Radiosynthesis of [¹⁸F]-GE180 was performed as previously described (Wickstrøm et al., 2014), with slight modifications (Brendel et al., 2016), a procedure yielding radiochemical purity exceeding 98%, and specific activity of 1400 ± 500 GBq/μmol at end of synthesis. Radiosynthesis of [¹⁸F]-florbetaben was performed as previously described (Rominger et al., 2013), a procedure yielding radiochemical purity exceeding 98%, and specific activity of 80 ± 20 GBq/μmol at the end of synthesis.

All experiments were carried out in compliance with the National Guidelines for Animal Protection, Germany with the approval of the regional Animal care committee of the Government of Oberbayern (Regierung Oberbayern), and were overseen by a veterinarian. Female animals were housed in a temperature- and humidity-controlled environment with a 12-h light–dark cycle, with free access to food (Ssniff, Soest, Germany) and water.

The transgenic B6.PS2APP (line B6.152H) is homozygous for both, the human presenilin (PS) 2, N141I mutation and the human amyloid precursor protein (APP) K670N, M671L mutation. The APP and PS2 transgenes are driven by mouse Thy-1 and mouse prion promoters, respectively. This line had been created by co-injection of both transgenes into C57Bl/6 zygotes (Richards et al., 2003). Homozygous B6.PS2APP mice show first appearance of plaques in the cerebral cortex and hippocampus at 5–6 months of age (Ozmen et al., 2009).

μPET with [¹⁸F]-GE180 and [¹⁸F]-florbetaben emission recordings (maximum 10 days apart) were obtained cross-sectionally in equal-sized groups (N = 8) of PS2APP and WT mice aged 5, 8, 13, and 16 months (±0.5 months). Group-size calculations were planned based on the assumption of a type I error α = 0.05, giving a power of 0.8, and with the possibility of substitution in case of dropouts. Within 1 week after completion of the final PET session, PS2APP mice were perfused with PBS while deeply anesthetized, and brains were removed for biochemical and immunohistochemical analyses. In case of dropouts (N = 7) additional age-matched mice were substituted for the dual μPET ([¹⁸F]-GE180 and [¹⁸F]-florbetaben) arm of the study, as equal group sizes are necessary for optimal statistical parametric mapping (SPM) analysis. For biochemical and immunohistochemical analyses in WT mice groups of N = 6 mice at 6 and 18 months of age were used without prior PET. A detailed overview of the different groups of mice and all modalities is provided in Table 1.

Frozen mouse cerebrum was cryo-pulverized in liquid nitrogen and sequentially extracted in Tris-buffered saline (TBS) followed by extraction with TBS-Triton (1%). In brief, 30 mg cryo-pulverized samples of brain tissue were homogenized in ice-cold TBS (10w/v) using a 26G needle and cleared by ultracentrifugation (186,000× g, 1 h, 4°C). The resulting pellet was washed once with TBS, extracted with TBS-Triton (TBS-T fraction) and lysate cleared again by ultracentrifugation (186,000× g, 1 h, 4°C).

For immunoblot analysis cryo-pulverized brains were extracted in immunoprecipitation lysis buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 5% glycerol) and cleared by centrifugation (17,000× g, 30 min, 4°C). Protein concentrations were measured using the bicinchoninic acid (BCA) method (Pierce). TREM2 was immunoprecipitated from 300 μg brain lysates using biotinylated anti-TREM2 (BAF1729; R&D Systems) and Streptavidin sepharose (GE Healthcare) overnight at 4°C. Streptavidin sepharose was washed three times with PBS and proteins eluted by boiling in 2× Laemmli sample buffer supplemented with beta mercaptoethanol for 10 min at 95°C.

Proteins were separated by SDS-PAGE and transferred either onto polyvinylidene difluoride membranes (Hybond P; Amersham Biosciences) or Nitrocellulose membranes (GE Healthcare) and probed with rat anti-mouse TREM2 (clone 5F4; Xiang et al., 2016), rabbit anti-GFAP (Dako), rabbit anti-IBA1 (Wako), rabbit anti-TSPO (Abcam) and anti-GAPDH (Sigma). Bound antibodies were visualized by corresponding HRP-conjugated secondary antibodies using enhanced chemiluminescence technique (Pierce). All buffers were supplemented with protease inhibitors (Sigma).

sTREM2 and cytokines (MSD proinflammatory panel) were quantified in TBS fractions from mouse brain using the Mesoscale platform similar to previously described methods (Kleinberger et al., 2014). The MSD proinflammatory panel was performed according to manufacturer’s recommendations. For sTREM2, streptavidin-coated 96-well small-spot plates were blocked overnight at 4°C in blocking buffer (3% bovine serum albumin and 0.05% Tween-20 in PBS, pH 7.4). For the detection of mouse sTREM2, plates were incubated for 1 h at RT with 0.25 μg/ml biotinylated polyclonal goat anti-mouse TREM2 capture antibody (R&D Systems; BAF1729) diluted in sample diluent (1% bovine serum albumin and 0.05% Tween-20 in PBS, pH 7.4). Plates were washed subsequently for four times with washing buffer (0.05% Tween-20 in PBS) and incubated for 2 h at RT with samples diluted 1:2 in assay buffer supplemented with protease inhibitors (Sigma). A recombinant mouse TREM2 protein (Hölzel Diagnostika) was diluted in assay buffer in a two-fold serial dilution and used for the standard curve (Concentration range: 2000–31.25 pg/ml). Plates were washed three times with washing buffer before incubation for 1 h at RT with 1 μg/ml rat monoclonal anti-TREM2 antibody (R&D Systems, MAB1729) diluted in assay diluent. After three additional washing steps, plates were incubated with 0.5 μg/ml SULFO-TAG-labeled anti-rat secondary antibody (Meso Scale Discovery) and incubated for 1 h at RT. Lastly, plates were washed three times with wash buffer followed and developed by adding 1X Meso Scale Discovery Read buffer. The light emission at 620 nm after electrochemical stimulation was measured using the Meso Scale Discovery Sector Imager 2400 reader. Calculation of the concentration of sTREM2 was performed with the MSD Discovery Workbench v4 software (MSD).

Total Aβ was quantified in TBS-Triton fractions similarly as described above using biotinylated 2D8 as capture antibody, mouse monoclonal 4G8 as detection antibody and SULF-TAG-labeled anti-mouse secondary antibodies. A recombinant human Aβ1–40 peptide (MSD) was diluted in assay buffer supplemented with 6% fetal calf serum in a three-fold serial dilution and used for the standard curve (Concentration range: 33–46 pg/ml).

Cerebral hemispheres intended for immunohistochemistry (random side) were fixed by immersion in 4% paraformaldehyde at 4°C for 24 h. Two representative 50 μm thick slices per animal were then cut in the axial plane using a vibratome (VT 1000 S, Leica, Wetzlar, Germany). Free-floating sections were permeabilized with 2% Triton X-100 overnight and blocked with 10% normal goat serum. We obtained immunohistochemical labeling of microglia using an Iba1 primary antibody (dilution; Wako, 1:200), and the A-21244 secondary antibody (Invitrogen, 1:500). The TSPO staining was obtained by incubation with anti-PBR antibody (Abcam, 1:300) and the A-11034 secondary antibody (Invitrogen, 1:500). Fibrillary Aβ plaques were stained with the fluorescent dye methoxy-X04 (0.01 mg ml⁻¹ in phosphate-buffered saline at pH 7.4 for 15 min) as described previously (Brendel et al., 2015b). The unbound dye was removed by three washing steps with PBS, and the slices were then mounted on microscope slides with fluorescent mounting medium (Dako, Germany). Images were acquired on an epi-fluorescence microscope (Axio Imager.M2 with ApoTome.2, Jena, Zeiss, Germany). Imaging of the whole slice was performed in tile scan mode, which allows automatic stitching of an array of fields of view. The area and number of plaques and microglia were automatically counted using Imaris software (Imaris 7.6.5; Bitplane, Zurich) in the entire cerebral cortex. These analyses were performed by an operator who was blind to the μPET results.

Univariate analysis of variance (ANOVA) with Tukey post hoc test was performed to compare values of different age groups in PS2APP mice or to compare PS2APP mice against WT littermates. Student’s t-test was used to compare results in groups of young and aged WT. Normal distribution of data was verified by the Kolmogorov-Smirnov test. For VOI-based correlation analyses of μPET-derived estimates for microglial activation and amyloidosis with sTREM2, Pearson’s coefficients of correlation (R) were calculated. Multiple regression analysis was performed to investigate the association of μPET-derived estimates for microglial activation and amyloidosis with sTREM2 levels with age as a fixed effect. The standardized regression coefficients (β) are reported. IBM SPSS Statistics (version 22.0; SPSS, Chicago, IL, USA) was used for all statistical tests. A threshold of p < 0.05 was considered to be significant for rejection of the null hypothesis. Standardized differences between PS2APP and WT mice were computed for the defined time points in PS2APP mice (under the assumption that measures in WT with only two time-points are developing linearly during aging) and divided by the standard deviation. For TBS Aβ levels at 5 and 8 months of age, a combined standard deviation was used as an approximation for both time points as the singular results were obviously distorted by the low standard deviation of mice aged 5 months.

Two independent studies have recently shown a significant increase of CSF sTREM2 as a function of age (Piccio et al., 2016; Suárez-Calvet et al., 2016b). Furthermore, transgenic amyloid mouse models indicate an age dependent increase of TREM2 mRNA expression in the brain (Matarin et al., 2015). Hence, we first investigated the levels of sTREM2 in post mortem brain homogenates of PS2APP mice (Richards et al., 2003; Ozmen et al., 2009) and WT littermates at different ages. ELISA based quantification of sTREM2 in TBS extracts showed a significant age-dependent increase of sTREM2 in the non-transgenic C57BL/6 mice (Figure 1A; +19.2%; p < 0.05), which is in line with an age-dependent increase of sTREM2 seen in cognitively normal humans (Suárez-Calvet et al., 2016b). Similarly, the TSPO μPET signal also showed a small but significant increase in aged WT mice (Figure 1B; +4.0%; p < 0.05).

In line with our previous μPET findings of increased fibrillar amyloid deposition and glial activation in PS2APP mice with age (Brendel et al., 2016), we observed a significant progressive increase of sTREM2 (5 vs. 8 months: +82%, p < 0.05; 5 vs. 13 months: +165%, p < 0.001; 5 vs. 16 months: +211%, p < 0.001; Figure 1A) which was far in excess of the age-dependent increases in WT animals. Quantitation of the proinflammatory cytokines Interleukin 1b, Interleukin 6 and KC/GRO (Interleukin 8 related protein) as well showed age-dependent increases in PS2APP, however to a lesser extent compared to sTREM2.

In parallel, the TSPO μPET signal indicated the known increase with age across the lifespan of PS2APP mice (Figure 1B). Total Aβ advanced with age at similar extent when compared to sTREM2 (5 vs. 8 months: +59%, p = n.s.; 5 vs. 13 months: +165%, p < 0.001; 5 vs. 16 months: +223%, p < 0.001; Figure 1C). Fibrillary plaque signal assessed by Aβ μPET showed a delayed increase in the entire forebrain from 8 months onwards (Figure 1D).

Figure 2 illustrates the standardized differences of biochemical and imaging measures for the PS2APP mouse model during their life course. For this purpose, age dependent differences between PS2APP and WT mice were computed (under the assumption that measures in WT with only two time-points are developing linearly during aging) and divided by the standard deviation in PS2APP mice. Here standardized differences of TSPO-PET and sTREM2 as well reached or even exceeded the level of fibrillar Aβ deposition and total Aβ. Standardized sTREM2 differences clearly exceeded those of common cytokines at all time points. In summary, the standardized timelines of sTREM2, TSPO-PET, Aβ-PET and total Aβ were characterized by a high concordance.

Together, these results suggest that microglia activation, as measured by TSPO μPET signal and sTREM2 measurement, increase with normal aging, and this increase is strongly augmented by amyloid pathology.

Given the age-dependent increases of the several biomarkers, we subsequently assessed the correlation between sTREM2 levels and μPET results. The concentration of sTREM2 in the brain showed a strong correlation with the TSPO μPET signal in PS2APP mice (R = 0.89, p < 0.001; Figure 3A). The same analysis in WT mice also indicated a positive correlation, although less pronounced (R = 0.72, p < 0.05; Figure 3B). A strong positive correlation was as well observed between sTREM2 level and fibrillar amyloidosis as assessed by Aβ μPET in PS2APP (R = 0.92, p < 0.001; Figure 3C) but not in WT mice (Figure 3D). The correlation between total Aβ and sTREM2 levels (R = 0.95; p < 0.001) further supported the strong association between sTREM2 and amyloidosis.

Next, we asked whether brain levels of sTREM2 are associated with TSPO and Aβ μPET signals independently from age. To this end, we performed a multiple regression analysis with sTREM2 as an outcome variable, with TSPO μPET and Aβ μPET as predictors, and with introduction of age as a covariate. In this analysis, forebrain TSPO μPET signal was significantly associated with sTREM2 (β = 0.40; p < 0.05), after adjusting for age; forebrain TSPO μPET signal and age together accounted for 84% of the variance in sTREM2 (F_(2,22) = 61.5, p < 0.0001, R² = 0.85, R²_Adjusted = 0.84). Forebrain Aβ μPET signal was also significantly associated with sTREM2 (β = 0.53; p < 0.01), after adjusting for age; forebrain Aβ μPET signal and age together accounted for 86% of the variance in sTREM2 (F_(2,22) = 77.5, p < 0.0001, R² = 0.88, R²_Adjusted = 0.86).

In summary, we found a strong correlation of sTREM2 levels and μPET findings of microglial activation and fibrillar amyloidosis in PS2APP mice, a correlation that was highly driven by age. Nonetheless, significant associations between sTREM2 levels and uptake of both μPET tracers still remained, even upon exclusion of age as a nuisance variable.

In the second step of image analysis, we used voxel-wise dual tracer μPET results for a more detailed mapping of the sequence of neuropathological alterations in the PS2APP mice. In particular, we used SPM to test for spatial and temporal relationships between neuroinflammation and amyloidosis as measured by μPET in transgenic vs. age-matched WT mice. Additional histochemical analyses were undertaken in order to support the detection of a very early stage of fibrillar amyloid deposition in the younger PS2APP mice, where detection thresholds of μPET might be too low to detect the presence of early pathology.

In this analysis, TSPO μPET revealed relevant microglial activation (11% of the total brain volume; p < 0.05, FDR-corrected) in PS2APP mice as young as 5 months, with main foci in the frontal and parietal cortices, and also in the cerebellar hemispheres (Figure 4A). At this early age, punctate fibrillar amyloid depositions were detected by methoxy-X04 histology in the frontal cortex of all PS2APP mice (Figure 5A), but these were not sufficiently pronounced to impart a [¹⁸F]-florbetaben Aβ μPET signal robust to correction for multiple comparisons.

At 8 months of age, nearly one third of the brain volume in PS2APP mice had increased TSPO μPET signal (32%; p < 0.05, FDR-corrected; Figure 4B). Most of the regions with new TSPO μPET specific binding arising between 5 and 8 months were in the frontal, piriform and entorhinal cortical areas. 7% of the brain volume had significantly elevated Aβ μPET signal at 8 months, with foci mainly in the frontal cortex and thalamus (p < 0.05, FDR-corrected; Figure 4B), of which total 3% of the total brain volume corresponded to regions with overlapping TSPO and fibrillar Aβ increases relative to corresponding μPET results in WT mice. Immunohistochemical analysis in the same animals showed numerous small fibrillar plaques, generally matching the regions with elevated amyloid μPET signal, and to microscopic examination, surrounding IBA-1 and TSPO positive cells (Figure 5B).

Together, these results indicate that microglial activation is already established in large parts of the brain in young PS2APP (5 and 8 months) mice, whereas fibrillar Aβ deposition is less pronounced and restricted to specific areas. Inflammatory processes clearly surround the establishment of early plaques, but occur as well in areas without manifest fibrillar Aβ deposition in young PS2APP mice, in which over-expressed Aβ is probably present only in soluble form.

Next, we mapped the regional progression of TSPO and Aβ μPET signals in PS2APP mice aged more than 8 months in order to discern the spatial relationship of the two biomarkers later in life. We noted the enlargement of areas with overlapping increases detected by dual tracer μPET in the forebrain, where the total brain volume of concerted microglial activation and amyloidosis increased from only 3% at 8 months to 24% at 13 months and 37% at 16 months (Figures 4B–D). Thus, more than one third of the entire brain volume, notably in the forebrain cortices, thalamus and hippocampus, manifested concerted microglial activation and amyloidosis at 16 months of age. This result was further strengthened by the increase in dice similarity coefficients from initially low levels (1% similarity at 5 months and 12% similarity at 8 months), progressing to scores indicating very high agreement between brain voxels with elevated TSPO and Aβ μPET binding in aged PS2APP mice (63% similarity at 13 months and 69% similarity at 16 months of age).

Immunohistochemical analyses confirmed the μPET findings and indicated an age-dependent increase of fibrillary Aβ plaques surrounded by numerous IBA-1 and TSPO positive cells (Figures 5C,D). The WT mice themselves indicated an age-dependent increase of IBA-1 positive cells (Figures 5E,F), confirming the findings of a slight age-dependent elevation of TSPO μPET signal together with increasing sTREM2 levels in aged WT mice (Figures 1A, 3B).

We conclude from these voxel-wise dual tracer μPET results that TSPO activity and fibrillar amyloidosis progress in a concerted manner in PS2APP mice aged more than 1 year, ultimately co-expressing in most of the forebrain. In contrast, WT mice showed only a subtle age-dependent neuroinflammation, as revealed by IBA-1 positive cells by immunohistochemistry. This result in WT mice confirmed the significant increases in TSPO μPET signal between 6 and 18 months of age.