OUP user menu

Frontal, midbrain and striatal dopaminergic function in early and advanced Parkinson's disease A 3D [18F]dopa-PET study

J. S. Rakshi, T. Uema, K. Ito, D. L. Bailey, P. K. Morrish, J. Ashburner, A. Dagher, I. H. Jenkins, K. J. Friston, D. J. Brooks
DOI: http://dx.doi.org/10.1093/brain/122.9.1637 1637-1650 First published online: 1 September 1999

Summary

We have studied focal changes in dopaminergic function throughout the brain volume in early and advanced Parkinson's disease by applying statistical parametric mapping (SPM) to 3D [18F]dopa-PET. Data from seven early hemi-Parkinson's disease and seven advanced bilateral Parkinson's disease patients were compared with that from 12 normal controls. Parametric images of [18F]dopa influx rate constant (Kio) were generated for each subject from dynamic 3D [18F]dopa datasets and transformed into standard stereotactic space. Significant changes in mean voxel [18F]dopa Kio values between the normal control group and each Parkinson's disease group were localized with SPM. Conventional region of interest analysis was also applied to comparable regions on the untransformed image datasets. In early left hemi-Parkinson's disease, significant extrastriatal increases in [18F]dopa Kio were observed in the left anterior cingulate gyrus and the dorsal midbrain region (P < 0.05, corrected) along with decreases in striatal [18F]dopa Kio. In advanced Parkinson's disease, significant extrastriatal decreases in [18F]dopa Kio were observed in the ventral and dorsal midbrain regions (P < 0.05, corrected). No significant changes in [18F]dopa Kio were observed in the anterior cingulate region. In a direct comparison between the early and late Parkinson's disease groups, we observed relative [18F]dopa Kio reductions in ventral and dorsal midbrain, and dorsal pontine regions along with striatal [18F]dopa Kio reductions. Similiar results were found with a region of interest approach, on non-transformed data, except for the focal midbrain [18F]dopa Kio increase seen in early Parkinson's disease. In conclusion, using SPM with [18F]dopa-PET, we have objectively localized changes in extrastriatal, pre-synaptic dopaminergic function in Parkinson's disease. The significance of the increased dopaminergic activity of anterior cingulate in early Parkinson's disease remains unclear, but may be compensatory. The [18F]dopa signal in dorsal midbrain and pontine regions suggests that [18F]dopa is taken up by serotonergic and noradrenergic neurons which also degenerate in advanced Parkinson's disease. This suggests, therefore, that Parkinson's disease is a monoaminergic neurodegenerative disorder.

  • Parkinson's disease
  • 3D [18F]dopa
  • focal changes
  • extrastriatal
  • progression
  • AADC = aromatic amino acid decarboxylase activity
  • COMT = catechol-O-methyl transferase
  • Kio = influx rate constant
  • MTGA = multiple time graphical analysis
  • rCBF = regional cerebral blood flow
  • SPM = statistical parametric mapping
  • UPDRS = Unified Parkinson's Disease Rating Scale

Introduction

[18F]dopa-PET provides an in vivo method for assessing the functional integrity of pre-synaptic dopaminergic function in the basal ganglia in Parkinson's disease (Garnett et al., 1983; Martin et al., 1989; Brooks et al., 1990a). It is known from post-mortem studies in Parkinson's disease that there is degeneration of nigral and mesial frontal dopaminergic terminals as well as striatal loss (Bernheimer et al., 1973; Scatton et al., 1982; Kish et al., 1988; Fearnley and Lees, 1991). However, to date, it has been difficult to study these changes with [18F]dopa-PET because of the lower specific uptake of [18F]dopa in these regions, coupled with the limitations of resolution and sensitivity of 2D [18F]dopa-PET scanning.

We have implemented and validated 3D acquisition, reconstruction and analysis of dynamic [18F]dopa-PET data (Rakshi et al., 1996), resulting in a fourfold increase in the sensitivity of brain signal to noise ratios in 3D compared with 2D acquired PET studies (Cherry et al., 1991; Bailey, 1992). In addition, the increased sensitivity enables a higher reconstructed resolution with a reduction in the partial volume effects leading to a further overall improvement in image quality. Therefore, 3D [18F]dopa-PET should be more sensitive and accurate at detecting focal changes within extrastriatal and striatal dopaminergic regions (see Fig. 1).

Fig. 1

3D data acquired by [18F]dopa-PET scan of a normal subject, demonstrating striatal and extrastriatal [18F]dopa uptake.

The conventional method of [18F]dopa analysis uses a region of interest approach with multiple time graphical analysis (MTGA) (Patlak and Blasberg, 1985) to sample putamen and caudate function. However, the size, shape and placement of these regions is arbitrary, producing inter- and intra-observer variation, and potential observer bias. Furthermore, in severe Parkinson's disease it can be difficult to identify the striatal boundaries for placement because of the reduced [18F]dopa uptake.

We have applied statistical parametric mapping (SPM) to 3D [18F]dopa-PET studies, enabling us to objectively localize focal changes in dopaminergic function in Parkinson's disease throughout the brain volume without having to make an a priori hypothesis as to their location. SPM was initially developed for PET activation studies and localizes significant differences in function in spatially normalized brain images on a voxel-by-voxel basis (Friston et al., 1991b). SPM has now been applied to a number of PET ligands including [11C]flumazenil to localize changes in GABAA receptor binding in focal epilepsy (Richardson et al., 1996) and to [11C]diprenorphine to study alterations in opioid receptor binding in Huntington's and Parkinson's disease (Weeks et al., 1995; Piccini et al., 1997).

In addition to applying SPM, we employed the standard region of interest and MTGA analytical approach (Brooks et al., 1990b) to sample striatal and midbrain regions in untransformed images and, subsequently, brain areas revealed by SPM to have altered dopaminergic function in Parkinson's disease.

Subjects and method

Subjects

Seven patients with early hemi-Parkinson's disease clinically affecting their left limbs [mean age 57 ± 5 years; mean motor Unified Parkinson's Disease Rating Scale (UPDRS) score 9 ± 3 (range 4–12); Hoehn and Yahr stage 1; and mean symptom duration 19 ± 5 months] and seven patients with advanced bilateral Parkinson's disease [five left dominant, two right dominant; mean age 54 ± 8 years; mean motor UPDRS score 41 ± 15 (range 22–66); Hoehn and Yahr stage 3–5; and mean symptom duration 12 ± 4 years] underwent [18F]dopa-PET scanning. Findings were compared with those of a group of 12 normal subjects, mean age 57 ± 11 years.

All patients fulfilled the UK Parkinson's disease Brain Bank criteria for idiopathic Parkinson's disease (Gibb and Lees, 1988). All Parkinson's disease patients were on regular levodopa therapy except for two of the early Parkinson's disease patients who were drug naive. In addition, three of the advanced Parkinson's disease patients were taking a dopamine agonist, and two a monoamine oxidase-B inhibitor.

Each patient had their medication stopped at least 12 h before their PET scan and was clinically assessed by a single observer, using the UPDRS and Hoehn and Yahr rating scales (Hoehn and Yahr, 1967; Fahn and Elton, 1987) just before their scan. Parkinson's disease patients with significant cognitive impairment, as assessed by the Folstein's Mini-Mental Test (Folstein et al., 1975), were excluded from the study (<28 out of 30). All controls had a normal neurological examination and showed no evidence of rest tremor, rigidity or bradykinesia. No normal volunteer was taking medication.

All patients and normal volunteers gave written informed consent, after a full explanation of the procedure. Permission to perform these studies was granted by the Ethics Committee of the Royal Postgraduate Medical School, London, UK and the Administration of Radioactive Substances Advisory Committee (ARSAC), UK.

Scanning protocol

In addition to having their medication stopped 12 h before their PET scan, all subjects were fasted on the morning of the scan and received an oral bolus of 150 mg carbidopa and 400 mg entacapone, a peripheral catechol-O-methyl transferase inhibitor (COMT; Orion Farmos Pharmaceuticals Espoo, Finland) 1 h before scanning (Sawle et al., 1994; Ishikawa et al., 1996a).

The [18F]dopa-PET scans were performed with an ECAT 953B neuroscanner (CTI/Siemens, Knoxville, Tenn., USA), with an axial field of view of 10.8 cm, and in 3D acquisition mode with protocols that have been previously reported (Bailey, 1992), yielding 31 planes with a slice separation of 3.4 mm and an average in-plane resolution of 6 mm full-width half-maximum. A correction for tissue attenuation of 511 KeV gamma radiation was measured with a 10 min 2D transmission scan performed prior to tracer injection and acquired using retractable 68Ga/ 68Ge sources.

[18F]Dopa (140–180 Mbq in 10 ml of normal saline solution) was infused intravenously over 30 s. Scanning began at the start of the tracer injection with a protocol of 25 time frames over 94 min (4 × 1 min, 3 × 2 min, 3 × 3 min, 15 × 5 min). The subjects were positioned such that the orbitomeatal line was parallel to the transaxial plane of the tomograph and head position was carefully monitored throughout the scan.

Data analysis

Two methods of analysis were employed: (i) SPM and (ii) standard region of interest approach with MTGA. Analysis of data was performed on a SUN Sparc 10 workstation (Sun Microsystems, Silicon Valley, Calif., USA) using ANALYZE 7.0 (Mayo Foundation, Baltimore, Md., USA) (Robb and Hanson, 1991) and in-house software written in IDL image analysis software (Research Systems, Inc., Boulder, Col., USA). Parametric images of [18F]dopa influx rate constants (Kio) were generated with IDL image analysis software. Image transformation and SPM analysis was performed using SPM software (SPM95; Wellcome Department of Cognitive Neurology, London, UK) implemented in Matlab (Mathsworks Inc., Sherborn, Mass., USA).

SPM analysis

Image transformation.

From each subject's dynamic 3D [18F]dopa-PET study, we generated [18F]dopa Kio on a voxel-by-voxel basis for all 31 image planes using the MTGA approach with an occipital tissue input function. Thus, we were able to produce parametric images of [18F]dopa Kio for each subject. Tissue count sampling was performed using time frames 14–25 (30–94 min) post-injection.

There is insufficent anatomical detail in parametric images of [18F]dopa Kio to enable accurate spatial normalization directly to the regional cerebral blood flow (rCBF) template within the SPM software (Friston et al., 1991a). We have, therefore, developed an indirect approach whereby we use the combined time frames 1–25 of the dynamic 3D [18F]dopa image to produce an integrated `add image' (0–94 min). This `add image' contains the earlier time frames which are blood flow dependent and so generates sufficient cortical as well as striatal detail to allow accurate spatial normalization to the SPM template (see Fig. 2). Applying this method we produced an `add image' for each subject and spatially normalized it to the rCBF template. The resulting transformation parameters were then applied to the corresponding subject's parametric image of [18F]dopa Kio, allowing all the parametric images to be transformed into the standard stereotaxic space of Talairach and Tournoux (Rakshi et al., 1998).

Fig. 2

Direct comparison of normal subject's spatially normalized [18F]dopa `add image' with the standard rCBF template displayed in SPM.

This then allowed comparisons to be made across scan datasets, in analogous voxel regions of the brain volume and to combine [18F]dopa-PET datasets from different subjects to perform group analyses. Following spatial normalization of the parametric images (Kio), a Gaussian kernel of 8 × 8 × 6 mm (full width half maximum in the x, y and z planes, respectively) was applied to remove high-frequency noise from the images.

SPM data analysis.

Categorical comparisons of mean [18F]dopa voxel Kio values between the Parkinson's disease and normal control groups were made applying SPM. Three between-group comparisons were made: (a) seven early left hemi-Parkinson's disease patients versus 12 normal controls; (b) seven advanced bilateral Parkinson's disease patients versus 12 normal controls; (c) seven early left hemi-Parkinson's disease patients versus the seven advanced bilateral Parkinson's disease patients.

Between-group comparisons were performed using appropriately weighted categorical contrasts to generate SPMs for both increases and decreases in mean voxel [18F]dopa Kio values on a voxel-by-voxel basis. No global normalization was applied since we used measured voxel [18F]dopa Kio values which were independent of rCBF. Significant differences in mean voxel [18F]dopa Kio values for the between-group comparisons (a), (b) and (c) were localized using SPM. The contrasts were used to derive between-group Z scores (unpaired t statistic transformed to the standard normal distribution) on a voxel-by-voxel basis using the general linear model (Friston et al., 1991b, 1995). The P values associated with these regional effects were corrected for multiple-dependent comparisons implicit in the SPM. Maps of Z scores surviving a threshold of >2.33 (P < 0.01) were further corrected for multiple comparisons. These surviving voxels were rendered on to a stereotactically normalized MRI (see Fig. 3).

Fig. 3

SPM{Z} transverse, sagittal and coronal maximum intensity projection maps rendered on to a stereotactically normalized MRI scan, showing areas of significant increases and decreases in [18F]dopa Kio uptake in A and D, early left hemi-Parkinson's disease (PD) compared with normal controls; in B and E, advanced bilateral Parkinson's disease compared with normal controls; and in C, advanced Parkinson's disease compared with early Parkinson's disease. Z score threshold for significance: P < 0.05, corrected.

Region of interest analysis

We applied standard region of interest and MTGA analysis to the original untransformed datasets to sample frontal, midbrain and striatal regions. The influx rate constants [18F]dopa Kios were calculated from the time frames 14–25 (30–94 min) post-injection for tissue count activity using occipital tissue counts as the input function. For the striatum, circular regions of 10 mm diameter were placed over each dorsal head of caudate and an elliptical region of 10 × 24 mm over each dorsal putamen aligned to the long axis on three contiguous transverse planes.

With the increased resolution and sensitivity of our 3D [18F]dopa-PET scanner, we were now able to identify the midbrain and frontal regions more clearly by visual image inspection and with reference to the stereotaxic atlas (Talairach and Tournoux, 1988). We sampled the midbrain region by placing a single circular region of 16 mm diameter, encompassing both ventral and dorsal midbrain regions, on two contiguous transverse planes. We retrospectively sampled the anterior cingulate region. This was achieved by initially identifying the region of interest on the stereotaxic atlas using the coordinates of the frontal focus determined from the SPM analysis. We then identified this region on the [18F]dopa-PET scan by the position of the anterior cingulate cortex in relation to the striatum on the PET image. However, because this method is imprecise, we used regions encompassing a larger volume than the SPM focus so that the region of interest would be included. Two circular regions of interest of diameter 16 mm for right and left anterior cingulate regions corresponding to the foci where significant increases in [18F]dopa Kio values were observed with SPM analysis were applied.

For the input function, we sampled right and left occipital lobes with two circular regions of 32 mm diameter placed on the same three contiguous planes as those selected for the dorsal striatal regions. All regions of interest were placed manually using the IDL software.

The mean [18F]dopa Kio values for each frontal, midbrain, putamen and caudate region in the early and advanced Parkinson's disease groups were compared with the [18F]dopa Kio values of normal subjects using Student's unpaired two-tailed t test.

Results

Frontal region

SPM analysis

In early left hemi-Parkinson's disease we observed a significant increase in [18F]dopa Kio in the the left anterior cingulate gyrus (Brodmann area 24/32) compared with the normal control group (Table 1 and Fig. 3A). There was also a similiar trend in the right cingulate but this did not reach statistical significance on correction (Z = 3.8, P < 0.076).

View this table:
Table 1

SPM findings for in between-group comparisons showing the locations of significant increases and decreases in [18F]dopa uptake in Parkinson's disease

Z scoreTalairach co-ordinatesP value (corrected)
xyz
Early Parkinson's disease group, n = 7; advanced Parkinson's disease group, n = 7; normal control group, n = 12.
Comparison of early left hemi-Parkinson's disease with normal control group
Increases
Left anterior cingulate gyrus4.98 –232240.001
Dorsal midbrain4.05 –4–38 –40.038
Decreases
Right putamen6.0726 –4 40.0001
Left putamen5.03–26 –8 40.001
Right anterior putamen4.121812 00.030
Dorsal body right caudate4.0022–16160.044
Comparison of advanced bilateral Parkinson's disease with normal control group
Decreases
Dorsal and ventral midbrain4.56 4–32–120.008
Left anterior putamen6.84–1814 40.0001
Left putamen6.81–26 –6 40.0001
Right putamen6.8024 4 40.0001
Dorsal body right caudate6.7818 6120.0001
Right anterior putamen6.632010 00.0001
Comparison of advanced bilateral Parkinson's disease with early left hemi-Parkinson's disease
Decreases
Dorsal pons5.27 2–36–240.001
Dorsal and ventral midbrain4.45–14–32 00.044
Left putamen4.42–24–10 –40.049

In advanced Parkinson's disease, no significant change in mean cingulate [18F]dopa Kio was observed, either when compared with the normal controls or directly with the early hemi-Parkinson's disease group.

Region of interest analysis

Following the unexpected frontal rise in [18F]dopa Kio in early Parkinson's disease we sampled the corresponding regions on the original untransformed datasets, independent of SPM. In early left hemi-Parkinson's disease, we found a 36% and 32% increase in mean right and left cingulate [18F]dopa Kio's, respectively (Table 2). There was no significant change in [18F]dopa Kio in the advanced Parkinson's disease group.

View this table:
Table 2

Mean regional right and left putamen, caudate, mesial frontal and midbrain Kio values in normal controls, early left hemi-Parkinson's disease and advanced bilateral Parkinson's disease

Kio/min
Right putamenLeft putamenRight caudateLeft caudateRight frontalLeft frontalMidbrain
n.s. = not significant.
Normal (n = 12)
Mean0.01710.01710.01680.01680.00160.00170.0079
SD0.00180.00180.00280.00230.00040.00040.0018
Range0.0145–0.02020.0145–0.02090.0135–0.02330.0137–0.02120.0012–0.00200.0012–0.00230.0062–0.00120
Hemi-Parkinson's disease (n = 7)
Mean0.00610.00960.01080.01230.00250.00250.0081
SD0.00050.00130.00160.00180.0007 0.00080.0019
Range0.0052–0.00690.0079–0.01090.0085–0.01330.0093–0.01480.0016–0.00370.0018–0.00350.0059–0.0110
P values0.00010.00010.00010.00010.0020.011n.s.
Bilateral Parkinson's disease (n = 7)
Mean0.00360.00510.00810.00920.00170.00190.0057
SD0.00130.00110.00180.00160.00060.00040.0007
Range0.0012–0.00520.0033–0.00620.0046–0.01020.0012–0.01090.0009–0.00270.0014–0.00250.0050–0.0067
P values0.00010.00010.00010.0001n.s. n.s. 0.006

Midbrain and pons regions

SPM analysis

In the early Parkinson's disease group we observed a significant increase in dorsal midbrain [18F]dopa Kio, compared with the normal control group (Table 1 and Fig. 3A).

In advanced bilateral Parkinson's disease there was a significant reduction in both ventral and dorsal midbrain [18F]dopa Kio compared with the normal control group involving the substantia nigra, midbrain tegmentum and dorsal raphe nuclei (Table 1 and Fig. 3B).

In the direct comparison of advanced bilateral Parkinson's disease with early hemi-Parkinson's disease significant reductions in mean [18F]dopa Kio were observed in the dorsal pons as well as the dorsal and ventral midbrain regions and striata (Table 1 and Fig. 3C).

Region of interest analysis

No significant changes in midbrain [18F]dopa Kio were found in early hemi-Parkinson's disease (Table 2).

In the advanced Parkinson's disease group there was a 28% reduction in mean midbrain [18F]dopa Kio at the level of the nigra compared with normal controls.

Striatal regions

SPM analysis

A comparison of the early left hemi-Parkinson's disease with the normal control group showed significant reductions in mean [18F]dopa Kio throughout the right and left putamen; these were more widespread and significant in the right putamen (Table 1 and Fig. 3D). In addition to the main putamen focus, there was also a second focus located in the right anterior putamen which extended into right head of caudate and a further separate focus in the dorsal body of the right caudate (Table 1 and Fig. 3D).

In advanced Parkinson's disease compared with the normal control group, more extensive and highly significant reductions in [18F]dopa Kio values were found throughout both the dorsal and ventral striata (Table 1 and Fig. 3E).

In the direct comparison of advanced bilateral Parkinson's disease (left-dominant) with early left hemi-Parkinson's disease, interestingly, only the relative further reduction in left putamen reached statistical significance at P < 0.05, corrected (Table 1). There was a trend, however, towards relative reductions throughout the striata (P < 0.05, uncorrected).

Region of interest analysis

In the early left hemi-Parkinson's disease group there was a 64% and 44% reduction in mean right and left putamen [18]dopa Kio, respectively (Table 2). There were corresponding 36% and 27% reductions in mean right and left caudate [18F]dopa Kio's (Table 2).

In the advanced bilateral Parkinson's disease group there were mean reductions of 79% and 70% in right and left putamen [18]dopa Kio values (Table 2). There were corresponding 52% and 45% reductions in right and left caudate [18F]dopa Kio values, respectively.

Discussion

We have applied SPM to spatially normalized parametric images of [18]dopa Kio, and localized similiar reductions in striatal [18]dopa Kio in early and advanced Parkinson's disease to those detected using the standard region of interest approach. SPM, however, enabled us to localize objectively focal [18F]dopa Kio changes in anterior cingulate, midbrain and pontine regions which could not have been predicted by visual inspection of [18F]dopa-PET images. This is because of their lower specific uptake of [18F]dopa.

Methodological issues

In practice, as with all methods and measurements, there are sources of error in the technique, and their understanding is required for the proper interpretation of the results. The main methodological issues are: 3D PET data acquisition, quantification and interpretation of [18F]dopa uptake, and the application of SPM to [18F]dopa datasets.

3D scanning has a great sensitivity advantage over 2D acquisition. The technique, however, is complex and some of the problems involved have yet to be completely resolved (Badawi, 1997). In 3D mode or volume imaging, the interplane septa of the PET scanner are removed and coincidence events may be detected by pairs of detectors located in any ring. This greatly increases the camera sensitivity to true coincident events, but also increases the sensitivity to scattered coincidences. One therefore has to employ a scatter correction method for 3D datasets (Cherry et al., 1993; Bailey and Meikle, 1994). Another difficulty relates to the requirement that all pairs of detectors in coincidence should have the same sensitivity, i.e. detector normalization. It has been shown that there is no unique normalization for both scatter and true coincidences or for all counts rates. Detector normalization is not perfect and is still being improved. Finally, because of the high count rates that detectors have to process in 3D acquisition, there is an increase in dead-time and random coincidence events detected. Ingenious methods to overcome these limiting factors in 3D mode are currently being developed.

Nevertheless, accepting the problems and limitations above, we have implemented 3D [18F]dopa-PET and successfully performed a validation experiment. We have also demonstrated improved image quality and reproducibility compared with our previous 2D protocol (Rakshi et al., 1996; Trebossen et al., 1996) and were able to completely discriminate both contralateral and ipsilateral putamen [18F]dopa Kio values in early hemi-Parkinson's disease from normal controls.

The objective of a parameter estimation using [18F]dopa-PET is to measure a biological variable of interest to the pathophysiology of Parkinson's disease. With [18F]dopa-PET, we would ideally like to measure k3, the rate constant which directly reflects aromatic amino acid decarboxylase activity (AADC) and thus pre-synaptic dopaminergic function (Dhawan et al., 1996). However, [18F]dopa is mainly peripherally metabolized in the presence of a peripheral AADC inhibitor (carbidopa), to 3-O-methyl-[18F]dopa by COMT (Boyes et al., 1986), which then crosses the blood–brain barrier and contributes to non-specific and striatal [18F] activity. This, along with the formation of other brain dopamine metabolites, makes direct kinetic modelling of k3 very complex (Gjedde et al., 1991; Kuwabara et al., 1995) and to date has not provided additional biological or clinical information on Parkinson's disease over simpler methods of deriving an [18F]dopa Kio reflecting AADC activity (Dhawan et al., 1996; Ishikawa et al., 1996b; Cumming et al., 1997; Ichise and Ballinger, 1997).

We have therefore employed the simpler but less specific MTGA approach to measure the net transfer of [18F]dopa across the blood–brain barrier as the [18F]dopa Kio, where Kio is the slope of the Patlak plot and reflects the rate of unidirectional striatal [18F]dopa accumulation into dopaminergic pre-synaptic terminals, its subsequent decarboxylation and vesicular storage. [18F]dopa Kio, however, is also influenced by factors other than those primarily related to [18F]dopa accumulation and decarboxylation, such as plasma and amino acid concentrations, peripheral and central COMT activities and cerebral blood flow. We have attempted to minimize these effects by fasting all subjects prior to scanning and by using a peripheral COMT inhibitor (Sawle et al., 1994). We have also used occipital counts for the input function to minimize errors due to attenuation correction, scan detector sensitivity or metabolite correction in 3D. Finally, it should be noted that brain [18F]dopa uptake is a measure of its capacity to decarboxylate exogenous dopa and, as such, measures pre-synaptic dopaminergic function. It does not provide a measure of the endogenous rate of dopamine synthesis.

Another important consideration is the effect of levodopa administration on [18F]dopa uptake in Parkinson's disease. We have found no significant difference in [18F]dopa uptake in early or advanced Parkinson's disease subjects scanned twice within 1 month, on one occasion taking their levodopa medication and on the other occasion off levodopa medication, stopped 12 h prior to their PET scan (R. Ceravalo, unpublished observations). The long-term effects of exposure to levodopa therapy on AADC activity in Parkinson's disease are, however, unknown.

Finally, it has been suggested that to improve region of interest placement, one can co-register each patient's individual MRI to their [18F]dopa-PET scan and place the regions on the co-registered image. We have not found this approach to signifiantly change [18F]dopa Kio values or improve reproducibilty when compared with the standard region of interest method (unpublished data). This may be because the various steps involved in image co-registration also introduce sources of error. With advancing PET technology, this method may ultimately improve the accuracy of region of interest placement.

With respect to SPM, there are three main issues: (i) accurate spatial normalization; (ii) applying Gaussian smoothing; and (iii) the assumption that the voxel [18F]dopa Kio values across the parametric images are normally distributed for all subject groups.

It is possible that the relative differences in contrast and intensity between [18F]dopa `add images' and the rCBF template might affect spatial normalization. The normalization process is partly dependent on these factors, but the degree to which it might be affected will also depend on the particular constraints of the normalization programme. In practice, however, the [18F]dopa `add images' closely resemble the rCBF template (Fig. 2) because the cortex and striatum have relatively high blood flow and the `add image' incorporates the earlier time frames which are highly blood flow dependent. Furthermore, if the spatial normalization was imprecise this would represent a substantial source of error variance and we would not have obtained high Z scores for [18F]dopa Kio changes in Parkinson's disease. Accurate spatial normalization is further supported by the SPM maps being accurately rendered on to the putamen and caudate regions of the standard normalized MRI, with the maximal scores identified in Talairach space corresponding to putamen and caudate structures (see Fig. 3).

In future, however, there is a strong case for developing an [18F]dopa template, rather than using an rCBF template for direct spatial normalization. SPM has now been successfully applied to a number of other PET radioligands (Weeks et al., 1995; Richardson et al., 1996; Piccini et al., 1997).

We arbitrarily applied minimal Gaussian smoothing (8 mm) in order to increase the signal to noise ratio and therefore the sensitivity. This step did not alter our results and could have been omitted. In general, irrespective of the size of the structure giving a signal, smoothing will increase signal to noise and therefore sensitivity, but will degrade resolution (Friston et al., 1995).

Finally, addressing the issue of the normal distribution of voxel [18F]dopa Kio values. This assumption is supported by no outliers being observed at the maximal Z score plots in the three Parkinson's disease studies and can be inferred from a normal distribution of region of interest [18F]dopa Kio values in normal and Parkinson's disease populations, both from this and previous PET studies. The unpaired t test is generally considered to be a robust statistical test, capable of handling suboptimal normally distributed data. At Z scores greater than 4.0, the permutation test is thought to be unnecessary (Holmes et al., 1996).

Extrastriatal changes in early Parkinson's disease

An unexpected finding revealed by SPM was increased [18F]dopa uptake in the anterior cingulate and dorsal midbrain regions in early hemi-Parkinson's disease (see Fig. 3A). In addition to the reasons already discussed, these extrastriatal [18F]dopa Kio changes are unlikely to be artefactual false positives because the anterior cingulate result was confirmed retrospectively, independent of stereotaxic transformation and SPM, by applying conventional region of interest analysis to the corresponding cingulate regions on the original untransformed images.

With region of interest analysis, we observed similiar mean right and left anterior cingulate [18F]dopa Kio increases but only a significant increase in the left anterior cingulate with SPM. (SPM showed a trend in right cingulate Z = 3.8, P < 0.076.) However, we applied a larger region of interest than the SPM focus which would have reduced the value of the higher left cingulate [18F]dopa Kio.

The significance of increased [18F]dopa Kio is unclear but suggests increased AADC activity. There are a number of possible explanations for this finding which are now discussed. Our knowledge of brain AADC activity in health and disease is limited because there are no direct in vivo human studies. Furthermore, there are no post-mortem studies in early hemi-Parkinson's disease to determine directly early striatal and extrastriatal dopaminergic changes. Therefore, the exact pathophysiological changes occurring in dopaminergic neurons in the very early stages of Parkinson's disease are unknown.

The rate limiting step of endogenous dopamine formation within pre-synaptic dopaminergic nerve terminals is tyrosine hydroxylase which converts l-tyrosine to l-dopa. Under normal physiological conditions AADC which converts l-dopa to dopamine is not thought to be regulated. However, a number of studies have suggested that AADC can be regulated under pathophysiological conditions. Human post-mortem brain studies have suggested that AADC may be up-regulated in dopamine neurons that are spared during ageing (Agid et al., 1987; Kish et al., 1995), and increased AADC activity has been demonstrated in patients with psychosis (Reith et al., 1994). It is possible, therefore, that AADC may undergo regional regulatory changes in Parkinson's disease.

The anterior cingulate observation (Brodmann area 24/32) corresponds neuroanatomically to the mesocortical region which receives dopaminergic projections from the ventral tegmental area A10 (Lindvall and Bjorklund, 1974; Simon and Le Moal, 1984). This projection forms part of the mesocorticolimbic dopaminergic system which is involved in motor, cognitive and behavourial functions (Glowinski et al., 1984; Le Moal and Simon, 1991). Dopaminergic projections to the cerebral cortex, however, also arise from dopamine neurons in the substantia nigra (Bjorklund and Lindvall, 1984). The mesocortical system, which includes the pre-frontal region, is distinct from other ascending dopaminergic projections, i.e. nigrostriatal and mesolimbic, and has unique intrinsic properties. For example, they completely lack or have greatly reduced numbers of autoreceptors on their cell bodies, dendrites and nerve terminals. They also have a higher neuronal firing rate, different activity pattern, higher dopamine turnover and a diminshed response to dopamine agonists/antagonists (Bannon and Roth, 1983).

Another interesting feature of the mesocortical system, which could possibly explain our PET finding, is the inverse relationship demonstrated between the mesocortical and nigrostriatal dopaminergic systems (Pycock et al., 1980). This was first demonstrated in animal studies where dopaminergic neurons projecting to the mesocortex were specifically lesioned in rats. This led to an increase in dopamine levels, its metabolites and D2 receptor binding and sensitivity in the striatum. Similiar findings have now been demonstrated in non-human primates (Roberts et al., 1994). It has therefore been proposed that the mesocortical dopaminergic system has an inhibitory regulatory effect on nigrostriatal dopaminergic systems. Conversely, nigrostriatal lesions could produce an increase in dopamine turnover associated with increased AADC activity in the mesocortex. Surprisingly, there are no studies investigating the reverse relationship.

A further unique feature of the mesocortical dopaminergic neurons is their preferential sensitivity to physical and psychological stress (Roth and Elsworth, 1995). Therefore our finding could instead represent a stress related response in early Parkinson's disease.

On a general dopaminergic perspective, it has been proposed that there are compensatory or adaptive processes in Parkinson's disease to counterbalance the initial nigral dopaminergic cell loss and that symptoms appear when dopamine depletion reaches a critical threshold and these adaptive processes fail (Zigmond et al., 1990). One such possible compensatory mechanism could involve increased synthesis and release of dopamine from residual midbrain dopaminergic neurons. Animal studies have shown that partial or unilateral midbrain dopaminergic lesions result in increased dopamine content and turnover in the remaining striatal dopaminergic terminals (Agid et al., 1973; Andersson et al., 1980; Melamed et al., 1982) perhaps related to the sprouting of residual dopaminergic nerve terminals (Blanchard et al., 1996).

In our [18F]dopa-PET study we have demonstrated increased [18F]dopa Kio suggestive of increased AADC activity. The question is whether this results from, or leads to, increased dopamine synthesis or turnover in early Parkinson's disease. If so, another alternative interpretation of our finding would be that up-regulation of AADC activity in the A10 projection is a compensatory or secondary effect, reflecting early dysfunction in that projection. One would then speculate that similiar up-regulation of AADC occurs in the nigrostiatal projection in pre-clinical Parkinson's disease, but by clinical presentation; although there is increased AADC in individual dopaminergic neurons, the total striatal AADC is reduced because of the continuing neuronal loss. [18F]dopa-PET would then underestimate dopaminergic neuronal loss in early disease. In summary, there is no clear explanation of our finding; however, there are a number of interesting possibilities which we have proposed.

The SPM finding of increased dorsal midbrain [18F]dopa Kio in early hemi-Parkinson's disease was not observed on region of interest analysis. SPM detected a specific midbrain area of increased [18F]dopa Kio, whereas region of interest sampled whole midbrain activity being unable to distinguish clearly between ventral and dorsal midbrain regions on an integrated `add image'. Although the focus occured at the level of the superior colliculus, it extended forward and may represent the rostral component of the raphe nuclei.

We now know that a single AADC enyzme catalyses the decarboxylation of both l-dopa in catecholaminergic neurons and 5-hyroxytryptophan in serotonergic neurons (Tison et al., 1991). [18F]dopa will therefore be taken up by both serotonergic and noradrenergic neurons and reflect their function within the brainstem, i.e. [18F]dopa-PET is not specific for dopaminergic neurons. With improved [18F]dopa-PET images, we have demonstrated high signal uptake in the dorsal brainstem corresponding to the regions of the serotonergic neurons of the median raphe nuclei, and the noradrenergic neurons of the locus coeruleus (Figs 1 and 3). Therefore, our dorsal midbrain finding may represent increased AADC activity in serotonergic neurons in this region.

In a previous PET study of early Parkinson's disease with [11C]WIN 35,428, a cocaine analogue which specifically labels the dopamine transporter, an 84% reduction in midbrain binding was reported (Frost et al., 1993). We did not observe a reduction in midbrain [18F]dopa uptake in our early hemi-Parkinson's disease cases. This may reflect either a milder degree of involvement at this stage (their five patients had more advanced and bilateral Parkinson's disease), and/or relative preservation of dopa decarboxylase activity compared with dopamine re-uptake site binding. Alternatively, it has been suggested that dopamine re-uptake sites may be down-regulated (Bannon et al., 1992).

Extrastriatal changes in advanced Parkinson's disease

No significant changes in anterior cingulate [18F]dopa Kio were observed with SPM or conventional region of interest analysis. This result is perhaps surprising given that post-mortem data have shown a reduction in frontal dopamine and its metabolites in Parkinson's disease (Scatton et al., 1982). One possible explanation could be that our advanced Parkinson's disease group did not have significant cognitive impairment. Alternatively, AADC activity may be maintained in mesocortical areas in advanced Parkinson's disease despite a fall in tyrosine hydroxylase activity and dopamine levels.

In the midbrain region significant decreases in [18F]dopa Kio were demonstrated by region of interest analysis and were shown by SPM to involve both ventral and dorsal regions, in particular, the substantia nigra, ventral tegmentum and the median raphe nuclei.

In the SPM comparison of advanced Parkinson's disease with early hemi-Parkinson's disease (reflecting Parkinson's disease progression) the most significant further [18F]dopa Kio reductions were in the lower dorsal midbrain to the upper dorsal pons regions, corresponding to dorsal raphe nuclei and

locus coeruleus, respectively (see Fig. 3C). Surprisingly, these reductions were much more significant than the further striatal losses and suggest marked monoaminergic rather than simply dopaminergic neuronal degeneration in Parkinson's disease. Their exact role in the clinical development and late complications of Parkinson's disease, e.g. motor fluctuations and freezing, and problems of long-term levodopa therapy, e.g. dyskinaesia, is unknown.

Striatal changes in Parkinson's disease

Employing 3D [18F]dopa-PET, we have demonstrated bilateral putamen dopaminergic dysfunction in early hemi-Parkinson's disease (Hoehn and Yahr stage 1). Furthermore, all individual putamen [18F]dopa Kio values contralateral to the clinically unaffected side fell outside the normal range. This contrasts with our previous 2D [18F]dopa-PET study of early hemi-Parkinson's disease where most corresponding putamen [18F]dopa Kio values were within the normal range (Morrish et al., 1995). In this study, we used a higher resolution 3D PET scanner with greater sensitivity and improved image quality. This resulted in an ~40% increase in normal striatal [18F]dopa Kio values and a wider separation from the Parkinson's disease range suggesting normal [18F]dopa Kio values were previously underestimated (Rakshi et al., 1996). Therefore, the conclusion, based on the results of the earlier 2D study, that the ipsilateral putamen in early hemi-Parkinson's disease is not significantly affected, is no longer supported. With further advances in PET technology, we may demonstrate even greater differences.

In Parkinson's disease, involvement but relative sparing of caudate dopaminergic function is well recognized; however, it is still unclear when and how it develops. With SPM, we detected a significant region of reduced [18F]dopa uptake in the dorsal body of the right caudate in early left hemi-Parkinson's disease. This suggests preferential targeting of the body as well as head of caudate in early Parkinson's disease. In support of our finding, a post-mortem study has demonstrated severe dopamine depletion in dorsorostral caudate regions comparable with dorsocaudal putamen losses in established Parkinson's disease (Kish et al., 1988).

In advanced bilateral Parkinson's disease, both SPM and region of interest analyses demonstrated greater and more extensive loss of [18F]dopa uptake throughout the putamen and caudate, compared with early hemi-Parkinson's disease. The differences in mean symptom duration and mean motor UPDRS scores between the early and severe Parkinson's disease patients were ~10 years and ~30 points, respectively. One might then have expected a greater difference in mean putamen and caudate [18F]dopa Kios between the two Parkinson's disease groups from region of interest analysis. However, the contralateral putamen in early hemi-Parkinson's disease has already lost 65% of the normal level of [18F]dopa uptake with our 3D measure and, therefore, further smaller reductions in [18F]dopa Kio are likely to be associated with large increases in motor disability. The relationship between motor UPDRS and [18F]dopa Kio is, therefore, likely to be non-linear, possibly exponential.

Conclusion

With recent advances in PET technology, [18F]dopa-PET has become more sensitive at detecting extrastriatal and pre-clinical changes in Parkinson's disease. Applying SPM and region of interest analysis to 3D [18F]dopa-PET data we have identified patterns of subregional dopaminergic dysfunction in cingulate, brainstem and striatal regions in Parkinson's disease. Our study suggests increased AADC activity in anterior cingulate and dorsal midbrain regions in early Parkinson's disease and significant serotonergic and noradrenergic as well as dopaminergic neuronal degeneration within the brainstem in advanced Parkinson's disease, in addition to the known striatal dopaminergic changes. Parkinson's disease, therefore, is most likely a monominergic neurodegenerative disorder. This may have important implications for future treatment strategies.

Acknowledgments

We would like to thank the chemists, Nigel Steel and Maria Constantinou for producing [18F]dopa; the radiographers, David Griffiths, Hope McDevitt and Andrew Blyth for assisting in scanning; and Paul Grasby, Jolanta Opacka-Juffry, Stavia Blunt and Sean Spence, of the MRC Cyclotron Unit, for their helpful comments and suggestions. J.S.R is supported by a grant from Smithkline Beecham, UK.

References

View Abstract