Protocadherin-αC2 is required for diffuse projections of serotonergic axons

Serotonergic axons extend diffuse projections throughout various brain areas, and serotonergic system disruption causes neuropsychiatric diseases. Loss of the cytoplasmic region of protocadherin-α (Pcdh-α) family proteins, products of the diverse clustered Pcdh genes, causes unbalanced distributions (densification and sparsification) of serotonergic axons in various target regions. However, which Pcdh-α member(s) are responsible for the phenotype is unknown. Here we demonstrated that Pcdh-αC2 (αC2), a Pcdh-α isoform, was highly expressed in serotonergic neurons, and was required for normal diffusion in single-axon-level analyses of serotonergic axons. The loss of αC2 from serotonergic neurons, but not from their target brain regions, led to unbalanced distributions of serotonergic axons. Our results suggest that αC2 expressed in serotonergic neurons is required for serotonergic axon diffusion in various brain areas. The αC2 extracellular domain displays homophilic binding activity, suggesting that its homophilic interaction between serotonergic axons regulates axonal density via αC2′s cytoplasmic domain.

axon from being too dense or too sparse and for promoting the diffuse projections of serotonergic axons. Here we propose that serotonergic axons predominantly use αC2 among the Pcdh-α members for appropriate axonal projection.
We previously reported that in Pcdha del (11-C2)/del (11-C2) mice, stochastic and weak expression patterns of α10 are severely altered to αC2-like constitutive and strong expression patterns in several brain regions including the cerebral cortex, hippocampus, and cerebellar Purkinje cells, and that the total expression levels of all of the Pcdh-α isoforms is maintained 14 . In contrast, in the raphe nuclei of these mice, the expression level of total Pcdh-α transcripts, which was detected by the common Pcdh-α CR exon probe, was markedly lower, and the expression level of α10 was grossly normal (Fig. 1g). These results indicated that the regulation of Pcdh-α gene expression in the raphe nuclei is different from that in the cerebral cortex, hippocampus, and cerebellum. αC2 is essential for normal serotonergic projections. Given that the Pcdha del(11-C2)/del (11-C2) mice exhibited abnormal serotonergic projections and αC2 was strongly expressed in the raphe nuclei, we speculated that αC2 was essential for normal serotonergic projections. To test this possibility, we generated ∆C2 alleles (Fig. 2a). αC2 KO (Pcdha ∆C2/∆C2 ) mice showed abnormal serotonergic projections in many target brain areas, compared with control mice (Fig. 2b,c). In control mice, serotonergic axons extended diffuse projections throughout various brain areas, including the anterior olfactory nucleus, the olfactory bulb, the cortex, and the superior colliculus ( Fig. 2b,e,g,i,k). In contrast, in these regions, αC2 KO mice exhibited both the densification and sparsification of serotonergic axons (Fig. 2c,f,h,j,l). In the spinal cord, the serotonergic axon density in αC2 KO mice was grossly normal, compared with control mice (Fig. 2m,n). In the hippocampus, the density of serotonergic axons in αC2 KO mice was significantly higher in the SLM of CA1, and lower in the SO of CA1 and the DG, compared with WT mice (Fig. 3a-e). We quantified the transcripts of Sert and other Pcdh-α members in whole brains by quantitative reverse transcription (qRT)-PCR, and observed no significant differences in their expression levels between the WT and αC2 KO mice (Fig. 3f). In situ hybridization analysis revealed that the expression levels of α11, α12, αC1, and αCR in αC2 KO mice were similar to those in WT mice in the raphe nuclei (Fig. 3g). In the raphe nuclei of the αC2 KO mice, the expression of αC2 was completely lost, although the expression of αC1 remained (Fig. 3f). These results indicated that αC2 expression is essential for the normal diffusion of serotonergic projections in the target brain regions. αC2 expression in serotonergic neurons is essential for normal serotonergic projections. Pcdh-α genes are expressed not only in serotonergic neurons, but also in other types of neurons in almost all brain regions, including the hippocampus 8 . Therefore, it was not clear whether the Pcdh-α expressed by serotonergic neurons or by other neurons in their target brain regions controls the serotonergic innervation. To address this question, we generated brain region-specific Pcdh-α deficient mice using the Cre-loxP system. In the Pcdha flox allele, loxP sequences were inserted upstream and downstream of the Pcdh-α gene cluster (Fig. 4a).
In dorsal telencephalon-specific Pcdh-α KO mice (Pcdha flox/flox ; Emx1 +/Cre ), Pcdh-α gene expression was lost in almost all of the cells in the hippocampus and cerebral cortex (Fig. 4b,c). In the dorsal telencephalon of Emx1 +/Cre mice, Cre recombinase is expressed in excitatory neurons but not in inhibitory neurons 15 . Thus, almost all of the cells expressing Pcdh-α in the hippocampus of the dorsal telencephalon-specific Pcdh-α KO mice were thought to be inhibitory neurons. The serotonergic projections in these mice were grossly normal in the hippocampus, compared with control mice (Pcdha flox/flox ) ( Fig. 4d-f). This result indicated that the expression of Pcdh-α genes in hippocampal excitatory neurons is dispensable for normal serotonergic projection.
biallelic αC2 exons but also hemi-allelic exons from α1 to αC1 are also deleted in the serotonergic neurons. We previously reported that Pcdha +/∆CR mice, which at least lack the expression of α1 to αC1 from the hemi-allele, show normal serotonergic projection in the hippocampus 8 . These results indicated that the loss of biallelic αC2 expression in serotonergic neurons leads to aberrant serotonergic projection in the hippocampus. causes serotonergic axon densification in the SLM of the hippocampus. To elucidate how the axonal densification occurs morphologically, we sparsely labeled serotonergic axons by injecting an adeno-associated virus vector (AAV-EF1α-DIO-tRFP-WPRE), in which Cre expression induces red fluorescent protein (RFP) expression, into the raphe nuclei of Sert-Cre mice 16 . Cre(+) cells in the raphe nuclei specifically expressed RFP (Fig. 6a). A portion of serotonergic axons was labeled by RFP in both control (Sert-Cre) and αC2 KO (Pcdha ∆C2/∆C2 ; Sert-Cre) mice ( Fig. 6b-d). We did not find any axons that were in continuous contact with each other, forming thick bundles with multiple axons, or clumps with other single axons in either the control (n = 5 sections, 2 mice) or αC2 KO mice (n = 5 sections, 4 mice). To analyze these axonal routes, we traced the RFP(+) axons passing     (Fig. 6c-f). The ratio of RFP(+) axons crossing L − 20 to those in L + 20 in the αC2 KO mice was higher than that in control mice [control, 82 axons at L − 20, and 71 axons at L + 20 (3 sections from 2 mice); KO, 121 axons at L − 20, and 46 axons at L + 20, Fisher's exact test, p = 0.0005, Fig. 6g]. This finding was consistent with the result of the SERT-immunopositive intensity experiment (Fig. 3a-e). Next, we analyzed the RFP(+) axons crossing both L − 20 and L + 20. The probability that RFP(+) axons passing through L − 20 also passed through L + 20 in the αC2 KO mice was significantly lower than that in control mice [control, 26 of 82 axons (3 sections from 2 mice); KO, 10 of 121 axons (3 sections from 4 mice), Fisher's exact test, p < 0.0001, Fig. 6g]. The probability that RFP(+) axons passing through L + 20 also passed through L − 20 in the αC2 KO mice was also significantly lower than that in control mice [control, 26 of 71 axons (3 sections from 2 mice); KO, 10 of 46 axons (3 sections from 4 mice), Fisher's exact test, p = 0.038, Fig. 6g]. These results indicated that the serotonergic axons in αC2 KO mice were more suppressed from crossing the SLM/SR boundary than those in control mice. These results suggested that αC2 in the serotonergic axons is required for their axonal diffusion.
αC2 protein is required for the outward diffusion of serotonergic axons in the olfactory bulb. Layers in the olfactory bulb are separately innervated by median and dorsal raphe nuclei: serotonergic neurons in the median raphe nucleus innervate the glomerular layer (GL), whereas those in the dorsal raphe nucleus innervate the granule cell layer (GCL) to the external plexiform layer (EPL) 17 . Here we focused on the median raphe-innervating layers. SERT(+) axons in αC2 KO mice were densified in the GCL, and sparsified in the EPL, compared to controls (Fig. 7a-c). To analyze these axonal routes, we traced the SERT(+) axons passing through the 20-µm inner (L − 20) line or 20-µm outer (L + 20) line from the boundary between the GCL and the inner plexiform layer (IPL) (Fig. 7d,e). The probability that SERT(+) axons passing through L − 20 also passed through L + 20 in the αC2 KO mice was significantly lower than in WT mice [WT, 20 of 95 axons (3 sections from 3 mice); KO, 9 of 115 axons (3 sections from 3 mice), Fisher's exact test, p = 0.0083, Fig. 7f]. In contrast, there was no significant difference in the probability that SERT(+) axons passing through L + 20 also passed through L − 20 between these genotypes (WT, 20 of 94 axons; KO, 9 of 48 axons, Fisher's exact test, p = 0.83, Fig. 7f). Although we cannot determine the axonal direction (outward or inward) from these results alone, these observations raised two possibilities: either the diffusion of outward axons from the GCL to IPL is suppressed, or the arborization of inward axons from the IPL to GCL is facilitated, in the GCL. Therefore, we analyzed SERT(+) axons in the GCL. In this region, there was no significant difference in the occurrence frequency of end points and branch points of SERT(+) axons between WT and αC2 KO mice (Fig. 7g). These results suggested that the outward diffusion of serotonergic axons from the GCL is suppressed.
The somata of granule cells in the GCL were densely clustered in both genotypes (Fig. 7a,c). Serotonergic axons ran through this cell-cluster niche in a manner avoiding the high-density cell clusters. In the GCL soma niche of αC2 KO mice, we frequently found serotonergic axons that ran parallel and partially in contact with each other (Fig. 7h,i). We did not detect this phenotype in WT mice. These results suggest that the loss of αC2 protein reduces the axon-axon repulsion of serotonergic axons in the olfactory bulb. As a result, the reduced axonal repulsion may dampen the outward diffusion of serotonergic axons. Around the GCL/IPL boundary, some SERT(+) axons returned to the GCL without contacting other SERT(+) axons (arrows in Fig. 7b,e). This observation suggests that loss of αC2 protein enhances the attraction of serotonergic axons to the GCL or their repulsion from the IPL. The αC2 protein may mediate not only the trans-homophilic interaction-driven repulsion of serotonergic axons, but also the trans-heterophilic interaction-driven axon guidance in target areas.

Discussion
Here we found that, among the Pcdh-α genes, αC2 was dominantly expressed in the serotonergic neurons of the raphe nuclei, and that the loss of only αC2, and not of α2 to α11, caused unbalanced distributions (densification and sparsification) of serotonergic axon density in various brain regions, including the hippocampus. The abnormalities of the serotonergic projections in Pcdha del(11-C2)/del (11-C2) and Pcdha ∆C2/∆C2 mice were very similar to those in Pcdha ∆CR/∆CR mice 8 . Thus, the diversity of the Pcdh-α genes is not required for the normal distribution of serotonergic axons. We also analyzed dorsal telencephalon-specific Pcdh-α KO mice and serotonergic neuron-specific αC2 KO mice. The former showed a normal distribution of serotonergic projections, whereas the latter showed unbalanced distributions similar to those of the αC2 KO mice. Analyses of serotonergic projections at the single-axon level revealed that the serotonergic-axon diffusion in αC2 KO mice is suppressed. These results indicated that the αC2 expression in serotonergic neurons is essential for the normal diffuse pattern of serotonergic axons in various target regions.
Recently, Chen et al. also analyzed Pcdh-α function in serotonergic system 18 . (1) They analyzed the expression of Pcdh-α, -β, and -γ by single-cell RNA-seq in serotonergic neurons, and showed that αC2 is intensely expressed in serotonergic neurons. (2) They showed that loss of αC1 and αC2 causes clumping of serotonergic axons in the hippocampus. (3) They showed that loss of Pcdhα in serotonergic neurons causes clumping of serotonergic axons in the hippocampus although loss of Pcdhα in the hippocampal neurons do not. (4) They individually labeled serotonergic axons in serotonergic neuron-specific Pcdh-α KO mice, and found clumped axons in the hippocampus and substantia innominata. From these results, they concluded that αC2 is required for axonal tiling and assembly of serotonergic axons.
Similar abnormal axonal clumps appear in Ia afferent axons in the spinal cord of Pcdh-γ KO mice 19 , and in triple KO mutants in γC3 to γC5 20 . Interestingly, γC3 to γC5 are constitutively expressed genes similar to αC2. However, analyses of conditional Pcdh-γ KO mice revealed that Pcdh-γ expressed in both Ia afferents and their targets (interneurons in the spinal cord) is required to prevent the clumping of projections during the formation of Ia afferent terminals 19 . In the serotonergic projection, αC2 was required not in the target neurons, but in the serotonergic neurons themselves. Therefore, the mechanism underlying the inhibition of axonal densification in serotonergic axons by αC2 is distinct from that in Ia afferents by Pcdh-γC3-C5.
We previously reported that the cytoplasmic tail of Pcdh-α is essential for normal serotonergic projection 8 . The cytoplasmic domain of αC2 is known to bind cytoplasmic signaling proteins of the focal adhesion kinase (FAK) family, FAK and PYK2 21 . The FAK family contributes to signaling cascades that regulate the growth cone motility, and is activated by several axon-guidance cues, including brain-derived neurotrophic factor (BDNF) 22,23 . Interestingly, BDNF can induce serotonergic axonal growth in the brain 24,25 . Stathmin family proteins, microtubule-destabilizing proteins, also bind to the cytoplasmic domain of Pcdh-α 26 . These proteins are involved in axonal elongation and branching 27 , and their phosphorylation is induced by BDNF 28,29 . Therefore, BDNF and stathmin family proteins are candidate molecules involved in the αC2 signaling cascade that controls serotonergic projections.
Glia cell line-derived neurotrophic factor (GDNF) enhances the neurite elongation of serotonergic neurons in vitro 30 . GDNF family receptor, which is expressed in serotonergic neurons 31 , forms a receptor complex with Ret tyrosine kinase, an essential molecule for GDNF signaling. Ret directly binds to α4, and enhances the phosphorylation of α4′s cytoplasmic domain, in a GDNF dosage-dependent manner 26 . This binding stabilizes both Ret and α4. In serotonergic neurons, the α4 expression level is considerably lower than that of αC2. Nevertheless, GDNF and Ret are also candidates for regulating αC2's function in the serotonergic projections.
A similar situation is seen for the Down syndrome cell adhesion molecule (DSCAM), which belongs to the immunoglobulin superfamily. Two closely related proteins (Dscam and Dscaml1) are expressed in vertebrates, and 19,008 proteins are generated by distinct combinations of alternative splicings of their homolog in Drosophila 32 . Each DSCAM homophilically mediates negative cell-cell interactions 33 , and these homophilic interactions are responsible for dendrite and axonal self-avoidance. In mice, Dscam is expressed in most retinal ganglion cells and prevents the fasciculation of their dendrites by self-avoidance. Similarly, Dscaml1 is expressed in rod bipolar cell dendrites and prevents the fasciculation and clumping of their dendrites and cell bodies by self-avoidance 34 . Thus, homophilic negative interactions between molecules of a single isoform among many diverse Dscams are sufficient to induce repulsion and prevent fasciculation in homotypic dendrites and axons. The diversity of the Pcdh-α genes is not required for the normal diffuse innervation of serotonergic axons. Therefore, our study could not reveal the role of the diversity of clustered Pcdh-α genes, but can support the possibility that the diversity of Pcdh-α isoforms mediates the axon and/or dendrite repulsion of cortical and hippocampal neurons, and of Purkinje cells of the cerebellum, which show diverse expression patterns 14,35 , like the Drosophila DSCAMs.
Azmitia and Segal analyzed serotonergic axonal routes in detail 36 . From the observation, they speculated that serotonergic projection is regulated by epiphytic guidance, that is, they rely on another group of fibers for their structural support. We found many serotonergic axons crossing GCL/IPL boundary in the olfactory bulb. Because granule cells in GCL project their dendrites to EPL, their may guide serotonergic axons. We found serotonergic axons returning to GCL from GCL/IPL boundary without entering IPL in αC2 KO mice (arrows in Fig. 7b,e). These results raise a possibility that αC2 in serotonergic axons is required to attach the other fibers and to find their appropriate routes and final target areas.
In this study, we reported that αC2 in serotonergic neurons enables serotonergic axons to extend diffuse projections in various brain areas. Understanding the mechanism by which αC2 promotes diffuse projections in the appropriate brain target regions should shed light on the cellular and molecular mechanisms important for homotypic axonal repulsion and serotonin-related developmental disorders.

Methods
Animal experiments. All of the experimental procedures were in accordance with the Guide for the Care and Use of Laboratory Animals of the Science Council of Japan and were approved by the Animal Experiment Committee of Osaka University or the National Institute of Genetics. Male or female mice in adulthood (>3 months old) were used in the axon-tracing analyses. In the other experiments, male or female mice at postnatal day 21 were used.
Analysis of serotonergic fiber density. The distribution of serotonergic axons using an anti-serotonin transporter (SERT) antibody (HTT-N77, a generous gift from Dr. Masahiko Watanabe, Hokkaido University, Japan), and the SERT-immunopositive fiber density in the hippocampus were analyzed as described previously with small modifications 8 . All photographs were taken with a DP-72 CCD camera (Olympus) mounted on a BX51 microscope (Olympus). The resolution of the photographs used to analyze serotonergic axon density was 1101 pixel/mm (Fig. 1e,f), or 1643 pixel/mm (Figs 3e, 4f and 5h). Prism software (GraphPad Software, Inc.) was used for the statistical analyses. Statistical comparisons among three mouse lines were done with one-factor ANOVA and Bonferroni's post hoc test. Statistical comparisons between two mouse lines were done with the Mann-Whitney U test. The criterion for significance was p < 0.05.

Analyses of serotonergic axons at the single-axon level.
To label serotonergic neurons that project to the hippocampus, we injected the adeno-associated virus vector pK207_AAV-EF1α-DIO-tRFP-WPRE into the median raphe nucleus (angle: 25°, AP: −4.7 mm, ML: 0.2~0.6 mm, DC 5.0 mm from bregma) of mice that were anesthetized by a Univentor 410 Anesthetisia Unit, using a stereotaxic apparatus (Narishige). 5HTT-Cre (Sert-Cre) mice 16 and Sert-Cre;Pcdha ∆C2/∆C2 mice were used as control and experimental groups, respectively. pK207 was constructed by inserting tRFP (pTurboRFP-N, evrogen) into the AscI/NheI site of pK168_AAV-EF1α-DIO-tTA-tRFP-WPRE (Adgene 85039) 43 . Virus was prepared as described previously 43 . More than 4 weeks after the virus injection, mice were transcardially perfused with 4% paraformaldehyde/0.1 M phosphate buffer (PB). The brains were removed and immersed in 30% sucrose/0.1 M PB, and 100-μm-thick sagittal sections in the hippocampus and 100-μm-thick coronal sections in the raphe nuclei were prepared with a microtome (Yamato Kohki). The sections were incubated in rabbit anti-SERT (1:1000, Frontier Institute) [and mouse anti-Cre antibodies (1:1000, Millipore) in some cases] in phosphate buffed saline (PBS) containing 1.5% normal goat serum and 0.25% Triton X-100 for >2 days. After washing with PBS, the sections were incubated in Alexa Fluor 488 anti-rabbit IgG (1:1000) [and Alexa Fluor 647 anti-mouse IgG antibody in some cases] and 4′,6-diamidino-2-phenylindole (DAPI, 1 μg/mg) for >2 days, and washed with PBS. From the sections, z-stack images were captured with a confocal microscope (Leica). In the images, the SLM/SR boundary (L0) was determined by the density difference in DAPI(+) nuclei (high in SLM, low in SR), and lines 20 μm away from the SLM/SR boundary were drawn on both the inner (L − 20) and outer (L + 20) sides. Z-stack images were smoothed by the 3D Gaussian lowpass filter function of ImageJ. All RFP(+) axons crossing both lines in one set of z-stack images (depth: 100 μm) were traced with the ImageJ plugin software, Simple Neurite Tracer 44 . For the statistical analysis, Fisher's exact test was used.
Sagittal sections of the olfactory bulb were also stained with the anti-SERT antibody and DAPI with the same method. From these sections, z-stack images were captured with the confocal microscope. In the images, the GCL/IPL boundary was determined by the density difference of DAPI(+) nuclei (high in GCL, low in IPL), and lines 20 μm apart from the GCL/IPL boundary were drawn on both the inner (L − 20) and outer (L + 20) sides. All SERT(+) axons crossing both lines in one set of z-stack images (depth: 50 μm) were traced with Simple Neurite Tracer. The length of SERT(+) axons in boxes (50 × 50 × 50 μm) in the GCL and EPL was traced and measured with Simple Neurite Tracer. For the statistical analysis, Welch's t-test was used. In the analyses of branch and end points in boxes (50 × 50 × 50 μm) in the GCL, an axon connecting with branch points was regarded as one axon, and its length was defined as its sum length. In the survival analyses of branch (or end) points, axons with branch (or end) points were dealt with as an event (death), and the other axons were regarded as being censored. For the statistical analyses, the log-rank test was used.