Isoform requirement of clustered protocadherin for preventing neuronal apoptosis and neonatal lethality

Summary Clustered protocadherin is a family of cell-surface recognition molecules implicated in neuronal connectivity that has a diverse isoform repertoire and homophilic binding specificity. Mice have 58 isoforms, encoded by Pcdhα, β, and γ gene clusters, and mutant mice lacking all isoforms died after birth, displaying massive neuronal apoptosis and synapse loss. The current hypothesis is that the three specific γC-type isoforms, especially γC4, are essential for the phenotype, raising the question about the necessity of isoform diversity. We generated TC mutant mice that expressed the three γC-type isoforms but lacked all the other 55 isoforms. The TC mutants died immediately after birth, showing massive neuronal death, and γC3 or γC4 expression did not prevent apoptosis. Restoring the α- and β-clusters with the three γC alleles rescued the phenotype, suggesting that along with the three γC-type isoforms, other isoforms are also required for the survival of neurons and individual mice.


INTRODUCTION
The rule governing the connectivity of a neural circuit is pivotal in constructing the brain, and the diversity of cell-surface recognition molecules has been implicated in the regulation of this connectivity. The nervous systems of insects and vertebrates have independently evolved different types of cell-surface recognition molecules with extraordinarily diverse isoforms, namely, Dscam1 in insects 1 and clustered protocadherins in vertebrates (cPcdh). [2][3][4] This suggests that the utilization of the isoform diversity, not the protein species itself, is essential in constructing the brain. Mice have 58 cPcdh isoforms encoded by three gene clusters, namely, Pcdha (14 isoforms), Pcdhb (22 isoforms), and Pcdhg (22 isoforms). 5,6 Individual neurons express a distinct combination of cPcdh isoform subsets in a stochastic manner. 7-10 cPcdh proteins form cis-dimers promiscuously, with preferences for heterologous dimers with other isoforms that increase the variety of recognition units. [11][12][13] cPcdh isoforms then interact strictly homophilically in trans at the cell surface of opposing neurons, such as at synapses, creating interaction specificity. [12][13][14][15][16][17] Therefore, cPcdh can create cell-surface identity for cell recognition, which leads to the hypothesis that cPcdh works as a synaptic partner-selection molecule. However, this has not been proven at the synapse level yet.
Gene knockout studies targeting each of the three cPcdh clusters have shown that cPcdh plays a role in multiple aspects of recognition events, including axonal projection, dendritic self-avoidance, dendritic arbor complexity, and synapse formation. [18][19][20][21][22][23][24][25][26][27][28][29][30][31] Mice lacking all 58 isoforms (Dabg mice) exhibit the most severe phenotype. They die immediately after birth due to massive neuronal death and synaptic loss in the brainstem and spinal cord. Genetically blocking apoptosis in Dabg mice by deleting the Bax gene cannot rescue neonatal lethality, synaptic defects, or neural circuit malfunction, suggesting that Dabg mice have an abnormally wired neural network. 32 A similar neonatal lethal phenotype is also observed in Dbg and Dg mice, whereas Da, Db, and Dab mice can survive, suggesting that Pcdhg plays a dominant role in neural network formation. However, since the phenotypic severity of neuronal death and synaptic loss increases with the number of deleted clusters (Dg<Dbg<Dabg), all three Pcdh clusters may cooperatively contribute to neuronal survival and functional neural circuit formation. 22 Among the 58 isoforms, the last two isoforms in the Pcdha cluster and the last three isoforms in the Pcdhg cluster are distinctly categorized as C-type isoforms based on sequence homology. 4 The C-type isoforms are more ubiquitously expressed (although not in all neurons) compared with the other 53 variable isoforms that are stochastically and combinatorially expressed in individual neurons. [8][9][10]18,28 Interestingly, the triple gC-type isoform knockout (TCKO), which lacks gC3, gC4, and gC5 isoforms, exhibits a phenotype similar to the Pcdhg null mutant, whereas the triple gA-type isoform knockout, which lacks gA1, gA2, and gA3 isoforms, shows no discernible abnormalities. 33,34 Subsequently, it was shown that gC4 is the only responsible and sufficient isoform for the survival of both neurons and individual mice. 35 CRISPR/Cas9-mediated disruption of gC4 caused neuronal apoptosis and neonatal lethality in mice, whereas the disruption of all the other g isoforms except gC4 resulted in a grossly normal phenotype. 35 This raised the question about the role and the necessity of the other 55 isoforms. Thus, in this study, we aimed to generate mutant mice that only express the three gC-type isoforms (TC for triple gC-type) but lacked all the other 55 isoforms including Pcdh a and b.
TC mutants died immediately following birth and exhibited massive apoptotic cell death in a specific brain region of the basal forebrain and in the large area of the brainstem that contains the reticular formation. On the other hand, ab/TC compound mutants that carry the TC allele complemented with a and b alleles survived to adulthood, indicating that the three gC-type isoforms in the TC allele were functional for individual survival. In situ hybridization (ISH) showed that the wild-type brain regions where apoptotic cell death occurred in TC mutants expressed both gC3 and gC4 isoforms and the stochastic isoforms from the Pcdha, b, and g clusters in combination. This result suggests that the essential gC4 isoform is always expressed with the other stochastic isoforms, and this combination is necessary for the survival of neurons and individual mice.

RESULTS
TC mutant neonates die at birth despite the expression of the three gC-type isoforms A genetically modified TC (triple gC-type) mutant mouse, which lacked 55 cPcdh isoforms from a1 to gA12 but retained only the three C-type isoforms of Pcdhg, was generated ( Figure 1). Briefly, we introduced loxP sites upstream of a1 (loxP-a1MV) 22 and downstream of gA12 ( Figure 1A; loxP-gA12/C3) by homologous recombination of targeting vectors ( Figure S1). A deletion allele lacking the 55 isoforms from a1 to gA12 was generated by Cre-induced meiotic recombination by crossing with mice carrying the Sycp-Cre transgene (Figures 1A and 1B). This deletion protocol also deleted the essential gene Taf7 located between the Pcdhb-and g-clusters, the loss of which is known to be early embryonic lethal. 36 To restore the additionally deleted Taf7 gene, TC mutant mice were crossed with transgenic mice that harbor the Taf7 transgene (TG taf7 ). 22 All mice used in this study, including the control (+/+:TG taf7 ) and mutants (TC, Dg, and Dabg), carry the Taf7 transgene.
We initially examined the expression of the retained triple gC-type isoforms in TC mutants by reverse transcription-polymerase chain reaction (RT-PCR) ( Figure 1C). We confirmed the expression of the gC isoforms and that the other isoforms in the deleted region (aCR, b3, b21, gA3, gB7) were not expressed in TC mutants ( Figure 1C). Since the deletion of the genomic region 5 0 upstream of the gC-type isoforms or the absence of the other 55 isoforms may affect and change the expression of the three gC-type isoforms, we conducted quantitative real-time PCR. The expression of the gC-type isoforms in TC mutants was comparable with that in control (+/+:TG taf7 ); no significant change in gC4 and gC5 was observed, whereas higher expression was noted in gC3 (Student's t test, p < 0.05; Figure 1D). We also examined the spatial expression of gC3 and gC4 mRNA by in situ hybridization (ISH). As shown in Figures 1F-1I, spatial expression patterns of gC3 and gC4 in control mice were retained and did not change in TC mutants. Both isoforms were widely expressed in the brain with more prominent expression of gC3 in the cerebral cortex and higher expression of gC4 in the thalamic region ( Figures 1F-1I). The expression of the Pcdhg protein was probed with the antibody against the constant region of Pcdhg. The antibody detected all Pcdhg protein isoforms in the control lysate, which were completely absent in the Dabg mutant. The antibody detected Pcdhg expression in the TC mutant ( Figure 1E), confirming that the three gC-type isoforms were translated into proteins. The amount of Pcdhg protein in the TC mutant was low, which is consistent with the result of quantitative real-time PCR. The large protein reduction was due to the deleted gAand gB-type isoforms and the remaining signal was due to the maintained expression of triple gC-type protein products ( Figure 1E).
Subsequently, we examined neonatal lethality, which was the salient common phenotype among all the mutant mice that lacked the critical gC4 isoform, such as the Dg, Dbg, Dabg, TCKO mutants, and DgC4 iScience Article mutant. 22,33,35,37 TC mutant mice were born alive; however, they exhibited acromphalus, a hunched posture, shallow breathing, slight movement, and no response to any touch of physical stimuli. Due to severely impaired breathing and blood circulation, the mutants died immediately after birth despite the (D) Quantitative real-time PCR showing the comparable expression of the three gC-type isoforms to control mice (+/+:TG taf7 ), although an increase in gC3 expression was noted. N = 3 animals per genotype. Error bars represent SEM *p < 0.05 by t-test.
(E) Western blot detection of Pcdhg isoforms in E18.5 brain lysates from TC mutants by the antibody against the constant region of Pcdhg (anti-gCR, indicated by red arrowheads). cPcdh-null mutants (Dabg) did not express Pcdhg, whereas TC mutants exhibited a large reduction but still detectable expression from the remaining gC3, gC4, and gC5 loci. iScience Article confirmed expression of the three gC-type isoforms (Figures 2A-2C; TC/TC, Videos S1, and S2). Their phenotype resembled that of Dabg and was more severe than that of Dg mutants, which exhibited a repetitive limb tremor and died within 12 h after birth (Video S1 in Hasegawa et al. 22 ). To confirm that the three gC-type isoforms in TC mutants were functional, we examined whether the TC allele could rescue the neonatal lethality of mutant mice that lacked the three gC-type isoforms. The mutant ab/mice lacking Pcdhg (which was generated by crossing abg/ab mice with abg/mice and retained only a single allele of Pcdhab) died after birth (Figures 2A-2C; ab/-). The behavioral defect of ab/mice was more severe than that of Dg mice (ab/ab). The mice exhibited little movement and little response to physical stimuli and died immediately after birth (Video S3). The introduction of a single TC allele into ab/mutant mice, such that they harbored a single allele of Pcdhab and a single allele of TC (ab/TC), rescued neonatal lethality (Figures 2A-2C; ab/TC). The ab/TC mice were born alive and behaved normally, providing proof of TC allele functionality (Video S4). The ab/TC mice survived beyond 7 months and were fertile, although their body weights were less than their littermates (at 7 weeks, the body weight of three ab/TC males was 18.5 g G 0.5 g, whereas the average weight of the other nine males was 23.1 g G 1.0 g; the average body weight of four ab/TC females was 16.8 g G 0.8 g, whereas the average weight of the other eleven females was 19.2 g G 1.4 g). The above results clearly showed that the deletion of 55 isoforms was neonatal lethal despite maintaining the functional TC allele and also showed that the three gC-type isoforms require the other isoforms in the Pcdha and Pcdhb clusters for the survival of the mice.

Massive apoptotic cell death in the brainstem reticular formation in TC mutants
The lack of the three gC-type isoforms (TCKO) was neonatal lethal but also caused massive apoptotic cell death in the spinal cord 33,37,38 and brainstem. 22 We, therefore, examined whether apoptotic cell death also occurs in TC mutants.
The spinal cords of TC mutants were thinner, suggesting a reduction in neuronal numbers ( Figures 2D-2F). Apoptosis was quantified by counting the remaining neurons of two representative neuronal types, namely, V1 inhibitory interneurons (FoxP2-expressing subsets) and V2a excitatory interneurons (Chx10-expressing subsets). The number of surviving FoxP2(+) interneurons and Chx10(+) interneurons in TC mutants was approximately 7.1 and 28.0% that of control mice, respectively, which was comparable to the reduction observed in Dabg mutant mice 22 (Figures 2I and 2J). Mutant mice carrying a combination of one TC-allele and one ab-allele (ab/TC mice) exhibited a normal spinal cord diameter. There was no statistically significant difference in the number of the FoxP2(+) and Chx10(+) interneurons between the control and ab/TC mice ( Figures 2H-2J), indicating that gC3, gC4, and gC5 in the TC-allele are functional for neuronal survival.
Next, we examined apoptotic cell death in the brain of an E18.5 TC mutant embryo. Figure 3A shows a sagittal section of the whole brain of a TC mutant embryo stained for cleaved-caspase-3 (CC3), a marker of apoptotic cells. Massive cell death was observed in the brainstem (midbrain, pons, medulla oblongata) ( Figure 3A). Apoptotic cells were also observed in several specific nuclei in the forebrain (Figures 3B-3E, higher magnification in Figure 4A; Table 1 (Table 1). Co-immunostaining of CC3 and the inhibitory neuronal marker GAD67 showed a correlation between the apoptotic cell death area and GAD67-enriched area. Apoptotic cells were normally observed in GAD67-enriched regions, such as the septum and ZI ( Figures 3B and 3E). Conversely, brainstem nuclei with weak GAD67 expression, such as the oculomotor nucleus, red nucleus, and the nucleus of the inferior colliculus, appeared to be devoid of (or had fewer) apoptotic cells ( Figure 4B). A subfraction of CC3-positive cells was also GAD67-positive, suggesting that inhibitory neurons were undergoing an apoptotic process ( Figure 4C). Quantitative analysis of apoptotic cell numbers or apoptotic cell area (including the area occupied by degenerating neuronal processes) showed that apoptosis also occurred with low frequency in control mice, whereas in TC mutant mice, the frequency increased by more than 10 times (Figures 4A, 4D and 4E). Among the brain regions, the midbrain reticular formation was the most severely affected. Analysis of apoptotic cell distribution suggests that apoptosis occurred in the brainstem reticular formation-centered interconnected neural networks with enriched inhibitory connections. The viability of ab/TC mice, in contrast to the neonatal death of TC mutants (Figures 2A-2C), suggested that the critical gC4 isoform needs to work in concert with the stochastic isoforms to exert its function. To elucidate the relationships of the spatial expression pattern of the stochastic isoforms and the gC4 isoform, we iScience Article conducted ISH of representative stochastic isoforms (a12, b22, gA3) and the gC4 isoform ( Figure 5). gC4 isoforms were highly expressed in the wide areas in the midbrain including PAG, midbrain reticular formation, and VTA ( Figure 5E), where massive apoptosis was observed in the TC mutant ( Figure 5A). Stochastic (a12, b22, gA3) isoforms were also expressed in similar areas in the midbrain of control mice, but the expression level of stochastic isoforms was generally very low and sparse ( Figures 5B-5D). Figures 5F-5T show representative examples of three brain areas where apoptosis was observed in TC mutants. As clearly shown in the higher magnification view of the septum (Figures 5F, 5I, 5L, 5O, and 5R), cortical amygdala ( Figures 5G, 5J, 5M, 5P, and 5S), and the reticular formation in the midbrain ( Figures 5H, 5K, 5N, 5Q, and 5T), the gC4 isoform was expressed in the vast majority of cells in these regions ( Figures 5I-5K), whereas the a12, b22, and gA3 isoforms were stochastically and sparsely expressed ( Figures 5L-5T). This expression pattern was also observed in other brain areas, such as the LHb and the VTA (data not shown). These results clearly show that the corresponding control brain regions where massive apoptosis occurred in TC mutants not only expressed the essential gC4 but also invariably expressed the stochastic isoforms simultaneously.
Finally, we examined whether apoptotic cells expressed the gC-type isoforms. For this purpose, we conducted dual immunohistochemistry (IHC) for CC3 and ISH for gC3 or gC4 mRNAs. The spatial expression pattern of the gC-type isoforms in the TC mutant did not change and was the same as that in control mice ( Figures 1F-1I). Figure 6 shows a higher magnification view of the ISH signal in the midbrain reticular formation for gC3 (Figures 6A and 6D) and gC4 (Figures 6G and 6J). The density of gC3-expressing cells was a little low compared with gC4, but quite a few cells expressed gC3 (Figures 6A and 6D). As the cells positive for CC3 were undergoing an apoptotic process, the mRNAs in these cells were potentially in the process of degradation. However, we found cells expressing both gC3 and CC3 (Figures 6C and 6F, higher magnification in 6MÀ6O) and cells expressing both gC4 and CC3 (Figures 6I and 6L, higher magnification in 6P-6R). This result suggests that the expression of either gC3 or gC4 did not ensure cell viability in the TC mutant.

DISCUSSION
We examined the phenotype of TC mutant mice lacking 55 cPcdh isoforms except the essential three gCtype isoforms. The mice died immediately after birth and exhibited massive neuronal death despite the presence of the functional TC allele. This finding suggests that the three gC-type isoforms were not iScience Article sufficient to rescue the neuronal defects of clustered Pcdh-null mice (Dabg), and other cPcdh isoforms are also required for the survival of neurons and individual mice.
Apoptosis was also observed in control mice in the same brain regions as TC, albeit the frequency of apoptotic cells was very low in control (Figures 4A and 4C). This suggests that apoptotic cell death is a iScience Article normal developmental process that occurs during neural network wiring. Massive apoptosis occurs in the absence of cPcdh (Dabg mouse), 22 suggesting that the default strategy of neural network formation in the brainstem and spinal cord is to exclude inappropriate cells by apoptosis and that cPcdh provides the survival signal. PcdhgC4 has been shown to be essential for the survival signal. 33,35 However, in TC mutants, the expression of PcdhgC4 did not ensure neuronal survival. At least the following three explanations are possible.
The first possibility is that PcdhgC4 is nonfunctional in TC mutants due to the failure in cell surface trafficking, It has been shown that certain cPcdh isoforms including PcdhgC4, the stochastic isoforms in Pcdha (a1-12) and PcdhaC1, cannot translocate to the cell surface by themselves (when expressed alone). PcdhgC4 requires another carrier isoform to form a cis-dimer to enable its translocation to the cell surface. 11,12,17 TC mutants express both gC3 and gC5 isoforms as potential carriers. However, gC5 expression was low at the prenatal stage, and the expression of gC3 was, to our surprise, rather sparse compared to gC4 in the apoptotic brain regions. Therefore, in TC mutants, the remaining gC3 and gC5 may not be enough for gC4 cell surface expression. This must be addressed in the future.
A second possibility is that the deleted isoforms in TC (Pcdha1 to PcdhgA12) are also critical for the survival of neurons and individual mice. Although the gC4 appeared as the only critical isoform, requirement of  39 We also observed the allele number-dependent effects of Pcdha and Pcdhb clusters on neonatal survival; the Dg mutant (ab/ab) exhibited repetitive limb tremor and died within 12 h, while the ab/mutant exhibited little movement and died immediately after birth. The requirement of gC4 might be attributable to it being the most dominantly expressed isoform in the apoptotic brain areas. The simultaneous loss of a bunch of other stochastic isoforms with low expression may also cause the lowering of the survival signal below the required threshold.
A third possibility is that the cis-dimer of PcdhgC4 and another isoform acts as the functional unit and exerts its antiapoptotic signal. A disadvantageous feature of gC4 is that it alone cannot be transported to the cell surface and requires another isoform (forming a cis-heterodimer) for its cell surface delivery. Therefore, the generation of the survival signal of gC4 is linked to the diversity of cell-surface recognition ability. Our expression study at E18 indicated that the essential gC4 isoform is always expressed with other stochastic isoforms, which is consistent with the above idea. The defect observed in TC mutants supports that the gC4 requires other isoforms (included in deleted 55 isoforms). Whether there is a specific carrier isoform, or in fact 55 isoform variety is required, awaits further study.
Neuronal types whose survival depends on cPcdh have been reported in many brain areas, such as spinal cord interneurons, brainstem neurons, cortical interneurons, retinal neurons. 22,32,34,[38][39][40] Here, we mapped the cPcdh-dependent neurons in the forebrain and midbrain at E18. The cell death areas include the MSDB, LHb, LHA, cortex-amygdala transition zone, ZI, VTA, PAG, and midbrain reticular formation, which were directly connected according to the previous studies. [41][42][43][44][45] Therefore, apoptotic cell death area appeared as a directly connected single mass of neural network. This suggests that the apoptotic cell death in TC mutants was correlated with network organization and may be correlated with its activity.
GABAergic interneurons are a major apoptotic cell type found in the Dg mutant. There exists a correlation between the severity of neuronal loss (reduction volume of the tissue) and abundance of GABAergic neurons (GAD2 expression). 22,34,37,38,40 We also found that the spatial distribution of CC3-positive apoptotic cells in the E18 TC mutant closely resembles that described in published ISH data for GAD1/GAD2 (available on Allen Brain Atlas; https://portal.brain-map.org). 46 The distribution of CC3-positive cells in TC mutants also resembles (although not exactly) the staining pattern of acetylcholine esterase (AChE) in the coronal plane of the E15/16 mouse brain (as reported in the mouse brain atlas by Jacobowitz and Abbott 47 ). These correlations suggest that the cell death area is enriched with GABAergic neurons and generally receives cholinergic input.
It has been shown that synchronized rhythmic activity, which is distinct from the activity of a mature circuit, occurs at many sites in the developing nervous system. 48 This activity has been extensively studied in the spinal cord; a wave of the synchronized rhythmic activity is propagated over the entire network, driven by both GABAergic and cholinergic inputs, and has been considered a general necessary program of neural circuit wiring. [48][49][50][51][52][53] Massive apoptosis in the developing spinal cord of Pcdh-deficient mice (such as TC and Dabg) occurred after the cessation of synchronized rhythmic activity. 22,53 The close spatial correlation of apoptotic cell distribution in TC mutants with that of GABAergic neurons and AChE and the temporal correlation of the onset of apoptosis and the cessation of a synchronized rhythmic activity suggest that embryonic rhythmic activity may be a prerequisite for the cPcdh-mediated survival signal. Determining whether synchronized neuronal activity plays a role in cell surface trafficking of cPcdh, which generates the survival signal, or whether the activity of the network whose wiring is regulated by cPcdh itself governs neuronal survival, requires further study. Taken together, the gC-type isoforms appeared to regulate neuronal survival by cooperating with other cPcdh isoforms, the molecular mechanism of which should be clarified in the future. iScience Article cis-dimer/multimer Pcdh complex, and the downstream signaling cascade need to be clarified in the future. The differential role of gC4 and other stochastic isoforms will be elucidated in the future by re-introducing stochastic isoforms to the TC allele.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We would like to thank Sonoko Hasegawa, Yukinori Inoue, and Yoshito Sakaij for their assistance with the animal models.

DECLARATION OF INTERESTS
The authors declare no competing interests. iScience 26, 105766, January 20, 2023 iScience Article iScience Article RESOURCE AVAILABILITY

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Takeshi Yagi (yagi@fbs.osaka-u.ac.jp).
Data and code availability d All data are available in the manuscript or supplemental information. d All codes are available in the supplemental information.
d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS Animals
All animal procedures were performed according to the Guide for the Care and Use of Laboratory Animals of the Science Council of Japan and were approved by the Animal Experiment Committee of Osaka University and the Institutional Animal Care and Use Committee of RIKEN Kobe Branch. Adult (beyond 2 months old) male and female Pcdh TC/+ :TG Taf7 , Pcdhabg del/+ :TG Taf7 (Hasegawa et al. 22 ), Pcdhg del/+ :TG Taf7 (Hasegawa et al. 22 ), and Pcdh +/+ :TG Taf (Hasegawa et al. 22 ) mice in a C57BL/ 6 background were used and maintained in the animal facility of Osaka University. Mice were housed in groups under a 12 h:12 h light:dark cycle. Control mice (+/+:TG taf7 ) were the littermates of heterozygous breeding for each genotype. All mouse strains used in this study were deposited in RIKEN BRC, Japan.

Oligonucleotides
For Southern blot probe A, B sequence, see Table S1 This study N/A For genotyping Pcdhabg TC allele, see Table S1 This study N/A For genotyping TG Taf7 allele, see Table S1 This study N/A For RT-PCR/real-time RT-PCR, see Table S1 This study N/A For amplifying in situ hybridization RNA probes, see Table S1 This study N/A

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analysis was conducted using Prism 7.05 (GraphPad, San Diego, CA). The data are expressed as the mean G SD. For the analysis in Figures 2I and 2J, one-way analysis of variance (ANOVA) and Tukey's post-hoc test was applied. For the analysis in Figures 4D and 4E, Mann-Whitney U-test was applied. p Values <0.05 were considered statistically significant. The details for each experiment including the number of animals are specified in the figure legends.

ll
OPEN ACCESS