³¹P-MRS of Healthy Human Brain: ATP synthesis, Metabolite Concentrations, pH, and T₁ Relaxation Times

Abstract

The conventional method for measuring brain ATP synthesis is ³¹P saturation transfer (ST), a technique typically dependent on prolonged pre-saturation at γ-ATP. In this study, ATP synthesis rate in resting human brain is evaluated using EBIT (Exchange Kinetics by Band Inversion Transfer), a technique based on slow recovery of γ-ATP magnetization in the absence of B₁ field following co-inversion of PCr and ATP resonances with a short adiabatic pulse. The unidirectional rate constant for the Pi → γ-ATP reaction is 0.21 ± 0.04 sec⁻¹ and the ATP synthesis rate is 9.9 ± 2.1 mmol/min/kg in human brain (n = 12 subjects), consistent with the results by ST. Therefore EBIT could be a useful alternative to ST in studying brain energy metabolism in normal physiology and under pathological conditions. In addition to ATP synthesis, all detectable ³¹P signals are analyzed to determine the brain concentration of phosphorus metabolites including UDPG at ~ 10 ppm, a previously reported resonance in liver tissues and now confirmed in human brain. Inversion recovery measurements indicate that UDPG, like its diphosphate analogue NAD, has a shorter apparent T₁ than monophosphates (Pi, PMEs and PDEs) but longer than triphosphate ATP, highlighting the significance of ³¹P-³¹P dipolar mechanism in T₁ relaxation of polyphosphates. Another interesting finding is the observation of ~40% shorter T₁ at intracellular Pi relative to extracellular Pi, attributed to the modulation by the intracellular phosphoryl exchange reaction Pi ↔ γ-ATP. The sufficiently separated intra- and extra-cellualr Pi signals also permit the distinction of pH between intra- and extra-cellular environments (pH 7.0 vs pH 7.4). In summary, the quantitative ³¹P MRS in combination with ATP synthesis, pH and T₁ relaxation measurements may offer a promising tool to detect biochemical alternations at early stages of brain dysfunctions and diseases.

Graphical abstract

INTRODUCTION

The rate of ATP synthesis may be measured in functioning tissue by ³¹P NMR magnetization transfer (MT) methods (1–3). However, due to presence of multiple competing magnetization exchange pathways and nature of MT pulsing technique, in vivo applications of ³¹P MT for ATP synthesis is generally challenging (3,4), especially in the brain where one has to consider additional limiting factors such as tight SAR restriction, and the presence of compartmentalized phosphates (2) and broad signal from less mobile tissue phosphates (5). So far, the only ³¹P MRS technique that has proven useful for measuring human brain ATP synthesis is saturation transfer (ST), a subset of MT method based on the ³¹P chemical exchange saturation transfer between Pi and γ-ATP (1). Technically, ST relies on selective irradiation of γ-ATP with a long RF pulse or pulse train, resulting in a reduced Pi signal through ATPase mediated reaction (Pi → γ-ATP). While ST is commonly used for probing ATP energy metabolism, it typically requires a prolonged irradiation at γ-ATP (5 – 9 sec) for two major reasons: 1) to build up a steady-state MT effect at Pi, so that the rate constant of ATP synthesis k_{Pi → γATP}, can be easily evaluated using a simplified formula (3); and 2) to overcome the rapid drainage of γ-ATP magnetization to other “unwanted” pathways, especially the exchange reaction PCr ↔ γ-ATP, which occurs at a much faster rate due to the larger pool size of PCr and the higher enzymatic activity of creatine kinase (CK) (2,4). Unfortunately, the prolonged RF pulsing is prone to undesirable side effects such as increasing SAR exposure, and off-resonance saturation, the so-called “spillover” effect (6,7). This latter artifact is expected to worsen at an increased magnetic field due to shortened T₁ and T₂ at γ-ATP (which demands a higher B₁ power for saturation) (4,8,9).

Selective inversion of the γ-ATP is an alternative approach to ST (15), but the magnitude of its magnetization transfer to Pi is small (4,10). However, when γ-ATP is co-inverted with PCr using a band inversion approach termed EBIT (Exchange Kinetics by Band Inversion Transfer), a multi-fold enhancement in MT is achieved, resulting in a Pi reduction comparable to that generated by ST but with much improved time efficiency (11). A key concept in EBIT is the role of PCr as a “magnetization buffer”. Upon co-inversion with γ-ATP by EBIT, PCr continuously replenishes a diminishing γ-ATP pool via the rapid exchange reaction PCr ↔ γ-ATP (11,12). This leads to an increased exchange time for the MT effect to build up between Pi and γ-ATP, equivalent to several seconds of ST saturation pulsing. The primary aim of this study is to evaluate the feasibility of this EBIT technique for measuring ATP synthesis rate in the human brain. Since band inversion transfer impacts much of the ³¹P NMR spectrum (all of the ATP signals plus phosphocreatine and UDP glucose), we also measured the apparent T₁ relaxation times of all observable phosphates and their concentrations.

METHODS

Human Subjects

The protocol was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center. Prior to the MRS study, informed written consent was obtained from all participants. Twelve subjects (7 male and 5 female), aged 43 ± 15 yr, BMI 24 ± 4, resting heart rate 70 ± 13, and peripheral capillary oxygen saturation (SpO2) 96 ± 2% participated the study. All subjects were in good general health with no history of peripheral vascular, systemic, myopathic, cancer, psychiatric or neurodegenerative diseases. Heart rate and blood oxygen saturation levels were monitored during the scan. The study was well-tolerated by all subjects.

MRS Protocol

All subjects were positioned head-first and supine in the MRI scanner (7T Achieva, Philips Healthcare, Cleveland, OH), with the back of the head positioned in the center of detection RF coil (Philips Healthcare, Cleveland, OH). The coil was a partial volume, double-tuned ¹H/³¹P quadrature TR coil consisting of two tilted, partially overlapping 10 cm loops. Axial, coronal, and sagittal T2-weighted turbo spin echo images were acquired for shimming voxel. Typical imaging parameters included field-of-view 180 × 180 mm (FOV), repetition time (t_R) 2.5 s, echo time (TE) 80 ms, turbo factor 15, in-plane spatial resolution 0.6 × 0.7 mm², slice thickness 6 mm, gap 1 mm, bandwidth 517 Hz, number of acquisitions (NA) one, and acquisition time 2.1 min. Second order ¹H-based automatic volume shimming was applied prior to ³¹P spectral acquisitions.

Quantitative ³¹P spectra were acquired using a non-localized block-shaped excitation pulse with B₁ 59 μT, pulse width 0.22 ms, flip angle 55°, and pulse dead-time DT (the delay prior to FID sampling for suppression of the broad ³¹P signals from less mobile tissue phosphates) 0.6 ms, with the transmitter frequency centered at 700 Hz upfield from the resonance of PCr. Precautions were taken to avoid signal contribution from the neck muscles by positioning the head chin-up, aligning the occipital lobe at the coil iso-center, and placing a supporting cushion pad underneath the neck. Other ³¹P NMR parameters include TR 30 sec, sampling points 4 K, zero-filled to 8 K prior to FT, 6 acquisition averages and scan time 3 min. For quantitative comparison of different ³¹P peaks, the transmitter excitation profile was obtained by monitoring ³¹P signal intensity against transmitter carrier frequency which was varied from downfield PC to upfield β-ATP in 600 Hz step to cover the whole chemical shift range of all metabolites. The chemical shifts of all ³¹P metabolites were referenced to PCr at 0 ppm. Gaussian apodization (6 Hz) was applied to each FID prior to Fourier transformation using the scanner software (SpectroView, Philips Healthcare).

The EBIT sequence consisted of an adiabatic inversion pulse, followed by a variable post-inversion delay time t, a hard 90° readout pulse, and a recovery period with TR = 30 sec. A total of 12 delay times (t_D = 3, 147, 341, 647, 1088, 1559, 2275, 3254, 4414, 5673, 7541 and 10000 msec) were used. The band-inversion pulse was a short trapezoid-shaped adiabatic pulse (pulse width pw = 42 msec, including 7 msec of pre- and post-ramp time), maximal B₁ 11.5 μT, inversion bandwidth of 2500 Hz, and centered at 150 Hz upfield from α-ATP at −7.5 ppm. The total data acquisition time for EBIT was 36 min for 6 acquisition average.

To measure the T₁ relaxation time by the inversion-recovery method, a 5-ms adiabatic inversion pulse with a 7000 Hz bandwidth was used to invert all ³¹P peaks. A total of 12 post-inversion delay times (t = 5, 385, 801, 1259, 1768, 2343, 3002, 3775, 4708, 5887, 7489 and 10000 msec) were used at a constant TR of 30 sec. A preparatory dummy scan was applied prior to data acquisition. This experiment was conducted on 6 subjects.

³¹P Spectral Analysis

The frequency-domain ³¹P spectra were baseline corrected, and each of the ³¹P peaks was fitted by a Voigt lineshape (a combination of Gaussian and Lorentzian lineshape) using ACD software (Advanced Chemistry Development, Inc., Toronto, Canada). The area of the fitted ³¹P peaks (S_m) was converted to ³¹P metabolite concentration (C_m), in reference to γ-ATP as an internal standard (C_γATP = 3 mM in occipital lobe of brain (1,2)), based on the following formula:

$$C_m=\frac{S_m}{S_{\gamma \mathit{\text{ATP}}}}\frac{{\rho }_{\gamma \mathit{\text{ATP}}}(\Delta f)}{{\rho }_m(\Delta f)}{C}_{\gamma \mathit{\text{ATP}}}$$ [1]

where ρ_m(Δf) represents the transmitter excitation profile for a given metabolite peak at offset frequency Δf (the difference between the resonance frequency of the ³¹P peak and the center frequency of the RF excitation pulse).

Evaluation of Rate Constant for ATP Synthesis

For the kinetic exchange reaction Pi↔ γ − ATP, the evolution of Pi Z-magnetization in the inversion delay time t, can be described by the Bloch-McConnell equation:

$${\dot{M}}_{Pi}(t)=\frac{M_{Pi}^o-M_{Pi}(t)}{T_{1,Pi}}-k_{Pi\to \gamma \mathit{\text{ATP}}}{M}_{Pi}(t)+k_{\gamma \mathit{\text{ATP}}\to Pi}{M}_{\gamma \text{ATP}}(t)$$ [2]

where ${\dot{M}}_{Pi}(t)$ is the rate of Pi Z-magnetization change at any given delay time t, and k denotes pseudo first-order kinetic rate constant, with the forward and reverse rate constants related by

$$k_{\gamma \mathit{\text{ATP}}\to \mathit{pi}}=k_{Pi\to \gamma \mathit{\text{ATP}}}\frac{M_{Pi}^o}{M_{\gamma \mathit{\text{ATP}}}^o}$$ [3]

in which $M_{Pi}^o$ and $M_{\gamma \mathit{\text{ATP}}}^o$ denote the Z-magnetization at Boltzmann thermal equilibrium for Pi and γ-ATP, respectively. Substituting k_γATP→Pi in Eq [2] by Eq [3] gives:

$${\dot{m}}_{Pi}(t)=\frac{1-m_{Pi}(t)}{T_{1,Pi}}-k_{Pi\to \gamma \mathit{\text{ATP}}}(m_{Pi}(t)-m_{\gamma \mathit{\text{ATP}}}(t))$$ [4]

where m is the normalized Z-magnetization defined by:

$$m_i(t)=M_i(t)/M_i^o(i=\text{Pi},\gamma \text{-ATP})$$ [5]

Since at m_Pi(t) = 1 at t = 0, the kinetic rate constant for ATP synthesis k_Pi→ATP can be evaluated by the following formula (derived from Equation [4]):

$$k_{Pi\to \gamma \mathit{\text{ATP}}}=\frac{{\dot{m}}_{Pi}(0)}{m_{\gamma \mathit{\text{ATP}}}(0)-1}$$ [6]

where m_γATP(0) is γ-ATP Z-magnetization (normalized) immediately after inversion, which can be obtained experimentally; ${\dot{m}}_{Pi}(0)$ is the initial rate of Pi Z-magnetization change, which can be derived by fitting Pi Z-magnetization according to the following bi-exponential equation (with satisfaction of the boundary conditions m_Pi|_{t = 0} = 1 and m_Pi|_t→∞ = 1):

$$m_{Pi}(t)=1-a(e^{-{\lambda }_{1}t}-e^{-{\lambda }_{2}t})$$ [7]

where a, λ₁ and λ₂ are three fitting constants. This leads to the following first-order derivative:

$${\dot{m}}_{Pi}(t)=a({\lambda }_1{e}^{-{\lambda }_{1}t}-{\lambda }_2{e}^{-{\lambda }_{2}t})$$ [8]

from which ${\dot{m}}_{Pi}(0)$ can be obtained at t = 0. Thus the rate constant k_Pi→γATP can be evaluated either by Eq [6] using the initial magnetization rate ${\dot{m}}_{Pi}(0)$ as derived from bi-exponential data fitting to $m_{Pi}(t)$, or by Eq [4] using ${\dot{m}}_{Pi}(t)$ data fitting to the measured (1 − m_Pi(t)) and (m_Pi(t) – m_γATP(t)).

Evaluation of pH in Brain

The intra- and extra-cellular pH values were evaluated from the chemical shift of the corresponding Pi peaks (δ_Pi, in ppm) in reference to PCr (δ = 0 ppm) using the following formula:

$$pH=pKa+\text{log}\frac{{\delta }_{Pi}-{\delta }_a}{{\delta }_b-{\delta }_{Pi}}$$ [9]

where the H₂PO₄⁻ ↔ H⁺ + HPO₄²⁻ deprotonation constant pKa = 6.73, and the ³¹P limiting shifts δ_a = 3.275 ppm (for acidic protonated species H₂PO₄⁻) and δ_b = 5.685 ppm (for basic deprotonated species HPO₄²⁻) were used in the data analysis.

Evaluation of Mg²⁺ and MgATP Concentration in Brain

The free Mg²⁺ concentration in brain was evaluated from the chemical shift difference between α- and β-ATP (δ_α−β, in ppm) using the following formula (13):

$$[M{g}^{2+}]=k_d\frac{{\delta }_{\mathit{\text{ATP}}}-{\delta }_{\alpha -\beta }}{{\delta }_{\alpha -\beta }-{\delta }_{\mathit{\text{MgATP}}}}$$ [10]

where the MgATP effective disassociation constant k_d = 0.05 mM and the limiting shift constants δ_ATP = 10.82 ppm and δ_MgATP = 8.32 ppm were used in the data analysis.

The total concentration of all species of MgATP complexes was calculated by:

$$[\mathit{\text{MgATP}}]=\frac{[M{g}^{2+}]}{k_d+[M{g}^{2+}]}{[\mathit{\text{ATP}}]}_t$$ [11]

where [ATP]_t represents the total concentration ATP in the brain (= 3 mM]).

Statistical Analysis

All data are reported as mean ± standard deviation, calculated using Matlab.

RESULTS

In Vivo Brain ³¹P Spectrum

A group-averaged (n = 12 subjects) in vivo ³¹P spectrum collected from resting human brain at 7T is shown in Figure 1a. After baseline correction (Figure 1b), the ³¹P signal intensities are evaluated by lineshape fitting (Figure 1c) and then converted to metabolite concentrations in reference to γ-ATP peak at −2.52 ppm assuming [ATP] = 3 mM (Table 1). The conversion takes into account the number of contributing phosphate groups and the non-uniformity of ³¹P excitation profile.

FIG. 1.

A 7T ³¹P MR spectrum, group-averaged for 12 subjects, acquired from resting human brain using a spatially nonselective single-pulse sequence with long TR of 30 sec, before (a) and after (b) baseline correction, and the result of ³¹P peak fitting (c). The fitting residual (bottom red trace) in (c) represents the spectral subtraction (b – c). Abbreviation: PE, phosphoethenolamine; PC, phosphocholine, GPE, glycerophosphoethanolamine; GPC, glycerophosphocholine; Piⁱⁿ and Pi^ex, intra- and extracellular inorganic phosphate; PCr, phosphocreatine; ATP, adenosine triphosphate; NAD, nicotinamide adenine dinucleotide; UDPG, uridine diphosphate glucose.

Table 1.

³¹P Chemical shift (δ), line-width (LW_1/2), apparent relaxation time (T_1,app), and concentration (Conc.) of ³¹P metabolites in resting human brain at 7T (N = 12 subjects)

	# of PO4 groups	δ (ppm)	LW1/2 (Hz)	T1,app (s)	Conc. (mM)
Monophosphates
PE	1	6.76 ± 0.05	71.1 ± 6.2	6.33 ± 1.10	2.27 ± 0.16
PC	1	6.24 ± 0.02	62.7 ± 13.2	4.31 ± 1.04	0.30 ± 0.12
Piex	1	5.24 ± 0.05	97.1 ± 27.7	5.80 ± 1.07	0.30 ± 0.09
Piin	1	4.82 ± 0.01	66.1 ± 5.3	3.70 ± 0.46	0.85 ± 0.11
GPE	1	3.50 ± 0.01	60.0 ± 5.5	6.79 ± 0.95	0.80 ± 0.18
GPC	1	2.95 ± 0.01	62.7 ± 4.4	5.82 ± 0.88	1.32 ± 0.18
PCr	1	[0]	48.4 ± 6.9	3.39 ± 0.17	4.37 ± 0.39
diphosphates
NAD	2	−8.21 ± 0.04	127.2 ± 17.8	2.07 ± 0.13	0.28 ± 0.13
UDPG	2	−9.72 ± 0.04	101.7 ± 10.2	2.95 ± 0.36	0.08 ± 0.04
triphosphates
ATP	3
γ-		−2.52 ± 0.01	122.6 ± 6.7	1.70 ± 0.15	[3.0]
α-		−7.56 ± 0.02	101.4 ± 3.3	1.35 ± 0.14	3.09 ± 0.23
β-		−16.15 ± 0.05	182.6 ± 11.9	1.13 ± 0.09	2.82 ± 0.25

The ³¹P spectrum at 7T is able to clearly detect a small peak upfield at −9.7 ppm (Figure 1 inset), previously observed but not assigned in the human brain (8), although a similar signal has been observed in other spectra from human subjects (14). Based on its chemical shift, it is assigned to a di-phosphate metabolite UDPG (0.08 mM), a glucose precursor for synthesis of glycogen (13). Between PME and PDE peaks, two distinct pools of Pi with an intensity ratio of ~ 1 : 3 can also be observed, assigned to intracellular (δ(Piⁱⁿ) = 4.82 ppm) and extracellular (δ (Pi^ex) = 5.24 ppm) inorganic phosphate (19), respectively. They correspond to a concentration of 0.85 mM (Piⁱⁿ) and 0.30 (Pi^ex), and a pH value of 6.99 (Piⁱⁿ) and 7.38 (Pi^ex).

ATP contributes three separate ³¹P signals at −2.5, −7.6 and −16.2 for γ-, α- and β-ATP, respectively. In principle, any one of them can be used as an internal reference to quantify other metabolites. However, given the fact that α-ATP signal is partially overlapped with neighboring NAD signal and that β-ATP signal is significantly broadened and away from other metabolite signals (Figure 1), we chose to use γ-ATP signal as the reference. By integral, the intensity ratio of γ- : α- : β-ATP signals is 1.00 : 1.03 ± 0.07 : 0.94 ± 0.09 after correction for excitation profile (Figure 4b). The slightly lower intensity of β-ATP relative to those of γ- and α-ATP is most likely due to T₂ effect (4).

FIG. 4.

(a) Plots of normalized ³¹P magnetization against the inversion delay time t for the evaluation of apparent T₁ relaxation time of brain metabolites. The data were from the signal intensity measurements in the inversion-recovery ³¹P spectra shown in Figure 3, and the solid curves represent the fitting based a mono-exponential process. Note the rapid relaxation of ATP, NAD and UDPG (P-metabolites with two or more coupling phosphate groups) as compared to PME, PDE, Pi and PCr (P-metabolites with a single phosphate group). (b) Excitation profile of the ³¹P spectrum showing the dependence of signal intensity on transmitter offset frequency. The excitation profile was used for correction of metabolite concentration based on signal intensity.

T₁ Relaxation Time of ³¹P Metabolites

The T₁ values of all ³¹P resonances are measured using the inversion recovery method (Figure 3). The dependence of ³¹P signal intensity on inversion delay time revealed a wide range of recovery rates for these different metabolites. Though the T₁ relaxation (by dipolar interaction, plus CSA) may not be the only mechanism governing the inversion recovery of every ³¹P signal, all follow reasonably well a mono-exponential trend (Figure 4a). For the 12 ³¹P signals observed, the recovery time constant is in the following order: β-ATP < α-ATP < γ-ATP < NAD < UDPG < PCr < Piⁱⁿ < PC < Pi^ex ~ GPC < PE ~ GPE. The observed trend is that the ³¹P signals of diphosphates (including NAD, UDPG) and triphospahte (α-, β- and γ-ATP) recover more rapidly than those of monophosphates (including PE, PC, Pi, GPE, GPC and PCr). The recovery rates reflect the T₁ relaxation times; however, for those ³¹P spins involving chemical exchange and/or cross-relaxation (such as Pi, PCr, α-, β- and γ-ATP), the observed recovery time constants represent the apparent T₁ relaxation times.

FIG. 3.

A typical ³¹P MR spectral series acquired from resting human brain at 7 T using a non-selective inversion-recovery sequence at TR 30 sec, NA = 6 and varying inversion delay time t. The insets showed the regional spectra enlarged for clear view of ³¹P signals of low-intensity in the chemical shift regions of 4.5 – 5.5 ppm (left) and −6.5 – −10.5 ppm (right); the spectra were colored coded from blue to red to indicate the gradual increase of the inversion delay time t. For comparison, the last trace of the spectral series represents the fully relaxed brain ³¹P MR spectrum acquired at TR 30 sec without applying the inversion pulse.

ATP Turnover Rate in Brain

In order to evaluate ATP synthesis, the magnetization exchange effect between γ-ATP and Pi is monitored using EBIT sequence at varying inversion delay times (Figure 5). The band inversion pulse is tailored to invert PCr, α-, β- and γ-ATP but not any of the downfield spins (Pi, PMEs and PDEs). For both Piⁱⁿ and Pi^ex, there is no measurable difference in signal intensity before and immediately after the band inversion. With an increase in inversion delay time, the Piⁱⁿ signal at 4.8 ppm clearly shows a trend of decreasing until it reaches an intensity minimum, occurred at t ~ 2.8 sec. After that time point, Piⁱⁿ starts to recover gradually toward its equilibrium state. The maximum Piⁱⁿ signal reduction is measured ~25% relative to its M° (Figure 5). In contrast, the P^ex signal at 5.2 ppm remains largely invariant, i.e., independent of the change in inversion delay time, indicating that the extracellular phosphate is not involved in magnetization exchange with γ-ATP.

FIG. 5.

A typical ³¹P MR spectral series acquired with EBIT sequence using a band inversion pulse selectively inverting PCr and ATP spins, and followed by a varying delay time t All data are shown with identical y-scale. The insets show the regional spectra in the chemical shift region of 2 – 8 ppm, enlarged for a clear view of magnetization transfer effect at intracellular inorganic phosphate (Piⁱⁿ) from γ-ATP due to inversion delay (green trace: inversion delay time t = 3 ms; red trace: t = 3.2 sec; and black trace: fully relaxed spectrum). For comparison, the last trace of the spectral series represents the fully relaxed brain ³¹P MR spectrum acquired at TR 30 sec without applying the inversion pulse.

For γ-ATP, the inversion fraction is 55% (m_γATP(0) = −0.55). After inversion, the recovery rate of γ-ATP signal lags behind α- and β-ATP, but it is faster than that of PCr (Figure 4). When Piⁱⁿ reaches its minimum intensity, the recovery of γ-ATP is at ~ 60% of its equilibrium magnetization (M°). According to Equation [4], at this time point, the net magnetization transfer between γ-ATP and Pi (= k (m_Pi − m_γATP)) is balanced by the Pi relaxation (= (1 – m_Pi)/T_1,Pi).

To evaluate the rate constant, k_Pi→γ-ATP, the m_Pi ~ t dataset collected for Piⁱⁿ is fitted to a bi-exponential equation (Equation [8]), yielding an initial rate constant ${\dot{m}}_{Pi}(0)=-0.31\pm 0.06$ (n = 12 subjects). Together with the measured γ-ATP inversion fraction (m_γATP(0) = −0.55), this leads to a k_Pi→γ-ATP value of 0.203 ± 0.043 sec⁻¹ ( $={\dot{m}}_{Pi}(0)/(1-m_{\gamma \mathit{\text{ATP}}}(0))$,Equation [6]), as comparable to k_Pi→γ-ATP = 0.214 ± 0.045 sec⁻¹ obtained by fitting ${\dot{m}}_{Pi}~t$ data to Equation [4], which also yields an intrinsic T₁ value of 6.71 ± 1.82 sec for Piⁱⁿ (n = 12 subjects). With an intracellular phosphate concentration of 0.85 mM (Table 1), the unidirectional flux from Piⁱⁿ to γ- ATP in the brain is measured 10.9 ± 2.3 mM/min, or 9.9 ± 2.1 mmol/min/kg wet weight converted with a brain tissue density of 1.1 g/ml.

DISCUSSION

³¹P Metabolite Measurements

In this ³¹P MRS study, quantitative spectral data are acquired from resting human brain at 7T using a FID-based, long-TR, single pulse-acquire sequence. One important finding in this brain ³¹P MR study is the observation of UDPG (Figures 1, 3 and 5) at −9.72 ppm. Due to its low abundance, the presence of UDPG in vivo and in vitro has been rarely recognized or commented upon (8,15–17). Our UDPG assignment is based on an early in vivo ³¹P study by Malloy et al (13) in rat liver in which a high UDPG ³¹P signal was detected 2.15 ppm upfield from α-ATP, matching exactly the chemical shift observed here in the brain at 7T. UDPG ³¹P signal has also been identified in human liver at 7T (14), and as well as in cancer cells ex vivo using high-resolution ³¹P NMR (18). In retrospect, UDPG signal, though very weak, has appeared in some of the early ³¹P NMR spectra acquired from the human brain (1,8).

T₁ Difference among Mono-, Di- and Tri-phosphates

The inversion recovery experiments reveal an interesting, previously unrecognized observation, that is, the (apparent) T₁ value of the brain metabolites decreases in the general order of triphosphate < diphosphate < monophosphate (Table 1). This is reflected in the faster recovery of all signals in the upfield region than those in the downfield regions (Figures 3 and 4), and the brain ³¹P T₁ values are consistent with the results previously reported in the skeletal muscle by us (4) and also by Bogner et al (9) at 7T. A possible explanation is that there may be a strong ³¹P dipole-dipole interaction within the di- and tri-phosphate metabolites (including ATP, NAD and UDPG) through the chemical bond –P–O–P–, while such a relaxation mechanism is apparently absent in the monophosphate metabolites (including PE, PC, GPE, GPC, Pi and PCr). Indeed, not only the T₁ of tri- and di-phosphates is ~2 – 4 fold shorter than that of monophosphates (Table 1), but even within the triphosphate ATP, the central β-³¹P spin is featured with a shorter T₁ (1.1 sec) than the terminal α- and γ-³¹P spins (1.4 – 1.7 sec, Table 1) as one would expect from dipolar relaxation mechanism.

Furthermore, of the six major monophosphates, both PCr and Pi have a reduced T₁ in comparison to those remaining metabolites including PE, PC, GPE and GPC. This phenomenon could be reasonably explained by the effect of the chemical exchange with γ-ATP, since γ-ATP has a short T₁ and is actively involved in phosphoryl exchange with both Pi and PCr, but not with PME and PDE metabolites.

In addition to dipolar mechanism, chemical shift anisotropy (CSA) might also play a role in ³¹P T₁ relaxation. Of all P-metabolites reported here, the most likely candidate with a relative large CSA contribution is PCr, since the P atom of PCr is bonded to one N plus three O atoms, rather than to four O atoms; the latter has a more symmetric structure with a higher degree of shielding isotropy as featured in all monophosphates other than PCr. This may explain a previous observation of a shorter intrinsic T₁ at PCr (5.4 sec) than at Pi (7.3 sec, ref(4)). It might also be part of the reason for the short apparent T₁ observed here for PCr (3.4 sec, Table 1), but discerning such a mechanism is complicated by the mixing of MT effect. As for tri- di-phosphates, it is likely that dipolar mechanism, rather than CSA, is more responsible for the observed short T₁. This is because all phosphate groups in these poly-phoshates have a similar isotropic structure P(O)₄, which is distincly different from the less isotropic structure NP(O)₃ as in PCr. Furthermore, the P-to-P spatial distance is only ~ 2.2 Å in the di- and tri-phosphates, well within the expected radius of dipolar interaction (typically ~ 5 Å).

Two Pools of Pi

Several ³¹P studies have reported the observation of two Pi signals in various organs (12,19–21). It is generally understood that the phenomenon is due to presence of two pools of Pi different in pH (ΔpH ~0.4). In the brain, the major upfield peak is assigned to intracellular Pi, and the minor downfield peak assigned to extracellular Pi (2). However, in skeletal muscle and perfused hearts, the minor Pi signal is considered to be from mitochondrial Pi, characterized by a shorter T₁ than the major Pi signal (19,21). In sharp contrast, for brain, the T₁ of the major Pi signal is 40% shorter than that of the minor Pi signal, due to being involved in chemical exchange with γ-ATP of short T₁ (see below). We also noticed that, for the major Pi signal, the T₁ is quite similar between the brain and skeletal muscle (3.7 ± 0.5 sec versus 4.3 ± 0.4 sec (19)). The major difference is in the minor Pi signal; the T₁ is about 4-fold longer in the brain (5.8 ± 1.1 sec, Table 1) than in the skeletal muscle (1.4 ± 0.5 sec, (19)).

EBIT Measurement of ATP Synthesis

The EBIT data in Figures 5 and 6 clearly illustrate the modulation of the intracellular Pi signal by γ-ATP inversion as anticipated by chemical exchange effect. In contrast, the extracellular Pi signals remain invariant (Figure 6 top), indicative of lack of such chemical exchange effect. In terms of MT effect, the maximal Piⁱⁿ reduction by EBIT is 25% in the brain, which is reached 2.8 sec after the band inversion (Figure 5), equivalent to the MT effect generated by a continuous saturation at γ-ATP for ~ 2 sec using a frequency-selective pulse train by conventional ST (1,2). In comparison, a steady-state Pi reduction of ~ 30 – 38% was reported in the brain using ST, achieved after a prolonged irradiation at γ-ATP for duration 6 – 9 sec (1,2). Since the Pi reduction by EBIT occurs during the inversion delay period in absence of any B₁ pulsing, one therefore can expect consistency in instrumental performance without concerns of potential issues that may be present with using prolonged repetitive saturation pulses (3).

FIG. 6.

(a) Comparison of intra- and extracellular Pi signal changes in response to increase in the inversion recovery delay time t. All data are shown with identical y-scale, and taken from the EBIT spectra shown in Figure 5. (b) Normalized Z-magnetizations versus delay time t for extracellular Pi (top) and intracellular Pi and γ-ATP (bottom, averaged for 12 subjects). The solid curves in the plots represent the data fitting using a bi-exponential equation (Eq. [7]) for Pi and a mono-exponential equation for γ-ATP.

Despite the difference in MT mechanism between EBIT and ST, both approaches yield comparable results for the ATP synthesis in the resting human brain. The unidirectional rate constant derived from EBIT data (k_Pi→γ-ATP = 0.21 ± 0.05 sec⁻¹), is in reasonable agreement with the results reported by Lei et al (0.17 ± 0.04 sec⁻¹, (1)) and Du et al (0.18 ± 0.05 sec⁻¹, (2)) using ST. The forward flux for ATP synthesis, calculated by product of k_Pi→γ-ATP and Piⁱⁿ, is 9.9 ± 2.1 mmol/min/kg, comparable to the ST results of 12.1 ± 2.8 mmol/min/kg (calculated using a total Pi concentration of 1.30 mM, (1)) and 8.8 ± 1.9 mmol/min/kg (calculated using Piⁱⁿ of 0.90 mM, (2)).

From EBIT data, similar k_Pi→γ-ATP values are obtained by two different ways: 1) after fitting the bi-exponential Eq [7] to data and using the initial slope of the fit (Eq [6]); and 2) by fitting all of m_Pi(t) and m_γATP(t) data to Eq [4]. The later approach is also able to evaluate T_1,Pi in addition to k_Pi→γ-ATP. While the initial slope approach has long been applied to evaluate k_Pi→γ-ATP in IT data analysis for the reaction PCr → γ-ATP (22–25), measuring k_Pi→γ-ATP is technically much more challenging in vivo since the MT effect by Pi→γ-ATP pathway is one order of magnitude smaller than that by the competitive PCr→γ-ATP reaction (1,4,10). Using PCr as a “magnetization buffer”, EBIT approach amplifies the MT effect by Pi→γ-ATP pathway, allowing k_Pi→γ-ATP measured in the human brain as in the skeletal muscle, in spite of the quantitative difference in PCr/ATP ratio (1.5 for the brain vs 4.5 for the skeletal muscle).

There are potential technical limitations in this study that have to be considered. First, although our data acquisition scheme is able to clearly illustrate small change in MT effect at varying inversion delay times, the use of long TR and a large number of data points for evaluation of k_Pi→γ-ATP is time consuming. To be practical for clinical application, a time-saving data acquisition scheme is needed which could be achieved by use of short TR and less data points. While TR can be shortened to speed up data acquisition, but the trade-off is the complexity of data analysis, since the ³¹P magnetization at the end of TR can no longer be taken as a constant (M_z^o) due to incomplete and delay-dependent recovery as previously described (4). Second, while the high SNR ³¹P data acquired in this study permit a more reliable determination signal intensity, the use of a surface coil is known to be susceptible to inhomogeneity in B₁ and sensitive to regions only sufficiently close to the surface of the RF coil, mainly the occipital-parietal lobes. But it is possible that the energy metabolism may vary in different brain regions and depend on local cellular density of neuron and astrocytes, an issue worthy to address in future studies using volume coil in combination with suitable spatial encoding techniques. Third, the use of non-localized ³¹P acquisition strategy may be potentially prone to contamination from non-brain tissues, such as the muscle in the skull. However, it is very likely that the muscle contribution is negligible, judging from the fact that the observed brain ³¹P spectrum is distinctly different from the muscle ³¹P spectrum acquired under identical conditions. Specifically, in the brain the total PME signal is more than 2-fold larger than Pi, while in skeletal muscle the PME signal appeared to be negligibly small in comparison to Pi (11). Moreover, in the brain the GPE/GPC ratio is ~1 : 1.5, while in the skeletal muscle, this ratio is ~1 : 15 (11), a 10-fold difference. Finally, though the term ATP has been used in this work, strictly speaking, all forms of NTP, including ATP, CTP, GTP and UTP, which are indistinguishable by in vivo ³¹P MRS (1), contribute to the measured unidirectional exchange rate constant.

In summary, we have also demonstrated that the rate of ATP synthesis in the human brain can be measured at 7T using EBIT. This technique could become an alternative to the conventional ST approach in the study of brain energy metabolism in normal physiology and under pathological conditions. In addition to ATP synthesis, we also measured the brain concentration of all detectable phosphorous metabolites including UDPG, an essential precursor for glycogen synthesis. Furthermore, an analysis of the inversion recovery data indicates that the apparent T₁ of ³¹P signals follows the order of monophosphate > diphosphate > triphospate, implying that dipolar mechanism may play a significant role in the T₁ shortening of tri- and diphosphates in reference to monophosphates. With the enhanced spectral resolution, the ³¹P MRS at 7T is also able to distinguish extra- from intra-cellular Pi pools, yielding important information about intra- and extracellular pH. Early diagnosis and follow-up of neurodegenerative diseases are often hampered by the lack of reliable biomarkers; the quantitative ³¹P MRS in combination with ATP synthesis, pH and T₁ relaxation measurements may offer a promising tool to detect biochemical alternations in such cases.

FIG. 2.

Brain concentration of P-metabolites (a) and Mg²⁺ (b), and intra- and extracellular pH (c), averaged for n = 12 subjects. Data quantification was based on the ³¹P peak fitting shown in Figure 1c, with P-metabolites concentrations evaluated by the integral of ³¹P peaks, [Mg²⁺] by chemical shift difference between α- and β-ATP, and pH by the chemical shift difference between Pi and PCr.

Abbreviations used

IT

inversion transfer

ST

saturation transfer

NOE

nuclear Overhauser effect

MT

magnetization transfer

SAR

specific absorption ratio

CSA

chemical shift anisotropy

EBIT

exchange kinetics by band inversion transfer

Eq

Equation

Pi

inorganic phosphate

ATP

adenosine triphosphate

ADP

adenosine diphosphate

PCr

phosphocreatine

PME

phosphomonoesters

PDE

phosphodiesters

NAD

Nicotinamide ddenine dinucleotide

UDPG

Uridine diphosphoglucose