Conflict of Interest: The authors have no conflict of interest to report.
Usefulness of Proton and Phosphorus MR Spectroscopic Imaging for Early Diagnosis of Parkinson's Disease
Article first published online: 10 DEC 2013
Copyright © 2013 by the American Society of Neuroimaging
Journal of Neuroimaging
Volume 25, Issue 1, pages 105–110, January/February 2015
How to Cite
Weiduschat, N., Mao, X., Beal, M. F., Nirenberg, M. J., Shungu, D. C. and Henchcliffe, C. (2015), Usefulness of Proton and Phosphorus MR Spectroscopic Imaging for Early Diagnosis of Parkinson's Disease. Journal of Neuroimaging, 25: 105–110. doi: 10.1111/jon.12074
- Issue published online: 8 JAN 2015
- Article first published online: 10 DEC 2013
- Manuscript Accepted: 29 SEP 2013
- Manuscript Revised: 16 AUG 2013
- Manuscript Received: 26 MAY 2013
- Dana Foundation
- Magnetic resonance spectroscopy;
- Parkinson's disease;
- mitochondrial dysfunction;
BACKGROUND AND PURPOSE
Cerebral mitochondrial dysfunction has been observed in Parkinson's disease (PD). If mitochondrial dysfunction is an early event contributing to PD development, then noninvasive techniques that detect disturbed energy metabolism in vivo might be useful tools for early diagnosis and treatment monitoring. In the present study, we tested the hypothesis that proton (1H) and phosphorus (31P) magnetic resonance spectroscopy (MRS) measures of brain metabolites are able to differentiate between individuals with early PD and healthy volunteers (HVs).
During this cross-sectional study including 20 subjects with early PD and 15 age-matched HV, ventricular lactate (anaerobic glycolysis); and regional levels of N-acetylaspartate (neuronal integrity); choline (membrane turnover); creatine (energy metabolism); ATP and other phosphate-containing compounds (oxidative phosphorylation) were determined using brain 1H and 31P MRS.
No metabolic abnormalities were detectable in early-stage PD patients. Metabolite concentrations were not related to age, disease duration, or Unified Parkinson's Disease Rating Scale motor scores.
In early PD, neither 1H nor 31P MRS were able to detect metabolic abnormalities, a finding that is in contrast to published data in more advanced PD cohorts. MRS under dynamic conditions might uncover latent energy deficits in early PD, thus warranting future study.
Background and Purpose
Parkinson's disease (PD) is a common neurodegenerative disorder characterized by degeneration of dopaminergic neurons in the substantia nigra and more widespread changes within the central and peripheral nervous system including aggregation of α-synuclein in the form of Lewy bodies and Lewy neurites. There is considerable evidence for mitochondrial dysfunction in PD, and diminished mitochondrial respiratory complex I activity has been reported in brain, skeletal muscle, and platelets.[2-4] Recent advances in PD genetics support this by linking the products of specific PD-associated genes such as parkin, α-synuclein, PTEN-induced putative kinase 1 (PINK1), DJ-1, and leucine-rich repeat kinase 2 (LRRK2) with mitochondrial function and oxidative stress.[5, 6]
If mitochondrial dysfunction is an early event contributing to or inducing PD, then techniques that enable detection of a disturbed energy metabolism in vivo would be a vital tool for early, even premotor, diagnosis, especially in those individuals with atypical clinical presentation. In clinical trials, these objective and noninvasive biomarkers might support treatment monitoring. Moreover, since PD is heterogeneous, defining endophenotypes may ultimately aid in improved selection of clinical trial participants, and in personalizing treatment, all the more important as mitochondrial dysfunction and oxidative stress have been identified as “druggable targets.”[7, 8]
It has been suggested that proton (1H) and phosphorus (31P) magnetic resonance spectroscopy (MRS), noninvasive techniques to measure levels of specific hydrogen- and phosphorus-containing compounds in vivo, are appropriate techniques to measure neuronal integrity and brain energy metabolism in PD.[9, 10] Metabolites quantifiable by 1H MRS include (a) N-acetylaspartate (NAA) as a marker of mitochondrial and neuronal integrity, (b) lactate indicating an upregulated anaerobic metabolism, (c) total creatine (tCr) as a marker of tissue energetics, and (d) total choline (tCho) reflecting membrane turnover and glial proliferation. Phosphorus MRS allows in vivo evaluation of compounds directly related to the energy metabolism of the brain including the low-energy metabolite free phosphate (Pi) and the high-energy phosphates (HEPs) adenosine triphosphate (ATP) and phosphocreatine (PCr). In PD, there are reports of reduced NAA and reduced HEP or HEP/Pi ratio,[10, 12] as well as elevated lactate, choline, and Pi,[12-14] while other studies failed to detect these changes.[15-17] Others have detected changes in HEP/Pi ratio only under conditions of increased neuronal activity. Thus, it remains unclear whether these metabolites are sensitive biomarkers, especially in very early stages of PD and in unmedicated individuals. In the present study, we tested the hypothesis that 1H and 31P MRS measures of brain metabolites at rest are able to differentiate between individuals with early PD and control subjects comprising age-matched healthy volunteers (HVs).
Twenty patients fulfilling the United Kingdom Parkinson's Disease Society Brain Bank criteria for clinical diagnosis of idiopathic PD with a modified Hoehn and Yahr stage of I or II and 15 age-matched HVs were recruited through the Parkinson's Disease & Movement Disorders Institute, an outpatient unit within the Department of Neurology at New York-Presbyterian Hospital/Weill Cornell Medical College (WCMC). Exclusion criteria were the inability to give informed consent or to undergo MRI (which includes the presence of neurostimulators); clinical diagnosis of depression; diagnosis of a neurodegenerative disorder other than PD; significant neurological illness such as stroke, tumor, and epilepsy; and diagnosis of any significant concomitant medical disease. Individuals with PD were evaluated using the Unified Parkinson's Disease Rating Scale (UPDRS). The Institutional Review Board of the WCMC approved the study and each participant gave written informed consent prior to inclusion.
MR Data Acquisition and Processing
All neuroimaging studies were conducted on a research-dedicated GE 3T EXCITE MR system (GE Medical Systems, Milwaukee, WI, USA) using a commercial, double-quadrature dual-tuned 1H/31P head coil (Clinical MR Solutions, LLC, Brookfield, WI). ROIs were the striatal area and the cortical gray matter (GM). In PD patients, the hemispheres were assigned ipsilateral and contralateral to the clinically most affected body side. In HVs, the mean value of metabolite levels was calculated from both hemispheres for each compound and location. The fractions of gray matter, white matter, and cerebrospinal fluid within the volumes of interest were not taken into account.
Each subject underwent a standard structural brain MRI study. A three-plane, low-resolution, high-speed scout imaging series was first obtained, followed by a series of high-resolution scans, consisting of a series of standard axial T1-, T2-, and spin density-weighted scans. These images were used to prescribe the voxels and slices for the 1H and 31P MRS data acquisitions. In addition, an axial fast fluid-attenuated inversion recovery (FLAIR) scan was performed to rule out potentially exclusionary focal brain lesions.
Multislice 1H MRSI Data Acquisition and Processing
Regional brain levels of NAA, tCr, tCho and, if present, lactate were obtained using a multislice 1H MRSI technique and the commercial double-quadrature dual-tuned 1H/31P head coil, as previously described.[22, 23] Briefly, the multislice MRSI method of Duyn et al employed for these acquisitions is a slice-interleaved spin echo sequence that incorporates octagonally tailored outer volume presaturation pulses for pericranial fat and tissue suppression, and a single water-selective radiofrequency pulse followed by strong spoiler gradients for water suppression. Each MRSI data set was recorded in approximately 16 minutes from four interleaved 15-mm brain slices—prescribed as shown in Figures 1(A) and (B)—with TE/TR 280/2,300 ms, field-of-view (FOV) 240 mm, 24 × 24 phase-encoding steps with circularly sampled k-space, 512 time-domain points, and 2,500 Hz spectral width. The resulting nominal MRSI voxel size was 1.0 × 1.0 × 1.5 cm3.
The 1H MRSI data thus recorded were transferred to an offline Sun Microsystems workstation for analysis by a blinded study investigator, who used in-house data analysis software written in Interactive Data Language (IDL; Exelis Visual Information Solutions, Boulder, CO, USA). The raw data were sorted by slice, zero-filled to a spatial matrix of 32 × 32 and twice along the acquisition domain (to 2,048 sample points), filtered with a Gauss–Lorentz window and a Hamming window along the time and spatial domains, respectively, and then processed by standard 3-dimensional fast Fourier transformation to yield an array of 32 × 32 spectra (see sample in Fig 1C). The spectral data were automatically corrected for susceptibility shifts due to slight variations in magnetic field strength across the brain, and then fitted in the frequency domain to obtain the peak area for each metabolite, which were expressed as ratios relative to root mean square of the background noise as previously described.[22, 23]
Single-Slice 31P MRSI Data Acquisition and Processing
All 31P MRSI data were recorded with the same dual-tuned double-quadrature 1H/31P volume head used for the 1H MRSI scans from a single 30-mm slice, prescribed to be coaxial with the central slice in the 1H MRSI lactate scan (Fig 1D), thereby ensuring overlap of regions of interest (ROIs) in the two scans for enhanced correlation (Fig 1E). A pulse sequence (“FIDCSI”) supplied by the instrument manufacturer and consisting of a single slice-selective radiofrequency pulse followed by a slice refocusing gradient and phase-encoding gradients was implemented with 14 × 14 phase-encoding steps, a FOV of 420 mm, 2,048 sample points, a 5 kHz spectral width, TR 1,000 ms, and eight excitations per phase-encoding step, to yield voxels with a nominal size of 3.0 × 3.0 × 3.0 cm3 in a total scan time of 26 minutes. The field homogeneity in the 31P MRSI slice was automatically optimized by the host computer on the water proton signal from the same slice detected with the 1H coil prior to initiating the 31P acquisition. The raw data matrix size was zero-filled to 32 × 32 spatial points prior to standard 3D Fourier transformation to yield a grid of spectral voxels (Fig 1F) that were selected for further postprocessing as for the proton data.
Statistical analyses were performed using IBM SPSS Statistics, version 18 (SPSS Inc., Chicago, IL, USA). Normality was tested using Shapiro–Wilk tests. HEP concentrations were calculated by adding the values for PCr and ATP. Metabolite levels and ratios as measured by MRS were compared using independent sample t-tests or, where appropriate, Mann-Whitney U tests, with a significance level of α < 0.05 without correction for multiple comparisons. For the group differences, 95% confidence intervals (CIs) were calculated.
For PD patients, Pearson or, where appropriate, Spearman correlation coefficients were calculated with Bonferroni correction for multiple comparisons to quantify the association of metabolite levels and clinical parameters. Additionally, metabolite levels between hemispheres were compared using t-tests for paired samples.
Participants with PD (10/20 were women, mean age 57.7 ± 12.0) had a mean clinical disease duration of 3.2 ± 1.8 years with a mean age at onset of 55.6 ± 12.0 years and a UPDRS score (UPDRS II + III) of 20.9 ± 12.0. Three of 20 patients were treated with levodopa alone, 3/20 with levodopa plus dopamine agonist, 5/20 with dopamine agonists alone, and 8/20 were not taking PD medications. Control subjects had a mean age of 58.5 ± 10.8 years without any significant age difference to the PD group (P = .824).
1HMRS was conducted in all participants; 31PMRS was conducted in all PD patients and in 12/15 healthy subjects. Due to movement artifacts, lactate levels could only be obtained in 9/20 patients and 10/15 HVs.
Metabolite concentrations as measured by 1H and 31P MRS were not significantly different in PD versus control subjects in any of the regions (Table 1). HEPs ATP and PCr, as well as their sum, were similar in PD and controls (in the striatal region: ATP 44.93 vs. 46.08, P = .76; PCr 32.61 vs. 31.85, P = .73; HEP 77.53 vs. 77.93, P = .94). Free phosphate (striatal Pi 8.03 vs. 7.92, P = .89; GM Pi 6.56 vs. 6.83, P = .66) and ventricular lactate (7.66 vs. 7.31, P = .54) did not differ between groups. For proton MRS measures NAA, tCho, and tCr, no significant differences were detected (NAA 22.58 vs. 22.06, P = .83; tCho 11.4 vs. 11.3, P = .94; tCr 9.9 vs. 9.77, P = .90). Figure 2 illustrates comparable ventricular lactate and striatal PCr/Pi levels in the two groups.
|Ventricle||Control Group||P||95% CI|
|Lactate||7.66 (1.20)||7.31 (1.20)||.54||−0.82; 1.50|
|Pi||7.92 (1.27)||8.03 (2.43)||.89||8.62 (2.89)||.36||−1.66; 1.44|
|PCr||31.85 (5.41)||32.61 (6.40)||.73||32.43 (6.94)||.81||−5.28; 3.76|
|ATP||46.08 (12.65)||44.93 (8.65)||.76||46.78 (9.22)||.86||−6.53; 8.84|
|PCr/Pi||4.09 (0.84)||4.33 (1.18)||.37||4.01 (1.07)||.53||−1.04; 0.55|
|HEP||77.93 (17.36)||77.53 (13.92)||.94||79.21 (15.21)||.83||−10.99; 11.78|
|NAA||22.58 (6.55)||22.06 (7.4)||.83||22.07 (6.94)||.83||−4.38; 5.42|
|Cho||11.40 (3.48)||11.3 (3.79)||.94||11.37 (3.56)||.98||−2.45; 2.64|
|Cr||9.90 (2.83)||9.77 (3.16)||.90||9.5 (2.84)||.69||−1.97; 2.23|
|NAA/Cr||2.30 (0.30)||2.27 (0.83)||.83||2.34 (0.32)||.66||−0.21; 0.25|
|Pi||6.83 (1.34)||6.56 (2.19)||.66||7.38 (2.24)||.45||−1.16; 1.71|
|PCr||30.74 (4.06)||31.31 (6.54)||.79||32.79 (6.27)||.32||−4.86; 3.73|
|ATP||36.83 (6.42)||34.19 (8.41)||.60||46.78 (9.22)||.45||−8.02; 2.74|
|PCr/Pi||4.58 (0.61)||5.17 (1.65)||.74||4.64 (1.01)||.87||−1.60; 0.43|
|HEP||64.93 (11.33)||68.14 (10.22)||.50||71.83 (14.66)||.18||−11.14; 4.73|
Within the PD group, clinical indices (disease duration, UPDRS score, age) did not correlate with metabolite concentrations in the striatal area and cortical GM. Metabolite levels in the hemisphere ipsilateral and contralateral to the clinically most affected body side were not significantly different. Finally, there was no significant difference between medicated (n = 12) or unmedicated (n = 8) PD patients for any of the metabolites, for age, age at onset or the UPDRS. However, disease duration was longer in medicated (4.08 ± 1.68 years) than in unmedicated (1.75 ± 1.04 years) individuals (P = .002).
In the present study, we investigated markers of neuronal integrity and energy metabolism at rest using proton and phosphorus MR spectroscopy in early stages of PD. We observed no differences between participants with PD and control subjects for any of the measured metabolites, or their ratios. Metabolite levels in the more and less affected hemispheres did not differ.
Using 31P MRS, some previous studies reported decreased ATP or HEP levels[10, 12] and increased Pi concentrations indicating energy failure, while others failed to replicate these results.[18, 24] With 1 H MRS, some studies detected reduced NAA in the basal ganglia reflecting neuronal and mitochondrial decline,[25, 26] although the majority of articles report no significant NAA differences to HV in accordance with our results.,[15-17],[24, 27] Also, for lactate, most investigators have failed to detect the lactate accumulation reported by Bowen et al,[10, 14, 27] although in our previous study, there was a significant lactate elevation in the PD group. Interestingly, a recent study by Brockmann et al reported NAA and choline reductions in GBA-associated PD without any changes in energy metabolism, while the same group found decreased HEP concentrations and normal NAA and choline in sporadic PD using an identical MRSI protocol. This suggests that clinical heterogeneity plays an important role and consideration of clinical parameters and potential subgroups is critical.
Accordingly, we propose that our nonconfirmatory results are due to the shorter disease duration, higher number of unmedicated individuals, and lower disease severity in our study compared with those previous studies that demonstrated changes typical for mitochondrial dysfunction and neurodegeneration (Table 2). Our sample size of 20 PD and 15 control subjects was larger than in prior studies investigating advanced PD stages, which suggests that statistical power was sufficient if early-stage PD patients had metabolic abnormalities comparable to PD patients in more advanced stages. However, due to movement artifacts, lactate concentrations were only available for 9/20 PD patients and 10/15 healthy controls, which limited the statistical power of this group comparison. Based on the calculated CIs for group differences in the present study, our data are theoretically consistent with metabolite deviations from the mean striatal concentration in HV of up to 21% in Pi, 16% in PCr, 19% in ATP, and 15% in HEP. However, the relative narrowness of the CIs around zero and the similarity of mean concentrations in the PD and HV group confirm that the null hypothesis should not be rejected.
|Present Study||Emir ||Groeger 2011||Hattingen ||Henchcliffe ||Hu ||Barbiroli |
|Disease severity (H&Y)||I n = 9||II n = 13||IIS-III n = 9||I/II n = 16||1.8 ± 0.7||2.45 ± 1||I n = 0|
|II n = 11||III/IV n = 13||II n = 2|
|III n = 7|
|IV n = 4|
|Disease duration in years||3.2 ± 1.8||∼6||4-25||I/II 6.4 ± 4.3||4.8 ± 1.8||5.9 ± 3.8||11.7 ± 4.9|
|III/IV10.5 ± 4.6|
In addition to clinical heterogeneity, methodological differences cause divergent results. Hattingen et al reported partial volume effects for 1HMRS in the order of 10%, which were not accounted for in the present and most prior studies.[12, 17, 18, 28, 29] Furthermore, most previous studies used metabolite ratios to creatine rather than absolute concentrations or ratios to background noise as in the present study. For example, Groeger et al reported a surprisingly elevated ratio of NAA/Cr in PD, commonly interpreted as NAA increase. However, the ratio might as well be explained by Cr depletion as previously reported or changes in both metabolites.
Although the clinical manifestations of PD occur when the neuropathological changes are already advanced, the metabolic changes might not be detectable at rest in mildly affected individuals due to compensatory mechanisms. One way to increase the sensitivity of MRS to detect metabolic abnormalities in PD might therefore be protocols including some form of cortical activation. Rango et al demonstrated that photic stimulation in PD patients led to a sharp decline of HEP in the recovery phase, whereas healthy controls exhibited an increase. Challenging energy metabolism by increasing the energy demand might uncover latent deficits, which are not detectable at rest. In addition, imaging of the energy metabolism under dynamic conditions reveals specific changes causally related to actual energy deficits and not merely to chronic neuronal decline.
Apart from the characterization of PD-related metabolic features, the correlation of these features with clinical markers is of interest. In our study, no significant correlations between MRS-derived data and age, disease duration, or UPDRS were found in accordance with previous studies.[28, 30] However, in demented PD patients, significant correlations between NAA or NAA/Cr and neuropsychological subtests have been reported.[28, 30] In nondemented PD patients Hu and colleagues found strong correlations between Pi/ATP and IQ reductions (r = 0.96, P < .001), full scale IQ (r = −0.82, P = .006), and verbal IQ (r = 0.7, P = .04). Neuropsychological testing was not conducted in the present study.
In summary, although proton and phosphorus MRS at rest are generally suitable tools to quantify neuronal integrity and oxidative phosphorylation deficits, metabolic changes in early and mildly affected PD patients may be too subtle to be detected reliably with current technology. Increasing the sensitivity and specificity of MRS markers in PD, by investigating metabolites under dynamic conditions or measuring neurotransmitter levels, might not only facilitate studies about the pathogenesis and natural history of PD, but also support the development and monitoring of therapies. This would be crucial in particular in the early disease stages, in which metabolic pathology is subtle, but therapeutic interventions are most promising. Furthermore, characterization of subgroups might improve selection of clinical trial participants and help to learn more about atypical clinical presentation and syndromic overlap with other neurodegenerative diseases. Finally, quantifying therapeutic effects in an objective, sensitive and possibly prompt manner would complement present measures such as clinical scales. All of these potential applications will become more important once neuroprotective treatments become available. In conclusion, our results highlight the unmet need for an early, sensitive, and quantitative biomarker in PD.
This research was funded by the Dana Foundation. CH and MN received support from the Parkinson's Disease Foundation as part of the PDF Research Center of Weill Cornell Medical College.
- 20UPDRS program members. Unified Parkinson's Disease Rating Scale. In: Fahn S, Marsden C, Goldstein M, Calne D, eds. Recent Developments in Parkinson's Disease. Florham Park, NJ: Macmillan Healthcare Information; 1987:153-163., ,