Protocadherin γC4 Promotes Neuronal Survival in the Mouse Retina Through Its Variable Cytoplasmic Domain

Abstract

Developmental apoptosis is an important mechanism for the formation of functional neural circuits. Distinct neuronal subtypes undergo apoptosis to a greater or lesser extent during development, although how this is regulated at the cell type level is unknown. The clustered protocadherins (cPcdhs) are ~ 60 homophilic cell adhesion molecules expressed from three contiguous gene clusters, which together encode the α-, β-, and γ-Pcdh families. Only one cPcdh isoform, γC4, is essential for survival in the mouse, given its role in attenuating the extent of developmental neuronal apoptosis. However, there is also evidence that other isoforms contribute to neuronal survival. Here, we focused on amacrine cell types in the mouse retina, using a series of genetic models to ascertain that γC4 alone accounts for the pro-survival function of the γ-Pcdhs, and that neuronal subtype dependence on γ-Pcdhs for survival correlates with expression of this single isoform. To test which domains of the protein were essential for this function, we employed a rescue approach with in vitro live cell imaging, finding that the unique variable cytoplasmic domain of γC4-not its adhesive extracellular cadherin repeats-is necessary and sufficient to promote neuronal survival.

Keywords: Apoptosis; G-Protocadherin; Mouse; Neural development; Retina.

Fig. 1

Thinning of the retinal inner nuclear layer in Pcdhg mutants. A A schematic representation of the mouse Pcdhg locus and the three alleles used in this study: Variable exons are categorized as γA, γB, γC based on sequence homology while constant exons are common to all isoforms. The Pcdhg^fcon3 allele includes loxp sites (blue triangles) flanking the GFP-fused third constant exon, and recombination of this allele disrupts all 22 isoforms. When crossed with Pax6α-Cre, the gene cluster is disrupted in the retina, which we refer to here as Pcdhg^RKO. The Pcdhg^C4KO allele harbors a 13 bp deletion in the Pcdhgc4 exon, resulting in a frame shift and loss of this isoform only. Pcdhg^1R1 resulted from a large deletion spanning from Pcdhga1 to Pcdhgc3, and a 1 bp deletion in Pcdhgc5. γC4 is the only isoform produced from this allele. B–E Retina cryosections from adult mice of the indicated genotypes were stained for DAPI to clearly identify the cellular retinal layers. The thickness of the inner nuclear layer (INL, black bars to the left of each panel) was sampled across the length of the section and averaged to a single number for each retina. Each point in F represents a single retina and bars represent mean values. Genotypes were compared using ANOVA with Tukey post hoc pairwise comparisons. Ns are indicated in parentheses in F. ** is p < 0.01, *** is p < 0.001, ns is not significant. Scale bar is 50 µm

Fig. 2

Loss of γC4 can account for the whole reduction in amacrine cells in Pcdhg^RKO/RKO mutants. Whole-mount adult retinas from the indicated genotypes were stained for A–E tyrosine hydroxylase (TH) to label dopaminergic ACs, F–J VGluT3 to label glutamatergic ACs, and K–T Chat to label starburst ACs (SACs). SACs are analyzed separately as the ON subtype (K–O) and the OFF subtype (P–T) based on their layer position. Densities were calculated for each AC type, and genotypes were compared by ANOVA with Tukey post hoc pairwise comparisons. Ns are indicated in parentheses on each graph. Controls are a mix of Pcdhg^RKO/+ and Pcdhg^+/+ retinas. * is p 0.05, ** is p < 0.01, *** is p < 0.001, **** is p < 0.0001, ns is not significant. Scale bar is 250 µm in A–D, 100 µm in other panels

Fig. 3

RNAScope analysis of Pcdhgc4 and Pcdhgc3 expression in SACs and VGluT3 + ACs. RNAscope was performed on cryosections from wild type retinas at the indicated ages for Vglut3 (closed arrowheads), Chat (open arrowheads), and A–C Pcdhgc4 or G–I Pcdhgc3. H-scores were calculated to approximate expression levels of both isoforms in each cell type (plotted in D–F for Pcdhgc4, J–L for Pcdhgc3). Means were compared by a Brown-Forsythe ANOVA and pairwise comparisons with Dunnett’s T3 multiple comparisons test. N = 4 retinas per condition. * is p 0.05, ** is p < 0.01, *** is p < 0.001, ns is not significant. Scale bar is 50 μm

Fig. 4

Slc35d3 + and Nefh + ACs have differential dependence on Pcdhg for survival. RNAScope was used to label A–D Nefh + and E–G Slc35d3 + ACs in the adult retinas of the indicated genotypes. Cells were counted and normalized to the length of the section to estimate density (D, H). Controls are a mix of Pcdhg^RKO/+ and Pcdhg^+/+ retinas. Genotypes were compared by ANOVA with Tukey post hoc pairwise comparisons. Ns are indicated in parentheses on each graph. * is p 0.05, *** is p < 0.001, ns is not significant. Scale bar is 50 μm

Fig. 5

RNAScope analysis of Pcdhgc4 expression in Slc35d3 + and Nefh + ACs. A–D RNAscope was performed on cryosections from wild type retinas at the indicated ages for Slc35d3 (red, closed arrowheads), Nefh (green, open arrowheads), and Pcdhgc4 (Cyan, lower panels). H-scores were calculated to approximate expression levels in both cell types (plotted in E–G). Means were compared by Student’s t-test. Ns are indicated in parentheses on each graph. * is p 0.05, ** is p < 0.01, *** is p < 0.001, ns is not significant. Scale bar is 50 µm. H The mean H-score for Pcdhgc4 at P10 in 7 AC types is plotted on the y-axis with the percent survival in Pcdhg^RKO/RKO on the x-axis. Error bars indicate standard deviation in x and y. Spearman correlation between H-score and % survival indicated a significant correlation with r =  − 0.786 and p = 0.048. I In a similar analysis using H-score for Pcdhgc3, Spearman correlation indicated no significant correlation with r = 0.600 and p = 0.242

Fig. 6

Long-term live cell imaging of γC4 promotion of neuronal survival. A Mixed retina cultures were transfected with cDNAs 2 days before imaging, then imaged for 48 h at 30-min intervals in the presence of NucView 488. B Bicistronic rescue constructs with a human synapsin promoter ensured that mCherry-positive cells were neurons expressing rescue cDNA. Mixed retinal neurons cultured from C–F control or G–R Pcdhg^C4KO/C4KO retinas were transfected with the C–J mCherry, L–O mCherry and full-length Pcdhgc4 cDNA, or P–S mCherry and truncated Pcdhgc4 cDNA without a constant domain. A spike in the green NucView signal in the nucleus of an mCherry-positive cell (e.g., J) indicated apoptosis, and was used to measure survival time (plotted in K, T). To test for rescue, survival in each condition was compared to matched (i.e., within the same 24-well plate) Pcdhg^C4KO/C4KO neurons transfected with mCherry by log-rank Mandel-Cox test. ** is p < 0.01, **** is p < 0.0001, scale bar is 50 µm. Ns are the number of neurons imaged per condition and are indicated in parentheses. C4FL and mCherry data from K are replotted in T for comparison

Fig. 7

γC4 promotes neuronal survival through its variable cytoplasmic domain. Live cell imaging of Pcdhg^C4KO/C4KO neurons transfected with A–D a C3/C4 construct comprised of the extracellular domain and transmembrane domain of γC3 and the cytoplasmic domain of γC4 and E–H the converse C4/C3 construct with the extracellular domain and transmembrane domain of γC4 and the intracellular domain of γC3. I Survival analysis demonstrates that the cytoplasmic domain of γC4 is essential to prevent apoptosis, while the extracellular domain is not. Neurons transfected with J–M the entire cytoplasmic domain only or N–Q the variable cytoplasmic domain only (excluding the constant domain) also rescued cell death (R). Membrane-targeting of these domains was not required. * is p < 0.05, ** is p < 0.01, **** is p < 0.0001, ns is not significant, by log-rank Mandel-Cox test compared with matched Pcdhg^C4KO/C4KO neurons transfected with mCherry. Scale bar is 50 µm. Ns are the number of neurons imaged per condition and are indicated in parentheses