The Evaluation of Rac1 Signaling as a Potential Therapeutic Target of Alzheimer’s Disease

The Small GTPase Rac1 is critical for various fundamental cellular processes, including cognitive functions. The cyclical activation and inactivation of Rac1, mediated by Rac guanine nucleotide exchange factors (RacGEFs) and Rac GTPase-activating proteins (RacGAPs), respectively, are essential for activating intracellular signaling pathways and controlling cellular processes. We have recently shown that the Alzheimer’s disease (AD) therapeutic drug donepezil activates the Rac1-PAK pathway in the nucleus accumbens (NAc) for enhanced aversive learning. Also, PAK activation itself in the NAc enhances aversive learning. As aversive learning allows short-term preliminary AD drug screening, here we tested whether sustained Rac1 activation by RacGAP inhibition can be used as an AD therapeutic strategy for improving AD-learning deficits based on aversive learning. We found that the RacGAP domain of breakpoint cluster region protein (Bcr) (Bcr-GAP) efficiently inhibited Rac1 activity in a membrane ruffling assay. We also found that, in striatal/accumbal primary neurons, Bcr knockdown by microRNA mimic-expressing adeno-associated virus (AAV-miRNA mimic) activated Rac1-PAK signaling, while Bcr-GAP-expressing AAV inactivated it. Furthermore, conditional knockdown of Bcr in the NAc of wild-type adult mice enhanced aversive learning, while Bcr-GAP expression in the NAc inhibited it. The findings indicate that Rac1 activation by RacGAP inhibition enhances aversive learning, implying the AD therapeutic potential of Rac1 signaling.

Recently, we have shown that the AD therapeutic drug donepezil, which increases brain acetylcholine levels, activates Rac1-PAK signaling in the striatum/nucleus accumbens (NAc), a mediator of emotional learning, for enhanced aversive learning in wild-type adult mice via the muscarinic acetylcholine receptor 1 (M1R)-PKC cascade [15]. Furthermore, the expression of the constitutively active PAK in NAc enhances aversive learning, indicating that Rac1-PAK signaling activation in NAc is sufficient for enhanced aversive learning. The above findings imply that Rac1 signaling is a potential AD therapeutic target for improving AD-associated learning deficits. In fact, aversive learning is used for preliminary AD drug screening because it allows short-term assessment of the effect of drugs against learning deficits caused by brain cholinergic dysfunction, which mimics AD-associated learning deficits [16]. To facilitate the development of AD therapeutic strategies, it is necessary to find an efficient way to activate Rac1 signaling. Considering that Rac1 activation by RacGEF is transient and that therapeutic drugs usually rely on inhibitors rather than activators, RacGAP inhibition can be a strategy for efficient Rac1 signaling activation. In this study, we tested whether Rac1 activation by RacGAP inhibition in the NAc enhances aversive learning to evaluate Rac1 signaling as a potential AD therapeutic target.
Here, we examined whether Rac1 activation by RacGAP inactivation affected aversive learning and found that conditional knockdown of breakpoint cluster region protein (Bcr), Rac1 GAP, in the NAc enhanced aversive learning. These findings indicate that Rac1-PAK signaling activation by RacGAP inhibition in the NAc enhances aversive learning.
To examine the protein expression of the five RacGAP candidates in the striatum/NAc of adult C57BL/6J mice (7-8 weeks old), ex vivo striatal/accumbal brain slices were prepared for subsequent immunoblotting analysis. Immunoblotting analysis of the brain slice lysates showed that all five RacGAP candidates were detected in adult mouse striatum/NAc at the protein level ( Figure 1A).
As not all RacGAPs show single specificity for Rac1 [17,24], we examined which RacGAPs efficiently inhibit Rac1 activity. Activated Rac1 is known to mediate membrane ruffling upon extracellular growth factor stimulation [25,26], and historically, growth factor-induced membrane ruffling has been used as an output to examine specificities for RhoGAPs in cells [27]. We therefore performed an HGF-induced membrane ruffling assay to screen for RacGAP candidates that efficiently inhibit Rac1 activity. To this end, MDCKII cells were transfected with plasmids expressing the RacGAP domain of each RacGAP candidate and were then stimulated with 100 pM of recombinant HGF (rHGF). Fifteen minutes after rHGF stimulation, the immunofluorescence analysis was performed, and the number of Rac GAP domain-expressing cells that showed membrane ruffling was counted. The expression of the ABR RacGAP domain (ABR-GAP) and the BCR RacGAP domain (BCR-GAP) significantly decreased the HGF-induced membrane ruffling to the same extent with non-stimulated MDCK II cells, whereas the expression of the SRGAP2 RacGAP domain (SRGAP2-GAP), SRGAP3 RacGAP domain (SRGAP3-GAP), or RICH2 RacGAP domain (RICH2-GAP) did not ( Figure 1B and Figure S1). These results indicate that BCR-GAP and ABR-GAP efficiently inhibit Rac1 activity in MDCKII cells. Therefore, we focused on Bcr and Abr for further analysis. One-way ANOVA followed by Dunnett's test, ns not significant, **** p < 0.0001. (C) Primary striatal/accumbal neurons were infected with EmGFP-control miRNA mimic-expressing AAV or EmGFP-Bcr miRNA mimic-expressing AAV at DIV 6. Bcr expression level was quantified by immunoblotting over three independent experiments. Upper panels show representative immunoblots. Quantification of the immunoblotting assay is shown in the bottom panels. The error bars represent the mean ± SEM of three independent experiments. One-way ANOVA followed by Dunnett's test, **** p < 0.0001. (D) Primary striatal/accumbal neurons were infected with EmGFP-control miRNA mimic-expressing AAV or EmGFP-Bcr miRNA mimic-expressing AAV at DIV 6. PAK autophosphorylation was quantified by immunoblotting over three independent experiments. Upper panels show representative immunoblots. Quantification of the immunoblotting assay is shown in the bottom panel. The error bars represent the mean ± SEM of three independent experiments. One-way ANOVA followed by Dunnett's test, * p < 0.05, ** p < 0.01. (E) Primary striatal/accumbal neurons were infected with EGFP-expressing AAV or EGFP-Bcr-GAP-expressing AAV. PAK autophosphorylation was quantified by immunoblotting over three independent experiments. Upper panels show representative immunoblots. Quantification of the immunoblotting assay is shown in the bottom panels. Student's t-test, **** p < 0.0001.
Bcr and Abr are large proteins that contain several different functional domains, including the RacGAP domain. Their RacGAP domains are well conserved, showing a high homology of 81% [28]. Bcr or Abr knockout has been shown to activate Rac1-PAK signaling in the whole brain in vivo, based on the increased PAK autophosphorylation [18]. They have also been shown to enhance spine formation in vivo, a critical factor for learning and memory [18]. Therefore, we investigated whether Bcr or Abr knockdown activates Rac1-PAK signaling in striatal/accumbal neurons. To this end, we utilized the engineered microRNA (miRNA) mimic system, in which a chemically synthesized double-stranded RNA molecule imitates the structure of endogenous miRNA (i.e., miR-155) and cleaves the open reading frame of complementary mRNA (for details, see Material and Methods). Two Bcr miRNA mimic sequences and two Abr miRNA mimic sequences were constructed into adenoassociated virus (AAV) plasmids. AAV-Bcr miRNA mimics or AAV-Abr miRNA mimics were transduced into primary striatal/accumbal neurons for immunoblotting analysis. Both Bcr miRNA mimics knocked down about 90% of the endogenous Bcr protein ( Figure 1C). Both Abr miRNA mimics also knocked down endogenous Abr protein ( Figure S2). However, their transduction caused cell death, despite the fact that neurons transduced with AAV-negative control or AAV-Bcr miRNA mimics did not when cultured under the same conditions. Thus, we hereafter focused on the effects of AAV-Bcr miRNA mimics.
To further confirm that Rac1-PAK activation by Bcr knockdown depends on its Rac-GAP activity, we transduced AAV-Flex-EGFP-Bcr-GAP and AAV-CaMKII-Cre into primary striatal/accumbal neurons. Immunoblotting analysis showed that Bcr-GAP domain expression significantly decreased PAK1 autophosphorylation at S144 ( Figure 1E). Collectively, the above results indicate that the Bcr RacGAP domain inhibits Rac1-PAK signaling in striatal/accumbal neurons and that Rac1-PAK activation by Bcr knockdown depends on its RacGAP activity.

Bcr Knockdown in the NAc Enhances Aversive Learning
To investigate the effect of Bcr knockdown on aversive learning, we used the Cre-Flex system to conditionally knockdown Bcr in accumbal neurons. In this system, AAV-CAGGS-Flex-EmGFP-Bcr miRNA mimic and AAV-CAMKII-Cre were co-injected into the NAc of adult C57BL/6J mice (7-8 weeks old), in which Cre recombinase inverts the Bcr miRNA mimic gene to the correct direction specifically in the NAc (Figure 2A). In this system, Bcr miRNA mimics significantly knocked down Bcr expression in the NAc in vivo ( Figure S3). Adult C57BL/6J mice (7-8 weeks old) were injected into the NAc with AAV-CAMKII-Cre, together with the indicated AAV vectors. Bottom panel shows a representative coronal brain slice showing the expression of AAV-CAGGS-Flex-EGFP-P2A 4 weeks after the injection. Scale bar, 1 mm. (B) Adult C57BL/6J mice (7-8 weeks old) were co-injected with AAV-CAMKII-Cre and AAV-CAGGS-Flex-EmGPF-Bcr miRNA mimics into the NAc and were subjected to passive avoidance test 4 weeks after the injection. AAV-mediated expression of Bcr miRNA mimics in the NAc increased the step-through latency 24 h after the mice received electric foot shock (0.47 mA, 4 s) during the passive avoidance test. Data are represented as the mean ± SEM (n = 8, where n is the number of mice for each group), One-way ANOVA followed by Dunnett's test, * p < 0.05. (C) Adult C57BL/6J mice (7-8 weeks old) were co-injected with AAV-CAMKII-Cre and AAV-CAGGS-Flex-EGPF-Bcr-GAP into the NAc and were subjected to passive avoidance test 4 weeks after the injection.AAV-mediated expression of Bcr-GAP in the NAc decreased the step-through latency 24 h after the mice received electric foot shock (0.8 mA, 5 s) during the passive avoidance test. Data are represented as the mean ± SEM (n = 8, where n is the number of mice for each group), Student's t test, ** p < 0.01.
Four weeks later, the injected mice were subjected to a passive avoidance test. In this test, mice were exposed to a mild electric foot shock (0.47 mA, 4 s) as an aversive stimulus in the dark component, where they prefer to stay in nature. Twenty-four hours later, step-through latency from the bright component to the dark component was measured.
As a result, Bcr knockdown in the NAc significantly extended the step through latency ( Figure 2B), indicating that Bcr knockdown enhances aversive learning.
To validate that the enhanced aversive learning by Bcr knockdown depends on its RacGAP activity, adult WT male mice were co-injected with AAV-CAGGS-Flex-EGFP-Bcr-GAP and AAV-CAMKII-Cre into the NAc (Figure 2A). Four weeks later, the AAVinjected mice were subjected to a passive avoidance test. In an experiment in which an electric foot shock (0.8 mA, 5 s) was used as an aversive stimulus 24 h after the shock, Bcr-GAP expression in the NAc significantly shortened the step through latency ( Figure 2C), indicating that Rac1-PAK inactivation by the Bcr RacGAP domain in the NAc inhibits aversive learning. Collectively, the above data indicates that the enhanced aversive learning by Bcr knockdown depends on its RacGAP activity.

Discussion
Dysfunctional Rac1 signaling has been implicated in AD pathogenesis and ADassociated learning deficits [32,33]. We have previously shown that the AD therapeutic drug donepezil activates Rac1-PAK signaling through the M1R-PKC cascade in striatal/accumbal neurons for enhanced aversive learning [15]. In this study, we found that Rac1-PAK signaling activation by RacGAP Bcr knockdown in the NAc during adulthood enhanced aversive learning. This finding implies that Rac1 signaling is an AD therapeutic potential target, as aversive learning is one of the outputs used for preliminary AD drug screening.
The role of RacGAPs in adult cognition has previously been assessed by global knockout during the embryonic stage [18,23]. However, it is unclear if RacGAPs during development or in adulthood are important in adult cognition because Rac1 is involved in both brain development and spine plasticity in adulthood. The present finding underlines the importance of Rac1 signaling activation during adulthood for adult cognition. It further implies the notion that Rac1 activation by RacGAP inhibition can be used as a therapeutic strategy for improving AD-associated learning deficits, considering that the onset of cognitive decline in AD occurs in adulthood.
As mentioned in the introduction, Rac1 signaling mediates various cellular processes through a wide range of downstream effectors. Historically, the characterization of actin reorganization by Rac1-PAK signaling has been well established [34], which has brought some mechanistic insights into AD pathogenesis, especially spine dysfunction [5,6]. Rac1 is also known to activate PI3K-AKT signaling downstream of GPCR [14], and recent findings show that its activation facilitates cell survival in various cell types [35][36][37]. Given that cell death is a major hallmark of AD, future investigations on whether Rac1 activation inhibits cell death in AD model mice would be of interest to further validate the AD therapeutic potential of Rac1 signaling. Recent findings that M1R-specific agonists inhibit cell death in prion-diseased mice and AD model mice [38,39] support this notion. In this context, we are currently investigating whether M1R-specific agonists activate Rac1 signaling.
The present study focused on Rac1 signaling in the NAc and aversive learning. However, AD patients also show declines in hippocampus-dependent learning and memory. We previously showed that donepezil activates PAK in the dorsal hippocampus and that PAK activation is involved in hippocampus-dependent recognition memory and associative learning [15]. Taken together with the finding that M1R-specific agonists facilitate hippocampus-dependent learning and memory [40], our previous findings imply that ACh-M1R-Rac1-PAK signaling is also involved in hippocampus-dependent learning and memory. For a full evaluation of the therapeutic potential of Rac1, further investigation on whether Rac1 signaling activation by RacGAP inhibition in the hippocampus enhances hippocampus-dependent learning and memory is currently underway.

Antibodies
The following antibodies were used for immunoblotting: mouse monoclonal anti-   Two Bcr engineered miRNA mimic oligos (NM001081412.2, miRNA-Bcr-895: 895 CTTC CTGAAGGACAACCTAAT 915 ; miRNA-Bcr-1582: 1582 CAAAGTGAGTGAGCTGGATCT 1602 ) and two Abr engineered miRNA mimic oligos (NM198895.2, miRNA-Abr-1111: 1111 TCTCC CAGAATTTCCTGTCTA 1131 ; miRNA-Abr-2320: 2320 TTGAGGAGTTGGGCATCTACA 2340 ) targeting the mouse BCR gene or mouse ABR gene were automatically designed by using BLOCK-iT RNAi Designer from Thermo Fisher SCIENTIFIC website (https://rnaidesigner. thermofisher.com/rnaiexpress/ (accessed on 4 December 2020)). The superscript numbers represent the first and last nucleotides of the target gene. As negative control, miRNA mimic (GTCTCCACGCGCAGTACATTT) that is predicted not to target any known vertebrate gene was used and purchased from Thermo Fisher SCIENTIFIC. The DNA oligos were annealed to generate double-stranded oligos in accordance with BLOCK-iT TM Pol II miR RNAi Expression Vector Kit user manual. The double-stranded oligos were cloned into pcDNA6.2-GW: EmGFP-miR vector, in which miRNA sequence derived from endogenous miR-155 sequence is inserted into the 3 -UTR of the EmGFP gene at BsaIrestriction enzyme site. The double-stranded oligos accompanied by the miRNA sequence were subcloned into the multiple cloning site (MCS) of the pAAV-CAGGS-Flex-EmGFP-miR-neg and pAAV-CAGGS-EmGFP-miR-neg at BamHIand XhoIrestriction enzyme site to allow AAV-mediated expression of Bcr miRNA mimics or Abr miRNA mimics.

Primary Neuron Culture
Primary striatal/accumbal neurons from E16 mouse embryos of ICR female mice were dissected on a 35 mm Petri dish on ice under a Leica stereo microscope as previously described with some modifications [41]. The dissected striatal/accumbal tissues were  (1 µM final dilution). AAV-miRNA mimics were transduced (MOI:10000) into striatal/accumbal neurons at DIV 6. Seventy-two hours after AAV transduction, half of the medium was aspirated and replaced with the same volume of fresh medium. The medium replacement was repeated every 2-3 days.

Cell Ruffling Assay
MDCK II cells were seeded at about 1.0 × 10 4 cells/well into 12 well plate containing PDL-coated cover glass and cultured in DMEM (Dulbecco's Modified Eagle's Medium-High Glucose, D5796-500ML, Sigma-Aldrich, St. Lois, MO, USA) containing 10% CS (BOVINE CALF SERUM, #SH30072.03, Thermo Scientific, Waltham, MA, USA) at 37 • C. Cells were transfected with the indicated vectors. Twenty-four hours after transfection, medium was changed to DMEM medium without serum, and cells were incubated at 37 • C for another 24 h. Forty-eight hours after transfection, cells were stimulated with 100 pM recombinant HGF (#082-09321, Fujifilm Wako, Osaka, Japan) for 15 min at 37 • C for subsequent immunohistochemical analysis. About 100 transfected cells at the outer edge of the colony were observed under confocal microscope (LSM780, Carl Zeiss, Jena, Germany). Among them, the number of transfected cells that show membrane ruffling was counted.

In Vivo Sample Preparation
In vivo sample preparation for immunoblot analysis was performed as previously described with some modifications [15]. After decapitation, mouse heads were immediately immersed in liquid nitrogen for~4 s, and the whole brains were removed. The brains were coronally sectioned (thickness: 2 mm) by using a mouse brain slicer matrix (#MBS-S1C, Brain Science Idea Co., Ltd., Osaka, Japan) chilled on ice. NAc was collected using 1.5 mm diameter biopsy punch (Kai corporation, Tokyo, Japan) on ice-cold plate. The tissue was snap-frozen in liquid nitrogen and stored at −80 • C until assayed. The sonication procedure was carried out as described in Section 4.7.

Passive Avoidance Test
Step-through passive avoidance test was performed as previously described with some modifications [15]. Shuttle box for passive avoidance testing (Model ENV-018MD, Med Associates Inc., St. Albans, VT, USA) was used in this study. The apparatus consists of two compartments: a bright compartment and a dark compartment (dimensions of each compartment: W 21.5 cm × D 17.0 cm × H 21.0 cm). The test consists of three phases: habituation phase, training phase, and testing phase. Each phase was carried out 24 h apart. The step-through latency was measured in each phase using MED-PC V software (Version 5.1, Brainscience Idea, Osaka, Japan). On day 1 (habituation phase), the mouse was put in the bright compartment. One minute later, the door opened, and the mouse was given 2 min to enter the dark compartment. On day 2 (training phase), the mouse was put in the bright compartment. One minute later, the mouse was allowed to enter the dark compartment. Once the mouse entered, the door closed automatically. Three seconds later, an electric foot shock (0.47 mA, 4 s or 0.80 mA, 5 s) was applied. The mouse was returned to the home cage 20 s later. On day 3 (testing phase), the mouse was put in the bright compartment, and the door opened 1 min later. The step-through latency was recorded up to 300 s.

Immunohistochemical Analysis
For membrane ruffling assay, MDCKII cells were fixed with 3.7% Formaldehyde/PBS for 10 min after HGF stimulation. Cells were rinsed with 1xPBS three times (10 min per wash) and were then permeabilized with 0.2% Triton X-100/PBS for 10 min at RT. After cells were washed with 1XPBS twice (10 min per wash), they were blocked with 1% BSA/PBS (filtered) for 30 min at RT. Cells were then stained for 1 h at RT or overnight at 4 • C with the indicated primary antibodies diluted with 1%BSA/PBS (filtered). Cells were washed three times with 1xPBS (10 min per wash) and incubated with Alexa Fluor 555-conjugated donkey anti-rabbit IgG secondary antibody for 1 h at RT. F-actin was visualized by staining with Alexa fluor 488-Phalloidin. After three washes with 1xPBS (10 min per wash), coverslips with cells were mounted onto the 76 mm × 26 mm slide glass (#S0318, MATSUNAMI, Osaka, Japan) using Fluoromount (#DBS-K024-25ML, DIAGNOS-TIC BIOSYSTEMS INC., Hague, The Netherlands). Image acquisition was performed using a confocal microscope (LSM780, Carl Zeiss, Jena, Germany) with a 63× objective lens (for Figure S1) or 20× objective lens (for Figure 1B) with identical gain and laser settings.
A total of 100 mm thick coronal brain slices were prepared using a vibratome (LECIA VT 2000S, Leica Microsystems, Wetzlar, Germany). The entire coronal brain section image was acquired by tile scanning using confocal microscope (LSM780, Carl Zeiss, Jena, Germany) with a 10 × objective lens.

Quantification and Statistical Analysis
Sample size was chosen according to previous studies [42,43]. No animals were excluded from the analyses. Animals were allocated randomly into experimental groups. The investigator was not blinded to group allocation during data collection, as only the investigator, who gave handling habituation, can access mice during behavior tests to avoid any types of stress.
All data are expressed as the mean ± standard error of the mean of at least three independent experiments. The statistical analysis was performed using Prism version 6 software (GraphPad, La Jolla, CA, USA). In all experiments, one-way ANOVA followed by Dunnett's multiple comparison test or Student's t test (two-tailed) was used. p values of <0.05 were considered statistically significant.