Assessment of brain beta-amyloid deposition in transgenic mouse models of Alzheimer’s disease with PET imaging agents ¹⁸F-flutemetamol and ¹⁸F-florbetaben

Recently, Aβ imaging with ¹⁸F labeled radiotracers has been widely used for patients with Alzheimer’s disease (AD). ¹⁸F-flutemetamol is ¹⁸F labeled analogue of ¹¹C-PiB produced by GE Healthcare (Buckinghamshire, UK) [2]. It has been useful in differentiating between patients with AD and healthy subjects with high specificity (96%) and sensitivity (93%) in the detection of AD, as well as high test–retest reliability [3, 4]. ¹⁸F-florbetaben is an ¹⁸F labeled polyethylene glycol stilbene derivative showing high in vitro affinity and specificity for β-amyloid plaques [3].

¹⁸F-florbetaben and ¹⁸F-flutemetamol are widely used for the diagnosis and monitoring of AD in a routine medical field. However, there are many unknown issues regarding the difference in tracer dynamics and biodistribution between ¹⁸F-flutemetamol and ¹⁸F-florbetaben. Because of the difficulty conducting the direct comparison between two tracers for humans due to the weighted exposure to radiation, a preclinical animal study is a good alternative option for a baseline study.

Recently, several imaging studies using newly developed ¹⁸F labeled Aβ PET tracers were reported in AD mouse models. In a previous study, the in vivo ¹⁸F-flutemetamol binding of Aβ deposits was tested in various AD mouse models [5]. In old APP23 mice, significant ¹⁸F-flutemetamol retention was observed in the brain. But, ¹⁸F-flutemetabmol did not show a outstanding advantage in APPswe-PS1dE9 and Tg2576 mice.

However, transgenic mice with various genetic backgrounds have been related with different pathologies, which make it difficult to interpret the overlapping study results [6]. Therefore, comparisons between β-amyloid imaging regarding AD mouse have to be accomplished with some caution as brain sizes and anatomic landmarks of target VOIs greatly affect accurate PET signal quantification. Until now, there has been no antecedent report comparing between ¹⁸F-flutemetamol and ¹⁸F-florbetaben images in an AD mouse model, so comparative conclusions draw special interest.

Herein, we tested a recently developed APPsw mouse model (C57BL/6-Tg(NSE-hAPPsw)Korl) enhancing expressing Swedish double mutation form of human APP (K670 N, M671L) under regulation of the neuron specific enolase (NSE) promoter. For this mouse model, there has been no attempt regarding its application for the evaluation of new Aβ imaging ligands. Hence, we performed a small animal study conducting direct comparisons between two ¹⁸F labeled Aβ PET tracers, ¹⁸F-flutemetamol and ¹⁸F-florbetaben in (C57BL/6-Tg(NSE-hAPPsw)Korl) mouse model in terms of following aspects: the ability to discriminate a transgenic from a control mouse, intensity of uptake and distribution pattern in visual images, difference of static ratio and kinetic parameters, bio-distribution and correlation with neuropathologic findings.

In this study, ¹⁸F-florbetaben and ¹⁸F-flutemetamol images could differentiate AD and control group on visual and SUVR analysis. The ¹⁸F-florbetaben group presented differences in K1 and k4 kinetic parameters between AD and control groups, although ¹⁸F-flutemetamol did not show difference. Several differences emerged between ¹⁸F-florbetaben and ¹⁸F-flutemetamol. ¹⁸F-florbetaben images showed more prominent visual uptake intensity and higher SUVR than the ¹⁸F-flutemetamol images did. Moreover, ¹⁸F-florbetaben PET images more correlated well with the thioflavin S staining. However, according to bio-distribution and kinetic results, ¹⁸F-flutemetamol is more actively metabolized than is ¹⁸F-florbetaben, suggesting that ¹⁸F-flutemetamol has faster transport from arterial plasma into the first tissue compartment and faster dissociation from the amyloid tracer complex.

In the static analysis data, the results were grossly consistent with a previous study [12]. In another ¹⁸F-florbetaben PET study, the SUVR in APPswe/PS2 at 5 months was 0.95 ± 0.04, and the SUVR in APPswe/PS1G384A mice at 5 months was 0.93 [12]. The traditional SUVR method measures the radioactivity ratio of brain target regions to reference tissue during a fixed time interval after injection of the tracer [11]. This relative quantitative approach for static PET data is practical for routine clinical setting. However, due to the kinetic compartment model for reversible binding radiotracers such as ¹⁸F-florbetaben or ¹⁸F-flutemetamol, the kinetic model reflects the available binding site density and also the perfusion signal and tracer clearance to and from brain tissue [11]. In this study, the 2 tissue compartment model with IDIF method was used, and the IDIF appears to be an attractive non-invasive alternative option obviating the need for arterial cannulation, blood handling and analysis [13–15]. Furthermore, to avoid the effects of non-specific binding, we prolonged the uptake time, resulting in a longer wash-out of non-specifically bound tracer. A clinical protocol for ¹⁸F-florbetaben involves a 90 min uptake periods [14]. A similar protocol was used in the previous APPPS1-21 mouse cohort study, allowing a 90-min uptake time [15].

The reasons for the disparity in imaging characteristics between ¹⁸F-florbetaben and ¹⁸F-flutemetamol are related to their chemico-physiological properties. ¹⁸F-florbetaben and ¹⁸F-flutemetamol belong to different families of imaging probes. ¹⁸F-flutemetamol is a member of the thioflavin derivatives imaging probe family [16], and ¹⁸F-florbetaben belongs to a different branch of imaging probe family, the trans-stilbene derivatives [16]. Because these two tracers belong to distinct chemical families, they showed differences in binding affinity. In the bio-distribution data, ¹⁸F-flutemetamol showed lower brain and higher peripheral organ uptake responsible for metabolite excretion compared with ¹⁸F-florbetaben. These findings and kinetic parameter results suggest that ¹⁸F-flutemetamol is more actively metabolized than is ¹⁸F-florbetaben. The tracer metabolites were more polar than were the parent molecules and therefore less able to enter the brain [17]. In a preclinical study comparing the pharmacokinetic characteristics of ¹⁸F-flutemetamol with that of ¹¹C-PiB, the metabolism of ¹⁸F-flutemetamol was faster than that of ¹¹C-PiB [18]. This finding can be explained by the higher lipophilicity of ¹⁸F-flutemetamol (logPC18 = 1.7) than that of ¹¹C-PiB (logPC18 = 1.2) [19]. In another study, the lipophilicity of ¹⁸F-florbetaben (Log Doct/PBS = 1.58) was higher than that of ¹¹C-PiB (Log Doct/PBS = 1.50) [20]. These results indirectly demonstrate that the rapid metabolism of ¹⁸F-flutemetamol could be explained by the higher lipophilicity of ¹⁸F-flutemetamol (logPC18 = 1.7) than of ¹⁸F-florbetaben (Log Doct/PBS = 1.58).

Previous studies reported that various transgenic animal models showed differences in binding affinity with imaging tracers and this phenomenon was thought to be related with variations in plaques configurations. In this study, 18-week-old AD transgenic mice carrying NSE-controlled APPswe, C57BL/6-Tg (NSE-hAPPsw) Korl were selected due to their rapid and robust amyloid plaque development at that early age [21]. In contrast, Tg2576 mice showed late onset and slower accumulation [5]. In another previous report using APPPS1 mice co-expressing L166P mutated Presenilin 1 under the control of a neuron-specific Thy1 promoter and KM670/671NL mutated amyloid precursor protein, cortical amyloidosis was reported at the age of 6–8 weeks [22]. In APPPS1-21 mice, amyloid was known to accumulate in a 4-week and cortical microglia increased threefold from 1 to 8 months of age [22]. Hence, APPPS1 mice are good for investigating the mechanism of amyloidosis and treatment strategies because of their early onset of amyloid deposition and convenient cross-breeding with other genetically engineered mouse models.

In humans, the APPswe gene caused early presentation of familial AD [23]. In the NSE-controlled APPswe mouse model, the Swedish double mutation at the 670/671 codon in the human APP gene under the control of the NSE promoter caused increased cleavage by the beta secretase and accelerated amyloid accumulation at young age [24]. Amyloid deposition in the NSE-controlled APPswe mouse induces subsequent neuronal apoptosis through the mechanisms of the mitogen-activated protein kinase (MAPK) and c-Jun N-terminal kinase (JNK) pathway or caspase-3 pathway [25, 26]. Although the mouse model in our study was relatively young, compelling evidence from previous studies regarding the dynamics of cerebral amyloidosis using a young APP mouse model indicates that NSE-augmented APPswe mice are suitable for the neuropathological phenotype of AD [27]. Moreover, we selected only male mice to control the effect of sex. Several studies reported the effect of sex on β-amyloid accumulation and AD phenotype. Latest studies have investigated the effects of sex on hippocampal atrophy in normal aging, MCI and AD [28]. Sex could regulate the relation of amyloid positivity and cognition [28]. Also, significant sex differences in pathology of 3xTg-AD mice suggested these differences may be due to organizational actions of sex hormones during development [29].

The results of this study are in contrary to those of antecedent ¹⁸F-FDDNP study that presented affinity for both amyloid and neurofibrillary tangles [30]. There was no increase in cortical uptake even in 13–15-month-old Tg2576 mice, even if technical issues, such as low spatial resolution, were regarded as the reasons for this negative PET finding [30]. However, in another ¹⁸F-FDDNP study in triple-transgenic rats, previous partial volume effects were overcome and contrasting results were observed; prominent uptake was presented in the frontal cortex and hippocampus [31]. In an ¹¹C-PiB study, old Tg2576 mice showed prominent cortical binding than did control mice [9]. Those paradoxical results were explained by the confounding effects of cortical perfusion and the low distribution of ¹¹C-PiB binding sites per plaque [32]. However, following the ¹¹C-PiB study with high specific activity overcame such confounding effects, so significant cortical uptake and excellent correlation between PET uptake and a pathologic amyloid burden were observed in APP23 mice compared with age-matched healthy controls [33]. Additionally, in a recent ¹¹C-PiB PET study of APP/PS1 mice, an outstanding correlation can be found between imaging results and the plaque burden measure obtained ex vivo and in vitro in the same animals [34]. In a previous preclinical imaging study comparing ¹⁸F-florbetaben and ¹¹C-PiB, which is of the same thioflavin T derivative family with ¹⁸F-flutemetamol, two aged AD mouse models with contrasting levels of amyloid deposition to high (APPPS 1-21) and low (BRI 1-42) target state were investigated [15]. Compared with control mice, APPPS 1-21 mice (high target state) presented prominent fibrillary amyloid accumulation in both ¹¹C-PiB and ¹⁸F-florbetaben, but the difference of uptake between AD and control mice was higher for ¹¹C-PiB than for ¹⁸F-florbetaben [15]. However, BRI1-42 mice (low target state) did not show enhanced tracer uptake [15]. Taking into consideration the difference in the mouse ages, our results broadly resemble their findings. Another ¹⁸F-florbetaben study using the same mouse cohort reported only a 14.5% difference between control and transgenic mice (5XFAD) found with ¹⁸F-florbetapir in comparison to a 21% difference found with ¹¹C-PiB [35].

Before the interpretation of ¹⁸F-florbetaben or ¹⁸F-flutemetamol images in clinical settings, preclinical approaches can provide baseline information regarding differences in the kinetic and metabolic properties of two tracers. The visual image and SUVR of ¹⁸F-florbetaben showed extensive cortical uptake in the same cohorts compared with ¹⁸F-flutemetamol images in both the AD and control groups. On the ¹⁸F-flutemetamol images, high lipophilicity and fast metabolism might complicate the analysis of PET data. In this study, the metabolism and kinetics of the tracer also have a great influence on the visual uptake of the amyloid tracers. Using these points, the human amyloid image should be read in consideration of the pharmacokinetic and metabolic properties of the tracer. Therefore, preclinical imaging might provide valuable information about the possibilities and limits of a given approach in humans by helping to better understand the in vivo binding characteristics of an imaging agent. The results of this study suggest that appropriate outcome measures are important in monitoring disease progression and response to therapeutic approaches in human settings. In this study, both tracers for VOI-based ratio analysis could discriminate the AD transgenic and control groups. However, on kinetic parameters from dynamic data, ¹⁸F-flutemetamol images could not be used as an indicator to distinguish between AD transgenic and control groups. Moreover, the detection of amyloid PET signal in this early aged mouse model used in this study suggests the sensitivity of the PET imaging bio-marker, suggesting the possibility of early detection of amyloid pathology before the manifestation of behavioral abnormalities.

There are several limitations that should be mentioned. First, the distribution patterns between ¹⁸F-florbetaben and ¹⁸F-flutemetamol were compared at a single time point. Therefore, the current data are insufficient to judge the superiority between the two tracers based. In a follow-up study, the scope of the analysis should be extended to cover the comparison of serial and chronological accumulation pattern between ¹⁸F-florbetaben and ¹⁸F-flutemetamol.

Second, there are some issues regarding methodological perspectives. Herein, for the shape of the merged atlas to match well with the skull CT, the thresholded CT image was manually fused with the magnetic resonance template. However, limitations could exist regarding the method of manual registration. More accurate, automatic algorithm is required in the further study. Additionally, partial volume correction was not conducted when defining the VOI in the blood input area, because the VOI size of the blood input area was larger than the volumetric PET spatial resolution (0.343 mm³). Therefore, the effect of the partial-volume correction should be investigated in a further study.

Moreover, in further studies, the different amyloid isoform structures and the range of fibrillarity influencing PET imaging results should be investigated. Amyloid plaques can be sub-classified according to the presence of dystrophic neuritis or reactive astrocytes and the morphological features as either diffuse, fibrillary or dense core types [36, 37]. In the thioflavin S image in our study, the thioflavin S positive plaque areas were predominantly diffuse rather than compact in terms of morphologic characteristic nature. Diffuse plaques are known to occur early in the disease course and to progress towards typical cored plaques [27, 38]. Dense-core plaques are often observed in AD mouse models, at an advanced age [39]. Morphological and biochemical compositional differences of plaques can influence the affinity binding sites for amyloid imaging tracers. Between ¹⁸F-florbetaben and ¹⁸F-flutemetamol, which tracer has higher binding affinity to diffuse type plaques? The answer to this question should be investigated in a further study including an in vitro binding assay. In addition, as plaques are amorphous three-dimensional configurations, further three-dimensional analysis of plaque structures with more precise detection stringency should be required.

Moreover, we have not performed Aβ 1-42 staining along with the Aβ 1-40 staining, in the follow-up of immunohistochemistry, because we simply wanted to demonstrate the establishment of AD mouse model and regarded Aβ 1-40 was better choice. Aβ 1-40 presents the most prominent Aβ isoform in the AD brain, while the Aβ 1-42 shows a substantial increase with specific forms of AD [40, 41]. Moreover, extraordinary expressions in AD mice carrying NSE-controlled APPsw presented that Aβ 1-40 was more prominent than Aβ 1-42 in the APPsw mice [42].

¹⁸F-florbetaben and ¹⁸F-flutemetamol images could distinguish between the AD and control group by both visual and SUVR-based analysis. The ¹⁸F-florbetaben and ¹⁸F-flutemetamol images showed disparate character in aspects of visual uptake intensity, quantitative parameters, bio-distribution and relations with neuropathological finding. ¹⁸F-flutemetamol was more actively metabolized than was ¹⁸F-florbetaben, although ¹⁸F-florbetaben presented higher visual uptake intensity, SUVR and close correlation with the pathology.