Creatine Kinase and ATP Synthase Reaction Rates in Human Frontal Lobe Measured by ³¹P Magnetization Transfer Spectroscopy at 4T

Abstract

The human frontal lobe is critical for cognitive function in the healthy brain. Many psychiatric disorders including schizophrenia and bipolar disorder are associated with apparent mitochondrial dysfunction and bioenergetic abnormalities in the frontal lobe. Therefore, measuring cerebral bioenergetics associated with creatine kinase (CK) and ATP synthase reactions could provide crucial information regarding the underlying molecular mechanisms associated with psychiatric disorders. In this study, the unidirectional forward chemical exchange metabolic fluxes of creatine kinase and ATP synthase reactions as well as reverse chemical exchange metabolic flux associated with ATP hydrolysis were determined at 4T by ³¹P magnetization transfer. The current experiments indicate that the kinetic network of PCr↔ATP↔Pi can be measured reliably in the human frontal lobe at 4T, which will enable detailed in vivo characterization of bioenergetic abnormalities in a variety of neuropsychiatric disorders.

INTRODUCTION

Schizophrenia and bipolar disorder are common and severe brain disorders with significant negative impact on the affected individual and on society. Although the exact etiology of these disorders is unknown, numerous in vivo studies have demonstrated that abnormalities in neural activity and energy metabolism exist in persons with schizophrenia and bipolar disorder (1–3). Using ¹H MRS, our group recently demonstrated that total creatine (tCr) concentrations are reduced in the anterior cingulate cortex and parieto-occipital cortex of patients with schizophrenia (4). Postmortem studies also have identified abnormalities in mitochondrial structure (5), dysfunctional oxidative phosphorylation (6), and altered mitochondria related gene expression(7) in schizophrenia and bipolar disorder. Energy production, which largely occurs in mitochondria, is critical for many metabolic and intracellular signaling pathways and for glutamate-glutamine cycling during neurotransmission. Thus, it has great bearing on brain function, including emotional, cognitive and sensorimotor processes. Therefore, quantifying energy metabolism rates in vivo may provide crucial information on the pathophysiology of schizophrenia and bipolar disorder and help identify novel targets for drug development.

Adenosine triphosphate (ATP), a high-energy phosphate (HEP) compound, is the universal “energy currency” in living cells (8). It is required to restore cell membrane ion gradients and for most signaling pathways, including intracellular and intercellular signaling. The majority of ATP is formed from adenosine diphosphate (ADP) and inorganic phosphate (Pi) in the mitochondria through oxidative phosphorylation catalyzed by the enzyme ATP synthase (ATP_syn) (9). This biochemical process is also coupled to the creatine kinase (CK) reaction linking to phosphocreatine (PCr), which acts as a HEP reservoir and carrier for transferring energy from the mitochondria to the sites of utilization in the cytosol, therefore maintaining a stable ATP level in living cells (10,11). These chemical exchange reactions of PCr↔ATP↔Pi play a fundamental role in cerebral bioenergetics and brain function. Cerebral ATP metabolic rates can be determined explicitly and noninvasively by the in vivo ³¹P magnetization transfer (MT) approach which directly reflects mitochondrial oxidative phosphorylation(12–15).

Although abnormal mitochondrial function in the frontal lobe appears to be associated with psychiatric diseases, CK and ATP_syn reaction rates have not previously been measured in this region in vivo. In the current study, we explored whether ATP synthesis rates catalyzed by CK and ATP_syn can be separately determined in the human frontal lobe at 4T scanner.

METHODS AND MATERIALS

Human subjects and MR imaging

Seventeen healthy human subjects without any psychiatric or substance use disorders (10 males and 7 females, 24.7±4.3 years old) were recruited for these studies. All subjects provided informed consent and the study procedure was approved by the McLean Hospital Institutional Review Board. The MR experiments were conducted using a 4T whole-body scanner interfaced with a Varian INOVA console. Brain anatomic imaging and shimming was conducted by a birdcage volume coil. After global shimming, a 6.5×6.5×5.0 cm³ voxel within the main sensitivity region of the 7-cm ³¹P surface coil applied over the frontal lobe was manually shimmed to further minimize magnetic field inhomogeneity (B₀).

In vivo ³¹P experiments

All the ³¹P signal was acquired from the sensitivity region of the surface coil except that signal from extra-cranial muscles was eliminated by outer-volume saturation (12). The ³¹P MT pulse sequence has been described in previous papers, with modification of NMR parameters for the current study. Briefly, a pulse train constructed with multiple hyperbolic Sech pulses with varied RF pulse amplitudes according to the BISTRO scheme (16) was used to saturate selectively the resonance peak of γ-ATP (one-site saturation for measurement of CK and ATP_syn reaction rate constants) and both resonances of PCr and Pi (two-site saturation for measurement of the chemical reaction rate constant of ATP hydrolysis to PCr and Pi) (17), in separate experiments. Note that the resonance peak at −2.5 ppm is composed of signal from all nucleoside triphosphates (including ATP, Guanosine-5′-triphosphate and others) and is sometimes referred to as the γ-NTP peak, however, the vast majority of the signal in this peak arises from ATP and therefore we will refer to it as the γ-ATP peak in this study.

Saturation time was controlled by varying the cycling number of the BISTRO pulse train. For the one-site saturation experiment, a hyperbolic Sech pulse (pulse width = 55 ms, frequency-selective bandwidth = 200 Hz) was applied and saturation time from 0 to 12.73 s was used. For the two-site saturation experiment, the RF pulse described above was modified to saturate PCr and Pi simultaneously. The saturation RF pulse with a single saturation frequency band was replaced by a hyperbolic Sech RF pulse (pulse width = 100 ms) with two identical saturation bands (140-Hz width for each) separated by 360-Hz between the band centers, which equals the chemical shift difference (in Hz) between the PCr and Pi resonance peaks at 4.0 Tesla. Saturation time was varied from 0 to 11.89 s.

The offset RF saturation effect caused by the imperfection of the frequency profile of BISTRO pulse trains was experimentally determined by measuring and comparing the magnetization of chemically non-exchangeable phosphate metabolites acquired in the presence of the BISTRO saturation pulse train at a selected saturation frequency and in the absence of the saturation pulse train. This measurement was performed with varied saturation times. The ratios of these two magnetizations provided the correction factors as a function of the saturation time. The correction factors were applied in the data processing of the saturated in vivo ³¹P spectra in this work.

A 200-μs hard RF pulse was used to excite the phosphate spins and its flip angle (nominal 90°) was adjusted to achieve an optimal NMR signal from the human frontal lobe. In vivo ³¹P spectra were acquired using the following acquisition parameters: 5000 Hz spectral width, 1024 complex data points, 32 and 24 signal averages for the one-site and two-site saturation experiment, respectively. All in vivo ³¹P spectra were acquired with a 14-s repetition time (TR) at the approximately fully relaxed condition. The specific absorption rate (SAR) was maintained below limit of the US Food and Drug Administration (FDA).

Due to the relatively long measurement time needed for conducting each in vivo ³¹P MT experiment, 17 subjects were divided into three sub-groups: (i) one-site saturation (n=9), (ii) two-site saturation (n=8, 6 of 8 overlapped with (i)), and (iii) the experiments for determining the offset RF saturation effect (n=6). The total measurement time for each experimental session was ≤ 70 min.

In order to delineate the ³¹P sensitivity region of the surface coil used in the current experiment setup, 1D profiles of Pi signal along three orthogonal dimensions were acquired from a phantom (14-cm diameter cylindrical bottle) with in-organic phosphate solution ([Pi] = 0.6 M and pH = 7.1).

Spectrum processing

The ³¹P spectra were analyzed in the time domain using the AMARES algorithm within jMRUI-software package (18). In this study, the initial 4 points of FID (the first 0.9 ms of the signal) were truncated to remove the broad component. Eight resonance peaks (PE: phosphoethanolamine; Pi: inorganic phosphate; GPC: glycerophosphocholine; GPE: glycerophosphoethanolamine; PCr: phosphocreatine; and three adenosine triphosphates: α-, β-, γ-ATP) were included in the basis set. The starting values of spectral resonances positions referred to PCr and line widths were initialed in the spectra fitting. The chemical shift of the PCr resonance peak was assigned to 0 ppm. The spectra base line and zero and first order phase were corrected.

Chemical exchange rates measured by progressive saturation transfer

The chemical exchange reactions among PCr, ATP and Pi were simply illustrated as the Eq. [1], where k_f and k_r are the pseudo first-order forward and reverse rate constant, respectively. For the experiments using progressive saturation on the γ-ATP resonance, the magnetizations of PCr or Pi are governed by the Eqs. [2a] and [2b](19).

$$\text{PCr}\underset{k_{\text{r}}^{\text{ATP}\to \text{PCr}}}{\overset{k_{\text{f}}^{\text{PCr}\to \text{ATP}}}{\rightleftarrows }}\gamma -\text{ATP}\underset{k_{\text{r}}^{\text{ATP}\to \text{Pi}}}{\overset{k_{\text{f}}^{\text{Pi}\to \text{ATP}}}{\leftrightarrows }}\text{Pi}$$ (1)

$${\text{M}}_{\text{s}}(\text{t})={\text{M}}_0\left[\left(\frac{\text{k}}{\alpha }\right)e^{-\alpha t}+\left(\frac{1}{\alpha {\text{T}}_1^{int}}\right)\right]$$ (2a)

$$\alpha =\frac{1}{T_1^{\mathit{app}}}=k+\frac{1}{T_1^{\mathit{int}}}$$ (2b)

Where M_s and M₀ are the magnetization of PCr, or Pi at saturation time (t) and Boltzmann thermal equilibrium condition, respectively; k is the pseudo first-order forward rate constant involving ATP production via the CK or ATP_syn reaction. $T_1^{\mathit{int}}$ and $T_1^{\mathit{app}}$ are the intrinsic and apparent spin-lattice relaxation times of PCr or Pi, respectively. Therefore k_f of CK and ATP_syn reactions, $T_1^{\mathit{int}}$ and $T_1^{\mathit{app}}$ of PCr and Pi can be determined by fitting the experimental data to a single exponential decay according to Eqs. [2a] and [2b] (14). Similarly, Eqs. [2a] and [2b] also can be used for the two-site (PCr and Pi) saturation experiment (17) except that M refers to magnetizations of γ-ATP. In this case, k is the pseudo first-order reverse rate constant ( $k_r^{\mathit{ATP}\to \mathit{PCr}+Pi}$) for ATP utilization via the ATP_ase and reverse CK reactions and

$$k_r^{\mathit{ATP}\to \mathit{PCr}+Pi}=k_r^{\mathit{ATP}\to \mathit{PCr}}+k_r^{\mathit{ATP}\to Pi}$$

$T_1^{\mathit{int}}$ and $T_1^{\mathit{app}}$ are the intrinsic and apparent spin-lattice relaxation time of γ-ATP, respectively.

Lastly, the chemical reaction flux (F) can be calculated by

$$F(\mu \mathit{mol}/g/\mathit{min})=60\times k\times [M]/1.1$$ (3)

Where, k is forward or reverse chemical exchange constant as described above (s⁻¹) and [M] is the metabolite concentration (μmol/ml) of PCr, Pi and ATP, respectively, according to the one-site or two-site saturation experiment. The chemical reaction fluxes were converted to μmol/g/min assuming brain tissue density of 1.1g/ml.

All data are reported as means +/− one standard deviation.

RESULTS

Phosphate metabolite concentrations and magnetization transfer results

Brain anatomic imaging (in the sagittal orientation) and experiment set-up are illustrated in Fig. 1. The ³¹P sensitivity region of the surface coil used in the current experiment was determined as ~160 cc, mainly covering the human frontal lobe. ³¹P spectra acquired from representative subjects for one-site (γ-ATP) and two-site (PCr and Pi) saturation are shown in Figs. 2 and 3, respectively. The bottom spectra in the Figs. 2 and 3 were acquired in the absence of saturation at approximately full relaxation, and were used to quantify the steady-state concentration of PCr, ATP and Pi. A spectral curve-fitting result is illustrated in Fig. 4, indicating an excellent spectral fitting quality. Based on the measured peak area ratios of PCr, γ-ATP and Pi and [ATP] = 3.0 μmol/ml taken from the literature (20, 21), PCr and Pi concentrations were determined and presented in Table 1. With the progressive increase in saturation time indicated in Fig. 5, magnetizations of PCr and Pi for one-site saturation and γ-ATP for two-site saturation gradually decreased and settled to a low steady-state at the longest saturation times. PCr and Pi magnetization decreased by about 63% and 33% at steady-state while that of γ-ATP decreased by about 37% (see Table 1).

Fig. 1.

T₂-weighted Brain anatomic imaging (in the sagittal orientation) with a 7-cm ³¹P surface coil. ³¹P surface coil position is illustrated by a 1-cm sphere with the saline solution at the center of ³¹P surface coil.

Fig. 2.

In vivo³¹P spectra with 10-Hz line broadening of a representative subject in the absence and presence of complete saturation (lowest spectrum and all higher ones, respectively) on the γ-ATP resonance (illustrated by the box) with varied saturation times (from 0 to 12.73 s, longer saturation times presented in progressively higher spectra). All resonance peaks are labeled in the lowest spectrum.

Fig. 3.

In vivo³¹P spectra with 10-Hz line broadening of a representative subject in the absence and presence of complete saturation (lowest spectrum and all higher ones, respectively) on the PCr and Pi resonances (illustrated by the boxes) with varied saturation times (0 to 11.89 s, longer saturation times presented in progressively higher spectra).

Fig. 4.

A ³¹P spectrum (bottom) acquired from the frontal lobe of a representative subject and its fitted spectrum (middle) and residue (upper). A line-broadening of 6 Hz was applied to display the spectra.

Table 1.

Results of Magnetization Saturation Transfer Measurements in the Human Frontal lobe

	Sat (γ-ATP)	Sat (PCr and Pi)
Subject #	N=9	N=8
Concentration (μmol/ml)	[PCr]=4.13±0.40; [Pi]=1.09±0.17	[PCr]=4.14±0.62; [Pi]=1.08±0.21
M*/M0 of PCr	0.37±0.07	—
M*/M0 of γ-ATP	—	0.63± 0.05
M*/M0 of Pi	0.67±0.11	—
Rate constant (s−1)	$k_f^{\mathit{PCr}\to \mathit{ATP}}=0.29\pm 0.04$	$k_r^{\mathit{ATP}\to \mathit{PCr}+Pi}=0.43\pm 0.07$
	$k_f^{Pi\to \mathit{ATP}}=0.16\pm 0.06$
T1 (s)	$T_1^{\mathit{app}}(\mathit{PCr})=2.20\pm 0.29$	$T_1^{\mathit{app}}(\gamma -\mathit{ATP})=0.82\pm 0.14$
	$T_1^{int}(\mathit{PCr})=5.80\pm 0.97$
	$T_1^{\mathit{app}}(Pi)=2.30\pm 0.58$	$T_1^{int}(\gamma -\mathit{ATP})=1.42\pm 0.21$
	$T_1^{int}(Pi)=3.56\pm 0.72$
Chemical exchange flux (μmol/g/min)	$F_f^{\mathit{PCr}\to \mathit{ATP}}=65.55\pm 12.11$	$F_r^{\mathit{ATP}\to \mathit{PCr}+Pi}=70.36\pm 11.45$
	$F_f^{Pi\to \mathit{ATP}}=8.88\pm 3.68$

Fig. 5.

Dependence of the averaged normalized resonance peaks integrals on the saturation time for one-site saturation (A, n=9) and two-site saturation (B, n=8), respectively. The peak integrals represent the normalized M_s/M₀ ratios with the correction of offset RF saturation effect. The apparent and intrinsic spin-lattice relaxation times of PCr and Pi, the forward rate constants ( $k_f^{\mathit{PCr}\to \mathit{ATP}}$ and $k_f^{Pi\to \mathit{ATP}}$), were determined by the regression analyses presented in (A). The apparent and intrinsic spin-lattice relaxation times of γ-ATP, the reverse rate constant ( $k_r^{\mathit{ATP}\to \mathit{PCr}+Pi}$), were determined by the least-square regression and presented in (B).

Chemical exchange rate constants and fluxes determined by magnetization transfer experiments

Figs. 5A shows the time dependency of the averaged and normalized resonance peak integrals of PCr and Pi on the saturation time during the progressive saturation experiment and their least-square regression curves according to Eqs. [2a] and [2b]. $T_1^{\text{int}}$ and $T_1^{\mathit{app}}$ of PCr and the forward exchange rate constant from PCr to ATP (i.e. $k_f^{\mathit{PCr}\to \mathit{ATP}}$) were determined as 2.20±0.29 s, 5.80±0.97 s and 0.29±0.04 s⁻¹, respectively. Similarly, $T_1^{\text{int}}$ and $T_1^{\mathit{app}}$ of Pi and the forward exchange rate constant from Pi to ATP (i.e., $k_f^{Pi\to \mathit{ATP}}$) were determined (see Table 1). With the measured concentrations of PCr and Pi, the unidirectional forward chemical exchange fluxes for CK and ATP_syn were calculated and all results are listed in Table 1.

Parallel to our γ-ATP saturation experiment, $k_r^{\mathit{ATP}\to \mathit{PCr}+Pi},T_1^{\text{int}}$ and $T_1^{\mathit{app}}$ of γ-ATP were determined by two-site saturation experiments (see Figs. 3 and 5B). Results including the reverse chemical exchange flux are listed in Table 1.

DISCUSSION

We have shown that it is feasible to calculate ATP synthesis and hydrolysis rates for the CK and ATP_syn/ATP_ase reactions in the human frontal lobe non-invasively at 4T. Our results are broadly consistent with previous work (20, 22–25) as described below, and we extend these findings to include a new brain area that has not been previously studied.

Previous work indicated that in vivo ³¹P MT provides a non-invasive approach for directly studying bioenergetics/mitochondrial function associated with brain activity changes (12–14, 24, 26). The major obstacle to this approach is the low signal-to-noise ratio (SNR) especially for ATP_syn reaction measurements because of the intrinsically low nuclear gyromagnetic ratio of ³¹P and the low concentration of Pi (~1mM). Therefore, to achieve sufficient SNR for ATP_syn reaction measurements in this preliminary study, a relatively large region covered by a 7-cm ³¹P surface coil was studied at 4T, where our group conducts patient-oriented MRS studies based on a private psychiatric hospital. Based on our results of 1D profiles of three orthogonal dimensions, the majority of ³¹P signal acquired by the 7-cm ³¹P coil come from the region of the frontal lobe including gray and white matter contributions. In this study we did not conduct an explicit comparison between 4T and our previous data collected at 7T. However, there are some interesting differences between 4T and 7T studies that deserve mention. First, the ³¹P signal from Pi in the intra-cellular and extra-cellular compartments and GPE/GPC (see Figs. 2 and 3) are resolved at 7T(14) while at 4T they were resolved only 20% of the time (3 subjects out of 17; see Figs. 3 and 4) despite very good shimming. In the current experiment, the line width of PCr is ~ 14 and 10 Hz without line broadening for the case of Figs. 2 and 3, respectively. Second, $T_1^{\text{int}}$ of PCr was 5.80 ± 0.97 s at 4T compared but was much shorter at 7T (4.89 ± 0.54 s) (14). An even shorter $T_1^{\text{int}}$ for PCr (3.83±0.57 s) was measured in the rat brain at 9.4T (13). The observed decreases in $T_1^{\text{int}}$ of PCr as magnetic field strength increases suggest that the ³¹P spin of PCr relaxation may be dominated by the chemical shift anisotropy (CSA) mechanism (27). By contrast, $T_1^{\text{int}}$ of Pi was calculated at 3.56 ± 0.72 s, 3.77 ± 0.53 s and 4.03 ± 0.64 s at 4T, 7T and 9.4T, respectively (13,14), with the significant difference between 4T and 7T on the one hand and 9.4T on the other (p=0.007, two-tailed t-test). This relationship demonstrates that the intrinsic T₁ of Pi increases with the magnetic field strength and the ³¹P spin of Pi relaxation may be dominated by dipolar interactions (27). $T_1^{\text{int}}$ of γ-ATP was calculated as 1.42 ± 0.21 s in the current 4T study, which is similar to observations at 7T and 9.4T (~1.35 s and ~1.24 s in human and rat brain, respectively) (13,14,28). Again, this observation indicates that the field dependence of phosphate T₁ is mainly due to two competing relaxation mechanisms, CSA and dipolar interactions. The intrinsic relaxation times of PCr, γ-ATP and Pi were compared in the current study, but similar trends have been observed in previous studies measuring T₁ via a conventional inversion recovery method and single exponential curve fitting (29,30). T₁ measurements for PCr, γ-ATP and Pi would be affected by the chemical exchange rate constants since the three metabolites are involved in chemical reactions with one another (31).

In addition to measuring the unidirectional forward chemical exchange fluxes of CK and ATP_syn reactions and simultaneously the intrinsic relaxation times of PCr and Pi in one set of experiments, we measured the reverse chemical exchange flux and the γ-ATP intrinsic relaxation time by saturating PCr and Pi simultaneously in another. The apparent total ATP hydrolysis rate constant, $k_r^{\mathit{ATP}\to \mathit{PCr}+Pi}$, were determined as 0.43 ± 0.07 s⁻¹. Using the unidirectional forward chemical exchange constants calculated from saturating γ-ATP and concentrations of PCr, ATP and Pi, as well as a chemical balance condition assumption, the reverse chemical exchange rate constant of ATP hydrolysis to PCr and Pi were deduced as 0.45 ± 0.06 s⁻¹, very similar to the value (0.43 ± 0.07 s⁻¹) directly determined by the two-site saturation experiment and not significant difference. Therefore, the current experiments measuring forward and reverse chemical reaction fluxes indicate that the chemical reactions among PCr, ATP and Pi are at an approximately balanced condition and the simple global chemical exchange model of PCr↔ATP↔Pi can be used to measure energy production and utilization in a combined ³¹P magnetization transfer strategy.

One methodological limitation of conventional ³¹P magnetization transfer approaches is the relatively long acquisition time (14, 19, 32–40), which prevents adequate temporal resolution for dynamic studies such as monitoring acute drug effects. For instance, the current measurements were performed with a 14-s repetition time at an approximately (although perhaps not completely) fully relaxed condition.

Some novel ³¹P MT approaches such as FAST and TRIST aim to measure the same chemical reaction fluxes (41, 42). Recently, we developed a ³¹P MT MRS approach (T₁^nom), aimed at rapidly mapping energy-ATP metabolic fluxes (28,43). Using this approach, only 2 spectra with optimized flip angles (β) and TR need be obtained to calculate chemical exchange fluxes of the CK and ATP_syn reactions as long as the intrinsic relaxation times of PCr, γ-ATP and Pi are known. The great advantage of this approach is that the acquisition time is significantly shorter, enabling improved SNR, increased spatial/temporal resolution, and higher reliability. Our current measurements lay the groundwork for rapid acquisition ³¹P MT MRS with the increased spatial and temporal resolution by establishing the necessary $T_1^{\text{int}}$ values for the relevant metabolites in the healthy human frontal lobe at 4T.

In summary, a ³¹P MT approach to measure ATP metabolic rates in the human frontal lobe was applied in the current study. Specifically, the unidirectional forward chemical exchange metabolic fluxes of the CK and ATP_syn reactions and reverse chemical exchange metabolic flux associated with ATP hydrolysis were determined at 4T. Thus, the current experiments indicate the kinetic network of PCr↔ATP↔Pi can be measured in human frontal lobe at 4T by using in vivo ³¹P MT. This work provides a foundation for future studies of bioenergetics/mitochondrial function in a variety of neuropsychiatric disorders at our 4T scanner.