Relationship of Striatal Dopamine Synthesis Capacity to Age and Cognition

Subjects.

Included in the study were 21 older (57–85; mean, 66.9) and 16 younger (20–30; mean, 22.8) cognitively intact adults (Table 1). Subjects were recruited via newspaper ads, flyers, and online postings. We excluded volunteers who were current smokers, hypertensive, or taking medications that could affect cognition or FMT signal. We also excluded all who were unable to undergo magnetic resonance imaging (MRI) or PET scanning for safety reasons or who had a history of neurological or psychological disorders, major systemic disease, drug or alcohol abuse, or serious head injury. Each subject underwent a full neuropsychiatric battery including tests of executive function, working memory, language, episodic verbal and spatial memory, motor response, and general cognitive function. All had a Mini Mental State Exam (Folstein et al., 1975) score of at least 26. Each scored low average or better for their age on the Wechsler Memory Scale (WMS) Auditory Immediate (Wechsler, 1987) age-adjusted summed score (which includes WMS Logical Memory Immediate score and WMS Verbal Paired Associates Immediate score), and on the Wechsler Adult Intelligence Digit Span (Wechsler, 1981) age-adjusted summed score (which includes Digit Span Forward and Digit Span Backward). No subject had a Geriatric Depression Rating (21 item) score >10 (Yesavage et al., 1982). All subjects provided written informed consent before enrolling in the study. Of the 21 older subjects in this study, 20 were reported on previously (Landau et al., 2008). Of the subjects included in the previous study, three were excluded from the current study: one whose PET data were unusable as previously reported, one for whom nearly all of the neuropsychological testing scores were unavailable, and one whose structural MRI scan was not useable for the partial volume correction (PVC) analysis performed here.

Frontal lobe function composite score.

Past studies have demonstrated that performance on certain tests thought to depend on frontal lobe function, such as working memory (Mehta et al., 2001; Erixon-Lindroth et al., 2005; Alfimova et al., 2007), verbal fluency (Erixon-Lindroth et al., 2005; Alfimova et al., 2007), and executive function (Vernaleken et al., 2007b; Lane et al., 2008), may be modulated by the dopamine system in healthy adults.

Our neuropsychological battery contained seven tests of frontal lobe function: Listening Span (Daneman and Carpenter, 1980) and Digit Span (Wechsler, 1981), which measure working memory; Controlled Oral Word Association (FAS) (Benton and Hamsher, 1989) and Category Fluency, which measure verbal fluency; and Stroop C Interference (Stroop, 1935), Wisconsin Card Sort Test (WCST) (Heaton, 1993), and Trail Making (Trails B-A) (Reitan, 1958), which measure executive function. Trails B-A scores were not available for two older subjects.

To evaluate the relationship between dopamine synthesis capacity and frontal lobe function, we created a composite cognitive score. Using a composite score derived from several similar test scores both reduces measure variability that might be associated with any one score, and eliminates the need to control for multiple comparisons. To identify tests measuring similar aspects of frontal lobe function, we prepared a covariance matrix containing scores on each test across all subjects. Five tests emerged that were highly correlated with at least three of the other test scores using a Pearson correlation (p < 0.05): Listening Span total recall, FAS, Category Fluency total, Stroop C Interference, and Trails B-A (Table 1). Scores on these tests were normalized as Z scores within the sample. Because a high score on Trails B-A represented worse cognition, whereas a high score on the other four tests represented better cognition, the signs on the Z scores for Trails B-A were reversed before the five scores were averaged to create a composite cognitive score. The composite cognitive score was significantly correlated with each of the five established tests of frontal lobe function from which it was derived (R² values ranged from 0.54 to 0.64), suggesting that it was a valid measure of frontal lobe-dependent cognitive function. However, although these tests are all thought to measure frontal lobe cognitive function, it is important to note that the tasks are complex and also require abilities such as sensory and motor function that may rely on brain regions other than those thought to subserve the frontal lobe-dependent cognitive function we are trying to measure. Furthermore, the timescale of PET measurements (minutes to hours) is very different from the second to millisecond timescales involved in these cognitive processes.

PET data acquisition.

FMT was synthesized at the Lawrence Berkeley National Laboratory as described previously (VanBrocklin et al., 2004). FMT is a well established PET tracer that has been used extensively to study animal models of aging (DeJesus et al., 2001; Eberling et al., 2002) and cognitive function (Cools et al., 2008; Landau et al., 2008), and to measure the effects of gene therapy related to AADC function (Eberling et al., 2008). Details on enzyme specificity, kinetics, and metabolism have been described previously (Srinivasan and Awapara, 1978; DeJesus et al., 1997; Jordan et al., 1997). Radiochemical synthesis for this study produced specific activities of ∼1000–2000 mCi/mmol; thus, a typical injection of 3 mCi for a PET imaging experiment would result in a mass dose of the compound of ∼3 μmol, which is unlikely to be pharmacologically active.

Subjects underwent PET scanning on a Siemens ECAT EXACT HR PET scanner in three-dimensional acquisition mode with retractable septae. Forty-seven parallel slices (4 mm thick; 3.6 mm in-plane voxel size) were acquired.

Subjects ingested 2.5 mg/kg carbidopa ∼60 min before FMT injection. They were positioned on the scanner bed with a pillow and padding to comfortably restrict head motion. After a 10 min transmission scan obtained for attenuation correction, ∼2.5–3.0 mCi of FMT were injected as a bolus into an antecubital vein and a dynamic acquisition sequence in three-dimensional mode was obtained: 4 × 1 min, 3 × 2 min, 3 × 3 min, and 14 × 5 min for a total of 89 min of scan time.

We reconstructed data using an OSEM (ordered subset expectation maximization) algorithm with weighted attenuation, an image size of 256 × 256, and six iterations with 16 subsets. A Gaussian filter with 6 mm full width at half-maximum was applied, with a scatter correction. Images were evaluated for subject motion and adequacy of statistical counts.

MRI scanning.

Within 6 months of each PET scan, an MRI scan was obtained using a Varian Unity/Inova whole-body 4.0 T scanner with a TEM send-and-receive head coil (MR Instruments). MPFLASH (magnetization-prepared fast low angle shot) T1-weighted volumetric scans were acquired for each subject (repetition time, 9 ms; echo time, 4.8 ms; field of view, 22.4 × 22.4 × 19.8 cm; matrix size, 256 × 256 × 128; resolution, 0.875 × 0.875 × 1.54 mm) to be used for region of interest (ROI) demarcation and brain matter masking for PVC of the PET data.

MRI analysis.

Bias was removed from each MRI scan using the “FSL Automatic Segmentation Tool” (FAST) from version 3.3 of Oxford Centre for Functional MRI of the Brain “(FMRIB) Software Library” (FSL) (Zhang et al., 2001). Nonlinear noise was then reduced using SUSAN (“Smallest Univalue Segment Assimilating Nucleus”) of FSL (Smith, 1992). A mask of MRI scan intracranial matter was automatically generated using the BET (“Brain Extraction Tool”) of FSL (Smith, 2002). We then used “FSL view” to manually refine individual intracranial masks, which were next applied to the MRI scans to exclude voxels that did not represent brain matter or CSF. FAST was used to segment the skull-stripped brains into brain matter versus CSF. A binary mask of the brain matter alone was created for use in the PVC of the PET data. In addition, we calculated the whole brain matter percentage for each subject, defined as the volume of the brain matter mask divided by the intracranial volume.

PET reference region.

Cerebellar masks were drawn in MRI native space on the coronal plane. The masks comprised two-thirds of the slices posterior to the last slice in which the superior cerebellar peduncles stretched superiorly. Including only these most posterior slices ensured that FMT signal from the substantia nigra and ventral tegmental area did not contaminate the reference region. The cerebellum was segmented automatically using FAST to exclude large white matter tracts. To determine whether FMT signal in the cerebellum contributed to our results, for each subject we calculated the average raw FMT cerebellar signal during the first, last, and center PET frames that were included in the final analysis. We compared values at each of the three time points across age groups using a two-tailed Mann–Whitney test. The older and younger groups did not have significantly different cerebellar FMT signal (frame 12, p = 0.82; frame 18, p = 0.61; frame 24, p = 0.30), signifying that choice of reference region did not influence our results.

PET data analysis.

To control for between-frame subject motion, we used six parameter (rigid body) realignment algorithms in Statistical Parametric Mapping (SPM2) (Wellcome Department of Cognitive Neurology, University College London, London, UK). Each PET frame was realigned to the summed image of the first 10 frames. We then used SPM2 to realign the individual MRI scans (along with their accompanying cerebellar gray matter masks and their brain matter masks) to their respective summed PET images, which included considerable anatomic detail from early frames. Next, K_i images were generated using a graphical analysis method for irreversible binding (Patlak and Blasberg, 1985). The approach used a simplified reference tissue model (Lammertsma and Hume, 1996), producing a K_i image that was scaled to the volume of distribution in the reference tissue (cerebellum); we used the term K_i for simplicity. Scaled K_i images were created from PET frames corresponding to 24 through 89 min, using the cerebellar gray matter masks as the reference regions.

Partial volume correction.

The accuracy of PET tracer measurements depends on the size of the object being measured relative to the effective resolution (point spread function) of the scanner. Apparent isotope concentrations are depressed for smaller items imaged in scanners with lower resolutions (Hoffman et al., 1979), a phenomenon known as partial volume error. These small items include image voxels that are composed partly of brain tissue (where a radiotracer signal is expected), and partly of CSF (where signal is not expected). Because of partial volume errors at brain/nonbrain boundaries, subjects with greater cortical atrophy or larger surface areas bordering on CSF typically demonstrate apparent PET signal attenuation, which may confound radiotracer parameter estimates. Correction for partial volume errors is necessary to render the data comparable (Tanna et al., 1991; Meltzer et al., 1996; Yanase et al., 2005; Park et al., 2006), especially when brain atrophy associated with aging may be present.

Evidence of the need for PVC may be found in several past in vivo studies of striatal dopamine synthesis having conflicting results in humans and nonhuman primates. Some of these studies have found age-related decreases in striatal PET signal using FDOPA or [¹¹C]DOPA as a PET tracer (Martin et al., 1989; Bhatt et al., 1991; Cordes et al., 1994; DeJesus et al., 2001; Harada et al., 2002; Ota et al., 2006), whereas others found no age effect (Sawle et al., 1990; Eidelberg et al., 1993; Ishikawa et al., 1996b). Some of these differences seem to be methodological. One PET study evaluated how differences in ROI size can influence the results of the FDOPA signal–age relationship. Specifically, striatal FDOPA signal declined significantly with age when PET-drawn ROIs that encompassed the full striatum were used. Conversely, when “small ROIs” composed of small circles throughout the center of the striatum were used, there was no significant effect of striatal FDOPA signal with age (Vingerhoets et al., 1994). One possible explanation for these contrasting results is that, using full striatal ROIs, there is a greater likelihood of partial voluming, which occurs in voxels that border on CSF or white matter. Such a partial volume effect would depress apparent PET signal more in atrophied brains, in which border voxels would compose a larger percentage of the ROI. Because brain atrophy is frequently associated with age (Raz et al., 1995), using full ROIs that have not been corrected for partial volume effects may result in an atrophy-related apparent decrease of PET signal with age and also may affect signal related to developmental striatal volume changes in young adults (Sowell et al., 1999; Bennett and Baird, 2006).

The current study used full striatal ROIs. However, unlike most previous studies, our FMT PET images were partial volume corrected to adjust for signal loss related to variable structure size. PVC was performed using a two-compartment model as described previously (Meltzer et al., 1999). Briefly, the regionally defined scanner point spread function specific to reconstructed data from our PET scanner was convolved with the binary brain matter mask. The corrected PET scan was then calculated as the observed PET signal divided by the convolved brain matter mask, which represented the percentage contribution of brain matter to signal in each voxel (Meltzer et al., 1999). All FMT signal values referenced in Results and Discussion are partial volume corrected unless otherwise noted.

Regions of interest.

Dorsal caudate (DCA) and dorsal putamen (DPUT) were drawn on the individual MRI scans using “FSL view.” All were drawn using guidelines outlined previously (Mawlawi et al., 2001). After individual MRI scans were coregistered to the corresponding PET images, an average intensity value for each ROI was calculated using the partial volume-corrected K_i image. This average intensity value represented the partial volume-corrected K_i for that ROI. The volumes of the ROIs also were calculated using FSL and were compared with FMT signal.

Depending on the extent to which variations in FMT signal within a brain region exist, the value from an ROI drawn on PET [as in some previous FMT studies (Cools et al., 2008; Landau et al., 2008)] and one drawn on MRI may be quite different. Conceptually, a PET-drawn ROI is one whose PET intensity has been visually thresholded such that only the strongest signal for each subject is included, whereas MRI-drawn ROIs are blind to PET intensity. As such, a PET-drawn ROI will tend to be smaller and will measure PET peak intensity. PET peak intensity will contribute to signal from an MRI-drawn ROI, but the extent of PET signal within the anatomical region will also contribute to the signal. If the percentage of a region having high PET signal were to change with age (e.g., a greater or lesser proportion of cells in a region contained detectable levels of enzyme activity), only an ROI drawn on MRI would identify that change. Although both methods are valid, they emphasize different aspects of the PET signal. MRI-drawn ROIs are necessary for unbiased results when partial volume correction is performed, as in the current study.

Statistics.

For comparisons between two groups or two variables whose data were not normally distributed according to a Jarque–Bera test of normality, we used nonparametric tests: the Spearman ρ rank correlation and the normal approximation of the two-tailed Mann–Whitney test. These tests were used for all comparisons unless otherwise noted. For normally distributed data, we used Pearson correlations or two-tailed unpaired t tests. To adjust for age in some comparisons, we used Spearman partial correlations (Schemper 1991). A Spearman partial correlation used here adjusts for age both in the dependent and independent variables in the same way that a Pearson multiple regression would. However, unlike a Pearson multiple regression, this test minimizes the effect of outliers if they exist and is valid even if the data are not normally distributed.