Skip to main content
NCBI home page
Journal of Cerebral Blood Flow & Metabolism logoLink to Journal of Cerebral Blood Flow & Metabolism
. 2014 Nov 5;35(1):157–165. doi: 10.1038/jcbfm.2014.194

Automated reference region extraction and population-based input function for brain [11C]TMSX PET image analyses

Eero Rissanen 1,2,*, Jouni Tuisku 1, Pauliina Luoto 1, Eveliina Arponen 1, Jarkko Johansson 1, Vesa Oikonen 1, Riitta Parkkola 3, Laura Airas 2, Juha O Rinne 1,2
PMCID: PMC4294409  PMID: 25370856

Abstract

[11C]TMSX ([7-N-methyl-11C]-(E)-8-(3,4,5-trimethoxystyryl)-1,3,7-trimethylxanthine) is a selective adenosine A2A receptor (A2AR) radioligand. In the central nervous system (CNS), A2AR are linked to dopamine D2 receptor function in striatum, but they are also important modulators of inflammation. The golden standard for kinetic modeling of brain [11C]TMSX positron emission tomography (PET) is to obtain arterial input function via arterial blood sampling. However, this method is laborious, prone to errors and unpleasant for study subjects. The aim of this work was to evaluate alternative input function acquisition methods for brain [11C]TMSX PET imaging. First, a noninvasive, automated method for the extraction of gray matter reference region using supervised clustering (SCgm) was developed. Second, a method for obtaining a population-based arterial input function (PBIF) was implemented. These methods were created using data from 28 study subjects (7 healthy controls, 12 multiple sclerosis patients, and 9 patients with Parkinson's disease). The results with PBIF correlated well with original plasma input, and the SCgm yielded similar results compared with cerebellum as a reference region. The clustering method for extracting reference region and the population-based approach for acquiring input for dynamic [11C]TMSX brain PET image analyses appear to be feasible and robust methods, that can be applied in patients with CNS pathology.

Keywords: A2AR, input function, PET, supervised clustering, [11C]TMSX

Introduction

[11C]TMSX ([7-N-methyl-11C]-(E)-8-(3,4,5-trimethoxystyryl)-1,3,7-trimethylxanthine) is a highly selective adenosine A2A receptor (A2AR) radioligand,1 that has been used in in vivo detection of human brain A2AR using positron emission tomography (PET).2, 3, 4, 5, 6, 7 In healthy humans, the A2AR are enriched in the striatum, with lower levels of expression in the cortical gray matter (GM) and cerebellum, and negligible expression in the white matter (WM).4, 8 In Parkinson's disease (PD), [11C]TMSX PET has shown a reduction of A2AR in the putamen contralateral to the clinically affected side in the early phase of the disease, followed by an increase after dopaminergic treatment.5 Adenosine A2A receptor antagonists are currently in active development for treatment of PD.9, 10, 11 In addition to their role in regulating neurotransmission in dopaminergic pathways, A2AR are essential effectors in inflammatory conditions in the central nervous system: A2AR are upregulated in the areas of inflammation and tissue damage, and possibly provide an endogenous route to limit inflammation.6 Moreover, in our recent study, we have shown an increase of A2AR in the WM and thalami of patients with secondary progressive multiple sclerosis (SPMS) using [11C]TMSX PET.7

Analyzing the binding of a radioligand in neurodegenerative diseases with wide spread pathologic assessment—as in multiple sclerosis (MS)—is not without challenges: for the estimation of specific binding (either as distribution volume ratio, DVR; or binding potential, BPND) in a target area, an anatomically distinct reference region, that is both assumingly free of disease pathologic assessment, and has negligible specific binding of the radioligand, cannot always be defined. Previous studies with A2AR radioligands have used the centrum semiovale,3, 4 cerebral cortex,5, 6 and the cerebellum12 as reference regions, which are not always feasible reference regions in MS patients because of possible disease pathologic assessment and thus specific binding occurring in these regions. The comparability of different studies with [11C]TMSX PET is so far hindered by the lack of optimal reference region, and the heterogeneity of the data-modeling methods. Moreover, despite being the golden standard for acquiring the plasma-input function for the kinetic modeling of neuroreceptor imaging, the evaluation of radiotracer activity and metabolism via arterial blood sampling during imaging is not only uncomfortable for the study subjects, but also susceptible for errors, and often logistically demanding.

To provide alternative input function acquisition methods for brain [11C]TMSX PET imaging, we implemented and compared two different approaches: first, a noninvasive, automated method for the extraction of GM reference region using supervised clustering, and second, a method for obtaining a population-based arterial input function. These methods were compared with the methodology using original, metabolite-corrected arterial input and cerebellum reference region. Corresponding clustering methods have previously been validated for [11C]PK1119513, 14, 15 and [11C]PIB16 brain PET image analyses, and population-based input as an option for individual arterial sampling has successfully been proposed for brain [11C]rolipram17 and [18F]FMPEP-d218 PET imaging, but neither of these methods have been reported for A2AR radioligands to date. In addition, we evaluated the ability of these methods to detect neuroinflammatory changes in the brain of SPMS patients compared with healthy controls.7

Materials and Methods

The study protocol was approved by the Ethical Committee of the Hospital District of Southwestern Finland, and the study was performed according to the ethical principles of the declaration of Helsinki.

For the development of the supervised cluster algorithm (SCA) and population-based input function (PBIF) for brain [11C]TMSX PET imaging, we used all the available [11C]TMSX PET imaging data with arterial blood samples from clinical study cohorts in our center consisting of 7 healthy controls (2 men and 5 women; mean age 49.3 years), 12 patients with MS, of whom 8 patients with SPMS (2 men and 6 women; mean age 49.4 years) and 4 with relapsing-remitting multiple sclerosis (RRMS; 1 man and 3 women; mean age 27.8), and 9 PD patients (4 men, 5 women; mean age 64.7). The main inclusion criteria for the MS patients was definite diagnosis of MS according to Poser et al19 or revised McDonald20 criteria, and for SPMS patients a disease duration of >5 years, established progression of the disease for >2 years, and no disease-modifying treatment for at least 3 months before enrollment were required. The RRMS group composed of newly diagnosed (within 12 months) patients with no initiated immunomodulatory treatment before evaluation. The PD group included patients with clinically defined diagnosis of idiopathic PD with no dyskinesia. The main exclusion criteria were an active neurologic disease or autoimmune disease other than MS or PD, or another significant comorbidity. The control subjects were healthy individuals with no known neurologic symptoms or diseases. Age limit for all study subjects was 18 to 75 years.

[11C]TMSX Radioligand Production

The radiochemical synthesis of [11C]TMSX, and the consequent arterial blood activity and metabolite analyses were performed according to the methods described in our previous work.7 The mean radiochemical purity of [11C]TMSX at the time of the ligand injection was 97.6%, and the mean specific radioactivity was 425.2 MBq/nmol (s.d. 160.17) corresponding to a mean mass of 0.5 μg (s.d. 0.2 μg). The mean (s.d.) injected dose of [11C]TMSX was 470 MBq (46) for controls, 459 MBq (75) for MS patients, and 483 MBq (26) for PD patients, with no significant differences between groups.

[11C]TMSX PET and Magnetic Resonance Imaging, Data Processing and Nonmetabolized Tracer Fraction Analyses

A dynamic 60-minute PET scan was performed using high resolution research tomograph scanner (CTI/Siemens, Knoxville, TN, USA)21 starting at the time of slow intravenous bolus injection of [11C]TMSX. The [11C]TMSX PET and magnetic resonance imaging (MRI), data reconstruction protocols, and the methodology for coregistrating the PET and MR images have been described in detail previously.7 The procedures used for the segmentation of WM and GM were performed according to the methods explained in our recent study on SPMS using [11C]PK11195.22 Individual binary masks of the cerebellum and cerebral cortical GM (neocortex; composed of bilateral frontal, parietal, occipital, and temporal regions of interest (ROIs) masked with the GM segment) were created using the automated anatomic labeling template after the procedure described in our former study.22 The corresponding time-activity curves (TACs) for cerebellum and neocortex masks, in addition to the manually delineated normal-appearing WM (NAWM), striatum and thalamus ROIs were extracted from dynamic PET data using the in-house developed software (Carimas 2.4, Turku PET Centre, Turku, Finland). The metabolite and activity analyses of [11C]TMSX from the arterial blood were performed according to the same methodology as in our previous study.7

Supervised Clustering of Reference Region from Brain [11C]TMSX PET Images

The automated reference region extraction from dynamic brain [11C]TMSX images was based on the methodology used in supervised clustering algorithm for brain [11C]PK11195 PET image analyses (SuperPK software; Imperial College, London, UK).14 Briefly, a set of predefined kinetic tissue class TACs describing the average kinetics of the radioligand in different types of brain tissue is first formed. Then a weighted linear combination of tissue class TACs is fitted to each voxelwise TAC of the dynamic PET image, which is standardised using frame-wise means and s.d.15 This procedure provides a weight coefficient for each PET image voxel representing the contribution of each kinetic class to the corresponding TAC.

For [11C]TMSX, four kinetic tissue class TACs were defined characterizing the average GM, WM, high-specific binding (HSB), and vasculature binding. The segmented GM and WM MRI volumes of healthy controls were first transformed into binary masks (thresholded using a value of 0.9) and then used for extracting WM and GM TACs from the PET image. The TACs representing HSB and vasculature binding were obtained from the PET image by using manually delineated ROIs drawn to the thalami of SPMS patients and venous sinuses of healthy controls, respectively. For evaluating the effect of different HSB class to clustering results, we derived another HSB class from ROIs drawn to the anterior putamen of the PD patients showing HSB. This alternative HSB class yielded similar results with no significant differences when compared with the MS HSB class (data not shown), and thus, only the results obtained using MS HSB class are regarded here. Finally, an individual set of kinetic class TACs were composed for each subject in a leave-one-out procedure by first standardising all class TACs to match the aforementioned standardised PET data, and then leaving out the specific subject before calculating the average standardised TAC for each class.

For reducing the possible effect of misclassification in the clustering process, the following strategy was applied: we first calculated for all the gray map voxels the ratio of the gray coefficient to the sum of all the weight coefficients. Then the gray coefficient map was thresholded to contain values above 0.25 of the gray ratio map.16 In addition, we excluded all those gray coefficient map voxels that had also HSB and vascular binding coefficients above zero. Finally, the remaining gray map coefficients were used to calculate a weighted average of the dynamic PET image, hence creating the clustered SCA-based gray matter reference region (SCgm) TAC.

Population-Based Input Function for Brain [11C]TMSX PET Image Analyses

The population-based plasma-input function was created using data from all 28 study subjects consisting of healthy volunteers (n=7), MS (n=12), and PD (n=9) patients. Each metabolite-corrected plasma TAC was standardised by the injected activity and the subject's body surface area using the Mosteller formula.23 Each standardised curve was then interpolated to have the same time points, shifted to the average peak time, and finally averaged groupwise with a similar leave-one-out procedure as described in the previous section. Although the variation of the arterial data between different cohorts was small, we noted differences in the peak heights of the group-specific average PBIFs (Supplementary Figure 1). Thus, the group-specific average PBIFs were preferred over a single-shared PBIF for all subgroups. The resulting group-specific average PBIFs were scaled individually to its original level with two alternative methods: (1) by using body surface area (noninvasive individually scaled PBIF; PBIFnis) and (2) by using an average of two metabolite-corrected arterial blood samples at 9 and 23 minutes that were tested to give the highest correlation with plasma area under the curve in these data (invasive individually scaled PBIF; PBIFis24).

Evaluation of the Specific Binding of [11C]TMSX

[11C]TMSX binding was evaluated using VT (distribution volume) and DVR for validating the PBIF inputs as an option for original arterial plasma input (OPIF), and SCgm as an alternative reference region to ROI-based cerebellum. Logan's method (from 10 to 60 minutes) with OPIF and both PBIFs were used for VT estimation, and similarly the reference region inputs from both cerebellum and SCgm were used for estimating DVR within the same time interval. The range of 10 to 60 minutes was selected from different timings, by considering the plasma-input model fits in the thalamus. For each subject the initial equilibrating time (t*) giving the smallest coefficient of variation for the VT estimate was selected and then the median of the resulting times was chosen.

First, the VT of the cerebellum and SCgm were compared using the data from healthy control subjects and patients with MS and PD. Second, an indirect estimate of DVR was evaluated in four target ROIs (thalamus, striatum, NAWM, and combined ROI of the neocortex) of healthy controls and SPMS patients using the target to reference VT ratio calculated with OPIF, PBIFnis, and PBIFis inputs. The VT results of this age- and sex-matched study have previously been published in our earlier study using OPIF.7 These results were also compared with direct DVR estimates from reference-input Logan25 (RefLog) using the aforementioned reference region inputs.

Statistical Analyses

The statistical analyses of the results were executed with SPSS software (SPSS Statistics, version 21; IBM, New York, NY, USA). The normality of the variable distributions was evaluated using Shapiro–Wilk test. The group characteristics were reported as mean and s.d. values if the distribution of the variables was normal, otherwise median and interquartile range was used. Because of few study subjects per group, and the nonnormal distribution of some of the variables, nonparametric Mann–Whitney U-test for the comparison between study groups and paired-sample Wilcoxon signed rank test for the comparison of VT and DVR evaluation methods within groups were used. For correlational analyses, either Pearson's or Spearman's test was used depending on the linearity, distribution and number of observations. Bland–Altman's method was applied to evaluate the agreement of the VT estimates between the original arterial input and PBIF methods.

Results

[11C]TMSX Radioligand Metabolism

The metabolism of [11C]TMSX remained slow throughout the imaging with the mean fraction of nonmetabolized ligand being 93.4% (s.d. 4.7) at the 50 minute's time point, and no significant differences between the groups were noted (mean (s.d.) percentages in controls 90.5 (6.7); in MS group 92.7 (4.6) and in PD group 95.5 (2.8)), even though the mean in PD group was five percentage points higher compared with controls.

Reference Region Extraction using Supervised Cluster Algorithm

The TACs of the four tissue classes for [11C]TMSX SCA reference region extraction are shown in Figure 1. The resulting coefficient maps of the SCgm showed similar distribution in all study groups with majority of the voxels contributing to the reference cluster situated in the cerebral cortices and to lesser extent in the cerebellar cortices (Figure 2). The TACs in the SCgm reference region showed lower uptake in the latter part of the curve when compared with the cerebellum. The average TACs of standardised SCgm and cerebellum are illustrated in Figure 3. The [11C]TMSX VT of SCgm and ROI-based cerebellum reference, estimated using both the OPIFnis and PBIFis (see below) are shown in Figure 4. These VT results are discussed in more detail in the following chapter.

Figure 1.

Figure 1

Open in a new tab

The time-activity curves (TACs) representing the average kinetics of the four predefined classes for [11C]TMSX ([7-N-methyl-11C]-(E)-8-(3,4,5-trimethoxystyryl)-1,3,7-trimethylxanthine) supervised cluster algorithm: gray and white matter, high-specific binding (HSB), and vasculature binding. (A) The kinetic classes are derived from the data, which are standardised by subtracting from each time point the frame average and dividing by the frame s.d. (B) The kinetic class TACs are illustrated using the standardised uptake values.

Figure 2.

Figure 2

Open in a new tab

Examples of the SCA gray matter weight coefficient maps in (A) a healthy control (male, 61 years), (B) SPMS patient (female, 39 years), and (C) PD patient (male, 75 years). The scaled color bars indicate the voxelwise ratio of the gray coefficient value to the sum of all weight coefficients, calculated for each voxel. PD, Parkinson's disease; SCA, supervised cluster algorithm; SPMS, secondary progressive multiple sclerosis.

Figure 3.

Figure 3

Open in a new tab

Average cerebellum and mean SCA-based gray matter reference region (SCgm) standardised time-activity curves of all 28 subjects.

Figure 4.

Figure 4

Open in a new tab

Comparison of the regional VT (correlation of distribution volume) results in the cerebellum (Cer) and clustered gray matter (SCgm) regions estimated using the Logan's method with original arterial plasma input (OPIF), noninvasively scaled population-based arterial input function (PBIFnis) and invasively scaled PBIF (PBIFis) in all subgroups visualized as boxplots showing median, minimum and maximum values and interquartile ranges. MS, multiple sclerosis; PD, Parkinson's disease.

Distribution Volumes of SCgm and Cerebellum Reference Regions

The VT of SCA-based SCgm and ROI-based cerebellum reference regions were compared in all study subjects by using OPIF and both PBIFs as input for the Logan's method. The noninvasive and invasive PBIFs yielded similar VT estimates with no statistically significant differences between the methods, and with moderate to good correlations in the cerebellum and SCgm when compared with the results of OPIF. For PBIFnis, Pearson's correlations to OPIF were 0.76 (P<0.001) in the cerebellum and 0.73 (P<0.001) in SCgm, and for PBIFis the correlations were 0.97 (P<0.001) in the cerebellum and 0.97 (P<0.001) in SCgm. The VT estimates calculated using PBIFis showed a slightly better concordance with the original plasma input than PBIFnis, the median (interquartile range) VT ratio to OPIF in both cerebellum and SCgm for all study subjects being 0.99 (0.91 to 1.06) for PBIFis and 0.99 (0.97 to 1.04) for PBIFnis. The VT values by subgroup and the correlations between the methods in all study subjects are shown in Figures 4 and 5, respectively.

Figure 5.

Figure 5

Open in a new tab

Correlations of distribution volumes (VTs) in the cerebellum using original arterial plasma input (OPIF) to VTs acquired with (A) noninvasively scaled population-based input function (PBIFnis) and (B) invasively scaled population-based input function (PBIFis) in all study subjects (healthy controls labeled with open squares, MS (multiple sclerosis) patients with open circles and PD (Parkinson's disease) patients with crosses), with Bland–Altman plots (on the left) showing relative differences in VT (relative to OPIF) and scatter plot (on the right) visualizing the correlations between the two methods.

In the PD group, the VT values were generally lower when compared with the healthy controls (Figure 4): in SCgm a significant difference was found with all inputs (OPIF: P=0.023; PBIFnis: P=0.023; and PBIFis: P=0.032), whereas in the cerebellum the difference was significant only with PBIFis (P=0.042), although a similar but nonsignificant trend was noted also with OPIF and PBIFnis (P=0.91 and 0.71, respectively). When comparing the healthy controls and MS patients, no significant differences between the groups were noted. However, when analyzing MS groups separately, the RRMS patients had significantly lower VT in SCgm with OPIF as compared with healthy controls (RRMS versus control group, median (interquartile range): 0.54 (0.51 to 0.57) versus 0.65 (0.60 to 0.70); P=0.042), whereas with other inputs and in the cerebellum no differences were found. No statistically significant differences in the reference region VTs were observed when comparing the controls and SPMS patients.

Performance of PBIFs and SCgm in the Evaluation of Specific [11C]TMSX Binding

The performance of PBIFs as input function, and SCgm as reference region in evaluating the specific binding of [11C]TMSX was tested in the clinical data of healthy controls and SPMS patients. The previously published VT results compared well with the results obtained by using PBIFs, although the VT estimates of both PBIFs had a small positive bias when comparing the group median values in different ROIs, whereas when evaluated as the median VT ratio to OPIF, the bias appeared negative for PBIFnis and negligibly positive for PBIFis (Table 1). As a measure of specific binding, DVRs from plasma-input Logan (with OPIF, PBIFnis, or PBIFis as input) and reference-input Logan with both SCgm and cerebellum ROI reference regions were estimated. Both PBIFs and the SCgm reference yielded similar results as the OPIF and the cerebellum reference. The VT results and corresponding DVR values with group differences are shown in Tables 1, and 2, respectively. The correlations of DVR estimated using PBIFnis, PBIFis, and reference Logan (RefLog) to the DVR estimates yielded using OPIF with both reference region options were highly significant in both groups in the neocortex, striatum, and the thalamus ROIs. For these regions the Spearman's correlation coefficients (ρ) were 0.99 to 1.00 (P<0.001 for all comparisons). In NAWM, all correlations to OPIF in the SPMS group were statistically significant (with cerebellum reference: ρ=0.99, P<0.001 for both PBIFnis and PBIFis, and ρ=0.8, P=0.017 for RefLog, and with SCgm reference: ρ=0.86, P=0.007 for PBIFnis and PBIFis, and ρ=0.91, P=0.002 for RefLog). Also, for the control group the correlations in NAWM were significant in all other comparisons (with cerebellum reference: ρ=0.82, P=0.02 for both PBIFnis and PBIFis, and with SCgm reference: ρ=0.93, P=0.003 for PBIFnis, PBIFis and RefLog) except for the borderline finding of ρ=0.75, P=0.05 for RefLog with cerebellum reference. The correlations and their significances are shown in more detail in the Supplementary Table 1.

Table 1. Distribution volumes (mL/cm3; median (CV%)) with Logan using original plasma input and population-based plasma inputs in healthy controls (n=7) and SPMS patients (n=8).

ROI Healthy controls
 SPMS patients
 Group differences, %
  OPIF PBIFnis PBIFis OPIF PBIFnis PBIFis OPIF PBIFnis PBIFis
VT (mL/cm3)
 Cerebellum 0.69 (11.6) 0.74 (22.6) 0.70 (10.5) 0.69 (17.0) 0.73 (12.3) 0.72 (13.2) −0.2 −0.6 3.5
 Biasa   0.92 (0.20) 1.00 (0.06)   0.99 (0.11) 1.01 (0.10)      
 SCgm 0.65 (15.2) 0.69 (18.3) 0.66 (6.0) 0.67 (17.0) 0.70 (11.7) 0.68 (19.1) 2.0 0.8 3.2
 Biasa   0.92 (0.20) 1.00 (0.06)   0.99 (0.11) 1.01 (0.10)      
 Neocortex 0.68 (15.0) 0.70 (17.2) 0.70 (6.1) 0.69 (20.4) 0.71 (13.6) 0.71 (21.8) 2.2 2 1.4
 Biasa   0.92 (0.20) 1.00 (0.06)   0.99 (0.11) 1.01 (0.10)      
 Striatum 1.03 (20.6) 1.11 (18.0) 1.09 (11.2) 1.09 (14.3) 1.14 (11.2) 1.10 (16.5) 5.9 2.8 0.8
 Biasa   0.92 (0.20) 1.00 (0.06)   0.99 (0.11) 1.01 (0.10)      
 Thalamus 0.78 (13.3) 0.83 (12.5) 0.78 (8.8) 0.89 (22.2) 0.92 (15.9) 0.92 (21.7) 13.9 11.2 17.8
 Biasa   0.92 (0.20) 1.00 (0.06)   1.00 (0.11) 1.01 (0.10)      
 NAWM 0.45 (14.2) 0.45 (14.6) 0.44 (16.8) 0.53 (19.4) 0.54 (9.8) 0.53 (12.7) 16.2b 22.2b 18.8
 Biasa   0.94 (0.23) 1.02 (0.04)   0.99 (0.09) 1.02 (0.08)      
Open in a new tab

CV, coefficient of variation; IQR, interquartile range; NAWM, normal-appearing white matter; OPIF, original arterial plasma input; PBIFis, invasively scaled population-based arterial input function; PBIFnis, noninvasively scaled PBIF; SPMS, secondary progressive multiple sclerosis; VT, correlation of distribution volume. CV% expressed here as |IQR|/median × 100.

a

Bias of PBIFnis and PBIFis reported as a median (CV%) VT ratio to OPIF.

b

Mann–Whitney U-test, group differences regarded statistically significant if P<0.05.

Table 2. Distribution volume ratios (median (CV%)) with Logan using different plasma inputs and reference Logan, and using the cerebellum and SCgm clustered gray matter (SCgm) reference regions (rr) in healthy controls (n=7) and SPMS patients (n=8).

ROI Healthy controls
 SPMS patients
 Group differences, %
  OPIF PBIFnis PBIFis RefLog OPIF PBIFnis PBIFis RefLog OPIF PBIFnis PBIFis RefLog
DVR
 Cerebellum rr
  Neocortex 0.96 (6.8) 0.97 (6.7) 0.97 (6.7) 0.96 (6.9) 0.98 (7.4) 0.98 (7.2) 0.98 (7.2) 0.98 (7.2) 2 1.7 1.7 1.8
  Striatum 1.49 (10.0) 1.49 (10.0) 1.49 (10.0) 1.49 (10.2) 1.59 (7.7) 1.59 (7.6) 1.59 (7.6) 1.59 (7.6) 6.3 6.5 6.5 6.5
  Thalamus 1.13 (7.3) 1.13 (7.6) 1.13 (7.6) 1.13 (7.6) 1.28 (9.2) 1.28 (9.3) 1.28 (9.3) 1.28 (9.3) 13.5a 13.6a 13.6a 13.6a
  NAWM 0.6 (17.1) 0.6 (18.3) 0.6 (18.3) 0.6 (17.6) 0.72 (9.1) 0.72 (6.8) 0.72 (6.8) 0.71 (7.1) 20.4 19.8 19.8 19.7
 SCgm rr
  Neocortex 1.02 (4.7) 1.02 (4.7) 1.02 (4.7) 1.02 (4.7) 1.03 (3.2) 1.03 (3.3) 1.03 (3.3) 1.03 (3.3) 0.8 0.8 0.8 0.8
  Striatum 1.67 (6.3) 1.67 (6.6) 1.67 (6.6) 1.67 (6.5) 1.66 (10.6) 1.66 (10.9) 1.66 (10.9) 1.66 (10.7) −0.4 −0.4 −0.3 −0.25
  Thalamus 1.2 (5.0) 1.2 (5.0) 1.19 (5.0) 1.19 (4.7) 1.33 (3.7) 1.33 (4.0) 1.33 (3.9) 1.33 (3.9) 11.1b 11.1b 11.1b 11.2b
  NAWM 0.65 (10.2) 0.68 (8.5) 0.68 (8.5) 0.66 (9.8) 0.75 (12.4) 0.76 (10.4) 0.76 (10.4) 0.77 (8.7) 15.7b 12.4b 12.4b 15.7b
Open in a new tab

CV, coefficient of variation; DVR, distribution volume ratio; IQR, interquartile range; NAWM, normal-appearing white matter; OPIF, original arterial plasma input; PBIFis, invasively scaled population-based arterial input function; PBIFnis, noninvasively scaled PBIF; SPMS, secondary progressive multiple sclerosis. Coefficient of variation (CV%) expressed here as |IQR|/median × 100.

a

Mann–Whitney U-test, group differences regarded statistically significant if P<0.05.

b

if P<0.01.

When comparing the VT values in the cerebellum and SCgm reference regions within the groups and within the different inputs, the VTs were lower in SCgm, with statistically significant differences in the MS group (P=0.06 for OPIF, PBIFnis, and PBIFis). There were also significant and borderline findings in PD group (P=0.04 for PBIFnis, and P=0.05 for OPIF and PBIFnis), and a trend observed in healthy controls (P=0.09 for OPIF and PBIFnis, and P=0.06 for PBIFis) for lower VTs in SCgm than in the cerebellum. These differences were also reflected in the examined DVR values of healthy controls and SPMS patients as higher DVRs with SCgm than with cerebellum (Table 2). The DVR estimates obtained using SCgm had also smaller variation in healthy controls compared with DVR estimates acquired using the cerebellum. For this reason, by using SCgm, the group differences are exposed in NAWM similarly as seen in the OPIF VT results, although the cerebellum DVR results showed a similar trend (P=0.072 for OPIF and PBIFs).

Discussion

Our aim was to evaluate the usability of two input function acquisition methods, that are novel for [11C]TMSX PET brain imaging. The automated method for the extraction of SCgm using supervised clustering was developed to replace the ROI-based method. The population-based arterial input method was implemented to possibly obviate the individual arterial sampling.

The PBIF methods estimated VT with good concordance to original metabolite-corrected plasma-input function, although with invasive scaling of the population-based input with two blood samples, the correlation to original plasma input was stronger and the variation was smaller when compared with the noninvasive scaling using body surface area. Thus, PBIF with invasive scaling can be used instead of arterial plasma-input function, and is preferrable to noninvasive scaling because of its better accuracy in VT estimation. However, when comparing the DVR estimates, the differences between original plasma input and both population-based inputs were negligible. Hence, using noninvasively scaled population-based input also appears feasible, especially because of its noninvasiveness.

The DVR estimates with reference-input Logan and clustered SCgm correlated well with the indirect DVR results from arterial input function and cerebellum reference in healthy controls and MS patients. The DVRs with SCgm were slightly higher than with cerebellum because of the lower reference region VT in SCgm. However, the variance in DVR with SCgm was smaller in healthy controls when compared with the cerebellum. In conclusion, supervised clustering of the SCgm is a robust option for traditional ROI-based method, and with reference-input models provides a noninvasive method for DVR estimation. Moreover, by using these noninvasive DVR estimation methods, we were able to confirm our previous findings of the increased [11C]TMSX binding in the NAWM of SPMS patients compared with healthy controls based on the VT estimates with arterial input function. In addition, the previous borderline finding of the increased binding in thalami of SPMS patients became more pronounced reaching statistical significance

Some caveats, however, need to be taken into account. Generally, PBIFs may be less sensitive at individual level when compared with OPIF, and hence, may not reveal more subtle group differences as well as OPIF. Also, although the differences between group-specific average PBIFs were subtle, using a shared average PBIF for all subject groups would likely add bias in the VT estimates.18 Therefore, we recommend the PBIFs of different patient groups and healthy controls to be derived independently. Even though PBIFnis is free from the methodological and logistical challenges related to OPIF, the scaling in PBIFis is still dependent on few blood samples. However, considerably fewer blood samples are needed for PBIFis than when using OPIF. Consequently, arterialized venous blood might be a feasible, less invasive option in this regard. Supervised clustering of the reference region is equally automated method as atlas ROI-based methods, but it is not dependent on reliable coregistration of the atlas template, which might be unsatisfactory with patients having atrophic brains. However, although no significant differences were observed in the clustered SCgm when an alternative HSB class was used, the results of SCA rely on the data from where the tissue kinetic classes are defined. The different image reconstruction protocols and PET scanner features effect on the shape of the kinetic classes, which might limit the usability of the method to data from different PET scanners.15 Therefore standardising the imaging procedures would make the method more widely applicable.

The methodology used for the evaluation of specific [11C]TMSX PET binding thus far has been very varied. The first reported brain [11C]TMSX PET study on healthy humans using arterial metabolite-corrected input and two-tissue compartmental model showed the lowest VTs in centrum semiovale, whereas the highest values were observed in the putamen, followed by head of caudate, thalamus and cerebellum, and cerebral cortices.3 Because of the low VT in the centrum semiovale, supported by the findings in a postmortem study showing negligibly low A2AR density in corresponding regions,8 centrum semiovale was regarded as the optimal reference region for future studies.3 Also in the following study,4 the VT was reported to be higher in cerebellum than in cerebral cortices, and the same order of magnitude was observed in BPND estimation using the centrum semiovale as the reference region. The rate of specific binding (SB, calculated as BP/VT) of [11C]TMSX was reported to exceed 30% in the cerebellum (53.0%) and cerebral cortex (37.8 in frontal, percentage, 42.1% in temporal, 42.7% in occipital, and 40.6% in parietal lobe) in addition to the high SB in the striatum (caudate nucleus 58.4%, anterior putamen 62.7%, posterior putamen 61.8%). The existence of specific A2AR binding in the cortices and cerebellum has also been proven with another A2AR radioligand, [11C]SCH442416, in primate26 and human27 studies; the administration of selective A2AR agonists (istradefylline and vipadenant, respectively) reduced the binding of the A2AR radioligand also in the cerebellum and cerebral cortex in addition to striatum. However, in later human PET studies, both cerebellum and cerebral cortex have been used as reference regions for the estimation of specific binding of [11C]TMSX5, 6 and [11C]SCH442416.12 It's indeed challenging to find a suitable reference region taken into account the requirements of the reference region: the nondisplaceable distribution volume should be constant across the regions and the region should be devoid of the target receptors. Since the specific binding estimates are dependable on the VT value in the reference region, the results are not fully comparable between different studies using various reference regions. Nevertheless, in our study the voxels representing HSB and vascular binding were additionally excluded from the clustered SCgm, and thus it should contain less specific A2AR binding. The method for defining the ROIs should also be taken into account when comparing the results from different studies. For instance, in the study by Mishina et al,5 reference region was defined by forming a combined ROI of several circular ROIs with 10 mm in diameter placed in frontal, temporal and occipital lobes. It is likely, that using this approach will include also subcortical WM within the ROIs thus affecting (i.e., lowering) the VT values.

For Logan's method, in addition to the standard metabolite-corrected arterial input,3, 7 alternative methods have been proposed for estimating the integrated plasma TAC to be used as input in brain [11C]TMSX PET studies: extracting the plasma-related information using independent component analysis,3 and intersectional searching algorithm with averaging and clustering of PET data to reduce the noise in measured PET data (robust EPISA)28 have been proposed. However, one of the disadvantages of these methods is the inability of intersectional searching algorithm to estimate the absolute integrated plasma TAC value. In addition, plasma-input methods are also affected by the fraction of radioactive metabolites. It has been reported, that using nonmetabolite corrected input underestimates the [11C]TMSX VT by ~5% when compared with metabolite-corrected plasma input.3 When using PBIFs, the correction for metabolites is included in the average population-based TAC, thus reducing the bias. In our data, the correlations between the original metabolite-corrected plasma-input function and the population-based inputs, both with invasive and noninvasive scaling, in VT and DVR were good and the group differences in these estimates were small. The variation in VT with the noninvasive scaling using body surface area was larger than with the invasive scaling using two arterial blood samples, consequently causing a decreased sensitivity to detect changes in VT. However, both methods yielded equivalent DVR estimates with robust correlations to indirect estimates of the original plasma input. Thus, both PBIFs appear as feasible methods.

For another A2AR PET ligand, [11C]SCH442416, presence of an additional vascular component with reversible binding has been proposed based on spectral analyses.29 Similar phenomenon has not been reported for [11C]TMSX in the literature, even though A2AR are known to be expressed on the vascular endothelial wall,30 and therefore it cannot be excluded that vascular binding would occur with [11C]TMSX. In our preliminary analyses with compartmental modeling, a blood volume fraction of 5% to 7% was observed (data not shown), being slightly higher than expected and possibly accounted by vascular binding with faster kinetics. Because the compartmental modeling is sensitive to input function shape, which might not be accurately reproduced by PBIF, we chose to use Logan's method for VT and DVR estimation. The Logan's method is independent of any particular model structure, and it uses the input function area under the curve, which can be more accurately reproduced using PBIFs compared with the actual shape of the original input function.18

Interestingly, there was a tendency for lower VT both in the cerebellum and SCgm in PD and RRMS groups when compared with control and SPMS groups in our study. It has been reported, that the binding of [11C]TMSX in the striatal A2AR does not change with age in healthy subjects, even though A1AR binding decreases with age measured with [11C]MPDX.6 However, the effects of age on the whole body biodistribution of [11C]TMSX are not available. In a whole body radiation dosimetry study in healthy controls with 120 minutes emission scanning, the highest [11C]TMSX uptake was seen in the liver for the whole period, followed by high uptakes in the heart and kidneys in the early phase and in the gallbladder an urinary bladder in later phase.31 Since A2AR densities in the cardiac and skeletal muscles have been reported to be greater in endurance-trained men compared with healthy men,32 it could be postulated that increasing age and physical disability caused by chronic diseases (such as MS and PD), could cause differences in the biodistribution when compared with healthy controls. However, VT estimates should be independent from the plasma clearance (and thus the biodistribution of the radioligand), because the tissue TAC is estimated in relation to the plasma activity. Moreover, the possible effects of age and disability would not solely explain the differences between the groups in our study, since PD patients being the oldest group had VT levels similar to RRMS patients with no significant disability.

It has been reported, that MS patients have significant A2AR upregulation in the peripheral blood lymphocytes compared with healthy controls, and that the upregulation of A2AR in MS was not affected by the disease subtype or the possible immunomodulatory medication used.33 Even though it could be hypothesized, that this might affect the ratio of availability of [11C]TMSX into brain in MS patients compared with healthy controls, it would not explain the observed differences in the reference region VTs between RRMS and SPMS patients in our study. Moreover, the original arterial input representing the free nonmetabolized fraction of the ligand in plasma is not biased by the possible metabolic differences between groups. In this study, age or disease state did not significantly alter the metabolism of the ligand, even though the metabolism rate was slightly lower in the PD group compared with healthy controls. On the other hand, we did not have age- and sex-matched controls for the RRMS and PD patients in our study, and the number of the study subjects, especially in the RRMS group, was relatively small, which limits the interpretation of the results, and the evaluation of DVR estimates in these groups were not deemed justifiable. Consequently, the supervised clustering of brain [11C]TMSX cannot be readily applied for older controls and patients with other neurologic diseases without further studies, even though the method's performance was robust for SPMS patients and healthy controls. Moreover, if PBIF was to be used in studies with matched controls and patients with other central nervous system pathologies, it should be confirmed first, that the disease state itself does not affect the metabolism of the radioligand.

Using the supervised clustering method on acquiring the reference region has several advantages. In addition to saving time on manual ROI drawing, the automated extraction is also free from operator bias. Also, as a result from the supervised classification process, the voxels in the clustered reference region have negligible specific binding, and thus free of disease pathologic assessment regarding the ligand used. In addition, segmentation and masking of the MR images can be used to exclude areas with disease pathologic assessment in MRI. Moreover, the SCgm performed equally well in our study when the DVR estimates were compared with indirect estimates of plasma-input Logan, thus providing a plausible noninvasive method for the estimation of specific [11C]TMSX binding when used with reference-input Logan.

Regarding the location and function of A2AR, the striatal receptors linked to the modulation of dopaminergic transmission have been called the classical or typical A2AR, whereas others (in the cerebral cortex, cerebellum, and thalami) are regarded as the atypical A2AR.4, 34, 35, 36 According to animal studies, the A2AR have also been reported to be associated with the modulation of striatopallidal GABA receptors,37 glutamate receptors in striatum38 and in astrocytes within striatum and cortex,39 and also with microglial activation.40 It can be hypothesized, that the increased A2AR binding in the NAWM and thalami in SPMS might be linked to the activation of microglia and astrocytes associated with the disease, thus providing an alternative route for developing antiinflammatory therapeutic interventions.

Taken together, [11C]TMSX brain PET imaging provides a means to study pathologic processes in neuroinflammatory and neurodegenerative diseases in vivo. However, applications for less invasive and more automated methods for the image analyses are called on to facilitate easier participation of the study subjects and to increase the reliability of analyses methods. According to our results, the population-based approach for acquiring input function for [11C]TMSX appears as a feasible and robust method and can be applied in patients with MS and PD. The proposed supervised clustering of reference region from [11C]TMSX PET images performed reliably for SPMS patients; however, further studies are needed to confirm the applicability of this method in other central nervous system pathologies.

Acknowledgments

The authors thank Dr Federico Turkheimer for providing the SuperPK software, which was used as a template for developing the [11C]TMSX SCA. The authors also thank the expertise and assistance of the radiochemistry, radiology, and laboratory personnel in the Turku PET Centre, and the tuition provided by Dr Jere Virta.

The authors declare no conflict of interest.

Footnotes

Supplementary Information accompanies the paper on the Journal of Cerebral Blood Flow & Metabolism website (http://www.nature.com/jcbfm)

This study was financially supported by the Finnish Academy, Turku University Hospital funds, Instrumentarium Science Foundation, Finnish MS Foundation, and the Finnish Parkinson Foundation. The research leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. HEALTH-F2-2011-278850 (INMiND).

Supplementary Material

Supplementary Information
Click here for additional data file. (55KB, doc)

References

  1. Ishiwata K, Wang WF, Kimura Y, Kawamura K, Ishii K. Preclinical studies on [11C]TMSX for mapping adenosine A2A receptors by positron emission tomography. Ann Nucl Med. 2003;17:205–211. doi: 10.1007/BF02990023. [DOI] [PubMed] [Google Scholar]
  2. Ishiwata K, Mishina M, Kimura Y, Oda K, Sasaki T, Ishii K. First visualization of adenosine A(2A) receptors in the human brain by positron emission tomography with [11C]TMSX. Synapse. 2005;55:133–136. doi: 10.1002/syn.20099. [DOI] [PubMed] [Google Scholar]
  3. Naganawa M, Kimura Y, Mishina M, Manabe Y, Chihara K, Oda K, et al. Quantification of adenosine A2A receptors in the human brain using [11C]TMSX and positron emission tomography. Eur J Nucl Med Mol Imaging. 2007;34:679–687. doi: 10.1007/s00259-006-0294-0. [DOI] [PubMed] [Google Scholar]
  4. Mishina M, Ishiwata K, Kimura Y, Naganawa M, Oda K, Kobayashi S, et al. Evaluation of distribution of adenosine A2A receptors in normal human brain measured with [11C]TMSX PET. Synapse. 2007;61:778–784. doi: 10.1002/syn.20423. [DOI] [PubMed] [Google Scholar]
  5. Mishina M, Ishiwata K, Naganawa M, Kimura Y, Kitamura S, Suzuki M, et al. Adenosine A(2A) receptors measured with [C]TMSX PET in the striata of Parkinson's disease patients. PLoS One. 2011;6:e17338. doi: 10.1371/journal.pone.0017338. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Mishina M, Kimura Y, Naganawa M, Ishii K, Oda K, Sakata M, et al. Differential effects of age on human striatal adenosine A(1) and A(2A) receptors. Synapse. 2012;66:832–839. doi: 10.1002/syn.21573. [DOI] [PubMed] [Google Scholar]
  7. Rissanen E, Virta JR, Paavilainen T, Tuisku J, Helin S, Luoto P, et al. Adenosine A2A receptors in secondary progressive multiple sclerosis: a [(11)C]TMSX brain PET study. J Cereb Blood Flow Metab. 2013;33:1394–1401. doi: 10.1038/jcbfm.2013.85. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Svenningsson P, Hall H, Sedvall G, Fredholm BB. Distribution of adenosine receptors in the postmortem human brain: an extended autoradiographic study. Synapse. 1997;27:322–335. doi: 10.1002/(SICI)1098-2396(199712)27:4<322::AID-SYN6>3.0.CO;2-E. [DOI] [PubMed] [Google Scholar]
  9. Kulisevsky J, Poyurovsky M. Adenosine A2A-receptor antagonism and pathophysiology of Parkinson's disease and drug-induced movement disorders. Eur Neurol. 2012;67:4–11. doi: 10.1159/000331768. [DOI] [PubMed] [Google Scholar]
  10. Mizuno Y, Kondo T. Adenosine A2A receptor antagonist istradefylline reduces daily OFF time in Parkinson's disease. Mov Disord. 2013;28:1138–1141. doi: 10.1002/mds.25418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Hauser RA, Olanow CW, Kiebutz KD, Pourcher E, Docu-Axelerad A, Lew M, et al. Tozadenant (SYN115) in patients with Parkinson's disease who have motor fluctuations on levodopa: a phase 2b, double-blind, randomised trial. Lancet Neurol. 2014;13:767–776. doi: 10.1016/S1474-4422(14)70148-6. [DOI] [PubMed] [Google Scholar]
  12. Ramlackhansingh AF, Bose SK, Ahmed I, Turkheimer FE, Pavese N, Brooks DJ. Adenosine 2A receptor availability in dyskinetic and nondyskinetic patients with Parkinson disease. Neurology. 2011;76:1811–1816. doi: 10.1212/WNL.0b013e31821ccce4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Tomasi G, Edison P, Bertoldo A, Roncaroli F, Singh P, Gerhard A, et al. Novel reference region model reveals increased microglial and reduced vascular binding of 11C-(R)-PK11195 in patients with Alzheimer's disease. J Nucl Med. 2008;49:1249–1256. doi: 10.2967/jnumed.108.050583. [DOI] [PubMed] [Google Scholar]
  14. Turkheimer FE, Edison P, Pavese N, Roncaroli F, Anderson AN, Hammers A, et al. Reference and target region modeling of [11C]-(R)-PK11195 brain studies. J Nucl Med. 2007;48:158–167. [PubMed] [Google Scholar]
  15. Yaqub M, van Berckel BN, Schuitemaker A, Hinz R, Turkheimer FE, Tomasi G, et al. Optimization of supervised cluster analysis for extracting reference tissue input curves in (R)-[(11)C]PK11195 brain PET studies. J Cereb Blood Flow Metab. 2012;32:1600–1608. doi: 10.1038/jcbfm.2012.59. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Ikoma Y, Edison P, Ramlackhansingh A, Brooks DJ, Turkheimer FE. Reference region automatic extraction in dynamic [(11)C]PIB. J Cereb Blood Flow Metab. 2013;33:1725–1731. doi: 10.1038/jcbfm.2013.133. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Zanotti-Fregonara P, Hines CS, Zoghbi SS, Liow JS, Zhang Y, Pike VW, et al. Population-based input function and image-derived input function for [(1)(1)C](R)-rolipram PET imaging: methodology, validation and application to the study of major depressive disorder. Neuroimage. 2012;63:1532–1541. doi: 10.1016/j.neuroimage.2012.08.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Zanotti-Fregonara P, Hirvonen J, Lyoo CH, Zoghbi SS, Rallis-Frutos D, Huestis MA, et al. Population-based input function modeling for [(18)F]FMPEP-d 2, an inverse agonist radioligand for cannabinoid CB1 receptors: validation in clinical studies. PLoS One. 2013;8:e60231. doi: 10.1371/journal.pone.0060231. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Poser CM, Paty DW, Scheinberg L, McDonald WI, Davis FA, Ebers GC, et al. New diagnostic criteria for multiple sclerosis: guidelines for research protocols. Ann Neurol. 1983;13:227–231. doi: 10.1002/ana.410130302. [DOI] [PubMed] [Google Scholar]
  20. Polman CH, Reingold SC, Edan G, Filippi M, Hartung HP, Kappos L, et al. Diagnostic criteria for multiple sclerosis: 2005 revisions to the ‘McDonald Criteria'. Ann Neurol. 2005;58:840–846. doi: 10.1002/ana.20703. [DOI] [PubMed] [Google Scholar]
  21. Heiss WD, Habedank B, Klein JC, Herholz K, Wienhard K, Lenox M, et al. Metabolic rates in small brain nuclei determined by high-resolution PET. J Nucl Med. 2004;45:1811–1815. [PubMed] [Google Scholar]
  22. Rissanen E, Tuisku J, Rokka J, Paavilainen T, Parkkola R, Rinne JO, et al. In vivo detection of diffuse inflammation in secondary progressive multiple sclerosis using PET imaging and the radioligand 11C-PK11195. J Nucl Med. 2014;55:939–944. doi: 10.2967/jnumed.113.131698. [DOI] [PubMed] [Google Scholar]
  23. Mosteller RD. Simplified calculation of body-surface area. N Engl J Med. 1987;317:1098. doi: 10.1056/NEJM198710223171717. [DOI] [PubMed] [Google Scholar]
  24. Takikawa S, Dhawan V, Spetsieris P, Robeson W, Chaly T, Dahl R, et al. Noninvasive quantitative fluorodeoxyglucose PET studies with an estimated input function derived from a population-based arterial blood curve. Radiology. 1993;188:131–136. doi: 10.1148/radiology.188.1.8511286. [DOI] [PubMed] [Google Scholar]
  25. Logan J, Fowler JS, Volkow ND, Wang GJ, Ding YS, Alexoff DL. Distribution volume ratios without blood sampling from graphical analysis of PET data. J Cereb Blood Flow Metab. 1996;16:834–840. doi: 10.1097/00004647-199609000-00008. [DOI] [PubMed] [Google Scholar]
  26. Moresco RM, Todde S, Belloli S, Simonelli P, Panzacchi A, Rigamonti M, et al. In vivo imaging of adenosine A2A receptors in rat and primate brain using [11C]SCH442416. Eur J Nucl Med Mol Imaging. 2005;32:405–413. doi: 10.1007/s00259-004-1688-5. [DOI] [PubMed] [Google Scholar]
  27. Brooks DJ, Papapetropoulos S, Vandenhende F, Tomic D, He P, Coppell A, et al. An open-label, positron emission tomography study to assess adenosine A2A brain receptor occupancy of vipadenant (BIIB014) at steady-state levels in healthy male volunteers. Clin Neuropharmacol. 2010;33:55–60. doi: 10.1097/WNF.0b013e3181d137d2. [DOI] [PubMed] [Google Scholar]
  28. Naganawa M, Kimura Y, Yano J, Mishina M, Yanagisawa M, Ishii K, et al. Robust estimation of the arterial input function for Logan plots using an intersectional searching algorithm and clustering in positron emission tomography for neuroreceptor imaging. Neuroimage. 2008;40:26–34. doi: 10.1016/j.neuroimage.2007.11.035. [DOI] [PubMed] [Google Scholar]
  29. Galazzo IB, Bose SK, Ramlackhansingh A, Ahmed I, Pavese N, Rizzo G, et al. Kinetic modeling of the adenosine A2A subtype receptor radioligand [11C]SCH442416 in humans. Neuroimage. 2010;52 (S1:S178. [Google Scholar]
  30. Takiguchi Y, Nagano M, Ikeda Y, Nakashima M. Early administration of YT-146, an adenosine A2 receptor agonist, inhibits neointimal thickening after rat femoral artery endothelium injury. Eur J Pharmacol. 1995;281:205–207. doi: 10.1016/0014-2999(95)00334-h. [DOI] [PubMed] [Google Scholar]
  31. Sakata M, Oda K, Toyohara J, Ishii K, Nariai T, Ishiwata K. Direct comparison of radiation dosimetry of six PET tracers using human whole-body imaging and murine biodistribution studies. Ann Nucl Med. 2013;27:285–296. doi: 10.1007/s12149-013-0685-9. [DOI] [PubMed] [Google Scholar]
  32. Mizuno M, Kimura Y, Tokizawa K, Ishii K, Oda K, Sasaki T, et al. Greater adenosine A(2A) receptor densities in cardiac and skeletal muscle in endurance-trained men: a [11C]TMSX PET study. Nucl Med Biol. 2005;32:831–836. doi: 10.1016/j.nucmedbio.2005.07.003. [DOI] [PubMed] [Google Scholar]
  33. Vincenzi F, Corciulo C, Targa M, Merighi S, Gessi S, Casetta I, et al. Multiple sclerosis lymphocytes upregulate A2A adenosine receptors that are antiinflammatory when stimulated. Eur J Immunol. 2013;43:2206–2216. doi: 10.1002/eji.201343314. [DOI] [PubMed] [Google Scholar]
  34. Fredholm BB, Svenningsson P. Adenosine-dopamine interactions: development of a concept and some comments on therapeutic possibilities. Neurology. 2003;61:S5–S9. doi: 10.1212/01.wnl.0000095204.89871.ff. [DOI] [PubMed] [Google Scholar]
  35. Lindstrom K, Ongini E, Fredholm BB. The selective adenosine A2A receptor antagonist SCH 58261 discriminates between two different binding sites for [3H]-CGS 21680 in the rat brain. Naunyn Schmiedebergs Arch Pharmacol. 1996;354:539–541. doi: 10.1007/BF00168448. [DOI] [PubMed] [Google Scholar]
  36. Ishiwata K, Noguchi J, Wakabayashi S, Shimada J, Ogi N, Nariai T, et al. 11C-labeled KF18446: a potential central nervous system adenosine A2a receptor ligand. J Nucl Med. 2000;41:345–354. [PubMed] [Google Scholar]
  37. Mori A, Shindou T. Modulation of GABAergic transmission in the striatopallidal system by adenosine A2A receptors: a potential mechanism for the antiparkinsonian effects of A2A antagonists. Neurology. 2003;61:S44–S48. doi: 10.1212/01.wnl.0000095211.71092.a0. [DOI] [PubMed] [Google Scholar]
  38. Popoli P, Frank C, Tebano MT, Potenza RL, Pintor A, Domenici MR, et al. Modulation of glutamate release and excitotoxicity by adenosine A2A receptors. Neurology. 2003;61:S69–S71. doi: 10.1212/01.wnl.0000095216.89483.a2. [DOI] [PubMed] [Google Scholar]
  39. Matos M, Augusto E, Agostinho P, Cunha RA, Chen JF. Antagonistic interaction between adenosine A2A receptors and Na+/K+-ATPase-alpha2 controlling glutamate uptake in astrocytes. J Neurosci. 2013;33:18492–18502. doi: 10.1523/JNEUROSCI.1828-13.2013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Orr AG, Orr AL, Li XJ, Gross RE, Traynelis SF. Adenosine A(2A) receptor mediates microglial process retraction. Nat Neurosci. 2009;12:872–878. doi: 10.1038/nn.2341. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information
Click here for additional data file. (55KB, doc)

Articles from Journal of Cerebral Blood Flow & Metabolism are provided here courtesy of SAGE Publications

RESOURCES