Developmental Epigenetic Modification Regulates Stochastic Expression of Clustered Protocadherin Genes, Generating Single Neuron Diversity

Neuron

Article Developmental Epigenetic Modification Regulates Stochastic Expression of Clustered Protocadherin Genes, Generating Single Neuron Diversity Shunsuke Toyoda,1,2 Masahumi Kawaguchi,1 Toshihiro Kobayashi,3,4 Etsuko Tarusawa,2,5 Tomoko Toyama,1 Masaki Okano,6 Masaaki Oda,6 Hiromitsu Nakauchi,3,4 Yumiko Yoshimura,5,7 Makoto Sanbo,2,8 Masumi Hirabayashi,2,7,8 Teruyoshi Hirayama,1,2 Takahiro Hirabayashi,1,2 and Takeshi Yagi1,2,* 1KOKORO-Biology

Group, Laboratories for Integrated Biology, Graduate School of Frontier Biosciences Science and Technology Agency-Core Research for Evolutional Science and Technology, CREST Osaka University, 1-3 Yamadaoka, Suita, Osaka 565-0871, Japan 3Division of Stem Cell Therapy, Center for Stem Cell Biology and Regenerative Medicine, Institute of Medical Science, University of Tokyo, 4-6-1 Shirokanedai, Minato-ku, Tokyo 108-8639, Japan 4Japan Science Technology Agency, ERATO, 4-6-1 Shirokanedai, Minato-ku, Tokyo 108-8639, Japan 5Division of Developmental Neurophysiology, National Institute for Physiological Sciences, National Institutes of Natural Sciences, Okazaki, Aichi 444-8787, Japan 6Laboratory for Mammalian Epigenetic Studies, RIKEN Center for Developmental Biology, Minatojima-minamimachi 2-2-3, Chuo-ku, Kobe, Hyogo 650-0047, Japan 7Department of Physiological Sciences, The Graduate University for Advanced Studies, Okazaki, Aichi 444-8585, Japan 8Section of Mammalian Transgenesis, Center for Genetic Analysis of Behavior, National Institute for Physiological Sciences, Okazaki, Aichi 444-8585, Japan *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2014.02.005 2Japan

SUMMARY

In the brain, enormous numbers of neurons have functional individuality and distinct circuit specificities. Clustered Protocadherins (Pcdhs), diversified cell-surface proteins, are stochastically expressed by alternative promoter choice and affect dendritic arborization in individual neurons. Here we found that the Pcdh promoters are differentially methylated by the de novo DNA methyltransferase Dnmt3b during early embryogenesis. To determine this methylation’s role in neurons, we produced chimeric mice from Dnmt3b-deficient induced pluripotent stem cells (iPSCs). Single-cell expression analysis revealed that individual Dnmt3b-deficient Purkinje cells expressed increased numbers of Pcdh isoforms; in vivo, they exhibited abnormal dendritic arborization. These results indicate that DNA methylation by Dnmt3b at early embryonic stages regulates the probability of expression for the stochastically expressed Pcdh isoforms. They also suggest a mechanism for a rare human recessive disease, the ICF (Immunodeficiency, Centromere instability, and Facial anomalies) syndrome, which is caused by Dnmt3b mutations. INTRODUCTION The mammalian brain contains enormous numbers of neurons that have distinct circuit specificities, and the stochastic expres94 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

sion of diversified cell-surface proteins is important for specifying individual neuronal identity (Yagi, 2013). In mice, each olfactory sensory neuron expresses only one of more than 1,000 odorant receptor genes, which is selected by common enhancer elements and allelic exclusion and permits the precise recognition of their axonal targets (Buck and Axel, 1991; Chess et al., 1994; Serizawa et al., 2003). In Drosophila, a subset of more than 19,000 isoforms of Down syndrome cell adhesion molecule (Dscam) 1 is alternatively spliced and expressed stochastically and combinatorially in each neuron; this diversity is required for self- and nonself recognition in neurite arborizations (Schmucker et al., 2000; Hattori et al., 2007; Matthews et al., 2007). Clustered protocadherins (Pcdhs) comprise a large subfamily of the cadherin superfamily (Zipursky and Sanes, 2010; Yagi, 2013; Chen and Maniatis, 2013). In mammals, more than 50 Pcdh genes are organized into three gene clusters, Pcdh-a, Pcdh-b, and Pcdh-g, which encode diversified transmembrane proteins that are predominantly expressed in the nervous system (Kohmura et al., 1998; Wu and Maniatis, 1999) (Figure 1A). In mice, the Pcdh-a and Pcdh-g clusters contain, respectively, 14 and 22 variable exons encoding the extracellular, transmembrane, and membrane-proximal part of the cytoplasmic domain, and three constant region (CR) exons encode the common portion of the cytoplasmic domain. In contrast, the Pcdh-b cluster consists of 22 single-exon genes. Each variable exon of the clustered Pcdhs is transcribed from its own promoter (Tasic et al., 2002; Wang et al., 2002a). Gene ablation experiments showed that Pcdh-a and Pcdh-g are required for neuronal survival, synapse formation, axonal targeting, dendritic arborization, and the self-avoidance of dendrites (Wang et al., 2002b; Weiner et al., 2005; Hasegawa

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

A

B

C

D

E

Figure 1. Correlation between Methylation of Pcdh Promoters and Gene Expression in the Brain (A) Genomic structure of the Pcdh gene clusters. V indicates variable exons; C indicates constant exons. (B) Genomic organization and representative expression patterns of the Pcdh-a cluster in wild-type (WT; blue) and Da11-C2 neurons (magenta). (C) Bisulfite sequencing of the Pcdh-a promoter regions in the cerebral cortex at P21. Each row represents a single DNA strand; the filled and open circles represent methylated and unmethylated CpG nucleotides, respectively. Average CpG methylation levels are indicated as a percentage. (D) Quantification of CpG methylation levels in P21 cerebral cortex. n = 3 mice for each genotype. Data show mean ±SEM (*p < 0.05; ***p < 0.001; t test). (E) Top, diagram showing positions of the restriction enzyme sites, probe (gray box), and a10 variable exon. Numbers indicate the length of each fragment (kb). B indicates BglII; H indicates HpaII (a methylation-sensitive enzyme that cleaves at CCGGs); M indicates MspI (a non-methylation-sensitive isoschizomer of HpaII). Bottom, methylation-sensitive Southern blot of genomic DNA extracted from WT and Da11-C2 cerebella at P21. See also Figure S2 and Table S1.

Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc. 95

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

et al., 2008; Katori et al., 2009; Chen et al., 2012; Garrett et al., 2012; Lefebvre et al., 2012; Suo et al., 2012). Moreover, homophilic interactions mediated by heterotetramers of the stochastically expressed clustered Pcdh isoforms at the cell surface could generate an enormous selection of potential combinations, thus allowing very specific cell-cell recognition (Schreiner and Weiner, 2010). In support of this idea, single-cell RT-PCR analyses of Purkinje neurons revealed that the Pcdh-a, -b, -gA, and -gB isoforms are expressed stochastically and monoallelically in individual neurons, whereas the C-type exons, including aC1, aC2, gC3, gC4, and gC5, are expressed constitutively from both alleles in most neurons; in addition, in situ hybridization analyses also showed scattered expression patterns of Pcdh isoforms in the differentiated neurons of various brain regions (Esumi et al., 2005; Kaneko et al., 2006; Noguchi et al., 2009; Hirano et al., 2012). Thus, the clustered Pcdhs have the potential, at least in concept, to specify the identity and circuitry of individual neurons. Epigenetics is sequence-independent transcriptional regulation that is essential for the establishment and inheritance of cellular identity during development (Bird, 2002). In neuroblastoma cell lines that express specific subsets of Pcdh-a isoforms as well as in individual neurons, the transcriptionally active promoters are mostly unmethylated, and the silent ones are highly methylated (Tasic et al., 2002; Kawaguchi et al., 2008). Furthermore, treatment with 5-azacytidine induces the expression of previously silenced isoforms (Kawaguchi et al., 2008). Therefore, the methylation patterns of the Pcdh promoter regions correlate with gene expression in vitro. However, the in vivo mechanisms for transcription regulation are not fully understood. Here we show that the promoter regions of stochastically expressed Pcdh isoforms are differentially methylated by the de novo DNA methyltransferase Dnmt3b (Okano et al., 1999), beginning on embryonic day (E) 3.5. We produced chimeric mice from Dnmt3b-knockout (KO) induced pluripotent stem cells (iPSCs), which survived to adulthood. However, Dnmt3b-KO Purkinje cells showed impaired dendritic arborization. We found that individual Dnmt3b-KO Purkinje cells expressed increased numbers of Pcdh isoforms at the single neuron level, although the overall expression level, at least within the Pcdh-a and -g clusters, was maintained. These results indicate that DNA methylation by Dnmt3b at early embryonic stages regulates both the probability of expression for the stochastically expressed Pcdh isoforms in single differentiated neurons and the dendritic arborization. RESULTS Methylation Patterns of Pcdh-a Promoter Regions Are Correlated with Isoform Expression in the Brain In wild-type (WT) brains, the Pcdh-a1 to -a12 isoforms are expressed stochastically and monoallelically in individual neurons, whereas the aC1 and aC2 isoforms are expressed constitutively and biallelically in most neurons (Figure 1B). However, deletion of the variable exons from a11 to aC2 (Da11-C2) causes the constitutive and biallelic expression of a10 (Noguchi et al., 2009) (Figure 1B). To reveal the relationship between promoter methylation and gene expression in vivo, we performed methylation analysis 96 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

of the Pcdh-a promoter regions in the cerebral cortex of Da11-C2 mice at postnatal day (P) 21. Bisulfite sequencing showed that in WT cortex, the promoter regions of a4, a9, and a10 were methylated differentially and mosaically, whereas those of aC1 and aC2 were mostly unmethylated (Figures 1C and 1D), consistent with our previous study (Kawaguchi et al., 2008). In contrast, the a10 promoter region in the Da11-C2 cortex was mostly unmethylated (Figures 1C and 1D). Methylation-sensitive Southern blot analysis for the a10 locus showed that HpaII digestion yielded four distinct bands in WT, but only a 1.5 kb band was detected in the Da11-C2 locus (Figure 1E), confirming the unmethylated status of a10 in Da11C2 brains. We further examined the promoter regions of a4 and a9, which are expressed stochastically in Da11-C2 neurons (Noguchi et al., 2009). Their promoters maintained differential methylation patterns, but at significantly reduced levels (Figures 1C and 1D). Thus, the promoter regions of stochastically expressed Pcdh isoforms are differentially methylated, and those of constitutively expressed Pcdh isoforms are mostly unmethylated. De Novo Methylation of Pcdh-a Promoter Regions during Development Given that the Pcdh-a genes are exclusively expressed in neural tissues (Kohmura et al., 1998), we compared the methylation patterns of their promoters in the cerebral cortex and liver of C57BL/6 mice at P21. Bisulfite sequencing revealed similar methylation patterns between them, with a tendency toward lower methylation levels in the liver (Figures 2A and 2B), suggesting that the differential methylation is not specific to the brain. Next, we examined the differential methylation patterns during development. In E3.5 blastocysts, all the Pcdh-a promoter regions analyzed here were hypomethylated, at less than 10% (Figures 2A and 2B). The methylation of the promoter regions of a1 to a12 gradually increased through E9.5, whereas those of aC1 and aC2 were demethylated and/or maintained in a hypomethylated state (Figures 2A and 2B). These results indicate that the promoter regions of a1 to a12, but not of aC1 and aC2, are de novo methylated during early embryonic stages, starting from E3.5. Dnmt3b Is Required for the De Novo Methylation of Pcdh Promoter Regions In mouse development, the DNA is globally demethylated from fertilization to implantation, and subsequent methylation patterns are established by the de novo DNA methyltransferases Dnmt3a and Dnmt3b (Okano et al., 1999). To test whether these two Dnmts are responsible for the de novo methylation of the Pcdh-a promoter regions, we examined the methylation states in Dnmt3a-KO and Dnmt3b-KO embryos at E9.5. Bisulfite sequencing revealed no obvious differences from WT in the promoter methylation states for Pcdh-a1 to -a12 in the Dnmt3a-KO embryos, but in the Dnmt3b-KO embryos the methylation was significantly reduced to less than 6.2%; in contrast, the hypomethylation of the promoter regions of aC1 and aC2 was unchanged (Figures 3A and 3D). To confirm these results, we performed methylation-sensitive Southern blot analysis for the a8 locus. In both WT and

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

A

B

Figure 2. DNA Methylation Profiles during Mouse Development (A) Bisulfite sequencing of the Pcdh-a promoter regions in C57BL/6 mice at several developmental stages. About 100 embryos for E3.5 blastocysts, single epiblasts at E7.5, single whole embryos at E9.5, and cerebral cortex and liver at P21 were analyzed. (B) Developmental changes in CpG methylation levels during development. Data are the mean ±SEM. n = 2 to 3 independent experiments. See also Figure S7 and Table S1.

Dnmt3a-KO embryos, 1.35 kb and 1.2 kb bands were strongly detected by HpaII digestion, whereas only a 0.8 kb band was detected in Dnmt3b-KO embryos (Figure 3B), confirming the

specific requirement for Dnmt3b. Moreover, the promoter regions of b22 and gA3 were methylated at similar levels in the WT and Dnmt3a-KO embryos but were hypomethylated in the Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc. 97

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

A

C

B

D

E

G

F

Figure 3. Dnmt3b Is Required for DNA Methylation in Pcdh Gene Clusters (A and C) Bisulfite sequencing of the promoter regions of the Pcdh-a (A) or the Pcdh-b and Pcdh-g (C) isoforms in WT, Dnmt3a-KO, and Dnmt3b-KO embryos at E9.5. (B) Top, schematic diagram of position of the restriction enzyme sites, probe (gray box), and a8 variable exon (black box). The numbers indicate the length of each fragment (kb). S indicates SacI; A indicates ApaI; H indicates HpaII; M indicates MspI. Bottom, methylation-sensitive Southern blot analysis for genomic DNA extracted from single embryos at E9.5 of Dnmt3a-KO, Dnmt3b-KO, and each WT littermate. (legend continued on next page)

98 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

Dnmt3b-KO embryos; the gC3 promoter regions were hypomethylated in all genotypes (Figures 3C and 3D). An intracisternal A-particle (IAP), an endogenous retrotransposon, is inserted between a7 and a8 (aIAP in Figure S1A, available online). The long-terminal repeat (LTR) of aIAP was highly methylated, to about 90%, even in Dnmt3b-KO embryos at E9.5, although the level in Dnmt3b-KO embryos was slightly lower than in WT (Figures S1B and S1C). Therefore, the methylation of the aIAP LTRs is independent of the methylation of Pcdh promoter regions. Previous studies showed that IAPs are not demethylated globally after fertilization (Lane et al., 2003) and are partially hypomethylated in Dnmt3b-KO embryos (Okano et al., 1999). Chromatin immunoprecipitation (ChIP) analysis in WT embryos at E7.5 revealed that Dnmt3b was more enriched than Dnmt3a at each Pcdh promoter region (Figure 3E), strongly suggesting that Dnmt3b can directly methylate each Pcdh promoter. Interestingly, Dnmt3b binding appeared, not only in the DNA methylated promoters (a1, a8, b22, and gA3), but also in the unmethylated ones (aC1, aC2, and gC3) (Figure 3E). Therefore, we next examined histone modifications. In E7.5 embryos, repressive di- and tri-methylations of histone 3 lysine 9 (H3K9me2 and H3K9me3) were relatively highly enriched at the promoter regions of a1 and a8 compared with those of aC1 and aC2 and at the gA3 promoter region compared with the gC3 promoter regions; similar enrichments were also observed for another repressive modification, H3K27me3, but not for the permissive H3K4me3 modification (Figure 3F). Notably, the aC2 promoter regions were enriched for both H3K27me3 and H3K4me3 modifications (Figure 3F). These results indicate that the DNA-methylated Pcdh promoters have repressive chromatin modifications, and the aC2 promoter has bivalent histone modifications. Taken together, we demonstrated that Dnmt3b is essential for de novo methylation across the three Pcdh clusters during early embryonic development and that different DNA methylation and histone modifications are established at each Pcdh promoter. De Novo Methylation of Pcdh Promoter Regions Depends on Gene Cluster Structure We evaluated the abnormal methylation in the nonneural tissue and embryo of Da11-C2 mice. Surprisingly, bisulfite sequencing showed that methylation in the Pcdh-a promoter regions in Da11C2 mice was altered in the liver at P21 (Figures S2A and S2B) as well as in the cortex (Figures 1C and 1D). These alterations were detectable in Da11-C2 embryos at E9.5 (Figures S2C and S2D), indicating that the de novo methylation by Dnmt3b in the Pcdh promoter regions is dependent on the gene cluster structure. Nonetheless, in WT embryo the methylation levels were negatively correlated with the number of CpGs per 100 bp

(r =
0.82; p = 0.012) (Figure 3G) except at a8 (attributable to the high methylation near aIAP) (Figure S1A), indicating that the CpG density links to the methylation frequency. Promoter Methylation Controls the Allocation of Pcdh Isoform Expression within Each Cluster To determine how differential methylation of the Pcdh promoter regions affects expression, we examined whole brains of Dnmt3b-KO mice at E13.5 (because these mice die by around E15.5) (Okano et al., 1999; Ueda et al., 2006). Quantitative RT-PCR (qRT-PCR) analysis showed that the aCR and gCR transcripts, which represent the total expression levels of the Pcdh-a and Pcdh-g isoforms, were, respectively, 1.2-fold higher or unchanged in the Dnmt3b-KO brains, whereas the effects on the first exons varied: some were upregulated, some downregulated, and some were unchanged (Figure 4A). Bisulfite sequencing showed that the Pcdh promoter regions analyzed here were almost entirely unmethylated in the Dnmt3b-KO brain (Figures S3A and S3B). Importantly, the expression changes of each isoform were highly correlated with the changes in its methylation level (Figure 4B). In particular, the methylation of a8, whose promoter was relatively highly methylated in WT brain, was reduced by almost 70% in the Dnmt3b-KO brain (Figures S3A and S3B), and its expression level increased 12-fold (Figure 4B). On the other hand, the expression levels of aC1, aC2, gC3, and gC4, whose promoter regions were hypomethylated in WT brain (Figures S3A and S3B), decreased to less than 50% of WT (Figure 4B). These results indicate that the methylation of Pcdh promoter regions regulates the distribution of expression across the isoforms but is less important for the total expression of isoforms at least within the Pcdh-a and -g clusters. To test whether the expression patterns of Pcdh isoforms were altered in the Dnmt3b-KO brain, we performed in situ hybridization analysis for gCR, which was shown to be highly expressed by qRT-PCR (Figure 4A). Although the signal intensity seemed slightly stronger than in WT, the overall pattern of gCR expression in the Dnmt3b-KO brain was similar, with postmitotic neurons in the cortical plate and other structures expressing it most strongly (Figure S4). This observation suggests that the overall pattern of Pcdh-g expressions in neurons is maintained in the Dnmt3b-KO brain. Generation of Chimeric Mice Using Dnmt3b-KO iPSCs The abnormal Pcdh isoform expression in the Dnmt3b-KO embryonic brain might be due to a change in the number of Pcdh isoforms expressed in individual neurons. To evaluate isoform expression in single neurons, we performed a mosaic experiment with Dnmt3b-KO neurons. Because the Pcdh cluster in Dnmt3b-KO embryonic stem cells (ESCs) had been already

(D) Quantification of the CpG methylation levels in whole embryos at E9.5. n = 6 embryos for WT, and n = 3 embryos for Dnmt3a-KO and Dnmt3b-KO. Data are the mean ±SEM. ***p < 0.001, Bonferroni’s post hoc test following two-way analysis of variance (ANOVA). (E and F) ChIP analysis of the Pcdh promoter regions using Dnmt3-specific (E) or histone-specific (F) antibodies in WT embryos at E7.5. Data are the mean ±SEM. n = 3 to 4 independent experiments. As the control, mitochondrial DNA 16S rRNA (negative control region for Dnmts), Rhox6 (highly DNA methylated region; Oda et al., 2006), and Gapdh were used. (G) Average methylation levels in WT embryos at E9.5 plotted against the number of CpGs per 100 bp. The Pearson’s correlation coefficient is shown in the graph. See also Figure S1 and Table S1 and S2.

Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc. 99

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

A

B

Figure 4. Promoter DNA Methylation Regulates the Allocation of Pcdh Isoform Expression within Each Pcdh Cluster (A) Quantitative RT-PCR analysis for the clustered Pcdh isoforms. aCR and gCR represent the total expression levels of the Pcdh-a and Pcdh-g isoforms, respectively. n = 4 whole brains at E13.5 for each genotype. Data are the mean ±SEM. *p < 0.05; **p < 0.01; ***p < 0.001; t test. (B) Scatter plots show the relationship between the fold change in expression level and the difference in DNA methylation level for various Pcdh isoforms in the Dnmt3b-KO brains. The Pearson’s correlation coefficient is shown in the graph. See also Figures S3 and S4 and Tables S1 and S3.

DNA methylated (Figure S5A), we therefore established iPSCs from Dnmt3b-KO embryos (Figure 3A) that ubiquitously expressed EGFP by crossing with a ‘‘green’’ mouse, and injected them into blastocysts of C57BL/6 mice to produce chimeras (Figure 5A). The resulting mice exhibited differing patterns of chimerism that could be assessed from the agouti coat color (Figure S5B). Although these mice survived to adulthood, some mice with high Dnmt3b-KO chimerism showed reduced body weight, growth retardation, and/or abnormal limb-clasping reflexes when suspended by their tails (Figures S5B–S5D). In the brain of the Dnmt3b-KO chimeras, the strong EGFP expression was observed in various brain regions (Figure S5E). For direct comparison, we also generated chimeric mice from iPSCs derived from MEFs of a WT green mouse (WT-EGFP). In both genotypes, similar EGFP signals were observed (1) in the pyramidal and inhibitory neurons and in the glia of the cerebral cortex and hippocampus (Figure S5F) and (2) in the Purkinje cells and 100 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

granule cells in the cerebellum (Figure S5G), indicating that Dnmt3b is dispensable for the generation of these cell types. Abnormal Dendritic Arborization in Dnmt3b-KO Purkinje Cells Given that Purkinje cells have elaborate dendritic arborizations with minimally overlapping dendritic branches (viewed from the parasagittal plane), we asked whether the dendrites of Dnmt3b-KO Purkinje cells developed normally. We obtained confocal z stack images of EGFP-labeled Purkinje cells from WT-EGFP and Dnmt3b-KO chimeras. These images were processed to generate 3D reconstructions, and the dendrites were traced by Imaris software (Figure 5B). The Dnmt3b-KO Purkinje cells frequently showed dendrites that formed bundles and that crossed one another (Figures 5B and 5C). To quantify the self-crossing in Dnmt3b-KO Purkinje cells, we counted the times of dendritic branches from a single Purkinje cell crossed one another in individual confocal planes and found double the number of crossings in Dnmt3b-KO Purkinje cells than in WT (Figure 5D). To determine whether these effects were cell autonomous, we injected biocytin into WT (EGFP-negative) and Dnmt3b-KO (EGFP-positive) Purkinje cells that were surrounded by Dnmt3b-KO cells in Dnmt3b-KO chimeras (Figure S5H). The number of dendritic self-crossings of WT Purkinje cells was similar to that of WT-EGFP Purkinje cells, whereas the Dnmt3b-KO Purkinje cells showed significantly more instances of dendritic crossing (Figure S5I). These findings suggest that Dnmt3b is required cell autonomously for avoidance of the dendritic self-crossing. To quantify the dendritic development in the absence of Dnmt3b in detail, we also measured the morphology of the Purkinje cell dendrites using Imaris software. In Dnmt3b-KO Purkinje cells, the number and total length of branches were significantly reduced (Figures 5E and 5F), whereas the area occupied by the dendritic arbors showed a bimodal distribution, which was significantly diminished in more than half of the Dnmt3b-KO Purkinje cells (Figure 5G). To evaluate the monoplanarity of the Purkinje cell arbors, we quantified the length and area occupied by the ‘‘overpassing’’ dendrites (i.e., those that crossed over other branches) (Figures 5B and 5C). In WT-EGFP Purkinje cells, the length and area occupied by overpassing dendrites was, respectively, 6.3% and 5.4% of the total, but these values were 24.5% and 31.4% in Dnmt3b-KO Purkinje cells (Figures 5H and 5I). In contrast, spine density was not altered by the Dnmt3b deficiency (Figure 5J). Taken together, these results indicate that Dnmt3b is required for normal dendritic patterning in Purkinje cells. Expression Profile of Pcdh Isoforms in Single Purkinje Cells Because the expression analysis of Pcdh isoforms in Dnmt3bKO brains at E13.5 suggested that they might be expressed constitutively in single Dnmt3b-KO neurons, we developed a method for high-throughput expression profiling of Pcdh-a isoforms. Using Dynamic Array chips (Fluidigm), a microfluidic real-time PCR system, we were able to quantify the expression of all the Pcdh-a isoforms in parallel at the single-cell level.

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

Due to the high homology between Pcdh-a isoforms, we improved the specificity of the system by using isoform-specific primers and TaqMan probes. For this experiment, the cerebellum from Dnmt3b-KO chimeras was dissociated, and individual EGFP-positive or -negative single Purkinje cells, identified by their morphology (Figure 6A), were picked up and subjected to real-time PCR analysis. As the control, we analyzed EGFP-positive single Purkinje cells collected from WT-EGFP mice. Most individual cells of all genotypes expressed Tubulinb3 (Tubb3), Purkinje cell protein 2 (Pcp2), and Glutamate decarboxylase 67 (Gad67), which are markers for neurons, Purkinje cells, and inhibitory neurons, respectively, but not Vesicular glutamate transporter 1 (VGlut1) or Glial fibrillary acidic protein (GFAP), which are markers for excitatory neurons and glia (Figure 6B), confirming that we had correctly identified the Dnmt3b-KO Purkinje cells. Notably, some cells exhibited the overlapping expression of different cellular markers and/or no expression of Pcp2, although they showed similar Pcdh-a gene expression patterns (Figure 6B). Therefore, we included these cells in further analyses. We found that Pcdh-aC1 and aC2 were expressed in most individual WT (25 cells) and WT-EGFP (7 cells) cells; in contrast, one to four isoforms from Pcdh-a1 to -a12 were expressed in each cell (Figures 6B and 6C). On the other hand, the individual Dnmt3b-KO cells expressed significantly higher numbers of Pcdh-a isoforms compared with WT and WT-EGFP cells (Figures 6B and 6C; p < 0.0001 for Bonferroni’s test following oneway ANOVA). Among 23 Dnmt3b-KO cells, 21 cells (91.3%) expressed more than seven isoforms, and five cells (21.7%) expressed all the Pcdh-a isoforms; only two cells (8.7%) maintained the expression of small numbers of isoforms as observed in WT cells (Figure 6C). Nonetheless, the overall expression of Pcdh-a isoforms (aCR) showed no obvious difference between genotypes (Figure 6B; p > 0.05 for one-way ANOVA), indicating that the allocation of Pcdh-a isoform expression was disrupted by the Dnmt3b deficiency. In addition, the expression frequencies of a1 to a12, but not of aC1 and aC2, were markedly increased in the Dnmt3b-KO cells (Figure 6D). These changes in expression frequency were highly correlated with the changes in expression levels in Dnmt3b-KO brains at E13.5 (Figure 6E), suggesting that the aberrant expression frequency of Pcdh-a isoforms in single cells is reflected in their overall levels of expression in the brain, as detected by qRTPCR analysis (Figures 4A and 4B). Moreover, the expression frequencies of stochastically expressed gA and gB isoforms were also increased in the Dnmt3b-KO cells, whereas the overall expression of Pcdh-g isoforms (gCR) was maintained (Figures S6A and S6B). Taken together, these results clearly indicate that the Dnmt3bmediated DNA methylation regulates the probability of expression of each Pcdh stochastically expressed isoform in single neurons, but it is not responsible for the overall expression of the genes within the Pcdh-a and -g clusters. Relationship between Enhancer Elements and Methylation of Promoter Regions Each Pcdh cluster is independently controlled by distinct transcriptional enhancer elements (Ribich et al., 2006; Kehayova

et al., 2011; Yokota et al., 2011) (Figure S7A). DNaseI hypersensitive site (HS) 5-1 is located downstream of the Pcdh-a cluster (Figure S7A), and its deletion (DHS5-1) differentially affects Pcdh-a isoform expression. To determine whether the enhancer-dependent expression affects methylation, we examined the methylation states of Pcdh-a promoter regions in DHS51 brains. Bisulfite sequencing revealed that the methylation pattern was similar to that of WT brain (Figures S7B and S7C). Notably, although the deletion of HS5-1 causes significant decreases in the expression of a10, a12, and aC1 (Yokota et al., 2011), only the methylation level of the a12 promoter region specifically increased, by nearly 2-fold (Figures S7B and S7C). A similar increase was observed in E9.5 embryo, but not in the liver (Figure S7D), suggesting that the deletion of HS5-1 leads to additional methylation preferentially in the brain, which may be partially owing to changes during de novo methylation. The expression of Pcdh-b isoforms is regulated by the Pcdh-b cluster control region (CCR), which is downstream of the Pcdh-g cluster (Figure S7A). The deletion of CCR (DCCR) greatly decreases the expression of almost all the Pcdh-b isoforms (Yokota et al., 2011). A methylation analysis of the DCCR brain revealed that the Pcdh-b promoter regions maintained differential methylation patterns, but the methylation levels, especially in the b3 promoter region, were increased (Figure S7E). Overall, these results suggest that the methylation of the Pcdh promoter regions is largely independent of the enhancer-dependent transcription; however, some specific promoter regions show increased methylation. DISCUSSION The clustered Pcdh genes are an attractive candidate for determining neuronal diversity at the molecular level, because distinct subsets of them are differentially expressed in each neuron. However, the biological significance of their epigenetic regulation in neurons remains largely unclear. Here we show that DNA methylation by Dnmt3b during early embryonic development is necessary for the differential expression of the clustered Pcdh isoforms, which generate neuronal diversity, and for the dendritic arborization of individual Purkinje cells. Role of DNA Methylation in Clustered Pcdh Expression Previous reports showed that each Pcdh cluster is independently regulated by cluster-specific enhancer elements (Ribich et al., 2006; Kehayova et al., 2011; Yokota et al., 2011). All Pcdh promoters except for aC2, gC4, and gC5 contain conserved sequence elements (CSEs) (Wu et al., 2001) that are required for promoter activation in vitro (Tasic et al., 2002). The chromatin organizers CCCTC-binding factor (CTCF) and Cohesin bind to CSEs and regulate the normal expression of the Pcdh isoforms by mediating the promoters’ interactions with the enhancer elements (Kehayova et al., 2011; Guo et al., 2012; Hirayama et al., 2012; Monahan et al., 2012). Interestingly, a recent report showed that the CTCF binding to the Pcdh promoter regions in a human neuroblastoma cell line, SK-N-SH, is dependent on CpG methylation (Guo et al., 2012). Genetic mutations of these elements or factors downregulate Pcdh expression in neurons. Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc. 101

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

A

B

C

D

E

F

G

H

I

J

Figure 5. Abnormal Dendritic Arborization in Dnmt3b-KO Purkinje Cells (A) Procedure for generating chimeric mice using Dnmt3b-KO iPSCs. Tg indicates transgenic; MEFs indicates mouse embryonic fibroblasts. (B) Immunostaining for EGFP and dendritic tracing of Purkinje cells in WT-EGFP and Dnmt3b-KO chimeras. Arrows point to representative overpassing dendrites. Scale bars show 50 mm. (legend continued on next page)

102 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

Figure 6. Gene Expression Profiles in Single Purkinje Cells

A

(A) Single Purkinje cells were picked up using a glass capillary under a fluorescence microscope. Arrow and arrowhead indicate EGFP-positive and -negative Purkinje cells, respectively. Left, bright-field; middle, fluorescence. Scale bars show 100 mm. (B) Heatmaps of the expression levels of single cells from three independent experiments. WT (n = 25) and Dnmt3b-KO cells (n = 23) collected from six Dnmt3b-KO chimeras at P15-P21, and WT-EGFP cells (n = 7) collected from two WT-EGFP chimeras at P18. (C and D) Quantification of the number (C) and frequency (D) of the Pcdh-a isoforms expressed in single cells. Samples are shown in (B). (E) Correlation between the change in expression level in the Dnmt3b-KO brain at E13.5 and the change in expression frequency in Dnmt3b-KO single cells. The Pearson’s correlation coefficient is shown in the graph. See also Figure S6 and Tables S4 and S5.

B

C

D

E

In contrast, we found that DNA methylation in the Pcdh promoter regions is not required for the total expression level, at least within the Pcdh-a and -g clusters; rather, it regulates the expression frequency of each Pcdh isoform. Related to this observation, when the frequency of expression of some Pcdh-a isoforms is altered, that of the other

isoforms necessarily changes to maintain the overall expression level (Noguchi et al., 2009). Our single-cell expression analysis revealed that the inactivation of Dnmt3b is sufficient to permit the expression of all the Pcdh-a isoforms and several Pcdh-g isoforms in individual Purkinje cells. In addition, the expression analysis in developing brain suggested that most Pcdh-b and Pcdh-g isoforms are also activated in single neurons by the Dnmt3b deficiency. Nonetheless, we also observed that some Pcdh-a and Pcdh-g isoforms remained repressed in some individual Dnmt3b-KO single cells. Although we cannot exclude the possibility that we failed to detect the expression of some isoforms, different epigenetic states, such as Dnmt3b-independent DNA methylation or histone modification, might have contributed to these differences. On the other hand, the Pcdh-a gene expression in nonneuronal cells was shown to be repressed by binding of the RE1 silencing transcription factor repressor complex to the enhancer elements HS5-1 (Kehayova et al., 2011), suggesting that the differential DNA methylation in Pcdh promoters is not responsible for the Pcdh repression in nonneural cells.

(C) High-magnification images of the boxed regions in (B). Scale bars show 10 mm. (D) Quantification of self-crossings detected in single confocal images. (E–G and J) Quantification of number (E), total length (F), total area (G), and spine density (J) of dendrites. (H and I) Quantification of length (H) and area (I) of overpassing dendrites. n = 9 cells (D)–(I); n = 70 branches of nine cells (J) from three mice at 6–8 weeks for each genotype. Data are the mean ±SEM. *p < 0.05; ***p < 0.001; t test or Mann-Whitney test. See also Figure S5.

Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc. 103

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

Mechanisms of DNA Methylation in Pcdh Clusters A previous genome-wide analysis reported that the Pcdh-a cluster is methylated during early embryogenesis (Borgel et al., 2010). Furthermore, embryonic stem cells, which have a DNA methylation signature similar to that of epiblast cells (Borgel et al., 2010), exhibit bivalent histone modifications in the Pcdh-g cluster (Mikkelsen et al., 2007). These findings support our results that the Pcdh clusters undergo epigenetic modifications at early embryonic stages. Although both Dnmt3a and Dnmt3b are essential for de novo methylation, they have distinct expression patterns and target specificities. Dnmt3b is predominantly expressed in epiblasts, whereas Dnmt3a is expressed more strongly in later stages (Okano et al., 1999), and a genome-wide analysis showed that gene clusters are methylated preferentially by Dnmt3b (Borgel et al., 2010). Here we found that Dnmt3b was more enriched than Dnmt3a at each Pcdh promoter in E7.5 embryos, and the Dnmt3b binding was found not only in the DNA methylated Pcdh promoters but also in the unmethylated ones. On the other hand, DNA-methylated Pcdh promoters had repressive histone modifications compared with the unmethylated Pcdh promoters. Dnmt3a and Dnmt3b binding does not always link to DNA methylation (Jin et al., 2012). H3K9 methylation directs Dnmt3b and is involved in the Dnmt3b-dependent DNA methylation at pericentric heterochromatin (Lehnertz et al., 2003). These findings suggest that differential patterns of histone modifications affect the DNA methylation in the Pcdh clusters. In addition, specific boundaries that regulate long-range DNA methylation were reported in tumor (Dallosso et al., 2009). However, the DNA methylation patterns of some pluripotent genes are largely determined by their proximal sequence elements (Lienert et al., 2011). Here we showed that the Pcdh promoters are differentially methylated by Dnmt3b, and the extent of methylation is highly correlated with the CpG density; nonetheless, deleting a11 to aC2 alters the methylation of the promoter regions. In particular, in the Da11-C2 brain the de novo methylation of the a10 promoter region, but not of the a9 promoter region, does not occur, and the a10 expression is changed to constitutive and biallelic in most neurons. Because the histone modifications at the aC2 promoter region were distinct from those of the other Pcdha isoforms, we speculate that the 30 downstream region protects the 30 variable exon from H3K9me2 and H3K9me3, which inhibit DNA methylation, similar to a previous observation (Lehnertz et al., 2003). Thus, our findings suggest that the establishment of differential DNA-methylation patterns in the Pcdh clusters is cooperatively regulated by the specificity of Dnmt3b, the gene cluster structure, and the sequence features. The deletion of the enhancer elements HS5-1 or CCR had little effect on methylation, suggesting that the establishment of methylation patterns in the Pcdh promoter regions is independent of enhancer-dependent transcription. However, our findings also suggested that the resulting transcriptional downregulation induces additional methylation that stably suppresses specific isoforms. Aberrant hypermethylation in the Pcdh clusters is reported in neuroblastoma cell lines and tumors (Tasic et al., 2002; Kawaguchi et al., 2008; Dallosso et al., 2009; Guo et al., 2012). Conversely, TET1, which converts methylcytosine to hydroxymethylcytosine, which leads to DNA demethylation, 104 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

binds the Pcdh clusters in ES cells (Xu et al., 2011). In addition, upregulation of the Pcdh isoforms coupled with hypomethylation of their promoter regions is associated with maternal care (McGowan et al., 2011), suggesting that demethylation can also occur in the Pcdh clusters. It will be interesting to determine whether established Pcdh methylation patterns can be changed in postmitotic cells. Epigenetic Functions in the Nervous System Strong Dnmt3b expression is observed in mitotic cells in the developing nervous system, but it is lost in postmitotic neurons (Okano et al., 1999; Feng et al., 2005; Watanabe et al., 2006). A rare human recessive disease, the ICF syndrome, which is characterized by immunodeficiency, centromere instability, and facial anomalies, is caused by hypomorphic Dnmt3b mutations; most patients also exhibit mental retardation (Hansen et al., 1999; Okano et al., 1999; Xu et al., 1999). However, little is known about Dnmt3b’s effects in the nervous system. Here we found that Dnmt3b is dispensable for neural differentiation, but Dnmt3b-KO Purkinje cells showed impaired dendritic arborization. Importantly, Pcdh-g KOs in Purkinje cells and retinal starburst amacrine cells lead to impaired dendritic self-avoidance, which is based on a contact-dependent repulsion that maximizes the receptive field (Lefebvre et al., 2012). A time-lapse observation of developing Purkinje cells revealed that contact-dependent branch retraction is mediated by protein kinase C (PKC) (Fujishima et al., 2012), and Pcdh-g regulates PKC signaling (Garrett et al., 2012). In addition, although it is far from the in vivo situation, biochemical and overexpression studies showed that heterotetrameric complexes of Pcdh isoforms mediate the isoform-specific homophilic binding activity (Schreiner and Weiner, 2010). In this context, our results lead us to speculate that, as the Dnmt3b-KO Purkinje cells express larger numbers of Pcdh isoforms across all three Pcdh clusters, the probability of the same heterotetrameric complexes forming on sister branches would be dramatically reduced. However, notably, because Dnmt3b has many target loci, it is possible that the phenotype of Dnmt3b deficiency resulted from defects caused by the misregulation of other target genes. Therefore, the mechanism underlying the dendritic defects in Dnmt3b-deficient Purkinje cells is an important question for future study. In human ICF cells, the Pcdh-gC3 gene is downregulated, even though its promoter region is hypomethylated (Jin et al., 2008). Although at present it is speculative, our findings suggest a potential mechanism in which epigenetic regulation in Pcdh clusters contributes to arbor morphology in Purkinje cells and that the altered methylation of these clusters is the mechanism underlying ICF syndrome. It will be important to clarify these issues in future studies. A Potential Mechanism for Cell-Lineage-Based Specification of Neuronal Identity Some recent studies have shown that clonal relationships between neurons inform their synapse formation and influence brain organization (Torii et al., 2009; Yu et al., 2009); in the visual cortex, clonally related neurons show similar stimulus selectivity (Li et al., 2012; Ohtsuki et al., 2012).

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

and some other types of neurons, the descendants of individual neural precursor cells migrate quite a distance, and they become intermingled with those of other neural precursors, which may have different epigenetic modifications. EXPERIMENTAL PROCEDURES Animals All procedures using animals 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 Osaka University, University of Tokyo, National Institute for Physiological Sciences, and RIKEN Center for Developmental Biology. See Supplemental Experimental Procedures for additional descriptions of the animals used. Preparation of Embryos Noon of the day on which the vaginal plug was detected in the morning was designated as E0.5. Blastocysts were collected at E3.5 by flushing the uterus with PBS. Epiblasts were isolated from extraembryonic ectoderm at E7.5. Whole embryos, excluding extraembryonic tissues, were dissected at E9.5. ESC Culture WT J1 and Dnmt3b-KO ESCs were maintained as described previously (Oda et al., 2006).

Figure 7. Schema for Epigenetic Regulation in Pcdh Clusters Schematic drawing of the Pcdh genes in each cell. Open and closed circles represent unmethylated and methylated CpGs, respectively. The red oval represents the cluster-specific enhancer element. In the model, blastocysts (unmethylated) give rise to embryonic cells. The blue and pink embryonic cells have different methylation patterns, which differentially regulate the isoform expression in individual neurons and may be preferentially inherited by the daughter cells (dotted arrows). See Discussion for more details.

On the basis of our findings, we suggest the following model for neuronal identification (Figure 7). The differential DNA methylation in the Pcdh promoters during early embryonic development differentially regulates the probability of the enhancer-dependent expression in individual neurons but is not responsible for it in nonneural cells. On the other hand, Dnmt3b deficiency causes hypomethylation of all the Pcdh promoters including those of Pcdh-a, Pcdh-b, and Pcdh-g, which can activate the transcription of all the Pcdh isoforms in single neurons, concomitant with abnormal dendritic arborization. In addition, because epigenetic marks can be inherited through cell division, this mechanism could provide epigenetic heterogeneity in individual neural progenitors, resulting in the preferential expression of the same Pcdh isoforms in sister neurons. Thus, our findings may indicate a potential mechanism for cell-lineage-based specification of neuronal identity. For Purkinje cells

Genomic DNA Preparation and Bisulfite Sequencing To extract genomic DNA (gDNA) from embryos, the QIAamp DNA Micro Kit (QIAGEN) was used, according to the manufacturer’s instructions. Approximately 100 blastocysts were pooled for each gDNA preparation for bisulfite sequencing. The gDNA of E7.5 and E9.5 embryos was isolated from individual embryos. The gDNA was prepared from cerebral cortex and liver by standard techniques. Bisulfite sequencing was carried out basically as previously described (Kawaguchi et al., 2008). Briefly, bisulfite treatment of gDNA was performed with an Epitect Bisulfite Kit (QIAGEN), according to the manufacturer’s instructions. The regions of interest were amplified by PCR and subcloned into pT7B vectors (Novagen). Each clone was sequenced by standard procedures. To analyze the methylation profile for the acquired data, we used the web-based tool QUMA (http://quma.cdb.riken.jp/) (Kumaki et al., 2008). The primer sequences are shown in Table S1. Due to difficulty of the PCR amplification, aC2 was analyzed using the 50 region of the exon (Kawaguchi et al., 2008). Methylation-Sensitive Southern Blot Analysis gDNA was digested by restriction enzymes (SacI and ApaI for a8 and BglII for a10), then by MspI or the CpG methylation-sensitive isoschizomer HpaII, and then subjected to Southern blot analysis. Probes that recognize the upstream regions of a8 (Kaneko et al., 2009) or a10 (Kawaguchi et al., 2008) were labeled with Fluorescein-High Prime (Roche) and recognized by an Anti-FluoresceinAP Fab fragment (Roche). The signals were detected by the CDP-Star detection reagent (GE Healthcare). ChIP Details are in the Supplemental Experimental Procedures. See also Table S2. qRT-PCR Analysis Total RNA was extracted with TRIZOL reagent (Invitrogen) from E13.5 brains, and 1 mg was treated with DNaseI (Takara). cDNA was synthesized with SuperscriptIII (Invitrogen) and random hexamers. Real-time PCR was performed on the ABI prism 7900HT Sequence Detection System (Applied Biosystems) using SYBR-Green and gene-specific primers (Table S3). All analyses were performed in duplicate. Expression levels were normalized to the Gapdh level, which was not significantly different between the WT and Dnmt3b-KO brains. In Situ Hybridization In situ hybridization was performed as described previously (Yokota et al., 2011). Heads of male embryos at E13.5 were embedded in O.C.T. compound

Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc. 105

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

(Sakuma). Fresh frozen sections (10 mm) were cut on a cryostat (Leica), followed by fixation with 4% paraformaldehyde for 10 min at room temperature. The sections were acetylated for 10 min and hybridized with digoxigenin (DIG)labeled antisense probes for gCR (Yokota et al., 2011) at 72 C. To elicit the probe signals, the sections were incubated with alkaline phosphatase-coupled anti-DIG antibodies for 1 hr at room temperature, and the enzymatic signals were visualized with nitroblue tetrazolium (NBT) and 5-bromo-4-chloro-3indolylphosphate. Immunohistochemistry Mice were anesthetized and transcardially perfused with ice-cold 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). The brain was removed and postfixed in the same fixative for 4 hr at 4 C and then transferred to 25% sucrose in phosphate buffer. Sagittal and coronal sections of 100 mm thickness were cut on a vibratome (Leica). For immunostaining, the sections were incubated in blocking buffer (20% Block Ace, 5% normal donkey serum, and 0.1% Triton X-100 in PBS) for 1 hr at room temperature and then in antibody dilution buffer (5% Block Ace, 5% normal donkey serum, and 0.1% Triton X-100 in PBS) containing primary antibody (1:1,000, rabbit anti-GFP; Invitrogen) overnight at 4 C. The sections were rinsed in PBS and incubated for 1 hr at room temperature with antibody dilution buffer containing Alexa-conjugated secondary antibodies. Slices were embedded with ProLong Gold antifade reagent (Life Technologies). Labeling of Purkinje Cells by Biocytin Injection Details are in the Supplemental Experimental Procedures. Image Acquisition and Analysis Images of whole brains or cerebella were obtained using a light microscope (BZ-9000; Keyence) at 43 magnification. Confocal images were obtained using an FV1000 confocal microscope (Olympus), with a 203 dry objective for images of the cerebral cortex and hippocampus, and a 603 oil immersion objective advanced at 1 mm intervals for z stack images of Purkinje cells. Dendrite self-crossings were quantified as the number of branch overlaps detected in a single confocal plane. 3D reconstruction and rendering were performed with Imaris software (Bitplane). The dendrites were traced and analyzed with the Filament tracer function. For the measurements of dendrites, Purkinje cells in lobules IV and V were used for the analysis. For the overpassing dendrites, both those that contacted the branches they crossed and those that did not contact them were analyzed. The spine density was quantified for highly branched dendrites in the regions of interest. Single-Cell Collection The cerebellum of 2- to 3-week-old chimeras was dissected and subjected to enzymatic digestion at 37 C for 30 min in 10 ml of dissociation solution containing 90 units of papain (Worthington), 0.002% DL-cysteine HCl (Sigma), 0.05% DNase I (Sigma), 0.1% bovine serum albumin (Sigma), and 0.05% glucose (Nacalai Tesque). The reaction was stopped by adding 10 ml of fetal bovine serum (GIBCO). The digested tissues were spun for 8 min at 300 3 g, and the pellets were then resuspended in Dulbecco’s modified Eagle’s medium (Sigma). To remove debris, the cells were filtered through a 100 mm cell strainer (Falcon). EGFP-positive or -negative single Purkinje cells were picked up by glass capillary under fluorescence and transferred to 200 ml or 500 ml tubes with 5 ml of 23 CellsDirect reaction mix (Invitrogen). The singlecell samples were immediately frozen and stored at
80 C. High-Throughput Single-Cell Expression Analysis Sequence-specific reverse transcription and preamplification were performed with gene-specific primers and SuperScript III RT Platinum Taq Mix (Invitrogen). The sample and assay mixes were loaded onto a 48.48 Dynamic Array (Fluidigm), and real-time PCR analysis was performed using the BioMark System (Fluidigm). See Tables S4 and S5 and Supplemental Experimental Procedures for details. Statistical Analysis Two groups were compared using an unpaired two-tailed Student’s t test, with equivalent variances determined by F test, or the Mann-Whitney nonpara-

106 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

metric test. To compare more than two groups, a one-way or two-way ANOVA was performed, followed by Bonferroni’s post hoc test to evaluate pairwise group differences. p < 0.05 was considered statistically significant. The analyses were performed by Prism 5.0 software (GraphPad). SUPPLEMENTAL INFORMATION Supplemental Information includes seven figures, five tables, and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi.org/10.1016/j.neuron.2014.02.005. ACKNOWLEDGMENTS We thank Y. Hiraoka, H. Asakawa, and F. Murakami (Osaka University) for the use of the BioMark system and the Imaris software. We also thank members of our laboratories for their support and helpful discussions. This work was supported by a Grant-in-Aid for Scientific Research (S) (JSPS), Innovative Areas ‘‘Mesoscopic Neurocircuitry’’ (number 25115720) (to T.Y.) and ‘‘Neural Diversity and Neocortical Organization’’ (number 25123711) (to T.H.), (Comprehensive Brain Science Network) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan; JST-CREST (to T.Y.); and in part by the Cooperative Study Program of National Institute for Physiological Sciences, Japan. Accepted: January 31, 2014 Published: April 2, 2014 REFERENCES Bird, A. (2002). DNA methylation patterns and epigenetic memory. Genes Dev. 16, 6–21. Borgel, J., Guibert, S., Li, Y., Chiba, H., Schu¨beler, D., Sasaki, H., Forne´, T., and Weber, M. (2010). Targets and dynamics of promoter DNA methylation during early mouse development. Nat. Genet. 42, 1093–1100. Buck, L., and Axel, R. (1991). A novel multigene family may encode odorant receptors: a molecular basis for odor recognition. Cell 65, 175–187. Chen, W.V., and Maniatis, T. (2013). Clustered protocadherins. Development 140, 3297–3302. Chen, W.V., Alvarez, F.J., Lefebvre, J.L., Friedman, B., Nwakeze, C., Geiman, E., Smith, C., Thu, C.A., Tapia, J.C., Tasic, B., et al. (2012). Functional significance of isoform diversification in the protocadherin gamma gene cluster. Neuron 75, 402–409. Chess, A., Simon, I., Cedar, H., and Axel, R. (1994). Allelic inactivation regulates olfactory receptor gene expression. Cell 78, 823–834. Dallosso, A.R., Hancock, A.L., Szemes, M., Moorwood, K., Chilukamarri, L., Tsai, H.H., Sarkar, A., Barasch, J., Vuononvirta, R., Jones, C., et al. (2009). Frequent long-range epigenetic silencing of protocadherin gene clusters on chromosome 5q31 in Wilms’ tumor. PLoS Genet. 5, e1000745. Esumi, S., Kakazu, N., Taguchi, Y., Hirayama, T., Sasaki, A., Hirabayashi, T., Koide, T., Kitsukawa, T., Hamada, S., and Yagi, T. (2005). Monoallelic yet combinatorial expression of variable exons of the protocadherin-alpha gene cluster in single neurons. Nat. Genet. 37, 171–176. Feng, J., Chang, H., Li, E., and Fan, G. (2005). Dynamic expression of de novo DNA methyltransferases Dnmt3a and Dnmt3b in the central nervous system. J. Neurosci. Res. 79, 734–746. Fujishima, K., Horie, R., Mochizuki, A., and Kengaku, M. (2012). Principles of branch dynamics governing shape characteristics of cerebellar Purkinje cell dendrites. Development 139, 3442–3455. Garrett, A.M., Schreiner, D., Lobas, M.A., and Weiner, J.A. (2012). g-protocadherins control cortical dendrite arborization by regulating the activity of a FAK/ PKC/MARCKS signaling pathway. Neuron 74, 269–276. Guo, Y., Monahan, K., Wu, H., Gertz, J., Varley, K.E., Li, W., Myers, R.M., Maniatis, T., and Wu, Q. (2012). CTCF/cohesin-mediated DNA looping is

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

required for protocadherin a promoter choice. Proc. Natl. Acad. Sci. USA 109, 21081–21086.

mediated histone H3 lysine 9 methylation directs DNA methylation to major satellite repeats at pericentric heterochromatin. Curr. Biol. 13, 1192–1200.

Hansen, R.S., Wijmenga, C., Luo, P., Stanek, A.M., Canfield, T.K., Weemaes, C.M., and Gartler, S.M. (1999). The DNMT3B DNA methyltransferase gene is mutated in the ICF immunodeficiency syndrome. Proc. Natl. Acad. Sci. USA 96, 14412–14417.

Li, Y., Lu, H., Cheng, P.L., Ge, S., Xu, H., Shi, S.H., and Dan, Y. (2012). Clonally related visual cortical neurons show similar stimulus feature selectivity. Nature 486, 118–121.

Hasegawa, S., Hamada, S., Kumode, Y., Esumi, S., Katori, S., Fukuda, E., Uchiyama, Y., Hirabayashi, T., Mombaerts, P., and Yagi, T. (2008). The protocadherin-alpha family is involved in axonal coalescence of olfactory sensory neurons into glomeruli of the olfactory bulb in mouse. Mol. Cell. Neurosci. 38, 66–79. Hattori, D., Demir, E., Kim, H.W., Viragh, E., Zipursky, S.L., and Dickson, B.J. (2007). Dscam diversity is essential for neuronal wiring and self-recognition. Nature 449, 223–227. Hirano, K., Kaneko, R., Izawa, T., Kawaguchi, M., Kitsukawa, T., and Yagi, T. (2012). Single-neuron diversity generated by Protocadherin-b cluster in mouse central and peripheral nervous systems. Front. Mol. Neurosci. 5, 90. Hirayama, T., Tarusawa, E., Yoshimura, Y., Galjart, N., and Yagi, T. (2012). CTCF is required for neural development and stochastic expression of clustered Pcdh genes in neurons. Cell Rep. 2, 345–357. Jin, B., Tao, Q., Peng, J., Soo, H.M., Wu, W., Ying, J., Fields, C.R., Delmas, A.L., Liu, X., Qiu, J., and Robertson, K.D. (2008). DNA methyltransferase 3B (DNMT3B) mutations in ICF syndrome lead to altered epigenetic modifications and aberrant expression of genes regulating development, neurogenesis and immune function. Hum. Mol. Genet. 17, 690–709. Jin, B., Ernst, J., Tiedemann, R.L., Xu, H., Sureshchandra, S., Kellis, M., Dalton, S., Liu, C., Choi, J.H., and Robertson, K.D. (2012). Linking DNA methyltransferases to epigenetic marks and nucleosome structure genome-wide in human tumor cells. Cell Rep 2, 1411–1424. Kaneko, R., Kato, H., Kawamura, Y., Esumi, S., Hirayama, T., Hirabayashi, T., and Yagi, T. (2006). Allelic gene regulation of Pcdh-alpha and Pcdh-gamma clusters involving both monoallelic and biallelic expression in single Purkinje cells. J. Biol. Chem. 281, 30551–30560. Kaneko, R., Kawaguchi, M., Toyama, T., Taguchi, Y., and Yagi, T. (2009). Expression levels of Protocadherin-alpha transcripts are decreased by nonsense-mediated mRNA decay with frameshift mutations and by high DNA methylation in their promoter regions. Gene 430, 86–94. Katori, S., Hamada, S., Noguchi, Y., Fukuda, E., Yamamoto, T., Yamamoto, H., Hasegawa, S., and Yagi, T. (2009). Protocadherin-alpha family is required for serotonergic projections to appropriately innervate target brain areas. J. Neurosci. 29, 9137–9147. Kawaguchi, M., Toyama, T., Kaneko, R., Hirayama, T., Kawamura, Y., and Yagi, T. (2008). Relationship between DNA methylation states and transcription of individual isoforms encoded by the protocadherin-alpha gene cluster. J. Biol. Chem. 283, 12064–12075. Kehayova, P., Monahan, K., Chen, W., and Maniatis, T. (2011). Regulatory elements required for the activation and repression of the protocadherin-alpha gene cluster. Proc. Natl. Acad. Sci. USA 108, 17195–17200. Kohmura, N., Senzaki, K., Hamada, S., Kai, N., Yasuda, R., Watanabe, M., Ishii, H., Yasuda, M., Mishina, M., and Yagi, T. (1998). Diversity revealed by a novel family of cadherins expressed in neurons at a synaptic complex. Neuron 20, 1137–1151. Kumaki, Y., Oda, M., and Okano, M. (2008). QUMA: quantification tool for methylation analysis. Nucleic Acids Res. 36 (Web Server issue), W170–W175. Lane, N., Dean, W., Erhardt, S., Hajkova, P., Surani, A., Walter, J., and Reik, W. (2003). Resistance of IAPs to methylation reprogramming may provide a mechanism for epigenetic inheritance in the mouse. Genesis 35, 88–93. Lefebvre, J.L., Kostadinov, D., Chen, W.V., Maniatis, T., and Sanes, J.R. (2012). Protocadherins mediate dendritic self-avoidance in the mammalian nervous system. Nature 488, 517–521. Lehnertz, B., Ueda, Y., Derijck, A.A., Braunschweig, U., Perez-Burgos, L., Kubicek, S., Chen, T., Li, E., Jenuwein, T., and Peters, A.H. (2003). Suv39h-

Lienert, F., Wirbelauer, C., Som, I., Dean, A., Mohn, F., and Schu¨beler, D. (2011). Identification of genetic elements that autonomously determine DNA methylation states. Nat. Genet. 43, 1091–1097. Matthews, B.J., Kim, M.E., Flanagan, J.J., Hattori, D., Clemens, J.C., Zipursky, S.L., and Grueber, W.B. (2007). Dendrite self-avoidance is controlled by Dscam. Cell 129, 593–604. McGowan, P.O., Suderman, M., Sasaki, A., Huang, T.C., Hallett, M., Meaney, M.J., and Szyf, M. (2011). Broad epigenetic signature of maternal care in the brain of adult rats. PLoS ONE 6, e14739. Mikkelsen, T.S., Ku, M., Jaffe, D.B., Issac, B., Lieberman, E., Giannoukos, G., Alvarez, P., Brockman, W., Kim, T.K., Koche, R.P., et al. (2007). Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature 448, 553–560. Monahan, K., Rudnick, N.D., Kehayova, P.D., Pauli, F., Newberry, K.M., Myers, R.M., and Maniatis, T. (2012). Role of CCCTC binding factor (CTCF) and cohesin in the generation of single-cell diversity of protocadherin-a gene expression. Proc. Natl. Acad. Sci. USA 109, 9125–9130. Noguchi, Y., Hirabayashi, T., Katori, S., Kawamura, Y., Sanbo, M., Hirabayashi, M., Kiyonari, H., Nakao, K., Uchimura, A., and Yagi, T. (2009). Total expression and dual gene-regulatory mechanisms maintained in deletions and duplications of the Pcdha cluster. J. Biol. Chem. 284, 32002–32014. Oda, M., Yamagiwa, A., Yamamoto, S., Nakayama, T., Tsumura, A., Sasaki, H., Nakao, K., Li, E., and Okano, M. (2006). DNA methylation regulates longrange gene silencing of an X-linked homeobox gene cluster in a lineage-specific manner. Genes Dev. 20, 3382–3394. Ohtsuki, G., Nishiyama, M., Yoshida, T., Murakami, T., Histed, M., Lois, C., and Ohki, K. (2012). Similarity of visual selectivity among clonally related neurons in visual cortex. Neuron 75, 65–72. Okano, M., Bell, D.W., Haber, D.A., and Li, E. (1999). DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell 99, 247–257. Ribich, S., Tasic, B., and Maniatis, T. (2006). Identification of long-range regulatory elements in the protocadherin-alpha gene cluster. Proc. Natl. Acad. Sci. USA 103, 19719–19724. Schmucker, D., Clemens, J.C., Shu, H., Worby, C.A., Xiao, J., Muda, M., Dixon, J.E., and Zipursky, S.L. (2000). Drosophila Dscam is an axon guidance receptor exhibiting extraordinary molecular diversity. Cell 101, 671–684. Schreiner, D., and Weiner, J.A. (2010). Combinatorial homophilic interaction between gamma-protocadherin multimers greatly expands the molecular diversity of cell adhesion. Proc. Natl. Acad. Sci. USA 107, 14893–14898. Serizawa, S., Miyamichi, K., Nakatani, H., Suzuki, M., Saito, M., Yoshihara, Y., and Sakano, H. (2003). Negative feedback regulation ensures the one receptor-one olfactory neuron rule in mouse. Science 302, 2088–2094. Suo, L., Lu, H., Ying, G., Capecchi, M.R., and Wu, Q. (2012). Protocadherin clusters and cell adhesion kinase regulate dendrite complexity through Rho GTPase. J. Mol. Cell Biol. 4, 362–376. Tasic, B., Nabholz, C.E., Baldwin, K.K., Kim, Y., Rueckert, E.H., Ribich, S.A., Cramer, P., Wu, Q., Axel, R., and Maniatis, T. (2002). Promoter choice determines splice site selection in protocadherin alpha and gamma pre-mRNA splicing. Mol. Cell 10, 21–33. Torii, M., Hashimoto-Torii, K., Levitt, P., and Rakic, P. (2009). Integration of neuronal clones in the radial cortical columns by EphA and ephrin-A signalling. Nature 461, 524–528. Ueda, Y., Okano, M., Williams, C., Chen, T., Georgopoulos, K., and Li, E. (2006). Roles for Dnmt3b in mammalian development: a mouse model for the ICF syndrome. Development 133, 1183–1192.

Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc. 107

Neuron Dnmt3b Is a Regulator for Neuronal Individuality

Wang, X., Su, H., and Bradley, A. (2002a). Molecular mechanisms governing Pcdh-gamma gene expression: evidence for a multiple promoter and cis-alternative splicing model. Genes Dev. 16, 1890–1905. Wang, X., Weiner, J.A., Levi, S., Craig, A.M., Bradley, A., and Sanes, J.R. (2002b). Gamma protocadherins are required for survival of spinal interneurons. Neuron 36, 843–854. Watanabe, D., Uchiyama, K., and Hanaoka, K. (2006). Transition of mouse de novo methyltransferases expression from Dnmt3b to Dnmt3a during neural progenitor cell development. Neuroscience 142, 727–737. Weiner, J.A., Wang, X., Tapia, J.C., and Sanes, J.R. (2005). Gamma protocadherins are required for synaptic development in the spinal cord. Proc. Natl. Acad. Sci. USA 102, 8–14. Wu, Q., and Maniatis, T. (1999). A striking organization of a large family of human neural cadherin-like cell adhesion genes. Cell 97, 779–790. Wu, Q., Zhang, T., Cheng, J.F., Kim, Y., Grimwood, J., Schmutz, J., Dickson, M., Noonan, J.P., Zhang, M.Q., Myers, R.M., and Maniatis, T. (2001). Comparative DNA sequence analysis of mouse and human protocadherin gene clusters. Genome Res. 11, 389–404.

108 Neuron 82, 94–108, April 2, 2014 ª2014 Elsevier Inc.

Xu, G.L., Bestor, T.H., Bourc’his, D., Hsieh, C.L., Tommerup, N., Bugge, M., Hulten, M., Qu, X., Russo, J.J., and Viegas-Pe´quignot, E. (1999). Chromosome instability and immunodeficiency syndrome caused by mutations in a DNA methyltransferase gene. Nature 402, 187–191. Xu, Y., Wu, F., Tan, L., Kong, L., Xiong, L., Deng, J., Barbera, A.J., Zheng, L., Zhang, H., Huang, S., et al. (2011). Genome-wide regulation of 5hmC, 5mC, and gene expression by Tet1 hydroxylase in mouse embryonic stem cells. Mol. Cell 42, 451–464. Yagi, T. (2013). Genetic basis of neuronal individuality in the mammalian brain. J. Neurogenet. 27, 97–105. Yokota, S., Hirayama, T., Hirano, K., Kaneko, R., Toyoda, S., Kawamura, Y., Hirabayashi, M., Hirabayashi, T., and Yagi, T. (2011). Identification of the cluster control region for the protocadherin-beta genes located beyond the protocadherin-gamma cluster. J. Biol. Chem. 286, 31885–31895. Yu, Y.C., Bultje, R.S., Wang, X., and Shi, S.H. (2009). Specific synapses develop preferentially among sister excitatory neurons in the neocortex. Nature 458, 501–504. Zipursky, S.L., and Sanes, J.R. (2010). Chemoaffinity revisited: dscams, protocadherins, and neural circuit assembly. Cell 143, 343–353.