PET imaging of [¹¹C]PBR28 in Parkinson’s disease patients does not indicate increased binding to TSPO despite reduced dopamine transporter binding

The study was approved by the Research Ethics Committee in Stockholm, Sweden, and the Radiation Safety Committee of Karolinska University Hospital, Stockholm, and was performed in accordance with the current amendments of the Declaration of Helsinki and International Conference on Harmonization/ Good Clinical Practice guidelines. Written informed consent was obtained from all participants. The study included 16 control subjects and 16 patients diagnosed with PD. Demographic and clinical information is presented in Table 1.

The study is an extension of a previous investigation assessing the effect of the myeloperoxidase inhibitor AZD3241 on TSPO binding in PD patients [15]. Patients were examined during the period May to October 2012 within this previous study supported by AstraZeneca. In the present extension, matched control subjects were examined between March 2015 and February 2016. Patients fulfilled the modified UK Parkinson’s Disease Society Brain Bank criteria for idiopathic PD [16]. Inclusion and exclusion criteria have previously been described [15]. The clinical evaluation of patients included collection of demographic and general clinical information, a standardized neurological examination, and assessment of disease severity and motor signs using the Unified Parkinson’s Disease Rating Scale (UPDRS) [17]. Genotyping for TSPO polymorphism [18] to assess the affinity status of [¹¹C]PBR28 was performed in accordance with previously described procedures [15, 19], using the TaqMan® assays (Applied Biosystems) and detection via ABI 7900HT SDS (Applied Biosystems).

All subjects underwent PET measurements with the TSPO radioligand [¹¹C]PBR28. In addition, for confirmation of nigrostriatal pathology, PD patients were examined using the DAT radioligand [¹⁸F]FE-PE2I. [¹⁸F]FE-PE2I binding was analysed by visual assessment of striatal DAT loss, and criteria for inclusion were the presence of asymmetrical decrease in striatal [¹⁸F]FE-PE2I binding according to the criteria reported in the literature [20, 21].

[¹¹C]PBR28 was prepared from its corresponding desmethyl-PBR28 precursor (PharmaSynth AS, Tartu, Estonia), essentially as described elsewhere [22]. [¹⁸F]FE-PE2I was prepared as previously described [23] from its corresponding tosyl precursor, TsOE-PE2I (PharmaSynth AS). Injected radioactivity, molar radioactivity and injected mass of [¹¹C]PBR28 are presented in Table 1. For [¹⁸F]FE-PE2I the injected radioactivity was 158–195 MBq. The molar radioactivity at the time of injection was 65–433 GBq/μmol, corresponding to an injected mass of 0.20–1.3 μg.

Prior to PET measurements, two anatomical brain MRI examinations were performed in each subject. The first (T2-weighted) images were used for clinical evaluation and exclusion of pathology, and were evaluated by a clinical neuroradiologist. The second (T1-weighted) images were used for coregistration with the PET images and spatial normalization. Images were acquired using a 1.5-T Siemens Avanto system in patients and a 3-T General Electric Discovery MR750 (GE, Milwaukee, WI, USA) system in control subjects. The T1-weighted MR images were reoriented according to the line defined by the anterior and posterior commissures, resampled and cropped to generate a 220 × 220 × 170 matrix with 1 mm³ voxels. The images were segmented into grey and white matter, and CSF using SPM8 software (Wellcome Department of Cognitive Neurology, UK) and coregistered with PET images.

An individual plaster helmet was made for each subject and used with a head fixation system to minimize head movement. A cannula was inserted into the left or right cubital vein. The radioligand was dissolved in sterile physiological phosphate buffer (pH 7.4) and injected as a bolus over 10 s. The cannula was then immediately flushed with 10 ml saline.

PET data were acquired with a high-resolution research tomograph (HRRT; Siemens/CTI) over 72 min for [¹¹C]PBR28 examinations in control subjects and 93 min for [¹¹C]PBR28 and [¹⁸F]FE-PE2I examinations in patients. The data acquisition protocol was shortened in control subjects based on previous experience indicating that 60 min data acquisition is sufficient for quantification of [¹¹C]PBR28 binding to the TSPO [19]. List-mode data were binned and reconstructed as previously described [24].

A catheter was inserted into the radial artery and arterial blood was collected continuously during the first 10 min using an automated blood sampling system (ABSS; Allogg AB, Sweden). In addition, arterial blood samples (2–4 ml) were drawn manually at approximately 2, 4, 6, 8, 10, 15, 20, 25, 30, 40, 50, 70 and 90 min after radioligand injection in patients and 2, 4, 6, 8, 10, 15, 20, 25, 30, 40, 50, 60 and 72 min after radioligand injection in control subjects. The blood sampling protocol in control subjects was optimized based on accumulated experience from previous analyses of [¹¹C]PBR28 data [19]. Radioactivity was measured for 10 s in a well counter cross-calibrated with the PET system. After centrifugation, 0.2 ml plasma was pipetted and plasma radioactivity was measured in a well counter.

Plasma radioactivity corresponding to unchanged [¹¹C]PBR28 was determined in arterial blood sampled at 4, 10, 20, 30, 40, 50, 70, and 90 min in patients and at 4, 10, 20, 30, 40, 50 and 72 min in control subjects. The plasma obtained after centrifugation of blood was deproteinized with acetonitrile and analysed by high-performance liquid chromatography [25, 26]. The free fraction of [¹¹C]PBR28 in plasma was measured essentially as previously described [27]. The time curve representing the plasma radioactivity of unchanged radioligand was generated based on data obtained from ABSS and manual blood and plasma samples according to previously described procedures [15, 28]. Correction for radioactive metabolites was performed using a population-based approach to fit the fraction of parent radioligand in the plasma using an empirical model consisting of a mixture of the Hill and Richards equations [29].

PET images were corrected for head movement using a frame-by-frame realignment algorithm, in which all frames were individually realigned to the first minute of data acquisition. Parametric images of the total distribution volume (V_T) for [¹¹C]PBR28 and binding potential (BP_ND) for [¹⁸F]FE-PE2I were generated as previously described [15] using the wavelet-aided parametric imaging approach [30] based on a multilinear version of Logan graphical analysis. Parametric images for [¹¹C]PBR28 were generated based on 63 min of data acquisition. The parametric images were spatially normalized based on parameters obtained from segmentation and coregistration of MR images. Regions of interest (ROIs) for analysis of [¹¹C]PBR28 binding in the cerebellum, limbic cortices and thalamus were defined using the automated anatomical labelling template [31], and ROIs for analysis of radioligand binding in nigrostriatal regions and tracts were defined using a [¹⁸F]FE-PE2I PET template according to previously described procedures [32]. In addition, the software package FreeSurfer v. 5.0 [33, 34] was applied for parcellation of brain structures for volumetric analysis based on individual MR images.

Differences in regional [¹¹C]PBR28 V_T between the two groups were also examined by voxel-based analysis, using SPM5 (Statistical Parametric Mapping, Wellcome Trust Centre for Neuroimaging). To assess differences in V_T between controls and PD patients a two-sample t test was used, in which the two groups were entered separately and TSPO genotype, age and gender were entered as covariates. As alternative methods to assess the regional binding of [¹¹C]PBR28, the averaged radioactivity (expressed as standardized uptake value, SUV) between 40 and 60 min of data acquisition and distribution volume ratios (DVR) for ROIs relative to the cerebellum were calculated as previously described [35].

The difference between control subjects and patients in terms of age was assessed using a t test for independent samples. Based on the Kolmogorov-Smirnov and Shapiro-Wilk tests, radiochemistry data (injected radioactivity, injected mass and molar radioactivity) were not considered to be normally distributed. Differences in these parameters between groups were consequently analysed using the Mann-Whitney U test. Group differences in regional [¹¹C]PBR28 binding (V_T, DVR or SUV) were analysed using two-way ANOVA, with TSPO genotype and diagnostic group as independent variables, and [¹¹C]PBR28 binding as the dependent variable. In addition, the strength of evidence in favour of the null hypothesis (no effect of group or genotype) or alternative hypothesis (effect of group or genotype) was estimated from the Bayes factors computed by Bayesian ANOVA. Pearson’s correlation coefficient was used to analyse the associations between V_T values for [¹¹C]PBR28 and BP_ND values for [¹⁸F]FE-PE2I in the putamen and substantia nigra of the more severely affected hemisphere based on clinical ratings. The threshold for statistical significance was set at P < 0.05.

There was no statistically significant difference in age between control subjects and PD patients (Table 1; P = 0.97). Based on genotype analysis of the TSPO rs6971 polymorphism, 16 subjects (8 control subjects and 8 PD patients) were identified as high-affinity binders (HABs) and the remaining 16 subjects were identified as mixed-affinity binders (MABs). In patients, the PET images of [¹⁸F]FE-PE2I indicated low DAT binding in striatal regions, confirming the diagnosis of PD (Fig. 1). The [¹⁸F]FE-PE2I BPs for the caudate nucleus (mean 1.6, SD 0.6) and putamen (mean 0.89, SD 0.3) were low relative to the corresponding values for a reference control group (n = 20) of the same age (mean 4.0, SD 0.6, for the caudate nucleus; mean 4.4, SD 0.6, for the putamen) [36]. Regional volumes of the brain structures analysed are presented in Supplementary Table S1.

The [¹¹C]PBR28 injected radioactivity did not differ significantly between patients and control subjects (P = 0.22). Molar radioactivity for [¹¹C]PBR28 was higher (P = 0.01) and injected mass of the compound was lower (P = 0.01) in patients than in control subjects (Table 1). No statistically significant correlation was found between molar radioactivity or injected mass and estimated [¹¹C]PBR28 V_T values in HABs (P > 0.13) or MABs (P > 0.15; Supplementary Fig. S1).

Time curves for the fraction of parent radioligand in plasma indicated a slower rate of metabolism in PD patients, with group differences reaching statistical significance (P < 0.05) for samples collected at 30, 40, 50 and 70 min after tracer injection (Supplementary Fig. S2). However, the area under the curve for the metabolite-corrected plasma radioactivity did not differ significantly between control subjects and PD patients (P = 0.75; Fig. S3a). Estimates of the free fraction in plasma were higher in patients (mean 0.11, SD 0.020) than in control subjects (mean 0.049, SD 0.020; P < 0.001; Fig. S3b).

Parametric images of the average V_T for [¹¹C]PBR28 in control subjects and PD patients are shown in Fig. 1. Global analysis of V_T for the whole brain region showed a main effect of genotype (P < 0.001; BF₁₀ 76,312), but no statistically significant effect of diagnostic group (P = 0.86; BF₁₀ 0.34), or a group by genotype interaction (P = 0.56; BF₁₀ 0.50). For all regions studied, the two-way ANOVA revealed a main effect of TSPO genotype with higher [¹¹C]PBR28 V_T in HABs than in MABs (P < 0.001; BF₁₀ >1,000), but no statistically significant effect of diagnostic group (P > 0.05; Fig. 2; Table S2). A Bayesian ANOVA provided evidence in favour of the hypothesis of no difference in V_T between the groups in the substantia nigra, putamen, nigrostriatal tract, limbic cortices, thalamus and cerebellum (BF₁₀ 0.34–0.43). For the caudate nucleus an effect of group at the trend level was found, with lower [¹¹C]PBR28 V_T in patients than in control subjects (P = 0.056; BF₁₀ 0.68). No statistically significant group by genotype interactions on [¹¹C]PBR28 V_T were found for any of the regions analysed (P > 0.15; BF₁₀ 0.41–0.88). Voxel-wise analysis of parametric maps for [¹¹C]PBR28 V_T revealed no clusters of statistically significant differences between control subjects and PD patients (results not shown).

Comparison of [¹¹C]PBR28 V_T values obtained using ROI definition in individual space showed a main effect of TSPO genotype (P < 0.001; BF₁₀ >570), but no statistically significant effect of diagnostic group (P > 0.13; BF₁₀ 0.34–0.50) or a group by genotype interaction (P > 0.13; BF₁₀ 0.44–0.92; Table S3). Analysis of the average SUV for the 40–60 min data acquisition (Table S4) showed a main effect of genotype (P < 0.001; BF₁₀ 113–2,168), but no statistically significant effect of diagnostic group (P > 0.23; BF₁₀ 0.34–0.45), or a group by genotype interaction (P > 0.27; BF₁₀ 0.40–0.62).

Analysis of the ratio of V_T for regions of interest relative to that for the cerebellum (DVR) revealed no statistically significant main effect of TSPO genotype for any of the regions studied (P > 0.05; BF₁₀ 0.34–0.93; Fig. S4). For the caudate nucleus, the nigrostriatal tract, limbic cortices and thalamus there were statistically significant effects of diagnostic group with lower DVR in PD patients than in controls (P < 0.05; BF₁₀ 2.4–13). A statistically significant group by genotype interaction on DVR was found for the caudate nucleus (P = 0.013; BF₁₀ 4.2), but not for other regions analysed (P > 0.05; BF₁₀ 0.47–1.7).

Analysing data from all patients, V_T values for [¹¹C]PBR28 did not correlate with [¹⁸F]FE-PE2I BP_ND values for binding to DAT in the putamen (r = −0.14, P = 0.62) or the substantia nigra (r = 0.28, P = 0.29; Fig. 3). Analysing data from HABs and MABs separately, no statistically significant correlations were found between binding to DAT and to TSPO: in HABs, r = 0.30, P = 0.48, in the putamen, and r = 0.56, P = 0.15, in the substantia nigra; in MABs, r = −0.24, P = 0.56, in the putamen, and r = 0.050, P = 0.91, in the substantia nigra. With regard to correlations between TSPO binding in the substantia nigra and DAT binding in the putamen, no statistically significant correlations were found when analysing data from all patients (r = −0.044, P = 0.87), or when analysing data from HABs and MABs separately (in HABs, r = 0.30, P = 0.46; in MABs, r = 0.0088, P = 0.98; Fig. 3).