CRISPR/Cas9 interrogation of the mouse Pcdhg gene cluster reveals a crucial isoform-specific role for Pcdhgc4

The mammalian Pcdhg gene cluster encodes a family of 22 cell adhesion molecules, the gamma-Protocadherins (γ-Pcdhs), critical for neuronal survival and neural circuit formation. The extent to which isoform diversity–a γ-Pcdh hallmark–is required for their functions remains unclear. We used a CRISPR/Cas9 approach to reduce isoform diversity, targeting each Pcdhg variable exon with pooled sgRNAs to generate an allelic series of 26 mouse lines with 1 to 21 isoforms disrupted via discrete indels at guide sites and/or larger deletions/rearrangements. Analysis of 5 mutant lines indicates that postnatal viability and neuronal survival do not require isoform diversity. Surprisingly, given reports that it might not independently engage in trans-interactions, we find that γC4, encoded by Pcdhgc4, is the only critical isoform. Because the human orthologue is the only PCDHG gene constrained in humans, our results indicate a conserved γC4 function that likely involves distinct molecular mechanisms.


INTRODUCTION 27
Cell-cell recognition via transmembrane cell adhesion molecules is essential for neural circuit 28 formation. With trillions of exquisitely specific synapses in the human brain, it has been suggested that 29 molecular diversity-achieved either via alternative gene splicing or combinatorial expression of large 30 adhesion molecule families-plays an important role (Zipursky and Sanes, 2010). In Drosophila, this is 31 exemplified by Dscam1, a single gene capable of generating 38,016 distinct protein isoforms through 32 alternative splicing (Schmucker et al., 2000). Dscam1 isoform diversity is essential for 33 neurodevelopmental processes including axon guidance, synapse specificity, and neurite self-avoidance 34 (Hattori et al., 2009(Hattori et al., , 2007Zhan et al., 2004, reviewed in Hattori et al., 2008. Mammalian Dscams, 35 despite important roles in neurodevelopment, do not generate such isoform diversity (Fuerst et al., 36 2009(Fuerst et al., 36 , 2008Garrett et al., 2018Garrett et al., , 2016. In this sense, the clustered protocadherins (cPcdhs) are the 37 diversity per se may be required. Second, there may be a high level of isoform redundancy such that any 86 one (or few) isoform(s) may suffice. Third, there may be unique roles for individual isoforms, such that 87 no other isoform can compensate for their loss. Some extant evidence exists for each of these 88 possibilities. The γC3-5 isoforms, but not γA1-3, were required for postnatal viability in mice (Chen et al., 89 2012). As γC5 is expressed later in the postnatal period (Frank et al., 2005), and γC4 is unable to localize Thus, no single model is likely to encompass all of the γ-Pcdhs' diverse functions. Additionally, 106 most prior studies have relied on mis-or over-expression of individual isoforms, which could in some 107 cases result in new, distinct phenotypes. Paralleling studies establishing the necessity of Dscam1 108 molecular diversity for specific neurodevelopmental roles in Drosophila, we used CRISPR/Cas9 genome 109 editing to simultaneously target the 22 Pcdhg variable exons in an unbiased manner, and created a new 110 allelic series of mouse mutants with reduced isoform diversity from the endogenous gene cluster. We 111 find, surprisingly, that only one isoform -γC4, which uniquely cannot mediate homophilic trans-112 interactions independently -is strictly required for postnatal viability and survival of the many neuronal 113 subsets shown previously to depend on the γ-Pcdhs. Our results: 1) show that some γ-Pcdh functions do 114 not require molecular diversity; 2) confirm that at least some γ-Pcdh isoforms have unique roles; and 3) 115 suggest that the regulation of neuronal survival may require novel mechanisms of cPcdh interaction 116 and/or signaling involving γC4. Pooled sgRNAs were concentrated, then combined with Cas9 mRNA and microinjected into C57BL/6J 131 mouse zygotes at three different concentrations (each individual guide at 50 ng sgRNA/µl, 10 ng 132 sgRNA/µl, or 5 ng sgRNA/µl: HI, MED, and LO, respectively), as described in Materials and Methods 133 ( Figure 1D). Microinjected zygotes were transferred into a total of 20 pseudopregnant females, resulting 134 in 100 live-born mice (16 from HI, 34 from MED, and 50 from LO). All 100 founders were screened at 7 135 Pcdhg exons by PCR with Sanger sequencing to detect frame-shifting mutations. From this initial screen, 136 15 founders exhibited some disruption (10 from HI and 5 from MED) and were designated for breeding. 137 Most disruptions were found in pups from HI or MED injections, so the 50 founders resulting from the 138 LO injections were not pursued further. The remaining 35 founders were further screened by PCR and 139 Sanger sequencing at the remaining 15 Pcdhg variable exons. In this way, a total of 31 founders were 140 identified that carried some constellation of mutations at the guide-targeted sites (14/16 HI, 17/34 141 MED). 142 Due to the likely mosaicism of the founders and the uncertainty of germline transmission of any 143 given mutation, we did not characterize the 31 identified founders more extensively. Rather, each was 144 crossed with wild-type C57BL/6J animals to generate G1 offspring for further analysis ( Figure 1D). Sperm 145 from male G1 mice was cryopreserved while somatic tissue was used for genotyping to identify lines of 146 interest carrying reduced diversity of Pcdhg variable exons. Ninety-four G1 offspring were screened for 147 heterozygous mutations using a custom amplicon assay from Illumina and Illumina MiSeq sequencing 148  Twenty-six distinct lines carrying unique constellations of mutations are represented in Table 1, derived  156 from 12 different founders. Most mouse lines (20 lines from 9 founders) resulted from microinjection of 157 from Anp32a which aligns to exons 4-7 of transcript Anp32a-201 without the intervening introns ( Figure  182 2B, Figure 1-figure supplement 1D). This phenomenon of the insertion of a transposable element along 183 with coding sequence from an early expressed gene has been previously described in CRISPR genome 184 editing (Ono et al., 2015). As this sequence was inserted 3' to the inverted exon Pcdhga5, there is no 185 associated transcription start site, and no protein product is expected. Altogether, at least 9 double-186 stranded breaks occurred, resulting in a frame-shifting 52 bp insertion into exon B2, and larger deletions 187 and rearrangements disrupting all the other γA and γB isoforms. As only the 3 γC V exons remained 188 intact in Pcdhg em12 , we renamed this allele Pcdhg 3R1 (3R1 hereafter; Figure 2B). 189 Whole genome sequencing from Pcdhg em35 mutants confirmed small frame-shifting deletions at 190 exons A12 and C5, as well as an in-frame deletion of 9 bp in exon C4. However, this more exhaustive 191 sequencing also revealed a large rearrangement undetected by the prior amplicon sequencing analysis. 192 Here, there was a ~94 kb deletion spanning the breakpoint from exon A1 to that of exon A11 (Figure 2-193 figure supplement 1B). Only exons B8 and C3 were unaffected by any mutation. As exon C4 still encoded 194 a nearly full-length protein lacking only 3 amino acids (residues 27-29 in the signal peptide), we renamed 195 the Pcdhg em35 allele Pcdhg 3R2 (3R2 hereafter, Figure 2C), as it represented the second allele identified 196 with 3 isoforms remaining. 197 In all three mutants analyzed thus far, rearrangements were identified by whole genome 198 sequencing that were undetected by the amplicon analysis ( Figure 2, Table 1 -Table supplement 1). 199 With this information, we re-analyzed the amplicon sequencing data by visual inspection of the paired 200 reads within the Integrative Genomics Viewer (IGV). We were able to find many, but not all, of the 201 junctions identified by the whole genome sequencing, but missed by BreaKmer analysis of the amplicon 202 sequencing (Figure 1-figure supplement 1B). Therefore, we manually inspected the alignments from 203 each of the other frozen lines. Additional rearrangements were identified where each of the paired ends 204 of multiple reads mapped to different exons (Table 1 -Table supplement  To exclude this possibility, we performed quantitative real-time PCR using cDNA from the 219 cerebral cortices of homozygous mutants and wild-type littermates ( Figure 3A). To detect specific 220 isoform transcripts, forward primers targeting the 3' end of each V exon were used with a reverse 221 primer in constant exon 1 (product spans 1 intron) or exon 2 (product spans 2 introns). To monitor total 222 levels of Pcdhg locus transcription, a forward primer in constant exon1 was used with a reverse primer 223 in constant exon 2 (spanning 1 intron, primer sequences in Figure 3 -figure supplement 2). Total locus 224 expression was significantly reduced in 3R2 homozygous mutants, but in both 3R1 and 13R1 mutant 225 cortex, expression levels were indistinguishable from controls. Furthermore, individual isoform 226 transcription was reduced only when mutations completely disrupted the exon by deletion or inversion. 227 Smaller indels generally had no effect on transcript expression level (e.g., A1, B1, A5, A7, B7, and C4 228 were not significantly reduced in 13R1 homozygous mutants, but A9, B6, and A10 were undetectable). 229 Additionally, transcription of several V exon fusions was detected, including A4 (fusion with A1) and A9 230 (fusion with A12) in 3R1 (predicted protein products encoded by these fused transcripts are listed in 231 Supplementary File 2). We also asked the extent to which mutations within the Pcdhg locus altered 232 isoform transcription from the Pcdha or Pcdhb clusters. While there were no significant changes 233 detected in 13R1 mutants, isoforms from the 3' end of the Pcdhb locus were expressed at significantly 234 higher levels in 3R1 and 3R2 homozygous mutants compared to controls (β11 in 3R1, β15 and β22 in 235 both 3R1 and 3R2). Expression of Pcdha cluster genes appeared to be unchanged (Figure 3-figure  236 supplement 1). 237 We also verified that these mRNA expression levels were reflected at the protein level, utilizing 238 a series of antibodies specific for particular γ-Pcdh protein isoforms (Lobas et al., 2012). Western blot 239 analysis of brain lysates from 13R1 neonates and 3R1 and 3R2 adults confirmed the presence of the 240 expected isoforms at the appropriate molecular weights: 13R1 brains expressed γA isoforms, γB2, and 241 γC3, but not γC4, while 3R1 and 3R2 brains expressed γC3 and γC4, but not γB2 or any γA isoforms 242 ( Figure 3B). Based on these analyses, we concluded that 13R1 homozygous mutants are not, in fact, 243 complete Pcdhg functional nulls, and that the similarity of their neonatally lethal phenotype to that of 244 nulls likely reflects the essential nature of a particular isoform lost in this line but present in both 3R1 245 and 3R2; that is, γC4, as described below. Before testing this conclusion in detail, we first asked whether 246 the neonatal lethality of 13R1 was accompanied by cellular phenotypes previously described in 247 Pcdhg del/del null mice, and whether the viable 3R1 and 3R2 lines lacked these phenotypes. 248 Excessive developmental neuronal apoptosis occurs in mutants exhibiting neonatal lethality 249  Having confirmed that the cellular phenotypes of 13R1 resemble the complete deletion of the 291 Pcdhg cluster, we next turned to utilizing the novel CRISPR mutant mouse lines to confirm which γ-Pcdh 292 isoforms were critical. We noted that homozygous 3R1 mutants survived and did not exhibit 293 exacerbated apoptosis despite lacking expression of any functional γA or γB isoforms, indicating that one 294 or more of the C3-C5 isoforms must be critical. Furthermore, 3R2 mutants survived and were 295 phenotypically normal without a functional Pcdhgc5 gene, whilst 13R1 homozygous mutants, which 296 exhibited neonatal lethality and exacerbated neuronal apoptosis uniquely harbored frame-shifting 297 mutations in PcdhgC4. In a separate study, we have derived and are analyzing a CRISPR-targeted mouse 298 line that specifically generated a Pcdhgc3 loss-of-function allele, and have found that they are viable and suggests that Pcdhgc4 encodes the sole cPcdh isoform essential for organismal survival. We generated 303 two additional mouse lines to confirm this conclusion. 304 First, we chose Pcdhg em8 from our list of mutants for cryorecovery (Table 1). This line harbored a 305 large deletion from within exon A1 to within exon C3 identified by visual inspection of the aligned reads 306 from the original amplicon sequencing, as well as a 1 bp frame-shifting deletion in exon C5. Upon 307 cryorecovery, we generated homozygous mutants and verified these mutations by linked-read whole 308 genome sequencing ( Figure Figure 7C). Together, these two additional mouse lines show that mutation of Pcdhgc4 alone 321 recapitulated the overt phenotype of losing the entire cluster, whilst expression of this single γ-Pcdh 322 isoform was sufficient to rescue viability, even when all 21 other Pcdhg isoforms were absent. 323

Pcdhgc4 is the crucial isoform for neuronal survival 324
To ask if these overt phenotypes of death vs. survival in C4KO and 1R1 mutants extended to 325 neuronal cell survival, we again analyzed spinal cords from P0 neonatal mutants. As expected from their 326 outward appearance, 1R1 homozygous spinal cords appeared grossly normal in overall size and neuronal 327 density (NeuN+ cells), with little if any reactive gliosis (GFAP+; Figure 8B is strictly required for a given function. Here, we used a CRISPR/Cas9 strategy to reduce γ-Pcdh isoform 371 diversity in an unbiased fashion, creating a new allelic series of mouse strains. In the course of analyzing 372 several new Pcdhg alleles, we discovered that the control of neuronal survival and postnatal viability is 373 best described by the third model. We found-surprisingly, given its inability to reach the cell surface and 374 mediate homophilic adhesion without other cPcdh cis-interaction partners-that γC4 is the sole 375 necessary and sufficient isoform. 376 Our strategy was to inject Cas9 mRNA along with 20 individual sgRNAs targeting each of the 377 Pcdhg V exons simultaneously. By screening G1 offspring from many founders, we collected an array of 378 mutants with distinct V exon mutation patterns, ranging from a single isoform disrupted (21 intact) to 21 379 isoforms disrupted (1 intact), all of which were cryopreserved. Each isoform was disrupted in at least 380 one mutant line (Table 1). Furthermore, there was evidence of double-stranded breaks at each sgRNA 381 site identifiable as indels or junctions between guide sites. Our G1 screening was done by next 382 generation sequencing of a custom amplicon array. This technique was effective for identifying indels 383 within each individual amplicon (e.g., smaller indels at each guide site), but was not exhaustive in 384 identifying rearrangements between guide sites, even when both sides of the junction were predicted to 385 be recognized by the amplicon array primers. This could be due to the new DNA sequence: 1) being 386 difficult to amplify within the array, and therefore not sequenced; 2) not being accurately aligned to the 387 reference genome; or 3) not being identified as a new junction by the bioinformatic algorithm 388 BreaKmer. Indeed, some junctions that were not recognized by BreaKmer were clearly identified by 389 visual inspection of the sequence alignments within IGV as paired reads mapping to two different exons 390 ( within a relatively small region should use other methods to identify rearrangements between guide 395 sites, such as targeted long-read sequencing (Bennett-Baker and Mueller, 2017). 396 One concern when using a pooled sgRNA approach such as ours is the compounding of potential 397 off-target mutations. To assess this, we sequenced the top 95 predicted off-target sites in G1 offspring 398 by amplicon sequencing. We did not identify indels attributable to Cas9 activity at any of these sites. 399 While not exhaustive, this analysis is consistent with recent studies suggesting that off-target mutations 400 from CRISPR/Cas9 are relatively rare in mouse model production (Iyer et  Thus, while the downstream mechanisms through which γ-Pcdhs promote neuronal survival 467 remain unknown, our findings suggest that either protein interactions specifically mediated by γC4, or a 468 unique localization for this isoform, will be involved. In 1R1 mutants, γC4 alone was not entirely 469 sufficient for normal neuron number in the spinal cord or retina (Figure 8). This could be explained 470 potentially by the significantly reduced expression levels of γC4 (and thus of total γ-Pcdh proteins) in 471 these mutants ( Figure 7C Table 2). 495

Generation of Pcdhg reduced diversity mutants 496
Guide RNA (sgRNA) sequences were designed to target the near 5' regions to the start codons of 497 each variable exon (Table 3). Guides were designed using the tool at crispr.mit.edu and were analyzed 498 using RGEN tools to minimize off target sites and to maximize the likelihood of frameshifting mutations 499 . 20 total guides were synthesized (IDT), as exons B4 and B5 500 had a high level of 5' homology, as did exons B6 and B7. Guides were received lyophilized, resuspended 501 in water, mixed in equal parts, then lyophilized again. This mixture was diluted and resuspended with S. 502 pyogenes Cas9 mRNA to generate three guide concentrations: 50 ng/µl for each guide, 10 ng/µl for each 503 guide, or 5 ng/µl for each guide. In each dilution, Cas9 mRNA was present at a concentration of 100 504 ng/µl. These mixtures were microinjected into C57BL/6J zygotes, which were subsequently implanted 505 into pseudopregnant female C57BL/6J mice. 506 The resulting live founders were screened by PCR and Sanger sequencing (primers in Table 4). 507 There were three iterations of screening. First, we chose 7 exons distributed across the locus for PCR 508 amplification and Sanger sequencing in 100 live founders. Heterozygous and homozygous indels were 509 identified by analyzing the sequence traces, and 15 founders were found to harbor some mutation. 50 510 of the 100 founders were from the lowest concentration of sgRNA injection (5 ng/µl). No mutations 511 were found in the first round of screening in these animals, and they were not analyzed further. In the 512 second round of screening, the 35 remaining founders were analyzed for indels at the other 15 exons. In 513 the third round, they were screened for rearrangements between exons by PCR with single forward 514 primers mixed with pooled reverse primers. PCR products detected above background level were 515 purified and sequenced using the forward primer. Mice carrying any mutation detected in any round of 516 screening were bred for one generation with C57BL/6J animals. The resulting G1 offspring were 517 screened by amplicon sequencing. Males were prioritized when available for ease of sperm 518 cryopreservation. 519 To generate Pcdhg C4KO mice, only sgRNA complimentary to exon C4 was microinjected (50 ng/µl, 520 Table 3) along with Cas9 mRNA (100 ng/µl). Of the resulting 16 live founders, 9 contained mosaic indels. 2018), we focused on indels that exceeded the following thresholds: QUAL ≥ 850; DP ≥ 90; QD ≥ 5. We 540 also filtered out indels detected in the wild type C57BL/6J control, as these reflect discrepancies from 541 the reference genome. Subsequent whole genome sequencing confirmed these mutations in the Pcdhg 542 locus, but not potential off-target mutations expected to be linked (e.g., Pcdhb7). Furthermore, one 543 mutation that did not exceed our threshold was detected by whole genome sequencing (A12 in 544 Pcdhg em35 ). Therefore, Table 1 -Table supplement 1 summarizes mutations detected in the Pcdhg locus 545 above and below the threshold, as indicated. 546 Because of the pooled amplicon reaction, we reasoned that large rearrangements between 547 guide sites could be detected by amplicon sequencing if enough sequence remained on either side of 548 the junction. To find these junctions we used BreaKmer (Abo et al., 2015). We were able to detect some 549 of these junctions with BreaKmer, but subsequent whole genome sequencing revealed rearrangements 550 missed by BreaKmer in each sequenced mouse line. In some cases, this could be because deletions 551 extended beyond the amplicon site on one or both sides of the junction. However, many junctions were 552 flanked by enough amplicon sequence to expect them to be detectible in this data set. Indeed, by 553 visually analyzing the read alignments in the Integrative Genomics Viewer (IGV, Broad Institute) we were 554 able to find some of these junctions when mating paired-end reads mapped to different exons (Figure 1-555 figure supplement 1B). We used this approach to identify additional junctions in the cryopreserved 556 mouse lines as indicated in Table 1 -Table supplement 1

. 557
Whole genome sequencing 558 Genomic DNA was isolated from homozygous mutants using the DNA Extraction from Fresh Frozen 559 Tissue protocol (10X Genomics). Briefly, nuclei were isolated from tissue, lysed using proteinase K, then 560 DNA purified using magnetic beads. Linked-read whole genome libraries were constructed using the 561 Genome Chip Kit v2 (10X Genomics). Briefly, high molecular weight DNA was partitioned into Gel Bead-562 In-EMulsions (GEMs) where unique barcoded primers were added to individual molecules of DNA. After 563 the GEMs were dissolved, Illumina specific sequencing primers and barcodes were added by isothermal 564 amplification, then library construction was completed via end repair, a-tailing, adapter ligation, and 565 amplification. Libraries were sequenced on a HiSeq X (Illumina) by Novogene (Sacramenta, CA) with 150 566 bp paired-end reads. Sequences were aligned using the Long Ranger analysis pipeline from 10X 567 Chromium and visualized in Loupe (10X Genomics) or IGV (Broad Institute). Paired-end reads were used 568 to find NHEJ junctions between guide sites, while reads spanning the junctions were analyzed to 569 uncover the specific sequences of these junctions. 570

Data deposition 571
All Illumina sequencing data is available at the Gene Expression Omnibus (GEO) with the 572 accession number XXXXX (in progress) or the Sequence Read Archive (SRA) with number XXXX. 573 Quantitative RT-PCR 574 RNA was isolated from cerebral cortex of animals at P0 (13R1 mutants and littermate controls) 575 or 2-12 weeks of age (other mutants and littermate controls) using Trizol reagent. Five µg of RNA per 576 sample was used to make cDNA using Superscript III (Invitrogen) according to the manufacturers 577 protocol. qPCR was performed in triplicate (technical replicates) using primers listed in Figure 3 - Figure  578 supplement 2 with SYBR Green PCR Master Mix. The relative abundance of each transcript was 579 calculated using the ∆∆Ct method, normalized to GAPDH and littermate controls. The relative levels of 580 each transcript in controls, 3R1, and 3R2 mutants were compared using an ANOVA with Tukey post-hoc 581 tests. 13R1 and 1R1 mutants were compared with littermate controls for each transcript using a t-test. 582

Western blotting 583
Whole brains at P0 or adult ages (3-8 months) were homogenized in RIPA buffer (0.1% SDS, 584 0.25% sodium deoxycholate, 1% NP-40, 150 mM NaCl, 50 mM Tris-HCl, pH 7.4, 5 mM NaF) plus protease 585 inhibitors (Roche Mini cOmplete) using a Dounce homogenizer and a Wheaton overhead stirrer. The 586 lysate was centrifuged at 16,000 X g for 15 min at 4 o C to remove cell debris and proteins were quantified 587 using a BCA assay kit (Pierce/Thermo Scientific). Thirty or 40 µg of protein was resolved on Mini-588 PROTEAN TGX precast 12% SDS-PAGE gels (Bio-Rad) and proteins were transferred to a nitrocellulose 589 membrane using the Bio-Rad Trans-Blot Turbo Transfer System. Membranes were blocked using 10% 590 skim milk for 1 hour and incubated with the primary antibody in buffer (2.5% BSA in TBS-T (Tris buffered 591 saline with 0.1% Tween-20) overnight at 4 o C. The following day, membranes were washed using TBS -T 592 and incubated in HRP-conjugated secondary antibody for 1 hour. Membranes were washed, developed 593 using the SuperSignal West Pico ECL reagents (Thermo Scientific), and imaged using a Li-Cor Odyssey Fc 594 Imaging system. 595 Immunofluorescence 596 Tissues were processed and stained as described previously (Garrett et al., 2016;Prasad et al., 597 2008). Briefly, neonatal spinal columns were removed and fixed by immersion in 4% paraformaldehyde 598 (PFA) for four hours at 4° C, followed by extensive washing in PBS and cryopreservation in 30% sucrose. 599 Eyes were enucleated and dissected to remove the cornea and lens, then fixed overnight by immersion 600 in 4% PFA at 4° C followed by cryopreservation in 30% sucrose. Tissue was embedded in OCT (Sakura-601 Finetek) and sectioned with a cryostat onto positively charged Superfrost Plus slides (Fisher Scientific). 602 After blocking in 2.5% bovine serum albumin with 0.1% Triton-X-100 in PBS, primary antibodies were 603 incubated on the slides overnight at 4° C in a humidified chamber, followed by secondary antibodies for 604 1 hour at room temperature. Whole mount retinas were stained free floating in primary antibody 605 diluted in blocking solution with 0.5% Triton-X-100 for 48-72 hours, and in secondary antibody for 24 606 hours. Sections were counter-stained with DAPI (4',6-diamidino-2-phenylindole), prior to mounting with 607 Fluoro-Gel mounting media (Electron Microscopy Services #17985-11). center and periphery (technical replicates), were averaged. These averaged values from at least six 635 retinas per genotype (biological replicates) were compared using an ANOVA with Tukey post-hoc tests. 636 Sample size was based on prior studies, as effective group sizes were known. Retinas were analyzed at 2 637 weeks of age, with the exception of 1R1 mutants and littermate controls, which were analyzed at 2 638 weeks and adult (3-6 months of age). Here, values were compared using a two-way ANOVA with age and 639 genotype as independent variables, followed by Tukey pairwise comparisons. We would like to thank the scientific services at the Jackson Laboratory for assistance throughout this 648 project, including Genetic Engineering Technologies, Microinjection, and Reproductive Sciences services 649 for the production and preservation of new mutants, the Genome Technologies service for sequencing, 650 and the Bioinformatics service for data analysis. We would also like to thank Kate Miers for assistance 651 with the mouse colony. there was a small frame-shifting insertion in exon B2 along with multiple deletions and rearrangements 957 between exons. For example, genomic DNA between the exon A5 and exon A6 guide sites was inverted 958 and inserted at the B6 guide site, followed by coding sequence from the gene Anp32a (purple box) and a 959 transposable element (green box). Only exons C3, C4, and C5 remain intact. C) In 3R2 mutants, a single 960 large deletion resulted in a fusion between guide sites at exon A1 and exon A11. cerebral cortex cDNA from 13R1 mutants (red), 3R1 mutants (blue) and 3R2 mutants (green) verified 967 that intact isoforms were expressed at expected levels. Expression of constant exons was significantly 968 reduced in 3R2 mutants only. 13R1 mutants were analyzed at P0, while 3R1 and 3R2 animals were 969 analyzed at P14, each with littermate controls (white). * = p < 0.05; ** = p < 0.01; *** = p < 0.001 by 970 Tukey post-hoc test comparing the indicated genotype with wild type. n=3-9 animals per genotype. Box 971 plots represent the median, first and third quartile, range, and outliers. B) Western blots of brain lysates 972 from 13R1 (at P0), 3R1, and 3R2 mutants (at 1 month of age) with littermate controls confirm the 973  analyzed as in Figure 4 (E-F, FoxP2 and G-H, Pax2). K-L) Ventral interneurons were drastically reduced in 1031 C4KO mutants; decreases in 1R1 mutants were statistically significant, but substantially more modest. I-1032 J) Apoptosis was increased in C4KO animals, but not in 1R1 mutants, as demonstrated by CC3 1033     Pcdhb isoforms at the 3' end of the cluster was increased in 1R1 mutants, consistent with the effect 1114 from large deletions in 3R1 and 3R2 mutants. B) Expression of the Pcdhg cluster reflected genomic 1115 mutations, including expression from the γA1-γC3 fusion. γC5 isoform expression was significantly 1116 reduced (disrupted by 1 base pair insertion), and total γ-constant expression was reduced by half. * = p < 1117 0.05; ** = p < 0.01; *** = p < 0.001 by student's t-test. n=3 animals per genotype. Box plots represent 1118 the median, first and third quartile, range, and outliers. 1119