Noninvasive optical activation of Flp recombinase for genetic manipulation in deep mouse brain regions

Spatiotemporal control of gene expression or labeling is a valuable strategy for identifying functions of genes within complex neural circuits. Here, we develop a highly light-sensitive and efficient photoactivatable Flp recombinase (PA-Flp) that is suitable for genetic manipulation in vivo. The highly light-sensitive property of PA-Flp is ideal for activation in deep mouse brain regions by illumination with a noninvasive light-emitting diode. In addition, PA-Flp can be extended to the Cre-lox system through a viral vector as Flp-dependent Cre expression platform, thereby activating both Flp and Cre. Finally, we demonstrate that PA-Flp–dependent, Cre-mediated Cav3.1 silencing in the medial septum increases object-exploration behavior in mice. Thus, PA-Flp is a noninvasive, highly efficient, and easy-to-use optogenetic module that offers a side-effect-free and expandable genetic manipulation tool for neuroscience research.

S tudies of complex brain functions require highly sophisticated and robust technologies that enable specific labeling and rapid genetic modification in live animals. A number of approaches for controlling the activity of genes or proteins in a spatiotemporal manner using light, small molecules, hormones and peptides have been developed for manipulating intact circuits or functions [1][2][3][4][5] . Cre/loxP recombination employing chemically inducible systems such as Cre-ER T2 , DD-Cre, and inducible promoter-directed Cre is the most commonly used in vivo genemodification system [6][7][8][9] . Other approaches include selective or conditional Cre-activation systems within subsets of green fluorescent protein (GFP)-expressing cells (Cre-DOG) 10 or dualpromoter-driven intersectional populations of cells 11 . However, these methods are limited by the considerable time and effort required to establish knock-in mouse lines and by constraints on spatiotemporal control, which relies on a limited set of available genetic promoters and transgenic mouse resources.
Beyond these constraints, optogenetic approaches enable to control the activity of genetically defined neurons in the mouse brain with high spatiotemporal resolution. Recently, two optogenetic modules of photoactivatable-Cre recombinase (PA-Cre) have been developed using split-Cre components, one in which each component is fused to CRY2 and CIB1 [12][13][14] , and the other in which they are fused to positive Magnet (pMag) and negative Magnet (nMag) 15 . In both cases, illumination with blue light induces heterodimerization. To date, however, an optogenetic module for gene-manipulation capable of revealing spatiotemporal functions of specific target genes in the mouse brain has remained out of reach.
Cre/loxP recombination has been established as an efficient genetic manipulation system in mammals, and constitutes the overwhelmingly prominent technology underlying currently existing conditional gene knock-out/in mouse resources 16 . Applications of the Flp/Frt system to sophisticated genetic manipulations in vivo, including in brain tissue, have developed at an accelerating pace 17,18 since most recent improvement in Flp recombinase to produce a thermostable, codon-optimized version (Flpo) that functions efficiently in mammals [19][20][21] . The combinatorial use of Cre and Flp in Cre/Flp lines, in which expression is driven by tissue-or cell-type-specific promoters, has enabled more detailed investigations of different cell populations through conditional or selective genetic mutagenesis in vivo 16,22 . To date, however, no such light-inducible Flp system has been developed. Accordingly, we sought to develop a photoactivatable Flp recombinase (PA-Flp) that takes full advantage of the high spatiotemporal control offered by light stimulation.
Here, we demonstrate that the utility of PA-Flp as a noninvasive in vivo optogenetic manipulation tool for use in the mouse brain, even applicable to deep brain structures reaching hippocampus or medial septum (MS) by external LED light illumination (1-2 mW mm −2 ). Furthermore, we engineer a Flpdependent Cre driver as a module without leaky Cre expression in viral vector system, finally showing noninvasive light-dependent, Cre-mediated Ca v 3.1 gene silencing in MS neurons, which results in increased objective exploration behavior.

Results
Development of PA-Flp. In designing PA-Flp, we employed a strategy in which Flpo was split into two pieces that reassemble in response to a light stimulus (Fig. 1a). Since a possible site at which Flpo could be split so as to restore activity upon reconstitution has not yet been reported, we first empirically addressed this question. To this end, we considered nine, aqueous-exposed loop regions of Flpo as possible sites, based on the structure of the full-length protein ( Supplementary Figure 1a-c). Next, we sought to introduce an efficient light-responsive module fused to each split-Flp site that satisfied the following four criteria: (1) lightinducible heterodimerization, (2) utilization of endogenous cofactors present at sufficient levels in mammals, (3) relatively small size to allow packaging in an adeno-associated virus (AAV), and (4) a high-affinity light-dependent interaction that minimally perturbs reassembly of split Flp. On the basis of these criteria, we selected the Magnet system (17.1 kDa) 23 as a suitable platform. This system offers a symmetry advantage compared with other light-responsive modules, such as CRY2 (65-70 kDa)-CIB1(21-22 kDa) pairs, which employ binding partners that are unbalanced in size.
In initial screens, we constructed fusions of each of the nine split-site variants of Flp fused at their C-terminus with nMagnetHigh1 (nMagH; nMagnet variants) or pMagnetHigh1 (pMagH; pMagnet variants), which are the strongest lightdependent dimerization pair among Magnet system 23 . To assess Flp activity in HEK293T cells, we used a Flp reporter plasmid in which FRT-stop-FRT-GFP is driven by a CMV promoter that produces a GFP signal upon catalytic activation of Flp. In these screens, we found that blue light-emitting diode (LED) illumination (470 ± 10 nm, 5-s pulse duration at 3-min intervals for 24 h) slightly increased Flp activity of split-site 1 (sp1:FlpN27/FlpC28) and split-site 7 (sp7:FlpN169/FlpC170) variants compared with the corresponding variants incubated in the dark (Supplementary Figure 1d). To further optimize the configuration, we next constructed sp1-and sp7-conjugated Magnet in both orientations, with or without a nuclear localization signal (NLS). Surprisingly, C-terminally fused split-FlpC showed dramatically increased (5-to 10-fold) recombination efficiency compared with N-terminally fused split-FlpC (see Supplementary Figure 1e compared with Supplementary Figure 1d). We termed sp1-and sp7-series with FlpNX-nMagH and pMagH-FlpCX orientation as PA-Flp1 or PA-Flp7, respectively. Additionally, we fused NLS Nterminally on PA-Flp1 or PA-Flp7 sites of FlpNX-nMagH, which showed that the potent NLS-tagging site for better efficacy is PA-Flp1 (Supplementary Figure 1d,e). Since, the efficiency of heterodimerization of between pMag and nMagH pair was comparable with pMagH and nMagH pair 23 , we also tested pMag among PA-Flp1 variants. Finally, we found that among (NLS)-Nflp1-nMagH with (NLS)-pMag(H)-(NLS)-Cflp1-(NLS) tested, NLS-Nflp1-nMagH with pMagH-Cflp1 showed the highest lightdependent Flp activity (Supplementary Figure 1e,f). Notably, sp1 of split-Flp could also be used with other heterodimerization modules, such as FKBP-FRB or PHR-CIBN, suggesting the universality of this split site (Supplementary Figure 1g). To further characterize the efficiency of PA-Flp1 or 7 variants, we tested them in stable cell lines bearing Frt-STOP-Frt to concurrently validate genome accessibility and light sensitivity. Ultimately, we chose one of PA-Flp1 variants, NLS-Nflp1-nMagH with pMagH-Cflp1 pair (hereafter simply PA-Flp), which showed high light sensitivity (EC 50 = 3.1 μW cm −2 ), fast recombination kinetics (t 1/2 = 1.1 h), and 2-to 8-fold induction compared with the other PA-Flp1 or PA-Flp7 variants tested (Supplementary Figure 2a, b).
Performance of PA-Flp in vivo. To verify PA-Flp in vivo, we electroporated PA-Flp with fDIO-YFP plasmids in embryonic day 15 (E15) embryonic mouse brains. Light illumination at P1 induced a substantial increase in Flp reporter (GFP) signal 2 days later compared with that observed in the dark (Fig. 1b) (Fig. 1c). We then evaluated whether PA-Flp is capable of manipulating endogenous target sites with high efficiency using a Flp reporter mice line (RCE:FRT, Gt(ROSA)26Sor tm1.2(CAG-EGFP) Fsh /Mmjax) 24 . To assess the effects of light-delivery conditions on the performance of PA-Flp in RCE:FRT mice, we delivered blue light (400μWmm −2 ) via an optic fiber for varying illumination times (5 s to 30 min). PA-Flp efficiency increased with increasing duration of light exposure up to 5 min, at which point efficiency saturated (Supplementary Figure 3b,c).
'Local genetic labeling' in the mouse brain by precise light targeting. Since the highly light-scattering nature of brain tissue requires more precise light-targeting conditions 25 , we sought to minimize light diffusion/scattering using thin optic fibers Through demonstrated that three-dimensionally restricted (~200 μm), anatomically defined areas could be genetically manipulated ( Fig. 1d-h).

PA-Flp activation by transcranial LED illumination.
To maximize the utility of the highly light-sensitive characteristics of PA-Flp, we tested its action by noninvasive light without prior implantation of an optic fiber. We found that light illumination using a light guide coupled LED (Ø 0.6 cm) at an intensity of 50 μW mm −2 for 30 min effectively delivered light to the cortex regions of mouse brain and produced highly efficient recombination performance (Supplementary Figure 7). Considering these results together with the captured images of noninvasive LED light penetrating through the mouse brain (Supplementary Figure 8), we anticipated that stronger light power (up to 5 mW mm −2 ) would provide sufficient light to activate PA-Flp at the depth of the hippocampus. Following PA-Flp expression in the hippocampus, we illuminated the mouse head with LED light at intensities of 0.1, 1, and 5 mW mm −2 for 30 s or 30 min with the fur removed, leaving the skull and skin intact. Surprisingly, we found that LED illumination with an intensity as low as 1 mW mm −2 for 30 s was sufficient for full activation of PA-Flp in the hippocampus (Fig. 2a-c). Stronger LED intensity (~5 mW mm −2 ) or increased illumination time (~30 min) did not lead to proportionally higher PA-Flp efficiency, indicating that LED illumination at 1 mW mm −2 for 30 s provided sufficient light to induce maximal PA-Flp activity (Supplementary Figure 9). To test PA-Flp activation in regions of the brain deeper than the hippocampus, we illuminated the medial septum (MS), a region 3.5-4.0 mm distal from the skull surface. These experiments showed that LED illumination at 2 mW mm −2 for 30 s was sufficient to activate PA-Flp in the MS ( Fig. 2d-f).
Extended utility of PA-Flp to Cre-loxP system. To extend the applicability of PA-Flp to the many existing loxP-flanked (floxed) lines, we investigated how to activate recombination at loxP sites using PA-Flp. To this end, we designed a Flp-dependent Cre driver (FdCd) by inserting Cre into an fDIO cassette (fDIO-Cre).
In initial experiments employing transient plasmid expression in vivo, we found that Flp activity was effectively transferred to activation of Cre (Supplementary Figure 10). However, packaging this fDIO-Cre construct in AAV or lentivirus resulted in unexpectedly high basal Cre activity in both cultured neurons and mouse brains (Supplementary Figure 11b-d). We hypothesized the viral nascent ITR/LTR promoter-driven transcripts with antisense direction generated a substantial number of Cre molecules to react with substrate DNA sequences under longexpression conditions, despite weakness of the ITR/LTR promoters [26][27][28] . To solve this problem, we designed a new FdCd candidate to block translation of ITR/LTR promoter-driven Cre (Supplementary Figure 11a).  (Fig. 4a). Our previous study revealed that silencing Ca v 3.1 T-type calcium channels in GABAergic neurons in the MS that project to the hippocampus increased object-exploration behavior in mice 31 . We thus validated this previous finding by silencing Ca v 3.1 specifically in the MS using light-triggered PA-FdCre. To evaluate objectexploration behavior, we first injected mice with PA-FdCre and floxed-stopped-shCa v 3.1, and then exposed them to LED illumination (shCa v 3.1 w/ LED mice) or room light conditions (shCa v 3.1 w/o LED mice); as a control, mice were injected with PA-FdCre and floxed-stopped-shControl (scrambled control shRNA construct) followed by LED illumination (shControl w/ LED mice) or room light conditions (shControl w/o LED mice). Thereafter, mice were exposed to novel objects in a familiar arena and monitored for 20 min (Fig. 4b). Tracking of mouse movements revealed enhanced object-exploration behavior by ShCa v 3.1 w/ LED mice compared with shCa v 3.1 w/o LED, shControl w/ LED, and shControl w/o LED mice (Fig. 4d), as confirmed by measurements of total exploration (Fig. 4e, f). Consistent with our previous observation 31 , the locomotor activity of shCa v 3.1 w/ LED mice, measured as total distance moved, was not significantly different from that of shCa v 3.1 w/o LED or shControl w/ LED, and shControl w/o LED mice during the habituation period of the object-exploration task (Fig. 4g). To assess the extent of gene silencing, we sacrificed mice after the completion of behavioral tests and measured Ca v 3.1 expression in MS neurons using immunohistochemistry. The proportion of Ca v 3.1-positive neurons in the MS decreased in shCa v 3.1 w/ LED mice compared with shCa v 3.1 w/o LED (Fig. 4c), indicating that PA-FdCre-mediated shCa v 3.1 silencing reduced endogenous Ca v 3.1 expression in the MS.

Discussion
In this study, we developed PA-Flp by searching out new split sites of Flp recombinase that was not previously identified, being capable of reconstitution to be active. We validated the highly light-sensitive, efficient performance of PA-Flp through precise light targeting by showing transgene expression within anatomically confined mouse brain regions. The concept of 'local genetic labeling' presented here suggests a new approach for genetically identifying subpopulations of cells defined by the spatial and temporal characteristics of light delivery. Importantly, genetic labeling is a permanent marking system that enables tracking of locally labeled subpopulations of cells, such as migrating, differentiating or proliferating cells, without loss of the labeled signal. PA-Flp activation through noninvasive light illumination deep inside the brain is advantageous in that it avoids chemical-or optic fiber implantation-mediated side effects, such as off-target cytotoxicity or physical lesions, that might influence animal physiology or behaviors 32,33 . Our optogenetic system is very easy to set up and simple to use, requiring only an LED (blue or coolwhite) source that satisfies light power and amount criteria, indicated above, for application to deep mouse brain regions, and it does not necessitate injury to the skull or skin for light delivery.
Finally, our technique provides expandable utilities for transgene expression system stringently upon Flp activity in vivo, by designing a viral vector for minimal 'leaky' expression influenced by viral nascent promoters, such as ITRs or LTRs, located at 5′ or 3′ end in AAV or Lentivirus. Accordingly, we showed that the PA-Flp-dependent Cre expression system called PA-FdCre is an expandable modality of PA-Flp, demonstrating the applicability of a light-inducible, Cre-dependent RNAi system to neurobehavioral research. The ability to support applications involving loxP-containing mouse lines or vector systems further highlights the versatility of PA-FdCre. Furthermore, PA-FdCre is a module of light-inducible Flp-and Cre-dual activation system, but Cre expression is activated by Flp activity along with the neuronspecific promoter-directed LF-FdCd as an intersectional manner. According to our observations, the probability of leaky Flp activity (attributed by spontaneous auto-assembly of split-Flp components) is far less likely to occur in driving two orthogonal reactions of Flp-FRT. Therefore, the design of gene expression cassette turned ON only under the sequential or intersectional Flp and Cre activation events 21 would offer greater reliability with robust amplitude of inducible gene regulation systems in vivo. Taken together, our findings indicate that PA-Flp is a highly
Stable cell line. A HEK293T cell line stably expressing Frt-stop-Frt-EGFP was produced by transduction of a lentiviral construct, prepared by PCR-amplifying a fragment from pCAFNF-GFP (Addgene ID:13772) and cloning it into a pLenti vector under the control of an EF1a promoter by Gibson assembly. Lentivirus was packaged by co-transfection of HEK293T cells on 15-cm plates with VSVG, pDelta, and payload constructs using polyethylenimine (#23966-2, Polyscience). After 4 h, the medium was changed and cells were cultured for an additional 72 h. Culture supernatants were collected at 48 and 72 h, pooled, and filtered through a 0.45-μm filter. HEK293T cells were transduced with lentivirus particles, and stable clones were selected by culturing with 300 µg/ml G418 (Geneticin) beginning 48 h after transduction. Single-cell clones of transduced HEK293T stably expressing Frt-stop-Frt-EGFP were collected by pipetting, after which recombination of Frt-stop-Frt-EGFP in each single cell clone lines was verified by transfection of Flp recombinase.
Cell culture, transfection, and live-cell imaging. HEK293T cells were acquired from the American Type Culture Collection (ATCC, Manassas, VA, USA) and were assessed for contamination using a PCR-based mycoplasma detection kit (Cell-Safe, BioMycoX). HEK293T cells were grown in high-glucose Dulbecco's Modified Eagle Medium (DMEM; PAA Laboratories GmbH) supplemented with 10% FBS (Invitrogen) and 1% penicillin-streptomycin (Invitrogen). Cells were incubated at 37°C in a humidified 10% CO 2 incubator. For experiments, dissociated HEK293T cells were plated at a density of 1 × 10 4 cells/well in 96-well plates and transfected in duplicate with 200 ng total DNA using a Fugen6 transfection kit (Promega) according to the manufacturer's protocol. Cells were stained with 500 ng/ml Hoechst 33342 (Invitrogen) for 30 min at 37°C before acquiring  images. Images were acquired at the indicated time points with an ImageXpress Micro XLS automated epifluorescence microscope (Molecular Devices) using a ×20 Plan Fluor objective and a 4.66 megapixel CMOS camera with a 16-bit readout. Image analysis was performed using MetaXpress software (Molecular Devices).
Histology. For fluorescence imaging, isolated mouse brains were fixed by incubating in 4% paraformaldehyde (PFA) in PBS for 1 d at 4°C. Brains in ice-cold PBS were sectioned into 50-60 μm coronal slices using a VT1200S vibratome (Leica). Slices were mounted in Fluoromount G (Southern Biotech), with or without DAPI (4′,6-diamidino-2-phenylindole). For immunohistochemistry, mouse brains were transcardially perfused with PBS followed by fixation in 4% PFA overnight at 4°C. Slices were incubated in blocking solution (5% normal goat serum, 0.3% triton-X in PBS) for 1.5 h and stained with primary antibody overnight at 4°C. Anti-HA-Tag rabbit mAb (1:1000, #3724, CST) and anti-Ca v 3.1 (1:200, #ACC-021, Alomone Lab) were used to detect HA-tagged PA-Flp and endogenous Ca v 3.1, respectively. After incubating with primary antibodies, slices were washed five times with 0.3% Triton-X in PBS, incubated with secondary antibody in blocking solution for 1.5 h at room temperature, and then washed five times in 0.3% Triton-X in PBS. Images were captured with a Nikon A1 confocal microscope.
Animals. C57BL/6J inbred mice were obtained from the Jackson Laboratory (JAX Mice and Services, Bar Harbor, ME, USA). The RCE:FRT Flp reporter line (Gt (ROSA)26Sor tm1.2(CAG-EGFP)Fsh/Mmjax ) 24 , was purchased from the Jackson Laboratory as cryopreserved sperm. Heterozygous mice were obtained by in vitro fertilization, performed by the KAIST animal facility. The Ai14 Cre reporter line (Gt(ROSA)26Sor tm14(CAG-tdTomato)Hze/J ) 37 was a gift from Dr. Shin (Institute of Basic Science, Korea). ROSA26 RCE:FRT/Ai14 mice were generated by crossing homozygous RCE:FRT and Ai14 mice; these Flp and Cre reporter lines have been previously characterized 16,[37][38][39] . Mice were maintained with free access to food and water under a 12/12-h light/dark cycle (light intensity in cages measured under 'room light' conditions, 2.5-5 μW mm −2 ). Animal care and experimental procedures followed the guidelines of the Institutional Animal Care and Use Committee of Institute for Basic Science.
Light source. A Ø 200 μm or 50-60 μm optic fiber (Doric) coupled to a blue diode 473-nm laser (MBL-III-473m; CNI) was used to deliver blue light to broad or local regions, respectively, within the hippocampal DG (see Supplementary Figure 4 for details). Immediately after AAV injection, optic fiber ferrules were implanted in approximately the upper 200-μm region of AAV-targeted sites. Custom-made (LCI, Korea) external LEDs (white or 470 nm blue), excited through a 0.28 cm 2 coupled fiber and controllable up to a power intensity of 10 mW mm −2 , were used for noninvasive light stimulation (See Supplementary Figure 9b for details of LED performance in the mouse brain). Light intensity was measured at the surface of the optic fiber or LED coupled-fiber end using the photodetector of a power meter (#PM120D; Thorlab). The light intensity per unit area (mW mm −2 ) from an optic fiber tip was defined as a measured value using the power meter.
Optical settings for local light delivery. Light stimulation of broad or local regions was performed using protocol settings described in Supplementary Figure 4a. Light-scattering effects of brain tissue were assessed by implanting optic fibers (Ø 200 μm) into the hippocampus of an anesthetized mouse and monitoring light-delivery patterns under various light-stimulation conditions using a dissecting microscope (Supplementary Figure 4b). The observation matched well with GFP (Flp reporter) signal patterns obtained from each light stimulation protocols. In addition to this, light diffusion emanating from the optic fiber tip was greatly reduced by introducing a tiny optic fiber (Ø 50-60 μm), combined with the use of optical settings with a low numerical aperture (NA; 0.1-0.22). Since total internal reflection occurs in conditions above critical angle within optic fibers, a lower NA results in the formation of more straightly oriented light path shapes (Supplementary Figure 4c), producing a light ray that exits the optic fiber tip with a lower diffusion angle (Supplementary Figure 4d). Tests of PA-Flp in the M1 region revealed a substantial amount of Flp leakage (~20%) under normal 12 h-light on/ off housing conditions, an effect that was likely attributable to the ultra lightsensitive characteristics of PA-Flp or possibly relatively high basal activity in the motor cortex cells where PA-Flp was expressed. Accordingly, we tested a PA-Flp version without the NLS tag (PA-Flp △NLS ), which is less light sensitive than PA-Flp ( Supplementary Figures 1e and 2a,b). Applied to the M1 region as a CAG promoter-driven form, PA-Flp △NLS exhibited significantly decreased basal Flp activity in the cortex (~5%) compared with that of PA-Flp (~20%). Local genetic labeling in the M1 region was validated using PA-Flp △NLS . In these experiments, an optic fiber (Ø 60 μm) was implanted by placing it in the upper part of layer V (Supplementary Figure 6a); PA-Flp △NLS expression was relatively broad across layers I-VI. Two light-intensity conditions (5 or 25 μW mm −2 ) and two illumination times (2 or 12 h) were used, and results were analyzed in layers I-III and V-VI regions. At an intensity of 5 μW mm −2 , PA-Flp △NLS efficiency was specifically increased in layer V−VI as the light illuminating duration was increased from 2 h (29.2% ± 1.5%) to 12 h (73.1% ± 3.0%). There was almost no impact of light diffusion or scattering patterns on layers I-III, whereas increasing the light intensity to 25 μW mm −2 resulted in considerable light diffusion or scattering patterns in all layers (Supplementary Figure 6b,c).
Data analysis. Images from ImageXpress were analyzed using CellProfiler 2.1.0. Cells in DAPI, mCherry, or GFP channels were detected using the 'Identify Primary, Secondary or Tertiary Objects' module. The recombination percentage was calculated by detecting GFP-positive (GFP+) cells among mCherry positive (mCh +) cells using intensity threshold settings. Images from a Nikon A1 confocal microscope were analyzed using Nikon imaging software (NIS-element AR 64-bit version 4.10; Laboratory Imaging) or MetaMorph software (version 7.8.1.0, MDS Analytical Technologies). In vivo-electroporated cells, each with fluorescence signals satisfying criteria of a mean intensity greater than 1000 arbitrary units (a.u.), diameter equivalent of 10-20 μm and circularity of 0.5-1.0, were detected automatically using the 'object count' tool in Nikon imaging software. Detected cells were assumed to be marker positive (+) cells for each fluorescence channel. The number of GFP+ (or YFP+) cells was then divided by the number of cells expressing both mCh and iRFP, and multiplied by 100 to obtain the percentage of GFP+/mCh+iRFP+ cells. The percentage of GFP+ cells among mCh+DAPI+ cells in AAV-infected tissues was calculated from automatic cell counts obtained using the 'Multi Wavelength Cell Scoring' tool in MetaMorph software. Measurements of the mean intensity (native GFP or TdTomato fluorescence) of individual cells in single confocal slices were taken from selected cell bodies or DAPI+ areas using 'Annotations and Measurements' or 'object count' tools in Nikon imaging software. A P-value < 0.05 was considered statistically significant.
Object-exploration task. The object-exploration task, used to detect differences in responses toward novel stimuli (novel objects) in a familiar arena, was performed according to previously described procedures 31,36 with minor modifications. Briefly, each mouse was habituated to a 20 × 32 × 14.5 cm experimental cage (identical to the home cage) containing bedding for 1 h. Locomotor activity during habituation was measured using Ethovision 3.1 software (Noldus Information Technology). After habituation, each mouse was removed from the experimental cage and briefly placed in a temporary cage. Three non-identical novel objects were introduced into the experimental cage, and the mouse was immediately reintroduced into this cage. Three different shapes of wood blocks-quarter circle (3.0 × 3.0 × 3.0 cm, 15 g), cylinder (3.0 cm diameter × 6.0 cm depth, 34 g), and cube (3.0 × 3.0 × 3.0 cm, 18 g)-were used as novel objects in this experiment. More than one object was used to maximize exploration. We confirmed that the mice showed no preference for any particular object. The object was placed at three different positions in the cage (Fig. 3h), and behavior was recorded for 20 min. Exploratory behavior was defined as satisfying any of the following criteria: (i) persistent projection of the nose toward the object; (ii) grabbing the object with fore limbs while keeping hind limbs fixed; (iii) touching the object using nose or whisker; or (iv) approaching such that the nose of the mouse and object were within 0.5 cm. Statistical significance was analyzed by one-way analysis of variance (ANOVA) or two-way repeated measures ANOVA followed by a Bonferroni post hoc test.

Data availability
The data supporting the findings of this study are available within the paper and its Supplementary Information files. Extra data are available from the corresponding author upon reasonable request.