[18F]Flotaza for Aβ Plaque Diagnostic Imaging: Evaluation in Postmortem Human Alzheimer’s Disease Brain Hippocampus and PET/CT Imaging in 5xFAD Transgenic Mice

Figure 2A shows the scan of the CA1 subiculum of one female AD subject, showing the GM and WM regions. Extensive amounts of neurotic, cored, and diffuse Aβ plaques (Figure 2B,C) were immunostained and consistent with previous reports of high levels of immunostained Aβ plaques in this brain region [19,22]. The brain slice in Figure 2A was labeled by [¹⁸F]flotaza and shows high levels of binding in the GM, while the WM regions were near background levels (Figure 2D). Analysis by QuPath of the immunostained slice in Figure 2B provided the pixel threshold image shown in Figure 2E. This was used to provide a quantitative assessment of Aβ positivity, which was plotted using quantitative measures of [¹⁸F]flotaza binding (Figure 2F; [17]). The regions of WM and GM (regions 1 through 7 highlighted in Figure 2D) were plotted, showing a high correlation between [¹⁸F]flotaza binding and Aβ positivity within the brain slice.

As seen in Figure 2G, all female AD subjects (n = 15) exhibited high [¹⁸F]flotaza binding in the GM regions and very low WM binding in the hippocampus. When compared to the other subjects, two subjects (AD 97-22 and AD 13-46) had lower levels of GM binding. Since all subjects were expected to have high Aβ plaque load (Table 1), these deviations may be due to heterogeneity within the specimens. Figure 2B,D,F suggest that the subiculum also has high levels of [¹⁸F]flotaza binding and is consistent with previous Aβ immunostain findings of the human subiculum [23]. It may also be noted that some of the variability in the [¹⁸F]flotaza binding may be due to variations in the tissue sections.

Male AD subjects also revealed extensive amounts of Aβ plaques. One subject shown in Figure 3B,C had an abundant amount of neurotic, cored, and diffuse Aβ plaques (Figure 3B,C). [¹⁸F]Flotaza was able to consistently bind to these Aβ plaque sites (Figure 3D). Through QuPath analysis, a generated pixel threshold image shown in Figure 3E was able to provide a quantitative assessment of Aβ positivity, which was plotted with quantitative measures of [¹⁸F]flotaza binding (Figure 3F). Both WM and GM regions (regions 1 through 8 highlighted in Figure 3D) were plotted with a high correlation between [¹⁸F]flotaza binding and Aβ positivity. All male AD subjects (n = 13) exhibited high [¹⁸F]flotaza binding in the GM regions, as seen in Figure 3G. Male AD subjects exhibited greater binding of [¹⁸F]flotaza in the WM regions compared to the female subjects. This higher WM binding (and the lower WM binding in female subjects) appeared to be consistent in all the subjects (Figure 2G and Figure 3G). Similar to the female subjects, some of the variability in the [¹⁸F]flotaza binding may be due to variations in the tissue sections.

The extent of [¹⁸F]flotaza binding in male and female subjects GM was similar, and the difference was not statistically significant, as seen in Figure 4A. A lack of male–female differences have been previously reported in postmortem hippocampal findings using immunostaining methods [23]. As expected, the difference between GM and WM in all males and all females was highly significant (p < 0.001, Figure 4A). Interestingly, the WM difference between males and females was highly significant (p < 0.001), with females showing significantly lower WM binding (Figure 4B). The greater WM binding in males resulted in lower GM/WM ratios (Figure 4C) compared to the female GM/WM ratios, which were higher because of the lower WM binding (Figure 4D). The GM/WM ratio in male AD subjects ranged from 9.2 to 89.3, whereas in the female AD subjects, it ranged from 12.5 to 216. There was a poor correlation of age with the [¹⁸F]flotaza GM/WM ratios both for males (R² = 0.0543; Spearman’s correlation ρ = 0.1821, p-value = 0.5489; Figure 4C) and females (R² = 0.1698; Spearman’s correlation ρ = −0.2238, p-value = 0.4196; Figure 4D).

Of the 32 CN subjects used in the study, only one subject had significant [¹⁸F]flotaza binding (Figure 5A,B). About five subjects showed some decipherable binding above the background without any GM binding preference (Figure 5C,D), while other subjects had some decipherable GM binding higher than WM binding (Figure 5E,F). All other control subjects had no binding whatsoever, suggesting the high specificity of [¹⁸F]flotaza binding to Aβ plaques. The absence of Aβ plaques in the brain slices of these remaining CN subjects was confirmed by the absence of any anti-Aβ immunostaining.

All the subjects (CN and AD) used in this study were categorized in Braak stages I, II, III, V, and VI (none of the subjects were found to be in the Braak IV stage). Figure 5H shows a plot of [¹⁸F]flotaza GM/WM binding ratio in all the subjects with respect to Braak staging. Females had the highest ratios, with Braak stage VI having the maximum average (average ratios ~80). Similarly, males had high ratios in Braak stages V and VI, but the average ratio was about half of those found for females. As indicated before, the lower male ratio of GM/WM is driven by the higher WM binding seen in males compared to the female AD subjects. Braak stages I–III did not have significant [¹⁸F]flotaza binding.

Transgenic 5xFAD mice and wild-type C57BL/6J mice brain slices were used to evaluate the total binding of [¹⁸F]flotaza. The imaging protocol for the autoradiographic study of mice was similar to that of postmortem human brains. These brain slices had the frontal cortex, lateral septal nuclei, thalamus, hippocampus, and cerebellum (Figure 6). Extensive binding of [¹⁸F]flotaza was observed in cortical regions, including the frontal cortex and anterior cingulate, lateral septal nuclei, thalamus, and hippocampus, with very little binding in the cerebellum (Figure 6A). The ranking order of binding was: thalamus > lateral septal nuclei = hippocampus > frontal cortex > striatum > cerebellum. Both male and female mice had a similar order of [¹⁸F]flotaza binding (Figure 6E).

The binding of [¹⁸F]flotaza in the various brain regions strongly correlated with anti-Aβ immunostains for Aβ plaques, hence confirming the binding of [¹⁸F]flotaza to the regions that contained high Aβ deposition (Figure 6B). Closer inspection of the hippocampal region (Figure 6C) revealed high Aβ positivity, with a significant amount of [¹⁸F]flotaza occurring in the CA1 and subiculum regions. The thalamus Aβ positivity (Figure 6D) was more diffuse in nature and not as high as the hippocampus, but the binding of [¹⁸F]flotaza in the thalamus was very high. There was a good correlation between Aβ positivity measured in the adjacent immunostained sections and [¹⁸F]flotaza binding in the different brain regions, except for the thalamus, which appears as an outlier with high levels of [¹⁸F]flotaza binding (Figure 6E). In the wild-type C57BL/6J mice brain slices lacking the Aβ plaques, there was no [¹⁸F]flotaza binding (Figure 6F,G).

The ratio of [¹⁸F]flotaza binding in different brain regions versus the cerebellum as a reference region was high. In descending order, the ratios for females and males were: thalamus (TH/CB = 65♀; 41♂), hippocampus (HP/CB = 43♀; 35♂), lateral septal nuclei (LSN/CB = 37♀; 33♂), frontal cortex (FC/CB = 26♀; 16♂), and striatum (ST/CB = 6♀; 3♂). In this preliminary in vitro study, female mice showed higher ratios in all brain regions compared to the males. The order of [¹⁸F]flotaza binding ratios in the different brain regions was the same for females and males.

Mice (5xFAD and C57BL/6J) were injected retro-orbitally under anesthesia with [¹⁸F]flotaza and tolerated the procedure well. Intraperitoneal administration of [¹⁸F]flotaza was not suitable because of its lipophilicity, causing insufficient absorption in the bloodstream. Retro-orbital administration allowed sufficient distribution, although a significant amount of [¹⁸F]flotaza radioactivity remained in the eye orbit. After an uptake period of 90 min post-injection, a whole-body PET/CT scan was acquired on both groups of mice. Figure 7A shows a sagittal whole-body distribution of [¹⁸F]flotaza in the 5xFAD mouse post retro-orbital administration. Major organs that accumulated the radioactivity included the liver, urinary bladder, kidneys, and gastrointestinal tract. Inset shows an image with a lower PET intensity to demarcate the liver and urinary bladder. This distribution of [¹⁸F]flotaza is similar to reports for other pegylated derivatives [21,26], where clearance of the radiotracer is seen via the kidney and the gastrointestinal tract to the urinary bladder. The uptake of [¹⁸F]flotaza in the brain is lower compared to the rest of the body. An intravenous injection of [¹⁸F]flotaza may help provide greater levels of the radiotracer in the bloodstream and help in greater brain uptake.

In the 5xFAD mice brain, there was a reasonable amount of uptake in regions that are known to contain Aβ plaques in these mice, as confirmed by our in vitro experiments (Figure 7). Regional brain uptake of [¹⁸F]flotaza was higher in the 5xFAD mice (Figure 7B) compared to the C57BL/6J mice (Figure 7C). The measured ratio of [¹⁸F]flotaza binding in the PET/CT images in different brain regions versus the cerebellum as a reference region was higher in the 5xFAD mice compared to C57BL/6J mice (Figure 7). The ratios were: TH/CB = 1.4 (5xFAD), 1.11 (C57BL/6J); HP/CB = 1.25 (5xFAD), 0.96 (C57BL/6J); FC/CB = 1.27 (5xFAD), 1.0 (C57BL/6J). The highest SUVR in the thalamus is consistent with in vitro results shown in Figure 6, although significantly lower in the in vivo PET experiment. Optimization of the in vivo binding in the 5xFAD mice will be carried out in future experiments.

After the PET study of the 5xFAD mice, the brains of select mice were excised and sectioned to provide horizontal brain slices. Because of the longer half-life of fluorine-18, sufficient radioactivity remains in the brain to provide adequate autoradiographs, as shown in Figure 8 and previously reported by us for other radiotracers. Brain slices were exposed on phosphor screens. After radioactive decay, select slices were immunostained for Aβ plaques (Figure 8A,B). The corresponding ex vivo binding of [¹⁸F]flotaza is seen in Figure 8C,D, and the scanned slices in Figure 8E,F.

The rank order of binding of [¹⁸F]flotaza in the different brain regions was similar to both in vitro and ex vivo experiments (Figure 6E and Figure 8G). The thalamus was still the highest binding region, followed by the hippocampus and frontal cortex. Significant binding in the colliculi became apparent in the ex vivo slices. There were regional variations within the cerebellum. Although it displayed the lowest amount of [¹⁸F]flotaza binding, the cerebellar WM (Figure 8D,F) had a significant amount of nonspecific binding. Brain regions to cerebellum ratios (Figure 8H, TH/CB = 1.89, FC/CB = 1.54, HP/CB = 1.70 and COL/CB = 1.74) in the ex vivo slices were significantly lower compared to the in vitro experiments (where alcohol was used in the buffer to wash away nonspecific binding) but was higher than the in vivo PET experiments. The ex vivo ratios were somewhat greater compared to the in vivo PET ratios for the various brain regions because of the greater accuracy of analysis of brain regions.

Fluorine-18 was purchased from PETNET, Inc. (Makati, Metro Manila) Fluorine-18 labeled [¹⁸F]flotaza was prepared as reported previously [14]. Capintec CRC-15R dose calibrator and Capintec Caprac-R well-counter were used for radioactivity measurements. Thin layer chromatography of radioligands was scanned on an AR-2000 imaging scanner (Eckart & Ziegler, Berlin, Germany). Cyclone phosphor autoradiographic imaging system (Packard Instruments Co., Meriden, CT, USA) and Optiquant Imaging System software, version 5.0 were used for analysis. Specialty chemicals PIB was purchased from ABX Inc., Radeberg, Germany.

Human postmortem brain tissue sections of HP (CA1 plus subiculum), 10 mm thick on Fisher slides, were obtained from Banner Sun Health Research Institute (BHRI), Sun City, AZ, brain tissue repository for in vitro experiments [25]. Brain tissue samples from AD and cognitively normal (CN) subjects were selected based on their cognitive status during life and the presence or absence of end-stage pathology. Age and gender-matched AD brain and CN brain tissue samples were used for the study. A total of CN, n = 32; 16 males (age 71–97) and 16 females (age 53–95) and AD, n = 28; 13 males (age 70–91) and 15 females (age 59–93) were used in this study (Table 1). Brain sections were stored at −80 °C. All human studies were approved by the respective Institutional Review Boards. All postmortem human brain studies were approved by the Institutional Biosafety Committee of the University of California, Irvine.

Purified [¹⁸F]flotaza was used for autoradiographic studies [14]. Human brain sections (10 μm thick) were placed in a glass chamber and preincubated in a PBS buffer for 10–15 min. The brain sections were placed in glass chambers and incubated with [¹⁸F]flotaza (approximately 74–111 kBq/mL; 0.5–0.75 nM; specific activity > 35 GBq/μmol)) in 40% ethanol–PBS buffer at 25 °C for 1.5 h. The slices were then washed with cold PBS buffer (1 × 5 min), 60% ethanol–PBS buffer (2 × 5 min), PBS buffer (1 × 5 min), and cold deionized water (2 min), respectively. Nonspecific binding was measured in the presence of 10 μM PIB. The brain sections were air-dried and exposed overnight on a phosphor film (Multisensitive Medium MS, PerkinElmer, Waltham, MA, USA). The apposed phosphor screens were read and analyzed by OptiQuant acquisition and the Cyclone Storage Phosphor System (Packard Instruments Co., Boston, MA, USA). The regions of interest (ROIs) were drawn on the slices, and the extent of binding of [¹⁸F]flotaza was measured in DLU/mm² using the OptiQuant acquisition and analysis program.

Immunostaining of all brain sections was carried out by the University of California-Irvine Pathology services using Ventana BenchMark Ultra protocols. For Aβ plaques, slices from all subjects were immunostained with anti-Aβ Biolegend 803015 (Biolegend, San Diego, CA, USA), which is reactive to amino acid residue 1–16 of β-amyloid. All IHC-stained slides were scanned using the Ventana Roche instrumentation and analyzed using QuPath [17,53,54]. QuPath (version QuPath-0.4.2) was used for quantitative analysis of scanned brain slices.

Using QuPath, a pixel threshold was created to outline the IHC images. Several annotations were made for Aβ plaques in the grey matter regions of the IHC brain slices of each subject. Approximately 20–25 annotations were made for Aβ plaques by means of visual identification of being an Aβ plaque. Negative annotations (approx. 15) with no Aβ plaques were drawn in each subject. The pixel classifier was run (using either random trees or artificial neural networks), and the entire brain slice of each subject was generated by the pixel classifier and saved as a new image. To measure the area of Aβ plaques for each new pixel-classified brain slice, regions of interest (ROI) for grey matter and white matter were drawn on the new classified image. This ROI was then run through the classifier again and asked to measure all annotations. The area of Aβ plaques in the ROI, total area, and percent of Aβ were obtained for each brain slice [17]. This percent positivity of Aβ plaques was used to correlate with autoradiography measures.

All animal studies were approved by the Institutional Animal Health Care and Use Committee of the University of California-Irvine. The C57BL/6 adult male and female mice used in this study (28–40 g; 12-month-old) were purchased from Jackson Laboratory and housed under controlled temperatures of 22 °C ± 1 °C in a 12-h light–dark cycle starting at 6:00 a.m. with water and food chow ad libitum. The 5xFAD transgenic line of mice (MMRRC hemizygous 4 male and 4 female; 12-month-old) were purchased from Jackson Laboratory (female mice weighed 20–28 g and male mice weighed 26–38 g). All mice were housed in standard cages.

All experiments were carried out in accordance with the Institutional Animal Care and Use Committee at the University of California, Irvine, and were consistent with Federal guidelines. Male and female hemizygous 5xFAD mice obtained from MMRRC JAX were used for in vitro and in vivo studies. Horizontal brain slices were sectioned (10 μm thickness) on a Leica 1850 Cryostat and collected on Fisher slides. The slides contained three to four brain sections placed in separate glass chambers (six slides per chamber) and were preincubated in a PBS buffer for 15 min. The preincubation buffer was discarded. The brain slices were treated with [¹⁸F]flotaza in 40% ethanol–PBS buffer pH 7.4 (60 mL, 5 kBq/mL). The chambers were incubated at 25 °C for 1.25 h. Nonspecific binding of [¹⁸F]flotaza was measured in separate chambers using IBETA (10 µM). The slices were then washed with cold PBS buffer, 90% ethanolic PBS buffer twice, PBS buffer, and cold water. The brain sections were air-dried and exposed overnight on a phosphor film. The regions of interest (ROIs) were drawn on the slices, and the extent of binding of [¹⁸F]flotaza was measured in DLU/mm².

Animals had free access to food and water during housing. All animals were fasted for 18–24 h prior to PET imaging. In preparation for the scans, the mice were induced into anesthesia with 3% isoflurane. Inveon preclinical Dedicated PET (Siemen’s Inc., Munich, Germany) was used for the MicroPET studies, which had a resolution of 1.45 mm [55]. The Inveon PET and MM CT scanners were placed in the “docked mode” for combined PET/CT experiments (Siemens Medical Solutions, Knoxville, TN, USA). A Sigma Delta anesthetic vaporizer (DRE, Louisville, KY, USA) was used to induce and maintain anesthesia during injections and PET/CT acquisitions. [¹⁸F]Flotaza was taken in 10% alcoholic sterile saline (0.9% NaCl injection, United States Pharmacopeia) for injections into mouse models for PET/CT studies. [¹⁸F]Flotaza was injected retro-orbitally (0.9 MBq) under 2% isoflurane anesthesia. The mice underwent 30-min PET scans (90 min postinjection) in a supine position accompanied by a 7-min CT scan, which was used for reconstruction and attenuation correction of PET images. The CT images were reconstructed with a cone beam algorithm (bilinear interpolation, Shepp–Logan filter) into 480 × 480 × 632 image arrays with a 206 μm pixel size. Following the reconstruction, the CT images were spatially transformed to match the PET images. In addition to being reconstructed into an image, these CT data were used for attenuation correction of PET images. PET imaging data were quantified as standard uptake values (SUVs), and using values of the nonspecific binding region, cerebellum, ratios (SUV ratios, SUVRs) of [¹⁸F]flotaza were calculated using our reported procedures for [¹²⁴I]IBETA [15].

After [¹⁸F]flotaza in vivo PET/CT experiments in control and 5xFAD mice, select mice were killed, and the brain was excised. The brains containing [¹⁸F]flotaza were rapidly frozen at −80 °C for approximately 10 min, followed by 10 min in the Leica cryotome at −20 °C. Similar procedures were previously reported for other radiotracers involved in correlative studies between in vivo and ex vivo measurements of brain uptake in small animal models [56]. Horizontal brain sections (10 µm to 40 μm thick) containing the thalamus, subiculum, cortex, striatum, hippocampus, and cerebellum were cut using the Leica cryotome. The sections were rapidly air-dried and subsequently exposed to phosphor films overnight. Adjacent sections were immunostained with Aβ-antibodies. The regions of interest (ROIs) were drawn on the slices, and the extent of binding of [¹⁸F]flotaza was measured in DLU/mm².

All regions of interest (ROIs) in the GM and WM autoradiographic images of [¹⁸F]flotaza were quantified using measures (DLU/mm²). The immunostained sections were analyzed using QuPath. GM and WM binding of [¹⁸F]flotaza in AD and CN subjects were measured, and GM/WM ratios of AD and CN were compared. Using the ratio method would be akin to in vivo PET methods, where the SUV between the target region is compared to a nonspecific binding reference region as a ratio and expressed as SUVR [15,19].

All in vivo images were analyzed using Inveon Research Workplace (IRW) software version 2.1, (Siemens Medical Solutions, Knoxville, TN, USA) and PMOD Software version 3.0 (PMOD Technologies, Zurich, Switzerland). Whole-body PET/CT images were analyzed using the IRW software for [¹⁸F]flotaza uptake and any other CT anomalies in the whole-body images. For additional brain quantitative analysis, brain images were also analyzed using PMOD, with PET images co-registered to a mouse brain MRI template [19]. The magnitude of [¹⁸F]flotaza in each volume of interest, VOI (in kBq/mL), was measured. The cerebellum was used as a reference region to calculate the ratio of target brain regions to reference regions.

Group differences between AD and CN subjects were assessed using average GM/WM ratios and were determined using Microsoft Excel 16 and GraphPad Prism 9. Statistical power was determined with Student’s t-test, and p-values of <0.05 were considered to indicate statistical significance. Spearman’s correlation was carried out to assess the effects of aging.

The data that supports the findings of this study are available from the corresponding author for discussions upon reasonable request.