Assessment of neuroinflammation in a mouse model of obesity and β-amyloidosis using PET

Obesity is a global health issue owing to high intake of refined carbohydrates and saturated fat and increasingly sedentary lifestyles [1]. Growing evidence suggests that obesity at midlife increases the risk of cognitive decline and Alzheimer’s disease (AD) in later life [2‐4]. AD pathogenesis, which begins decades prior to the onset of symptoms [5, 6], is characterized by the accumulation of the toxic beta amyloid (Aβ) peptides, formation of neurofibrillary tangles, chronic inflammation, and neuronal loss [5, 7]. Numerous imaging modalities including structural and functional magnetic resonance imaging (MRI) and positron emission tomography (PET) assessment of cerebral glucose metabolism and amyloid burden have proven diagnostically useful in mild cognitive impairment (MCI) and AD [8]. However, the relationship between imaging biomarkers and early midlife pathogenic changes leading to AD, such as obesity, has remained elusive. Since obesity-driven inflammation and metabolic impairment have been identified as important mechanisms accelerating AD pathogenesis and functional impairment [9‐13], they may represent potential early-stage imaging markers of cognitive decline and AD risk.

Neuroinflammation has been identified as a contributor to both the initiation and progression of AD. Accumulation of soluble Aβ is suggested to induce inflammatory processes in AD, activating glial cells and potentiating the production of proinflammatory cytokines [14]. In obesity, chronic systemic inflammation triggers neuroinflammation, thereby exacerbating Aβ-induced degenerative cascades and AD progression [15]. The mitochondrial translocator protein (TSPO) is markedly upregulated in activated glia in inflammatory disease including AD, believed to play a critical role in mediating glial responses to AD pathogenesis. In vitro studies using transgenic mouse models of AD indicated that TSPO expression is linked with neuronal toxicity and death [16, 17]. ¹¹C-PBR-28 is a second-generation TSPO ligand with high affinity and penetration of the blood-brain barrier compared to the prototypic first-generation TSPO ligand, ¹¹C-PK-11195 [18]. Previous studies have demonstrated the usefulness of ¹¹C-PBR-28 in subjects with AD [19] and multiple sclerosis [20] and in animal models of aging [21] and cerebral ischemia [22]. However, the detection of early-stage inflammatory changes associated with obesity-related cognitive decline and AD risk has not been investigated using TSPO ligands.

Cerebral glucose metabolism measured by ¹⁸F-fluoro-2-deoxy-d-glucose-PET (FDG PET) is widely used in AD as an index of neuronal loss [23]. Although cerebral metabolic changes prior to the development of neuronal loss in AD remains poorly understood, emerging evidence suggests that hypometabolism may be preceded by a hypermetabolic phase [24‐27]. It is unclear if this early hypermetabolic stage reflects neuronal hyperactivity or inflammatory changes or if the hyperactivity is a cause (e.g., activity-driven Aβ deposition) or result of AD pathogenesis (e.g., Aβ-induced seizure activity or inflammation) [26, 27]. Transgenic mouse models of amyloidosis which do not develop neuronal loss may be useful tools for investigating metabolic changes associated early stages of Aβ pathogenesis [28‐31]. However, use of ¹⁸F-FDG PET in transgenic animal models may be hampered by the overexpression of amyloid precursor protein (APP) and non-Aβ fragments of APP, which are known to modify synaptic and metabolic activity [28‐31]. To circumvent this, here, we have used Aβ-infusion to investigate interactions between Aβ- and obesity-driven inflammatory and metabolic abnormalities in vivo using ¹¹C-PBR-28 and ¹⁸F-FDG PET, respectively, and determine the relationship between these imaging biomarkers, underlying pathogenic changes and functional outcomes.

In this study, we investigated obesity-driven inflammatory and metabolic abnormalities in an Aβ-infusion model of β-amyloidosis using PET. Obesity and hyperglycemia were evident in the high-fat fed mice after 1 month of feeding, accompanied by significantly increased abdominal and subcutaneous fat deposition and impaired glucose tolerance at sacrifice (Fig. 2b–d). Mice were infused with either vehicle or oligomeric human Aβ₄₂ 1 month after feeding initiation, once obesity was already evident in high-fat fed groups (Fig. 2a). In vivo TSPO signals were assessed at baseline, 2, and 3 months after feeding initiation via ¹¹C-PBR-28 PET, while cerebral metabolism was assessed 2.5 months after feeding initiation via ¹⁸F-FDG PET (Fig. 2a). Learning and memory and anxiety-related behaviors were assessed 3 months after feeding initiation, prior to the final ¹¹C-PBR-28 PET scan (Fig. 2a).

Using ¹¹C-PBR-28-PET, we detected elevated TSPO associated with activated astrocytes and microglia resulting from Aβ-induced inflammation. The Aβ-induced inflammatory response was transient, with TSPO signals no longer significantly elevated at the completion of the experiment. However, the magnitude and duration of the Aβ-induced inflammatory response was exacerbated by obesity, even though obesity alone did not increase TSPO inflammatory signals. Previous studies using TSPO as a marker of inflammation in transgenic mouse models of β-amyloidosis have also reported elevated TSPO signals detected by PET, although only in aged mice at late stages of neuropathology [17, 34, 35]. What remains unclear is to what extent TSPO expression represents deleterious versus trophic glial responses. While many studies have focused on TSPO as a marker of deleterious gliosis, reporting good correlations between microglial markers and TSPO PET signals [36], some evidence suggests that TSPO-positive glial activation may be associated with trophic glial functions including Aβ clearance [37]. To address this, in the current study, the systemic and hippocampal cytokine profile was examined to determine which specific inflammatory factors were associated with in vivo TSPO signals. Serum levels of numerous obesity-linked chemokines promoting adipocyte proliferation (TIMP-1 [38]) and adipose macrophage infiltration (CXCL1 [39, 40], CXCL2 [41], CXCL12 [42], CCL3 [43], and CCL5 [44]) were found to be significant predictors of in vivo hippocampal TSPO signals. Interestingly, in the hippocampus, levels of G-CSF [45, 46], ICAM-1 [47], and CXCL12 [48] and IL-1ra [49], which have been identified as inflammatory factors important in the regulation of amyloid clearance, significantly predicted in vivo hippocampal TSPO signals. Identification of the TPSO-associated inflammatory profile is necessary in order to understand the specific glial functions associated with TSPO signals, which is essential for the validation of TSPO as an effective biomarker of β-amyloidosis in AD.

In contrast to clinical observations in advanced AD, in the current study, β-amyloidosis and cognitive impairment were associated with increased cerebral metabolism as assessed by ¹⁸F-FDG PET. Numerous other studies have also associated hypermetabolism with early stages of amyloid pathology in transgenic mouse models of β-amyloidosis [28‐31]. Luo et al. reported hypermetabolism in young Tg2576 mice, prior to the development of plaques [29]. Previous studies have suggested hypermetabolism may result from increased gliosis in the absence of neuronal loss; however, we observed poor overlap between regions of hyperactivity and TSPO signals and no association between cerebral metabolism and markers of gliosis. Our findings suggest that hypermetabolic changes cannot be explained by gliosis alone. Multiple metabolic disturbances like altered glucose uptake and glycolytic metabolism has been mechanistically implicated in β-amyloidosis [50] and Aβ-induced astrocytic and neuronal hyperactivity and seizure activity [51‐54]. In very young AD mutation carriers, approximately 25 years before the expected onset of symptoms, region-specific hypermetabolism is observed prior to the development of hypometabolism [24]. Likewise, amyloid deposition has been found to be associated with hypermetabolism in normal aging and MCI [25‐27]. Functional MRI (fMRI) studies have also demonstrated increased brain activation in normal controls with amyloid deposition [55], in MCI [56‐60] and in subjects at genetic risk for AD [61‐63]. Cerebral hypermetabolism may be induced in early stages of β-amyloidosis, reflecting Aβ-induced pathogenic mechanisms including glial and neuronal hyperactivity, preceding neuronal loss.

In contrast to our previous findings [13], in the current study, obesity alone did not result in a significant deficit in learning and memory performance in the object recognition task. The lack of any deficit in obese mice in the current study is likely a consequence of poor performance of vehicle-treated control mice, which exhibited around 60 % recognition. From our previous studies, we expect the performance of control mice to be around 75 % in this task. We hypothesize that the reduced performance observed in control mice in the current study may be the consequence of repeated interventions including the pump and cannula implantation surgery and imaging experiments, with the stress on the procedures potentially impairing cognitive performance. This reduced the overall sensitivity of the object recognition test, making it difficult to detect impairment resulting from the experimental manipulations.

Interestingly, we did not observe any significant effect of obesity alone on markers of neuroinflammation, cerebral metabolism, or cognitive function. Findings from previous studies are mixed, with some studies reporting increased cortical or hippocampal markers of inflammation in high-fat fed, obese mice [64‐66]; others report modest region-specific effects or no detectable effect on the hippocampus [67, 68]. These differences may reflect differences in the severity of metabolic symptoms between mouse strains, diet composition (e.g., cholesterol and caloric and macronutrient content), and duration of treatment exposure. However, this may not be detectable by PET due to methodological issues.

In the current study, obesity induced some limitations including not only logistical and image quality issues due to scatter and attenuation but also in validation of analysis models where large differences in body weight affect radioligand distribution throughout the body. SUV has been widely validated as an outcome measure for ¹⁸F-FDG uptake adjusting for differences in body weight. However, hyperglycemia is associated with reduced cerebral ¹⁸F-FDG uptake due to increased competition to cross the blood-brain barrier, likely resulting in an underestimation of cerebral metabolism in obese groups [69‐71]. In the case of ¹¹C-PBR-28, systemic inflammation and increased TSPO binding in peripheral tissues of obese animals [72] may also have significant impacts on brain uptake. These factors are likely contributors to the reduction in in vivo brain ¹¹C-PBR-28 signals observed in obese mice; however, good agreement was observed between in vivo TSPO signals and immunoreactivity, validating the %ID/g for quantifying TSPO in living brains.

Obesity combined with Aβ infusion was found to increase both TSPO-PET signals and cerebral glucose metabolism. These signals can predict learning and memory performance. In vivo TSPO signals were confirmed by ex vivo analyses and correlated with inflammatory markers of hippocampal gliosis and serum levels of cytokine and chemokines. In contrast, since hypermetabolic regions did not directly correspond to regions of increased TSPO signals, it may reflect early pathogenic effects of Aβ toxicity on neurons rather than inflammation. These findings demonstrate the potential usefulness of TSPO and cerebral glucose metabolism for the investigation of interactions between midlife lifestyle factors and Aβ amyloidosis with the aid of neuroimaging.