Dynamic interleaved ¹H/³¹P STEAM MRS at 3 Tesla using a pneumatic force-controlled plantar flexion exercise rig

Abstract

1. Objective

To develop a measurement method for interleaved acquisition of ¹H and ³¹P STEAM localised spectra of exercising human calf muscle.

2. Materials and Methods

A nonmagnetic exercise rig with a pneumatic piston and sensors for force and pedal angle was constructed to enable plantar flexion measured in the 3 Tesla MR scanner, which holds the dual tuned (¹H,³¹P) surface coil used for signal transmission and reception.

3. Results

³¹P spectra acquired in interleaved mode benefit from higher SNR (factor of 1.34± 0.06 for PCr) compared to standard acquisition due to the Nuclear Overhauser effect (NOE) and substantial PCr/P_i changes during exercise can be observed in ³¹P spectra. ¹H spectral quality is equal to that in single mode experiments and allows Cr2 changes to be monitored.

4. Conclusion

The feasibility of dynamic interleaved localised ¹H and ³¹P spectroscopy during plantar flexion exercise has been demonstrated using a custom-built pneumatic system for muscle activation. This opens the possibility of studying the dynamics of metabolism with multi nuclear MRS in a single run.

Introduction

Proton and phosphorus magnetic resonance spectra can provide complementary information about tissue metabolism. Acquiring two sets of spectra in consecutive experiments is time consuming and may result in different test conditions, e.g. due to fatigue during studies of exercising muscle. Whilst ³¹P MRS has long been used for investigation of exercising muscle, interest in ¹H spectroscopy of muscle tissue has increased during the last few years [1]. As the resonance frequencies of ¹H and ³¹P differ by a factor of 2.47, ¹H and ³¹P spectra can, in principle, be acquired simultaneously without interference, i.e. hydrogen and phosphorus nuclei can be excited by RF pulses on their respective resonance frequencies independently (disregarding heteronuclear coupling effects) and spectra may be acquired synchronously, if hardware permits. When B₀ gradients are used they always act on both nuclei synchronously and simultaneous experiments are no longer fully independent. Interleaving multi-nuclear acquisitions is therfore a more versatile approach than simultaneous acquisition. Here, “interleaved” means nesting ¹H and ³¹P MRS in such way that excitation and acquisition of one nucleus’ spectra is accomplished during the recovery period, T_R, of the other.

Interleaved NMR spectroscopy experiments have been used by various groups for monitoring myoglobin oxygenation and high energy phosphates in exercising human muscle in vivo [2–4]. In these studies localisation of the signal is defined by the sensitive volume of relatively small (d = 2.8−7 cm) surface loop coils. In contrast to previous studies, an interleaved STEAM sequence was developed to acquire gradient-localised ³¹P and ¹H spectra in a single experiment, which has the further benefit of NOE enhancement of the ³¹P spectra. The main advantage of localised versus non-localised spectroscopy is increased specificity: The signal originates only from a distinct region of interest, e.g. an exercising muscle, excluding surrounding tissue which may be activated to a lesser extent or not at all. In addition, the feasibility of dynamic interleaved localised ¹H and ³¹P spectroscopy during plantar flexion exercise is demonstrated using a novel custom built pneumatic system for muscle activation.

Methods

Sequence design

A pulse sequence for interleaved acquisition of ¹H- and ³¹P spectra with STEAM localisation for both nuclei was implemented on a 3T Bruker Medspec whole-body scanner, utilising MultiScanControl-Tool (Bruker Biospin, Ettlingen, Germany). A standard doubletuned surface coil (d = 10 cm), manufactured by Bruker, was used for RF transmission and reception. A CHESS scheme was used for water suppression. Volume selective shimming was achieved with Bruker’s implementation of FASTMAP [5], selecting a 5×5×5 cm³ volume of interest (VOI).

In this implementation, all sequence parameters, including size, shape and position of the voxels, can be chosen independently for the respective nuclei. The freedom to select different VOIs does not pose any restrictions (as voxels may also be set to be identical if required) and follows inherently from the necessity to rescale gradients by gyromagnetic ratios γ1_H/γ31_p in pulse sequence programming. For large flip angles, as used in a STEAM experiment (α = 90°) optimum signal to noise ratio (SNR) per unit time is obtained for T_R = 1.27 · T₁ [6]. This also applies when using a surface coil, as the pulse angle is adjusted to 90° within the VOI (which is small compared to the coil’s dimensions) by seeking the maximum signal achieved with a STEAM acquisition, for each subject. Since longitudinal relaxation times of ³¹P and ¹H metabolites differ significantly, the T_R should preferably be chosen individually for each nucleus. T₁s of ³¹P metabolites are relatively long, in the range 3 s (α-NTP) to 6 s (PCr) at 3 Tesla [7], compared to the T₁ of ¹H metabolites which are typically of the order of 1 s, although they can be as short as 300 ms for lipids [8]. To account for this, the sequence design allows the acquisition of an arbitrary number of ¹H spectra between excitations on the ³¹P channel, as illustrated in Fig. 1. Voxels were positioned in the medial head of m. gastrocnemius (see Fig. 2), because plantar flexion exercise with a straight knee activates primarily the gastrocnemius muscles [9, 10]. The ¹H VOI was 2.7 cm³ and the ³¹P VOI was 31.4 cm³ to partly compensate for the 15 times lower ³¹P sensitivity. The ¹H voxel was not chosen to be the same size as the ³¹P VOI to avoid broad ¹H line widths which would render spectra unusable, but both VOIs were positioned in the same muscle. A short echo time T_E = 7.5 ms, T_M = 30 ms, receiver bandwidth BW = 2500 Hz and 1024 data points were chosen for both nuclei.

Fig. 1.

Schematic representation of the interleaved ¹H and ³¹P STEAM sequence. Several ¹H spectra (four in this example) can be acquired per ³¹P acquisition T_R to optimise SNR of each spectrum despite the large differences in T₁. Rectangles schematically represent RF and B₀ gradient pulses and periods of plantar flexion.

Fig. 2.

Typical position of the volumes of interest for interleaved STEAM spectroscopy in human gastrocnemius muscle. ¹H: V = 2.7 cm³, ³¹P (due to lower sensitivity): V = 43.1 cm³

For calculation of phosphocreatine (PCr), inorganic phosphate (P_i), total creatine (Cr) and pH changes, data were quantified using the AMARES [11] time domain fit routine, as incorporated in the jMRUI software package [12].

Exercise Rig

A non-magnetic rig (Fig. 3) was constructed for plantar flexion exercise with defined constant force during NMR studies. The calf of the extended leg lies on the surface coil which is countersunk in a plywood frame, while the subject pushes a pedal against a pneumatic piston (displacement V = 1.7 l) which generates the force. The lower leg and foot is strapped to the rig and pedal with three belts (white straps in Fig. 3) to minimise motion of the calf. By varying the pneumatic pressure the pedal force can be adjusted arbitrarily, e.g. to match the subject’s maximum voluntary contraction force (MVC) – even remotely and during the NMR measurement, which is useful for multi-power (e.g. ramp) exercise protocols. Exercise prameters are recorded continuously by force and angle sensors on the pedal (see Fig. 3 and Results section), an LED bar indicates the pedal angle for subject feedback. Good linearity of force over the pedal’s angular range is maintained by a large reservoir air tank and the piston’s and pedal’s geometric arrangement. A throttle valve controls the pedal’s recoil.

Fig. 3.

Non-magnetic exercise rig with pneumatic piston, RF coil and sensors for monitoring pedal force and angle.

To reduce calf motion (a potential source of artifacts) during acquisition subjects were instructed to exercise only during periods without RF excitation and reception, i.e. not to exert force on the pedal during and shortly after gradient noise. Before starting the experiment they were trained in a test run. The optimum T_R for ³¹P is 8 s. Although optimum SNR for ¹H would, in principle, allow shorter repetition times, T_R = 2 s leaves 1.2 s between data acquisition and subsequent ¹H excitation; long enough to allow plantar flexion. The most suitable number of ¹H acquisitions between consecutive ³¹P acquisitions is therefore four. After 3 consecutive plantar flexions and ¹H acquisitions, subjects paused once, allowing ³¹P acquisition without motion. Verbal support was given during the experiment whenever necessary.

Subjects

Healthy subjects (n = 7, 6 male, 1 female), aged 31.5±9.2 years, body mass index 23.6 ± 2.5 kg/m² (mean ± SD) were examined. Each subject’s maximum test force was estimated using a Biodex Multi-Joint-System 3 Pro (Biodex Medical Systems, Inc., N.Y., USA) dynamometer, one week before NMR examination. Written informed consent to the protocol, which was approved by the local ethics committee, was obtained from all subjects.

Results

The performance of the interleaved ¹H/³¹P STEAM sequence was evaluated using a two-compartment test object [7]. Contamination of the spectra by signals originating from outside the nominal VOI was C = 0 ± 2% under fully relaxed conditions and C = 3 ± 2% for short T_R, i.e. T_R = 1 s ≪ T1.

Typical ¹H and ³¹P in vivo spectra of exercising muscle are shown in Fig. 4. After a baseline period of 2 minutes at rest, plantar flexion was executed for 6.5 min at 50% MVC (adjusted individually for each subject via air pressure in the reservoir tank) followed by 15 min of recovery. ¹H spectra of resting human calf muscle acquired in a subgroup of n = 4 subjects for in vivo evaluation of the interleaved sequence were equivalent to standard STEAM experiments, whilst SNR in ³¹P spectra acquired interleaved with ¹H spectroscopy was higher by a factor of 1.34 ± 0.06 for PCr and 1.06 ± 0.13 for P_i due to NOE.

Fig. 4.

¹H (left) and ³¹P (right) STEAM spectra from human gastrocnemius muscle at rest, exercise and recovery, acquired in interleaved mode. ¹H spectra (a) were acquired with T_E = 7.5 ms, T_R = 2 s, VOI = 2.7 ml, and averaged to blocks of 64 for display (2’08” per spectrum). The insert shows a magnification (×3) of the Cr2 resonance. ³¹P spectra (b) were acquired with T_E = 7.5 ms, T_R = 8 s, VOI = 31.4 ml, 2 avgs, yielding a time resolution of 16 s. Relative PCr change: 74 %, pH_rest = 7.1, pH_ex = 6.7. (Only the first 50 of 96 ³¹P spectra are shown, for clarity).

The repetition time was 2 s for ¹H spectra (left) and T_R = 8 s for ³¹P spectra (right).

Spectral quality is sufficient to clearly identify peaks of creatine (Cr), choline (TMA) and extra- and intramyocellular lipids (EMCL, IMCL) from ¹H spectra and P_i and PCr from ³¹P spectra. Changes in the EMCL signal in some subjects can be attributed to motion, as tissue with higher EMCL concentration may contaminate the VOI at the onset of exercise despite fixation of the calf and foot and confinement of plantar flexion to periods without NMR acquisitions. The variation of the Cr CH₃ peak area, given as standard error, was however as low as 1.7 %. In ³¹P spectra substantial PCr/P_i changes, as expected during exercise, can be observed without perceptible deterioration of signals, e.g. by motion or artifacts caused by the pneumatic exercise rig (Figs. 4 and 5). Similar to the total creatine CH₃ signal in ¹H spectra, the variation of the sum of PCr and P_i signals was only 2% during the measurement, which indicates that motion did not affect spectral quality. This exercise setup produced a mean (± SD) pH decrease from 7.12 ± 0.11 to 6.94 ± 0.19 and a mean relative PCr depletion of 63 ± 18% after 6.5 min of exercise at 50% MVC (n = 4).

Fig. 5.

Pedal angle and force (F₁, F₂: output of two strain sensors located on the pedal, below the subject’s ball of the foot) recorded during an interleaved ¹H/³¹P spectroscopy experiment. Grey bars indicate periods between the first RF pulse and the end of acquisition of the STEAM sequence, as recorded via the NMR scanner’s trigger output. (Dark grey: ³¹P, light grey: ¹H) Note that there is no motion during NMR acquisitions in this example.

Whilst 6.5 min of moderate exercise is neither expected to deplete IMCL nor alter total creatine concentration, the insert in Fig. 4 shows a drop in peak amplitude of the CH₂ signal of Cr alone during exercise and its recovery post exercise, concomitant with the PCr time course in ³¹P spectra (see Fig. 4), which can not be fully explained by variations in line width.

Discussion

Interleaved MRS experiments have been used to investigate metabolism in muscle tissue [2–4] with surface coils and have been demonstrated with a combination of TEM and surface coils for ¹H, ¹H-{¹³C}, ³¹P STEAM and ISIS NMR spectroscopy in the human brain [13, 14]. In this study we present a method for interleaved acquisition of ¹H and ³¹P NMR spectra with STEAM localisation for both nuclei using a dual tuned surface coil and its application to exercising human calf muscle during plantar flexion in a 3 Tesla whole-body MR system. STEAM was chosen because it is a robust single shot sequence that performs better with the surface coil than e.g. PRESS, especially when large voxels are used for ³¹P spectroscopy (which was also verified on test objects). Localised spectroscopy has the advantage of increased specificity compared to non-localised measuremts. The signal originates only from the region of interest, i.e. one can focus on the muscle which exercises most and exclude surrounding tissue which may be activated to a lesser degree (adjacent muscles) or not at all (skin, fat- or connective tissue). SNR is lower as a consequence of smaller tissue volume contributing to the signal. Our ³¹P VOIs were typically 30 − 40 cm³, while the sensitive volume of a 10 cm surface coil is a hemisphere with a volume of ca. 260 cm³, significantly larger than our VOI. On the other hand, line width and shape, baseline and spectral resolution of multiplets is clearly better in localised spectroscopy.

Note that our ³¹P spectra were acquired with long T_R = 8 s, to achieve optimum SNR for PCr and P_i the metabolites of major interest during exercise. The intensity of metabolite signals with shorter T₁s (NTP) is therefore reduced compared to short T_R measurements, where in fact PCr and P_i are partially saturated and need significant correction for T₁ relaxation.

Small VOIs are favourable in MRS because line width can generally be reduced by decreasing voxel sizes [15]. This is particularly important for ¹H spectra of muscle tissue [16], which are characterised by orientation-dependent splitting of resonances [1]. Characteristic line splitting may become less clear in the accumulation of signal from regions with locally different fibre orientation in a large voxel, as fibre orientation is mostly, but not completely uniform across a muscle. Water suppression is also less efficient for large VOIs, due to the inhomogeneous B₁-field of surface coils. In ³¹P MR spectroscopy SNR is inherently lower, and a big voxel (≳ 30 cm³), covering a large fraction of the exercising muscle under investigation seems more appropriate. Studies mapping T₂ of exercising muscle revealed that recruitment does not vary significantly across a single muscle [17]. We can therefore justifiably assume that our interleaved measurements of ¹H and ³¹P metabolites are representative of the whole medial gastrocnemius muscle. Contributions from adjacent muscles, known to be recruited to a highly different extent during exercise [9, 10, 17] can still be largely excluded from the VOI due to the efficient suppression of contamination [7] of the ³¹P STEAM sequence. If improved time resolution is desired, repetition times may be reduced, with the proviso that exercise be confined to periods without NMR signal excitation or acquisition to minimise motion. Moreover, for metabolites with long T₁ at 3 T, such as PCr [7], this decreases SNR per unit time and may introduce complications in quantification due to T₁ weighting and chemical exchange [18] for other metabolites. The quality of ¹H spectra acquired in interleaved mode here is equivalent to single mode experiments, while the corresponding ³¹P spectra benefit from higher SNR compared to standard acquisition due to the Nuclear Overhauser effect (NOE). The extent of NOE enhancement depends on RF power irradiation and is usually accomplished by saturating the water resonance with continuous wave or WALTZ irradiation [19, 20]. In the experiment described here, the NOE is induced by the RF pulses of the ¹H STEAM sequence which is interleaved with ³¹P STEAM acquisitions. Since neither ¹H pulse amplitude nor duty cycle were maximised to achieve ¹H saturation, but ¹H RF pulse power was optimised for ¹H STEAM signal amplitude and water suppression, the observed NOE is not the maximum achievable. NOE enhancements between 43% (NTP) and 64% (PCr) in calf muscle have been reported in a 1.5T study [19]. In brain, 25-30% enhancement was achieved for PCr at 1.5 and 7T [20,21], P_i signal was increased by 15.7% in this 7T study. These results match our nonoptimised NOE enhancement factors.

Longitudinal relaxation times of PCr and P_i, in human calf muscle are 6.4 ± 0.2 s and 5.2 ± 1.0 s [7], resulting in saturation factors of 0.71 and 0.78, respectively, for T_R = 8 s. This partly compensates the different NOE factors from 1.34 and 1.06 to calibration factors of 0.96 and 0.83, neglecting any chemical exchange effects.

Whilst substantial changes in PCr and P_i are observed in ³¹P spectra, the concentration of metabolites in ¹H spectra (creatine methyl resonance (Cr3), TMA, IMCL) are not expected to change significantly during 6.5 minutes of exercise with 50% MVC. Intramyocellular lipid depletion in muscle becomes measurable with ¹H MRS after prolonged submaximal exercise, e.g. running for several hours [22, 23]. Similar to findings in isometric exercise of tibialis anterior muscle under ischemia [24], a drop in peak amplitude of the methylene group of creatine (Cr2) during exercise and its recovery post exercise, not fully explained by altered line width, can be observed in ¹H spectra in this work (see insert in Fig. 4) simultaneously with the PCr time course observed in ³¹P spectra, which were acquired in the same experiment. During strenuous exercise protocols – and also sporadically at lower exercise intensities in some subjects – a resonance attributed to acetyl carnitine (AcCt) [25] was described in ¹H spectra of human muscle. Interleaved acquisition of ¹H and ³¹P spectra can, in principle, be used to simultaneously acquire time courses of AcCt and high energy phosphates during a high workload exercise protocol.

A localised double quantum filter sequence for lactate detection has recently been incorporated into the interleaved sequence [26] and is currently being tested for in vivo use to quantify changes in lactate concentration which will, together with information from ³¹P spectra acquired simultaneously, be valuable in studying the regulation of glycolysis and cellular acid-base buffering mechanisms.

In summary, we have demonstrated the feasibility of time-resolved localised ¹H and ³¹P spectroscopy of exercising skeletal muscle in a single interleaved experiment. The benefits of this approach are reduced total measurement time and NOE enhancement of ³¹P spectra. Most significantly it opens the possibility of studying the dynamics of metabolism with multi nuclear MRS in a single run, e.g. examining PCr depletion and resynthesis with ³¹P MRS and the concomitant time course of the Cr2 resonance in ¹H spectra. The method has been tested on human calf muscle during exercise on a custom-built pneumatic ergometer and shown to be effective.