Dysregulated protocadherin-pathway activity as an intrinsic defect in induced pluripotent stem cell-derived cortical interneurons from subjects with schizophrenia

Abstract

We generated cortical interneurons (cINs) from induced pluripotent stem cells derived from 14 healthy controls and 14 subjects with schizophrenia. Both healthy control cINs and schizophrenia cINs were authentic, fired spontaneously, received functional excitatory inputs from host neurons, and induced GABA-mediated inhibition in host neurons in vivo. However, schizophrenia cINs had dysregulated expression of protocadherin genes, which lie within documented schizophrenia loci. Mice lacking protocadherin-α showed defective arborization and synaptic density of prefrontal cortex cINs and behavioral abnormalities. Schizophrenia cINs similarly showed defects in synaptic density and arborization that were reversed by inhibitors of protein kinase C, a downstream kinase in the protocadherin pathway. These findings reveal an intrinsic abnormality in schizophrenia cINs in the absence of any circuit-driven pathology. They also demonstrate the utility of homogenous and functional populations of a relevant neuronal subtype for probing pathogenesis mechanisms during development.

Figure 1.. SCZ and HC iPSCs efficiently generate homogeneous population of cINs.

The lines used in each experiment are summarized in Supplementary Table 3. (a) Table of subjects analyzed in pilot study. (b) Immunocytochemistry analysis of human PSC markers, Oct4 and Tra-1–60, in generated iPSCs (Scale bar = 200 μm). Differentiation was repeated at least 3 times with comparable results. (c) Differentiation scheme of cINs from human iPSCs. SRM: serum replacement media, LDN: 100 nM LDN193189, SB: 10 μM SB431542, SAG: 0.1 μM Smoothened agonist, and IWP2: 5 μM Inhibitor of Wnt production-2. (d) Immunocytochemistry analysis for expression of β-Tubulin III, Sox6, and GAD1 in generated cINs after 8 weeks’ differentiation (Scale bar = 50 μm). Differentiation was repeated at least 3 times with comparable results. (e) Cell counting analysis after 8 weeks’ differentiation. Data are presented as mean ±SEM from three independent differentiations (n = 3 differentiations). There were no significant differences among different lines based on one-way ANOVA (β-Tubulin III p = 0.2626, Sox6 p = 0.3802, and GAD1 p = 0.5072). (f) Real-time PCR analysis of different cell types (H9 hESC: n = 3 batches, hiPSC: n = 6 batches from 2 iPSC lines, HC cIN: n = 9 batches from 3 lines, SCZ cIN: n = 9 batches from 3 lines, and glutamatergic neuron: n = 6 batches from 2 lines). Results were normalized using Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) expression levels and presented as mean ±SEM. See also Supplementary Fig. 1–2.

Figure 2.. Neuronal and synaptic properties of cINs transplanted into mouse cerebral cortex.

The lines used in each experiment are summarized in supplementary Table 3. (a) Transplantation scheme of cINs into Nod Scid mice. Eight-week old cINs infected with LV-Syn-ChR2-GFP were transplanted into cerebral cortices. Brain slices were prepared for electrophysiological experiments 7 months after grafting. (b) Left: image showing ChR2-GFP-expressing grafted cINs (green) in the cerebral cortex. Right: patch-clamp recordings were performed with grafted cells, which generated ChR2-mediated photocurrents and were labeled with biocytin-streptavidin-Alexa 568 (scale bar = 10 µm). (c) Voltage pulses from –115 mV to 85 mV with 20 mV increments induced transient inward (Na⁺) and sustained outward currents (K⁺) in a grafted cINs. Na⁺ currents were quantified as peak inward currents induced by depolarization to –35 mV (red). K⁺ currents were quantified as sustained outward currents induced by depolarization to 85 mV (blue). (d) AP firing induced by current injection in a grafted cell at approximately –85 mV. (e) Summary plot of AP firings in transplanted cINs in the HC (15 cells) and SCZ groups (16 cells). Repeated measures two-way ANOVA was used for analysis (Main effect of groups, p = 0.77; group × current interaction, p = 1.00). Error bars are the SEM. (f) Left: spontaneous AP firings recorded at RMP in a grafted cIN. Right: average AP firings. (g) Spontaneous EPSCs recorded in a grafted cell at –85 mV (left) and blocked by NBQX (right). (h) Photostimulation activates ChR2-expressing grafted cINs, generating IPSCs in GFP^– host neurons labeled with biocytin (red, scale bar = 10 µm). (i) Image showing a GFP⁺ grafted human cIN (green) and a GFP^– host cortical neuron labeled with biocytin (red). (j) Photostimulation induced IPSCs recorded at 0 mV in a host neuron (black) and inhibited by SR95531 (10 mM, blue). (k) IPSCs induced by photostimulations (40 mW/mm²) and recorded in host neurons in the HC (black) and SCZ groups (red). (l) Left: comparison of the proportion of host cells with IPSCs in the HC (15 cells out of 21 cells examined) and SCZ groups (13 cells our of 20 cells examined; p = 0.66, χ2 test). Right: comparison of IPSC amplitude in the HC (15 cells) and SCZ groups (13 cells; p = 0.37). Error bars are the SEM. See also Supplementary Fig. 3–5. For detailed statistics information, see Supplementary Table 15.

Figure 3.. PCDHA2 expression is significantly altered in developing SCZ cINs.

The lines used in each experiment are summarized in Supplementary Table 3. The breakdown data are summarized in Supplementary Table 4–5. (a) RNA-seq transcriptome profiling of cINs derived from 4 HC vs. 4 SCZ iPSCs in three independent differentiations (n = 12 differentiations). Gene expression is shown as reads per kilobase of transcript per million mapped reads (RPKM). Differential expression was analyzed in R by the Voom function in Limma with adjusted multiple testing. Center and error bars show mean ± SEM. (b) PCDHA2 expression is significantly decreased in SCZ cINs. Data were collected by RNA-seq transcriptome profiling of cINs derived from 4 HC vs. 4 SCZ iPSCs in three independent differentiations (n = 12 differentiations). Differential expression was analyzed in R by the Voom function in Limma with adjusted multiple testing. Center and error bars show mean ± SEM. (c) Quantitative real-time PCR analysis of PCDHA2 expression in cINs derived from 4 HC iPSCs vs. 4 SCZ iPSCs in three independent differentiations (n = 12 differentiations). Data were normalized by GAPDH expression level. Center and error bars show mean ± SEM. (d) The differentiation scheme for glutamatergic neurons. Human iPSCs were plated with Lentivirus expressing inducible Ngn2-puroR and treated with Doxycycline (DOX) until day 10. Cells were treated with Puromycin on day 2 and were used for analysis on day 14. (e) Immunocytochemistry analysis of iPSC-derived glutamatergic neurons (scale bar = 20 μm). This analysis was repeated in 2 batches of differentiation with similar results. (f) Quantitative real-time PCR analysis of PCDHA2 expression in glutamatergic neurons derived from 4 HC vs. 4 SCZ iPSCs in two independent differentiations (n = 8 differentiations). PCDHA2 expression level was normalized using that of GAPDH. Center and error bars show mean ± SEM. See also Supplementary Fig. 6–7. For detailed statistics information, see Supplementary Table 15.

Figure 4.. Multiple PCDHA family members are affected in developing SCZ cINs.

The lines used in each experiment are summarized in Supplementary Table 3. The breakdown data are summarized in Supplementary Table 6. (a) Table of subjects in the expanded cohort. (b) Changes in multiple PCDHA gene expression, analyzed by RNA-seq of cINs derived from 14 HC vs. 14 SCZ iPSCs in two independent differentiations (n = 28 differentiations). Expression level was shown as transcripts per kilobase million (TPM). Differentially expressed genes were analyzed by DESeq2 and Sleuth (Wald test for significance testing). Center and error bars show mean ± SEM. (c) Quantitative real-time PCR analysis of expression of multiple PCDHA family members in cINs derived from 14 HC vs. 14 SCZ iPSCs in two independent differentiations (n = 28 differentiations). The expression levels of each gene were normalized by that of the GAPDH gene. Center and error bars show mean ± SEM. (d) Correlation between genotype and gene expression level. R designates reference allele (G) and A designates alternate allele (A) of rs7445192. Gene expression levels were normalized by β-actin and grouped by genotype. The difference between the RR group and the AA group was analyzed by the two-tailed unpaired t-test (n = 6 for RR and n = 8 for AA; PCDHA2 p = 0.5102, PCDHA3 p = 0.4551, PCDHA6 p = 0.7251 and PCDHA8 p = 0.3548). Center and error bars show mean ± SEM. (e) Heat map depicting expression changes of protocadherin genes in SCZ cINs. Blue color depicts downregulation in SCZ cINs and red depicts upregulation. Numbers in colored box indicate expression change in log2 fold change. (f) eQTL diagram of different protocadherin genes, showing the presence of multiple eQTLs for protocadherin family members within SCZ risk loci. Green bar designates SCZ risk loci (Chr5: 140,023,664–140,222,664). Red dots are significant cis-eQTLs for the given gene at FDR<5%. More detailed information on this analysis can be found in the Commonmind portal (https://www.synapse.org/#!Synapse:syn2759792/wiki/69613). See also Supplementary Fig. 8. For detailed statistics information, see Supplementary Table 16.

Figure 5.. Pcdha KO (Pcdha ^Δα/Δα) leads to cIN abnormalities in the PFC and behavioral abnormalities.

The breakdown data for each mouse are summarized in Supplementary Table 7–8. (a) Arborization analysis of PV⁺ cINs in PFC of littermate wildtype (WT) or Pcdha KO (Pcdha ^Δα/Δα) mice. Arborization within 150 µm of the cell body was traced and analyzed using ImageJ with the Neuron J plugin and is presented as mean ±SEM (n = 100 neurons per group). (b-c) Analyses of inhibitory and excitatory synapses in PFC of WT and Pcdha ^Δα/Δα mice using Imaris software. For inhibitory synapse analysis on PV⁺ cINs, the data are presented as the number of inhibitory synapses (juxtaposed PV⁺VGAT⁺ puncta and Gephyrin⁺ puncta) per 100 µm² of PV⁺ cINs. For excitatory synapse analysis on PV⁺ cINs, the data are presented as the number of excitatory synapses (juxtaposed PV⁺PSD95⁺ puncta and Vglut1⁺ puncta) per 100 µm² of PV⁺ cINs. Data are presented as mean ±SEM (n = 25 116 µm × 116 µm images per group). (d) Pcdha ^ΔCR/ΔCR mice show a deficit in PPI. Amplitudes of startle response and percentage of PPI are shown. The data were analyzed using the two-tailed unpaired t-test. Data are shown as mean ±SEM (WT: n = 17 mice, ΔCR: n = 18 mice). (e) Pcdha ^{ΔBneo/ΔBneo} mice show a deficit in PPI. Amplitudes of the startle response and the percentage of PPI are shown. The data were analyzed using the two-tailed unpaired t-test. Data are shown as mean ±SEM for the indicated numbers of mice (n = 13 mice for each group). See also Supplementary Fig. 9. For detailed statistics information, see Supplementary Table 17.

Figure 6.. Developing SCZ cINs show arborization deficit relevant to protocadherin dysregulation in vitro.

The lines used in each experiment are summarized in Supplementary Table 3. The breakdown data for each line are summarized in Supplementary Tables 9–11. (a) Arborization analysis of HC vs. SCZ cINs infected with a limiting titer of GFP-expressing lentivirus. Images were analyzed using ImageJ with the Neuron J plugin. (n = 180 neurons). Center and error bars show mean ± SEM. (b) Arborization analysis of PKC inhibitor-treated HC vs. SCZ cINs. cINs infected with a limiting titer of GFP-expressing lentivirus were treated with PKC inhibitor GO6893 for 12 days and analyzed using ImageJ with the Neuron J plugin (n = 160 neurons). Center and error bars show mean ± SEM. (c) Arborization analysis of siRNA transfected HC vs. SCZ cINs (two independent differentiations). cINs infected with a limiting titer of GFP-expressing lentivirus were transfected with siRNA against PCDHA, PCDHG, or a mixture (PCDHA+PCDHG) and analyzed using ImageJ with the Neuron J plugin (n=60 neurons). Center and error bars show mean ± SEM. The asterisk indicates p < 0.05 compared to the negative control. See also Supplementary Fig. 10. For detailed statistics information, see Supplementary Table 17–18.

Figure 7.. Developing SCZ cINs show synaptic deficit in vivo after transplantation into mice brains.

The lines used in each experiment are summarized in Supplementary Table 3. The breakdown data for each line are summarized in Supplementary Table 12. (a) Schematic diagram of transplantation analysis of HC vs. SCZ cINs. (b) 3D Lightsheet microscopy images of an iDISCO+-cleared mouse cortex transplanted with iPSC-derived human cINs, stained with human cytoplasm antibody. Major grid in reference frame is 300 µm. This result is from a single whole brain staining. (c-d) Recognition of grafted neurons by Imaris software shown by green dots. Note that the cells in the front look bigger with bigger green dots and the cells in the back look smaller with smaller green dots. Graphs depict the location of grafted cells along the X, Y, Z axes and distance from the injection site. This result is from a single whole brain staining. (e) Immunocytochemistry analysis of transplanted cINs 7 months after grafting (scale bar = 50 µm). Left: MEF2C⁺ neurons among human NCAM⁺ grafted neurons (p = 0.7737, chi-square test). Right: SST⁺ neurons among human NCAM⁺ grafted neurons (p = 0.6575, chi-square test). White arrowheads indicate double-positive neurons. (f) Inhibitory synapse analysis on transplanted GFP⁺ cINs 4 months after grafting. The data are presented as the number of inhibitory synapses (juxtaposed GFP⁺VGAT⁺ puncta and Gephyrin⁺ puncta) per 100 µm² of GFP⁺ cINs. Center and error bars show mean ± SEM (n = 14 116 µm × 116 µm images per group). (g) Excitatory synapse analysis on transplanted GFP⁺ cINs 4 months after grafting. The data are presented as the number of excitatory synapses (juxtaposed GFP⁺PSD95⁺ puncta and Vglut1⁺ puncta) per 100 µm² of GFP⁺ cINs. Center and error bars show mean ± SEM (n = 14 116 µm × 116 µm images per group). See also Supplementary Fig. 11. For detailed statistics information, see Supplementary Table 17.

Figure 8.. SCZ cINs in postmortem tissue show similar deficit as in developmental SCZ cINs.

The breakdown data for each line are summarized in Supplementary Table 13–14. (a) Table of subjects used in the postmortem study. (b) Arborization analysis of layer 3 PV⁺ cINs in HC and SCZ postmortem PFC. Arborization within 150 µm of the cell body was traced and analyzed using ImageJ with the Neuron J plugin. Data are presented as mean ±SEM (n = 160 neurons). (c-d) Synapses analysis of layer 3 PV⁺ cINs in HC and SCZ postmortem PFC using Imaris software. For inhibitory synapse analysis on PV⁺ cINs, the data are presented as the number of inhibitory synapses (juxtaposed PV⁺VGAT⁺ puncta and Gephyrin⁺ puncta) per 100 µm² of PV⁺ cINs. For excitatory synapse analysis on PV⁺ cINs, the data are presented as the number of excitatory synapses (juxtaposed PV⁺PSD95⁺ puncta and Vglut1⁺ puncta) per 100 µm² of PV⁺ cINs. Data are presented as mean ±SEM (n = 32 116 µm × 116 µm images). See also Supplementary Fig. 12. For detailed statistics information, see Supplementary Table 17.