OUP user menu

Phosphorus and proton magnetic resonance spectroscopy demonstrates mitochondrial dysfunction in early and advanced Parkinson's disease

Elke Hattingen , Jörg Magerkurth , Ulrich Pilatus , Anne Mozer , Carola Seifried , Helmuth Steinmetz , Friedhelm Zanella , Rüdiger Hilker
DOI: http://dx.doi.org/10.1093/brain/awp293 3285-3297 First published online: 1 December 2009

Summary

Mitochondrial dysfunction hypothetically contributes to neuronal degeneration in patients with Parkinson's disease. While several in vitro data exist, the measurement of cerebral mitochondrial dysfunction in living patients with Parkinson's disease is challenging. Anatomical magnetic resonance imaging combined with phosphorus and proton magnetic resonance spectroscopic imaging provides information about the functional integrity of mitochondria in specific brain areas. We measured partial volume corrected concentrations of low-energy metabolites and high-energy phosphates with sufficient resolution to focus on pathology related target areas in Parkinson's disease. Combined phosphorus and proton magnetic resonance spectroscopic imaging in the mesostriatal region was performed in 16 early and 13 advanced patients with Parkinson's disease and compared to 19 age-matched controls at 3 Tesla. In the putamen and midbrain of both Parkinson's disease groups, we found a bilateral reduction of high-energy phosphates such as adenosine triphophosphate and phosphocreatine as final acceptors of energy from mitochondrial oxidative phosphorylation. In contrast, low-energy metabolites such as adenosine diphophosphate and inorganic phosphate were within normal ranges. These results provide strong in vivo evidence that mitochondrial dysfunction of mesostriatal neurons is a central and persistent phenomenon in the pathogenesis cascade of Parkinson's disease which occurs early in the course of the disease.

  • Parkinson's disease
  • MR spectroscopy
  • 31P MRS
  • energy metabolism
  • mitochondrial dysfunction

Introduction

Parkinson's disease is the most common neurodegenerative movement disorder with a very high socioeconomic impact. The clinical symptoms of early stage Parkinson's disease mainly result from progressive degeneration of dopaminergic neurons in the substantia nigra and other monoaminergic cell groups in the brainstem (Braak et al., 2003) accompanied by increased microglial activation and intraneural accumulation of proteins called Lewy bodies (Forno, 1996).

Though Parkinson's disease is a sporadic condition of uncertain aetiology, there is some evidence that mitochondrial dysfunction considerably contributes to the pathogenesis of the disorder (Schapira, 2008). Since the brain is the organ system which is most reliant on mitochondrial energy supply, it is particularly susceptible to mitochondrial dysfunction (Wallace, 2005). Post-mortem studies of brain tissue reported diminished cortical and nigral complex I activity in the mitochondrial electron transport system in Parkinson's disease (Schapira, 1994; Keeney et al., 2006). Further studies observed mitochondrial complex I deficiency in blood cells and muscle tissue of patients with Parkinson's disease (Parker et al., 1989; Shoffner et al., 1991; Yoshino et al., 1992; Schapira et al., 1999). Recently, familial monogenetic parkinsonism has been shown to result from autosomal mutations in gene loci encoding for proteins which are involved in mitochondrial functioning (Polymeropoulos et al., 1997; Hyun et al., 2002; Palacino et al., 2004; Valente et al., 2004). Therefore, the question arises whether disturbances of energy metabolism can also be detected in living Parkinsonian humans. This is of special interest in the early stage of Parkinson's disease when possible disease modifying therapeutic interventions targeted at mitochondrial functioning are most promising. To provide answers to this question, phosphorus (31P) magnetic resonance spectroscopic imaging (MRSI) is a suitable tool since this technique allows direct monitoring of the energy metabolism of the brain (Ross and Bluml, 2001). Consequently, cerebral oxidative energy metabolism has been studied with 31P MRSI in mitochondrial cythopathies (Eleff et al., 1990; Montagna et al., 1992; Barbiroli et al., 1993), but also in Parkinson's disease (Montagna et al., 1993; Barbiroli et al., 1999; Hu et al., 2000; Rango et al., 2006). Most of these studies evaluated adenosine diphosphate (ADP) concentration as a widely accepted major regulator in mitochondrial activity (Chance et al., 1955, 1986) from the phosphocreatine/adenosine triphosphate (PCr/ATP) ratio, assuming constant values for ATP (3 mmol/l) (Eleff et al., 1990; Barbiroli et al., 1993, 1999; Montagna et al., 1992, 1993). In mitochondrial cythopathies, decreased PCr as well as increased inorganic phosphate (Pi) and ADP was reported, whereas in patients with advanced Parkinson's disease only increased inorganic phosphate/ATP ratios but no change in PCr/ATP ratio were found (Montagna et al., 1993; Barbiroli et al., 1999). Thus an altered phosphate metabolism rather than mitochondrial malfunction was suggested (Barbiroli et al., 1999).

Two previous studies investigated ATP in Parkinson's disease rather than the PCr/ATP ratio (Hu et al., 2000; Rango et al., 2006; for review see Henchcliffe et al., 2008). Rango and colleagues observed a decrease of occipital high-energy phosphates including ATP during the recovery period after visual stimulation in patients with Parkinson's disease but not in controls (Rango et al., 2006); while Hu et al., (2000) found decreased β-ATP ratios in a large midline voxel including the meso-diencephalon in rest condition.

In this study we obtained a more comprehensive picture of energy metabolism with a combined 31P and 1H MRSI study, which provides absolute ADP, ATP and PCr concentrations in two well-defined cohorts of patients with early and advanced Parkinson's disease. We used spectroscopic imaging at a 3 Tesla scanner with a double tuned 1H/31P-volume coil which allowed us to focus on the substantia nigra and on the putamen since these are anatomical key areas in the neurodegenerative process of the dopaminergic system in Parkinson's disease.

Materials and methods

Study subjects

The study was approved by the local ethics committee and informed consent was obtained from each participant prior to inclusion. We investigated 29 patients with Parkinson's disease and 19 age-matched healthy controls without neurological, psychiatric or systemic diseases. Exclusion criteria were alcohol and substance abuse, dementia, history of epilepsy or stroke, structural brain disease, brain surgery and severe arterial hypertension or diabetes mellitus. The diagnosis of Parkinson's disease was established according to the UK Brain Bank criteria (Hughes et al., 1992). Patients with Parkinson's disease were stratified according to the Hoehn and Yahr scale into early (Hoehn and Yahr I and II, n = 16) and advanced disease (Hoehn and Yahr III and IV, n = 13). Gender was almost matched for early disease and healthy controls while patients with advanced disease were mainly male. Demographic and clinical data of the Parkinson's disease cohorts are given in Table 1.

View this table:
Table 1

Demographic and clinical data of patients with Parkinson's disease and controls

Stage of diseaseControlsHoehn and Yahr I–IIHoehn and Yahr III–IV
Age (years)n191613
Mean and SD62.9 ± 7.962.4 ± 9.266.7 ± 7.6
Median and range66 (50–78)60 (45–77)68 (48–76)
Sexn9 males/10 females11 males/5 females12 males/1 female
Duration of disease (years)Mean and SD6.4 ± 4.310.5 ± 4.6
Median and range5.5 (2–17)10 (7–20)
Age at onset of Parkinson's disease (years)Mean and SD56 ± 957 ± 6
Type of parkinsonismAkinetic (a) Equivalence (e) Tremor (t)8 a/5 e/3 t6 a/6 e/1 t
Side of Parkinson's disease onsetn7 right/9 left7 right/6 left
Levodopa equivalence doseamg347 ± 244680 ± 371
Agonist monotherapyn52
Without therapyn20
Unified Parkinson disease rating scale III off stateb17 ± 630 ± 7
  • a One milligram of pergolide, 1 mg of lisuride, 1 mg of pramipexole, 2 mg of carbergoline, 5 mg of ropinirole, 10 mg of bromocriptine, 10 mg of apomorphine, 20 mg of dihydroergocriptine = 100 mg levodopa.

  • b Obtained at least 12 h after cessation of antiparkinsonian medication (practically defined off).

Magnetic resonance spectroscopy imaging

MRSI of the brain was performed on a 3 T whole body system (Magnetom Trio, Siemens Medical AG, Erlangen, Germany) with a double tuned 1H/31P volume head coil (Rapid Biomedical, Würzburg, Germany). Antiparkinsonian medication was stopped 12 h prior to MRSI. Two MRSI slices were recorded with 1H MRSI: an axial slice including the putamen and a coronal slice aligned on the dorsal line of the pontine and midbrain tegmentum. A weighted circular phase encoding scheme was used with repetition time 1500 ms, echo time 30 ms and 2 acquisitions. The volume of interest was selected by a combination of point resolved selective spectroscopy and outer volume suppression. For the coronal orientation a 28 × 28 matrix with 240 × 240 mm2 field of view and 10 mm slice thickness was acquired. The axial slice was recorded with a 16 × 16 matrix, field of view of 240 × 240 mm2 and 12 mm slice thickness. Before Fourier transformation, the matrix was extrapolated to 48 × 48 (coronal) or 32 × 32 (axial) resulting in a 5 × 5 mm2 or 7.5 × 7.5 mm2 in plane grid size. For 31P spectroscopy, a 3D MRSI sequence with wideband alternating phase low power technique for zero residual splitting proton decoupling was used with flip angle 60°, repetition time 2000 ms, echo time 2.3 ms and 10 acquisitions. An axial slab of 100 mm aligned to the axial 1H MRSI slice was recorded employing circular phase encoding with a weighted acquisition scheme on a 10 × 10 × 8 matrix, field of view of 300 × 300 × 200 mm3, nominal voxel size of 30 × 30 × 25 mm3. The matrix size was extrapolated to a 20 × 20 × 16 matrix resulting in a series of axial slices with 12.5 mm thickness and 15 × 15 mm2 in plane grid size. For taking into account partial volume effects originating from the cerebrospinal fluid, the fraction of grey and white matter contributing signal to each voxel was calculated as described in literature (Hetherington et al., 1996). In brief, segmented T1-weighted 3D images of the total brain were aligned to the MRSI slab followed by digital filtering to mimic the effect of the poor point spread function and resolution caused by the limited number of phase encoding steps (Gasparovic et al., 2006). The resulting parameter maps provided a value for the partial volume of grey and white matter for each MRSI voxel. The basic data for tissue segmentation were obtained from a sagittal oriented magnetization prepared rapid gradient echo sequence performed with 1.3 mm isotropic resolution in 4 min and 8 s scan time. In order to match these datasets with the MRSI scans, a short magnetization prepared rapid gradient echo was measured aligned to the MRSI slab. The shorter acquisition time of 1 min and 39 s was achieved by reduced resolution (2 mm isotropic). Oversampling of 38% in the head-to-foot direction was applied to avoid aliasing. The whole magnetic resonance examination took ∼45 min.

Data processing

Data were processed offline on a Linux workstation. Spatially transformed MRSI from the scanner console were resampled to the required grid resolution by Fourier transformation and zero-filling, and inverse Fourier transformation using MATLAB (The Mathworks, Inc., Natick, MA, USA). Segmentation of anatomical data was performed with the vbm5 extension (http://dbm.neuro.uni-jena.de/vbm/vbm5-for-spm5/) in Statistical Parametric Mapping-5 (http://www.fil.ion.ucl.ac.uk/spm/). The MRSI-aligned low quality 3D images were registered to the high quality anatomy by rotational and translational transformations (6 degrees of freedom) with Functional Magnetic Resonance Imaging of the Brain's (FMRIB) Linear Image Registration Tool, which performs linear inter- and intra-modal registration within the FMRIB Software Library (http://www.fmrib.ox.ac.uk/fsl/). The inverse registration matrix was then used to transform the segmented data onto the MRSI coordinate system in order to obtain the anatomical information for each MRSI voxel.

Down-sampling and filtering of anatomical information to MRSI resolution was performed with MATLAB according to the procedure previously described by Gasparovic et al. (2006). In addition to filtering, the dataset was multiplied before spatial Fourier transformation with a linear increasing phase in up to three directions causing a ‘grid shift’ which corrects for different offsets of the 31P slab and the 1H slice. The 31P spectra were analysed in the time domain with the jMRUI software tool (Version 3.0, available at http://www.mrui.uab.es) employing a non-linear least square fitting algorithm (AMARES) (Vanhamme et al., 1997). The time domain model function was composed of 14 exponentially decaying sinusoids in the frequency domain. Six of those, which had identical damping, were corresponding to peaks assigned PCr, phosphocholine, glycero-phosphocholine, phosphoethanolamine, glycero-phosphoethanolamine and Pi. PCr was adjusted to 0 ppm and constraints for the chemical shifts of the other signals except for Pi were applied as a fixed difference with regard to the position of PCr. ATP was represented by seven exponentially damped sinusoids, defining each multiplet by the respective number of peaks with identical damping and adequate amplitude ratios. The coupling constant was fixed at 18 Hz. One signal with a fixed chemical shift of 2.24 ppm and maximum line width of 50 Hz was used to account for potential macromolecule signals in the phosphodiester region.

The 1H MRSI spectra were fitted in the frequency domain by a linear combination of a set of model spectra including the main metabolites total choline (tCho), total creatine and total N-acetylaspartate using the commercially available software tool LCModel (Provencher, downloadable test version at http://s-provencher.com) (Provencher, 1993). Baseline correction was performed including macromolecules. Absolute concentrations were calculated by referring to an independent measurement with a spherical phantom of 15 cm diameter containing a solution of 20 mmol/l creatine and 20 mmol/l phosphate. For 1H MRSI, the in vivo sequence was used but repetition time was increased to 10 s. For 31P MRSI, 2D MRSI was performed instead of a 3D approach to avoid errors due to the poor point spread function. A full 16 × 16 matrix (field of view 240 × 240 mm2) was acquired with 90° flip angle and a repetition time of 40 s. The coil loading, which was obtained from the radiofrequency power required to achieve a 90° pulse for 1H was adjusted with sodium chloride to match the mean coil loading of the in vivo studies. For 1H MRSI, correction terms for longitudinal (T1) and transversal relaxations were applied as described by our group recently (Hattingen et al., 2008). The 31P data were corrected for the reduced flip angle (60°) and partial saturation using T1s of 2.4 s for PCr, 1.1 s for Pi, and 1.0 s for ATP resulting in intensity correction terms of 1.6, 1.27 and 1.24, respectively. The relaxation times were within the error range of published data measured at 1.5 (Buchli et al., 1994) and 4 T (Hetherington et al., 2001). The calibration of signal intensity was done with the phantom replacement method (Michaelis et al., 1993). Coil loading was taken into account as a covariate in the statistical analysis.

Anatomical parameters for the two target regions are summarized in Figs 1 and 2. Figure 1a shows the axial MRSI slice for 1H MRSI and 31P MRSI in a sagittal image. The respective grid resolution is shown for 1H in Fig. 1c and for 31P in Fig. 1d. The orange framed voxels from the putamen of each hemisphere were studied. Representative spectra are shown in Fig. 1b. The selection of corresponding 1H and 31P data for the midbrain are depicted in Fig. 2. The sagittal image (Fig. 2a) depicts the orientation of the coronal 1H and the axial 31P slices.

Figure 1

Anatomical details and MRSI slice orientation for putamen. (a) Sagittal image showing the orientation and feet-head extension of the axial 2D 1H MRSI (red borders). The respective slice of the 3D 31P MRSI data (green bar) was adjusted by grid shift to match position of the 1H MRSI slice (see Material and methods section). (b) Representative 31P spectra (upper row) and 1H spectra from the target areas. The red line marks the result of the fitting procedure. (c) Axial slice with grid representing the resolution of 31P MRSI data. An in-plane grid shift was applied to obtain anatomically matched midlines of the 31P MRSI slice and the 1H MRSI slice (b) The dorsal part of the putamen from each hemisphere was selected as target region (orange squared voxel). (d) Axial slice with grid representing the resolution of 1H MRSI data. The orange marked area matches the respective region in (c).

Figure 2

Anatomical details and MRSI slice orientation for midbrain. (a) Sagittal image showing the orientation and extensions of the 2D 1H MRSI (red bar) slice and an appropriate slice of the 3D 31P MRSI data (green bar). The 31P MRSI slice was adjusted to the midbrain by grid shift. (b) Representative 31P spectra (upper row) and 1H spectra from the target areas. The red line shows the result of the fitting procedure. (c) Axial slice with grid representing anatomy and resolution of 31P MRSI data. An in-plane grid shift was applied to position the centre in the cerebral peduncle as indicated by the broken white lines. The red line marks the intersection with the centre of the 1H MRSI slice. The orange squared voxels represent the target region. (d) Coronal 1H MRSI slice with the target voxels. Six voxels matching the target volume in (c) are framed in orange. The green line marks the centre of the intersection with the 31P slice (c).

All spectra from the selected voxels were inspected visually. Bad data were discarded according to the following criteria: (i) remaining signals in 31P residual data; (ii) either huge baseline modulation or extensive line broadening that obscures the discrimination between total choline and total creatine in 1H data and (iii) the Cramer-Rao lower bounds (%SD) of the non-linear fit exceeded 20% for both total creatine and total choline, in 1H data. The remaining data were used for further analysis.

In healthy control brains, lactate is not detectable by 1H MRSI with the data acquisition scheme described (Isobe et al., 2007). However, anaerobic glycolysis (i.e. an excess of glycolysis over oxidative phosphorylation) can lead to elevated and therefore detectable lactate levels (Prichard et al., 1991). The LCModel software provides a value for the lactate concentration and an estimation of its accuracy is given (%SD). Each complete MRSI data set was screened for lactate SD values <20%. The presence of a clear positive doublet signal at 1.3 ppm in the spectra was then verified. Voxels fulfilling these criteria were considered to have elevated lactate levels.

Calculated parameters

The pH was determined from the signal position of inorganic phosphate (δPi) and the concentration of magnesium (Mg)2+. The latter was estimated from the chemical shift difference between the β-ATP and the phosphocreatine signal (δβ) according to the formula given by Iotti et al. (1996, 2000) using the tool provided by the authors (http://www.cermiv.unibo.it/):

  1. pMg = 4.24 − log10 [(18.58 + δβ)0.42/(−15.74 − δβ)0.84]

  2. pH = 6.706 − 0.0307 [Mg] + log10 [(δPi − 3.245)/(5.778 − δPi)]

The following parameters were calculated from measured values (Petroff et al., 1985; Pietz et al., 2003; Iotti et al., 2005):

  1. ADP concentrations were calculated based on the equilibrium constant of the creatine kinase reaction KCK (Iotti et al., 2005): Embedded Image The value of KCK was estimated according to the calculation presented in the paper of Iotti et al. (2005).

  2. Unphosphorylated choline concentration [Cho] was calculated as difference Embedded Image

  3. [High-energy phosphates] = [ATP] + [PCr];

  4. [Low-energy metabolites] = [ADP] + ([total creatine] − [PCr]) + [Pi].

Metabolite concentration differences were calculated for each volume of interest in individual datasets. Since tissue concentrations were calculated, a correction for different partial volume effects in both MRSI modalities was performed. This was achieved by transforming the volume concentrations of the 1H detectable metabolites to values which are related to the respective tissue fraction of the 31P MRSI data using the ratio of 31P tissue fraction to 1H tissue fraction as a correction factor. Metabolite intensities were also corrected for the dependence on coil loading with linear regression.

Statistical methods

Statistical analysis was performed with STATISTICA (version 7.1, StatSoft, Tulsa, OK, USA). Mean values for each metabolite were calculated for every volume of interest separately, for both cerebral hemispheres. The latter were assigned ipsi- and contralateral to the clinically most affected side of the body in patients with Parkinson's disease. For healthy subjects, the data of both hemispheres were averaged.

In addition, a mean value for the tissue fraction of the respective area was calculated as the sum of grey- and white-matter fractions The statistical analysis was based on the General Linear Model using multivariate analysis of variance (ANOVA) with coil loading and tissue contents for both modalities as covariates. According to the a priori hypothesis of the study, the concentrations of metabolites related to cerebral energy metabolism (ATP, PCr, Pi, ADP) were analysed for significant difference with ANOVA with post hoc correction for multiple comparisons (Scheffé test) using target region and control/early (Hoehn and Yahr I/II) and advanced Parkinson's disease (Hoehn and Yahr III/IV) as categorical variables. Changes in pH were analysed with ANOVA without covariates, since these parameters depend on signal position only. Significant changes (P < 0.05) for each variable and localization were reported. For all other metabolites not directly related to energy metabolism significance levels were estimated by contrast analysis with univariate ANOVA. Side related concentration differences were analysed by multivariate ANOVA with repeated measurements. Metabolite differences were tested for significance using a post hoc Scheffé test.

Results

The amounts of metabolites divided by the volume of the region of interest are listed in Tables 2 and 3. The resulting concentrations (in mmol/l) still require correction for partial volume effects to depict tissue concentrations. The partial volume was determined from registered anatomical images (see Material and methods section) and is listed as tissue fractions for 31P and 1H voxels in separate rows in Table 2. As expected, there is a considerable amount of cerebrospinal fluid (25%) in the 31P target area of the midbrain. Due to the high resolution and careful visual alignment to the anatomy, partial volume effects for 1H MRSI are in the order of 10%. No significant difference was detected between control and the two stages of the disease. Significance was calculated taking into account partial volume effects and coil loading using the general linear model. Significant changes against the control group are printed in bold (multivariate post hoc Scheffé test). The other metabolites are shown in Table 3. The terms ipsilateral and contralateral define the hemisphere in reference to the clinically most affected body side. Significant changes (univariate ANOVA) are printed in bold. Adjusted means calculated by the General Linear Model approach are shown in Figs 3–5 for energy metabolites.

View this table:
Table 2

Concentrations of energy metabolites in different brain areas ipsilateral and contralateral to the clinically most affected body side at different stages of disease compared to healthy controls

PutamenMidbrain
ControlContralateralIpsilateralControlContralateralIpsilateral
Hoehn and YahrNAI–IIIII–IVI–IIIII–IVNAI–IIIII–IVI–IIIII–IV
n for 31P data191591591915111511
n for 1H and combined data19158158171410159
Tissue fraction for 31P0.99 (±0.095)0.94 (±0.086)0.092 (±0.079)0.96 (±0.096)0.92 (±0.091)0.80 (±0.087)0.75 (±0.079)0.74 (±0.106)0.76 (±0.079)0.74 (±0.108)
Tissue fraction for 1H0.98 (±0.057)0.98 (±0.079)1.00 (±0.077)0.99 (±0.062)1.02 (±0.052)0.92 (±0.032)0.90 (±0.046)0.87 (±0.063)0.91 (±0.039)0.89 (±0.058)
Phosphocreatine3.31 (±0.44)2.72 (±0.40)2.61 (±0.37)2.84 (±0.41)2.62 (±0.35)2.79 (±0.39)2.66 (±0.36)2.48 (±0.44)2.67 (±0.37)2.44 (±0.40)
ATP1.86 (±0.22)1.48 (±0.23)1.35 (±0.28)1.49 (±0.31)1.44 (±0.33)0.88 (±0.27)0.65 (±0.23)0.61 (±0.23)0.66 (±0.26)0.69 (±0.26)
ADP0.102 (±0.021)0.087 (±0.020)0.090 (±0.027)0.088 (±0.024)0.087 (±0.016)0.027 (±0.013)0.024 (±0.009)0.025 (±0.027)0.022 ±(0.009)0.026 (±0.018)
Inorganic phosphate1.16 (±0.22)1.02 (±0.22)1.02 (±0.29)1.12 (±0.33)1.04 (±0.21)0.97 (±0.21)0.87 (±0.14)0.92 (±0.20)0.86 (±0.24)0.92 (±0.20)
pH6.924 (±0.009)6.921 (±0.011)6.934 (±0.016)6.934 (±0.017)6.932 (±0.014)6.938 (±0.013)6.939 (±0.015)6.952 (±0.027)6.935 (±0.036)6.949 (±0.030)
HEP5.17 (±0.60)4.20 (±0.54)3.96 (±0.48)4.33 (±0.64)4.06 (±0.54)3.67 (±0.55)3.32 (±0.54)3.10 (±0.49)3.33 (±0.58)3.14 (±0.46)
LEM6.92 (±0.96)6.14 (±0.95)6.55 (±0.85)6.51 (±1.16)6.43 (±1.38)3.86 (±0.89)3.97 (±0.98)3.45 (±1.56)3.79 (±1.33)3.40 (±0.90)
  • NA = not applicable. Values indicate amount of metabolite (given in mmol) divided by the volume of the region of interest (given in l). The resulting concentrations (in mmol/l) still require correction for partial volume effects to depict tissue concentrations. The partial volume was determined from registered anatomical images (see Materials and methods section). Coil loading might also affect the MRSI detected concentrations. Means (in arbitrary units) were 168.8 (±12.8), 168.0 (±14.3) and 178.6 (±10.2) for control, mild and severe group, respectively. High energy metabolites (HEP) were determined as sum of ATP and PCr concentrations. ADP and low energy metabolites (LEM) were calculated by combining 31P and 1H data (see Materials and methods section). pH was determined from the chemical shift difference of inorganic phosphate and phosphocreatine, as described in Materials and methods section. Energy metabolism related metabolites with significant differences are printed in bold. Significance was determined using post hoc Scheffé Test (see Materials and methods section).

View this table:
Table 3

Concentrations of metabolites not involved in energy metabolism in different brain areas ipsilateral and contralateral to the clinically most affected body side at different stages of disease compared to healthy controls

PutamenMidbrain
ControlContralateralIpsilateralControlContralateralIpsilateral
Hoehn and YahrNAI–IIIII–IVI–IIIII–IVNAI–IIIII–IVI–IIIII–IV
Phosphocholine0.18 (±0.053)0.12 (±0.064)0.16 (±0.060)0.19 (±0.12)0.16 (±0.085)0.16 (±0.07)0.17 (±0.07)0.19 (±0.07)0.17 (±0.07)0.17 (±0.07)
Glycero-phosphocholine1.64 (±0.22)1.43 (±0.19)1.55 (±0.33)1.48 (±0.28)1.44 (±0.22)1.21 (±0.20)1.18 (±0.22)1.21 (±0.22)1.20 (±0.24)1.19 (±0.21)
Choline1.11 (±0.050)1.07 (±0.46)0.84 (±0.40)1.02 (±0.44)0.92 (±0.45)0.89 (±0.44)1.07 (±0.55)0.75 (±0.75)0.76 (±0.40)0.64 (±0.48)
Phosphoethanolamine1.04 (±0.15)0.83 (±0.20)0.82 (±0.17)0.88 (±0.20)0.44 (±0.14)0.92 (±0.17)0.87 (±0.17)0.74 (±0.14)0.86 (±0.19)0.74 (±0.15)
Glycero-phosphoethanolamine1.45 (±0.25)1.16 (±0.27)1.43 (±0.27)1.27 (±0.25)1.28 (±0.13)1.20 (±0.25)1.18 (±0.20)1.28 (±0.35)1.19 (±0.25)1.22 (±0.28)
Total choline2.89 (±0.43)2.74 (±0.49)2.75 (±0.24)2.78 (±0.50)2.78 (±0.47)2.63 (±0.43)2.85 (±0.60)2.58 (±0.52)2.57 (±0.54)2.36 (±0.44)
Total creatine8.98 (±0.98)8.19 (±1.26)8.39 (±0.74)8.51 (±1.48)8.41 (±1.69)6.48 (±0.91)6.91 (±1.05)5.71 (±1.45)6.75 (±1.54)5.68 (±1.13)
Total N-Acetyl-Aspartate13.44 (±1.50)13.00 (±1.70)12.37 (±0.89)13.92 (±1.94)12.38 (±1.61)12.69 (±1.45)13.16 (±2.15)11.52 (±1.75)12.91 (±2.37)11.06 (±2.31)
  • NA = not applicable. Values indicate amount of metabolite (given in mmol) divided by the volume of the region of interest (given in l). The resulting concentrations (in mmol/l) still require correction for partial volume effects listed in Table 2 to depict tissue concentrations. Coil loading might also affect the 1H MRSI detected concentrations. Means (in arbitrary units) were 168.8 (±12.8), 168.0 (±14.3) and 178.6 (±10.2) for control, mild and severe group, respectively. Metabolite differences were tested for significance by contrast analysis according to the General Linear Model (see Materials and methods section) Values with P < 0.05 are printed in bold. Choline was calculated by combining 31P and 1H data (see Materials and methods section).

Figure 3

Concentration of energy metabolism related metabolites and pH in the putamen. Left column shows data from contralateral, right column shows data from the ipsilateral hemisphere. Adjusted means, corrected for partial volume and coil loading (see Materials and methods section) are given in mmol/l. The bar represents 95% confidence interval according to univariate ANOVA. Asterisk marks significant difference compared to controls (na) (P < 0.05).

Figure 4

Concentration of energy metabolism related metabolites and pH in the midbrain. Left column shows data from contralateral, right column shows data from the ipsilateral hemisphere. Adjusted means, corrected for partial volume and coil loading (see Materials and methods section) are given in mmol/l. The bar represents 95% confidence interval according to univariate ANOVA. Asterisk marks significant difference compared to controls (na) (P < 0.05).

Figure 5

Concentration of high-energy phosphates ([ATP]+[PCr]) in the putamen. Left column shows data from contralateral, right column shows data from the ipsilateral hemisphere. Adjusted means, corrected for partial volume and coil loading (see Materials and methods section) are given in mmol/l. The bar represents 95% confidence interval according to univariate ANOVA. Asterisk marks significant difference compared to controls (na) (P < 0.05).

ATP was decreased versus controls in the putamen and in the midbrain of patients with early and advanced stage of Parkinson's disease. All mean differences in ATP reached statistical significance (P < 0.05) with the exception of the ipsilateral midbrain in advanced Parkinson's disease. PCr was significantly reduced in the putamen of both hemispheres in the early and advanced patient cohorts (P < 0.05) and showed a consistent similar trend in the ipsi- and contralateral midbrain. High-energy phosphates were significantly decreased in all contralateral areas for the early and advanced stage. For ipsilateral hemisphere, the decrease was only significant in the putamen. In contrast, ADP, Pi and low-energy metabolites concentration did not show a consistent trend in the two target regions with a slight decrease for ADP in the putamen and rather an increase in the midbrain but the differences were small and did not reach statistical significance. The pH was significantly increased in the contralateral putamen for the advanced stage of Parkinson's disease. Lactate was detected in two patients and one healthy subject, all other subjects did not show any sign of elevated lactate.

For other metabolites (Fig. 6) decreased concentrations of the membrane lipid precursors (phosphoethanolamine, PCho) and degradation products (glycero-phosphocholine, glycero-phosphoethanolamine) were found in early Parkinson's disease, but not in the advanced stage of the disease.

Figure 6

Concentration of membrane metabolism related metabolites in the putamen. Left column shows data from contralateral, right column shows data from the ipsilateral hemisphere. Adjusted means, corrected for partial volume and coil loading (see Materials and methods section) are given in millmoles per litre. The bar represents 95% confidence interval according to univariate ANOVA. na = controls; Cho = choline; PCho = phosphocholine; GPC = glycero-phosphocholine.

Discussion

This is the first quantitative 31P MRSI and 1H MRSI study which investigates brain energy metabolism measuring the concentration of total creatine, PCr, Pi, ATP and ADP in the midbrain and putamen of early and advanced patients with Parkinson's disease. The data demonstrate defects of mitochondrial oxidative phosphorylation with depletion of high-energy phosphates in the mesostriatal region occurring early in the course of the disorder. They are in line with the previous 31P study which investigated a large midline voxel including the basal ganglia and brainstem (Hu et al., 2000). This study measured the metabolite values either in terms of ratios to ATP or to the integrated signal intensity of the entire 31P spectrum. Since ATP contributes significantly to the total amount of phosphate, the decrease in ATP might simulate the increase of other metabolites. Hu et al. (2000) reported a significant decrease of ATP in the central voxel as well as a temporo-parietal rise of Pi, both measured as ratio to total 31P. The authors suggested a shift to anaerobic metabolism. In a more recent study, Rango et al. (2006) measured the concentration of high-energy phosphates in the visual cortex before, during and following visual stimulation. The data were referenced to a standard solution which was recorded together with the in vivo data to avoid the use of ATP as internal standard. They observed a decrease compared to controls, but only in the recovery period. At rest the ATP decrease may be compensated and latent in this area which is not a primary target region of pathology in Parkinson's disease.

The most important function of the mitochondria is to assure sufficient energy supply of the neuron, particularly by the regeneration of high energy phosphates via oxidative phosphorylation (Erecinka and Silver, 1989). Therefore, the significant depletion of high energy metabolites in the putamen (ATP and PCr) and in the midbrain (ATP) indicates mitochondrial dysfunction in the mesostriatal dopaminergic neurons in early and advanced Parkinson's disease. In the putamen and midbrain, we observed a characteristic pattern of MRSI changes with decreased high- but unchanged low-energy metabolites. The significant reduction of high-energy phosphates argues for an insufficient ATP production due to impaired oxidative phosphorylation in Parkinson's disease. The metabolism of ATP and PCr is tightly coupled via the enzyme creatine phosphokinase that generates ATP by the transfer of a phosphate group from PCr to ADP under conditions of increased energy demand or decreased ATP production through oxidative phosphorylation. PCr is a high-energy source for rapid retrieval of ATP and, therefore, a sensitive marker for mitochondrial functioning. The reduced putaminal PCr concentration in patients with Parkinson's disease might represent an insufficient compensatory mechanism to maintain adequate ATP levels and indicate a restricted capacity of the neuron to cope with further energy demands.

In contrast to ATP, the low-energy metabolites unphosphorylated creatine, ADP and Pi were unchanged in patients with Parkinson's disease compared to controls. This is consistent with previous findings in patients with Parkinson's disease and with progressive supranuclear palsy (Montagna et al., 1993; Stamelou et al., 2009). However, in some encephalomyopathies and in chronic hypoperfusion elevated ADP concentrations were found (Eleff et al., 1990; Montagna et al., 1992; Barbiroli et al., 1993; Lodi et al., 1994; Hattingen et al., 2009). ADP might be a regulator for temporary oxidative stress situations as it might appear under chronic hypoperfusion (Petroff et al., 1985) or relapsing phases of mitochondriopathies rather than for chronic dysfunction of oxidative phosphorylation. Also, there is some evidence that the most important factor in determining ATP synthesis is not the level of ADP but the oxygen concentration and the state of the membrane (Reynafarje and Ferreira, 2008). We also did not find increased lactate levels or alterations of intracellular pH in most of our patients suggesting a slowly progressive disease process which is not accompanied by ineffective anaerobic energy production in the long-term.

Beyond the characterization of a Parkinson's disease-related pattern of disturbed neuronal energy metabolism, our study design permitted access to the evolution of mitochondrial dysfunction during the course of Parkinson's disease. All significant reductions were present in both hemispheres, with the exception of high-energy phosphates in the midbrain and did not show differences between patients with early and advanced Parkinson's disease. These findings indicate that even patients with early Parkinson's disease (Hoehn and Yahr stage I/II), with clearly lateralized motor symptoms, exhibit a significant reduction of putamen high-energy phosphates in the less affected hemisphere with a less pronounced dopaminergic cell loss. Therefore, we conclude that mitochondrial dysfunction is a rather early occurring and subsequently persistent event in the pathophysiology of dopaminergic degeneration in Parkinson's disease. However, it has to be emphasized that the midbrain results have to be interpreted with caution in view of the limited spatial resolution of 31P MRSI and a considerable amount of signal spreading between the two sides. These potential partial volume effects might have obscured the rather mild PCr decrease in the midbrain in contrast to the more prominent and robust ATP decrease in this region. A PCr decrease may also result in an decrease of the total creatine, which is in turn measurable with 1H MRSI. O’Neill et al. (2002) reported a decrease in total creatine concentration in the midbrain of patients with Parkinson's disease. We could only find a significant decrease of total creatine in the putamen, contralateral to the most affected side, and a trend of total creatine decrease in the midbrain of patients in the advanced stage of Parkinson's disease. Unfortunately, many publications dealing with 1H MRSI in Parkinson's disease discussed metabolite concentration changes as intensity ratios to total creatine (Davie et al., 1995; Cruz et al., 1997; Federico et al., 1997; Tedeschi et al., 1997; Choe et al., 1998; Simoes-Ribeiro et al., 1998; Firbank et al., 2002).

It is worthwhile to note that we also found changes in membrane related metabolites obtained from 31P MRSI, namely a decreased concentration of membrane lipid precursors (phosphoethanolamine, PCho) and degradation products (glycero-phosphocholine, glycero-phosphoethanolamine) in the putamen of patients with early Parkinson's disease. This might indicate reduced membrane turnover rates due to the impaired energy metabolism.

Our MRSI in vivo findings corroborate the pathogenetic relevance of mitochondrial dysfunction in Parkinson's disease which was first supposed by Schapira et al. (1989), who discovered a mitochondrial complex I deficiency of the respiratory chain in the substantia nigra post-mortem (Schapira et al., 1989). Complex I deficiency as a systemic defect due to genetic or environmental causes is present in nearly 25% of patients with Parkinson's disease (Schapira, 2008). This deficiency has been shown to render neurons more vulnerable to apoptosis and excitotoxicity contributing to cellular dysfunction and neuronal death in Parkinson's disease (Beal, 1998; Green and Kroemer, 2004). Oxidative damage has been shown to impair proteasomal ubiquitination and degradation of proteins, leading to alpha-synuclein aggregation in Parkinson's disease in the form of characteristic Lewy bodies (Jenner, 2003). The ubiquitin proteosomal system degrades intracellular proteins by several ATP-dependent enzymes. Thus, decreased ATP production due to impaired oxidative phosphorylation causes a decreased catalytic function of the ubiquitin proteosomal system and subsequent accumulation of damaged proteins (Hoglinger et al., 2003). In turn, alpha-synuclein overexpression is associated with a mitochondrial load of this protein, lower cytosolic ATP levels and an increased susceptibility to cell death (Shavali et al., 2008). Mutations in the PARK6 gene locus, which are associated with familial autosomal recessive parkinsonism in humans, are associated with mitochondrial dysfunction, dopaminergic cell loss and reduced ATP levels in Drosophila melanogaster (Clark et al., 2006). Taken together, mitochondrial dysfunction seems to be a converging pathway of multifactorial mechanisms in the neurodegenerative cascade which is linked with oxidative stress, diminished proteosomal activity, neuroinflammation and excitotoxicity as major determinants of pathogenesis in Parkinson's disease.

The MRSI data underline the possible usefulness of mitochondria targeted therapies in Parkinson's disease. Similar to our findings, a recent study revealed decreased high-energy phosphates and unchanged low-energy metabolites in the basal ganglia of progressive supranuclear palsy patients, which was less severe after the oral administration of the antioxidant coenzyme Q10 acting as a cofactor of complex I of the mitochondrial respiratory chain (Stamelou et al., 2008). Despite controversial debate, treatment with coenzyme Q10 has been shown to slow functional decline in a previous pilot study on 80 Parkinson's disease patients (Shults et al., 2002) but did not display a symptomatic effect given as an adjunct in midstage patients (Storch et al., 2007). Moreover, it did not meet the pre-specified criteria for futility in another trial (Investigators, 2007). The combined administration of coenzyme Q10 and creatine showed additive neuroprotective effects against dopamine depletion in the striatum and loss of tyrosine hydroxylase neurons in the substantia nigra pars compacta following chronic subcutaneous administration of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine in a mouse model of Parkinson's disease (Yang et al., 2009). Therefore, a large clinical trial on the potential of coenzyme Q10 as a disease modifying agent in patients with Parkinson's disease is currently planned.

In conclusion, combined 1H/31P MR spectroscopy revealed a significant decrease of high energy phosphates in the mesostriatal region of patients with Parkinson's disease in vivo, which was similarly present in early and advanced stages of the disease. The data strongly support the hypothesis that mitochondrial dysfunction is involved early in the pathogenesis of Parkinson's disease and a major constituent of the pathogenetic cascade.

Funding

Bundesministerium für Bildung und Forschung (Brain Imaging Center Frankfurt); contract number: DLR 01GO0203 and Deutsche Forschungsgemeinschaft; contract number: ZA 233/1-1.

Footnotes

  • Abbreviations:
    Abbreviations
    ADP
    adenosine diphosphate
    ATP
    adenosine triphosphate
    MRSI
    magnetic resonance spectroscopic imaging
    PCr
    phosphocreatine
    Pi
    inorganic phosphate

References

View Abstract