRS measured k f CK by surgical biopsy
In a cohort of 25 subjects listed for clinically indicated surgery for either severe aortic stenosis with preserved (n = 18) or impaired (n = 4) left ventricular ejection fraction (LVEF ≥ or < 55%, respectively), severe primary mitral regurgitation (n = 2), or atrial myxoma (n = 1), k CK f was measured by supine TRiST and compared with CK activity measured ex vivo from surgical LV biopsies.
All subjects preoperatively underwent the TRiST MRS protocol in the supine position as described in the previous section. k f CK was measured for the most apical voxel identified as purely myocardium on 1 H localisers. Intra-operative biopsies from LV septal endocardium were obtained by the operator 10-15 min after cardiopulmonary bypass was established, then immediately placed into liquid nitrogen and stored at −80°C.
For the measurement of CK activity, a heaped spatula full of frozen, crushed LV tissue was combined with CK-NAC reagent (Thermo Fisher Scientific catalogue code TR14010) and the prescribed series of reactions were monitored using a spectrophotometer to measure the absorbance of NADH at 340 nm and 37°C over 3 min. 24-26 CK activity (IU/mL) was calculated from the rate of change in absorbance of NADH, corrected for reaction volume and an assay-specific correction factor, averaged over three runs and normalised to Lowry protein (mg/mL). Results are presented as CK activity (IU/mg protein). The MRS-measured CK rate constant was then correlated with biopsy-measured CK activity.

| Validation of the stress k f CK measurement (StreST) in healthy volunteers
The validity of the final reduced-time k CK f measurement (from steps 5 and 6 of the full StreST protocol) was tested in six healthy volunteers (all males, 31 ± 9 years old, 75 ± 8 kg). After the initial TRiST measurement (steps 1-4), the follow-on measurement (steps 5 and 6) was made without repositioning and with the subject still at rest (ie a "null stress" control condition). The PCr-matched filtered signal-to-noise ratio (SNR) of the control acquisition (step 4 in Table 2) was determined. The k CK f , T ′ 1 , T * 1 and Q were reported. Reproducibility of the PCr amplitude of individual scans was assessed from the fourth and fifth scans, which were acquired with identical protocols in this validation step (ie corresponding to rest and dobutamine-stress scans in patients). Two methods of measuring M ′ PCr were compared: (i) by saturation-correction in TRiST; and (ii) directly from the sixth StreST step (see Table 2). The correlation and Bland-Altman statistics for these two k CK f measurements were computed.

| StreST in obese subjects and age-matched controls
As many cardiac patients are obese, to allow the measurement to be validated in a real-world population, the full StreST protocol (steps 1-6), including dobutamine-induced stress during the second measurement, was performed in age-matched obese and normal-weight volunteers.
For stress scans, dobutamine was administered intravenously, starting at 5 μg kg −1 min −1 , and increasing the infusion rate every 3 min until a target heart rate of 65% maximum heart rate [ie 220 (age in years)] was achieved; this target heart rate was then maintained at a steady state for 18 min while the additional StreST measurements (steps 5-6) were made. Spectra from cardiac slices were selected using the transverse 1 H localisers for analysis as described above. The data from the most apical slice containing only myocardium and blood (but not skeletal muscle) were also analysed separately. The coil to slice distance was ≤70 mm for these slices.

| Literature values
The results of the survey of literature k f CK are provided in Table 3. The inter-study mean ± SD k f CK values were 0.27 ± 0.04 s −1 (skeletal muscle) and 0.32 ± 0.07 s −1 (myocardium).

| Skeletal muscle (calf)
In all subjects, seven or more slices were identified in the transverse 1 H localiser images as containing mainly skeletal muscle. The mean (± SD) PCr SNR in the control saturation acquisition (step 4) was 45 ± 32. Example spectra are shown in Figure 2a.
Consistent T ′ 1 and k CK f values were found across the five slices corresponding to 20-70 mm from the coil in all subjects. The average T ′ 1 in these slices was 2.2 ± 0.4 s, and k CK f was 0.25 ± 0.03 s −1 . In the two slices furthest from the coil (~70-80 mm), which also contained the tibia and the highest amount of subcutaneous fat, T ′ 1 was higher and k CK f lower (Figure 2b,e). T * 1 was less consistent across slices and between the subjects (Figure 2c).
Complete saturation (>95% saturation) of γ-ATP was observed in all subjects, and in all slices except the two furthest from the coil; in these slices the residual γ-ATP level was 12 ± 3% of the control saturation scan. The ratio of the control-saturation PCr peak to the no-saturation PCr peak (Q, a measure of direct saturation of PCr by DANTE) was >0.5 for depths from 30-80 mm (Figure 2d). In the closest slices to the coil (10 and 20 mm) Q was <0.5, ie substantial direct saturation occurred.
Results from this subsection and others are summarised in Table S1.

| Myocardium in prone position
From the 10 healthy volunteers scanned in the prone position, 29 slices were identified as corresponding to myocardium in the transverse 1 H localisers and had sufficient SNR for analysis (PCr SNR >10 in the control scan).
The all-slice mean ± SD k f CK was 0.29 ± 0.09 s −1 . Analysing only the most anterior purely myocardial slice in each subject (10 slices) gave a mean k f CK of 0.32 ± 0.15 s −1 .

| Myocardium in supine position
The same 10 healthy volunteers were also scanned in the supine position. In this dataset, 30 slices were identified as corresponding to myocardium in the transverse 1 H localisers and had sufficient SNR for analysis.
The all-slice mean ± SD k f CK was 0.15 ± 0.10 s −1 . Analysing only the most anterior purely myocardial slice in each subject gave a mean k f CK of 0.24 ± 0.12 s −1 .

| Effect of intra-scan B 0 fluctuation
The single subject measurement of B 0 established that the mean range of γB 0 experienced in the apical myocardium due to cardiac motion in a 31 P experiment is 34.3 Hz (supine) and 34.6 Hz (prone), and due to respiratory motion is 66.7 Hz (supine) and 36.1 Hz (prone). (Figures S1 and S2). As the range of B 0 variation was increased in the simulations, the amount of time during the DANTE saturation pulse when M z,γ-ATP = 0 decreased (ie γ-ATP saturation was not achieved at all times), even though the SNR of the residual γ-ATP peak in TRiST steps 2 and 3 remained very low: SNR <2.5 (Figure 3a,b). With increasing B 0 fluctuation amplitude and decreasing γ-ATP saturation, the measured k f CK also decreased (Figure 3c). At the level of the estimated B 0 variation due to respiration in our study, the measured k f CK was simulated to be 0.61 times the true k f CK in a supine position and 0.85 times the true k f CK in the prone position. The linearity of the ratio of measured k f CK /true k f CK decreases with increasing B 0 variation ( Figure 3d).

| Validation of MRS measured k f CK by surgical biopsy
For the 25 subjects listed for clinically indicated surgery, mean (± SD) k f CK was 0.21 ± 0.10 s −1 and mean (± SD) biopsy-measured CK activity was

| Validation of the stress k f CK measurement (StreST) in healthy volunteers
All the per-subject and mean k f CK values from the myocardial and skeletal muscle voxels of the six healthy volunteer rest-rest ("null stress" control) StreST scans are plotted in Figure 4. In these scans, 36 slices were identified as corresponding to myocardium in the transverse 1 H localisers and had sufficient SNR for analysis (PCr SNR >10 in the control scan). The all-slice mean (± SD) PCr SNR was 16 ± 9, T 1 ' was 2.9 ± 1.0 s, and T 1 * was 4.8 ± 1.8 s. The all-slice mean k f CK of the first measurement (TRiST) was 0.14 ± 0.08 s −1 and the all-slice mean of the second measurement (dobutamine was not administered for this validation experiment) was 0.22 ± 0.14 s −1 . A per-slice comparison of these k f CK measurements yielded a correlation of 0.51 (Figure 5a). Bland-Altman (Figure 5b)   StreST was performed without dobutamine stress for this validation study. Different markers denote different subjects. In (c) the inter-subject mean and SD k f CK is shown for TRiST (blue: Also the first t measurement of StreST) and the second measurement of StreST (red) The coil reference voltage, measuring the degree of coil loading, varied by <10% for all six subjects, and was within 25% of the values measured in the skeletal muscle validation.

| StreST in obese subjects and age-matched controls
In both obese and normal-weight volunteers (34 in total) the average k f CK in all myocardial slices (with PCr SNR >10) was 0.12 ± 0.08 s −1 . The average PCr SNR was 14 ± 9 across the 209 slices analysed.
In the subjects that underwent both rest and stress measurements, the mean k f CK at rest was 0.16 ± 0.07 s −1 (obese) and 0.15 ± 0.09 s −1 (normal weight). Under stress the values were 0.17 ± 0.11 s −1 (obese) and 0.17 ± 0.15 s −1 (normal weight). This data is shown in Figure 6b.
The T 1 * of the two cohorts was 5.69 ± 1.43 s for obese and 4.67 ± 1.92 s for normal-weight subjects ( Figure 6c); this difference is statistically significantly (Student's t-test, p = 0.02).

| DISCUSSION
We have implemented a new StreST protocol for measuring human CK rate constants in the human heart during dobutamine-induced stress. In so doing, we have also implemented the published TRiST protocol measuring k f CK at rest for the first time on a Siemens scanner, and using a commercially available coil. We have tested StreST (and hence also TRiST) in calf and cardiac muscle and applied it in the hearts of normal volunteers and obese subjects. We have demonstrated a correlation between our MRS measured value of k f CK and CK activity in human LV biopsies.
Measurements in calf muscle show that our implementation of TRiST measures k f CK are in line with literature values up to 70 mm from coil.
The coil loading changed by up to 25% between skeletal muscle and the thorax. Therefore, we expected accurate myocardial measurement in cardiac slices ≤70 mm from the coil, ie we expected that k f CK in apical cardiac slices could be measured robustly. This is corroborated by a Monte   (Table 3). In the supine position, the measured k f CK throughout this study is much lower than the paired prone estimate, the literature estimate of 0.32 s −1 , and our own 7 T k f CK estimate (0.35 ± 0.05 s −1 ). 9 It is therefore likely that the absolute value of k f CK measured in a supine position is an underestimate. However, simulations of the effect of B 0 variation during respiratory and cardiac cycles and correlation with biopsy-measured CK activity in 25 patients indicate that despite the low absolute value of supine MRS-measured k f CK , trends in our measured k f CK values are still meaningful, that is, increases or decreases in measured k f CK correspond to real increases or decreases.
We invested considerable effort in studying the possible causes of the lower supine TRiST k f CK measurements. A thorough validation of the sequence timings was performed in the vendor simulation environment and by capturing the live waveforms of the triggered sequence using a digital oscilloscope on the scanner. Data shared from Johns Hopkins were used to validate our analysis process, which performed comparably with the Johns Hopkins analysis. Bloch simulations of the TRiST method indicated that if constant steady-state saturation of γ-ATP is not maintained completely throughout the mixing time, the measured k f CK will underestimate the true k f CK by a predictable scaling that is approximately linear for modest B 0 fluctuation amplitudes ( Figure S5). Note that this effect can occur even when the γ-ATP peak is well suppressed in the observed saturated spectra. It is proposed that this is produced by B 0 shifts, due to respiration or cardiac motion, intermittently shifting the γ-ATP resonance away from the target selective saturation frequency. We have shown that the range of B 0 experienced in the myocardium is raised in this experiment when the subject is supine rather than prone (~60 Hz range vs 30 Hz). The choice of supine scanning was necessitated in this study for subject safety during pharmacological stress and will be required in our institute for further studies using StreST in patients with established cardiac diseases. Scanning supine also helps coil placement and matching.
The effect of B 0 shifts due to respiration was found, by simulation, to decrease the measured k f CK by~1.6 times for supine scans. This factor was found to be constant for all values of k f CK as long as the B 0 shifts did not exceed a range of 80 Hz. Above this level the effect is nonlinear, decreasing the sensitivity of TRiST to changes in k f CK . Our simulations also suggest that even in the prone position the true value of k f CK is likely to be underestimated by the TRiST method. At the measured amplitude of B 0 fluctuation the correction remains mostly linear and so relative changes in k f CK are preserved for both prone and supine scanning.  7 compared with that measured in this study, for all cardiac slices, the most apical cardiac slices, and the apical cardiac slices from normal-weight and obese subjects StreST reduces the time of the consecutive measurement from 40 min to 20 min by assuming that the subject's T 1 * is constant, which makes it feasible to measure k f CK during dobutamine infusion at 3 T. Previously, Weiss et al used an adaptation of the FAST protocol to measure k f CK during stress in 13 min at 1.5 T. 1 The validation of StreST applied without dobutamine showed that the method is able to reliably measure the same k f CK in a reduced duration in the most apical, high SNR voxels. It is therefore likely that the assumption of static between-scan T 1 * is reasonable. In voxels with low SNR or experiencing high direct saturation (low Q, eg skeletal muscle) the reduced duration measurement does not match the full TRiST measurement and introduces high variance.
The average PCr T 1 * calculated, as per Equation 2, was different for the two cohorts: normal-weight and obese (p = 0.02). This suggests that to accurately measure stress k f CK , T 1 * must either be determined per subject, as in StreST, or per cohort in a pilot study designed to measure T 1 *. We do not recommend assuming a single PCr T 1 * for all human subjects.
Like TRiST, StreST has diagnostic potential for noninvasively assessing the CK system and by extension a subject's contractile reserve. 27 Sensitivity to contractile reserve would be valuable in patients who are not able to undergo conventional stress testing, eg severe valvular heart disease. The CK system is also a major mechanism for controlling cytosolic [ADP]. A raised cytosolic [ADP] at stress contributes to increased LV enddiastolic pressure and diastolic dysfunction. 28 A raised LV end-diastolic pressure is characteristic of heart failure with preserved ejection fraction, which comprises approximately half of all clinically presenting heart failure cases.
Cardiac positron emission tomography (PET) can also measure myocardial metabolic reaction kinetics through the uptake of tracers. 29 It is able to confirm viability in suspected hibernating myocardium using glucose tracers. 30 PET is able to detect uptake in ingressing inflammatory cells and has emerging roles in the detection of prosthetic valve endocarditis 31 and inflammatory atherosclerotic coronary and carotid plaques. 32,33 However, the onward metabolism of the tracer after uptake cannot be assessed and it is not possible to distinguish which cell type is responsible using PET alone. The MRS technique presented here is specific to CK-expressing cells, ie cardiomyocytes. Cardiac MR(S) and PET measure similar information with differing trade-offs in temporal and spatial resolution. The use of both in tandem could offer complementary information. 30 StreST was demonstrated in a control cohort, as well as an obese cohort. The TRiST component of the protocol was run successfully on 34 out of 35 initial subjects. The full StreST protocol was completed by 17 out of 24 subjects, with five subjects electing not to complete due to discomfort, and two scans were stopped after exceeding the local limit on maximum scan duration. The mean ± SD time of a complete StreST protocol was 103 ± 7 min; steps 1-6 of StreST take 64 min in total. The TRiST and StreST techniques are being applied in ongoing studies, building on the initial cohort scans in this work.

| CONCLUSION
In this work, we introduced an extended StreST protocol that enables measurement of k f CK during a 20-min dobutamine infusion at 3 T. We also independently implemented the TRiST protocol on a Siemens 3 T scanner using commercially available hardware. We compared TRiST measured in the prone and supine position and provided a non-MR validation of MR-measured k f CK . We showed by simulations that respiratory-induced motion can lead to incomplete γ-ATP saturation during the saturation-transfer phase of the TRiST sequence, even in the case where the γ-ATP peak is absent from the saturated spectra. Linear correction can compensate for these effects for light to moderate B 0 -field fluctuation amplitudes.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.