J-difference editing (MEGA) of lactate in the human brain at 3T

Abstract

Purpose:

The need to detect and quantify brain lactate accurately by MRS has stimulated the development of editing sequences based on J coupling effects. In J-difference editing of lactate, threonine can be co-edited and it contaminates lactate estimates due to the spectral proximity of the coupling partners of their methyl protons. We therefore implemented narrow-band editing 180° pulses (E180) in MEGA-PRESS acquisitions to resolve separately the 1.3-ppm resonances of lactate and threonine.

Methods:

Two 45.3-ms rectangular E180 pulses, which had negligible effects 0.15-ppm away from the carrier frequency, were implemented in a MEGA-PRESS sequence with TE 139 ms. Three acquisitions were designed to selectively edit lactate and threonine, in which the E180 pulses were tuned to 4.1 ppm, 4.25 ppm, and a frequency far off resonance. Editing performance was validated with numerical analyses and acquisitions from phantoms. The narrow-band E180 MEGA and another MEGA-PRESS sequence with broad-band E180 pulses were evaluated in six healthy subjects.

Results:

The 45.3-ms E180 MEGA offered a difference-edited lactate signal with lower intensity and reduced contamination from threonine compared to the broad-band E180 MEGA. The 45.3 ms E180 pulse had MEGA editing effects over a frequency range larger than seen in the singlet-resonance inversion profile. Lactate and threonine in healthy brain were both estimated to be 0.4±0.1 mM, with reference to N-acetylaspartate at 12 mM.

Conclusion:

Narrow-band E180 MEGA editing minimizes threonine contamination of lactate spectra and may improve the ability to detect modest changes in lactate levels.

INTRODUCTION

Brain lactate plays important roles in several functions including energy supply and cell signaling for neuronal activities, and changes in lactate levels are prominent in several neurological diseases.^{1, 2} The capability to measure lactate noninvasively and accurately by MRS would be of value clinically, but in-vivo detection of lactate signal without contamination from other compounds remains a challenge.

MRS of lactate is commonly focused on the CH₃ proton doublet resonance at 1.3 ppm. Direct measurement of this low-concentration metabolite (< 1 mM) by standard, short-TE MRS is often elusive because of the low signal strength and spectral overlaps with neighboring abundant signals largely from macromolecules. As an alternative, a long echo time approach (e.g., TE ≈ 144 ms) is often used for lactate measurement, in which the lactate 1.3 ppm resonance gives rise to an inverted echo peak due to the effect of its J coupling to the CH proton spin resonance at 4.1 ppm. The macromolecular signals, which undergo much faster transverse relaxation,³ are substantially attenuated and as a result, the spectrum is markedly simplified, thereby improving the detection of lactate compared to short-TE MRS. In practice, however, some residual broad signals are present in the proximity of the lactate 1.3 ppm resonance in long-TE spectra, complicating data processing. Accurate modeling of these interfering multiplet resonances, which may be attributed to small peptides and/or macromolecules,^{3, 4} is required for resolving the lactate and neighboring signals reliably.

Further simplification of spectra may be achievable by means of spectral editing approaches incorporating the J-coupling feature of the lactate spin system. While double-quantum filtering is powerful in terms of single-shot editing capability and consequently minimal artifacts from potential subject motion, absence of other metabolite signals in the filtered spectrum is a major drawback. A single-quantum based J-difference editing method reported in early studies⁵ also erases all resonances other than the target resonance of a single metabolite. MEGA (Mescher-Garwood),⁶ which is also a version of single-quantum based J-difference editing, is currently well established and widely used for selectively detecting low-concentration metabolites having weakly-coupled spins, such as GABA, glutathione, and lactate.^7–9 This may be largely because the spectrum without editing preserves full spectral information while the signal of the target weak resonance is emphasized in the difference-edited spectrum.

For J-difference editing of lactate, the prominent resonance at 1.3 ppm may be targeted by manipulating its J evolution via selective 180° rotations of the coupling partner at 4.1 ppm. Threonine also has a CH₃ proton resonance at 1.3 ppm, coupled to the CH proton resonance at 4.25 ppm, only 0.15 ppm away from the lactate CH proton resonance (Figure 1A).^{10, 11} Thus, threonine can be co-edited but contaminates lactate editing to a degree that depends on the effect of the editing 180° pulse on the threonine CH proton resonance. A prior study reported selective editing of lactate and threonine, in which ~75 ms long dual- and triple-band 180° RF pulses (bandwidth 15 Hz) were implemented within a semi-LASER sequence, and subtractions among three subspectra were used for separating the lactate and threonine 1.3 ppm resonances.¹² Limitations of this approach were the high susceptibility of the editing to potential frequency drifts during the triple-refocused editing scan, and absence of metabolite signals other than those from lactate, threonine, and creatine in subspectra. Selective MEGA editing of lactate and threonine has not been reported to date.

FIGURE 1.

Numerical simulations for MEGA editing of lactate (Lac) and threonine (Thr). (A) Numerically calculated pulse-acquire spectra of lactate and threonine at equal concentrations. The spectra are broadened to singlet FWHM of 2 Hz. (B) Schematic diagram of 45.3-ms E180 MEGA-PRESS (TE 139 ms, TE₁ 24 ms, and TE₂ 115 ms), where the 45.3 ms E180 pulse had a rectangular envelope (bandwidth 17.6 Hz and B₁ 0.26 μT). The time intervals between the E180 pulses was 64.8 ms. The slice-selective 180° RF pulses were 13.2 ms long (bandwidth 1.3 kHz and B₁ 13.5 μT). The spoiling gradient pulses were all 2.4 ms long with 30 mT/m strength. (C) Bloch-simulated inversion profile of the 45.3-ms E180 and density-matrix simulation for the MEGA-PRESS. For each of lactate and threonine, three subspectra were calculated for E180 carrier frequency at 4.1 ppm (subscan A), 4.25 ppm (subscan B), and 100 ppm (subscan C), from which difference spectra, (A–C)/2 and (B–C)/2, were calculated. (D) Bloch-simulated inversion profile of a 14-ms Gaussian E180, truncated at 10%, (bandwidth 81.8 Hz and B₁ 1.48 μT) and density-matrix simulation of the 14-ms E180 MEGA-PRESS (TE 139 ms, TE₁ 24 ms, and TE₂ 115 ms). For each of lactate and threonine, two subspectra were calculated for E180 carrier frequency at 4.1 ppm (subscan X) and 100 ppm (subscan Y), from which a difference spectrum, (X – Y)/2, was calculated. (E) Localized pulse-acquire spectra of lactate, threonine, and an arbitrary uncoupled 3-proton resonance. These localized zero-TE spectra of lactate and threonine were calculated from their unlocalized pulse-acquire spectra multiplied by the ratio of a PRESS singlet to an unlocalized pulse-acquire singlet. All spectra in (C), (D), and (E) were broadened to singlet FWHM of 6 Hz, an approximate in-vivo linewidth. Dashed lines are drawn at 4.1 and 4.25 ppm in the inversion profile figures.

Here we report selective editing of the lactate and threonine 1.3 ppm resonances by MEGA-PRESS. Narrow-band editing 180° RF pulses are introduced for selective rotation of the 4.1 and 4.25 ppm resonances of lactate and threonine in subscans. The MEGA editing performance of the highly selective editing 180° pulses is discussed, with numerical and phantom analyses, as a function of the editing pulse carrier frequency. Evaluation of lactate and threonine in six healthy subjects is presented.

METHODS

MEGA-PRESS sequence design

Two 45.3-ms rectangular editing 180° RF pulses were implemented before and after the second 180° pulse of PRESS (MEGA-PRESS) (Figure 1B). The editing 180° pulse (E180), with a bandwidth of 17.6 Hz at half amplitude, had negligible effects on resonances 0.15 ppm away from the carrier frequency at 3T (1 ppm = 127.75 Hz). Three MEGA runs were designed such that the E180 pulse is tuned at 4.1 ppm, 4.25 ppm, and a frequency far off resonance in subscans A, B, and C, respectively (Figure 1C), and subsequently difference spectra, (A–C)/2 and (B–C)/2, give rise to editing of lactate and threonine, respectively. The slice-selective 90° and 180° RF pulses (vendor supplied) were 9.8 and 13.2 ms long at B₁ of 13.5 μT (bandwidth 4.2 and 1.3 kHz, respectively). The TE of the MEGA-PRESS sequence was 139 ms (TE₁ = 24 ms and TE₂ = 115 ms).

For comparison, a more conventional MEGA-PRESS sequence was designed, in which the E180 pulses had a 14-ms Gaussian envelope, truncated at 10% (bandwidth = 81.8 Hz). This MEGA scheme was designed to evaluate co-editing of threonine when the broad-band E180 pulse was tuned to 4.1 ppm for lactate editing. The slice-selective RF pulses and inter-RF pulse timings of this 14-ms E180 MEGA were kept the same as those of the 45.3-ms E180 MEGA.

Density-matrix simulations

Density-matrix simulations were conducted to assess the lactate editing efficacy of the MEGA-PRESS sequences. The density operator evolution during the sequence was numerically calculated using a product-operator based transformation-matrix algorithm,¹³ programmed with Matlab (The MathWorks Inc.). Transformation matrices were created for the slice-selective 90° and 180° pulses and the E180 pulses, incorporating the actual RF and gradient pulse envelopes, and used for calculating edit-on, edit-off, and difference spectra, as described in a prior study.¹⁴

Experimental

¹H MR experiments were carried out on a Philips Elition 3T human scanner, equipped with a whole-body coil for RF transmission and a 32-channel head coil for reception. Phantom and in-vivo MEGA experiments were conducted with the same RF pulses as described in simulations.

In-vitro MR experiments were undertaken in two phantom solutions (pH 7.2; room temperature): one with 20-mM lactate and 20-mM N-acetylaspartate (NAA), and another with 40-mM threonine and 40-mM NAA. Short-TE STEAM spectra were acquired from a 25×25×25 mm³ voxel at the center of the spherical phantoms (TE 9 ms, TM 11 ms, and TR 12 s). For the 45.3-ms E180 MEGA editing, three subscans were performed with E180 carrier at 4.1, 4.25, and 100 ppm (subspectra A, B, and C, respectively). For the 14-ms E180 MEGA, two subspectra were acquired with E180 tuned to 4.1 and 100 ppm (subspectra X and Y, respectively). In addition, the dependence of MEGA water and lactate signals on the 45.3-ms E180 carrier frequency was tested by incrementing the E180 carrier frequency from −22 to +22 Hz with respect to the water resonance and the lactate methine proton resonance. An oscilloscope was used to monitor the 45.3-ms E180 MEGA sequence to ensure no significant RF drops occurred during the long RF pulses (Supporting Information Figure S1). TR was set at 2 s in all MEGA-PRESS scans.

In-vivo MEGA scans were performed in six healthy subjects (3 males and 3 females; age 33±11 years). The MR protocol was approved by the Institutional Review Board of Vanderbilt University Medical Center. Written informed consent was obtained from the subjects prior to the MR scans. In each subject, both 45.3 and 14 ms E180 MEGA data were acquired from a 30×27×27 mm³ voxel positioned in the parietal brain. FIDs were recorded in multiple blocks, each averaging 16 acquisitions of water-suppressed RF-phase-cycled FIDs. The 45.3 and 14 ms E180 MEGA data were recorded in 18 and 12 blocks (scan durations of 9.6 and 6.4 min, with TR 2 s), respectively. At the beginning of each block, an unsuppressed water signal was acquired from the voxel and used to determine and update the reference frequency for the block using a vendor-supplied tool (Frequency stabilization). A four-pulse variable-flip-angle scheme was used for water signal suppression.¹⁵ First and second order shimming were conducted using a vendor-supplied tool (PB-volume). The carrier frequency of the slice-selective RF pulses was set to 2.7 ppm. MRS data were recorded with sweep width 2.5 kHz and number of sampling points 2048.

Multi-channel combination and eddy-current compensation for the sixteen FIDs of each block were undertaken using the unsuppressed water signal as reference by the scanner built-in tool, resulting in an FID for each block. The FIDs were then processed off-line for additional correction for frequency drifts and RF phase using the NAA and creatine singlets as references. Following apodization with a 1-Hz exponential function and averaging of Fourier-transformed subscan data, difference spectra, (A–C)/2 and (B–C)/2, were calculated in the 45.3-ms E180 MEGA and a difference spectrum (X–Y)/2 in the 14-ms E180 MEGA. Spectral fitting was performed, with LCModel software,¹⁶ on the edit-off and difference spectra. The basis sets comprised in-house calculated model spectra of lactate, threonine, MM12, MM14, and 16 metabolites as shown in Supporting Information Figure S2. Here MM12 and MM14 were macromolecular species, which were modeled to have CH₃ proton resonances at 1.43 and 1.24 ppm, respectively coupled to 4.16 and 4.23 ppm CH proton resonances.^{3, 4} Published chemical-shift and J coupling constants were used in calculating the basis signals.¹⁰ The MM12 and MM14 basis signals were 5 times broader than metabolite basis signals.³ The spectral fitting was conducted between 0.2 and 4.0 ppm, without LCModel built-in macromolecule and lipid bases (NSIMUL = 0) in the fitting of all edit-off and difference-edited spectra. Cramér-Rao lower bounds (CRLB) were returned by LCModel as percentage standard deviations (SD) of metabolite signal measures. Metabolites were quantified in millimolar units, with reference to tNAA (NAA + NAAG) at 12 mM,¹⁷ assuming identical T₁ and T₂ relaxation effects between metabolites. Data are presented as means ± SD.

To achieve proper phasing in LCModel fitting of difference spectra, artificial singlets were added at 2.01 and 3.03 ppm, which had the same FWHMs as those of the NAA and creatine (Cr) singlets in the edit-off spectrum. Lactate was excluded in the basis set for (B–C)/2 spectra and threonine was excluded in the basis sets for all other spectra.

RESULTS

Numerical simulations indicated that, when the 45.3-ms rectangular E180 pulses are tuned to 4.1 ppm in the TE 139-ms MEGA-PRESS (subscan A), the lactate CH₃ proton spins gave rise to a positive signal at 1.3 ppm whose peak amplitude was 61% with respect to an unedited zero-TE lactate signal for singlet FWHM of 6 Hz, ignoring T₂ relaxation effects (Figure 1C). When the E180 was turned off (subscan C), the lactate 1.3 ppm signal was inverted with magnitude 50% of the unedited signal. As a result, a difference spectrum, (A–C)/2, gave a positive lactate signal, with amplitude 54% of the unedited signal. The E180 had negligible effects at 4.25 ppm in the subscan A and consequently the threonine spins underwent nearly ordinary J evolution, leading to an inverted peak at 1.3 ppm, which was similar to that in subscan C. Following an (A–C)/2 subtraction action, a small co-edited threonine signal was present at 1.3 ppm, whose amplitude and area were 23% and 25% with respect to the difference-edited lactate signal, respectively. In contrast, when the E180 was tuned to 4.25 ppm (subscan B), the polarities of lactate and threonine signals were respectively negative and positive, thereby leading to small lactate and large threonine signals in the (B–C)/2 difference spectrum, in which the lactate signal amplitude and area were 22% and 27% relative to the threonine signal, respectively.

When the 14-ms Gaussian E180 was tuned to 4.1 ppm, the threonine 4.25 ppm resonance was largely influenced (86%) (Figure 1D) and as a result, the threonine 1.3 ppm resonance was substantially co-edited. The co-edited threonine signal amplitude was 89% with respect to the edited lactate signal amplitude. Compared to this 14-ms E180 MEGA, the lactate editing yield of the 45.3-ms E180 MEGA was smaller (76%). The lactate and threonine signals were overall larger in edit-on spectra than in edit-off spectra (i.e., lactate in A vs. C, threonine in B vs. C, and both metabolites in X vs. Y), which was due to the chemical-shift voxel displacements arising from the finite bandwidth (1.3 kHz) of the slice-selective 180° RF pulses.

The calculated lactate and threonine signals in Figure 1A agreed well with short-TE STEAM spectra (TE 9 ms) from phantom solutions (Figure 2A,B). Also, the strengths and polarities of the lactate and threonine signals in simulated MEGA sub- and difference spectra, shown in Figures 1C and 1D, were very similar to phantom results. For the lactate-NAA phantom, when the signals were broadened to NAA FWHM of 6 Hz, the lactate signal amplitudes were 28% and 7% with respect to the NAA singlet amplitude in (A–C)/2 and (B–C)/2 spectra, respectively (Figure 2C), as predicted in simulations. For the threonine-NAA phantom case, the threonine peak amplitude was 6% and 30% of the NAA singlet amplitude in (A–C)/2 and (B–C)/2 spectra, respectively (Figure 2E). For the 14-ms E180 MEGA, lactate and threonine were both largely edited, giving similar signal strengths in difference spectra, (X–Y)/2 (Figure 2D,F). The edited lactate signal amplitude in 45.3-ms E180 MEGA was 77% of that in 14-ms E180 MEGA (Figure 2C,D).

FIGURE 2.

Phantom data from MEGA editing of lactate (Lac) and threonine (Thr). (A,B) TE 9 ms STEAM spectra from lactate and threonine phantom solutions (32 averages; TR 2 s). The spectra were broadened to NAA singlet FWHM of 2 Hz. (C,D) MEGA spectra from the lactate-NAA phantom, obtained with 45.3-ms rectangular E180 MEGA and 14-ms Gaussian E180 MEGA. (E,F) MEGA spectra from the threonine-NAA phantom, obtained with 45.3-ms and 14-ms E180 MEGA schemes. The sub- and difference spectra were all broadened to NAA singlet of 6 Hz. The notations of MEGA sub- and difference spectra are identical to those in Figure 1. MEGA spectra are normalized to the NAA singlet amplitude for each phantom. Each subspectrum in C - F was acquired with 32 averages (TR 2 s).

Simulations and experiments showed that, due to the multiplet splitting of the lactate CH proton resonance, MEGA E180 pulses have editing effects over a frequency range larger than seen in the inversion profile of the E180. The difference-edited lactate signal was maximal with the 45.3-ms E180 carrier at 4.1 ppm and decreased progressively as the E180 carrier deviated from 4.1 ppm (Figure 3B and Supporting Information Figure S3). The decreases with changing carrier were notably slower in the lactate signal than in the inversion profile. The frequency range for amplitude of 98% or higher was more than 2-fold larger in the lactate edited signal than in the inversion profile (0.052 vs. 0.024 ppm).

FIGURE 3.

Simulated and phantom signals of lactate in 45.3-ms E180 MEGA. A green line denotes the refocusing efficiency (or an adapted inversion profile) of the 45.3-ms E180, i.e., 1 – (M_z + 1)/2, where M_z is the inversion profile shown in Figure 1C. Blue and red lines represent the dependence of numerically calculated difference-edited lactate peak amplitude and area on the E180 carrier frequency. Blue circles and red asterisks present in-vitro difference-edited lactate peak amplitude and area versus E180 carrier frequency (8 averages in each subscan of each carrier frequency). The simulated and experimental lactate data were normalized to the largest value in each case. The lactate peak amplitudes and areas were obtained from phantom and simulated spectra shown in Supporting Information Figure S3. A calculated quartet of the lactate CH proton spin resonance, broadened to singlet FWHM of 1 Hz, is shown inside the figure.

Figure 4 presents in-vivo MEGA editing results in a healthy subject together with voxel positioning in the parietal brain. For the 45.3-ms E180 MEGA (Figure 4A), the subspectra A and C were similar in terms of metabolite signal strengths and pattern and the resulting difference spectrum (A–C)/2 was devoid of large peaks. A doublet signal was clearly discernible at 1.3 ppm in the difference spectrum. Spectral fitting of the signal with a lactate basis gave a lactate concentration of 0.5 mM. The (B–C)/2 difference spectrum showed a somewhat different pattern. Relatively broad signals were present between 1.1 and 1.5 ppm. Spectral analysis of the signals using threonine, MM12 and MM14 bases resulted in 0.4 mM threonine, 0.8 mM MM12, and 0.8 mM MM14. A relatively large co-edited signal was observable at ~2.61 ppm, which was attributed to NAA CH₂ protons. For the 14-ms E180 MEGA (Figure 4B), a relatively large signal was observed between 1.1 and 1.5 ppm, which may be a sum of lactate, threonine, MM12, and MM14 signals. Fitting of the signals without threonine in the basis set resulted in 0.9 mM lactate+, 0.6 mM MM12, and 0.8 mM MM14. For both 45.3-ms and 14-ms E180 MEGA methods, the inverted signals at ~1.3 ppm in the edit-off spectra (C and Y) were fitted with lactate, MM12 and MM14, without threonine in the basis set. The estimates of the compounds were inconsistent between the data, perhaps due to some differences of the signal pattern around 1.3 ppm between the spectra.

FIGURE 4.

In-vivo MEGA edited spectra from the parietal brain of a healthy subject are presented with voxel positioning. Subspectra A, B, and C of 45.3-ms E180 MEGA (A) and subspectra X and Y of 14-ms E180 MEGA (B) are stacked together with LCModel fitting results of their edit-off and difference spectra (96 averages for each subspectrum). LCModel-returned signals of lactate, threonine, MM12 (macromolecular 1.24 ppm resonance), and MM14 (macromolecular 1.43 ppm resonance) are displayed with their concentration estimates and CRLBs. Lactate was excluded in the fitting of the (B–C)/2 data while threonine was excluded in other fittings. Spectra were normalized to the tNAA singlet amplitude in each MEGA method. Difference spectra are 5-fold magnified. Lactate+ in edit-off and (X–Y)/2 spectra denotes lactate estimation with threonine contamination.

The two MEGA editing methods were tested in the parietal brain in six healthy volunteers. For the 12 edit-off spectra (Supporting Information Figure S4), the mean FWHMs of the NAA and tCr (3.03 ppm) singlets were 6.3±0.3 and 7.0±0.3 Hz and the mean signal-to-noise ratios (SNR) of tNAA and tCr singlets were 147±21 and 81±11, respectively. Here, the SNR was the ratio of the peak amplitudes to the SD of the residuals between 1.0 and 4.0 ppm and the FWHMs were calculated from the LCModel-returned signals. In each of the (A–C)/2 difference spectra from the six subjects, a signal was discernible at 1.3 ppm (Figure 5A). From spectral fitting of the signals using a lactate basis, the lactate concentration was estimated to be 0.4±0.1 mM (N = 6). All (B–C)/2 spectra showed broad signals between 1.1 and 1.5 ppm, in contrast to the (A–C)/2 spectra. Fitting of these signals using a basis set without lactate resulted in a threonine level of 0.4±0.1 mM. The (X–Y)/2 spectra were fitted using a basis set without threonine. The estimated lactate+ level was 0.8±0.1 mM, significantly higher than the lactate estimation from the 45.3-ms E180 MEGA (P < .001).

FIGURE 5.

In-vivo MEGA difference spectra from six healthy subjects are presented together with spectral fitting results. Lactate and threonine signals, returned by LCModel, are shown with their concentration estimates and CRLBs. Lactate was excluded in the fitting of the (B–C)/2 data while threonine was excluded in the fittings of (A–C)/2 and (X–Y)/2 data. Spectra were normalized to the tNAA singlet amplitude in each subject.

DISCUSSION

We report here our measurements of lactate and threonine in the human brain as measured with a novel, narrow-band E180 MEGA-PRESS acquisition at 3T. Our data suggest that selective inversion of the lactate 4.1 ppm and threonine 4.25 ppm resonances can be achieved with 45.3-ms rectangular E180 pulses within the commonly used lactate MEGA editing echo time frame (e.g., 144 ms), improving the purity of edited lactate and threonine signals. While two types of edit-on scans were used for measuring both lactate and threonine in the present study, a lactate-only MEGA scheme is feasible with a single edit-on scan, in which narrow-band 180° pulses are tuned to the lactate CH proton resonance. The MEGA-edited signal yield was reduced due to the use of narrow-band editing 180° pulses, but the capability to generate a high-purity lactate signal would outweigh the reduced editing yield, making it possible to identify modest changes in lactate level more effectively.

The inversion of the lactate CH proton spin polarization by the MEGA editing 180° RF pulses influences the editing yield of the 1.3 ppm resonance. The 45.3 ms pulse, whose half-amplitude inversion bandwidth is 17.6 Hz in uncoupled spins, can only bring about partial inversion of the CH proton spins (Figure 6A), thereby leading to a reduced edit-on lactate signal at 1.3 ppm and eventually reduction in editing yield (Figure 6B). It appears that the 45.3 ms rectangular RF pulse can induce inversion of the lactate CH proton spin polarization at 77% when tuned at 4.1 ppm while 100% inversion of a singlet resonance is achievable by the pulse. The ratio 77% agrees well with our theoretical and phantom edited lactate signal ratios between the 45.3 and 14 ms E180 pulses (Figures 1 and 2). Given the large quartet frequency range (~20 Hz), a few Hz change in the editing pulse carrier frequency may not cause a significant decrease in the degree of inversion of the lactate CH proton spin polarization, thereby leading to a wider MEGA editing frequency range compared to the singlet-resonance inversion profile. In contrast, when the singlet-resonance inversion bandwidth is much larger than the lactate quartet frequency range (e.g., > 100 Hz), the inversion profile for the lactate multiplet resonance is about the same as that for an uncoupled resonance.

FIGURE 6.

(A) Density-matrix simulated inversion profiles of various Gaussian 180° RF pulses are shown for the lactate CH proton quartet resonance at 4.1 ppm (blue line) and an artificial 4.1 ppm uncoupled spin resonance (red line). The duration and truncation percentage of the pulse and the singlet resonance inversion bandwidth are shown at the top of each subfigure. Modeling lactate as an A₃X spin system, the inversion efficiency of the lactate 4.1 ppm resonance was obtained by calculating the coefficient of X_z in the density operator following the action of the RF pulse on the thermal equilibrium longitudinal polarization of the resonance, X_z, for various RF carrier frequencies. A similar method was used for calculating the inversion profile on the singlet resonance. Note that the singlet-resonance inversion profile of the 45.3 ms rectangular pulse shown here is identical to the profile shown in Figure 1C. (B) Density-matrix simulated edit-on, edit-off, and difference spectra of lactate, obtained with an editing pulse carrier frequency at 4.1 ppm, are shown for the editing 180° RF pulses in (A).

Precise determination of frequency and minimization of frequency drifts are important for MEGA editing, particularly when a highly selective editing RF pulse is used. In our study, the 45.3 ms E180 MEGA to 14 ms E180 MEGA edited lactate signal ratio was about the same between simulation and phantom data (76% – 77%) (Figures 1 and 2). For both phantom and in-vivo scans, sixteen FIDs in each block were acquired with an updated editing pulse carrier frequency. As a result, the frequency variations of large singlets (NAA and Cr) during the 10-min in-vivo scans were measured to be within ±2 Hz. Also, the NAA estimation from co-edited NAA aspartate signal, which was the largest in difference spectra, was fairly consistent across the three types of edited data from six subjects (10.1±0.9, 9.7±0.7, and 9.4±0.6 mM) and agreed well with the NAA estimation from edit-off spectra (10.7±0.7 and 10.9±0.6 mM). Taken together, issues with frequency determination and frequency drifts were unlikely considerable in our study.

While many prior lactate MRS studies utilized TE of 144 ms, at which the inphase coherence of the 1.3 ppm resonance is maximized, the TE of the present study was set at 139 ms. Given that the peak amplitude is of high interest in MRS, we optimized the TE and inter-RF timings for difference-edited lactate peak amplitude for in-vivo singlet FWHM (5 – 7 Hz) and for a published lactate T₂ of 240 ms¹⁸, similarly as in a prior study¹⁹. Our simulations indicated that the difference-edited lactate peak area was maximum at TE of 144 ms, but the difference-edited lactate peak amplitude was maximum at TE 139 ms, which was the shortest possible TE in the 45.3 ms E180 MEGA sequence for given RF and gradient pulse durations. Taking into account the T₂ relaxation effects, in-vivo linewidth, and the finite bandwidth of slice-selective 180° RF pulses (1.3 kHz), the TE and inter-E180 pulse timings were set to those used in the study. In addition, the 14 ms Gaussian 180° pulse was chosen as a comparative editing pulse since the pulse, with bandwidth of ~82 Hz, has essentially full inversion effects on the lactate CH proton spins and thus gives much higher editing efficiency than the 45.3 ms rectangular pulse. Also, the 14 ms pulse has negligible effects on the choline singlet resonance and thus the resulting cancelation of the choline signal as well as the creatine and NAA singlets in difference spectra can be used as a measure of the cancelation performance.

In this study, we report the results from the spectral fitting that was conducted without both lactate and threonine in the basis set. When the metabolites were both included in the basis set, most of cases resulted in zero (or very low) estimation of threonine while the entire signal was assigned to lactate. The threonine C2-proton resonance at 3.58 ppm, which was partially co-edited, was not sufficiently strong for decomposition of lactate and threonine signals. Decomposition of the 1.3 ppm composite signals into lactate and threonine was not achievable in the present study.

Given that a symmetric tuning approach was used for eliminating co-editing of macromolecules in MEGA editing of GABA,²⁰ one may consider MEGA editing of lactate with symmetric tuning of editing 180° pulses in alternate scans (i.e., 4.1 and 4.4 ppm) to cancel out co-editing of threonine. It appears that a larger-bandwidth editing pulse (e.g, bandwidth up to 32 Hz) can be used for the symmetric tuning approach, enjoying an increase in lactate editing yield (by ~20%) compared to our 45.3-ms E180 MEGA, as indicated in a computer simulation (data not shown). In practice, however, the effect of the E180 is very sensitive to frequency variations in the shoulder region of the inversion profile, so in-vivo application of the symmetric tuning approach would require exceptionally careful management of frequency. For this reason, a symmetric tuning approach was not pursued in the present study.

Several prior studies used MEGA for measuring brain lactate and reported lactate levels with reference to tCr. The lactate/tCr ratio in our data from the narrow-band E180 MEGA was 0.06±0.01. This is 2 – 3 fold smaller compared to the results of the prior MEGA studies, in which the bandwidths of the editing 180° pulses were large (> 60 Hz) and consequently the lactate measurements could contain contaminations from threonine,^{9, 21, 22}. The lactate estimate of the current study (~0.4 mM) is in good agreement with the result in a prior study of non-MEGA difference editing of lactate and threonine.¹² Our threonine estimation is slightly lower compared to the prior non-MEGA study (0.4 vs. 0.6 mM), but higher than the result (~0.3 mM) from a multiple-bond threonine editing study in rat brain,¹¹ which focused on the threonine 3.58 ppm resonance. Threonine measures from human brain biopsy were 0.3 – 0.4 mM.²³ In our narrow-band E180 MEGA data, the spectral pattern around 1.3 ppm in the (B–C)/2 difference spectra was broad, indicating that several compounds were co-edited in the threonine editing. Further study is required to resolve a small threonine peak from the overlapping signals in MEGA editing of threonine. As an alternative, editing strategies such as localized 2D NMR and Hartmann-Hahn polarization transfer may be considered to overcome the spectral complexities.^{11, 24, 25}

Lastly, the MM14 estimation from (A–C)/2 spectra was very low compared to that from (B–C)/2 spectra (0.1 vs. 0.6 mM). It is most likely that the low MM14 estimation in (A–C)/2 spectra was an underestimation arising from the weakness of the co-edited MM14 signals at 1.43 ppm in the spectra, similarly as zero estimations of glutamate and glutamine from (A–C)/2 and (B–C)/2 spectra (Supporting Information Figures S2 and S5).

CONCLUSION

We have demonstrated selective editing of the overlapping 1.3 ppm resonances of lactate and threonine in the human brain, achieved by means of narrow-band E180 MEGA at 3T in vivo. Brain lactate and threonine levels were both estimated to be approximately 0.5 mM. The frequency range over which the narrow-band E180 was in effect was relatively large, alleviating the vulnerability of the editing efficiency to frequency drifts. Reduction in the lactate editing yield in narrow-band E180 MEGA may be outweighed by the improved purity of the measures. The proposed editing scheme has potential for reliable evaluation of changes in lactate level in brain diseases and/or after interventions, in which alterations in lactate level are modest.

Supplementary Material

Supporting Information

FIGURE S1. Oscilloscope captures of the 45.3-ms long rectangular E180 MEGA-PRESS sequence. The display in (A) is vertically magnified in (B). The RF pulse intensity was maintained without noticeable RF drops during the two 45.3-ms rectangular pulses. The RF field intensity was 13.5 μT for the PRESS 90° and 180° pulses and 0.26 μT for the 45.3-ms rectangular pulses. The RF pulses were monitored with a pick-up coil (~3 cm diameter) that was put on the patient table inside the bore of the magnet.

FIGURE S2. LCModel-returned signals are shown together with their concentration estimates for the difference spectra of Figure 4.

FIGURE S3. Phantom and simulated lactate spectra of the 45.3-ms E180 MEGA are shown for various E180 carrier frequencies (3.92 – 4.28 ppm). The lactate peak amplitude and area presented in Figure 3 were calculated from these spectra. Spectra are shown for 1.12 – 1.52 ppm. For the phantom data, 8 signals were averaged in each subscan.

FIGURE S4. In vivo edit-off spectra from six subjects are shown together with LCModel fits. The spectra in the lower panel are zoomed for 1 – 1.7 ppm region.

FIGURE S5. Calculated edit-on, edit-off, and difference spectra of NAA (2.01 ppm), MM12, MM14, lactate and threonine at equal concentrations are shown for 45.3 and 14 ms E180 MEGA sequences. The spectra were broadened to singlet FWHM of 6 Hz for NAA, lactate and threonine, and 30 Hz for MM12 and MM14.