**3. Discussion**

ABCG2 was shown to be co-localized with ABCB1 at the BBB, where both transporters limit the brain distribution of many therapeutic drugs [19,37]. Studies in transgenic mice have provided evidence for functional redundancy between ABCB1 and ABCG2 at the BBB [19,37]. In the absence of ABCB1, the transport capacity of ABCG2 usually suffices to restrict the brain distribution of dual ABCB1/ABCG2 substrates and vice versa. Only when both transporters are genetically knocked out or pharmacologically inhibited, do dual ABCB1/ABCG2 substrates show unrestricted brain distribution. Both ABCB1 and ABCG2 have been identified as Aβ transporters [5–9,22–24] and the abundance and activity of cerebral ABCB1 were found to be decreased in AD [10–14], which may be caused by an Aβ-induced ubiquitination, internalization and proteasomal degradation of ABCB1 [38]. It is tempting to speculate that a similar functional redundancy between ABCB1 and ABCG2 exists with respect to Aβ export as described for restricting the brain distribution of small-molecule drugs [19,37]. Xiong et al. have shown an upregulation of ABCG2 in the brains of AD patients and AD mouse models (3XTg and Tg-SwDI) by means of immunohistochemistry and Western blot [22]. Moreover, these authors used optical imaging to show that the brain concentration of fluorescent-labeled Aβ after i.v. injection was

higher in *Abcg2(*−/−*)* mice as compared with wild-type mice, which confirmed Aβ transport by mouse ABCG2 [22]. Other studies, however, were not able to confirm an ABCG2 upregulation in AD brains by using either immunohistochemistry [10–12] or quantitative targeted proteomics [25,26].

To further address these questions, in the present study we examined both the abundance and activity of cerebral ABCG2 in a commonly employed β-amyloidosis mouse model (APP/PS1-21) [32]. This mouse model rapidly develops extensive cerebral Aβ deposits from an age of 2 months onwards [32,33] and has been used in a series of previous studies conducted in our laboratory to assess the activity of ABC transporters implicated in brain Aβ clearance with PET [15,33,39]. In a previous study we used immunohistochemical staining to show that the abundance of ABCB1 is decreased in the brains of APP/PS1-21 mice (i.e., in the hippocampus and the cortex) relative to wild-type mice of the same age range [33]. In the present study, using a comparable methodology we found a reduction in the number of ABCG2-stained microvessels in APP/PS1-21 mice (Figures 1 and 2). This reduction was not only found in the hippocampus, a brain region with high Aβ load, but also in the cerebellum, in which Aβ load is negligible [33]. Our analysis was not able to differentiate whether the reduced abundance of ABCG2 in the brains of APP/PS1-21 mice was caused by a decrease in vascular density [40] or by a decrease in the density of the transporter [11]. Therefore, additional experiments with a different methodology (e.g., Western blot analysis of isolated brain microvessels) will be needed to further examine the regional differences in ABCG2 observed in the present study.

To assess the consequences of these parallel reductions in the abundance of both ABCB1 and ABCG2, utilizing PET imaging, we studied the brain distribution of two radiotracers that are dual ABCB1/ABCG2 substrates ([11C]tariquidar and [11C]erlotinib). As a first approach, we used a previously developed PET protocol dedicated to measuring cerebral ABCG2 activity [28–30]. This protocol uses the dual ABCB1/ABCG2 substrate [11C]tariquidar [31] co-administered with a pharmacological dose of unlabeled tariquidar (12 mg/kg), which leads to complete saturation of ABCB1 activity while ABCG2 remains fully active, thereby ABCG2 selectivity is achieved. The attainment of ABCG2 selectivity is enabled by the great difference in half-maximum inhibitory concentrations (IC50) of tariquidar for in vitro inhibition of its own transport by ABCB1 (IC<sup>50</sup> = 17.1 nM) and ABCG2 (IC<sup>50</sup> = 310.4 nM) [30]. To reveal the activity of ABCG2, [11C]tariquidar PET scans were performed without and with pretreatment with the ABCG2 inhibitor Ko143 [34] at a dose that only partially inhibits ABCG2. The employed dose of Ko143 (5 mg/kg) was selected based on a previous dose-response curve generated in *Abcb1a*/*b (*−/−*)* mice, which provided a half-maximum effect dose of Ko143 of 4.98 mg to enhance brain uptake of [11C]tariquidar [28]. We determined *K*p,brain as the outcome parameter of the brain distribution of [11C]tariquidar, which was in a similar range in both APP/PS1-21 and wild-type mice after 5 mg/kg Ko143 (*K*p,brain range: 6–8, see Figure 5) as in *Abcb1a*/*b (*−/−*)* mice pretreated with 5 mg/kg Ko143 [28]. Maximum brain uptake of [11C]tariquidar amounted to a *K*p,brain of approximately 15 in *Abcb1a*/*b (*−/−*)* mice pretreated with 15 mg/kg Ko143, which was comparable to the brain uptake of [11C]tariquidar in *Abcb1a*/*b (*−/−*)Abcg2(*−/−*)* mice [28]. Taken together, the present data as well as the previous Ko143 dose-response data strongly suggest that the dose of Ko143 employed in our study (5 mg/kg) led to only partial ABCG2 inhibition at the mouse BBB. The experimental paradigm of studying transporter activity with a radiolabeled substrate by employing an inhibitor administered at a dose that only partially inhibits the transporter has been successfully employed by our group and by others to measure the activity of ABCB1 in the rodent and human brain with the ABCB1 substrate (*R*)-[11C]verapamil [33,41–44]. For instance, only PET scans after partial inhibition of ABCB1 revealed significant differences in the brain distribution of (*R*)-[11C]verapamil in APP/PS1-21 versus wild-type mice, while no differences were observed in baseline scans without ABCB1 inhibition [33]. In contrast to this previous study, we failed to detect differences in [11C]tariquidar brain distribution between APP/PS1-21 and wild-type mice, both under conditions of full ABCG2 activity and partial ABCG2 inhibition (Figure 5).

To confirm this apparent lack of difference in ABCG2 activity between APP/PS1-21 and wild-type mice, we also performed PET imaging in APP/PS1-21 and wild-type mice with a second dual ABCB1/ABCG2 substrate radiotracer. For this we used [11C]erlotinib, which is structurally identical to the epidermal growth factor receptor (EGFR)-targeting tyrosine kinase inhibitor erlotinib, a clinically used drug for the treatment of non-small cell lung cancer. Patients with this type of cancer often develop brain metastases, which are difficult to treat with erlotinib, most likely due to its low brain distribution caused by ABCB1/ABCG2-efflux transport at the BBB. We have previously shown that [ <sup>11</sup>C]erlotinib is transported by mouse ABCB1 and ABCG2 and can be used to assess the functional redundancy between ABCB1 and ABCG2 at the mouse BBB [35]. To mimic the clinical use of this drug we co-injected [11C]erlotinib with a pharmacological dose of unlabeled erlotinib (2 mg/kg) [36]. Similar to the results obtained with [11C]tariquidar, we found no differences in the brain distribution of [ <sup>11</sup>C]erlotinib between APP/PS1-21 and wild-type mice (Figure 6).

Our results are in line with and extend previous studies that failed to detect differences in the brain distribution of different drugs (including several ABCB1 substrates) between different β-amyloidosis mouse models and wild-type mice [15,45–47]. For instance, Gustafsson et al. found no differences in the unbound brain-plasma concentration ratios (*K*p,uu,brain) of the ABCB1 substrates paliperidone and digoxin between tg-APPArcSwe and age-matched control mice [46]. Similarly, Mehta et al. found no changes in the brain distribution of the ABCB1 substrates loperamide, verapamil and digoxin in triple transgenic AD mice harboring three mutant genes (*APPswe*, *PS-1M146V* and *tauP301L*), despite a reduction in the abundance of ABCB1 in this mouse model as shown by Western blot analysis of isolated cerebral microvessels [45]. This was explained by Mehta et al. by a thickening of the basement membrane of the BBB in AD mice, which may have impeded transcellular diffusion and thereby counteracted the reduction in the abundance of ABCB1. An alternative explanation for these previous findings may be that the changes in ABCB1 abundance in the AD mouse models were too low to cause significant changes in the brain distribution of ABCB1 substrates. ABCB1 is a high-capacity transporter, whose abundance needs to be reduced by >50% to see >2-fold changes in the brain distribution of its substrates [48]. In line with this hypothesis, we have shown that the brain distribution of the ABCB1 substrate radiotracers [11C]*N*-desmethyl-loperamide and (*R*)-[11C]verapamil was increased by only 1.1- and 1.5-fold, respectively, in heterozygous *Abcb1a*/*b* knockout mice (*Abcb1a*/*b (*+/*–)*), which have a 50% reduction in the abundance of ABCB1 at the BBB, as compared with wild-type mice [49]. In contrast, in homozygous *Abcb1a*/*b* knockout mice (*Abcb1a*/*b* (−/−) ), which completely lack ABCB1, the brain distribution of [11C]*N*-desmethyl-loperamide and (*R*)-[11C]verapamil was increased by 2.8 and 3.9- fold, respectively, relative to wild-type mice [49]. In our present study, we extended previous findings related to ABCB1 substrates [15,45–47] to dual ABCB1/ABCG2 substrates and showed that despite a concomitant reduction in the abundance of cerebral ABCB1 and ABCG2 in APP/PS1-21 mice as revealed by immunohistochemistry, the brain distribution of the dual ABCB1/ABCG2 substrates [ <sup>11</sup>C]tariquidar and [11C]erlotinib was unaltered. While caution is warranted in extrapolating these results to humans, our results suggest that the brain distribution of clinically used ABCB1/ABCG2 substrate drugs may be unaffected by possible disease-induced alterations in transporter abundances at the BBB of AD patients.

Limitations of our study include the low number of animals used for immunohistochemical analysis, our inability to differentiate between a reduction in vascular density and a reduction in the density of ABCG2, and the lack of further experimental data to confirm the reduction in cerebral ABCG2 (e.g., Western blot analysis of ABCG2 in isolated brain microvessels).
