In the field of neurodegenerative diseases, mice are the most commonly used animals for their biological features that are similar to those of humans, easily manipulated genetics to mimic human conditions and diseases, and a relatively short life span (1.5–2 years). Because mice do not develop AD naturally, transgenic mice are generated to recapitulate certain AD pathological features to fit the research needs (Table 1). The genes associated with the early onset of AD have been the main targets for transgenic manipulations.

Compared to mice, rats are easier to handle and have larger brain sizes for easier surgical operation and imaging analysis (Ellenbroek and Youn, 2016). Both genetic and non-genetic rat AD models have been developed (Table 3). However, unlike transgenic AD mouse models, not as many AD rat models are available for scientific research and rats appear to be more resilient to AD pathology than mice (Charreau et al., 1996).

Non-human primates such as rhesus macaques and marmosets are not known to develop AD but do accumulate Aβ deposits and show tauopathy in their aged brains (Paspalas et al., 2018; Haque and Levey, 2019; Arnsten et al., 2021b; Datta et al., 2021; Leslie et al., 2021). Intracranial injection of Aβ₄₂ and thiorphan, an inhibitor of neprilysin that is responsible for Aβ clearance, has been employed to generate an AD model in middle-aged (16–17 years) rhesus monkeys (Li et al., 2010). Significant intracellular accumulation of Aβ was found in the neurons of the basal ganglia, cortex, and hippocampus, accompanied by neuronal atrophy and loss. Two injections of an adeno-associated virus expressing a double tau mutation (AAV-P301L/S320F) in the left hemisphere of rhesus monkeys result in misfolded tau propagation similar to that in humans. Tau spreading is accompanied by robust neuroinflammatory response driven by TREM2 + microglia, with biomarkers of inflammation and neuronal loss in cerebrospinal fluid and plasma (Beckman et al., 2021).

Other non-primate large animals used for AD modeling include domestic animals such as dogs and cats as well as farm animals including pigs, sheep, and cows. Aged dogs develop plaque pathology and cerebral amyloid angiopathy (Yu et al., 2011), as well as a dementia-like syndrome resembling human AD (Prpar Mihevc and Majdic, 2019; Abey et al., 2021). Tau dysfunction and tangles have been reported and associated with cognitive decline (Yu et al., 2011; Schmidt et al., 2015; Smolek et al., 2016). Cats also develop plaques, tangles, and brain atrophy along with cognitive decline as they age (Chambers et al., 2015; Fiock et al., 2020). Two transgenic pig models of AD have been reported using minipigs. The first one carries an hAPP transgene with the Swedish mutation driven by the human BDGFβ promoter, resulting in high levels of brain-specific Aβ expression (Kragh et al., 2009), and the second minipig model carries three copies of a transgene expressing the 695 variant of hAPP with the Swedish mutation and a human PSEN1 transgene with the M146L mutation (Jakobsen et al., 2016). Intraneuronal accumulation of Aβ_1–42 was detected in two pigs: one at 10 months and one at 18 months. Plaque- and tangle-like pathologies have also been seen after traumatic brain injury (TBI) in pigs (Hoffe and Holahan, 2019). Tau pathology and Aβ plaques have been identified in aged sheep and goats as well (Braak et al., 1994).

The development and validation of the first-in-class Aβ PET radiotracer, the thioflavin T-derived Pittsburgh compound B ([¹¹C]PIB or PIB), was a milestone in AD imaging. It not only allows the direct in vivo visualization and quantification of Aβ plaque in living subjects (Klunk et al., 2004) but also paves the road for the development and FDA approval of its ¹⁸F-labeled analog ([¹⁸F]flutemetamol), the stilbene derivative [¹⁸F]florbetaben, and the styrylpyridine derivative [¹⁸F]florbetapir, the use of which have become impactful in AD clinical trials and diagnosis. The intrinsic fluorescent characteristics of these imaging probes and their analogs allow for the microscopic assessment of their binding selectivity and binding preference to different forms of Aβ plaques and Aβ plaques at different locations (e.g., parenchymal and cerebral amyloid angiopathy, CAA) (Bacskai et al., 2003; Fodero-Tavoletti et al., 2012).

Many Aβ imaging tracers have been evaluated using multiple different AD animal models, mainly in AD mice. PIB has been tested in AD mice of APPswe, APP/PS1, 3 × Tg, 5 × FAD, Tg2576, and APP23 (Ni, 2021). Initial reports on PIB binding in Tg2576 and APP/PS1 mice at advanced ages were negative, even with abundant Aβ pathology (Klunk et al., 2005; Toyama et al., 2005); while PIB binding in APP23 mice was positive (Maeda S. et al., 2007). These data led to the hypothesis that the paucity of high-affinity binding sites for PIB in murine Aβ plaques requires very high molar activity PIB for successful imaging in murine AD models. Snellman et al. (2013) compared PIB uptake longitudinally in the brains of multiple AD mouse models and found higher PIB uptake in the cortex of APP23 mice compared with the wild-type controls and no difference in APP/PS1 and Tg2576 mice with their corresponding controls, consistent with previous results from other groups (Figure 2A). They also compared the thioflavin-T staining patterns and found that APP23 mice form large and compact human-like Aβ deposits, whereas Tg2576 mice and APP/PS1 mice form sparse fibrillar deposits. The results suggest that PIB binding is highly dependent on the AD model and the associated higher-order fibrillar structure rather than simple β sheets. At a microscopic level, the Aβ plaques formed in early onset autosomal dominant AD and sporadic AD brains have different levels of non-fibrillar Aβ species (Querol-Vilaseca et al., 2019), and the Aβ deposits in familial AD, sporadic AD, and cerebral amyloid angiopathy manifest different conformations (Condello et al., 2018). Further understanding of the interactions of the imaging probes with amyloid plaques of different forms will help with the development of probes targeting the various forms of misfolded Aβ proteins in the brain (Biancalana and Koide, 2010).

One of the biggest advantages of small animal PET imaging is the longitudinal tracking of the pathogenesis and therapeutic effects of experimental drugs. This was demonstrated by the longitudinal PET imaging studies in AD animal models (Maeda J. et al., 2007; Deleye et al., 2017; Snellman et al., 2017). The challenges in imaging Aβ plaques in AD animal models are due to the different forms of plaques and disposition patterns in different animal models and at different ages of the same animals (Snellman et al., 2013). Other challenges are the quantification of the PET signals. For the quantitative analysis of human Aβ PET imaging data, the cerebellum was chosen as the reference region to generate distribution volume ratio (DVR) or standardized uptake value ratio (SUVR) because of the lack of specific binding of PIB in the human cerebellum (Lopresti et al., 2005; Price et al., 2005). However, there are emerging effective drugs targeting other pathological pathways and that do not alter Aβ plaque levels, e.g., Fyn inhibitor and mGluR5 silent allosteric modulator (SAM) (Kaufman et al., 2015; Haas et al., 2017). Thus, the objective assessment of their treatment effects needs different imaging biomarkers that are closely related to synaptic/functional recovery rather than Aβ plaque levels. The current consensus considers Aβ oligomers as the primary cause of the neurotoxicity derived from abnormal amyloidosis. Thus, the development of an imaging agent targeting Aβ oligomers is highly desirable, albeit challenging.

Large molecules such as antibodies have been developed for Aβ PET imaging. Sehlin et al. (2016) and Fang et al. (2019) engineered the ¹²⁴I-labeled Aβ antibodies [¹²⁴I]mAb158 and Di-scFv [¹²⁴I]3D6-8D3, respectively, to detect the soluble Aβ in the tg-ArcSwe (AβPP E693G) and Swedish (AβPP KM670/671NL) mouse models with clearly visualized Aβ in the brain. The brain PET imaging shows a correlation between the PET signal and the levels of soluble Aβ aggregates. An increased SUVR of 2.2–3.5 in AD mice brains was reported compared to the wild-type brains. The evaluations of other amyloid imaging tracers in AD models have been summarized nicely in a recent review (Ni, 2021).

Targeting another hallmark of AD, tau neurofibrillary tangle is an extremely exciting area of PET tracer development. Tau is an axonally enriched microtubule-associated protein (MAP) that accumulates in the temporal and parietal neocortex in AD brains (Hung et al., 2016; DeTure and Dickson, 2019). Tau exists as six different isoforms, which contain either 3 or 4 microtubule-binding repeats (3R or 4R). The hyperphosphorylation and aggregation of tau with different repeats are involved in different neurodegenerative diseases, e.g., AD (3R/4R), Pick disease (3R), and progressive supranuclear palsy (4R). Postmortem histopathological studies demonstrated that NFTs are a better index of disease severity and progression than Aβ for patients with AD (Shoghi-Jadid et al., 2002). Tau pathology appears earlier than the Aβ plaque in human brains (Arnsten et al., 2021a). Tau-PET imaging allows the detection of tauopathy and highly predicts subsequent cognitive decline in both asymptomatic and symptomatic individuals (Leuzy et al., 2019; Wang and Edison, 2019; Beyer and Brendel, 2021).

Brain [¹⁸F]FDG PET primarily indicates synaptic activity. [¹⁸F]FDG uptake strongly correlates at autopsy with levels of the synaptic vesicle protein synaptophysin (Rocher et al., 2003). The degree and regional extent of hypometabolism measured by [¹⁸F]FDG-PET roughly correlate with the overall severity of cognitive impairment in AD. There is a close correlation between the regional accumulation of a tau-PET tracer ([¹⁸F]AV1451) and [¹⁸F]FDG hypometabolism (Rubinski et al., 2020). Along with amyloid imaging, [¹⁸F]FDG PET has been applied in multiple AD rodent models such as APPswe (Tg2576), 5 × FAD, APP/PS1, 3 × Tg, Tg4-42, TASTPM mice, and McGill-R-Thy1-APP rats (Waldron et al., 2015b; Bouter et al., 2018; Bouter and Bouter, 2019). Varying [¹⁸F]FDG PET results were found in Tg2576 mice. No differences in cerebral glucose metabolism were found in Tg2576 compared to WT mice in Kunter’s study (Kuntner et al., 2009), while Luo et al. (2012) found an increase in the FDG uptake in 7-month-old Tg2576, and Coleman et al. (2017) reported reduced FDG uptake in 18-month-old mice. Two separate [¹⁸F]FDG PET studies using 12-month-old APPPS1-21 mice reached the same conclusion that FDG uptake was reduced in the brain (Waldron et al., 2015a; Takkinen et al., 2017). Using APP/PS1 mice, both Poisnel et al. (2012) and Li et al. (2016) showed an age-dependent increase in glucose metabolism. In addition, PS2APP mice showed increased [¹⁸F]FDG uptake at 5 and 16 month (Brendel et al., 2016) and TASTPM mice were found to have decreased FDG uptake at 9 and 14 months of age (Waldron et al., 2015a,2017; Deleye et al., 2016). Contradictory results were reported in 5xFAD mice, with Rojas et al. (2013) reporting increased uptake of [¹⁸F]FDG in 11-month-old 5xFAD, and Macdonald et al. (2014) showing decreased uptake in 13-month-old mice. Sancheti et al. (2013) also reported decreased FDG uptake in 3xTg mice.

Clinical FDG PET imaging studies have shown promise in detecting early AD as neurodegeneration in certain brain regions (temporoparietal predominantly) is reflected by hypometabolism of FDG (Cohen and Klunk, 2014), and the hypometabolism pattern could serve as a predictive biomarker for conversion from MCI to AD dementia (Sala et al., 2020) earlier than MRI (Laforce<suffix>Jr.</suffix>, Soucy et al., 2018). However, there has been no suitable method to distinguish the FDG signal contributed by neuronal activity and immune cell activation, and thus the FDG PET signal could theoretically be influenced by two opposing forces, i.e., hypometabolism and neuroinflammation, at certain stages of AD pathogenesis and progression. With the recent development of PET imaging methods for synapse density (see section “PET Imaging of Synaptic Vesicle Glycoprotein 2A” for SV2A PET) and neuroinflammation (see section “PET Imaging of Neuroinflammation” for PET imaging of neuroinflammation), we are at a stage where we could potentially quantitatively attribute the FDG signals to synaptic and glial activities. This is of relevance in cases of MCI patients who show a positive correlation between Aβ PET and FDG PET.

Synaptic Vesicle Glycoprotein 2A is ubiquitously expressed in the neurons of the central nervous system and is widely used as one of the synaptic density biomarkers. Loss of synapses in the hippocampus and prefrontal cortex is implicated as an early pathological event in AD before the appearance of Aβ plaques and tau tangles and increasingly worsened during AD progression (Cai et al., 2019; Jackson et al., 2019).

[¹¹C]Levetiracetam was first developed but was not pursued in further imaging study (Cai et al., 2014). Nevertheless, it encouraged the development of SV2A ligands with much higher affinities, including [¹¹C]UCB-A (Estrada et al., 2016), [¹⁸F]UCB-H (Warnock et al., 2014; Bahri et al., 2017; Becker et al., 2017), and [¹¹C]/[¹⁸F]UCB-J (Cai et al., 2019; Li et al., 2019b). Among these, [¹¹C]UCB-J exhibited high brain uptake, fast and reversible tissue binding kinetics, and high specific binding signals in both non-human primates and humans (Finnema et al., 2016; Nabulsi et al., 2016). Most recently, [¹⁸F]SynVesT-1 (also known as [¹⁸F]SDM-8 (Li et al., 2019a) and [¹⁸F]MNI-1126 (Constantinescu et al., 2019) are developed and evaluated in non-human primates and humans (Li et al., 2021; Naganawa et al., 2021). Using APP/PS1 mice, Toyonaga et al. showed decreased [¹¹C]UCB-J uptake as compared to the WT mice, and treatment with the tyrosine kinase Fyn inhibitor saracatinib reversed this effect (Toyonaga et al., 2019). Sadasivam et al. (2019, 2021) found a lower [¹⁸F]SynVesT-1 signal in the whole brain of APP/PS1 mice, compared with wild-type mice (Figure 2B). However, in a study using [¹¹C]UCB-J in the tg-ArcSwe model and wild-type mice, a small but non-significant difference (∼5%) was found between the two groups, presumably due to large inter-animal variability (Xiong et al., 2021).

Microglia are macrophages in the brain that play an important role in neuroinflammation in AD. PET imaging of biomarkers of microglia provides insights into the time course of AD pathology. However, the diverse phenotypes of activated microglia and their different roles over the course of the AD trajectory make it challenging to develop radiotracers specific for neuroinflammation in AD.

The 18kDa translocator protein (TSPO) has been widely studied as a biomarker for microglial activation for over 20 years. Early radiotracers had disadvantages of low brain penetrability, low binding affinity for TSPO, the short half-life of the radioisotope, and sensitivity of binding affinity to gene polymorphisms (Zhou R. et al., 2021). [¹⁸F]DPA714 is one of the more recent radiotracers developed for TSPO. [¹⁸F]DPA714 was evaluated in APP/PS1 mice at different months to determine the role of microglia in the pathogenesis of AD neuroinflammation (Hu et al., 2020). Higher [¹⁸F]DPA714 uptake was noted in the cortex and hippocampus of 12–13 and 15–16-months-old but not younger AD mice compared with control mice. Another longitudinal PET study in APP23 mice used [¹⁸F]GE180 for TSPO imaging and [¹¹C]PIB for assessing amyloid deposition in ex vivo autoradiography experiments (Lopez-Picon et al., 2018). The APP23 model was chosen because of high [¹¹C]PIB binding in the brain of model mice compared with other AD models such as APP/PS1. AD mice were imaged with [¹⁸F]GE-180 at 17, 20, and 26 months of age. The binding of [¹⁸F]GE-180 plateaued in the frontal cortex and hippocampus regions in the early stage of AD, but amyloidosis increased throughout the later stages of AD (17–26 months of age). Thus, [¹⁸F]GE-180 appeared to be useful for tracking TSPO/neuroinflammation in early-stage AD but not for monitoring disease progression.

Compared with TSPO, colony-stimulating factor 1 receptor (CSF1R) expression in the brain is predominantly localized to microglia and low in other cell types. [¹¹C]CPPC was developed from a potent CSF1R inhibitor with an IC₅₀ of 0.8 nM (Horti et al., 2019) and evaluated in a mouse model of AD-related amyloidosis-overexpressing APP with Swedish and Indiana mutations (Melnikova et al., 2013). [¹¹C]CPPC had about 30% higher uptake in the cortex of AD mice compared with control mice at 40 min post injection. Significantly higher uptake in the hippocampus and cerebellum was also observed in the AD mice. Additionally, increased expression of CSF1R after LPS treatment and about 50% specific binding of [¹¹C]CPPC in LPS-treated mice were observed relative to sham controls. Autoradiography studies with [³H]CPPC demonstrated the lack of specificity of [³H]CPPC in brain tissues of LPS-treated Sprague-Dawley rats (Knight et al., 2021). In another study, the imaging performance of [¹¹C]CPPC was compared with that of [¹¹C]GW2580 in mouse models of acute and chronic neuroinflammation and a rhesus monkey (Zhou X. et al., 2021). In WT vs. APP-KI mice, [¹¹C]GW2580 demonstrated higher sensitivity than [¹¹C]CPPC, shown by a greater increase in [¹¹C]GW2580 uptake in the neocortex, forebrain, and striatum of APP-KI mice compared with that in WT based on SUVR measurements at 60–90 min. Blocking studies in the monkey showed higher specificity for [¹¹C]GW2580 over [¹¹C]CPPC.

TREM2 is a relatively new biomarker for microglial activation. Bispecific antibody scaffolds that bind to transferrin to enter the brain and to TREM2 were chemically conjugated and radiolabeled with relatively longer-lived radioisotopes. One example is ¹²⁴I-mAb1729-scFv8D3_CL, which was evaluated in Arc-Swe transgenic AD mice (Meier et al., 2021). While areas under the curve (AUC) for ¹²⁴I-mAb1729-scFv8D3_CL at 24–72 h post injection were higher in caudate, cortex, thalamus, and hippocampus of AD mice compared with control mice, significant differences in SUVs were not observed for the individual imaging timepoints. However, ex vivo binding studies through autoradiography with the radiotracer showed significant differences between the animal models. The lack of a significant difference in vivo was then attributed to the increased blood residence time of ¹²⁴I-mAb1729-scFv8D3_CL. Thus, radiolabeling of smaller antibody fragments is desirable to address the slow pharmacokinetic issue.

Another new biomarker for AD is the purinergic P2X ligand-gated ion channel type 7 receptor (P2X7R), which is involved in triggering parts of the AD neurodegenerative processes. P2X7R activates microglia in acute AD models (Sanz et al., 2009) and increases the production of chemokines mediated by Aβ peptide in chronic AD models (Martin et al., 2019). [¹⁸F]JNJ-64413739 was evaluated in a rat model of acute neuroinflammation. The uptake of [¹⁸F]JNJ-64413739 was found to be elevated in the LPS-treated site of the rat brain compared with the contralateral hemisphere of the brain treated with PBS (Berdyyeva et al., 2019). Biomarkers for neuroinflammation, such as higher mRNA levels of P2X7R, TSPO, and Aif1, were associated with the LPS-treated site. It was acknowledged that LPS treatment as a model of neuroinflammation is considered extreme and that this novel tracer warrants evaluation in rodent models of AD and other neurodegenerative diseases. Other PET tracers for P2X7R and other biomarkers of neuroinflammation are reviewed by Zhou R. et al. (2021). So far, the development of neuroinflammation imaging agents has been focused on targeting microglial activation, largely ignoring the other glial cell types. It would be instrumental to be able to distinguish the protective microglial activation in early AD from the later destructive phenotype to guide the proper timing of anti-inflammatory treatments.

Glutamate is the major excitatory neurotransmitter in the brain and acts on the ionotropic glutamate receptors (iGluRs) and metabotropic glutamate receptors (mGluRs) to regulate synaptic plasticity. iGluRs comprise three subfamilies: α-amino-3- hydroxy-5-methyl-4-isoxasolepropionic acid (AMPA) receptors, kainate receptors, and NMDARs (Traynelis et al., 2010). mGluRs are a family of G-protein-coupled receptors with 8 subtypes, mGluR1-8. Both iGluRs and mGluRs are found to be involved in synaptic malfunctions in AD (Avila et al., 2017; Foster et al., 2017; Wang and Reddy, 2017; Liu et al., 2019; Srivastava et al., 2020).

Several imaging tracers for glutamate receptors have been developed. So far, only the mGluR5 radiotracer [¹⁸F]FPEB has been evaluated in 5 × FAD mice (Lee et al., 2019), APP/PS1 mice (Varlow et al., 2020), and Tg-ArcSwe mice (Fang et al., 2017), with conflicting results: compared to wild-type animals, uptake of [¹⁸F]FPEB was found to be lower in 5 × FAD mice, with no difference in Tg-ArcSwe mice, and increased in APP/PS1 mice. Shimojo et al. (2020) observed that radioactivity signals derived from the other mGluR5 tracer (E)-[¹¹C]ABP688 were unaltered relative to controls at 2 months of age in rTg4510 mice but then gradually declined with aging in parallel with progressive brain atrophy.

The deficit in cholinergic neurotransmission is a prominent pathophysiological feature in AD. Dramatic loss of cholinergic neurons located in the basal forebrain increased levels of α₇ nicotinic acetylcholine receptor (α₇ nAChR) (Ikonomovic et al., 2009; Marutle et al., 2013) and decreased levels of M1 muscarinic acetylcholine receptor (M1 mAChR) (Yi et al., 2020) were found in the cortical regions of human AD brains (Ferreira-Vieira et al., 2016). The following PET imaging agents for cholinergic targets have been developed: 1) [¹¹C]NS14492 (Ettrup et al., 2011), [¹¹C](R)-MeQAA (Nishiyama et al., 2015), and [¹⁸F]ASEM (Gao et al., 2013) for α7 nAChR; 2) [¹¹C](+)3-MPB (Yamamoto et al., 2011) and [¹⁸F]fluorobenzyl-dexetimide (Rowe et al., 2021) for mAChR; 3) [¹¹C]LSN3172176 for M1 mAChR (Nabulsi et al., 2019); and 4) [¹¹C]MK-6884 M4 mAChR (Tong et al., 2020). Uptake of the α₇ nAChR tracer [¹¹C](R)-MeQAA was found to be increased in aged monkeys (Nishiyama et al., 2015), and lower uptake of the other α₇ nAChR tracer [¹⁸F]ASEM was seen in aged TgF334 rats compared with wild-type rats (Chaney et al., 2021). No difference was noted in the brain uptake of the acetylcholine esterase tracer [¹¹C]MP4A between the APP23 and wild-type mice at 10–13 months of age (Heneka et al., 2006).

Altered expression of endogenous cannabinoid receptor 2 (CB2), histaminergic receptors, sigma receptors, adenosine receptors (A1A and A2A receptors), and enzymes (BACE1, caspase 3, aspartryl cathepsin, and TrkB/C), as well as abnormalities in dopamine and serotonin neurotransmission, have been noted in AD. These provide additional targets for PET radiotracer development for AD imaging. The CB2 radiotracers [¹¹C]A-836339 have been tested in the LPS-induced neuroinflammation mouse model and the Appswe/PS1/dE9 mouse model (Horti et al., 2010), while [¹⁸F]JHU94620 has been tested in the LPS-induced neuroinflammation mouse model, which shows high-affinity binding to CB₂R and sufficient selectivity over CB₁R. A few tracers targeting caspase 3 and aspartryl capthepsin have been tested in AD model mice, with the majority of the tracers mainly tested in human subjects.

There has been great interest in imaging the compromised BBB in animal models of amyloidosis and patients with AD, as there is evidence of damaged BBB at the early stages of AD in patients and animal models. PET imaging of specific transporters and receptors expressed at the BBB was recently reviewed thoroughly by Ni⁹⁸.