Alzheimer’s disease (AD) is the most common form of dementia accounting for 60%–80% of all cases. The neurodegenerative disease is characterized by behavior and cognitive symptoms due to functional and structural abnormalities including accumulation of amyloid-β (Aβ) protein, neurofibrillary tangles and neuronal loss (Holtzman et al., 2011).

According to the amyloid cascade hypothesis the deposition of Aβ is the causative event of AD pathology (Hardy and Allsop, 1991). Aβ is generated through sequential cleavage of the amyloid precursor protein (APP) by β- and γ-secretases (Selkoe, 1998; Zhang et al., 2011). Next to Aβ peptides starting with an aspartate as the first amino acid, a variety of different N-truncated Aβ peptides have been identified in AD brains. N-terminal deletion enhances neurotoxicity and Aβ aggregation (Pike et al., 1995; Bayer and Wirths, 2014). There is substantial evidence that N-terminal truncated peptides play a key role in AD and they are highly abundant in brains of patients diagnosed with sporadic and familial AD (Jawhar et al., 2011; Bayer and Wirths, 2014; Dunys et al., 2018). Aβ4–42 is a particular abundant Aβ species in AD and one of the major fractions in the cortex and hippocampus of AD patients (Masters et al., 1985; Portelius et al., 2010). In order to further elucidate the toxicity of Aβ4–42 the transgenic mouse model Tg4–42 was generated. This AD mouse model overexpresses Aβ4–42 without human APP overexpression and without any mutations related to familial AD (Bouter et al., 2013). N-truncated Aβ4–42 triggers age-dependent neuron loss and behavior deficits comparable to AD-typical memory dysfunction albeit without plaque formation and neurofibrillary tangles (Antonios et al., 2015). While hippocampal atrophy caused by neuron death is one of the major hallmarks of AD progression only a small fraction of AD models show this key feature of AD (Bayer and Wirths, 2010).

The development of new treatment strategies against AD relies to a great extent on the use of transgenic mouse models. The massive neuron loss coupled with age- and disease-related cognitive decline make the Tg4–42 model well suited for preclinical testing of AD-relevant therapeutics. However, a suitable mouse model for preclinical trials also requires the validation of the same in vivo biomarkers that can be used in patients. To date, biomarkers for AD include measurement of Aβ42, total (t-tau) and phosphorylated (p-tau) in cerebrospinal fluid as well as positron-emission tomography (PET) and magnetic resonance imaging (MRI; Jack et al., 2013, 2018). PET displays a non-invasive imaging tool for the detection of Aβ deposition with amyloid PET or measuring cerebral glucose metabolism with ¹⁸F-Fluorodeoxyglucose (¹⁸F-FDG)-PET. ¹⁸F-FDG-PET is able to detect neuronal dysfunction and neurodegeneration in vivo as brain glucose metabolism is determined by synaptic activity mainly in order to restore membrane potentials (Kadekaro et al., 1985). In neurodegenerative disorders synaptic activity and glucose metabolism can be disturbed long before neuron loss occurs. ¹⁸F-FDG-PET can detect those changes in order to identify cerebral areas with reduced glucose metabolism. In the clinical routine ¹⁸F-FDG-PET is used to differentiate between different neurodegenerative diseases as certain patterns of reduced glucose uptake are specific for various disorders. A meta-analysis of Bloudek et al. (2011) evaluating FDG-PET in the diagnosis of AD showed a sensitivity of 91% and a specificity of 86%. Patients show an early and progressive reduction of glucose metabolism in the posterior cingulum, precuneus and temporoparietal cortex and hippocampus (Mosconi, 2013). Small animal PET scanners allow measuring the same molecular processes in animal models as in humans. Ideally functional, structural and molecular changes observed in patients should also be detected in a similar way in an animal model of AD.

In this study we used in vivo imaging with ¹⁸F-FDG-PET to evaluate the glucose metabolism in Tg4–42 mice under resting conditions. The aim of the present report was to investigate if the glucose metabolism could be an early biomarker to distinguish Tg4–42 animals from wild-type (WT) mice.

In order to develop and evaluate new therapeutic approaches an early diagnosis of AD is crucial. Brain imaging displays a promising tool for a non-invasive early detection of AD. PET is an imaging technique that allows in vivo monitoring of pathological and physiological metabolic processes (Shukla and Kumar, 2006). In clinical routine, ¹⁸F-FDG-PET is an established method for differential diagnosis of clinically suspected indistinguishable dementia including AD. The target of ¹⁸F-FDG is the cerebral glucose metabolism which is mainly determined by synaptic activity. Neuronal dysfunction occurs long before neuron loss and can therefore function as an early marker of neurodegenerative diseases. The use of ¹⁸F-FDG-PET in order to assess dementias relies on the detection of cerebral areas with reduced glucose metabolism. Disease-specific patterns of lower ¹⁸F-FDG uptake help identifying different forms of dementia. Typical AD cases already show reduced ¹⁸F-FDG uptake within the posterior cingulate cortex in early stages (Minoshima et al., 1997; Matsuda, 2001; Jack et al., 2013). Impaired glucose metabolism then proceeds to the precuneus, posterior temporal and parietal cortex, often in a bilateral matter. In advanced stages of AD hypometabolism can also extend to the frontal lobe (Minoshima et al., 1997; Silverman et al., 2001; Foster et al., 2007; Teune et al., 2010). In preclinical research assessment of new therapeutic strategies in a non-invasive way is essential. Brain imaging with ¹⁸F-FDG-PET appears suitable for longitudinal therapy monitoring, showing neuron loss and neuronal dysfunction as major hallmarks of AD in vivo. However, earlier approaches using ¹⁸F-FDG-PET to study brain disorders in mice displayed some obstacles. One issue is the brain size relative to the resolution of the preclinical PET scanners as the volume of the mouse brain is about 3,000 times lower compared to humans (Welch et al., 2013). However, modern scanners and new reconstruction engines show a spatial resolution around 700 μm allowing detailed structural information of the brain with an exact distribution of the tracer. Another issue is the quantification of regional ¹⁸F-FDG uptake. With the use of a standardized mouse brain atlas VOI template and MRI co-registration this issue can also be resolved providing reliable and reproducible data.

A preclinical model for AD should display many features of the disease similar to the patient in order to achieve a successful translation into the clinic. Using ¹⁸F-FDG-PET for translational research of AD, a suitable mouse model should show similar changes in glucose metabolism as seen in AD patients. In the present study, Tg4–42 mice displayed changes in ¹⁸F-FDG-PET according to its AD pathologies. Aged Tg4–42 mice showed distinctly impaired glucose metabolism in all cortical areas correlating with the observed memory and neuron loss. While young Tg4–42 mice did not show significant differences in whole brain uptake compared to WT, glucose metabolism was reduced in the hippocampus, forebrain, hypothalamus, amygdala and midbrain. These regional changes in ¹⁸F-FDG uptake occurred even before memory deficits or neuron loss. We could previously show in electrophysiology experiments that young Tg4–42 mice display neuronal dysfunction and altered synaptic short-term plasticity (Dietrich et al., 2013). Our current in vivo results in the Tg4–42 model are also well in line with FDG-PET studies performed in AD patients where synaptic activity and glucose metabolism can be disturbed long before neuron loss and atrophy occur (Mosconi et al., 2010). The Tg4–42 mouse model shows changes in brain metabolism and synaptic dysfunction in an early disease stage similar to early changes observed in patients with AD.

Interestingly, cerebral metabolism seems to highly vary between different mouse models for AD. Different animal models of amyloidosis have shown unaltered, decreased as well as increased glucose metabolism (Kuntner et al., 2009; Luo et al., 2012; Li et al., 2016; Takkinen et al., 2017; Waldron et al., 2017). Cerebral glucose metabolism in AD mice has been previously studied in different AD mouse models using ¹⁸F-FDG or [¹⁴C]-2-deoxyglucose autoradiography. Similar to our finding several previous studies showed cerebral hypometabolism ex-vivo using autoradiography in different transgenic APP mouse models including PDAPP and the triple-transgenetic mouse model 3xTg (Valla et al., 2008; Nicholson et al., 2010). However, autoradiography requires euthanasia of the animals and therefore is not suitable for longitudinal studies or drug evaluations. The clinical phenotype of decreased ¹⁸F-FDG has also been reported in vivo in several AD models (Waldron et al., 2015, 2017). Notably, other AD mouse models did show no differences in brain metabolism or even cerebral hypermetabolism (Kuntner et al., 2009; Luo et al., 2012; Li et al., 2016; Takkinen et al., 2017). For example, an increased regional glucose metabolism was detected in the commonly used APP/PS1 mice at 2- and 12-months of age (Poisnel et al., 2012; Li et al., 2016). Furthermore, Rojas et al. (2013) reported increased ¹⁸F-FDG binding ratios relative to the cerebellum in aged 5XFAD mice. They hypothesized that the lack of glucose hypometabolism in many AD mouse models may be due to increased glia uptake of ¹⁸F-FDG. Amyloid plaques are pro-inflammatory agents and increased activated astroglia and microglia around them could lead to increased cerebral ¹⁸F-FDG uptake (Rojas et al., 2013; Waldron et al., 2017). However, AD patients display a decreased brain glucose metabolism despite the presence of amyloid plaques and increased gliosis (Herholz et al., 2007; Mosconi et al., 2010; Marcus et al., 2014). We could reaffirm this in the Tg4–42 model as mice also show increased inflammation and hypometabolism.

The lack of cerebral hypometabolism in many AD mouse models may also be due to the lack of neuron loss in most AD models. While atrophy and neuron loss are one of the major neuropathological hallmarks of AD, many transgenic AD mouse models lack this key feature (Bayer and Wirths, 2010; Schaeffer et al., 2011). Some of the earliest damages in AD brains have been described in the hippocampus. Padurariu et al. (2012) reported a decrease in neuronal density especially in the CA1 and CA3 regions of the hippocampus, although the effects in the CA1 area were more severe. The hippocampus is a brain area that is known to be highly affected in AD and contributes to the learning and memory deficits observed in AD patients. Therefore, we examined exemplary the neuron numbers in the hippocampus and extended previous findings demonstrating that Tg4–42 display a severe neuron loss not only in the CA1 region of the hippocampus but also in CA2 and CA3 region and the DG. The impaired spatial reference and recognition memory in aged Tg4–42 mice can be attributed to the neuronal disfunction, glucose metabolic impairments and the observed neuron loss.

Inconsistencies in previous ¹⁸F-FDG-PET AD animal studies might also be explained by different PET scanners, image acquisition and analysis methods. There is no standard for small animal brain imaging which results in high inter-study variations. Variants include tracer application, image acquisition and quantification methods. The effect of blood glucose on ¹⁸F-FDG uptake in the brain has also to be taken into consideration. Normalizing the SUV for blood glucose appears a suitable method to compensate for inequalities in glucose levels and fasting durations as cerebral ¹⁸F-FDG uptake shows an inverse relationship with blood glucose levels (Wong et al., 2011). Earlier studies demonstrated in repeated scans that brain ¹⁸F-FDG uptake values are well stabilized by glucose normalization (Wong et al., 2011; Deleye et al., 2016; Coleman et al., 2017).

It has to be noted while a wide range of animal models of AD have been generated and contributed to a better understanding of the pathogenesis of AD none of the models mimics the full range of pathological and degenerative alternations observed in AD patients (Götz and Ittner, 2008; Elder et al., 2010; Schaeffer et al., 2011; Webster et al., 2014). Most of these models only show certain aspects of the disease including amyloid plaques, neurofibrillary tangles, altered glucose metabolism, neuronal loss, gliosis or cognitive deficits. For example, the majority of transgenic mouse models expressing human APP show a robust plaque pathology but most of them lack a significant neuron loss (Duyckaerts et al., 2008; Cavanaugh et al., 2014). The Tg4–42 model used in this study shows key pathological features of AD including soluble amyloid Aβ oligomers, intraneuronal Aβ, glucose hypometabolism, neuron loss and memory deficits. However, it lacks neurofibrillary tangles and a robust plaque load pathology while showing region-specific intraneuronal Aβ deposits. Therefore, in vivo imaging with amyloid (or tau) PET is not applicable to this model. On a side note, growing evidence suggests that intraneuronal Aβ accumulation contributes to the pathological events in AD and is one of the first neurodegenerative alterations in the AD brain (Fernández-Vizarra et al., 2004). Grundke-Iqbal et al. (1989) already reported nearly 30 years ago that intracellular Aβ precedes Aβ plaque formation in AD patients. In addition, it has been shown that intracellular Aβ also appears before neurofibrillary tangles (Gouras et al., 2000; Fernández-Vizarra et al., 2004). Furthermore, it could be demonstrated that mainly Aβ42 variants accumulate within neurons in AD vulnerable regions (Aoki et al., 2008).

The evaluation of new therapeutic strategies relies heavily on suitable animal models that mimic the pathological changes in AD. Here we could show that Tg4–42 mice display an age-dependent altered glucose metabolism accompanied with severe neuron loss and memory deficits. Similar to AD patients a significant regional decrease could be detected early in the disease progression in young Tg4–42 mice. The altered glucose metabolism coupled with age- and disease related cognitive decline of the Tg4–42 mouse model make it well-suited for utilization of preclinical testing of AD-relevant therapeutics. ¹⁸F-FDG-PET is a robust tool to display changes and monitor therapeutic effects in vivo. Repeated scans of the same animal allow a follow up during a long-term therapy improving accuracy by reducing the effect of inter-animal variances and reducing the total number of required animals. However, it is essential that animals are tested under standardized conditions with low variations between single animals and groups. High inter-study differences should be avoided in future studies in which a standard imaging protocol would be highly beneficial.

Overall, Tg4–42 could be a useful AD model to monitor the effects of various therapeutic strategies in vivo using ¹⁸F-FDG uptake as a therapeutic readout.

CB performed experiments, analyzed data, contributed to experimental design and wrote the manuscript. PH performed experiments and analyzed data. TF, CI, MS and NB performed experiments. CS and TB participated in the discussion of the results. YB conceived and designed the project, performed experiments, analyzed data and wrote the manuscript. All authors contributed to revising the manuscript and approved the final version.

University Medicine Göttingen has been granted a patent on the Tg4–42 mouse model, on which TB is listed as an inventor. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.