Background
The regional cerebral binding of adenosine A
2A receptor (A
2AR) antagonists, [7-methyl-
11C]-(
E)-8-(3,4,5-trimethoxystyryl)-1,3,7-trimethylxanthine (
11C-TMSX) [
1] and
11C-KW-6002 [
2],[
3], were quantitatively investigated
in vivo in healthy human.
11C-TMSX has been evaluated in human brain studies in not only healthy human controls [
4]-[
7] but also in drug-naïve Parkinson's disease patients before and after therapy [
8]. An aging effect on A
2AR was also evaluated using
11C-TMSX PET [
9]. Several outcome measures, such as distribution volume (
VT), distribution volume ratio (DVR), and binding potential (BP
ND), can be used to detect changes of A
2AR binding due to disease progression or therapeutic treatment. The reproducibility of these measures is important to conduct studies to detect change in A
2AR. Given that A
2AR distribution is heterogeneous, with only a very small amount of extrastriatal-specific binding, either the frontal cortex in the rat [
3], the cerebellum in the monkey [
10],[
11], or centrum semiovale [
7] or cerebral cortex [
8],[
9] in human, was used as a reference region to estimate non-displaceable binding of
11C-TMSX. Since only a few
postmortem human brain studies and blocking studies are available, it is not clear which region is a suitable reference region.
The aim of this paper was to assess the test-retest reproducibility of PET outcome measures, VT and BPND, with the centrum semiovale and cerebral cortex as candidate reference regions.
Methods
Human subjects
All studies were performed under a protocol approved by the Ethics Committee of the Tokyo Metropolitan Institute of Gerontology. Five healthy, male subjects participated in this study (mean age ± SD, 22.4 ± 2.6 years old, range: 21 to 27 years old). Subjects were all right-handed and screened for history of neurological, psychiatric, and physical diseases. All subjects did not have a history of alcoholism and not on any medications to affect brain function. Caffeine intake was not allowed for at least 12 h prior to PET scanning. Written informed consent was obtained from all subjects after receiving an explanation of the study. Magnetic resonance (MR) images were acquired on all subjects to eliminate those with any brain abnormalities and to place regions of interest (ROIs) on PET images. The MR imaging was conducted with three-dimensional spoiled gradient-recalled echo (SPGR) imaging on a SIGNA 1.5 Tesla machine (General Electric, Waukesha, WI, USA) [
6].
Radiochemistry
Radiosynthesis of
11C-TMSX followed the literature procedure [
12]. All procedures were conducted under dim light to prevent photoisomerization of
11C-TMSX. The radiochemical purity of
11C-TMSX was >99%.
PET acquisition
Each subject underwent two
11C-TMSX brain PET scans on two different days, and time of scanning was identical for test and retest scans of each individual subject, in order to remove the influence of circadian rhythm. The inter-scan interval was 28 to 35 days. Dynamic PET images were acquired in the Positron Medical Center, Tokyo Metropolitan Institute of Gerontology with the SET-2400 W PET scanner (Shimadzu, Kyoto, Japan), which acquires 63 slices (3.125-mm slice separation) with a spatial resolution of 4.4 mm full width at half maximum (FWHM) and a z-axis resolution of 6.5 mm FWHM [
13]. Prior to the scan, a 5-min
68Ga/
68Ge transmission scan was conducted for attenuation correction.
11C-TMSX was injected intravenously over 60 s. Emission data were collected in two-dimensional mode for 1 h in 27 frames of increasing duration (6 × 10 s; 3 × 30 s; 5 × 1 min; 5 × 2.5 min; 8 × 5 min). Head movement was minimized with an air cushion. The dynamic images were reconstructed by the filtered back-projection method using a Butterworth filter (second-order low-pass filter, cutoff frequency was 1.25 cycles/cm) with corrections for scatter and randoms.
In advance of each scan, an arterial catheter was inserted into the radial artery for blood sampling. After radiotracer injection, arterial blood samples were manually collected every 10 s for the first 2 min and thereafter at longer intervals, 2.25, 2.5, 3, 5, 7, 10, 15, 20, 30, 40, 50, and 60 min post-injection. A total of 24 samples were obtained per scan. Whole blood and plasma were counted in a cross-calibrated well-type gamma-counter (BSS-1, Shimadzu, Kyoto, Japan). An additional venous blood sample was taken before 11C-TMSX administration, which was used for the in vitro assessment of the fraction of 11C-TMSX in plasma bound to plasma proteins (fP). Arterial blood sampling was not available in one subject. Thus, a total of five subjects were included in reference region analyses and four subjects were also analyzed using plasma data.
The fraction of intact radioligand to total plasma activity was determined from blood samples collected at 3, 10, 20, 30, 40, and 60 min after injection by high-performance liquid chromatography (HPLC). The blood was centrifuged at × 7,000 g for 1 min at 4°C to obtain the plasma, which was denatured with an equivalent volume of acetonitrile in an ice-water bath. The suspension was centrifuged under the same conditions and divided into soluble and precipitable fractions. The precipitate was resuspended in 2 vol. of 50% aqueous acetonitrile followed by centrifugation. The recovery yield of the radioactivity in the two soluble fractions was 98.7%. Two soluble fractions were combined, and into this solution, an equivalent volume of a solution of 50-mM aqueous acetic acid and 50-mM aqueous sodium acetate (pH 4.5; 50/50, v/v) was added. After centrifugation of the samples as described above, the supernatant was loaded onto a Nova-Pak C8 column equipped in an RCM 8 × 10 module (8 mm diameter × 100 mm length; Millipore-Waters, Milford, MA, USA). The mobile phase was a mixture of acetonitrile, 50-mM aqueous acetic acid and 50-mM aqueous sodium acetate (pH 4.5; 4/3/3, v/v/v) at a flow rate of 2 mL/min. The elution profile was detected with a radioactivity monitor (FLO-ONE 150TR; Packard Instrument, Meriden, CT, USA). The retention time of 11C-TMSX was 6.2 min. The recovery in the eluate of the injected radioactivity was essentially quantitative. The six measured parent fractions were fitted by a sum of exponential functions. The metabolite-corrected plasma curve was generated as the product of the total plasma activity and the fitted parent fraction curve.
Individual fP values were determined by ultrafiltration. Prior to administration of 11C-TMSX, approximately 6 mL of blood was taken from each subject. A reference blood sample was created by adding 22.9 ± 15.7 MBq (at the time of administration, range: 10.1 to 49.5 MBq of 11C-TMSX in approximately 60 μL to this blood sample and incubated for 10 min at 37°C). Following centrifugation (2,000 g at room temperature for 3 min), triplicates of 400 μL aliquots of plasma sample were pipetted into ultrafiltration tubes (Microcon-30, 30 kDa, Merck Millipore, Billerica, MA, USA), and centrifuged at room temperature (14 min at 14,000 g). The free fraction fP was calculated as the ratio of activity in the ultrafiltrate to the total plasma. The amount of nonspecific binding of 11C-TMSX to the filter was also determined by applying the same procedure to a sample created by addition of 11C-TMSX to saline.
Image analysis
Regions of interest were defined by manually drawing circles using the registered MR images as additional reference. The details are written in [
6],[
14]. Time-activity curves (TACs) were generated for eight ROIs: anterior putamen, posterior putamen, putamen, caudate head, thalamus, cerebellum, centrum semiovale, and cerebral cortex. The putamen ROI consists of the anterior and posterior putamen subregions. The cerebral cortex ROI included the frontal, temporal, and occipital cortices.
In the present study, the cerebral cortex and centrum semiovale were chosen as candidate reference regions. For
11C-TMSX kinetic analysis, the cerebellum was not used as a reference region, because A
2AR binding in our previous human study [
7] was higher in the cerebellum than in neocortical regions. In a previous human autoradiographic study [
15], the density of A
2ARs in the frontal cortex was found to be low, as that in the temporal and occipital cortices.
Outcome measures
The DVR has been used in our previous study on an aging effect of A
2AR in human brain [
8],[
9]. In this study, the two additional outcome measures,
VT and BP
ND, were estimated. The definition of the outcome measures is described in [
16]. Regional TACs were analyzed using the Logan graphical analysis (LGA) with input function and reference tissue (two-parameter version) [
17],[
18] to estimate the outcome parameters of
VT and BP
ND. Starting time (
t*) was set to 10 min post-injection [
7].
Statistical analyses
The test-retest reproducibility was statistically evaluated according to the following three criteria: signed test-retest variability (TRV), absolute test-retest variability (aTRV), and intra-class correlation coefficient (ICC). TRV was calculated as the difference between the test and retest measurements, divided by the mean of the test and retest values (2 × (
ptest −
pretest)/(
ptest +
pretest)). aTRV was calculated as the absolute value of TRV (2 × |
ptest −
pretest|/(
ptest +
pretest)). TRV indicates whether there is a systematic trend between the test and retest scans. The test-retest reliability of the two parameter measurements was the ICC calculated using the following equation [
19]:
where BSMSS and WSMSS are the mean sum of squares between subjects and within subjects, respectively. In the test-retest study, an ICC value ranges from −1 (no reliability) to 1 (maximum reliability) [
20],[
21]. Sample sizes were calculated to detect a 20-percent difference in BP
ND between independent groups (two-tails
t-test) using the software G*power 3.1 [
22]. The confidence level was set to be 5% (
P < 0.05) and statistical power to 0.8. The mean of the test scans was used as the mean of baseline scans, and the SDs of the baseline and blocking scans were assumed to be same as the SDs of the test scans. All statistical parameters except for power analysis were calculated with MATLAB Version 7.12.0.635 (the MathWorks Inc., Natick, MA, USA) and Microsoft Excel 2010 (Microsoft, Redmond, WA, USA).
Discussion
The plasma free fraction (
fP) was measured in this study, allowing for correction of
VT values. This correction by
fP is useful if
fP can be measured reliably and if there is substantial intra-subject variation. In our measurements, the
fP was consistently low (<3%), with evidence that
11C-TMSX stuck to the ultrafiltration tubes, which may lead to underestimation of
fp. However, the
fP value measured in [
23] was 9.1% ± 0.4% (
n = 6, human) by the ultrafiltration method. Such a discrepancy might be attributable to high stick factor in our data. Note that the stick factor was not reported in [
23]. Another possibility is the difference in the preparation of the injection solution. Finally, the signed and absolute TRVs were larger for
VT/
fP compared to those of
VT. Hence, the normalization of
VT by
fP did not reduce variability in this case.
Inter-subject variability (% coefficient of variation (COV)) of
VT at retest scans were lower (approximately 7%) than that of the test scans (approximately 16%), while no significant difference was observed in the injected dose, specific activity, and
fP of
11C-TMSX. Another possibility to explain the difference in the inter-subject variability is a difference in the metabolism of the tracer. The subjects were controlled for caffeine intake, but not for smoking habituations. Nicotine consumption might change the metabolism of
11C-TMSX as seen in the study with adenosine A
1 receptor ligand
18 F-CPFPX [
24]. In a retrospective investigation, it turned out that subjects consisted of a nonsmoker, a smoker (blood sampling was not available), and three subjects with unknown status. However, the parent fraction of
11C-TMSX was very high and well reproducible (Figure
1B). Therefore, we concluded that a change in the metabolism speed was not a reason to increase the inter-subject variability. In contrast to
VT, such a difference in the inter-subject variability did not exist in BP
ND. The difference in the %COV between test and retest scans might come from errors included in the input function measurement.
The test-retest variability and reliability of
VT were good (aTRV ≤10%, ICC >0.6) across regions except for the thalamus (aTRV: 13% and ICC: 0.27). For BP
ND, a good absolute TRV was seen in the high A
2A regions (putamen and caudate). However, lower-binding regions (BP
ND < 0.4) showed high aTRV (>15%) and low ICC values; this is not surprising, since BP
ND is small in those regions. We examined the test-retest variability data of
VT and BP
ND from a number of radioligands. The aTRV of
VT of
11C-TMSX (8% averaged across all regions) was comparable to that of other radioligands used to study dopamine and adenosine receptors. The reported aTRV values of
VT were 5% to 11% (average: 7%) with
11C-FLB457 [
25] for dopamine D
2/3 receptor and 12% to 14% (average: 13%) with
18 F-CPFPX [
26] for adenosine A
1 receptor. The aTRV of BP
ND with
11C-TMSX was comparable to that with
11C-FLB457 (6% to 15%) and larger than that with
18 F-FPFPX (3% to 9%).
Given the good reproducibility of
VT,
11C-TMSX should be suitable for use in receptor occupancy studies with input function. The range of
VT values was not wide across regions (0.70 to 1.46 mL/cm
3). However, using the occupancy plot [
27] is feasible using the regions with a narrow range of
VT values with
11C-GSK931145 for glycine type 1 transporter (0.43 to 0.79 mL/cm
3) [
28] and
18 F-CPFPX for adenosine A
1 receptor (0.42 to 0.82 mL/cm
3) [
29]. Note that the occupancy plot assumes that the receptor occupancies are uniform in all regions of interest. Previous reports [
1],[
30] suggest that some regions might have an ‘atypical’ binding. Therefore, we need to carefully choose regions used for the occupancy plot with
11C-TMSX. Another possible way for estimating receptor occupancy is to estimate a relationship between blocking dose (or plasma level) and
VT for each region [
31]. This second method can be used even if all regions have the same baseline
VT.
The test-retest variability of BPND values using the cerebral cortex as a reference region was larger than those using the centrum semiovale. In the striatum, the high A2AR-binding region, the aTRVs of BPND were 5% in the putamen and 19% in the caudate head using the cerebral cortex as a reference region. On the other hand, the aTRVs of BPND were 3% in the putamen and 13% in the caudate head using the centrum semiovale as a reference region. This is partly because the BPND value was smaller using the cerebral cortex as a reference region.
The thalamus showed a low reproducibility of both
VT and BP
ND values. Moreover, while the mean distribution volume in the thalamus was high, a
postmortem study with
3H-SCH58261 [
15] showed that A
2AR density is low. The uptake in the thalamus is considered to be ‘atypical’ binding [
1],[
30], which is different from classical A
2AR binding. This low reproducibility in the thalamus may be partly due to such an ‘atypical’ binding. Thus, given the low reproducibility and ‘atypical’ binding, the thalamus should be carefully considered in further clinical research.
Using either the cerebral cortex or centrum semiovale as a reference region, reference LGA and LGA with input function provided similar BPND values. The TRV and aTRV of BPND were slightly smaller using the reference LGA. Not surprisingly, the reference tissue model is not affected by errors in the measurement of input function. This suggests that the reference LGA can be useful for further studies.
There are two limitations of this study: unknown optimal reference region for
11C-TMSX and small sample size. As far as we know, the only available A
2AR blocking study using an antagonist radiotracer
in vivo in human brain is a
11C-KW-6002 PET study with varying dose of cold KW-6002 [
3] However, blocking results in the centrum semiovale and neocortical regions were not included in the report. Thus, the suitability of the cerebral cortex or central semiovale as a reference region has yet to be determined by blocking or occupancy studies. Due to a lack of blocking study and
postmortem study in the regions with low A
2AR density, the region with lowest
VT was chosen. For the SPECT A
2AR tracer
123I-MNI-420, while a reference region is not yet validated, a test-retest reproducibility of BP
ND was evaluated to facilitate the comparison between
123I-MNI-420 and other A
2AR radiotracers [
32]. We also took an exploratory approach to calculate BP
ND values using candidate reference regions in order to evaluate BP
ND reproducibility. However, the determination of the reference region is most desirable in order to establish the utility of
11C-TMSX for PET imaging. In this study, we evaluated outcome measures with input function in four subjects. We examined sample sizes for test-retest human studies using other radioligands. As far as we know, the minimum sample size is three subjects for test-retest protocol (
18 F-MK-6577 [
33] for glycine transporter type 1 and
123I-MNI-420 [
32] for A
2AR).
Competing interests
The authors declare that they have no competing interest.
Authors’ contributions
Contributions to the conception of the study and its design were made by MN, MM, KIi, and KIa. The experiments were conducted by MM, MS, KO, MH, KIi, and KIa. MS was responsible for measuring plasma free fraction. KIa was responsible for tracer synthesis and metabolite analysis. MN performed the analysis and wrote the manuscript. MM, MS, KO, MH, KIi, and KIa helped in the discussions and drafting of the manuscript. All authors read and approved the final manuscripts.