Expansion of stochastic expression repertoire by tandem duplication in mouse Protocadherin-α cluster

Tandem duplications are concentrated within the Pcdh cluster throughout vertebrate evolution and as copy number variations (CNVs) in human populations, but the effects of tandem duplication in the Pcdh cluster remain elusive. To investigate the effects of tandem duplication in the Pcdh cluster, here we generated and analyzed a new line of the Pcdh cluster mutant mice. In the mutant allele, a 218-kb region containing the Pcdh-α2 to Pcdh-αc2 variable exons with their promoters was duplicated and the individual duplicated Pcdh isoforms can be disctinguished. The individual duplicated Pcdh-α isoforms showed diverse expression level with stochastic expression manner, even though those have an identical promoter sequence. Interestingly, the 5′-located duplicated Pcdh-αc2, which is constitutively expressed in the wild-type brain, shifted to stochastic expression accompanied by increased DNA methylation. These results demonstrate that tandem duplication in the Pcdh cluster expands the stochastic expression repertoire irrespective of sequence divergence.

The diversity and sophisticated gene regulation exhibited by the Pcdh cluster genes are important for normal development of the nervous system 11,12,21,22 . The Pcdh cluster genes, which encode a group of diverse cadherin-related transmembrane proteins, are expressed mainly in the nervous system, and gene regulation mechanisms in the Pcdh clusters include both constitutive and stochastic expression in single neurons [23][24][25][26][27] . Gene ablation studies showed that Pcdh-a and Pcdh-c are required for neuronal survival, synapse formation, axonal targeting, dendritic arborization, and selfavoidance of dendrites 11,12,[28][29][30] . These findings led to the suggestion that the Pcdh cluster genes are likely candidates for the individualization of neurons in the vertebrate brain, which would be generated through the stochastic expression of these genes 11,12,21,22 . Although these findings suggest that stochastic expression in the Pcdh cluster is important in neurodevelopment, the evolutionary origin of the stochastic expression in the Pcdh cluster has remained a mystery.
To better understand the evolutionary origin of the stochastic expression and CNVs' effect in the Pcdh cluster, here we focused on the effect of tandem duplication in the mouse Pcdh-a cluster. In the present study, we engineered a targeted tandem duplication within the mouse Pcdh-a cluster, a situation somewhat comparable to that occurring during vertebrate evolution and in current human populations. The individual Pcdh-a isoforms transcribed from each duplicate exon can be distinguished in the mutant mice, enabling us to determine the manner by which their expression was regulated. The individual duplicated Pcdh-a isoforms showed diverse expression level with stochastic expression manner, even though those have an identical promoter sequence. Surprisingly, the duplicated Pcdh-ac2 isoform, which shows constitutive expression in the wild-type allele, shifted to stochastic expression accompanied by increased DNA methylation. Our results demonstrate that tandem duplication in the Pcdh cluster expands the stochastic expression repertoire irrespective of sequence divergence.

Results
Targeted tandem duplication in the mouse Pcdh-a cluster. To study the consequences of tandem duplication in the Pcdh gene cluster, we generated a targeted tandem duplication in the mouse Pcdh-a cluster using the inter-strain targeted meiotic recombination (iTAMERE) system 31 . The wild-type mouse Pcdh-a cluster contains 14 large 'variable' exons, each of which encodes a cadherin-like type I membrane protein consisting of extracellular domains, a transmembrane domain, and a proximal cytoplasmic domain. Each variable exon is expressed from its own promoter and spliced to three short 'constant' exons, which encode a shared, 152-amino acid C-terminal domain (Fig. 1a) 16 . In the mutant allele [hereafter called dup(2-c2) or simply dup], a 218-kb region containing the Pcdh-a2 to Pcdh-ac2 variable exons with their promoters was duplicated; the 59-located duplicate was derived from the C57BL/6J (B6) strain and the 39located one was derived from the CBA strain. We selected this particular region for duplication, because it contains both stochastically and constitutively expressed Pcdh-a isoforms, and because several of the isoforms involve a single nucleotide polymorphism (SNP) between the B6 and CBA strains in the coding region (Fig. 1b, c,  d). The isoforms with a SNP between the B6 and CBA strains are Pcdh-a3, Pcdh-a5, Pcdh-a6, Pcdh-a7, Pcdh-a9, Pcdh-a10, Pcdh-a12, and Pcdh-ac2, and among these isoforms, only Pcdh-a12 has a polymorphism in its promoter region. Although the individual duplicated genes in the previous Pcdh-a duplication lines cannot be distinguished 25 , the individual duplicated genes derived from these eight Pcdh-as in the dup(2-c2) allele can be distinguished from each other. Therefore, the dup(2-c2) mice for the first time enables the expressional analysis of the individual duplicated isoforms in the mouse Pcdh-a cluster.
Before using the iTAMERE system, we individually inserted two loxP sites into the variable region of the Pcdh-a cluster with the same orientation. First, the ''G16Neo'' allele 25 was generated by inserting a loxP site between the Pcdh-a1 and Pcdh-a2 exons of the CBA allele in the TT2 ES embryonic stem (ES) cell line, which is on a CBAxB6 F1 genetic background (Fig. 1b). Second, a loxP site was inserted between the Pcdh-ac2 exon and the first exon of the constant region (CR1) in the Pcdh-a cluster to generate the ''SR'' allele ( Fig. 1c and Supplementary Fig. S1). To insert this loxP site into the B6 allele, we used the RENKA ES cell line, which is on a pure B6 genetic background 32 .
To duplicate the sequence between the loxP site of the G16Neo allele and that of the SR allele using the Cre-loxP system, we obtained male mice that possessed the G16Neo and SR alleles (G16Neo/SR) and the Sycp1-Cre transgene, which elicits Cre recombinase expression specifically in the testis 25 . The male mice were crossed with B6 female mice, and the genotypes of the F1 pups were analyzed by PCR. The minority of F1 pups carried the dup(2-c2) allele, in which exons Pcdh-a2 to Pcdh-ac2 were duplicated (Fig. 1d), or the del(2-c2) allele, in which exons Pcdh-a2 to Pcdh-ac2 were deleted (data not shown). F1 pups carrying these duplication or deletion alleles were obtained at 5.8% (4 of 69 pups) and 1.4% (1 of 69 pups), respectively. We then analyzed the tail DNA of the duplication-containing mice by PCR to detect the Cre-mediated duplication alleles, and sequenced the PCR products to confirm the presence of the predicted junction sequences generated by the Cre-mediated site-specific recombination events. Animals homozygous for the duplicated allele (Pcdha dup(2-c2)/dup(2-c2) ) were obtained by crossing heterozygous (Pcdha wt/dup(2-c2) ) parents.
The Pcdha dup(2-c2)/dup(2-c2) pups were born with the expected Mendelian distribution (Supplementary Table S1), developed normally to adulthood, and were fertile. Histochemical analysis with Nissl staining revealed an apparently normal gross anatomy of the Pcdha dup(2-c2)/dup(2-c2) mouse brain (Fig. 1g left and Supplementary  Fig. S2a). This finding was further supported by cytochrome oxidase staining showing a normal barrel structure in the Pcdha dup(2-c2)/dup(2-c2) mice ( Supplementary Fig. S2a). Furthermore, neural pathway and serotonergic axon analyses by anti-neurofilament and anti-SERT staining, respectively, showed no obvious differences between the genotypes (Fig. 1g middle and Supplementary Fig. S2b). Finally, the distribution of c-fos mRNA, a well-known marker for neuronal activity, was also similar between the genotypes ( Supplementary Fig.  S2c). These findings suggested that the tandem duplication of exons Pcdh-a2 to Pcdh-ac2 does not result in any deleterious effects on mouse development or brain morphogenesis. Tandem duplication maintains the expression level of neighboring genes. Previous studies have indicated that tandem duplication may alter not only the expression of genes within the duplication boundaries but also of genes located in their genomic neighborhoods 33 . Prompted by these observations, we first quantified the expression levels of transcripts in the vicinity of the Pcdh-a cluster in the cerebellum of 4-week-old Pcdha dup(2-c2)/dup (2-c2) mice. The analyzed non-Pcdh genes included Wdr55, Dnd1, Hars, Zmat2, and Vault, located about 130-kb , 170-kb upstream from the duplication's 59 boundary and Slc25a2, Taf7, Diap1, and Hdac3, located about 480-kb , 780-kb downstream from the duplication's 39 boundary (Fig. 2a). We found that the duplication did not alter the expression of the Pcdh-b, Pcdh-c, or non-Pcdh genes, except for a small effect on Pcdh-b22 (Fig. 2b). These results indicated that the tandem duplication of exons Pcdh-a2 to Pcdh-ac2 did not exert long-range effects on the regulation of neighboring genes.
Next, the expression levels of these duplicated Pcdh-as were quantified by qRT-PCR and cloning-mediated SNP analysis, which is highly sensitive and yields quantitative data (Fig. 2d). The expression level of the spliced CR transcripts, which are common to all the 59located and 39-located duplicated Pcdh-as, was unchanged in the Pcdha dup(2-c2)/dup(2-c2) mice. There were no significant differences in the expression levels of most of the 39-located duplicated Pcdh-as (Pcdh-a3, Pcdh-a6, Pcdh-a7, Pcdh-a10, Pcdh-a12, and Pcdh-ac2) compared to wild-type. In contrast, the expression levels of all the 59located duplicated Pcdh-as and the 39-located duplicated Pcdh-a5 and Pcdh-a9 genes were significantly reduced compared to wildtype. These observations revealed that the expression level of the total Pcdh-a genes was maintained, while that of individual isoforms was altered, indicating that expressional re-allocation occurred in the dup(2-c2) allele.
Interestingly, the expression levels of the 59-located duplicated Pcdh-as were significantly lower than those of their duplicated 39located counterparts. This observation suggested that, despite their identical promoter sequences, each 59-located and 39-located duplicated Pcdh-a receives distinct gene-regulation influences. Taken together, these findings indicate that tandem duplication alters the manner of gene regulation in the Pcdh-a cluster.
To analyze the expression of Pcdh-a3, Pcdh-a5, and Pcdh-a7 in the dup(2-c2) allele, single Purkinje cells from 4-week-old Pcdha JF1/dup(2-c2) mice were picked up by glass capillary. Complementary DNA of Pcdh-a3, Pcdh-a5, Pcdh-a7, and Pcp-2 (a marker for Purkinje cells) was synthesized from the single-cell samples in the same tube, and the resulting cDNA was then divided into three tubes and subjected to separate, first-round multiplex PCR analysis. The second round of PCR amplification was carried out individually for each tube and used nested primers for the Pcdh-a3, Pcdh-a5, Pcdh-a7 genes, and for Pcp-2. Finally, each PCR product was subjected to direct sequencing to determine from which exon the transcript was derived: i.e., the wild-type (JF1) allele or the 59-located (dup-59) or 39-located (dup-39) exons in the dup(2-c2) allele.
Of the 163 single Purkinje cells analyzed, 45 yielded PCR amplicons of Pcdh-a3, Pcdh-a5, or Pcdh-a7 from the same exons in all three tubes, and all 45 cells were positive for Pcp-2, confirming that they were differentiated Purkinje cells. In addition to three of three specific transcripts from the same exon, some cells showed one or two of three transcripts (for example, Pcdh-a3 in cell #1-37); these findings suggested that the amounts of corresponding transcripts in these cells were low, and therefore we excluded these cells from the following analysis.
The 59-located Pcdh-ac2 acquired stochastic expression upon tandem duplication. To investigate whether the 59-located and 39located duplicated Pcdh-ac2 retained its constitutive expression at the single-neuron level, single-cell RT-PCR and SNP analysis was carried out (Fig. 4). Of the 16 single Purkinje cells analyzed, all showed the Pcdh-ac2 transcript from the wild-type (JF1) and 39located duplicated exon in all three tubes. They were also all positive for Pcp-2, confirming that they were differentiated Purkinje cells. These results suggested that the Pcdh-ac2 wild-type exon (JF1) and 39-located duplicated exon were expressed constitutively in differentiated Purkinje cells.
Because bisulfite sequencing was unable to discriminate between the two Pcdh-ac2s in the dup(2-c2) allele, we further analyzed the DNA methylation using HpaII digestion-mediated DNA methylation analysis (Fig. 5c and 5d). The HpaII-resistant fraction, containing methylated CCGG, predominantly included the 59-located duplicated Pcdh-ac2 genomic DNA. The DNA methylation level on the 59-located duplicated Pcdh-ac2 was higher than that on the 39-located and wild-type Pcdh-ac2.
To identify the region with increased DNA methylation more precisely, DNA methylation around the 59-located Pcdh-a12 promoter was examined (Fig. 6c). No significant differences in the DNA methylation level were observed for the 4-kb upstream region (38.5% , 55.0%, for the wild-type, the 59-located, and 39-located exon) or for the 39-region exon (around 65%, for wild-type and the mixture of the 59-located and 39-located exon), suggesting that the increase in DNA methylation was specific for the promoter. Therefore, as in Pcdh-ac1 and Pcdh-ac2, DNA hypomethylation in the Pcdh-a12 promoter is dependent on the cluster structure, and tandem duplication directly increased the DNA methylation of the 59-located Pcdh-a12.
To gain further insight into the cluster-dependent regulation of the DNA methylation, we next examined the methylation during development (Fig. 6d). Interestingly, while the DNA methylation level on Pcdh-a1 and Pcdh-a6 showed similar for all the loci, the DNA methylation level on the 59-located Pcdh-a12 promoter was higher than that on the wild-type and 39-located Pcdh-a12 promoter. The DNA methylation level on the 59-located Pcdh-a12 resembled that of Pcdh-a1 and Pcdh-a6 throughout development. The increased DNA methylation on the 59-located Pcdh-a12 promoter was also observed in the tail but not in the liver (Fig. 6e), suggesting that an organ-dependent DNA methylation regulator influenced the DNA methylation on the 59-located Pcdh-a12 promoter.
Taken together, these results suggest that 39-located genes in the wild-type Pcdh-a cluster, Pcdh-a12, Pcdh-ac1, and Pcdh-ac2, are hypomethylated in a cluster-structure dependent manner, and that these DNA hypomethylations were masked by position shift to 59location, which is caused by tandem duplication in the dup(2-c2) allele. This mechanism probably underlies the lower expression level of the the 59-located Pcdh-a12 duplicate and the stochastic expression of the 59-located Pcdh-ac2 duplicate in the dup(2-c2) allele.

Discussion
Tandem duplications are concentrated within the Pcdh cluster throughout vertebrate evolution and as CNVs in human populations 3,19,20 , but the effects of tandem duplication in the Pcdh cluster remain elusive. Here we revealed a critical role for tandem duplica-  tion in the Pcdh cluster gene regulation. The individual duplicated Pcdh-a isoforms showed diverse expression level with stochastic expression manner, even though those have an identical promoter sequence. Interestingly, the 59-located duplicated Pcdh-ac2, which is constitutively expressed in the wild-type brain, shifted to stochastic expression upon tandem duplication, accompanied by increased DNA methylation. These observations suggest that tandem duplication has been beneficial for the acquisition of the stochastic expression and the expansion of its repertoire through vertebrate evolution and in human populations (Fig. 7).
What are the mechanisms by which individual duplicated Pcdhas, each of which has an identical promoter sequence, are expressed stochastically? The present data are consistent with our previous findings, which is that the stochastic expression of Pcdh-a cluster genes is governed by enhancer and promoter DNA methylation 25,38 . The enhancers for the Pcdh-a genes, named HS5-1 and HS7, are located downstream of the Pcdh-a cluster [39][40][41] , and the present data suggest that the HS5-1 and/or HS7 enhancers in the duplicated allele are still effective at even greater distances than in the wild-type allele (distance between the HS5-1 enhancer and the most distal promoter (Pcdh-a1) in the wild-type allele was ,280 kb; in the dup(2-c2) allele, it was ,500 kb). However, the present finding of lower expression of the 59-located duplicated Pcdh-a genes than the 39located duplicated Pcdh-a genes in the dup(2-c2) allele indicates that Each circle represents a methylated (black) or unmethylated (white) CpG dinucleotide. Each row represents a single clone. A primer set was designed to amplify the region corresponding to the promoter (Pcdh-a1 and Pcdh-a12) or 59 region of the exon (Pcdh-a6) containing the SNP. In addition, the 4-kb upstream region and 39 region of Pcdh-a12 were analyzed. The percentage below each methylation pattern indicates the CpG methylation rate for each region. (d) Changes in DNA methylation levels during development. DNA methylation levels of the Pcdh-a1 promoter (top), 59 region of the Pcdh-a6 exon (middle), and Pcdh-a12 promoter (bottom) were analyzed by bisulfite sequencing of sperm and of mice at E3.5, E7.5, E9.5, E12.5, 4-weeks old, and 3months old. a P 5 0.0071 compared with wt, P 5 0.0002 compared with the 39-located Pcdh-a12, b P , 0.0001 compared with wt, P , 0.0001 compared with the 39-located Pcdh-a12, c P 5 0.0005 compared with wt, P 5 0.0264 compared with the 39-located Pcdh-a12, d P 5 0.0167 compared with wt, P 5 0.0086 compared with the 39-located Pcdh-a12. (e) DNA methylation levels in the liver and tail of 4-week-old mice. DNA methylation levels at the Pcdh-a1 promoter (top), 59 region of the Pcdh-a6 exon (middle), and Pcdh-a12 promoter (bottom) were analyzed by bisulfite sequencing. e P 5 0.0003 compared liver with tail of the 59-located Pcdh-a12.
www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 6263 | DOI: 10.1038/srep06263 the longer distance between the HS5-1 enhancer and the promoter lowers the probability of expression. The present data strongly support the ''enhancer sharing and stochastic promoter competition'' model, in which a single enhancer stochastically governs the expression of the Pcdh cluster genes 25 .
Stochastic promoter competition has also been suggested for other gene clusters, such as the olfactory receptor MOR28 cluster 10,42 and the primate red and green-pigment genes 43 . Furthermore, nonstochastic promoter competition has been suggested for other gene clusters, such as the Hoxd gene cluster 44 , aand b-globin gene cluster 8,9 , and zebrafish red opsin genes 45 ; these gene clusters show temporally and spatially organized expression. Thus, promoter competition is widely distributed through gene clusters. We argue that the characteristics of the enhancer and/or promoter add further sophistication to gene regulatory systems.
Here we found that the 59-located duplicated Pcdh-ac2, which is constitutively expressed in the wild-type brain, acquired stochastic expression upon tandem duplication, accompanied by increased DNA methylation. These results provide supportive evidence for previous findings that mosaic DNA methylation states are correlated with the stochastic expression of Pcdh-a isoforms in wild-type mouse brain 36 . Previous studies described the suppression of promoter DNA methylation, which locates the region 39 proximal to the Pcdh-a cluster (Pcdh-a12, Pcdh-ac1 and Pcdh-ac2 in wild-type mouse brain and PCDHA13, PCDHAC1, and PCDHAC2 in normal human kidney and human kidney tumor) 34,36,38 . Our recent results suggested that the establishment of mosaic DNA methylation patterns in the Pcdh clusters is cooperatively regulated by the specificity of Dnmt3b, the gene cluster structure, the enhancer element, and the sequence features 38 . Another possible mechanism is an altered enhancer-promoter interaction 40,46 . Collectively, the mechanism underlying the shift of the 59-located duplicated Pcdh-ac2 to stochastic expression is an important question for future study.
The experiments described here reveal some role of tandem duplication on gene regulatory differentiation, that include expression level divergences of the duplicated Pcdh-a genes and changes from constitutive to stochastic manner of the 59-located duplicated Pcdh-ac2 gene. The data suggest that the current state of the Pcdh-a cluster in mammals, in terms of how it is expressed, was shaped by tandem duplication and distance from promoter to enhancer. Furthermore, although it is not clear that Pcdh-a gene number or stochastic expression have been strongly selected during vertebrate evolution, the present results may imply the evolutionary history of Pcdh cluster gene regulation. Previous reports have suggested that Pcdh cluster evolution included successive tandem duplications and sequence divergences 16,17,19 . Here, we propose a model in which stochastic expression of the duplicated Pcdh genes is immediately acquired after tandem duplication, which precedes sequence divergence (Fig. 7).
The human PCDH cluster is particularly rich in CNVs, including duplications and deletions 3,20,47 (see Database of Genome Variants: http://dgv.tcag.ca/gb2/gbrowse/dgv2_hg18/?name5chr5:140050001. 141050000). The present study showed that the tandem duplication resulted in healthy mice with a macroscopically normal brain. This result can be explained in part by the maintenance of the expression levels and distribution patterns of the total Pcdh-a transcript, by the re-allocation of Pcdh-a isoform expression, and by the maintenance of the expression levels of neighboring genes, upon duplication. Thus, it is likely that following various CNV events in the human PCDH cluster, the total expression level, dual gene-regulatory mechanisms, and stochastic expression of the human PCDH genes are maintained. This robustness may provide the predominant reason for the frequent CNVs in the human PCDH cluster. One study reported that there is no phenotypic link between a CNV in the human PCDH cluster, a 16.7-kb deletion affecting PCDHA8-A10, and psychiatric disorders 47 . Recently, a de novo gene disruption in PCDHA13 was reported in autism 48 . Furthermore, Anitha A. et al. reported strong genetic evidence of PCDHA as a potential candidate gene for autism 49 . The PCDHA cluster is also a candidate locus for bipolar disorder 50 . Furthermore, deletion of PCDHA1-PCDHA9 is associated with higher brain function, such as music perception 51 . It will be interesting to investigate the effects of these genetic mutations on the PCDHA expression and neural circuit formation.
Recent human genome analyses revealed that tandem duplications contribute to human phenotypes, including many psychiatric disorders, color vision, Parkinson's disease, and Rheumatoid arthritis 1,3,52 . For example, duplications of 7q36.3, which contains the vasoactive intestinal peptide receptor gene VIPR2, confer significant risk for schizophrenia, and VIPR2 mRNA levels are increased differently among duplication carriers 53 . These findings suggest the importance of conducting detailed investigations addressing the effects of duplications on gene regulation. The iTAMERE approach enables careful analyses directed toward understanding the etiology of CNVassociated human disorders.
The vertebrate Pcdh cluster shows remarkable similarity to the Drosophila Dscam1 gene, within which tandem duplication is frequent throughout its evolutionary history 54 . Importantly, recent studies demonstrated that the stochastic gene regulation in Pcdh and Dscam1 play important roles in neural circuit development by providing a source for cell surface diversity 11,12 . These findings suggest essential roles for tandem duplications in the evolution of vertebrate and invertebrate nervous systems. Further studies aimed at dissecting fine-scale neural circuits in the Pcdha dup(2-c2)/dup(2-c2) mice will improve our understanding how tandem duplication in the Pcdha cluster contribute to brain function. Hypothetical evolutionary history from clustered Pcdh-a genes. Tandem duplication (orange shading) creates duplicated exons in the Pcdh-a cluster. Both of the duplicate stochastically expressed Pcdh-a isoforms (those with black-grey arrows) retain their stochastic expression. In contrast, from constitutively expressed isoforms (thick red arrow), the 39-located duplicated Pcdh-a maintains its constitutive expression, but the 59-located one shifts to stochastic expression accompanied by higher DNA methylation. Mutations in exons and/or regulatory sequences generate diverse coding exons, pseudogenes, and altered gene regulatory systems. Boxes, Pcdh-a exons. Dotted box, pseudogene. Methods Animals. B6 mice were purchased from Charles River Japan. The wild-type mouse strain JF1 was obtained from the National Institute for Genetics (Mishima, Shizuoka, Japan). All animals were maintained in a specific pathogen-free space under a 12-h light/dark regimen. Experimental procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals of the Science Council of Japan and approved by the Animal Experiment Committee of Gunma University and Osaka University.
Generation of Pcdha dup(2-c2)/dup(2-c2) mice. By crossing SR mice (see Supplemental Experimental Procedures) with Sycp1-Cre transgenic mice 25 , we generated SR mice carrying the Sycp1-Cre transgene. We crossed these mice with mice bearing the G16Neo allele, then selected male offspring bearing the SR allele, G16Neo allele, and the Sycp1-Cre transgene (Pcdha SR/G16Neo , Sycp1-Cre). We then crossed Pcdha SR/G16Neo , Sycp1-Cre male and B6 female, and genotyped the pups using genomic DNA extracted from the tail. Some of these pups carried the duplicated [dup(2-c2)] or deleted [del(2-c2)] allele as a result of TAMERE in the testis 31 . To identify the Pcdha wt/dup(2-c2) mice, we performed Southern blotting and PCR analyses ( Fig. 1e and  1f). The BamHI-digested genomic DNA from the tail was subjected to Southern blotting using probe A (the same probe used to identify the SR allele). A band at 8.6 kb indicated the wild-type allele, and a band at 12.4 kb indicated the dup(2-c2) allele. Pcdha dup(2-c2)/dup(2-c2) mice were obtained by crossing Pcdha wt/dup(2-c2) parents, which were backcrossed with B6 for more than three generations.
In situ hybridization. In situ hybridization (ISH) was performed essentially as described previously 27 . The details are provided in the Supplemental Experimental Procedures.
Expression analysis in cerebellum. The cerebellum was dissected from 4-week-old mice and immediately frozen in liquid nitrogen. The tissue was homogenized with a Polytron homogenizer. Total RNA was isolated using RNeasy (Qiagen), according to the supplier's recommendations. To obtain cDNA, 2.5 mg of the total RNA was treated with DNase I (Takara) and reverse transcribed with SuperscriptIII reverse transcriptase (Invitrogen) using random primers in a 40-ml reaction volume.
For the SNP analyses, PCR for each Pcdh-a isoform was performed using 0.4 ml of cDNA from the cerebellum of a 4-week-old mouse as a template. The primer sequences used for the SNP analyses are shown in Supplementary Table S2. For direct SNP analyses, the PCR products were sequenced using a standard method. For the cloning-mediated SNP analysis, the PCR products were cloned into pT7-Blue (Novagen), white colonies were randomly picked, and individual clones were sequenced using a standard method. The SNPs used for the direct sequencing analysis are shown in Supplementary Table S3.
The qRT-PCR was performed with SYBR Premix ExTaq II (Takara) using the 7500Fast Real-Time PCR system (Applied Biosystems) or LightCycler480 (Roche). The primer sequences used for qRT-PCR are shown in Supplementary Table S2. All data shown are normalized to beta 2 microglobulin. The efficiency of all the primer pairs was confirmed by performing reactions with serially diluted samples. The specificity of all the primer pairs was confirmed by analyzing the dissociation curve.
Split single-cell RT-PCR. Single-cell RT-PCR was performed essentially as described previously, with small modifications 24,55 . The details are provided in the Supplemental Experimental Procedures.
DNA methylation analysis by bisulfite sequencing and HpaII digestion. The genomic DNA was prepared with the QIAamp DNA Micro Kit (Qiagen) from the cerebellum of a 4-week-old or 3-month-old mouse, or from sperm, or with the EpiTect Plus LyseAll Kit (Qiagen) from the brain of an E12.5 embryo, head of an E9.5 embryo, E7.5 whole embryo, or blastocyst (E3.5), according to the supplier's recommendations. Sperm was obtained from the cauda epididymides of adult male mice. Embryos were obtained by the natural breeding of Pcdha wt/dup(2-c2) parents. The morning of the vaginal plug was designated E0.5. Bisulfite conversion was performed with the EpiTect Plus DNA Bisulfite Kit (Qiagen), according to the supplier's recommendations. We used ''MethPrimer'' (http://www.urogene.org/methprimer/) to design primers for use on bisulfite-treated DNA 56 . The primers used for DNA amplification are listed in Supplementary Table S2. The first PCR program consisted of 95uC for 3 min, 25 cycles of 95uC for 30 s, 60uC for 30 s, 72uC for 30 s, and a final extension of 72uC for 7 min. The second PCR was carried out using the same program as the first, except that 32 cycles were performed. To obtain the methylation profile from the acquired data, we used the web-based tool, ''QUMA'' (http://quma.cdb. riken.jp/) 57 .
In the HpaII digestion-mediated DNA methylation analysis, HpaII was used to distinguish between methylated (undigested) and unmethylated (digested) HpaII/ MspI sites, whereas MspI digested both methylated and unmethylated HpaII/MspI sites. The HpaII/MspI-digested DNA was subjected to PCR analysis to amplify the SNP-containing regions. We mixed 500 ng of DNA with 10 U HpaII or 10 U MspI, and restriction buffer in a 10-ml reaction volume. Samples were incubated at 37uC overnight, and we used 40 ng of the digested DNA for the PCR reaction. The primers are listed in Supplementary Table S2. The PCR program consisted of 95uC for 3 min, 30 cycles of 95uC for 30 s, 60uC for 30 s, 72uC for 1 min, and a final extension of 72uC for 7 min. The PCR products were sequenced using a standard method. The results shown are representative of two independent experiments. Statistical analysis. Statistical analysis was performed using GraphPad Prism Version 6.0 (GraphPad Software, La Jolla, CA, USA) and was performed using oneway ANOVA followed by Bonferroni's post hoc test, or using an unpaired two-tailed Student's t test if applicable. All data are expressed as the mean 6 S.E.M.