Targeting Neurons with Functional Oxytocin Receptors: A Novel Set of Simple Knock-In Mouse Lines for Oxytocin Receptor Visualization and Manipulation

Abstract The neuropeptide oxytocin (Oxt) plays important roles in modulating social behaviors. Oxt receptor (Oxtr) is abundantly expressed in the brain and its relationship to socio-behavioral controls has been extensively studied using mouse brains. Several genetic tools to visualize and/or manipulate Oxtr-expressing cells, such as fluorescent reporters and Cre recombinase drivers, have been generated by ES-cell based gene targeting or bacterial artificial chromosome (BAC) transgenesis. However, these mouse lines displayed some differences in their Oxtr expression profiles probably because of the complex context and integrity of their genomic configurations in each line. Here, we apply our sophisticated genome-editing techniques to the Oxtr locus, systematically generating a series of knock-in mouse lines, in which its endogenous transcriptional regulations are intactly preserved and evaluate their expression profiles to ensure the reliability of our new tools. We employ the epitope tagging strategy, with which C-terminally fused tags can be detected by highly specific antibodies, to successfully visualize the Oxtr protein distribution on the neural membrane with super-resolution imaging for the first time. By using T2A self-cleaving peptide sequences, we also induce proper expressions of tdTomato reporter, codon-improved Cre recombinase (iCre), and spatiotemporally inducible Cre-ERT2 in Oxtr-expressing neurons. Electrophysiological recordings from tdTomato-positive cells in the reporter mice support the validity of our tool design. Retro-orbital injections of AAV-PHP.eB vector into the Cre line further enabled visualization of recombinase activities in the appropriate brain regions. Moreover, the first-time Cre-ERT2 line drives Cre-mediated recombination in a spatiotemporally controlled manner on tamoxifen (TMX) administration. These tools thus provide an excellent resource for future functional studies in Oxt-responsive neurons and should prove of broad interest in the field.

Some studies have used an antibody (called OXTR-2) generated in a research laboratory to detect Oxtr proteins (Marlin et al., 2015;Mitre et al., 2016). Unfortunately, this antibody is not available for the public to replicate the experimental findings. In addition, because of the general difficulty in generating antibodies to lipophilic GPCRs with low epitope exposures, no anti-Oxtr antibody suitable for immunohistochemistry is commercially available. From these reasons, the Oxtr-Venus knock-in mouse line was generated (Yoshida et al., 2009) and used for receptor visualization (Dölen et al., 2013;Hung et al., 2017;Newmaster et al., 2020). However, contrary to our expectations, expressions of the Venus reporter in the brain was reported to largely decrease during development (Newmaster et al., 2020). In this reporter line, the first coding exon had been replaced with Venus sequences (Extended Data Fig. 1-1b). The resulting knock-in allele lacks its endogenous configurations of intron and 39 untranslated region (UTR) sequences, both of which likely contain transcriptional regulatory elements essential for the spatiotemporally coordinated Oxtr expressions. This might considerably affect the Venus reporter expression profiles.
Cre recombinase driver mice also play pivotal roles in both conditional knock-out experiments and viral vectordependent functional analyses. Although two Oxtr-Cre knock-in lines using internal ribosome entry site (IRES) sequences for their bicistronic expressions have been generated (Hidema et al., 2016;Ryan et al., 2017;Extended Data Fig. 1-1c,d), those expression profiles reported were relatively sparse and did not well overlap with each other, probably because of the different context and integrity of their genomic configurations. In addition, the Oxtr-T2A-Cre-D line recently became available from The Jackson Laboratory (B6.Cg-Oxtr tm1.1(Cre)Hze /J, stock #031303), generated as a part of project in Allen Institute for Brain Science (Daigle et al., 2018;Extended Data Fig. 1-1e), yet its detailed expression profile has not been described.
Endogenously, the Oxtr gene locus consists of two coding exons divided by a long intron (Extended Data Fig.  1-1a). In prairie vole and human, some single-nucleotide polymorphisms (SNPs) in this long intron have been reported to be associated with Oxtr expression densities or social behavioral phenotypes (King et al., 2016;Jurek and Neumann, 2018). The authors suggested that these SNPs might overlap with cis-regulatory elements in the intron (King et al., 2016). The 39 UTR of the Oxtr gene has also been reported to contain nucleotide variations that are related to social behavioral phenotypes in humans (Jurek and Neumann, 2018). Considering these reports, it must be critical to design knock-in lines by maximally preserving the original transcriptional regulations, so as to precisely recapitulate endogenous Oxtr-expression profiles (Extended Data Fig.1-1 and Gerfen et al., 2013).
For proteins with no antibodies available for immunohistological detection, the epitope tagging strategy could be a promising breakthrough (Brizzard, 2008). Although the previously reported Oxtr-IRES-Cre line was designed to harbor an HA-tag at the C terminus of the receptor protein (Extended Data Fig. 1-1c), the epitope tag could not be detected in the brain tissues probably because of its relatively lower expression levels, caused by the non-native exon/intron configuration (Hidema et al., 2016). In this study, we first employed the novel PAtag system in which human podoplanin-derived epitope tag is recognized by its highly specific monoclonal antibody (Fujii et al., 2014). Since this system has been successfully applied to many in vitro procedures, such as protein purification and visualization, we anticipated that it would also work in vivo to determine the subcellular distributions of endogenous Oxtr protein in neurons. We next used the HA-tag system, because its commercially available antibody (Cell Signaling Technology, C29F4, #3724) has been successfully adopted by in vivo genome editing technique, SLENDR, to clearly map the subcellular localization of endogenous proteins in brain tissues with low background staining (Mikuni et al., 2016).
To achieve these concepts, here we apply our refined genome-editing skills to the Oxtr locus to generate four novel knock-in mouse lines, and evaluate their expression profiles to ensure the reliability in vivo.

Animals
All animal procedures were performed in accordance with the Animal Ethics Committee's regulations at the National Institute of Neuroscience. B6C3F1 female mice for fertilized eggs collection and ICR female mice as surrogate mothers were purchased from SLC Japan. Onecell stage zygotes were obtained by mating B6C3F1 stud males with super-ovulated females. Pronuclear injections and mouse transgenesis experiments were performed by standard protocols (Nagy et al., 2003). C57BL/6J mice for back-crossing were purchased from Charles River Japan. B6.Cg-Gt(ROSA)26Sor tm9(CAG-tdTomato)Hze /J (also referred to as Ai9, stock #007909) were introduced from The Jackson Laboratory.

Preparation of CRISPR components
To insert epitope-tag, reporter, and recombinase sequences into the Oxtr locus via the CRISPR/Cas9 system, Figure 1. A series of Oxtr knock-in mouse lines generated by CRISPR/Cas9 genome editing. a, Targeting strategy to insert a functional cassette just upstream from the translational stop codon (TGA) of mouse Oxtr gene locus is schematically outlined. Oxtr gene has two coding regions shown as solid black bars that are divided by a long intron. Its 39 UTR shown as an open bar is also relatively long. PAM and guide sequences to recruit CRISPR components onto the Oxtr stop codon are depicted at the top. Solid gray bar indicates a knock-in cassette. Note that the endogenous Oxtr gene configuration including the long intron and 39 UTR are intactly preserved in the resulting knock-in allele even after the genome editing. For comparison with already-existing knock-in lines, see Extended Data Figure 1-1. b, Four knock-in alleles generated in this study are schematically depicted. Epitope tag sequences (1ÂPA, 3ÂPA, or 3ÂHA) are fused 39 to the Oxtr coding sequences. T2A self-cleaving peptide sequences are employed for bicistronic functional cassette expression (tdTomato, iCre, and ERT2) in Oxtr-expressing cells. Complete genomic DNA sequences and genotyping primer sequences for four knock-in lines are supplied as .dna files (Extended Data 1 DNA sequence files 1-1, 1-2, 1-3, 1-4). Detailed genotyping PCR conditions are listed in Extended Data Figure 1-2. the guide RNA sequences closest to the stop codon TGA in the second exon (depicted in Fig. 1a) were selected. Two parts of the CRISPR guide RNA, crRNA and tracrRNA, were chemically synthesized (FASMAC) and recombinant Cas9 protein (EnGen Cas9 NLS S. pyogenes) was purchased from New England Biolabs. For pronuclear injections, equimolar crRNA and tracrRNA were mixed and annealed (94°C, 2 min, then at room temperature for 10 min).
Generation of the Oxtr 13PA-tdTom mouse line To insert 1ÂPA-tag, T2A self-cleaving peptide, and tdTomato sequences just upstream from the stop codon (Fig. 1b), the targeting vector containing those sequences flanked by 2.9-kb homology arms on both sides was constructed by the In-Fusion cloning system (Takara Bio). The mixture of annealed crRNA/tracrRNA, Cas9 protein, and the targeting vector was injected into mouse fertilized eggs at the concentrations of 38, 50, and 25 ng/ml, and the eggs were transferred into the oviduct of surrogate mothers to raise pups.

Generation of the Oxtr 33PA mouse line
To insert three tandem copies of PA-tag sequences just upstream from the stop codon (Fig. 1b), 3ÂPA-tag sequences flanked by 129-and 118-bp homology arms were artificially synthesized (Integrated DNA Technologies). To avoid unwanted excision of PA-tag sequences by the internal recombination among three copies, synonymous substitutions were introduced into each tag sequences. By using this synthesized DNA as the template, long single-stranded DNA (ssDNA) donors were prepared with the Phospho-PCR method (Guide-it Long ssDNA Production System, Takara Bio; Inoue et al., 2021). The mixture of annealed crRNA/ tracrRNA, Cas9 protein, and long ssDNA donor was injected into mouse zygotes at the concentrations of 38, 50, and 25 ng/ml, and the zygotes were transferred into the oviduct of surrogate mothers to raise pups.
Generation of the Oxtr 13PA-iCre mouse line To insert 1ÂPA-tag sequences, T2A self-cleaving peptide sequences, and codon-improved Cre recombinase (iCre; Shimshek et al., 2002) sequences just upstream from the stop codon (Fig. 1b), the targeting vector containing those sequences flanked by 245-and 294-bp homology arms was constructed via the In-Fusion cloning system. By using this vector as the template, long ssDNA donors were synthesized by the Phospho-PCR method. The mixture of annealed crRNA/tracrRNA, Cas9 protein, and long ssDNA donor was injected into mouse zygotes at the concentrations of 38, 50, and 18.5 ng/ml, and the zygotes were transferred into the oviduct of surrogate mothers to raise pups.
Generation of the Oxtr 33HA-iCreERT2 mouse line To insert three copies of HA-tag, T2A self-cleaving peptide, iCre, and mutated ligand binding domain of the human estrogen receptor (ERT2) sequences (Indra et al., 1999) just upstream from the stop codon (Fig. 1b), the targeting vector containing those sequences flanked by 2-kb homology arms on both sides was constructed by combining artificial gene synthesis (Integrated DNA Technologies) and In-Fusion cloning system. Synonymous base substitutions were introduced into each HA-tag sequence to obviate unwanted internal recombination among them. The mixture of crRNA/tracrRNA complex, Cas9 protein, and the targeting vector was injected into mouse 1-cell embryos at the concentrations of 50, 100, and 15 ng/ml, and the embryos were transferred into the oviduct of surrogate mothers to raise pups.

Validation of the founder knock-in mice and their backcrossing
Genomic DNAs were prepared from newborns' tail by proteinase K treatment in lysis buffer (Takara Bio). Knock-in mice were screened by PCR using Tks Gflex DNA polymerase (Takara Bio) or PrimeSTAR GXL DNA polymerase (Takara Bio). To exclude the possibilities that the donor DNAs had been randomly integrated into the genome, a subset of genotyping primers was designed external to the donor DNAs' homology arms. For knock-in candidates, PCR products were analyzed by Sanger sequencing to confirm carrying correct insertions. Obtained founders were backcrossed with C57BL/6J to get F1 heterozygous progenies, which were again sequence-verified to harbor the designed knock-in allele. F1 heterozygous mice were then intercrossed to generate homozygous knock-in mice. For histologic, electrophysiological, and viral analyses, homozygous knock-in mice after F3 generations were used.

Antibodies and viral vectors
Primary antibodies and dilutions used were guinea pig anti-tdTomato (

Immunohistochemistry
Brain samples were prepared from 2-to 11-week-old male homozygous knock-in mice and immunohistochemically analyzed by standard protocols. Briefly, brains were intracardially perfused with 4% paraformaldehyde (PFA) in PBS (pH 7.4), postfixed with the same fixative for 16 h, and cryo-protected with 30% sucrose in PBS, then embedded in OCT compound (Sakura Finetek Japan) to be frozen. Brains were sliced into 40-mm coronal sections using a Leica cryostat. Sections were treated in blocking buffer (10% normal donkey serum, 0.3% Triton X-100 in PBS) for 1 h at room temperature with gentle agitation. Sections were then incubated with primary antibodies in blocking solution (3% normal donkey serum, 0.3% Triton X-100 in PBS) overnight at 4°C. After being washed with PBS, sections were treated with appropriate Alexa Fluorconjugated secondary antibodies (Thermo Fisher Scientific, Abcam, or Jackson ImmunoResearch) and DAPI for 1 h at room temperature. After washing, sections were mounted onto glass slides with Permafluor (Thermo Fisher Scientific) or VECTASHIELD Vibrance antifade mounting medium (Vector Laboratories). Fluorescent images were obtained on a KEYENCE BZ-X710. For confocal and super-resolution imaging, an OLYMPUS SpinSR10 was used to take zstacked images. Movies were constructed by using an OLYMPUS cellSens software.

Intravenous injection of AAV-PHP.eB vector to the
After anesthesia with isoflurane (4%), 100 ml of viral solution (AAV-PHP.eB CBh_FLEX-GFP WPRE; 2.5 Â 10 12 vg/ml) was intravenously injected into the retro-orbital sinus of Oxtr 1ÂPA-iCre/1 mice using a 1.0-ml syringe with a 27-gauge needle. Four weeks after the injection, the mice were anesthetized with sodium pentobarbital (50 mg/kg body weight, intraperitoneal) and transcardially perfused with PBS followed by 4% PFA in 0.1 M phosphate buffer (pH 7.4). The brains were removed, postfixed in the same fixative overnight at 4°C, and immersed in 15% and 30% sucrose in PBS overnight for cryoprotection. Free-floating sections (40 mm) were cut using a sliding microtome (REM-710; Yamato Kohki Industrial Co, Ltd.) and incubated in PBS with 0.3% Triton X-100 and 1% bovine serum albumin (BSA) for 1 h at room temperature. Then, the sections were incubated for 12 h at room temperature with primary guinea pig anti-GFP antibody (1:1000; Frontier Institute) diluted in PBS with 0.3% Triton X-100, 2% BSA. The sections were washed in PBS and incubated for 2 h in Alexa Fluor 488-conjugaed species-specific secondary antibodies (Thermo Fisher Scientific) diluted 1:500 in PBS with 0.3% Triton X-100 and 2% BSA. Fluorescent images of GFP were captured using a fluorescent microscope (BZ-X710, KEYENCE).
Induction of iCre activity in the Oxtr 33HA-iCreERT2 line by oral administration of TMX Oxtr 3ÂHA-iCreERT2: : Gt(ROSA)Sor Ai9 double heterozygous males at four weeks were fed with powdered diet supplemented with tamoxifen (TMX; Sigma-Aldrich T5648) at the concentration of 2 mg/g for five consecutive days by using feeding jars (Yoshinobu et al., 2021;n = 5). For the control experiments, double heterozygous males at four weeks were fed with normal diet for five consecutive days (n = 3). After five more days, brains were sampled at five weeks and immunostained as described above. To analyze effects of chronic TMX treatment on Oxt and Oxtr expressions, five-week-old homozygous Oxtr 1ÂPA-tdTom males were fed with the same protocol (n = 3). Normal diet was fed to mice with the same genotype as the control (n = 3). Brains were sampled at six weeks for immunostaining.

Material availability
Four knock-in mouse lines generated in this study will be made available with a materials transfer agreement.

Seamless insertions of epitope tag and functional gene cassettes into the Oxtr locus via CRISPR/Cas9 genome editing
To insert the epitope tag or reporter/recombinase sequences in-frame and 59 to the stop codon of the Oxtr gene second exon, we selected the guide RNA sequences depicted in Figure 1a. The targeting strategies were basically the same for four knock-in lines; one or three copies of epitope tag sequences were directly fused 39 to the Oxtr coding sequences. For bicistronic expression, we inserted T2A self-cleaving peptide sequences followed by functional gene cassettes (tdTomato reporter or Cre recombinase coding sequences). We used the T2A sequences which had been successfully used for knockin mouse generations, to simultaneously express two separate proteins at nearly equimolar amounts from the same mRNA (Trichas et al., 2008;Tang et al., 2009).
We next employed the cloning-free CRISPR/Cas9 system in which chemically synthesized guide RNAs (crRNA and tracrRNA) were combined with recombinant Cas9 nuclease protein to induce highly efficient genome editing in mouse fertilized eggs (Aida et al., 2015;Inoue et al., 2021), and successfully generated one founder mouse for each knock-in design. We crossed these knock-in founders with C57BL/6J mice to obtain heterozygous F1 generations, which were sequence-verified to ensure that they correctly harbored the designed knock-in cassettes. We then intercrossed F1 heterozygous mice to obtain homozygous F2 generations. For histologic, electrophysiological, and viral analyses, we prepared homozygous knock-in mice after F3 generations. Complete genomic DNA sequences and genotyping primer sequences for the four knock-in lines are supplied as .dna files (Extended Data 1 DNA sequence files). Detailed genotyping PCR conditions are listed in Extended Data Figure 1-2.
The epitope tagging strategy precisely visualized the endogenous Oxtr protein expression profiles and ensured the reliability of the knock-in strategy As depicted in Figures 1b, 2a, we first designed a dualpurpose Oxtr 1ÂPA-tdTom allele, in which one copy of PAtag can monitor endogenous Oxtr protein localization and the portion of the tdTomato cassette serves as a bright fluorescent reporter. To evaluate PA-tag and tdTomato expression profiles in this line, we used homozygous male brains at two weeks [postnatal day (P)14] because Oxtr expression levels in mouse brains have been reported to peak at around two to three weeks (Newmaster et al., 2020). In the figures, we mainly presented the section equivalent to bregma À1.82 mm, as this level contains several well-known regions that have been reported to express Oxtr. Although tdTomato fluorescence could be detected without immunostaining as described later (Extended Data Fig. 3-1a,b), we double-stained the sections with anti-tdTomato and anti-PA-tag antibodies to validate their colocalization. As a result, we observed abundant tdTomato reporter signals in the cortical region (Newmaster et al., 2020), hippocampal dentate gyrus (DG) and anterior CA2 (aCA2)/CA3 (Raam et al., 2017), paraventricular thalamus (PVT; Newmaster et al., 2020;Watarai et al., 2020), central nucleus of amygdala (CeA; Yoshida et al., 2009;Mitre et al., 2016), dorsal endopiriform nucleus (DEn; Sharma et al., 2019), and posteromedial cortical amygdaloid area (PMCo; Newmaster et al., 2020;Fig. 2d), where considerable level of Oxtr expression has already been reported. Consistently, we also detected the presence of endogenous Oxtr protein by the Cterminally fused PA-tag in the same regions (Fig. 2e); however, the signal intensity was not strong enough for higher magnification, as described below. These matched results indicated that the epitope tagging and T2A-dependent bicistronic strategies we employed properly functioned as expected. We therefore applied the same design for the Oxtr 1ÂPA-iCre allele (Fig. 1b).
In the magnified confocal image of the primary somatosensory cortex barrel field (S1BF) in the Oxtr 1ÂPA-tdTom brain (Fig. 2h), tdTomato and PA-tag signals showed significant colocalization. However, we noticed that the use of anti-PA-tag antibody resulted in high background staining within brain tissue. To increase the signal-tonoise (S/N) ratio, we next generated the Oxtr 3ÂPA allele in which three tandem copies of PA-tags were simply fused to the C terminus of Oxtr (Figs. 1b, 2b). Consequently, the PA-tag signals from this 3ÂPA allele (Fig. 2f) perfectly coincided with those from 1ÂPA in the Oxtr 1ÂPA-tdTom allele (Fig. 2e), and its S/N ratio was improved. On comparing the magnified confocal images of the S1BF, the PA-tag signals in the Oxtr 3ÂPA line (Fig. 2i) were much clearer than those in the Oxtr 1ÂPA-tdTom allele (Fig. 2h), along with relatively lower background staining. In Figure 2i-l, we focused on SST-expressing interneurons in Layer IV, since Oxtr-expressing interneurons have been found in some brain areas to recruit local inhibitory circuits, which might be a major mode of oxytocinergic modulation in the brain (Froemke and Young, 2021), as described in Discussion.
We further designed a dual-purpose Oxtr 3ÂHA-iCreERT2 allele (Figs. 1b, 2c). We newly employed three copies of HA-tags for endogenous Oxtr protein visualization, because a good antibody with lower background staining in the brain tissue was commercially available. To induce spatiotemporally controlled Cre recombinase expressions in mouse brains, we fused a mutated ligand-binding domain of the human estrogen receptor (ERT2; Indra et al., 1999) with iCre recombinase. The detailed results for iCre inductions are described later. As shown in Figure 2g, the HA-tag signals from this 3ÂHA allele nicely coincided with those from the 1ÂPA and 3ÂPA alleles (Fig. 2e,f). On the magnified confocal image of Oxtr 3ÂHA-iCreERT2 brain shown in Figure 2k, we observed clear HA-tag signals, with relatively low background staining. The fact that three independently generated epitope-tagged lines resulted in exactly the same expression profiles confirms the reliability of our knock-in strategies.
Oxtr 33PA and Oxtr 33HA-iCreERT2 mouse lines enabled visualization of endogenous Oxtr protein at the subcellular level We next employed spinning-disk confocal microscopy technology to observe epitope tagged Oxtr-expressing neurons at super-resolution. First, we obtained the Zstacked 3-D image of a 3ÂPA-tagged neuron in the S1BF, which was identified as an SST-positive interneuron ( Fig.  2j; Movie 1). We successfully detected the distribution of PA-tagged endogenous Oxtr proteins on the plasma membrane, as expected as a member of GPCRs, whereas SST was localized in the cytoplasm. This is the first evidence showing that Oxtr proteins reside both on the cell body and on nerve fibers. The super-resolution image from a 3ÂHA-tagged neuron further replicated the above observation, supporting the reliability of this approach ( Fig. 2l; Movie 2).

C-terminally fused epitope tags did not disrupt the Oxtr protein function in vivo
Although our experiments have confirmed the advantage of epitope tagging in visualizing endogenous protein localization, the concern remained that C-terminally fused tags may alter Oxtr protein conformation and/or stability, affecting its functions. Since Oxt plays essential roles in lactation, pups from Oxtr knock-out mothers (Oxtr À/À ) have been reported to die soon after birth, because of the mother's defects in breastfeeding (Takayanagi et al., 2005;Hidema et al., 2016; Extended Data Fig. 2-1a). Therefore, we confirmed that all the homozygous mothers of our four knock-in lines had successfully nursed their pups. As shown in Extended Data Figure 2-1, the pups' digestive tracts were filled with milk (Extended tdTomato reporter and iCre-ERT2 recombinase are bicistronically expressed in the cytoplasm by employing the T2A self-cleaving peptide system. d, e, Expression profiles of tdTomato reporter and 1ÂPA-tag in the Oxtr 1ÂPA-tdTom homozygous brain from two-week-old male are arranged (equivalent to bregma À1.82 mm). tdTomato expressions are observed in the areas that have been reported to express Oxtr, such as the S1BF, hippocampal DG and aCA2/CA3, PVT, CeA, DEn, and PMCo. Notably, PA-tag signals are also observed in the same regions, indicating the T2A-dependent bicistronic expression system works well. f, Expression profiles of 3ÂPA-tag in the Oxtr 3ÂPA homozygous brain from two-week-old male nicely match up with those of tdTomato in d and 1ÂPA-tag in e. g, Expression profiles of 3ÂHA-tag in the Oxtr 3ÂHA-iCreERT2 homozygous brain from two-week-old male well coincide with those of 1xPA-tag in e and 3ÂPA-tag in f. Note that independently generated three epitope-tagged lines result in the same expression profiles, ensuring the reliability of knock-in strategies. h, Magnified confocal image of the boxed area (S1BF, Layer IV) in d, e indicates tdTomato signals nicely coincide with PA-tag staining. However, this line is not suitable for higher magnification because of its unwanted background staining of PA-tag antibody. i. Enlarged confocal image of the boxed area (S1BF, Layer IV) in f shows clearer signals derived from 3ÂPA-tag than those from 1ÂPA-tag in h. j, Superresolution image of a SST-positive PA-tagged Oxtr-expressing neuron indicates PA-tag signals are observed both on the surface of cell body and on the neurites, whereas SST is localized in the cytoplasm. 3-D movie of this interneuron is supplied as Movie 1. k, Magnified confocal image of the boxed area (S1BF, Layer IV) in g shows much clearer signals derived from 3ÂHA-tag than those from 1ÂPA-tag in h. l, Super-resolution image of an SST-positive HA-tagged Oxtr-expressing neuron indicates HA-tag signals are observed on the surface of cell body and on the neurites, reconfirming the result from 3xPA-tag knock-in line in j. 3-D movie of this interneuron is supplied as Movie 2. Data Fig. 2-1b,c) and the mother's mammary glands were well developed (Extended Data Fig. 2-1d), indicating that the peripheral Oxt signaling pathway was intact in the homozygous females. In addition, we performed electrophysiological recordings in tdTomatopositive neurons from the homozygous Oxtr 1ÂPA-tdTom male, to confirm their responsiveness to agonist administration (details are shown in Fig. 3 and described later), supporting that 1ÂPA tagged Oxtr proteins could sufficiently mediate Oxt signals in the brain. These results indicate that fusion of one or three copies of the PA-tag or HA-tag to the C terminus of Oxtr did not affect receptor functions in vivo. It has already been reported that a much larger tdTomato protein fused C-terminally to the k opioid receptor, which is also a member of the GPCR family, did not disrupt neuronal circuitries in the knock-in mouse brain (Chen et al., 2020). Therefore, we did not expect that the relatively smaller epitope tags we had used in the present study would affect Oxtr functions. Consistent with this assumption, all the knock-in lines we generated could be maintained by crossing homozygous males with homozygous females (Extended Data Fig. 2-1e,f), without the need for laborious genotyping procedures. This is a notable advantage of our new lines.

The tdTomato reporter in the Oxtr 13PA-tdTom line precisely labeled Oxtr-expressing neurons
Since tdTomato has been developed as a nonaggregating, extremely bright red fluorescent protein suitable for live imaging (Shaner et al., 2004), we were able to detect tdTomato signals in our homozygous Oxtr 1ÂPA-tdTom brains without immunostaining (Extended Data Fig. 3-1a,  b), allowing us to perform patch-clamp recordings on these cells (Fig. 3a,b). We hypothesized that tdTomato-positive neurons should correctly respond to an Oxtr agonist, TGOT, if the tdTomato expressions were faithfully regulated parallel to endogenous Oxtr expressions.
A subset of both glutamatergic and GABAergic neurons in the PFC has been reported to express Oxtr, which modulate female sociosexual behavior (Nakajima et al., 2014) and social memory (Tan et al., 2019). We first investigated the distribution of tdTomato-positive neurons in the PFC using immunohistochemistry, particularly in the prelimbic cortex (PrL) of Oxtr 1ÂPA-tdTom homozygous brain (Extended Data Fig. 3-1). We found that tdTomato-positive neurons mainly resided in Layers II and V of the PrL (Extended Data Fig. 3-1c,d). Consistent with a previous report (Tan et al., 2019), we identified the neurons in Layer II by Cux1 expression as glutamatergic neurons (Extended Data Fig. 3-1e,f), while the neurons in Layer V were confirmed by SST expression as GABAergic interneurons (Extended Data Fig. 3-1g,h).
To test whether these tdTomato-positive neurons in Oxtr 1ÂPA-tdTom mice faithfully respond to TGOT application, we used four-to six-week-old homozygous Oxtr 1ÂPA-tdTom males and perform in vitro electrophysiological analyses using a cell-attached patch-clamp configuration (Fig. 3a,b). We tested the neurons in Layers II and V separately (Fig. 3a), to test the possibility that they might have distinct responsiveness to Oxt ligands to modulate synaptic sites or membrane potentials. As the results, in tdTomato-positive Layer II neurons, TGOT significantly increased the spontaneous firing rate from 2.7 6 1.0 to 7.6 6 2.6 Hz (n = 6; p = 0.0312, Wilcoxon test; Fig.  3c,d). In tdTomato-positive Layer V neurons, TGOT also considerably increased the spontaneous firing rate from 0.99 6 0.7 to 7.3 6 3.3 Hz (n = 6; p = 0.0312, Wilcoxon test; Fig. 3e,f). Additionally, we confirmed that these TGOT-induced increases in spontaneous firing rates in both Layer II and V neurons were inhibited by the presence of an Oxtr antagonist, OTA (Extended Data Fig.  3-2a,b). These results clearly demonstrate that tdTomatopositive cells in Oxtr 1ÂPA-tdTom mice express "functional" Oxtr and appropriately respond to its ligand, even in the presence of a C-terminally fused PA-tag. Hence, tdTomato reporter signals in our Oxtr 1ÂPA-tdTom line are a reliable marker for Oxtr-expressing cells.
The Oxtr 13PA-iCre line sufficiently drove recombinase activities in the Oxtr-expressing neurons To newly generate an effective Cre driver, we employed the 2A peptide-based bicistronic expression system (Trichas et al., 2008) and codon-improved Cre recombinase (iCre) sequences (Shimshek et al., 2002). For the evaluation of Cre activities in the brain, we first crossed our Oxtr 1ÂPA-iCre line with the well-known Cre reporter, Ai9 (Madisen et al., 2010) to obtain double heterozygous mice. As depicted in Figure 4a, Cre-dependent stop signal excision induces tdTomato expression from the Ai9 allele. We analyzed tdTomato expression profiles in Oxtr 1ÂPA-iCre/1 ::Gt(ROSA)Sor Ai9/1 male brains at P7, P14, and P21. We observed tdTomato signals already at P7 (Fig. 4b) and found that the reporter signals became more abundant at P14 and P21 (Fig. 4c,d). Once recombination occurs, the reporter gene remains recombined thereafter, and consequently, the reporter is expressed even in the absence of ongoing recombinase expressions. In other words, a significant proportion of tdTomato expression in adult stages does not reflect real-time Cre activity but instead reflects a cumulative effect of earlier events. Therefore, we concluded that the Ai9 reporter was unsuitable for our purpose of evaluating Cre expressions in adult brains.
We next tried to immunostain Cre proteins in the brain tissue, using a commercially available anti-Cre recombinase antibody (clone 2D8, MAB3120, MERCK). This antibody has been reported to be successful in staining Cre proteins in transgenic lines carrying multiple copies of transgene (Mizuguchi et al., 2012). However, we could not detect Cre in the brain tissues of our iCre knock-in mouse line using this antibody.
To circumvent these issues, we employed AAV-PHP.eB vector harboring CBh-FLEX-GFP (kindly gifted by Hirokazu Hirai, Brain/MINDS project). AAV-PHP.eB has been developed to achieve efficient transport across the bloodbrain barrier in adult mammalian brains for broad gene delivery (Chan et al., 2017). As depicted in Figure 4e, GFP expressions can be induced by Cre-dependent inversion. We therefore retro-orbitally injected AAV-PHP. , and P21 are arranged (equivalent to bregma À1.82 mm). Constitutive tdTomato expressions from the Gt(ROSA)Sor locus after Cre-dependent stop signal excision gradually accumulate over time. tdTomato signals are observed already at P7 and then become more abundant at P14 and P21. e, Schematics of Oxtr 1ÂPA-iCre allele and AAV-PHP.eB vector harboring CBh_FLEX-GFP. GFP expression can be induced by Cre-dependent inversion. f, Expression profiles of GFP in the Oxtr 1ÂPA-iCre heterozygous brain (bregma À1.82 mm) from 11-week-old male to which AAV-PHP.eB vector has been retro-rbitally injected at six weeks. Other section levels are shown in Extended Data Figure 4-1. g-j, Enlarged images of the boxed areas in f are arranged. GFP expressions are observed in the areas that have been reported to express Oxtr, such as the S1BF, hippocampal DG and aCA2/CA3, PVT, CeA, and DEn, indicating this mouse line sufficiently drives Cre activities in Oxtr-expressing cells. k, Schematic diagram of Oxtr 1ÂPA-tdTom allele is depicted. l, tdTomato expression profiles in the Oxtr 1ÂPA-tdTom homozygous brain (bregma À1.82 mm) from 11-week-old male are arranged. Other section levels are shown in Extended Data Figure 4-1. m-p, Enlarged images of the boxed areas in l are arranged. These tdTomato expression profiles correspond to the GFP expressions shown in g-j induced by iCre on AAV-PHPeB administration. eB vector into six-week-old heterozygous males and analyzed their brains after four weeks. As the results, we observed abundant GFP signals in the cortical region, hippocampal DG and aCA2/CA3, PVT, CeA, DEn, and so forth, where Oxtr expression has previously been reported ( Fig. 4f-j, bregma À1.82 mm). Other section levels are shown in Extended Data Figure 4-1.
As we had designed the Oxtr 1ÂPA-tdTom and Oxtr 1ÂPA-iCre alleles based on the same concepts (Figs. 1b, 4e,k), their expression profiles should closely resemble each other. Predictably, we found that the tdTomato expression profiles from 11-week-old homozygous Oxtr 1ÂPA-tdTom male brains (Fig. 4l-p) were quite similar to those of iCre visualized by AAV-PHP.eB administration (Fig. 4f-j). Other section levels, shown in Extended Data Figure 4-1, confirm the trend. These results ensured the reliability of our new Cre driver line.
A new genetic tool, Oxtr 33HA-iCreERT2 line, could be used to target Oxtr-expressing neurons in a spatiotemporally controlled manner We next investigated whether the Oxtr 3ÂHA-iCreERT2 allele properly functioned as an inducible Cre driver. line. a, Schematics of Oxtr 3ÂHA-iCreERT2 allele and Gt(ROSA)Sor Ai9 allele are depicted. TMX binds to iCre-ERT2 expressed in the cytoplasm, resulting in iCre translocation into the nucleus to flip out stop signal from Ai9 allele and induce tdTomato expression. b, Schematic diagram of experimental protocol. Oxtr 3ÂHA-iCreERT2 ::Gt(ROSA)Sor Ai9 double heterozygous mice can freely access to TMX-supplemented diet from day 0 to day 5. At day 10, brains are sampled and immunostained. For control experiments, TMX-free normal diet is fed to same genotype mice. As shown in Extended Data Figure 5-1, 5-d TMX treatment has no effects on Oxt/Oxtr expression profiles in five-week-old males. c, TMX-free normal diet does not induce any leaky tdTomato signals in the brain from five-week-old male (bregma À1.82 mm). d, TMX-supplemented diet induces tdTomato expressions in the cortical region, hippocampal DG and aCA2/CA3, PVT, CeA, DEn, and PMCo, where Oxtr expressions has been reported. The density of the tdTomato-positive cells is much sparser than that induced by the constitutive iCre line shown in Figure 4d To detect Cre activity in the brain, we crossed the Oxtr 3ÂHA-iCreERT2 line with Ai9 to obtain double heterozygous mice. As depicted in Figure 5a, TMX binds to iCre-ERT2 expressed in the cytoplasm, resulting in iCre translocation into the nucleus to flip out the stop signal from Ai9 allele and elicit tdTomato expressions. We orally administered TMX to four-week-old males for five consecutive days by feeding them a TMX-supplemented powdered diet (Fig. 5b). For the control experiments, we fed four-week-old males with normal diet for five consecutive days. After five more days, we sampled and analyzed their brains at the level of bregma À1.82 mm. As shown in Figure 5c, the normal diet did not induce any leaky tdTomato reporter expression. Contrarily, the TMXcontaining diet successfully induced iCre recombinase activity in the cortical region, hippocampal DG and aCA2/ CA3, PVT, CeA, DEn, and PMCo, where Oxtr expression has previously been reported (Fig. 5d-i). In the constitutive iCre line, too many cells were labeled with tdTomato to be able to clearly distinguish each neuron (Fig. 4d). However, reporter expression was only sparsely induced in the inducible iCre line, allowing the identification of cell morphology at single-cell resolution ( Fig. 5d-i). As a result, we observed two types of Oxtr-expressing neurons with different morphologies and properties in the S1BF (Fig. 5i). Consistent with the previous reports (Nakajima et al., 2014;Tan et al., 2019), some tdTomato-labeled neurons in Layer II turned out to be Cux1-positive glutamatergic neurons (Fig. 5j-l), while other tdTomato-labeled neurons in Layer IV were identified as SST-positive interneurons ( Fig. 5m-o).
These results demonstrated that the first-time inducible-Cre driver for Oxtr-expressing neurons efficiently drives Cre-mediated recombination in a spatiotemporally controlled manner on TMX administration.
Since subcutaneous implantations of a pellet containing estrogen into gonadectomized rodents have been reported to affect Oxt/Oxtr transcript levels (Bale and Dorsa, 1995;Vanya et al., 1997;Nomura et al., 2002;Patisaul et al., 2003;Garcia et al., 2017) and radioactively labeled OTA binding to Oxtr (Young et al., 1998), chronic administrations of TMX, a selective estrogen receptor modulator, might change Oxt/Oxtr expression levels in the Oxtr 3ÂHA-iCreERT2 line. To analyze the effect of 5-d oral TMX treatment on Oxt/Oxtr expressions, we used homozygous Oxtr 1ÂPA-tdTom mice, in which real-time Oxtr expressions can be monitored as tdTomato signals, and examined Oxt-expressing neurons by immunostaining (Extended Data Fig. 5-1a). Although long-term (12 d to three months) estrogen treatments have been reported to significantly increase Oxt transcript levels in the paraventricular hypothalamic nucleus (PVN; Nomura et al., 2002;Patisaul et al., 2003) and supraoptic nucleus (SON; Garcia et al., 2017), our relatively short-term (5 d) TMX treatment did not affect Oxt expression levels in these nuclei (Extended Data Fig. 5-1b). In addition, while long-term estrogen treatments have been reported to affect Oxtr transcript levels in the bed nucleus of stria terminalis (BNST; Garcia et al., 2017), ventromedial hypothalamus (VMH; Bale and Dorsa, 1995;Vanya et al., 1997), and medial amygdala (MeA; Garcia et al., 2017), our 5-d TMX treatment had no effect on Oxtr expression profiles in these brain areas (Extended Data Fig. 5-1c,d). Furthermore, while radioactively labeled OTA binding to Oxtr has been reported to increase in the DEn and PMCo by long-term estrogen treatment (Young et al., 1998), Oxtr expression levels itself were not affected on our TMX administrations (Extended Data Fig. 5-1d). From these results, we concluded that 5-d TMX treatments have little effect on Oxt/ Oxtr expression profiles in young male mice. However, we need to carefully interpret the results obtained from Oxtr 3ÂHA-iCreERT2 line by chronic TMX administrations because estrogens play key roles in mediating social behaviors (Laredo et al., 2014;Ervin et al., 2015) as described in Discussion.

Discussion
Genetic tools that provide access to Oxtr-expressing cells remain limited. To improve this situation, we have newly generated a series of genome-edited mouse lines that are to make a valuable contribution to the growing research on Oxt. We have created simple and seamless knock-in designs that maximally preserve the endogenous transcriptional regulation of the Oxtr gene.
The first achievement of the present study was the successful use of epitope tags to visualize Oxtr protein distributions at the subcellular level for the first time. Two strategies, the RNAscope-based in situ hybridization and the Oxtr-Venus knock-in mice, have previously been used for Oxtr visualization. However, these methods detect transcript expression, but not Oxtr protein localization per se. To overcome these limitations, we generated epitope tag knock-in mice by genome editing. We employed the PA-tag, a 12-aa epitope derived from human podoplanin, which could be detected using the rat monoclonal antibody, NZ-1 (Fujii et al., 2014). This epitope-antibody combination has been reported to show a high binding affinity of K D = 4.9 Â 10 À10 (M), which is much stronger than those of HA, FLAG, and other affinity tag systems, and which has been successfully used to many biological procedures in vitro (Fujii et al., 2014). In addition, we have previously reported the applicability of the PA-tag system in mouse brains and testes in in vivo studies (Arimura et al., 2020;Kimura et al., 2021). Here, we triplicated the PAtag to improve the S/N ratio of immunostaining, by which the background staining in brain tissues were reduced, allowing us to successfully carry out super-resolution imaging of Oxtr protein. We also used the HA-tag because it has been successfully employed for SLENDR, which enabled in vivo protein labeling via genome editing in mammalian brains (Mikuni et al., 2016). Similarly, using three copies of the HA-tag enabled us to reconfirm the result from the 3ÂPA tag knock-in brains.
Although we have roughly demonstrated that Oxtr proteins are present not only on cell bodies but also on neurites, it remains unclear on which part of neurites (axons, dendrites, or both) they reside. To date, three types of oxytocinergic modulation on neurons have been proposed (Froemke and Young, 2021). As the first-order modulation, Oxt directly depolarizes excitatory cells that express Oxtr. In the second-order modulation, Oxt acts on inhibitory interneurons expressing Oxtr, to regulate synaptic inhibitions. Furthermore, in the thirdorder modulation, Oxtr is believed to reside on axon terminals, where the presynaptic Oxtr activation enhances transmitter release and indirectly impacts on other excitatory transmissions (Dölen et al., 2013;Nardou et al., 2019). However, the presence of Oxtr in axon terminals has not yet been proven. Therefore, our in vivo super-resolution imaging could serve as a significant step toward deciphering the subcellular distribution of Oxtr protein in such studies.
The second achievement of the present study was the development of an Oxtr reporter line expressing the bright red fluorescent protein, tdTomato, which is suitable for live imaging. In vitro electrophysiological analyses were successfully conducted on tdTomato-positive neurons in our reporter line. Such analyses may not have been possible in the Oxtr-Venus line (Yoshida et al., 2009), because expression levels of Venus-positive cells have been reported to largely decline after three weeks (Newmaster et al., 2020). In addition, tdTomato's red fluorescence could be advantageous when calcium indicators such as the genetically encoded GCaMPs are used, as they mainly contain GFP and emit green fluorescence which overlaps with Venus signals.
Third, we generated both constitutive and inducible Cre drivers for Oxtr-expressing neurons. IRES elements, used in the existing two Oxtr-IRES-Cre knock-in lines, have been reported to occasionally cause lower expression of the downstream cistron because of factors such as experimental cell types or cloned gene cassettes (Ibrahimi et al., 2009). Accordingly, we employed the 2A peptide system, which allowed nearly equimolar expression of two cistrons (Trichas et al., 2008;Tang et al., 2009). As we paid much attention to maintain the endogenous Oxtr gene configuration in our knock-in allele, the design of our Oxtr-T2A-iCre is indeed different from that of Oxtr-T2A-Cre-D, available from The Jackson Laboratory (Daigle et al., 2018). While the latter inserted an exogenous bovine growth hormone poly A signal and ignored the Oxtr's own endogenous poly A sequence (Extended Data Fig. 1-1e), our knock-in mice simply made use of the endogenous poly A signal instead (Fig. 1b). Unlike the systematically generated Cre allele as a part of a big project, our iCre knock-in allele was carefully customized for the Oxtr locus, which is a promising development for future Oxt studies. In addition, our inducible Cre line was suitably responsive to TMX administration. Although intraperitoneal injection is the most popular TMX administration protocol, it may sometimes be toxic to cause the death of test animals. Therefore, we employed an oral administration protocol (Andersson et al., 2010;Yoshinobu et al., 2021) instead, to ensure the safe and sufficient induction of Cre activity.
TMX is a potent and clinically used selective estrogen receptor modulator that blocks effects of estrogens in breast tissues but facilitates estrogen actions in others such as uterus tissues (Osborne, 1998). Thus, chronic TMX treatment has the possibility to interfere with estrogen function in multiple tissues including the brain. Estrogens are involved in social behavior at multiple levels, from the detection and integration of social information to more complex behavior such as social preference, aggression, and social learning and memory (Ervin et al., 2015). As an example, in the hippocampus, estrogens are locally synthesized and modulate memory-related synaptic plasticity by the rapid nonclassical indirect pathway via synapse-localized estrogen receptors (Mukai et al., 2010;Laredo et al., 2014). We hence need to carefully consider the potential interference by TMX to the behavioral outcomes from Oxtr manipulations in using Oxtr 3ÂHA-iCreERT2 mice.
Oxt and arginine vasopressin (Avp) are evolutionally related molecules that differ only in two of the 9 aa. Since their receptors also show significant structural homology to one another, the selectivity of Oxt and Avp for their own receptor is not absolute. Considerable receptor crosstalk has been reported to occur among Oxt, Avp, and their receptors (Carter, 2017;Song and Albers, 2018). While exogenously administrated Avp can act on Oxtr in higher concentration and so can do Oxt on Avp receptor 1a (Avpr1a), limited lines of evidence indicate that endogenously released Oxt and Avp can produce functionally significant responses by activating each other's receptors (Song and Albers, 2018). Based on receptor autoradiography approach with relatively low resolution, Oxtr and Avpr1a appear to show distinct and nonoverlapping expression in the rodent brain (Dumais and Veenema, 2016). However, it remains still unknown whether the localization of Oxtr and Avpr1a is mutually exclusive at the synaptic levels, and thus the possibility of crosstalk following synaptic release of these peptides should rigorously be examined. In this context, our new genetic tools could play a part in clarifying these receptors' distribution with much higher resolution, deepening our understanding of ligand-receptor interactions for Oxt and Avp. Generation of Avpr1a genome-edited mouse lines might further help precisely visualize its endogenous expression profiles.
Oxt is expected as a potential therapeutic for social deficits represented by autism spectrum disorders. Managing socio-emotional behavior in health and disease to increase resilience to chronic stress has become much more important, especially given the ongoing COVID-19 pandemic (Grinevich and Neumann, 2021). Although the preclinical trials of exogenous Oxt administration via intranasal delivery are currently in progress (Quintana et al., 2021), second-generation strategies to potentiate endogenous Oxt signaling should be preferred in the future (Ford and Young, 2021). Moreover, a fundamental shift from the traditional brain-wide chronic pharmacological intervention to a circuit-specific approach is now anticipated, owing to significant advances in our understanding of the neural circuit mechanisms underlying social behavior in animal models (Ford and Young, 2021). Hence, to establish dependable and effective Oxt-based therapeutic interventions, we need reliable animal models to help uncover the mechanisms underlying Oxt-Oxtr circuitry. Our new genetic tools would play a significant role in meeting this demand and we are going to make the four knock-in lines generated in this study quickly and widely