OUP user menu

Cortical dysfunction in non-demented Parkinson's disease patients
A combined 31P-MRS and 18FDG-PET study

M. T. M. Hu, S. D. Taylor-Robinson, K. Ray Chaudhuri, J. D. Bell, C. Labbé, V. J. Cunningham, M. J. Koepp, A. Hammers, R. G. Morris, N. Turjanski, D. J. Brooks
DOI: http://dx.doi.org/10.1093/brain/123.2.340 340-352 First published online: 1 February 2000


Regional cerebral phosphorus-31 magnetic resonance spectroscopy (31P-MRS) was performed in 10 non- demented Parkinson's disease patients and nine age-matched control subjects. Five of the patients undergoing 31P-MRS and four additional Parkinson's disease patients had cerebral 2-[18F]fluoro-2-deoxy-D-glucose PET (18FDG-PET), the results of which were compared with those of eight age-matched control subjects. All Parkinson's disease patients underwent neuropsychological testing including performance and verbal subtests of the Wechsler Adult Intelligence Scale—Revised, Boston Naming Test, Controlled Oral Word Association test (FAS Test) and California Learning Test to exclude clinical dementia. 31P MR spectra from right and left temporo-parietal cortex, occipital cortex and a central voxel incorporating basal ganglia and brainstem were obtained. 31P MR peak area ratios of signals from phosphomonoesters (PMEs), inorganic phosphate (Pi), phosphodiesters (PDEs), α-ATP, γ-ATP and phosphocreatine (PCr) relative to β-ATP were measured. Relative percentage peak areas of PMEs, Pi, PDEs, PCr, and α-, β- and γ-ATP signals were also measured with respect to the total 31P-MRS signal. Significant bilateral increases in the Pi/β-ATP ratio were found in temporoparietal cortex (P = 0.002 right and P = 0.014 left cortex) for the non-demented Parkinson's disease patients compared with controls. In the right temporoparietal cortex, there was also a significant increase in the mean relative percentage Pi (P = 0.001). 18FDG-PET revealed absolute bilateral reductions in glucose metabolism after partial volume effect correction in posterior parietal and temporal cortical grey matter (P < 0.01 and P < 0.05, respectively) for the Parkinson's disease group, using both volume of interest analysis and statistical parametric mapping. There were significant correlations between right temporoparietal Pi/β-ATP ratios and estimated reductions in performance IQ (r = 0.96, P < 0.001). Left temporoparietal Pi/β-ATP ratios correlated with full scale IQ and verbal IQ (r = −0.82, P = 0.006, r = −0.86, P = 0.003, respectively). In summary, temporoparietal cortical hypometabolism was seen in non-demented Parkinson's disease patients with both 31P-MRS and 18FDG-PET, suggesting that both glycolytic and oxidative pathways are impaired. This dysfunction may reflect either the presence of primary cortical pathology or deafferentation of striato-cortical projections. 31P-MRS and 18FDG-PET may both provide useful predictors of future cognitive impairment in a subset of Parkinson's disease patients who go on to develop dementia.

  • Parkinson's disease
  • magnetic resonance spectroscopy
  • PET
  • glucose
  • dementia
  • ADP = adenosine diphosphate
  • ATP = adenosine triphosphate
  • Cho = choline
  • CSI = chemical shift imaging
  • CVLT = California Verbal Learning Test
  • DLB = dementia with Lewy bodies
  • 18FDG-PET = cerebral 2-[18F]fluoro-2-deoxy-d-glucose PET
  • 1H-MRS = proton magnetic resonance spectroscopy
  • H&Y = Hoehn and Yahr
  • MNI = Montreal Neurological Institute
  • 31P-MRS = phosphorus-31 magnetic resonance spectroscopy
  • MSA = multiple system atrophy
  • NAA = N-acetylaspartate
  • PCr = phosphocreatine
  • NART = National Adult Reading Test
  • Pi = inorganic phosphate
  • PDE = phosphodiesters
  • PME = phosphomonoesters
  • PVC = partial volume effect correction
  • rCMRGlc = regional cerebral metabolic rate for glucose
  • SPM = statistical parametric mapping
  • VOI = volume of interest
  • WAIS-R = Wechsler Adult Intelligence Scale—Revised


The prevalence of clinical dementia is significantly higher in Parkinson's disease than in the general population—estimates range from 10 to 40%—and the incidence increases with age (Brown and Marsden, 1984; Mayeux et al., 1992; Tison et al., 1995). Pathological processes responsible for cognitive impairment in Parkinson's disease patients are multifaceted and may include striatal and extrastriatal dopamine deficiency, loss of ascending noradrenergic, cholinergic and serotonergic projections to cortex, disruption of corticostriatal connections and the presence of co-existent Alzheimer's pathology, vascular infarcts or cortical Lewy bodies (Lees and Smith, 1983; Brown et al., 1984; Cooper et al., 1991). However, specific cognitive deficits are recognized in clinically non-demented Parkinson's disease patients even early in the course of the disease; impairment of visuospatial capacity, attentional control, planning, short-term memory and immediate recall have all been reported (Lees and Smith, 1983; Brown et al., 1984; Cooper et al., 1991; Sagar, 1996). Many of these deficits are thought to arise from frontal lobe dysfunction, but the neurochemical and neuropathological substrates remain unclear. Not all patients with Parkinson's disease and cognitive impairment go on to develop frank dementia. Risk factors for dementia at diagnosis may include being elderly, rapidly progressive motor disability, poor verbal fluency and poor ability on visuospatial subtests of the Wechsler Adult Intelligence Scale (WAIS) (Sagar, 1996; Mahieux et al., 1998). The challenge is to identify early risk factors for the development of cortical pathology so that neuroprotection can be developed and targeted at appropriate patients.

Phosphorus-31 magnetic resonance spectroscopy (31P-MRS) is a non-invasive technique that allows cerebral metabolism to be studied in vivo providing information on levels of cerebral phospholipids and high energy phosphates such as phosphocreatine (PCr) and adenosine triphosphate (ATP) (Coutts et al., 1989). This technique has been used to study Alzheimer's dementia (Pettegrew et al., 1988; Brown et al., 1989), but not Parkinson's disease with dementia. Cerebral 2-[18F]fluoro-2-deoxy-d-glucose PET (18FDG-PET) has demonstrated reduced glucose metabolism in frontal and temporoparietal association areas in Alzheimer's dementia (Goto et al., 1993; Vander Borght et al., 1997), Parkinson's disease with dementia (Peppard et al., 1992; Goto et al., 1993; Vander Borght et al., 1997) and dementia with Lewy bodies (DLB) (Minoshima et al., 1997). To our knowledge, 31P-MRS and PET techniques have not, as yet, been combined to study resting cortical function in Parkinson's disease patients in the absence of overt clinical dementia. Using proton magnetic resonance spectrocopy (1H-MRS), we reported previously temporoparietal cortical reductions in N-acetylaspartate/creatine (NAA/Cr) ratios in 17 non-demented Parkinson's disease patients which correlated with estimated cognitive decline (Hu et al., 1999). Following on from these findings, we have combined 31P-MRS and 18FDG-PET techniques with neuropsychological testing to investigate cortical function in a subgroup of the original 1H-MRS cohort of non-demented Parkinson's disease patients.


Patients and control subjects

Ten Parkinson's disease patients (mean age 62.2 ± 7.0 years) and nine age-matched healthy volunteers (mean age 56.8 ± 7.7 years) participated in the 31P-MRS study. Informed consent was obtained from each subject according to the declaration of Helsinki (1991) and the study was approved by the ethics committee of Imperial College School of Medicine/Hammersmith, Queen Charlotte's and Chelsea and Acton hospitals. All patients fulfilled the United Kingdom Parkinson's Disease Society Brain Bank criteria for clinical diagnosis of idiopathic Parkinson's disease (Hughes et al., 1992). The patients were rated Hoehn and Yahr (H&Y) stage I–IV in the `off' state following overnight withdrawal of medication (mean 2.45 ± 1.0), and had a mean clinical disease duration of 5.9 ± 3.8 years (range 2.2–15.0 years). All patients were receiving levodopa medication and four were taking additional dopamine agonist therapy including apomorphine, pergolide and cabergoline. Of the 10 patients who had 31P-MRS, seven also had 18FDG-PET within 3 months. Unfortunately, arterial blood sampling was not possible in two of the seven patients undergoing 18FDG-PET, hence these two patients were excluded from the final analysis. A further four Parkinson's disease patients who had not had 31P-MRS, matched for age, disease duration and severity to the above group, also underwent 18FDG-PET. Findings in these nine Parkinson's disease patients (mean age 64.4 ± 8.3 years, mean clinical disease duration 7.4 ± 5.3 years, mean H&P;Y 2.8 ± 1.0) were compared with those of eight separate healthy age-matched controls (mean age 61.0 ± 5.5 years). On the same day as MRS or PET, all Parkinson's disease patients underwent neuropsychological testing (in the `on' state while receiving anti-parkinsonian medication). All patients also completed a Geriatric Depression Score (Yesavage et al., 1983) and a Beck Depression Inventory (Beck, 1978). Patients experiencing symptoms suggestive of DLB, such as visual hallucinations, paranoid delusions or fluctuating confusion (McKeith et al., 1996), were excluded from the study.


Nine of the ten Parkinson's disease patients underwent a T1-weighted volumetric MRI [TR (retention time) 21 ms, TE (echo time) 6 ms, 128 contiguous 1.3 mm thick sagittal images] using a 1.0 T Picker HPQ scanner on the same day as MRS, and this was inspected visually for evidence of cortical and subcortical atrophy and used to correct the 18FDG-PET volume of interest (VOI) analysis for partial volume effects. For both MRI and MRS examinations, patients were scanned in the `off' phase following overnight withdrawal of medication. This avoided movement artefacts arising from dyskinesias. Two Parkinson's disease patients were severely disabled and required apomorphine injections before PET but were clinically `off' at the time of the scan.


31P-MRS of the brain was performed at the Magnetic Resonance unit of the Hammersmith Hospital using a Picker prototype spectroscopy system, based on a whole-body magnet (Oxford Magnet Technology, Oxford, UK), operating at 1.5 T. A 1H/31P enveloping birdcage coil was used, which comfortably encompassed the entire head. The proton signal was used for shimming and to acquire T1-weighted axial images in order to position a 4 cm transverse slice at the level of the basal ganglia (Taylor-Robinson et al., 1999). The three-dimensional chemical shift imaging (3D CSI) technique (TR 5000 ms) was used to obtain spectra from multiple contiguous voxels covering all the brain in the selected slice. CSI resolution was 40 × 40 × 40 mm, giving a voxel size of 64 cm3. Total examination time was ~60 min. Four VOIs covering right and left temporoparietal (predominantly temporal) cortex, occipital cortex and a central voxel including the basal ganglia (predominantly thalamus and globus pallidus) and brainstem (including substantia nigra) were analysed in each patient (see Fig.1). In the Parkinson's disease patients, it was not possible to sample right temporoparietal cortex in two patients, and left temporoparietal cortex in one patient due to voxel contamination by skull and scalp tissue. The more posterior temporoparietal VOIs were not analysed, as these frequently were contaminated by signal from skull and scalp tissue. It was also not possible to sample frontal lobe spectra consistently in this study owing to inhomogeneity arising from frontal eye fields and sinuses.

Fig. 1

Basal ganglia slice through a Parkinson's disease patient brain with a 31P 3D CSI grid applied showing volume of interest location. Volumes of interest are right temporoparietal cortex (A), central voxel incorporating thalamus and brainstem (B), left temporoparietal cortex (C) and occipital cortex (D).

The 31P MR spectra were convolved with a cosine filter in all three spatial directions, a 30 ms exponential filter and were phased manually. The baseline roll was removed using a knowledge-based algorithm (Saeed, 1995). A manual baseline correction was used, where necessary, and peak areas of phosphomonoester (PME), inorganic phosphate (Pi), phosphodiester (PDE), α-ATP, γ-ATP, β-ATP and PCr signals were measured using the NMR1® spectral processing program. Intracellular pH was calculated from the chemical shift of Pi relative to PCr in each voxel (Coutts et al., 1989). The data were fitted to inverse polynomial functions, since overlapping resonances in the 31P MR spectrum precluded the use of bell-shaped functions such as Lorenzian or Gaussian curve fitting. Inverse polynomial functions gave the best fit in terms of both curve fitting and reproducible results. The peak area of each 31P MR resonance was expressed as a percentage of the total 31P MR signal area, and as a ratio of the β-ATP signal area (Taylor-Robinson et al., 1999). All results were rechecked by a blinded observer (S.D.T-R.).


All patients and volunteers fasted overnight prior to 18FDG-PET. All anti-parkinsonian medications were discontinued 12 h prior to the PET scan, with the exception of one patient who was severely disabled and required apomorphine injections, but was clinically `off' at the time of the scan. 18FDG-PET was performed in 3D mode with the septa retracted using a Siemens 953b/CTI PET camera (CTI, Knoxville, Tenn., USA). The performance characteristics have been described previously (Spinks et al., 1992) and the axial resolution is 5 mm (Bailey, 1992). Studies were performed with the subject's eyes open in a dimly lit room with minimal auditory stimulation. Controls received 5.01 ± 0.41 mCi (185.4 ± 15.2 MBq) and subjects 5.20 ± 0.13 mCi (192.4 ± 4.8 MBq) of FDG by intravenous injection. The time course of plasma 18F radioactivity was determined by continuous on-line sampling of radial arterial blood (Ranicar et al., 1991). A dynamic 3D series, consisting of 21 frames over 60 min, was acquired over the whole brain volume. Voxel-by-voxel parametric images of regional cerebral metabolic rate for glucose (rCMRGlc) were produced from brain uptake and plasma input functions using spectral analysis (Cunningham and Jones, 1993).

18FDG-PET images were analysed using both VOI and statistical parametric mapping (SPM; Wellcome Department of Cognitive Neurology, London, UK) approaches. We used Analyze© version 7.0 and Matlab (Mathworks Inc., Sherborn, Mass., USA) to perform image manipulation and measurements. An anatomical template that divided the entire brain into 42 cortical and subcortical VOIs was defined on the Montreal Neurological Institute (MNI) brain template (SPM96). This VOI template was then transformed and co-aligned to the individual patient's MRI. The basal ganglia (caudate, lentiform and thalamic nuclei) and motor, orbitofrontal and dorsolateral prefrontal cortex were defined individually on the patient's MRI brain, and superimposed onto the individualized VOI template. The high resolution volume acquisition MRI scans were then segmented automatically into probability images of grey matter, white matter and CSF using a clustering, maximum likelihood `Mixture Model' algorithm (Hartigan, 1975). After co-registering the probability images and VOIs to the parametric rCMRGlc images (Woods et al., 1993), estimation of partial volume effects causing overspill of radioactivity into VOIs from surrounding tissue due to the lower resolution of PET was calculated as described and validated autoradiographically elsewhere (Koepp et al., 1997, 1998; Labbé et al., 1998). We report only the estimated grey matter activity contributions to the neocortical VOIs. At all stages, the template normalization and image co-registration were checked visually for anatomical accuracy using Analyze. SPM96 software (Friston et al., 1991) was used to transform and co-align 18FDG-PET add images to an 18FDG-PET add image template already in MNI space. These normalization parameters were then applied to computed rCMRGlc images, and smoothing of 10 × 10 × 10 mm was applied. Statistical analysis was performed on the normalized, smoothed rCMRGlc images using a single subject with replication of conditions technique, with a height threshold of P < 0.01 uncorrected and a P = 0.05 extent threshold.

Statistical analysis

Comparisons between Parkinson's disease patient and control groups were made using Student's sample t-test statistics. Correlations between 31P-MRS metabolite ratios and rCMRGlc values, and between 31P-MRS metabolite ratios/rCMRGlc values and neuropsychological testing were interrogated with the Pearson rank statistic. Because of the number of comparisons made, a P-value ≤0.01 was considered significant, while a P-value between 0.05 and 0.01 was reported as a trend.


31P MR spectroscopy

The cerebral 31P MR spectrum from a healthy volunteer contains at least seven resonances which can be attributed to PMEs, Pi, PDEs, PCr, γ-ATP, α-ATP and β-ATP (Bottomley et al., 1984). The PME and PDE peaks are multi-component; the α-ATP peak contains contributions from α-ADP and NADH and the γ-ATP peak contains contributions from β-ADP (Iles et al., 1985). Representative spectra from the temporoparietal cortex of a Parkinson's disease patient and a healthy volunteer are shown in Fig. 2.

Fig. 2

Representative 31P MR spectra from temporoparietal cortex of (A) a normal volunteer and (B) a Parkinson's disease patient.

In voxels localized to temporoparietal cortex, there were significant bilateral increases in the mean Pi/β-ATP ratio of the Parkinson's disease patients compared with healthy volunteers (P = 0.002 right, P = 0.014 left cortex). In right temporoparietal cortex, three Parkinson's disease patients had values that fell >2.5 SD above the normal mean, in left temporoparietal cortex two Parkinson's disease patients had values that fell 2.5 SD, and four >2 SD above the mean (Fig. 3). There was also a significant increase in the mean relative percentage of Pi signal in right temporoparietal cortex (P = 0.001), with a non-significant trend towards an increase in the left temporoparietal cortex (P = 0.079) (Table 1). In right temporoparietal cortex, one patient had a relative percentage Pi that fell 2.5 SD above the normal mean, and five patients had values that fell 2 SD above the normal mean. In left temporoparietal cortex, one patient had a relative percentage Pi that was 2 SD above the normal mean (Fig. 4). No other significant differences, including pH values, were observed in the remaining cortical voxels of the Parkinson's disease cohort.

View this table:
Table 1

Comparison of temporoparietal and central voxel metabolite ratios and percentage of total 31P signal in the patient and control populations

Metabolite ratioRegionParkinson's population mean (SD) (n = 10)Control population mean (SD) (n = 9)
T-P = temporoparietal cortex.
PME/β-ATPRight T-P 0.95 (0.23) 0.79 (0.34)
Left T-P 0.87 (0.34) 0.80 (0.34)
PDE/β-ATPRight T-P 4.71 (0.82) 4.46 (1.12)
Left T-P 5.27 (1.11) 4.49 (1.60)
Pi/β-ATPRight T-P 1.07 (0.24) 0.66 (0.22)**P = 0.002
Left T-P 0.96 (0.27) 0.62 (0.24)*P = 0.014
PCr/β-ATPRight T-P 1.63 (0.45) 1.40 (0.29)
Left T-P 1.82 (0.58) 1.80 (0.76)
PME/β-ATPCentral voxel 1.00 (0.32) 0.67 (0.18)*P = 0.013
PDE/β-ATPCentral voxel 6.38 (2.42) 4.07 (1.23)*P = 0.019
Pi/β-ATPCentral voxel 1.00 (0.32) 0.56 (0.14)**P = 0.002
PCr/β-ATPCentral voxel 1.84 (0.83) 1.15 (0.29)*P =0.039
Metabolite %
%PMERight T-P 7.93 (2.37) 6.88 (1.93)
Left T-P 6.43 (2.00) 5.23 (1.72)
%PDERight T-P39.64 (4.78)40.63 (3.32)
Left T-P39.48 (1.45)38.51 (4.22)
%PiRight T-P 8.73 (1.17) 5.87 (1.55)**P = 0.001
Left T-P 7.19 (1.30) 5.94 (1.51)
% PCrRight T-P13.36 (3.34)12.16 (1.78)
Left T-P13.46 (2.08)14.00 (3.1)
%β-ATPRight T-P 8.34 (1.09) 9.52 (2.00)
Left T-P 7.83 (1.80) 9.69 (3.81)
%PMECentral voxel 6.86 (1.11) 6.88 (1.64)
%PDECentral voxel43.08 (4.28)40.56 (4.65)
%PiCentral voxel 7.22 (2.63) 5.84 (1.25)
%PCrCentral voxel12.45 (2.69)11.18 (2.12)
%β-ATPCentral voxel 7.41 (2.10)10.66 (2.73)**P = 0.01
Fig. 3

Scatter plot of Pi/β-ATP ratios in temporoparietal cortex of normal volunteers (diamonds) and Parkinson's disease (PD) patients (squares).

Fig. 4

Scatter plot of relative percentage Pi values in temporoparietal cortex of normal volunteers (diamonds) and Parkinson's disease (PD) patients (squares).

In the central voxel incorporating thalamus, globus pallidus and brainstem (including substantia nigra), we found significant decreases in the mean relative percentage of β-ATP (P = 0.01), with associated increases in mean PME/β-ATP, PDE/β-ATP, PCr/β-ATP ratios (P < 0.05) and Pi/β-ATP ratios (P = 0.002) (Table 1). There were no significant correlations between central voxel 31P signals and age, disease duration or `off `motor UPDRS scores. No other significant differences in 31P resonances, ratios or pH were observed with this central voxel.


Results of 18FDG-PET are summarized in Table 2. Significant bilateral rCMRGlc reductions were found after partial volume effect correction (PVC) in the mean posterior temporal and posterior parietal grey matter (P < 0.05 and P < 0.01, respectively) of Parkinson's disease patients compared with controls. Three out of nine Parkinson's disease patients had either right or left posterior parietal rCMRGlc values (in two patients these reductions were bilateral) that fell >2 SD below the normal mean after PVC (Fig. 5). One patient also had left posterior temporal rCMRGlc values that fell >2 SD below the normal mean after PVC (Fig. 6). None of the patients had rCMRGlc values that fell 2.5 SD below the normal mean following PVC. The mean combined temporal and parietal lobe grey matter rCMRGlc of the Parkinson's disease group was also bilaterally significantly reduced when compared with normal volunteers (P < 0.02 right and left temporoparietal cortex). Other areas with reduced rCMRGlc were right lateral and left medial occipital lobe grey matter (P < 0.05), and mean right posterior cingulate PVC grey matter rCMRGlc (P < 0.05), while left posterior cingulate grey matter showed a non-significant trend towards reduction (P = 0.08). There were no significant differences in basal ganglia rCMRGlc between Parkinson's disease patients and volunteers. Overall, the mean PVC grey matter rCMRGlc for all the regions of interest sampled was reduced in the Parkinson's disease group compared with controls (P < 0.05).

View this table:
Table 2

Comparison of partial volume corrected absolute grey matter CMRGlc values (mg/min/100 ml) for regions of interest analysis in PD patients and healthy volunteers

RegionNormal right mean (SD) (n = 8)PD right mean (SD) (n = 9)Normal left mean (SD) (n = 8)PD left mean (SD) (n = 9)
DLPF = dorsolateral prefrontal cortex.
Caudate 7.81 (1.72) 6.93 (1.15) 8.02 (1.56) 7.11 (1.36)
Lentiform 7.03 (2.07) 6.72 (1.25) 6.89 (2.04) 6.88 (1.18)
Thalamus 6.22 (1.73) 5.86 (1.35) 6.51 (1.49) 6.05 (1.32)
Ant. temporal 6.66 (1.65) 5.69 (0.57) 6.39 (1.36) 5.62 (0.68)
Mid. temporal 8.15 (1.27) 6.95 (1.05) 8.19 (1.15) 8.19 (1.05)
Post. temporal 8.09 (1.24) 6.78 (0.79)*P = 0.018 8.08 (1.01) 7.1 (0.73)*P = 0.024
Ant. parietal10.01 (1.32) 9.22 (1.05) 9.98 (1.18) 9.00 (0.90)
Post. parietal10.77 (1.49) 8.88 (1.11)**P = 0.00911.39 (1.49) 9.36 (1.09)**P = 0.006
DLPF10.96 (1.79) 9.65 (1.32)11.12 (1.68) 9.67 (1.35)
Orbitofrontal7.68 (1.02) 6.54 (1.24) 7.35 (1.04) 6.40 (1.13)
Motor11.66 (1.48)10.89 (1.16)11.22 (1.52)10.64 (1.45)
Frontal10.83 (1.38) 9.94 (1.17)11.17 (1.31)10.16 (1.35)
Med. occipital8.97 (2.04) 7.44 (1.17) 9.66 (1.99) 7.74 (1.01)*P = 0.022
Lat. occipital 9.21 (2.00) 7.07 (1.63)*P = 0.028 9.03 (1.64) 7.71 (1.18)
Cerebellum 5.20 (1.33) 4.99 (0.77) 5.37 (1.49) 5.02 (0.89)
Ant. cingulate 7.18 (0.73) 6.65 (1.44) 5.71 (1.83) 5.84 (1.31)
Post. cingulate10.46 (1.37) 9.17 (0.91)*P = 0.03510.81 (1.62) 9.14 (2.03)
Insula 6.07 (0.89) 5.92 (0.63) 6.25 (0.54) 6.22 (0.84)
Global 9.39 (1.22) 8.16 (0.84)*P = 0.028 9.39 (1.22) 8.16 (0.84)*P = 0.028
Fig. 5

Scatter plot of rCMRGlc values in posterior parietal cortex following partial volume effect correction in normal volunteers (diamonds) and Parkinson's disease (PD) patients (squares).

Fig. 6

Scatter plot of rCMRGlc values in posterior temporal cortex following partial volume effect correction in normal volunteers (diamonds) and Parkinson's disease (PD) patients (squares).

Results of SPM are shown in Fig. 7. This also localized significant bilateral reductions in rCMRGlc in the posterior temporal and parietal cortex of Parkinson's disease patients compared with controls. There were no significant correlations between rCMRGlc values and 31P metabolite signals in the five patients scanned with both techniques.

Fig. 7

Absolute reductions in glucose metabolism in Parkinson's disease patients compared with normal volunteers using SPM.

Cognitive testing

Results of neuropsychological testing are summarized in Table 3. With the exception of one patient who was being treated for depression, none of the Parkinson's disease patients were clinically depressed. All the Parkinson's disease patients scored 24 or more on the Folstein Abbreviated Mini-mental Test (29.0 ± 0.9 mean ± SD), with no patients fitting clinical criteria for frank dementia. Overall, the mean verbal IQ of the patients was in the superior range, and mean performance and full scale IQ in the average and high average range, respectively. When the National Adult Reading Test (NART) estimates of full scale, verbal and performance IQ in each patient were compared with their measured IQ values in a paired t-test, performance IQ alone was significantly reduced compared with pre-morbid estimates (P < 0.001). Performance on subtests of the WAIS-R measuring visuospatial performance, such as Block Design and Object Assembly, was selectively impaired in the Parkinson's disease patients, unlike performance on Verbal subtests. The patients performed poorly on the FAS Test (Controlled Oral Word Association Test) (Spreen and Strauss, 1991), a test of executive functioning (P < 0.001 compared with normal controls) and also on memory functioning tested using the CVLT (P = 0.0025 compared with normal controls). When the reductions in verbal, performance and full scale IQ were estimated by subtracting calculated IQ values from pre-morbid scores on the NART, performance IQ was impaired more than full scale IQ, whereas the verbal IQ actually showed a higher level compared with pre-morbid estimates, possibly due to the NART underestimating intelligence in the more able patients.

View this table:
Table 3

Neuropsychology results of the Parkinson's disease patients undergoing 31P-MRS compared with age-matched control data

Neuropsychological testParkinson's disease (n = 10)Control data
Results are quoted as mean ± standard deviation. All WAIS-R subtest values are quoted as age-scaled scores. The difference in verbal, performance and full scale IQ was calculated as the NART pre-morbid estimate of IQ minus the actual IQ as calculated from the WAIS-R. **P < 0.001, *P < 0.003 significance on t-test comparison with normal controls. Control reference ranges are quoted as means for male and female, for 60–64 year age group, with education <12 years where applicable (Nelson, 1982; Van Gorp et al., 1986; Spreen and Strauss, 1991; Paolo et al., 1997).
Folstein AMT 29.0 ± 0.929.2 ± 1.3
Full scale IQ112.1 ± 10.9 100 ± 15
Verbal IQ121.9 ± 16.4 100 ± 15
Performance IQ 97.2 ± 9.2 100 ± 15
NART full scale IQ115.7 ± 9.1 100 ± 15
NART verbal IQ113.6 ± 8.1 100 ± 15
NART performance IQ114.5 ± 8.0 100 ± 15
Difference verbal IQ−9.2 ± 12.9
Difference performance IQ 17.3 ± 10.1
Difference full scale IQ 3.6 ± 9.0
Digit Span 10.2 ± 1.7510.0 ± 3.0
Vocabulary 12.2 ± 2.910.0 ± 3.0
Comprehension 15.3 ± 3.010.0 ± 3.0
Similarities 14.6 ± 2.510.0 ± 3.0
Block Design 9.9 ± 1.510.0 ± 3.0
Object Assembly 8.8 ± 2.810.0 ± 3.0
Boston Naming Test 58.5 ± 5.156.7 ± 3.0
FAS 29.0 ± 0.9**37.6 ± 10.9
CVLT 37.5 ± 11.4*47.9 ± 10.0
Beck Depression Inventory 9.3 ± 8.7
Geriatric Depression Scale 3.4 ± 2.7

Cognitive testing and 31P-MRS/18FDG-PET

Significant correlations were found between Pi/β-ATP ratios in right temporoparietal cortex and estimated reductions in performance IQ (r = 0.96, P < 0.001) (Fig. 8). In left temporoparietal cortex, significant correlations were found between Pi/β-ATP ratios and full scale IQ (r = –0.82, P = 0.006) and verbal IQ (r = −0.86, P = 0.003), with a trend to correlations between estimated reductions in verbal IQ (r = 0.70, P = 0.04). Left temporoparietal Pi/β-ATP ratios and individual subtests of the WAIS-R including Digit Span (r = –0.79, P = 0.01) and Similarities (r = –0.82, P = 0.007) were also correlated, with a trend toward correlations between the relative percentage of Pi signal from left temporoparietal cortex and NART pre-morbid estimates of verbal and full scale IQ (r = –0.68, P = 0.046 and r = –0.67, P = 0.047, respectively). No significant correlations were found between reductions in grey matter glucose utilization and any neuropsychological test.

Fig. 8

Correlation between right temporoparietal cortex Pi/β-ATP ratio and estimated reductions in performance IQ in the Parkinson's disease group.

Volumetric MRI imaging

This showed on inspection that three out of 10 Parkinson's disease patients had mild global cerebral atrophy when compared with 10 age-matched normal controls.


This is the first study to combine both 31P-MRS and 18FDG-PET in the study of non-demented Parkinson's disease patients and to correlate regional brain metabolism with neuropsychological testing. These two techniques have been combined previously to study Alzheimer's dementia patients (Murphy et al., 1993), where significant glucose hypometabolism in the presence of normal 31P-MRS was demonstrated. No correlations with dementia severity were found. In our study, both 31P-MRS and 18FDG-PET demonstrated significant bilateral abnormalities of temporoparietal cortical function in non-demented Parkinson's disease patients. 31P-MRS changes correlated significantly with neuropsychological measures of global cognitive decline and individual neuropsychological tests assessing language function.

31P-MRS is particularly relevant for in vivo clinical studies as resonances of central importance in oxidative metabolism such as PCr, ATP and Pi are readily observed, and intracellular pH can be measured, the chemical shift of Pi being pH dependent. However, the Pi signal represents only ~40% of the total intracellular levels of Pi (Iles et al., 1985) because the Pi bound to the mitochondrial inner membrane is not mobile and thus is MR invisible. Levels of MR-measurable metabolites such as Pi, ATP and ADP reflect oxidative phosphorylation status, and a rise in the Pi/β-ATP ratio reflects impairment of this bioenergetic pathway (Iles et al., 1985).

Our results suggest that several bioenergetic abnormalities are occurring in the temporoparietal cortex of non- demented Parkinson's disease patients. Compared with normal volunteers, we found significant increases in the Pi/β-ATP ratio and Pi resonance, when expressed as a percentage of the total MR-detectable phosphorus signal. No change was observed in the PCr, PDE, PME and α-, β- and γ-ATP resonances. Under normal circumstances, PCr acts as an energy reservoir in brain tissue, being a `high energy' phosphate source for ATP generation (Conn et al., 1987). In conditions such as hypoxia, the requirements for ATP cannot be met by oxidative phosphorylation in the mitochondria (Nioka et al., 1987). PCr is therefore utilized in these conditions and, as PCr falls, Pi increases, while reductions in ATP are minimized because of the buffering effect of creatine kinase. Our finding of an elevation in both Pi/β-ATP ratios and relative percentage Pi signal could reflect a shift in Pi from mitochondria to cytoplasm (and hence an increase in MR-detectable Pi signal) brought about by impaired mitochondrial oxidative phosphorylation and a resultant compensatory change in anaerobic glycolysis in the cytoplasm in order to maintain ATP/PCr levels. As the percentage levels of ATP and PCr were unaltered in our study, we conclude that the system must be compensated. This is in contrast to acute conditions such as ischaemic stroke where ATP and PCr levels are reduced (Levine et al., 1992). We previously have studied the same group of patients with 1H-MRS and demonstrated significant reductions in temporoparietal NAA/Cr ratios, while choline (Cho) to Cr ratios remained constant, providing circumstantial evidence that the reduced NAA/Cr ratio was occurring secondarily to reduced NAA synthesis (Hu et al., 1999). NAA is synthesized by the mitochondria (Bates et al., 1996) and is present almost exclusively within neurons and their processes within adult brain (Birken and Oldendorf, 1989; Vion-Dury et al., 1994). These findings, therefore, provide further evidence of impaired cortical mitochondrial function in this group of non-demented Parkinson's disease patients.

All our patients except one were right handed. In the left and, therefore, dominant hemisphere, significant correlations were found between Pi/β-ATP ratios and verbal and full scale IQ, as well as Similarities and Digit Span subtests of the WAIS-R, which are all measures of dominant temporoparietal function (Kolb and Whishaw, 1990). Significant correlations were found between right temporoparietal Pi/β-ATP ratios and estimated reductions in performance IQ. The non-dominant temporoparietal lobe is involved in tests of visuospatial function such as the Block Design and Object Assembly subtests of the WAIS-R (Kolb and Whishaw, 1990), which are reflected in the performance IQ score. This may explain the correlations found between estimated reductions in performance IQ and the right temporoparietal Pi/β-ATP ratio. As far as we are aware, this is the first study to correlate 31P-MRS with neuropsychological testing in non-demented Parkinson's disease patients.

In our study, 18FDG-PET demonstrated significant bilateral temporoparietal reductions in grey matter glucose metabolism which failed to correlate with measures of neuropsychological testing. Animal studies have shown that cerebral glucose metabolism primarily reflects afferent synaptic activity (Sokoloff, 1977; Auker et al., 1983), hence the reductions seen may reflect dysfunction of afferent synaptic or interneuronal connections in the temporoparietal cortex in Parkinson's disease. Corticostriatal connections between the parietal and temporal cortex, and ipsilateral striatum may have a role in the preparation and kinematic coding of movement (see Hu et al., 1999 for references). Disruption of these circuits in Parkinson's disease with reductions in cortical inputs may account for some of the temporoparietal metabolic abnormalities we have observed. Previous 18FDG-PET studies in demented Parkinson's disease patients have shown global reductions in frontal and temporoparietal glucose metabolism similar to that seen in Alzheimer's dementia patients (Peppard et al., 1992; Goto et al., 1993; Vander Borght et al., 1997). The magnitude of the reductions in rCMRGlc reported in demented Parkinson's disease patients ranged from 28 to 40% of normal subject values, in comparison with the smaller reductions in temporoparietal and occipital rCMRGlc (13–23%) we found in non-demented Parkinson's disease patients. With progression of disease and development of dementia, rCMRGlc decreased particularly in parietal and occipital cortex (Piert et al., 1996). Neuropathological studies have shown that additional occipital hypometabolism may distinguish patients with DLB from patients with Alzheimer's pathology (Minoshima et al., 1997). It is interesting to note that as well as temporoparietal grey matter rCMRGlc reductions, we also found significant reductions in lateral and medial occipital, as well as posterior cingulate grey matter rCMRGlc metabolism in the Parkinson's disease patients. In agreement with our results, previous 18FDG and oxygen-15 PET studies have demonstrated absolute reductions in temporoparietal glucose and oxidative metabolism in non-demented and demented Parkinson's disease patients (Lenzi et al., 1979; Piert et al., 1996). Eidelberg and colleagues have shown patterns of relative glucose hypometabolism in lateral frontal, paracentral, inferior parietal and parieto-occipital areas of non-demented Parkinson's disease patients (Eidelberg et al., 1990, 1994). In this study, we applied a novel MRI-based technique to calculate partial volume effect-corrected absolute values for grey matter rCMRGlc, results of which are broadly in agreement with the above studies.

The areas of temporoparietal rCMRGlc reduction appear to be more posterior than the corresponding temporoparietal cortical areas sampled with 31P-MRS. However, we were unable to sample the most posterior temporoparietal cortex with 31P-MRS due to technical limitations. Furthermore, we were unable to demonstrate frontal lobe dysfunction in Parkinson's disease with the 31P-MRS technique as we were prevented from sampling frontal cortex due to technical limitations (see Methods). 18FDG-PET showed a trend towards reduced resting rCMRGlc in the frontal regions which did not reach significance in our study. This does not exclude the presence of frontal dysfunction. Playford and colleagues have shown failure of frontal activation in Parkinson's disease although resting regional cerebral blood flow was normal (Playford et al., 1992).

The significant correlations demonstrated between 31P-MRS measures and neuropsychological testing in the absence of parallel correlations between 18FDG-PET and neuropsychological measures is intriguing. One explanation might be differences in variance across the 18FDG-PET and 31P MR data set, although this does not appear to be the case, as both rCMRGlc values and 31P peak ratios and percentages have similar distributions about the mean. A second possibility may be that different Parkinson's disease and control populations were used for the 18FDG-PET and 31P-MRS studies. Hence, although the populations were closely matched for age and disease severity, the variance may have contributed to the difference in results. A final explanation may be that deficits in oxidative phosphorylation are more pertinent to cognitive function than deficits in glycolysis.

Our 31P-MRS findings for the central voxel, incorporating predominantly thalamus, globus pallidus, lateral ventricles, brainstem (including substantia nigra) and adjacent white matter, are more difficult to interpret. We found significant decreases in the relative percentage of β-ATP and associated increases in PME, PDE, PCr and Pi/β-ATP ratios. In the adult brain, ATP levels begin to fall when the PCr pool is reduced by 50% (Tsuji et al., 1995), reflecting decompensation of the oxidative phosphorylation system. Much controversy exists concerning the contribution of mitochondrial dysfunction to the pathogenesis of Parkinson's disease. Several post-mortem studies have reported moderate reductions in mitochondrial complex I respiratory chain activity in the substantia nigra of Parkinson's disease patients compared with age-matched controls (Schapira et al., 1992; DiMauro, 1993). Interpretation of the results found in the central voxel are limited by its 64 cm3 size; however, the substantia nigra and its projections have a high density of mitochondria, and one intriguing possibility for these findings is that they reflect impaired mitochondrial function within the substantia nigra of Parkinson's disease patients. However, the nigral contribution to the total 31P-MRS signal will probably be small due to its relatively small size. Dopamine and its metabolites have been shown to produce either no inhibition or very weak inhibition of mitochondrial complex I activity and NADH-linked mitochondrial respiration (Morikawa et al., 1996; McLaughlin et al., 1998), suggesting that dopaminergic therapy in the Parkinson's disease group alone could not account for the abnormalities found. Alternatively, impaired mitochondrial function in the thalamic nuclei due to excessive inhibitory input from the basal ganglia output nuclei (substantia nigra pars reticulata and internal globus pallidus) postulated to occur in Parkinson's disease (DeLong, 1990) might also explain these results.

We were not able to perform absolute quantification of 31P-MRS metabolites in this study as this requires further sequences and would have prolonged the examination time unacceptably for the patients. However, by expressing results as a percentage of the total 31P signal which remains unchanged (Hope et al., 1984), an internal reference is, in effect, used (Taylor-Robinson et al., 1999). It seems unlikely that our results can be explained by an increased degree of atrophy in the Parkinson's disease group for several reasons. (i) Volumetric fine slice MRIs performed in nine of the 10 Parkinson's disease patients showed no significant differences in cerebral atrophy when inspected visually and compared with eight age-matched controls. (ii) The concentration of Pi and β-ATP in CSF is an order of magnitude smaller than the concentration in brain, and relatively small differences in the proportion of CSF in the MRS volume should have little, if any, effect on ratios or percentages (Siegel et al., 1989). (iii) The concentration of β-ATP and Pi is similar in both grey and white matter (Buchli et al., 1994), hence it is unlikely that the differences reported here are due to different quantities of grey and white matter in the VOI. (iv) Partial volume-corrected rCMRGlc values were reduced in the temporoparietal cortex of the Parkinson's disease group compared with normal subjects.

Further larger scale, longitudinal studies using combined 31P-MRS and 18FDG-PET studies to investigate cortical function in Parkinson's disease should address whether the changes we have found in non-demented patients with Parkinson's disease are useful predictors for the development of future dementia, or whether these changes are a common finding in Parkinson's disease uncomplicated by dementia. If the former is the case, these techniques may be useful in targeting neuroprotective strategies at appropriate patients.


We thank all the radiographers at the MRC Cyclotron Unit, Hammersmith Hospital for their help and support, and Dr Eraldo Paulesu, Neurologist and Neuropsychologist, PET Centre, San Raffaele Hospital, Milan for his generous provision of the 18FDG-PET template. M.T.M.H. is supported by an Action Research Training Fellowship, Action Research, UK. The MRC and Picker International supported work in the Robert Steiner MR Unit.


View Abstract