Novel genetic features of human and mouse Purkinje cell differentiation defined by comparative transcriptomics

Abstract

Comparative transcriptomics between differentiating human pluripotent stem cells (hPSCs) and developing mouse neurons offers a powerful approach to compare genetic and epigenetic pathways in human and mouse neurons. To analyze human Purkinje cell (PC) differentiation, we optimized a protocol to generate human pluripotent stem cell-derived Purkinje cells (hPSC-PCs) that formed synapses when cultured with mouse cerebellar glia and granule cells and fired large calcium currents, measured with the genetically encoded calcium indicator jRGECO1a. To directly compare global gene expression of hPSC-PCs with developing mouse PCs, we used translating ribosomal affinity purification (TRAP). As a first step, we used Tg(Pcp2-L10a-Egfp) TRAP mice to profile actively transcribed genes in developing postnatal mouse PCs and used metagene projection to identify the most salient patterns of PC gene expression over time. We then created a transgenic Pcp2-L10a-Egfp TRAP hPSC line to profile gene expression in differentiating hPSC-PCs, finding that the key gene expression pathways of differentiated hPSC-PCs most closely matched those of late juvenile mouse PCs (P21). Comparative bioinformatics identified classical PC gene signatures as well as novel mitochondrial and autophagy gene pathways during the differentiation of both mouse and human PCs. In addition, we identified genes expressed in hPSC-PCs but not mouse PCs and confirmed protein expression of a novel human PC gene, CD40LG, expressed in both hPSC-PCs and native human cerebellar tissue. This study therefore provides a direct comparison of hPSC-PC and mouse PC gene expression and a robust method for generating differentiated hPSC-PCs with human-specific gene expression for modeling developmental and degenerative cerebellar disorders.

Keywords: Purkinje cell; human pluripotent stem cells; transcriptomics.

Fig. 1.

Differentiation of hPSCs to PCs. (A) Schematic of the first phase of differentiation. (B) Immunolabeling of a cryosection of a representative neural aggregate after 6 d of differentiation; 10 µm z projection. (Scale bar: 50 µm.) (C) Immunolabeling of an attached neural aggregate after 10 d of differentiation; 18 µm z projection. (Scale bar: 100 µm.) (D) Immunolabeling of an attached neural aggregate after 22 d of differentiation; 18 µm z projection. (Scale bar: 100 µm.) (E) Immunolabeling of an attached neural aggregate after 24 d of differentiation. (Scale bar: 50 µm.) (F) A representative histogram and quantification of flow cytometry for PCP2 after 24 d of differentiation. Positive signal is >0.01% of secondary control. Error bars represent SD. (G) Schematic of the second phase of differentiation. (H) Immunolabeling of hPSC-PCs after isolation and coculture with mouse glia cells and GCs for an additional >89 d. The panels from left to right show 7 µm z projection, 5 µm z projection, 5 µm z projection, and single optical section. (Scale bars: 50 µm.) (I) Live imaging of genetically encoded calcium indicator jRGECO1a and hPSC-PC reporter Pcp2-mGFP after 100 d of coculture with mouse glia cells and GCs and a trace of the change in jRGECO1a fluorescence (ΔF/Fο) over time. (J) Representative trace of jRGECO1a fluorescence in the presence of the Na⁺ channel antagonist TTX after 101 d of coculture with mouse glia cells and GCs. (K) Representative trace of jRGECO1a fluorescence in the presence of the glutamate receptor antagonist CNQX after 100 d of coculture with mouse glia cells and GCs.

Fig. 2.

Comparison of hPSC-PC gene expression with mouse PC gene expression over development. (A) Schematic of the lentiviral construct used to create the PCP2-EGFP-L10a TRAP hPSC line and immunolabeling of the construct in differentiating hPSC-PCs; 4 µm z projections. (Scale bars: 50 µm.) (B) Schematic depicting timing of TRAP RNA isolation from differentiation hPSC-PCs. (C) Principle component analysis of differentiating hPSC-PCs after 24 and +95 d as well as undifferentiated hESC lines (H1, RUES2). (D) Heat map showing median scaled ssGSEA enrichment scores of gene expression levels in mouse PCs over postnatal development for gene sets defined as log₂ 4-fold (16-fold) change between day 24 and day +95 in hPSC-PCs. Day 24 hPSC-PCs are most similar to P0 mouse PCs (P = 0.0014). Day +95 PCs are most similar to P21 mouse PCs (P = 1.74x10⁻³²).

Fig. 3.

NMF analysis of mouse PC developmental gene expression and comparison with hPSC-PCs. (A) Metagenes defined by NMF analysis of mouse PC gene expression over postnatal development. (B) GO expression signatures with a correlation coefficient to a metagene in A of >0.85. Listed next to the terms is the metagene with highest correlation. (C) Expression patterns of representative genes from each of the major GO categories found in B.

Fig. 4.

Differences between hPSC-PCs and mouse PCs. (A) Expression patterns of thyroid hormone signaling pathway. (B) Expression levels of primate-specific genes in hPSC-PCs at day 24 and day +95. Error bars are SEM. (C) Expression levels of genes up-regulated in hPSC-PCs but not expressed in mouse PCs. For hPSC-PCs, background was set at 2 TPMs (dotted lines). For mouse PCs, background was set at microarray intensity of six (dotted lines). The negative control NANOG and the positive control CALB1 are shown for reference. Error bars are SEM. (D) Expression of CD40lg in hPSC-PCs and 5-d human cerebellum but not P7 mouse cerebellum. (Scale bars: 50 µm.)