Targeted expression of step-function opsins in transgenic rats for optogenetic studies

Rats are excellent animal models for experimental neuroscience. However, the application of optogenetics in rats has been hindered because of the limited number of established transgenic rat strains. To accomplish cell-type specific targeting of an optimized optogenetic molecular tool, we generated ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats that conditionally express the step-function mutant channelrhodopsin ChRFR(C167A) under the control of extrinsic Cre recombinase. In primary cultured cortical neurons derived from this reporter rat, only Cre-positive cells expressing ChRFR(C167A) became bi-stable, that is, their excitability was enhanced by blue light and returned to the baseline by yellow~red light. In bigenic pups carrying the Phox2B-Cre driver, ChRFR(C167A) was specifically expressed in the rostral parafacial respiratory group (pFRG) in the medulla, where endogenous Phox2b immunoreactivity was detected. These neurons were sensitive to blue light with an increase in the firing frequency. Thus, this transgenic rat actuator/reporter system should facilitate optogenetic studies involving the effective in vivo manipulation of the activities of specific cell fractions using light of minimal intensity.

Actuator/Reporter gene expression by the transfection of exogenous Cre gene. To examine the conditional expression of ChRFR(C167A)-Venus, we injected adeno-associated virus 2 (AAV2)-Cre virus vectors into the striatum or hippocampus of heterozygous rats (n = 3) obtained by crossing male ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats with wild-type female Long-Evans (LE) rats. At 3 weeks post-injection, Venus fluorescence was detected at all injection sites, and the signals were closely confined to the injection points ( Fig. 2A, Supplementary Fig. S2). By contrast, the expression of Venus was not detected on the other side of the hemisphere, which underwent PBS injection. At higher magnification, Cre signals were clearly detectable in the nuclei, and Venus signals were accumulated in the plasma membrane of those cells; importantly, these signals did not merge with each other (Fig. 2B).
Robustness of the actuator/reporter system. To validate the Cre-dependent expression in this actuator/reporter line, primary cultured cortical neurons of ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats were transfected with plasmids containing mCherry-NCre 28 . The mCherry-positive nuclei and Venus-positive plasma membranes were detected only in cells (n = 28) prepared from Tg-positive embryos (n = 5). By contrast, cells (n = 59) expressing mCherry-NCre prepared from Tg-negative siblings (n = 6) did not express ChRFR(C167A)-Venus (Fig. 3A,B). An unexpected STOP excision in loxP-based reporter animals is referred to as leaky expression 29,30 . To detect leaky expression in this actuator/reporter system, the DNA template was extracted from homogenized whole brains of ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats, and each allele in the BAC construct, with (497 bp) or without the STOP sequence (314 bp), was amplified using PCR. As a positive control, the DNA template extracted from the medullary tissue of bigenic pups obtained by crossing ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats and Phox2b_tTA-2A-Cre driver rats (Fig. 3C) was used. Only a single band corresponding to the allele with the STOP sequence was detected in the actuator/ reporter rat, indicating that background recombination was negligible (Fig. 3D). In subsequent experiments, a subpopulation of cortical neurons expressing ChRFR(C167A)-Venus under the control of the CaMKIIα promoter was used for the whole-cell patch clamp experiments.
Regulation of neuronal activity using two colours of light. It was well anticipated that the excitability of a neuron from the ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats could be manipulated by light when it expressed Cre. The excitability of a neuron in the primary culture was tested by current ramp injections under whole-cell current clamp (Fig. 4A). At first, the slope of a ramp was adjusted to generate several action potentials. Next, the same current ramp was injected after illumination with blue light (500 ms, 438 ± 24 nm) and SCIeNTIFIC RepORTs | (2018) 8:5435 | DOI:10.1038/s41598-018-23810-8 then red light (8 s, 643 ± 20 nm, 6.8 mWmm −2 ) starting after the current ramp injection. In every Cre-positive neuron (n = 16), the membrane potential was depolarized solely by blue light in a manner dependent on the power ( Supplementary Fig. S3). Although this depolarization was accompanied by action potentials in two of them, it was not in the others (n = 14) even at the maximal power of light (3 mWmm −2 ). However, the time to evoke the first action potential (t AP ) was significantly reduced during the current ramp injection by the preceding blue light in the Cre-positive neurons, but not in the Cre-negative neurons (Fig. 4B,C). On the other hand, when the primary culture was made from the littermate embryo without the transgenic allele, neither the depolarization nor the reduction of t AP was induced by any light even in the Cre-positive neurons (Fig. 4D).
Directive expression using Phox2B Cre-driver rats. Phox2b is a paired-like homeobox-encoding gene and a key regulator of autonomic neural crest derivatives and placode-derived visceral sensory ganglia. Phox2b is also expressed in pre-inspiratory (Pre-I) neurons in the parafacial respiratory group, constituting one of the respiratory rhythm generators in the medulla of newborn rats 31,32 . We previously reported that the expression of PHOX2B is a remarkable feature of Pre-I neurons and a useful marker for identifying the cytoarchitecture of neurons in the pFRG in the newborn rat medulla 33 . Newborn pups expressing ChRFR(C167A)-Venus under the control of Phox2b enhancer/promoter were obtained by crossing ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats with Phox2b_tTA-2A-Cre-driver rats (Fig. 3C). The functional expression of ChRFR(C167A) was evaluated in the newborn, bigenic rat for the following three issues. First, we investigated whether the Venus-positive The transgene construct, which was inserted into the mouse ROSA26 BAC clone, comprised the CAG promoter, a cassette for the neomycin resistance gene with four STOP signal repeats (pA) flanked by loxP sites, and a sequence containing the ChRFR-C167A open reading frame. After Cre-mediated recombination, ChRFR(C167A)-Venus expression was under the regulation of the generic CAG promoter. (B) PCR analysis of genomic DNA. The 726-bp band, which is recognized by the designed primer sets targeting the PGK-neo cassette sequence between two loxP sequences, was present in the ChRFR(C167A)-Venus reporter rats but not in the wild-type rats. A 100-bp DNA ladder was used as marker (M). A full uncropped image is available as Supplementary Figure S1A signals driven by Cre recombinase could be observed in Pre-I neurons of the pFRG. A Venus-positive signal was detected in the rostral pFRG in the superficial area just ventral to the facial nucleus (FN) in the medulla, where endogenous Phox2b immunoreactivity was detected (Fig. 5). Second, we compared the expression pattern of ChRFR(C167A)-Venus in Phox2b-Cre/floxed ChRFR(C167A) with that of tdTomato in Phox2b-Cre/floxed STOP tdTomato bigenic neonatal rats 15 (Supplementary Figs S4, S5). The expression of ChRFR(C167A)-Venus was consistently detected, and the expression patterns of the actuator were coincident with those of tdTomato throughout the brainstem-spinal cord sections. The reliability of this Cre-dependent recombination was further evaluated using VGAT-Cre BAC rat (NBRP-0839, Supplementary Fig. S6). Even in other brain regions such as the hippocampus, cortex, cerebellum, and striatum in the bigenic rats, the expression of ChRFR(C167A)-Venus was successfully induced and coincident with the distribution of GABAergic neurons. Third, the effects of photostimulation were examined on the Phox2b positive pFRG neurons in the ex vivo brainstem-spinal cord preparation (n = 6, Fig. 6). The Phox2b-positive neurons were depolarized in the membrane potential with increased excitability during and after photostimulation by blue LED. Indeed, the firing frequency was significantly increased upon blue light stimulation. The neuronal excitability was up-regulated for several seconds after the end of photostimulation, but returned to baseline in tens of seconds (Fig. 6D).  The primary cultured cortical neurons were prepared from both Tg-positive and negative littermates, and mCherry-NCre was transfected using calcium phosphate method; Venus fluorescence (green) and mCherry-NCre (magenta). (B) Comparison of the fraction of doublepositive cells with ChRFR(C167A)-Venus and mCherry-NCre between littermates (Tg, n = 5 pups; wild-type, n = 6 pups). For each group, 25 fields of interest were randomly chosen, and the cell number was counted in a double-blind manner. *p < 0.001, Student's paired t-test. (C) Structure of the Phox2b_tTA-2A-Cre BAC transgene. The exon-intron structure of the Phox2b gene is shown in the upper part. The first exon that contains noncoding (a white rectangle) and coding (a black rectangle) regions, the second exon (a black rectangle), and the third exon which contains coding (a black rectangle) and noncoding (a white rectangle) regions are indicated as I, II, and III, respectively. The recombination targeted the first exon (around the ATG site of Phox2b gene) by inserting the tTA, 2 A peptide, and Cre recombinase with nuclear localization signal (NLS) coding sequences attached the rabbit ß globin gene polyadenylation sequence at 3′ end. (D) PCR assay of the genomic DNA extracted from the homogenized whole brain of a ChRFR(C167A)-Venus actuator/reporter rat. The allele of BAC with STOP sequence was detectable as 497 bp band. In the bigenic rat with Phox2B-Cre driver, 314 bp band became manifest as expected by the removal of Neo-STOP sequence. A full uncropped image is available as Supplementary Figure S1B.
the Phox2b_tTA-2A-NLS-Cre BAC and VGAT-Cre BAC transgenic strain. Our results indicate that the cells carrying the active promoter could be selectively labelled with Venus fluorescence and effectively depolarized by light by combining these actuator/reporter rats with an appropriate driver rat strain. Although ChRFR(C167A) was well characterized in the photocurrent kinetics and its photocurrent was more dependent on Na + and Ca 2+ than H + like ChR1 and 2 34,35 (Supplementary Fig. S8A and S8B, Supplementary Table S2), the validity of this system should be checked in each case for the practical expression in various subpopulations of neurons. It should also be noted that the ON/OFF kinetics of a photocurrent is dependent on the cellular environment such as temperature and pH 2 as well as the molecular state of rhodopsin 36,37 .
Rats and mice have frequently been used as fundamental model organisms for biomedical research. Although similar in many aspects, there are several obvious differences between these rodent species in functional anatomy, physiology, and pharmacology 38 . Interestingly, the characteristics of Phox2b-expressing cells in the parafacial region of neonatal rats are basically similar to those in mouse, but the roles in respiratory rhythm generation during development are distinct 39 . Thus, the establishment of a reliable transgenic system in rats for further studies has been desired, and the two BAC Tg lines introduced herein, namely, ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats and Phox2b_tTA-2A-Cre driver rats, would contribute to expanding the repertoire and offer an attractive means for precise optogenetic targeting.
The application of an optogenetic approach to larger animal models has been challenging because the light has to be widely delivered to activate an entire substructure of larger size. To overcome this problem, the animal model has to accommodate a substantial number of modified optic fibres that deliver strong light into a wider region in the tissue, or the photocurrent properties, including the light sensitivity of the opsins, have to be optimized. In one of the well-characterized step-function variants of ChRs, ChRFR(C167A), the magnitude of the photocurrent was approximated by a function of the turning-on (ON) rate constant, which is equivalent to τ ON −1 , and the irradiation time 25 . Therefore, even light at a weak power could effectively depolarize the neuronal membrane to become excitable with prolongation of the irradiation time. In the present study, ChRFR(C167A) was highly sensitive to blue light (50% activation; 0.08 mWmm −2 , Supplementary Fig. S3), which was less than 1/10 of the power required for 50% activation of native channelrhodopsin-2 (1.3 mWmm −2 ) 40 . When all synaptic inputs of the cultured cortical neurons from our ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats were pharmacologically blocked, the light-evoked depolarization was often subthreshold even with the maximal power in the Cre-positive neurons. However, the excitability of a neuron was actually increased by light as evidenced by the reduction of t AP . That is, the light should cause some of the sub-threshold EPSPs to become excitable in neurons in the network. Indeed, the irradiation of weak blue light (0.1-0.2 mWmm −2 ) onto the brainstem-spinal cord of rat neonates was sufficient to enhance the excitability of neurons in the pFRG (Fig. 6). To our knowledge, this actuator/reporter system is the first successful use of a step-function mutant channelrhodopsin in a transgenic rat. With the temporally precise tuning of the membrane potentials, this rat line is optimized for an experimental context that requires modulation of the spontaneous firing rate rather than the generation of every action potential 23,41 . For instance, the transition from a hyperpolarized to a depolarized state of the membrane potential is usually accompanied by a peculiar pattern of electrical activity intrinsic to its neuronal subclass 42,43 . The relevance of the periodic oscillations in a genetically defined cell population to the specific behavioural outcome could be directly tested by simply alternating the delivery of blue or red light into the targeted region 44 .
In summary, this actuator/reporter rat strain enables the control of identified neuronal activity by light under the visualization of Venus fluorescence. It appears to be suitable to optically manipulate deep in the tissue, since the spectral sensitivity of ChRFR(C167A) is more red-shifted than the SFOs derived from ChR2 25 and the brain tissue preferentially absorbs short wavelengths of light 45 . With the above-mentioned advantages and strengths of the relatively large body size and complex behaviour of rats, the transgenic lines reported herein would facilitate a variety of medical studies, including neuroscience.

Construction of ROSA26/CAG-floxedSTOP-ChRFR(C167A)-Venus BAC. The ROSA26/CAG-floxed
STOP-ChRFR(C167A)-Venus BAC was generated using a BAC recombineering method 15,46 with a BAC clone containing the mouse ROSA26 locus. Briefly, a BAC clone (RP23-244D9) from a C57BL/6 J mouse genomic BAC library (BACPAC Resource Center, Oakland, CA, USA) spanning ~183 kb upstream of ROSA26 exon 1 to ~34 kb downstream of ROSA26 exon 3 (total size: ~226 kb) was used. To modify the Rosa26 locus, a cassette containing the following components was constructed: a CAG promoter derived from pCAGGS 47 , loxP, FRT-flanked Kan/Neo cassette (from J. Takeda, Osaka University), 4 × poly(A) (from M. Yamamoto, Gunma University), ChRFR-C167A cDNA 48 , and SV40 poly(A). Through lambda red protein-mediated homologous recombination in E. coli EL250 49 , the above-mentioned cassette flanked by homologous fragments was inserted into the ROSA26 BAC clone at the desired location. The following primers were used to generate 5′ and 3′ homology regions: R5-FS, gtcgaCGTCGTCTGATTGGCTCTC and R5-RX, ctcgaGACTGGAGTTGCAGATCAC for the 5′ homology region; and R3-FS, gtcgACAGTGTCGCGAGTTAGA and R3-RX, ctcgagCACCTGAACTTTG-CATTCC for the 3′ homology region. The recombinants were identified after screening for kanamycin resistance, followed by PCR analysis. The following primers were used to assess the integrity of the recombination: R5upF2, CGTCTCGTCGCTGATTGGCTTC and CAG-R2, CCGTAAATAGTCCACCCATTGACG for the 5′ site; and G/RFPtail-F, CATGGACGAGCTGTACAAG and R3dwnR2, ATGCCATGAGTCAAGCCAG for the 3′ site. The loxP site and lox511 site in the pBACe3.6 backbone were removed, and the PI-SceI site was inserted using a BAC recombineering method with the PIsac-mLK cassette (from R. Kaneko, Gunma University) or the d4PI511zeo cassette (from R. Kaneko, Gunma University), respectively.
Detection of transgene recombination. Genomic DNA was extracted using a standard method 51 or simple alkali isolation. In the alkali extraction, the animal tissue was immersed in 50 mM NaOH and subsequently vortexed. After heating for 10 min at 95 °C, the solution was neutralized using 1 M Tris-HCl, followed by centrifugation. Only the supernatant was used for the subsequent PCR experiments. Genomic DNA was used as the template for PCR to detect the transgene using the following primer: 5′-CTATGACTGGGCACAACAGACAAT-3′ (in the Neo resistance gene-coding region). To evaluate the effectiveness of recombination, DNA was extracted from homogenized whole brains. Each allele with/without a STOP sequence was PCR-amplified using either set of primers: pCX1624F, CTAGAGCCTCTGCTAACC and PGK-R, GACGTGCTACTTCCATTTGTCAC for the STOP-remaining allele (PCR product: 497 bp); or pCX1624F, CTAGAGCCTCTGCTAACC and ChR + 136 R, CTCGGTGGAAGACGTAATCAGG for the STOP-deleted allele (PCR product: 314 bp). The PCR fragments were amplified at 95 °C for 3 min, followed by 35 cycles at 95 °C for 30 sec, 60 °C for 30 sec, and 72 °C for 1 min using KOD FX Neo (TOYOBO, Osaka, Japan). GeneRuler 100 bp DNA ladder (Fermentas, Burlington, ON, Canada) was used as marker for electrophoresis.
Primary culture of cortical neurons. Female LE rats were mated with male ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats, and cortical neurons were isolated from the embryos at day 16. Cortical tissue was dissociated using a 2.5% Trypsin and 0.5% DNase I mixture and grown in culture medium (Sumitomo Bakelite, Tokyo, Japan) under a 5% CO 2 atmosphere at 37 °C. The expression plasmid mCherry-NCre was transiently transfected in cortical neurons using calcium phosphate transfection at 6-7 days in vitro (DIV). Electrophysiological recordings were subsequently conducted at 21-25 DIV in neurons identified to express mCherry fluorescence under a conventional epifluorescence system.
To evaluate the recombination efficiency, 25 fields (218.9 μm 2 ) of interest showing mCherry positive nuclei from wells (Nunc, cat. no. 176740) in each group were selected. Double blind counting was applied to detect cells positive for both Venus enhanced by Alexa-488 and mCherry enhanced by Alexa-546.
Electrophysiology. All experiments were conducted at room temperature (23 ± 2 °C). Photocurrents were recorded as previously described 3 Table S1). To investigate the excitability of a neuron the current ramp injection of 0.4-2 nA/s was made twice with an interval of 6 s through the patch electrode to depolarize the membrane potential above the threshold of generating action potentials. Optics. Irradiation was performed using a SpectraX light engine (Lumencor Inc., Beaverton, Oregon, USA) controlled by computer software (pCLAMP10.3, Molecular Devices) at wavelengths (nm, >90% of the maximum, maximal irradiance): 438 ± 24 (max. 3.0 mWmm −2 , blue) and 643 ± 20 (max. 6.8 mWmm −2 , red). The power density (irradiance) of the light was directly measured under microscopy by a visible light-sensing thermopile (MIR-101Q, SSC Co., Ltd., Kuwana City, Japan). To excite the primary culture neurons, blue irradiation was attenuated using ND filters at 50%, 25% and 5%. Closing kinetics of ChRFR(C167A) was evaluated using red light at the maximal irradiance ( Supplementary Fig. S7). It followed a single exponential function with an OFF time constant (τ OFF ) of 1080 ± 56 ms with a negligible component of the steady state photocurrent (I max = −251 ± 46 nA, C = −17 ± 5 nA, n = 11). Therefore, the light-induced depolarization was almost completely attenuated by irradiating red light at the maximal irradiance for 8 s; the photocurrent was expected to be attenuated by 93.4% at the end of red light irradiation. BAC (referred to as Phox2b_tTA-2A-Cre) transgenic construct was generated after integrating tandem cassettes of genes encoding a tetracycline transactivator (pTet-Off-Advanced, Clontech, Mountain View, CA), 2 A peptide, and Cre recombinase with a nuclear localization signal into clone 95M11 derived from the CHORI RP-24 C57BL/6 J (B6) mouse genomic library using the Red/ET recombination system (Gene Bridges GmbH, Heidelberg, Germany) 54 . A previous study showed that the 95M11 BAC clone contained the necessary regulatory regions required for the proper expression of the Phox2b gene in mice 55 . Phox2b_tTA-2A-Cre transgenic rats were generated by the pronuclear injection of Wistar rat embryos (Charles River Laboratory Japan Inc., Japan). Transgenic founders carrying the transgenic construct were assessed using Southern blotting. No obvious gross phenotypic differences were apparent between the transgene-positive and transgene-negative littermates. Several transgenic founder rats were bred with Wistar rats. Subsequent analyses were performed using one line (line 3_19). Phox2b_tTA-2A-Cre rats were mated with ROSA26/CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats, and the newborn rats were used in subsequent experiments.

Generation of transgenic
All animal experiments were approved by the Tohoku University Committee for Animal Experiments (Approval No. 2014LsA-023) and the Animal Research Committee of Showa University (56010). Experiments were performed in accordance with the Guidelines for Animal Experiments and Related Activities of Tohoku University and the guiding principles of the Physiological Society of Japan and the National Institutes of Health (NIH), USA. The number of animals in the present study was kept to a minimum. To minimize suffering, the rats were deeply anaesthetized with isoflurane in a glass bottle until nociceptive reflexes induced by tail pinch were completely abolished. Animals were provided access to food and water ad libitum and maintained under a 12-hour light-dark cycle.
Ex vivo electrophysiology. Experiments were performed with brainstem-spinal cord preparations from newborn pups (0 to 3 days old) obtained after crossing Phox2b_tTA-2A-Cre RecBAC rats with ROSA26/ CAG-floxed STOP-ChRFR(C167A)-Venus BAC rats. The newborn rats were deeply anaesthetized with isoflurane, and the brainstem and spinal cord were isolated as previously described [56][57][58] . In most experiments, the preparations were cut transversely at a level just rostral to the anterior inferior cerebellar artery, corresponding to the level between the roots of the sixth cranial nerve and the lower border of the trapezoid body. The preparations were continuously superfused with artificial cerebrospinal fluid (ACSF 56 ) (composition [in mM]: 124 NaCl, 5 KCl, 1.2 KH 2 PO 4 , 2.4 CaCl 2 , 1.3 MgCl 2 , 26 NaHCO 3 , 30 glucose, equilibrated with 95% O 2 and 5% CO 2 , pH 7.4) at a rate of 2.5-3 ml/min in a 2-ml chamber and maintained at a temperature of 25-26 °C. The inspiratory activity corresponding to phrenic nerve activity was monitored from the fourth cervical ventral root (C4). The rostral ventrolateral medulla corresponding to the pFRG was photostimulated by blue LED (460-470 nm, 0.1-0.2 mWmm −2 ) via an optic fibre with a 0.25-mm outer diameter for up to 35 s (50 ms duration/150 ms interval). The membrane potential was −44.8 ± 3.9 mV, and input resistance was 683 ± 205 MΩ (n = 6). The recorded neurons located in the rostral pFRG were labelled with Lucifer Yellow and identified as Phox2b positive. All data analyses were performed using LabChart 7 Pro software (ADInstruments, Castle Hill, Australia).
Immunohistochemistry. Adult rat brains were resected three weeks after AAV2-Cre injection and promptly sectioned at 20 μm using a VT 1000S vibratome (Leica). Tissue slices were fixed with 4% paraformaldehyde, reacted with anti-Cre (1:500, MAB3120, Millipore) and anti-GFP (1:1000, 04404-84, Nacalai) antibodies, followed by secondary antibodies conjugated with Alexa Fluor 546 (1:200, Z25004, Molecular Probes, Eugene, OR, USA) and Alexa Fluor 488 (1:200, A11006, Molecular Probes, Eugene, OR, USA), respectively. Immunoreactivity was assessed under conventional fluorescence microscopy (Axiovert200, Carl Zeiss) or confocal microscopy (LSM510META, Carl Zeiss or FV1200, Olympus). The brainstem-spinal cords of rat neonates were also subjected to immunostaining after electrophysiological study. The preparations were fixed in 4% paraformaldehyde/0.1 M phosphate-buffered saline (PBS, pH 7.4) at 4 °C for 3 hrs. and immersed in 18% sucrose/ PBS, embedded in Tissue-Tek optimal cutting temperature (OCT) compound (Sakura Finetek, Torrance, CA), frozen on dry ice, and subsequently cut into 18-μm-thick sections, followed by immunofluorescence imaging. Anti-Phox2b (1:2000 dilution 33 ) and anti-Lucifer Yellow (Thermo Fisher Scientific/Invitrogen, Waltham, MA) were used for the primary antibodies, and Cy3-conjugated anti-guinea pig (Jackson ImmunoResearch, Baltimore Pike, West Grove, PA) and Alexa Fluor 488 anti-rabbit (Molecular Probes/Thermo Fisher Scientific, Eugene, OR) were used for the secondary antibodies. Images of the immunofluorescence samples were obtained using the 4× or 20× objectives of a BX51 fluorescence microscope (Olympus Optical). In each figure, the top is the dorsal side. The experiments were performed several times using different neonates, and the results were compatible between samples. Representative results are shown in the figures. Statistical analysis. All data in the text, figures and tables are expressed as mean ± SEM and were evaluated with the Mann-Whitney U-test for statistical significance unless otherwise noted. It was judged as statistically insignificant when P > 0.05.