Image-Derived Blood Normalization of Antibody-Based TREM2 PET in Mouse Models of Amyloidosis and Myocardial Infarction

Abstract

The triggering receptor expressed on myeloid cells 2 (TREM2) plays a pivotal role in the activation of myeloid cells and is currently being investigated as a potential therapeutic target in several diseases. In this study, we established enhanced quantification of PET images of a ⁶⁴Cu-labeled antibody-based PET radiotracer as a noninvasive tool for the assessment of TREM2 expression in the brain and peripheral organs of mice. We used TREM2 knockout mice that lack target expression to investigate data-driven blood normalization of PET images against percentage of injected dose normalization. Methods: TREM2 knockout and wild-type mice (n = 11 each) were injected with the radiotracer [⁶⁴Cu]Cu-NODAGA-ATV:4D9 (ATV is antibody transport vehicle). Twenty hours after injection, TREM2 PET was conducted and blood samples were collected. A voxelwise analysis with statistical parametric mapping served to determine voxels that correlate with ex vivo blood radioactivity levels. Furthermore, TREM2 PET signals were compared between mice with and those without TREM2 expression using image-derived blood normalization. Correlation with TREM2 protein expression levels in the lung, liver, spleen, and bone marrow was used to validate organ-specific PET results. Disease models of brain amyloidosis and myocardial infarction were investigated to test for the value of image-derived normalization in mice. Results: Blood radioactivity levels derived from a statistical parametric mapping-derived region of interest demonstrated a robust correlation with radioactivity measurements obtained from ex vivo blood samples. Voxelwise clusters of TREM2 PET signals were more robustly detected after blood normalization of the PET images. Significant voxelwise clusters of TREM2 PET signals in peripheral organs correlated with TREM2 protein expression levels. Furthermore, image-derived normalization enhanced the significance of voxelwise clusters of TREM2 in the brains of App ^SAA;TfR^mu/hu mice, as well as the TREM2 signal in the myocardial infarct region. Both strongly correlated with ex vivo autoradiography. Conclusion: Normalization of PET images to account for blood levels enhanced the detection of TREM2. This improved methodology for TREM2 PET analysis provides a promising basis for future assessments of TREM2 imaging.

Keywords: 64Cu; ATV:4D9; PET; SPM; TREM2.

Graphical abstract

FIGURE 1.

Correlation between radioactivity concentration in ex vivo blood samples and radioactivity concentration derived from manually placed VOIs (A) or SPM-driven VOIs (B) in PET images of WT and TREM2 knockout mice (n = 11 each). Linear regression, 95% CI. Signals resolved for genotype and correlations are presented in Supplemental Figure 4. (C) SPM image correlating individual PET images and ex vivo blood levels projected on CT. Multiple regression. Resulting SPM image with P < 0.05 and cluster size of >50 voxels is presented in Supplemental Figure 2B.

FIGURE 2.

Voxelwise group comparison of TREM2 radiotracer uptake in PET images of WT vs. TREM2 knockout mice (n = 11 each) after SPM-derived blood level normalization (A) or %ID normalization (B). Two-sample t test, P < 0.05, cluster size of >50 voxels projected onto CT. Average PET images are presented in Supplemental Figure 5.

FIGURE 3.

(A) 3-dimensional CT-based segmentation of mouse, displaying bones (beige), bone marrow (purple), lung (pink), liver (brown), and spleen (green). (B) T values derived from SPM images of group comparisons of WT vs. TREM2 knockout mice (n = 11 each) using SPM-derived blood normalization or %ID normalization. Mean ± SD. (C) T values derived from blood normalization from (B) and mouse TREM2 protein levels detected by MSD in WT mice (n = 3). Mean ± SD.

FIGURE 4.

(A) Voxelwise group comparison of TREM2 radiotracer uptake in brains of App^SAA;TfR^mu/hu vs. WT;TfR^mu/hu mice (n = 6 each) of PET images after SPM-derived blood level normalization or %ID normalization. Two-sample t test, P < 0.05, cluster size of >50 voxels, projected onto a mouse MRI. Average PET images are presented in Supplemental Figure 11. (B) Ex vivo autoradiography of representative App^SAA;TfR^mu/hu brain slice. Additional images are available in Supplemental Figure 12. (C) Immunohistochemistry of adjacent brain slice (DNA/4′,6-diamidino-2-phenylindole [blue], microglia/Iba1 [yellow], and TREM2 [red]). Individual fluorescence channels are shown at plaque level in Supplemental Figure 13. Red arrows indicate TREM2-rich signals. WM = white matter.

FIGURE 5.

(A) Representative PET images of mice with left anterior descending (LAD) ligation and sham mouse, normalized to blood levels using segmentation from carotid artery. (B) Corresponding ex vivo autoradiography images. Quantification of tracer signal in PET and ex vivo autoradiography is available in Supplemental Figure 15. A = anterior; P = posterior.