Modeling considerations for 11C-CUMI-101, an agonist radiotracer for imaging serotonin 1A receptor in vivo with PET

Abstract

Several lines of evidence demonstrate involvement of serotonin 1A receptors (5-HT1ARs) in the pathophysiology of neuropsychiatric disorders such as depression, suicidal behavior, schizophrenia, and Alzheimer's disease. We recently published the synthesis and initial evaluation of [O-methyl-11C]2-(4-(4-(2-methoxyphenyl)piperazin-1-yl)butyl)-4-methyl-1,2,4-triazine-3,5(2H,4H)dione (11C-MMP), a 5-HT1AR agonist. Here we determine the optimal modeling parameters for 11C-MMP under its new name, 11C-CUMI-101, in Papio anubis.

Methods: PET scans were performed on 2 adult male P. anubis; 166.5 MBq +/- 43.0 (4.50 +/- 1.16 mCi) of 11C-CUMI-101 were injected as an intravenous bolus, and emission data were collected for 120 min in 3-dimensional mode. We evaluated 4 different models (1- and 2-tissue compartment iterative and noniterative kinetic models, basis pursuit, and likelihood estimation in graphical analysis [LEGA]), using binding potential (BPF = Bmax/Kd) (Bmax = maximum number of binding sites; Kd = dissociation constant) as the outcome measure. Arterial blood sampling and metabolite-corrected arterial input function were used for full quantification of BPF. To assess the performance of each model, we compared results using 6 different metrics (percentage difference, within-subject mean sum of squares [WSMSS] for reproducibility; variance across subjects, intraclass correlation coefficient [ICC] for reliability; identifiability based on bootstrap resampling of residuals; and time stability analysis to determine minimal required scanning time) at each of 6 different scanning durations. Models were also evaluated on scans acquired after injecting the 5-HT1A antagonist [N-(2-(4-(2-methoxyphenyl)-1-piperazinyl)ethyl)-N-(2-pyridinyl) cyclohexane carboxamide] [WAY100635] 0.5 mg/kg, intravenous) and the 5-HT1A agonist 8-hydroxy-2-(di-n-propylamino)tetralin) [8-OH-DPAT] 2 mg/kg, intravenous).

Results: All metabolites are more polar than 11C-CUMI-101, and no significant change in metabolites was observed in the blocking studies. The free fraction is 59% +/- 3%. We determined that 100 min of scanning time is adequate and that for the region-of-interest (ROI)-level analysis, the LEGA model gives the best results. The median test-retest percentage difference for BPF is 11.15% +/- 4.82% across all regions, WSMSS = 2.66, variance = 6.07, ICC = 0.43, and bootstrap identifiability = 0.59. Preadministration of WAY100635 and 8-OH-DPAT resulted in 87% and 76% average reductions in BPF values, respectively, across ROIs.

Conclusion: On the basis of the measurable free fraction, high affinity and selectivity, adequate blood-brain permeability, and favorable plasma and brain kinetics, 11C-CUMI-101 is an excellent candidate for imaging high-affinity 5-HT1ARs in humans.

FIGURE 1

(A) Fraction of plasma activity associated with unmetabolized ¹¹C-CUMI-101 in baboon plasma during a single scan. Fitted line is estimated from the Hill model (1 − At^B/[t^B + C]). (B) Total and metabolite-corrected plasma radioactivity. Fitted line represents the 3-exponential function fit to the data. Error bars represent weights calculated by the delta method.

FIGURE 2

Regional time–activity curves and corresponding least-squares minimized fits to 3 different models. (A) 2TCNI. (B) Basis pursuit. (C and D) LEGA (native and transformed space, respectively). CER = cerebellum; ACN = anterior cingulate; TEM = temporal cortex.

FIGURE 3

Modeling metrics comparing 2TCNI, basis pursuit, and LEGA models at 6 different scan durations. (A) PD. (B) WSMSS. (C) Variance. (D) ICC. (E) Identifiability. For all measures, medians are taken across all scans and all ROIs, for both subjects. (A) Error bars represent average deviation from the median. Asterisk in A indicates that the modeling of dorsal raphe did not converge for basis pursuit for the 100-min scan time (for a single study) and thus was left out of the analysis.

FIGURE 4

Mean percentage change (PD) in V_T (A) and BP_F (B) of ¹¹C-CUMI-101 after blockade with 8-OH-DPAT and WAY-100635. Values were derived from the LEGA model with a 100-min scan duration. ACN = anterior cingulate; AMY = amygdala; CER = cerebellum; DRN = dorsal raphe; HIP = hippocampus; INS = insula; PFC = prefrontal cortex; TEM = temporal cortex.

FIGURE 5

(First row) Sagittal, coronal, and axial parametric PET images of ¹¹C-CUMI-101 BP_F values in baboon brain. (Second and third rows): PET images after blockade by 8-OH-DPAT and WAY-100635, respectively. (Fourth row) Corresponding coregistered MR images. On the MR images, the hippocampus has been highlighted.

FIGURE 6

Correlation plots of V_T values across various modeling approaches including ROI-analysis vs. voxel-based analyses. (A) 2TCNI voxel vs. 2TCNI ROI (R² = 0.97). (B) 2TCNI ROI vs. LEGA ROI (R² = 0.97). (C) Basis pursuit ROI vs. LEGA ROI (R² = 0.97). (D) 2TCNI voxel vs. LEGA ROI (R² of 1.00). Line of identity has been added for reference. Data reflect all ROIs in all studies.