The transcription factor ZEB1 regulates stem cell self-renewal and cell fate in the adult hippocampus

Abstract

Radial glia-like (RGL) stem cells persist in the adult mammalian hippocampus, where they generate new neurons and astrocytes throughout life. The process of adult neurogenesis is well documented, but cell-autonomous factors regulating neuronal and astroglial differentiation are incompletely understood. Here, we evaluate the functions of the transcription factor zinc-finger E-box binding homeobox 1 (ZEB1) in adult hippocampal RGL cells using a conditional-inducible mouse model. We find that ZEB1 is necessary for self-renewal of active RGL cells. Genetic deletion of Zeb1 causes a shift toward symmetric cell division that consumes the RGL cell and generates pro-neuronal progenies, resulting in an increase of newborn neurons and a decrease of newly generated astrocytes. We identify ZEB1 as positive regulator of the ets-domain transcription factor ETV5 that is critical for asymmetric division.

Keywords: Cre-loxP; EMT; animal model; astrogliogenesis; asymmetrical division; epithelial-mesenchymal transition; gliogenesis; lineage specification; neural stem cell; neurogenesis.

Graphical abstract

Figure 1

Expression of ZEB1 in the adult mouse hippocampus (A) Overview of whole DG with co-staining of GFAP and ZEB1. (B) GFAP and ZEB1 are co-expressed in SGZ RGL cells (B′) and in mature astrocytes (hilus, B″). (C) SOX9 and ZEB1 are co-expressed in RGL cells and astrocytes within the GCL. (D) Quiescent (q) RGL cells are mostly negative for ZEB1 (arrow in D′, left bar graph), while the majority of ZEB1⁺ cells in the SGZ constitute active (a) RGL cells (arrow in D″, middle bar graph) or IPCs (arrowhead in D′, right bar graph). (E) Fraction of qRGL versus aRGL cells out of all ZEB1⁺ RGL cells. (F and G) ZEB1 is absent in DCX⁺ neuroblasts (F, arrows), as well as in NeuN⁺ granule neurons (G, arrow). Dots represent individual mice (minimum of two sections analyzed per animal); numerical data are shown as mean ± SEM. Scale bars, 100 μm (A); 20 μm (B–D, F, and G); 10 μm (insets). GCL, granule cell layer; ML, molecular layer. See also Figure S1.

Figure 2

Validation of the conditional-inducible Zeb1 knockout model (A) Breeding strategy to generate inducible control and Zeb1 knockout mice with endogenous tdTOM reporter expression. (B) Tamoxifen (TAM) was administered daily for 5 consecutive days. Tissue was harvested at the indicated time points post-induction (black arrows). (C) Induction of tdTOM in GFAP⁺ RGL cells was comparable in both models. (D) Representative image depicting overlap of tdTOM in GFAP⁺ RGL cells within the SGZ. (E) Representative images showing RGL cells are ZEB1⁺ in control mice (arrow), but ZEB1⁻ following TAM administration in Zeb1^−/− mice (arrow). (F) Quantification of ZEB1 expression in GFAP⁺tdTOM⁺ RGL cells in the SGZ of control and Zeb1^−/− mice 1 day post-induction. (G) Comparison of ZEB1-expressing RGL cells at 4 weeks post-induction in control, Zeb1^+/−, and Zeb1^−/− mice. Dots represent individual mice (minimum of two sections analyzed per animal); numerical data are shown as mean ± SEM. Scale bars: 100 μm (D); 20 μm (F); 10 μm (insets).

Figure 3

Effects of Zeb1 loss in RGL cells and IPCs (A and B) Representative images of quiescent (GFAP⁺MCM2⁻tdTOM⁺; arrowheads) and activated (GFAP⁺MCM2⁺tdTOM⁺; arrows) RGL cells in the SGZ of control and Zeb1^−/− mice at 1 day (A) and 4 weeks (B) post-induction. (C) Numbers of quiescent RGL cells at 1 day and 4 weeks post-Zeb1 deletion. (D) Numbers of activated RGL cells at 1 day and 4 weeks post-Zeb1 deletion. (E and F) Representative images and quantification of RGL cells at 8 (E) and 12 weeks (F) post-induction. (G) Numbers of GFAP-MCM2⁺tdTOM⁺ IPCs at 1 day and 4 weeks post-Zeb1 deletion. (H and I) Representative images at 4 weeks post-induction (H) and quantification of TBR2⁺ IPCs at 2, 4, and 12 weeks post-Zeb1 deletion (I). (J–L) Summary graphs depict quiescent/activated (left) and total (right) RGL populations in control and Zeb1^−/− mice (J). Summary graphs outline the temporal changes of GFAP⁻MCM2⁺ (K) and TBR2⁺ (L) IPCs. Dots represent individual mice (minimum two sections analyzed per animal), except in (J)–(L), where dots represent averages. Numerical data are shown as mean ± SEM. Dashed lines in images demarcate DG boundaries. Scale bars: 20 μm (insets: 10 μm).

Figure 4

Effects of Zeb1 loss in newborn neurons (A) Representative images of DCX⁺tdTOM⁺ neuroblasts (arrows) at 2 weeks post-induction. (B) Quantification of DCX⁺tdTOM⁺ neuroblasts in the DG of control and Zeb1^−/− mice at 2, 4, and 8 weeks post-induction. (C) Summary graph of neuroblast changes over time. (D) Representative images of NeuN⁺ granule neurons (arrows) at 4 weeks post-induction. (E) Numbers of NeuN⁺tdTOM⁺ granule neurons at 4, 8, and 12 weeks post-induction. (F) Summary graph showing granule neuron changes over time. (G) 2D projection of a two-photon image z stack showing a typical granule cell filled with Alexa 488 via the patch-clamp recording electrode. Representative recordings from tdTOM expressing Zeb1^−/− (red) and control (blue) DGGCs. (H) Scatterplots show resting membrane potential (R_m), input resistance (R_N), membrane time constant (τ_m), and membrane capacitance (C_m) for individual DGGCs overlaid with the mean for each group. Additional graphs and Neurolucida traces are in Figure S3. Dots represent individual mice (minimum of two sections analyzed per animal; B and E), average values (C and F), or individual neurons (4 mice/genotype, H). Numerical data are shown as mean ± SEM. Scale bars: 20 μm (insets: 10 μm). See also Figures S2 and S3.

Figure 5

Effects of ZEB1 loss in astrocytes (A) Representative images at 4 weeks post-induction, identifying SOX9⁺ SGZ astrocytes (insets). (B) Representative images at 8 weeks post-induction, identifying S100β⁺ SGZ astrocytes (arrows). (C) Quantification of SOX9⁺tdTOM⁺ non-RGL astrocytes in the SGZ (left) at 4 and 8 weeks post-induction and of S100β⁺tdTOM⁺ SGZ astrocytes (right) at 8 and 12 weeks post-induction. (D) Quantification of SOX9⁺tdTOM⁺ (left) and S100β⁺tdTOM⁺ (right) astrocytes in the DG. (E) Fraction of apoptotic GFAP⁺ RGL at 1 day post-induction. (F) Fraction of SOX9⁺ apoptotic astrocytes at 4 weeks post-induction. Dots represent individual mice (minimum of two sections analyzed per animal); numerical data are shown as mean ± SEM. Scale bars: 20 μm (insets: 10 μm).

Figure 6

Analysis of RGL cell clones (A) Mice were injected with low-dose TAM (0.05 mg), and recombination was assessed 4 weeks post-induction. (B) Representative images of clones at 4 weeks post-induction. (C) Relative frequency of quiescent (containing only an RGL cell [single R]), active (containing an RGL cell and any other cell type [R+X]), and depleted (containing only lineage-restricted cells [no R]) clones (n = 38 [control, from 12 hippocampi] versus 35 [Zeb1^−/−, from 10 hippocampi]). (D) Frequencies of active clone subtypes (relative to all clones; n = 26 [control] versus 14 [Zeb1^−/−]). Clone subtypes are neurogenic (RGL cell and neurons [RN]), astrogliogenic (RGL cell and astrocyte [RA]), bi-lineage (RGL cell, neuron(s) and astrocyte [RAN]), or self-renewing (two RGL cells [RR]). (E) Frequencies of bi-lineage (RAN) versus neuron-only-producing (RN) clones across active clones (containing RGL cell; left) and all clones (right). (F) Ratio of clones containing two cells versus clones containing five cells. (G) Representative images of cleavage plane orientation in RGL cells undergoing asymmetric (top) or symmetric (bottom) division in control and Zeb1^−/− mice. Dashed lines indicate SGZ-hilus border. (H) Quantification of RGL cell division angles, binned into 30° groups. (I) Representative images from in vitro time-lapse imaging of primary adult hippocampal cells. (J) Quantification of dividing versus non-dividing clones. (K) Numbers of cell divisions per clone. (L) Numbers of surviving cells per clone. (M) Ratio of symmetric to asymmetric divisions across all clones. Dots represent individual clones from 6–7 mice/genotype (C and D), individual cells from 7–8 mice/genotype (H), and individual cells from 5–6 mice/genotype (K and L). Numerical data shown as mean ± SEM. Red line in (K) and (L) represents median. Scale bars: 10 μm (B and G); 20 μm (I). A, astrocyte; N, neuron; R, RGL cell. See also Figure S4.

Figure 7

ZEB1 directly regulates expression of ETV5 (A) Workflow for narrowing down the list of candidates relevant for asymmetric division. (B) Predicted ZEB1 binding sites with a p value of 10⁻³ (based on the JASPAR database; Fornes et al., 2020) in the Etv5 promoter region. (C) ChIP of the Etv5 promoter after pulldown with ZEB1 from hippocampal neurosphere cultures. (D) Western blot of ETV5 and ZEB1 in hippocampal Zeb1^−/− and control neurospheres. (E) Immunofluorescence staining for ETV5 in RGL cells of control and Zeb1^−/− mice. (F) Quantification of ETV5 expression in control RGL cells. (G) Quantification of ETV5⁺GFAP⁺tdTOM⁺ cells with RGL morphology in the SGZ at 1 day post-induction. (H) Percentage of ETV5⁺GFAP⁺tdTOM⁺ RGL cells out of total GFAP⁺tdTOM⁺ RGL cells at 1 day post-induction. (I) Western blot of ETV5 in hippocampal Zeb1^−/− neurospheres transduced with a lentiviral ETV5 expression vector and control Zeb1^−/− cultures. (J) Ratio of asymmetric to symmetric divisions quantified from time-lapse imaging of primary adult hippocampal cells. (K) Quantification of GFAP⁺ and DCX⁺ cells after live-cell imaging. Dots represent individual mice (minimum of two sections analyzed per animal); numerical data are shown as mean ± SEM. Scale bars: 10 μm. GBM, glioblastoma; HC, hippocampus. See also Figure S5 and Table S1.