ABSTRACT
Mutations in components of the exon junction complex (EJC) are associated with neurodevelopment and disease. In particular, reduced levels of the RNA helicase EIF4A3 cause Richieri-Costa-Pereira syndrome (RCPS) and copy number variations are linked to intellectual disability. Consistent with this, Eif4a3 haploinsufficient mice are microcephalic. Altogether, this implicates EIF4A3 in cortical development; however, the underlying mechanisms are poorly understood. Here, we use mouse and human models to demonstrate that EIF4A3 promotes cortical development by controlling progenitor mitosis, cell fate and survival. Eif4a3 haploinsufficiency in mice causes extensive cell death and impairs neurogenesis. Using Eif4a3;p53 compound mice, we show that apoptosis has the most impact on early neurogenesis, while additional p53-independent mechanisms contribute to later stages. Live imaging of mouse and human neural progenitors reveals that Eif4a3 controls mitosis length, which influences progeny fate and viability. These phenotypes are conserved, as cortical organoids derived from RCPS iPSCs exhibit aberrant neurogenesis. Finally, using rescue experiments we show that EIF4A3 controls neuron generation via the EJC. Altogether, our study demonstrates that EIF4A3 mediates neurogenesis by controlling mitosis duration and cell survival, implicating new mechanisms that underlie EJC-mediated disorders.
INTRODUCTION
The neocortex is responsible for higher-order processes, including cognition, memory and motor control. In mice, corticogenesis initiates at mid-gestation around embryonic day 10 (E10), when neuroepithelial progenitors divide symmetrically to expand the progenitor pool (Lodato and Arlotta, 2015; Silver, 2019). These cells then become radial glial cells (RGCs), which can either undergo self-renewing divisions to produce more RGCs or give rise to intermediate progenitors (IPs) and excitatory pyramidal neurons. IPs are also a major source of excitatory neurons. These excitatory neurons are produced in an ‘inside out’ fashion, with early-born neurons forming the deep layers (VI and V) and late-born neurons generating the superficial layers (IV-II/III). Radially migrating neurons use the RGC scaffold to reach their final location in the cortical plate (CP) (Silver, 2019; Vaid and Huttner, 2022). In humans, these key steps of corticogenesis are largely conserved, although there are some differences in cell behavior and composition. Disruption of these processes can result in severe neurodevelopmental disorders, including microcephaly and intellectual disability (Jayaraman et al., 2018).
Previous studies implicate the RNA-binding exon junction complex (EJC) in cortical development and disease. The EJC is composed of three core proteins, EIF4A3, MAGOH and RBM8A, and is a central regulator of RNA metabolism, including nonsense-mediated decay (NMD), RNA localization and translation (Fig. 1A) (Le Hir et al., 2016; Mao et al., 2016; Nott et al., 2004; Palacios et al., 2004). Mutations in core EJC components cause diverse neurodevelopmental disorders (Asthana et al., 2022; Martin et al., 2022; McMahon et al., 2016). EIF4A3 and RBM8A copy number gains and losses are both linked to intellectual disability (Nguyen et al., 2013). RBM8A loss-of-function mutations cause thrombocytopenia with absent radii (TAR syndrome), a congenital disorder associated with severe microcephaly (Albers et al., 2012). Additionally, decreased EIF4A3 expression underlies the developmental disorder Richieri-Costa-Pereira syndrome (RCPS), in which individuals exhibit craniofacial malformations as well as microcephaly (Bertola et al., 2018; Favaro et al., 2014; Hsia et al., 2018). This disorder is largely due to repetitive 18-20 nucleotide motifs in the 5′UTR of EIF4A3. How mutations in EJC components cause these neurodevelopmental disorders remains an important question. In particular, although neural crest cell dysfunction is implicated in the developmental etiology of RCPS (Miller et al., 2017), whether neurogenesis is impaired is unknown.
Genetic depletion of the EJC component Eif4a3 in mice causes microcephaly, substantial apoptosis and progenitor loss. (A) Cartoon illustrating canonical functions of the nucleo-cytoplasmic exon junction complex (EJC) in RNA metabolism. The RNA-binding EJC is composed of core components EIF4A3 (blue), MAGOH (orange) and RBM8A (purple). (B) Top: cartoon of a mouse embryo, embryonic brain and coronal cortical section. Bottom: cartoon of a human cortical organoid, section and dissociated organoid culture. (C) Simplified neurogenesis cartoon. At the onset of neurogenesis, neuroepithelial progenitors (purple) self-renew and then produce radial glial progenitor cells (RGCs, green), which can self-renew, generate neuron-producing intermediate progenitors (IPs, orange) or neurons (blue and pink). This study asks what are the cellular and developmental mechanisms by which EIF4A3 regulates neurogenesis? (D) Images of P0 whole-mount brains from indicated genotypes. (E) Images of E13.5 coronal sections stained using DAPI (white) with brain borders indicated by dotted yellow lines. (F) Representative images of E13.5 coronal sections stained using DAPI (gray), and for CC3 (green) and TUJ1 (purple). Zoomed panel is of the region outlined in yellow in the middle panel. (G) Quantification of TUJ1+ and CC3+ double-positive cells relative to all CC3+ cells at E13.5, n=3 embryos per genotype. (H,J) Representative images of E13.5 coronal sections stained with DAPI (blue) and (H) SOX2 (orange) or (J) TBR2 (green). Dotted lines indicate the cortex boundary. (I,K) Quantification of (I) SOX2+ and (K) TBR2+ cells relative to all cells (DAPI) at E13.5, n=6 embryos per genotype. (G,I,K) Unpaired two-tailed Student's t-test: **P<0.01, ****P<0.0001. Individual dots indicate individual biological replicates. Data are mean±s.d. Scale bars: 1000 µm in D; 200 µm in E; 100 µm in F (left and middle), H, J; 10 µm in F (right).
Genetic depletion of the EJC component Eif4a3 in mice causes microcephaly, substantial apoptosis and progenitor loss. (A) Cartoon illustrating canonical functions of the nucleo-cytoplasmic exon junction complex (EJC) in RNA metabolism. The RNA-binding EJC is composed of core components EIF4A3 (blue), MAGOH (orange) and RBM8A (purple). (B) Top: cartoon of a mouse embryo, embryonic brain and coronal cortical section. Bottom: cartoon of a human cortical organoid, section and dissociated organoid culture. (C) Simplified neurogenesis cartoon. At the onset of neurogenesis, neuroepithelial progenitors (purple) self-renew and then produce radial glial progenitor cells (RGCs, green), which can self-renew, generate neuron-producing intermediate progenitors (IPs, orange) or neurons (blue and pink). This study asks what are the cellular and developmental mechanisms by which EIF4A3 regulates neurogenesis? (D) Images of P0 whole-mount brains from indicated genotypes. (E) Images of E13.5 coronal sections stained using DAPI (white) with brain borders indicated by dotted yellow lines. (F) Representative images of E13.5 coronal sections stained using DAPI (gray), and for CC3 (green) and TUJ1 (purple). Zoomed panel is of the region outlined in yellow in the middle panel. (G) Quantification of TUJ1+ and CC3+ double-positive cells relative to all CC3+ cells at E13.5, n=3 embryos per genotype. (H,J) Representative images of E13.5 coronal sections stained with DAPI (blue) and (H) SOX2 (orange) or (J) TBR2 (green). Dotted lines indicate the cortex boundary. (I,K) Quantification of (I) SOX2+ and (K) TBR2+ cells relative to all cells (DAPI) at E13.5, n=6 embryos per genotype. (G,I,K) Unpaired two-tailed Student's t-test: **P<0.01, ****P<0.0001. Individual dots indicate individual biological replicates. Data are mean±s.d. Scale bars: 1000 µm in D; 200 µm in E; 100 µm in F (left and middle), H, J; 10 µm in F (right).
In mouse models, haploinsufficiency for each of the core EJC components in RGCs causes microcephaly, suggesting that defective neurogenesis may contribute to EJC disorders (Mao et al., 2015, 2016; Silver et al., 2010). Transcriptomic and proteomic analyses of these embryonic brains reveal converging dysregulation of common pathways, including p53 (Mao et al., 2016). Furthermore, p53 (Trp53) genetic deletion in the Eif4a3, Magoh and Rbm8a mutants indicate apoptosis contributes to the microcephaly, although the mechanisms are unknown (Mao et al., 2016). Collectively, these common phenotypes suggest core EJC components may work together to control neurogenesis. Previous studies of Magoh mutants also implicate mitosis dysregulation in the development of microcephaly, but whether this is the case for all EJC components is unknown. This is of interest, given that all three core components are essential for mitosis in immortalized cells (Silver et al., 2010).
In this study, we use mouse models as well as human cortical organoids to understand how reduced levels of Eif4a3, the helicase component of the EJC, influence neurogenesis and cause RCPS (Fig. 1B,C). Using live imaging of mouse and human progenitors, we discover conserved roles for Eif4a3 in mitosis, cell fate and cell death. We find that p53 ablation in Eif4a3 haploinsufficient mutant brains rescues deep-layer neuron number, but not upper-layer neurons, suggesting that both p53-dependent and -independent pathways govern progenitor behavior and cell composition. Using rescue experiments, we establish EJC-dependent roles of EIF4A3 in neurogenesis. Our study reveals essential roles for cortical progenitor mitosis duration in EIF4A3-mediated neurodevelopmental pathologies.
RESULTS
Eif4a3 haploinsufficiency in mice causes defects in neurogenesis
We have previously shown that Eif4a3 haploinsufficiency leads to microcephaly and to altered progenitor and neuron number at early stages of neurogenesis (E11.5 and E12.5) (Mao et al., 2016). To further understand how EIF4A3 controls neurogenesis and contributes to microcephaly, we employed these previously generated Eif4a3lox/+ mice and crossed them to Emx1-Cre. This strategy removes one copy of Eif4a3 from RGCs and their progeny, beginning at E9.5 (Chou et al., 2009; Gorski et al., 2002), resulting in Eif4a3 conditional heterozygous brains (Eif4a3 cHET). We analyzed Eif4a3 cHET brains at postnatal day 0 (P0) to assess gross cortical size at the end of neurogenesis. At P0, whole-mount Eif4a3 cHET (Emx1-Cre;Eif4a3lox/+) brains were severely microcephalic (Fig. 1D). This corroborates our previous findings at E18.5, indicating that Eif4a3 is crucial for proper brain size (Mao et al., 2016).
We next evaluated neurogenesis and cortical size, focusing on mid-neurogenesis stages. The cortices of E13.5 Eif4a3 cHET brains were markedly thinner (Fig. 1E). As microcephaly is frequently associated with cell death, we assessed apoptosis in the Eif4a3 cHET brains. Immunostaining for cleaved caspase 3 (CC3) revealed massive apoptosis in Eif4a3 cHET cortices (Fig. 1F). Quantification of CC3 and TUJ1 (β-tubulin, a neuronal marker) double-positive cells relative to all CC3-positive cells revealed apoptosis of both neurons (TUJ1+) and progenitors (TUJ1−), with a slight bias to progenitors at this stage (Fig. 1G). In Magoh mutant brains, apoptosis is associated with increased DNA damage (Silver et al., 2010). Consistent with this, immunostaining for γH2AX in E13.5 Eif4a3 haploinsufficient sections revealed extensive DNA damage (Fig. S1A).
Next, to analyze the impact of Eif4a3 haploinsufficiency upon neurogenesis, we quantified RGC and IP populations. E13.5 Eif4a3 cHET brains showed a 20% reduction in the composition of RGCs (SOX2+/DAPI) compared with control brains (Fig. 1H,I). Moreover, the composition of IPs (TBR2+/DAPI) was significantly reduced by 36% (Fig. 1J,K). In addition, the overall number of both progenitor populations was reduced (Fig. S1B,C). These findings corroborate previous observations of apoptosis and neurogenesis defects in younger Eif4a3 haploinsufficient brains (Mao et al., 2016). Taken together, these analyses indicate that Eif4a3 controls progenitor number and viability of cortical cells during different stages of neurogenesis.
Eif4a3 haploinsufficient progenitors are delayed in mitosis and undergo fewer viable divisions
We next sought to understand how EIF4A3 controls progenitor number. EIF4A3, MAGOH and RBM8A are each essential for mitosis in immortalized cells and EJC cHET E11.5 brains each have increased mitotic indexes (Mao et al., 2015, 2016; Silver et al., 2010). Additionally, Magoh haploinsufficiency prolongs progenitor mitosis, correlated with altered neurogenesis (Pilaz et al., 2016; Sheehan et al., 2020). Moreover, pharmacology studies indicate this prolonged mitosis can causally impact cell fate in the developing cortex (Mitchell-Dick et al., 2019; Pilaz et al., 2016). Given this, we postulated that EIF4A3 may likewise influence progenitor number by affecting mitosis.
To directly test this possibility, we employed a live imaging paradigm established in our lab (Mitchell-Dick et al., 2019; Pilaz et al., 2016; Sheehan et al., 2020) (Fig. 2A). Primary progenitor cultures were generated from control and Eif4a3 cHET cortices at E12.5 (Pilaz et al., 2016), a stage when neurogenesis and apoptosis defects are evident (Mao et al., 2016). Live imaging was performed for 20 h, with images captured every 10 min. The mean mitosis duration in Eif4a3 cHET progenitors was 1.8-fold longer than control, taking 47 min on average compared with 26 min (Fig. 2B,C). This is consistent with the higher mitotic index of Eif4a3 cHET brains (Mao et al., 2016) and the 1.8-fold longer mitoses of Magoh haploinsufficient progenitors relative to control at E12.5 (Pilaz et al., 2016). Taken together, these analyses demonstrate that Eif4a3 haploinsufficient progenitors exhibit prolonged mitosis in vitro, in addition to mitotic defects in vivo (Mao et al., 2016), similar to Magoh mutants (Pilaz et al., 2016; Silver et al., 2010).
Live imaging reveals Eif4a3 haploinsufficient progenitors are delayed in mitosis and undergo fewer viable divisions. (A) Live-imaging paradigm for monitoring the fate of progenitors. Adapted, with permission, from Pilaz et al. (2016). (B,C) Quantification of average mitosis duration (B) and mitosis distribution (C) for control (black) and Eif4a3 cHET (blue) progenitors. (D) Schematic illustrating mitotic catastrophe (unsuccessful mitosis and subsequent cell death) with live imaging DIC snapshots, at indicated times (t; in min). (E,F) Quantification of the proportion of successful divisions (white) and mitotic catastrophe (purple) among (E) all progeny or (F) relative to mitosis duration. (B) Mann–Whitney two-tailed non-parametric test: ****P<0.0001. Violin plots with median±quartiles. (C,E,F) χ2 analysis with post-hoc Bonferroni adjusted P-values represented by asterisks. ***P<0.001; ns, not significant. Data are mean±s.e.m. Two live-imaging sessions and two litters; n=4 control embryos (126 cells); n=5 cHET embryos (199 cells). Scale bars: 10 μm.
Live imaging reveals Eif4a3 haploinsufficient progenitors are delayed in mitosis and undergo fewer viable divisions. (A) Live-imaging paradigm for monitoring the fate of progenitors. Adapted, with permission, from Pilaz et al. (2016). (B,C) Quantification of average mitosis duration (B) and mitosis distribution (C) for control (black) and Eif4a3 cHET (blue) progenitors. (D) Schematic illustrating mitotic catastrophe (unsuccessful mitosis and subsequent cell death) with live imaging DIC snapshots, at indicated times (t; in min). (E,F) Quantification of the proportion of successful divisions (white) and mitotic catastrophe (purple) among (E) all progeny or (F) relative to mitosis duration. (B) Mann–Whitney two-tailed non-parametric test: ****P<0.0001. Violin plots with median±quartiles. (C,E,F) χ2 analysis with post-hoc Bonferroni adjusted P-values represented by asterisks. ***P<0.001; ns, not significant. Data are mean±s.e.m. Two live-imaging sessions and two litters; n=4 control embryos (126 cells); n=5 cHET embryos (199 cells). Scale bars: 10 μm.
We next quantified the ability of Eif4a3 cHET progenitors to complete mitosis. We used the term ‘successful division’ to define progenitor divisions that generated two daughter cells and the term ‘mitotic catastrophe’ when dividing progenitors underwent a prolonged mitosis and subsequently either senesced or died (Fig. 2D) (Blagosklonny, 2007; Castedo et al., 2004). In comparison with control, 82% of Eif4a3 cHET progenitors successfully completed mitosis, with the remainder showing mitotic catastrophe (Fig. 2E). Given the link between mitotic delay and mitotic catastrophe (Sheehan et al., 2020), we next asked whether progenitors undergoing catastrophe were mitotically delayed (Fig. 2F). The vast majority of control progenitors completed mitosis in 30 min or less and were hence classified as ‘normal division’ (Fig. 2C). Mitotically delayed Eif4a3 cHET progenitors were significantly more likely to undergo mitotic catastrophe, with 28% of delayed cHET progenitors failing to complete mitosis compared with both the non-delayed cHET (1%) and control progenitors (0%) (Fig. 2F). Altogether, these findings indicate that Eif4a3 controls progenitor mitosis duration and that mitotically delayed progenitors are more likely to die before division.
Mitotically delayed Eif4a3 haploinsufficient progenitors generate fewer progenitors, more neurons and apoptotic progeny
The vast majority of Eif4a3 cHET progenitors successfully completed mitosis. Given the neurogenesis defects observed in vivo (Fig. 1) and links between mitosis duration and altered cell fate (Mitchell-Dick et al., 2019; Pilaz et al., 2016; Sheehan et al., 2020), we assessed whether Eif4a3 cHET progenitors delayed in mitosis produced altered progeny. To achieve this, we quantified viability of newborn progeny after live imaging by assaying their morphology using DIC microscopy (Fig. 3A). Overall, Eif4a3 cHET progenitors produced significantly more apoptotic progeny than control progenitors (Fig. 3B,C), and this significantly correlated with mitotic delay (Fig. 3D). These data show that Eif4a3 is required for the generation of viable progeny and that this correlates with mitotic duration.
Mitotically delayed Eif4a3 haploinsufficient progenitors produce apoptotic progeny, more neurons and fewer progenitors. (A) Live imaging DIC snapshots at indicated times (t; in min) depicting a dividing progenitor that produces apoptotic progeny, with a cartoon representation. (B) Distribution of divisions with apoptotic progeny relative to mitosis duration for control (white) and Eif4a3 cHET (blue) progenitors. (C,D) Quantification of progenitors with surviving progeny (white) or progeny undergoing apoptosis (red) relative to mitosis duration. (E) Schematic illustrating proliferative (SOX2+ or TBR2+ progeny), asymmetric neurogenic (SOX2+ and TUJ1+ progeny; TBR2+ and TUJ1+ progeny) and symmetric neurogenic divisions (both TUJ1+ progeny), with representative images of live and fixed analysis of progeny cell fate. (F) Distribution of divisions with neuronal progeny relative to mitosis duration for control (white) and Eif4a3 cHET (blue) progenitors. (G) Quantification of the proportion of proliferative (purple), asymmetric neurogenic (blue) or symmetric neurogenic (orange) divisions among (G) all progeny or (H) relative to mitosis duration. χ2 analysis with post-hoc Bonferroni adjusted P-values represented by asterisks; *P<0.05, **P<0.01, ***P<0.001; ns, not significant. Data are mean±s.e.m. Two live-imaging sessions, two litters; n=4 control embryos (126 cells); n=5 cHET embryos (199 cells). Scale bars: 10 μm.
Mitotically delayed Eif4a3 haploinsufficient progenitors produce apoptotic progeny, more neurons and fewer progenitors. (A) Live imaging DIC snapshots at indicated times (t; in min) depicting a dividing progenitor that produces apoptotic progeny, with a cartoon representation. (B) Distribution of divisions with apoptotic progeny relative to mitosis duration for control (white) and Eif4a3 cHET (blue) progenitors. (C,D) Quantification of progenitors with surviving progeny (white) or progeny undergoing apoptosis (red) relative to mitosis duration. (E) Schematic illustrating proliferative (SOX2+ or TBR2+ progeny), asymmetric neurogenic (SOX2+ and TUJ1+ progeny; TBR2+ and TUJ1+ progeny) and symmetric neurogenic divisions (both TUJ1+ progeny), with representative images of live and fixed analysis of progeny cell fate. (F) Distribution of divisions with neuronal progeny relative to mitosis duration for control (white) and Eif4a3 cHET (blue) progenitors. (G) Quantification of the proportion of proliferative (purple), asymmetric neurogenic (blue) or symmetric neurogenic (orange) divisions among (G) all progeny or (H) relative to mitosis duration. χ2 analysis with post-hoc Bonferroni adjusted P-values represented by asterisks; *P<0.05, **P<0.01, ***P<0.001; ns, not significant. Data are mean±s.e.m. Two live-imaging sessions, two litters; n=4 control embryos (126 cells); n=5 cHET embryos (199 cells). Scale bars: 10 μm.
We next assessed the extent to which Eif4a3 haploinsufficiency impacts the fate of surviving progeny and whether there is a link to mitotic duration. To achieve this, fixed progeny were stained after live imaging to identify progenitor divisions as either proliferative (only SOX2+ or TBR2+ progeny), asymmetric neurogenic (SOX2+ and TUJ1+ progeny; TBR2+ and TUJ1+ progeny) or symmetric neurogenic (both TUJ1+ progeny) divisions (Fig. 3E). In comparison with E12.5 control and non-delayed Eif4a3 cHET progenitors, delayed Eif4a3 cHET progenitors generated significantly more neuronal progeny (Fig. 3F-H), phenocopying Magoh+/− mutant progenitors and progenitors treated with mitotic inhibitors (Mitchell-Dick et al., 2019; Pilaz et al., 2016). This phenotype was most striking in delayed Eif4a3 cHET progenitors compared with non-delayed or control progenitors (Fig. 3H). Cell fate was also further analyzed in subcategories to discriminate RGC and IP, generating divisions that revealed that delayed progenitors underwent significantly fewer symmetric RGC divisions and more direct symmetric neurogenic divisions (Fig. S1D-F). Although there were trends towards fewer IP-generating symmetric divisions in delayed Eif4a3 cHET progenitors (Fig. S1E,F), this difference was insignificant, likely due to the low number of IP divisions at this stage. Taken together, these data demonstrate that delayed mitosis in Eif4a3 haploinsufficient progenitors is associated with altered progeny cell fate and viability.
p53-dependent and -independent mechanisms explain Eif4a3 cHET microcephaly and neurogenesis
Our data collectively indicate that Eif4a3 controls progenitor mitosis, which is associated with altered composition of progenitors and neurons and significant apoptosis (Mao et al., 2016). A key question is the extent to which cell composition and cortical thickness changes in Eif4a3 haploinsufficient brains can be explained by apoptosis. To disentangle cell fate and cell death, we crossed Emx1-Cre;Eif4a3lox/+ (Mao et al., 2016) mice onto a p53lox/lox background, which eliminates apoptosis (Marino et al., 2000). Indeed, apoptosis in p53;Eif4a3 compound mutant (Emx1-Cre;Eif4a3lox/+;p53lox/lox) P0 brains was similar to that of controls but distinct from the cHET, which showed many CC3+ cells, particularly in the CP (Fig. S2A-C). There were no significant differences in cortical size between Emx1-Cre;p53lox/lox cKO and control brains (either Emx1-Cre; Eif4a3+/+ or Emx1-Cre; Eif4a3+/+;p53lox/+) (Fig. S2D,E). This indicates that P53 loss alone does not grossly affect cortical size. In contrast, as expected, Eif4a3 cHET P0 brains were strikingly reduced in size, measuring, on average, 55% smaller than the Emx1-Cre controls and missing most of the pallium (Fig. 4A,B). Cortical thickness in cHET brains was also significantly reduced by 54% relative to controls (Fig. 4C). However, neither cortical area nor thickness was fully restored in the compound mutant, suggesting that additional factors independent of p53 and apoptosis contribute to cortical size (Fig. 4A-C).
Loss of p53 partially rescues microcephaly and neuron number associated with Eif4a3 haploinsufficiency. (A) Images of P0 whole-mount brains from indicated genotypes (top). Low-magnification images of P0 coronal sections, stained for DAPI (white). (B,C) Quantification of cortical area (B) and cortical thickness (C) in P0 brains with indicated genotypes. (D,F,H,J) Representative P0 sections of indicated genotypes from the areas outlined in magenta in A stained for (D) TBR1 (green), (F) CTIP2 (magenta), (H) RORβ (green) and (J) LHX2 (green). (E,G,I,K) Quantification of (E) TBR1+, (G) CTIP2+, (I) RORβ+, (K) LHX2+ cell density at P0. Dotted lines indicate the dorsal cortex boundary. ANOVA with Tukey post-hoc: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; ns, not significant. Individual dots indicate individual biological replicates. (E) n=3 control, n=4 cHET and n=3 p53;Eif4a3 compound mutants. (G) n=5 ctrl, n=4 cHET and n=4 p53;Eif4a3 compound mutants. (I) n=4 ctrl, n=3 cHET and n=3 p53;Eif4a3 compound mutants. (K) n=4 ctrl, n=5 cHET and n=4 p53;Eif4a3 compound mutants. Data are mean±s.d. Scale bars: 1000 µm in A (top); 500 µm in A (bottom); 100 µm in D,F,H,J.
Loss of p53 partially rescues microcephaly and neuron number associated with Eif4a3 haploinsufficiency. (A) Images of P0 whole-mount brains from indicated genotypes (top). Low-magnification images of P0 coronal sections, stained for DAPI (white). (B,C) Quantification of cortical area (B) and cortical thickness (C) in P0 brains with indicated genotypes. (D,F,H,J) Representative P0 sections of indicated genotypes from the areas outlined in magenta in A stained for (D) TBR1 (green), (F) CTIP2 (magenta), (H) RORβ (green) and (J) LHX2 (green). (E,G,I,K) Quantification of (E) TBR1+, (G) CTIP2+, (I) RORβ+, (K) LHX2+ cell density at P0. Dotted lines indicate the dorsal cortex boundary. ANOVA with Tukey post-hoc: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; ns, not significant. Individual dots indicate individual biological replicates. (E) n=3 control, n=4 cHET and n=3 p53;Eif4a3 compound mutants. (G) n=5 ctrl, n=4 cHET and n=4 p53;Eif4a3 compound mutants. (I) n=4 ctrl, n=3 cHET and n=3 p53;Eif4a3 compound mutants. (K) n=4 ctrl, n=5 cHET and n=4 p53;Eif4a3 compound mutants. Data are mean±s.d. Scale bars: 1000 µm in A (top); 500 µm in A (bottom); 100 µm in D,F,H,J.
We next used this genetic paradigm to assess the extent to which apoptosis explains altered neuronal composition of Eif4a3 haploinsufficient brains. To achieve this, we first quantified deep-layer early-born neurons (Greig et al., 2013) (Fig. 4D-K). In the Eif4a3 cHET brains, immunostaining revealed 37% fewer TBR1+ (layer VI) and 45% fewer CTIP2+ (layer V/VI) deep-layer neurons compared with control (Fig. 4D-G). Notably, both neuron populations were significantly rescued in the p53;Eif4a3 compound mutant brains, measured by composition or by total number of cells (Fig. 4D-G, Fig. S2F,G). The distribution of these neurons in cHET mutants was comparable with controls, with some slight differences relative to the p53;Eif4a3 compound mutant (Fig. S2J,K). Altogether, these data suggest that reduced numbers of deep-layer neurons are largely explained by p53 activation and apoptosis.
To examine later born superficial layer neurons (Greig et al., 2013), we performed immunostaining for RORβ (layer IV) and LHX2 (layer II/III) (Fig. 4H-K). Compared with control, Eif4a3 cHET brains had 61% fewer RORβ+ neurons, while p53;Eif4a3 compound mutant brains still had 53% fewer RORβ+ neurons, indicating these neurons were not significantly rescued (Fig. 4H-K). Similarly, LHX2+ neurons were markedly reduced by 70% in Eif4a3 cHET brains compared with control, but only partially rescued in the p53;Eif4a3 compound mutant brains (41% fewer compared with control) (Fig. 4J-K). Similar trends were found when the overall number of neurons was quantified (Fig. S2H,I). Thus, the populations of both LHX2 and RORβ neurons were partially, but not completely, rescued in the p53 mutant background. The distribution of LHX2 was altered in cHET mutants compared with controls, showing more neurons in both outer two bins. This was rescued in the compound mutant (Fig. 2SL,M). Taken together, these results indicate that loss and organization of superficial neurons in Eif4a3 mutants is due, in part, to p53-mediated cell death but, additionally, that p53-independent mechanisms contribute to their genesis.
We next sought to understand whether differential cell death of progenitors at early and later stages of development explains the subtype-specific neuronal phenotypes at P0. The genesis of layer VI and V neurons peaks from E12.5 to E13.5, and genesis of layers IV and II/III peaks from E14.5 to E16.5 (Greig et al., 2013). Thus, we first used our compound mutant brains to evaluate progenitors at E13.5 to understand genesis of deep-layer neurons. Staining for CC3 in the p53;Eif4a3 compound mutant brains at E13.5 verified that apoptosis is fully rescued at this stage (Fig. S3A). Likewise, the cortical thickness of control and p53;Eif4a3 brains was similar (Fig. 5A,B). The composition and total number of RGCs and IPs were also both rescued in the p53;Eif4a3 compound brains (Fig. 5C-F, Fig. S3B,C). These findings show that in E13.5 Eif4a3 haploinsufficient brains, p53-mediated cell death is largely responsible for reduced cortical thickness and progenitor number.
Genetic p53 ablation differentially rescues Eif4a3-dependent cortical thickness and progenitors at E13.5 versus E16.5. (A) Low-magnification images of E13.5 coronal sections, stained using DAPI (white). (B) Quantification of E13.5 cortical thickness for indicated genotypes. n=4 ctrl, n=3 cHET and n=6 p53;Eif4a3 compound mutants. (C,D) Representative images of E13.5 coronal sections stained for (C) SOX2 (orange) and DAPI (blue) (C) or for TBR2 (green) and DAPI (blue) (D). (E,F) Quantification of (E) SOX2+ cells and (F) TBR2+ cell density at E13.5. (E) n=10 ctrl, n=9 cHET and n=7 p53;Eif4a3 compound mutants. (F) n=6 ctrl, n=9 cHET and n=7 p53;Eif4a3 compound mutants. (G) Low-magnification images of E16.5 coronal sections, stained for DAPI (white). (H) Quantification of E16.5 cortical thickness for indicated genotypes. n=7 ctrl, n=4 cHET and n=3 p53;Eif4a3 compound mutants. (I,K) Representative images of E16.5 coronal sections stained for SOX2 (magenta) and DAPI (white) (I) or for TBR2 (green) and DAPI (white) (K). (J,L) Quantification of (J) SOX2+ or (L) TBR2+ cell density at E16.5. n=7 ctrl, n=8 cHET and n=5 p53;Eif4a3 compound mutants. Dotted lines indicate the dorsal cortex boundary. ANOVA with Tukey post-hoc: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001, ns, not significant. Individual dots indicate individual biological replicates. Data are mean±s.d. Scale bars: 200 µm in A; 100 µm in C,D,G,I,K.
Genetic p53 ablation differentially rescues Eif4a3-dependent cortical thickness and progenitors at E13.5 versus E16.5. (A) Low-magnification images of E13.5 coronal sections, stained using DAPI (white). (B) Quantification of E13.5 cortical thickness for indicated genotypes. n=4 ctrl, n=3 cHET and n=6 p53;Eif4a3 compound mutants. (C,D) Representative images of E13.5 coronal sections stained for (C) SOX2 (orange) and DAPI (blue) (C) or for TBR2 (green) and DAPI (blue) (D). (E,F) Quantification of (E) SOX2+ cells and (F) TBR2+ cell density at E13.5. (E) n=10 ctrl, n=9 cHET and n=7 p53;Eif4a3 compound mutants. (F) n=6 ctrl, n=9 cHET and n=7 p53;Eif4a3 compound mutants. (G) Low-magnification images of E16.5 coronal sections, stained for DAPI (white). (H) Quantification of E16.5 cortical thickness for indicated genotypes. n=7 ctrl, n=4 cHET and n=3 p53;Eif4a3 compound mutants. (I,K) Representative images of E16.5 coronal sections stained for SOX2 (magenta) and DAPI (white) (I) or for TBR2 (green) and DAPI (white) (K). (J,L) Quantification of (J) SOX2+ or (L) TBR2+ cell density at E16.5. n=7 ctrl, n=8 cHET and n=5 p53;Eif4a3 compound mutants. Dotted lines indicate the dorsal cortex boundary. ANOVA with Tukey post-hoc: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001, ns, not significant. Individual dots indicate individual biological replicates. Data are mean±s.d. Scale bars: 200 µm in A; 100 µm in C,D,G,I,K.
We next examined E16.5, when progenitors produce layer IV and II/III neurons (Greig et al., 2013). We first measured cortical area of whole-mount brains and thickness in cortical sections. The cortex was almost entirely absent in E16.5 Eif4a3 cHET brains and contained CC3+ apoptotic cells (Fig. 5G, Fig. S3D). In the p53;Eif4a3 compound mutants, although apoptosis was not evident (Fig. S3D), cortical thickness was only partially restored compared with control (Fig. 5H). This suggests that, unlike at E13.5, at E16.5, apoptosis is insufficient to fully explain the microcephaly associated with Eif4a3 haploinsufficiency.
We next assessed the extent to which apoptosis explains reduced RGC and IP populations in Eif4a3 haploinsufficient brains at E16.5 (Fig. 5I-L). The composition of both progenitor populations was significantly reduced in the Eif4a3 cHET brains, consistent with E13.5 (Fig. 5E,F,J,L). In the p53;Eif4a3 compound mutant brains, IPs were significantly rescued in both cell density and total number (Fig. 5K,L, Fig. S3G). In contrast, RGCs were only partially rescued (Figs 5I,J, 3F). Collectively, these data suggest that the reduction of upper-layer neurons at P0 is likely due to both apoptosis and additional defects in cell fate specification. Altogether, our data provide evidence of p53-dependent and p53-independent mechanisms of EIF4A3 over the course of development.
EIF4A3 depletion from human NPCs results in fewer proliferative and viable divisions
Having found that Eif4a3 is essential for neurogenesis in our mouse model, we next asked whether this requirement is conserved in humans. To achieve this, we dissociated neural progenitor cells (NPCs) from D35 cortical organoids, electroporated these with siRNAs against EIF4A3 or scrambled siRNAs (control), and allowed 2 days for knockdown (Fig. 6A). We quantified a 68% reduction in EIF4A3 in electroporated cells after 48 h (Fig. 6B). We then quantified mitosis and neurogenesis of individual progenitors using a similar live imaging paradigm to that in Fig. 3. Notably, EIF4A3-deficient hNPCs were significantly delayed in mitosis compared with control siRNA-treated progenitors (Fig. 6C). The mean mitosis duration of control NPCs was 36 min, whereas it was 54 min in EIF4A3-deficient NPCs. To quantify a potential link between mitosis duration and progeny of these progenitors, we classified normal hNPC mitosis duration as ≤40 min, as 91% of control cells completed mitosis in 40 min or less (Fig. 6D). Eighty-eight percent of EIF4A3-deficient NPC divisions that produced apoptotic progeny were mitotically delayed (Fig. 6E). Indeed, in comparison with either control NPCs or non-delayed EIF4A3-deficient NPCs, delayed EIF4A3-deficient NPCs produced significantly more apoptotic progeny (Fig. 6F,G). Mitotic delay was also associated with altered cell fate, evidenced by immunostaining progeny for Ki67 (proliferative marker) and Tuj1 (neuronal marker). Delayed EIF4A3-deficient hNPCs produced significantly more neuronal progeny compared with non-delayed progenitors (Fig. 6H,I). These findings parallel the mouse data, demonstrating that EIF4A3 controls mitosis of human neural progenitors, which is linked to altered fate of progeny. Hence, the requirement of EIF4A3 for neurogenesis is evolutionarily conserved.
EIF4A3 depletion from human NPCs causes mitotic delay and altered progeny fate and survival. (A) Schematic depicting the paradigm for live imaging of human neural progenitors (hNPCs) dissociated from day 35 cortical organoids and electroporated with control (scrambled) or EIF4A3 siRNAs for 48 h, imaged for 24 h and subsequently stained to assign progeny cell fate. (B) qPCR quantification of EIF4A3 mRNA levels, normalized to TBP, in siRNA-treated hNPCs. (C,D) Quantification of average mitosis duration (C) and distribution (D) for control (black) and EIF4A3 siRNA-treated (blue) hNPCs. (E) Distribution of divisions with apoptotic progeny relative to mitosis duration of control (black) and EIF4A3 siRNA-treated (blue) hNPCs. (F,G) Quantification of the proportion of hNPC divisions with progeny surviving (white) or undergoing apoptosis (red) (F), and relative to mitosis duration (G). (H) Distribution of divisions with neuronal progeny relative to mitosis duration for control (black) and EIF4A3 siRNA-treated (blue) hNPCs. (I) Quantification of the proportion of proliferative (purple), asymmetric neurogenic (blue) or symmetric neurogenic (orange) divisions relative to mitosis duration. (B) Unpaired two-tailed Student's t-test: ****P<0.0001. (C) Mann–Whitney two-tailed non-parametric test: ****P<0.0001. Violin plots of median±quartiles. (D-I) χ2 analysis with post-hoc Bonferroni adjusted P-values. *P<0.05, **P<0.01, ***P<0.001; ns, not significant. Data are mean±s.e.m. There were two live-imaging sessions and four independent electroporations per genotype; n=58 control cells; n=63 EIF4A3 KD cells.
EIF4A3 depletion from human NPCs causes mitotic delay and altered progeny fate and survival. (A) Schematic depicting the paradigm for live imaging of human neural progenitors (hNPCs) dissociated from day 35 cortical organoids and electroporated with control (scrambled) or EIF4A3 siRNAs for 48 h, imaged for 24 h and subsequently stained to assign progeny cell fate. (B) qPCR quantification of EIF4A3 mRNA levels, normalized to TBP, in siRNA-treated hNPCs. (C,D) Quantification of average mitosis duration (C) and distribution (D) for control (black) and EIF4A3 siRNA-treated (blue) hNPCs. (E) Distribution of divisions with apoptotic progeny relative to mitosis duration of control (black) and EIF4A3 siRNA-treated (blue) hNPCs. (F,G) Quantification of the proportion of hNPC divisions with progeny surviving (white) or undergoing apoptosis (red) (F), and relative to mitosis duration (G). (H) Distribution of divisions with neuronal progeny relative to mitosis duration for control (black) and EIF4A3 siRNA-treated (blue) hNPCs. (I) Quantification of the proportion of proliferative (purple), asymmetric neurogenic (blue) or symmetric neurogenic (orange) divisions relative to mitosis duration. (B) Unpaired two-tailed Student's t-test: ****P<0.0001. (C) Mann–Whitney two-tailed non-parametric test: ****P<0.0001. Violin plots of median±quartiles. (D-I) χ2 analysis with post-hoc Bonferroni adjusted P-values. *P<0.05, **P<0.01, ***P<0.001; ns, not significant. Data are mean±s.e.m. There were two live-imaging sessions and four independent electroporations per genotype; n=58 control cells; n=63 EIF4A3 KD cells.
Human RCPS cortical organoids exhibit altered neurogenesis
Human EIF4A3 hypomorphic mutations cause Richieri-Costa-Pereira syndrome (RCPS), a rare developmental disorder associated with microcephaly, and language and learning disability, among other clincial presentations (Bertola et al., 2018; Favaro et al., 2014; Hsia et al., 2018). How reduced EIF4A3 expression causes RCPS neurological defects is largely unknown. To address this gap, we used previously characterized patient-derived iPSCs to generate cortical organoids that model early human brain development (Alsina et al., 2022 preprint; Miller et al., 2017; Yoon et al., 2019) (Fig. 7A). We have previously shown that in day 35 (D35) organoids, EIF4A3 is reduced by about 50% (Alsina et al., 2022 preprint). To assess progenitor number, we immunostained organoid sections for SOX2. RCPS organoids at D35 had 35% fewer RGCs compared with control (Fig. 7B,C). The RGCs remaining in RCPS organoids showed an increased mitotic index (1.5-fold) compared with control. Thus, although RCPS patient-derived organoids contain fewer progenitors, the surviving progenitors show defects in mitosis.
EIF4A3 RCPS human cortical organoids exhibit neurogenesis defects. (A) Schematic representing the generation of cortical organoids and organoid sections from iPSCs derived from three control people (ctrl; green, light blue and dark blue), one isogenic line generated from a patient (black) and three people with Richieri-Costa-Pereira syndrome (RCPS) (orange, pink and purple). Modified, with permission, from Alsina et al. (2022 preprint). (B) Representative images of ctrl and RCPS D35 organoids stained for SOX2 (magenta), for PH3 (green) and with DAPI (blue). The area shown at higher magnification on the right is outlined in yellow in the left image. (C,D) Quantification of SOX2+ cells relative to all DAPI+ cells (C) or PH3+ SOX2+ cells relative to all SOX2+ cells (D) in organoid sections. (C) n=9 control and n=10 RCPS. (D) n=10 ctrl and n=11 RCPS. (E) Schematic illustrating the EdU labeling paradigm. (F) Representative images of ctrl and RCPS D30 organoids pulsed with EdU for 2 days and stained for EdU (magenta), for Ki67 (green) and with DAPI (blue). The area shown at higher magnification on the right is outlined in yellow in the middle image. (G) Quantification of EdU+ Ki67+ cells relative to all labeled EdU+ cells. n=15 control, n=14 RCPS. (H) Representative bright-field images of ctrl and RCPS D70-80 organoids. (I) Quantification of ctrl and RCPS organoid area at D70-80. n=15 control, n=25 RCPS. (C,D,G,I) Two or more independent differentiations per experiment. Individual dots indicate results from one organoid. Data are mean±s.d. Unpaired two-tailed Student's t-test: *P<0.05, **P<0.01, ****P<0.0001. Scale bars: 100 µm in B (left and middle), F (left and middle), H; 50 µm in F (right); 10 µm in B (right).
EIF4A3 RCPS human cortical organoids exhibit neurogenesis defects. (A) Schematic representing the generation of cortical organoids and organoid sections from iPSCs derived from three control people (ctrl; green, light blue and dark blue), one isogenic line generated from a patient (black) and three people with Richieri-Costa-Pereira syndrome (RCPS) (orange, pink and purple). Modified, with permission, from Alsina et al. (2022 preprint). (B) Representative images of ctrl and RCPS D35 organoids stained for SOX2 (magenta), for PH3 (green) and with DAPI (blue). The area shown at higher magnification on the right is outlined in yellow in the left image. (C,D) Quantification of SOX2+ cells relative to all DAPI+ cells (C) or PH3+ SOX2+ cells relative to all SOX2+ cells (D) in organoid sections. (C) n=9 control and n=10 RCPS. (D) n=10 ctrl and n=11 RCPS. (E) Schematic illustrating the EdU labeling paradigm. (F) Representative images of ctrl and RCPS D30 organoids pulsed with EdU for 2 days and stained for EdU (magenta), for Ki67 (green) and with DAPI (blue). The area shown at higher magnification on the right is outlined in yellow in the middle image. (G) Quantification of EdU+ Ki67+ cells relative to all labeled EdU+ cells. n=15 control, n=14 RCPS. (H) Representative bright-field images of ctrl and RCPS D70-80 organoids. (I) Quantification of ctrl and RCPS organoid area at D70-80. n=15 control, n=25 RCPS. (C,D,G,I) Two or more independent differentiations per experiment. Individual dots indicate results from one organoid. Data are mean±s.d. Unpaired two-tailed Student's t-test: *P<0.05, **P<0.01, ****P<0.0001. Scale bars: 100 µm in B (left and middle), F (left and middle), H; 50 µm in F (right); 10 µm in B (right).
To assess neuronal generation in RCPS organoids, we quantified cell cycle exit. To achieve this, we performed a 48 h pulse of the nucleotide analog EdU, which is incorporated in proliferating cells during S phase (Fig. 7E). Quantification of EdU and Ki67 revealed significantly more cells that have exited the cell cycle (EdU+Ki67−/EdU+) in the RCPS organoids (Fig. 7F,G). These data are consistent with the finding that RCPS organoids have fewer progenitors, and indicate that Eif4a3 regulation of cell fate is conserved in mouse and human.
Given the microcephaly observed in some individuals with RCPS and in Eif4a3 cHETs, we next asked whether these neurogenesis defects were associated with reduced size of RCPS organoids. Indeed, RPCS organoids at D70-80 were 38% smaller on average compared with control organoids (Fig. 7H,I). The smaller overall size of RCPS organoids is consistent with altered neurogenesis, further mirroring the phenotypes of Eif4a3 haploinsufficient brains.
EIF4A3 acts in an EJC-dependent manner to control newborn neuron number
EIF4A3 is a canonical regulator of RNA metabolism, as one of the three core components of the EJC. Although individual EJC mutants show similar brain phenotypes and transcriptome changes (Mao et al., 2016), the extent to which EIF4A3 actually functions in this complex during neurogenesis is unknown. To address this, we used a point mutant, EIF4A3T163D, previously shown to abolish EIF4A3 binding to the MAGOH-RBM8A heterodimer and RNA (Ryu et al., 2019), but competent to bind directly to microtubules (Alsina et al., 2022 preprint). We asked whether EIF4A3WT or EIF4A3T163D is sufficient to rescue Eif4a3 cHET neurogenesis phenotypes (Fig. 8A). Primary progenitor cultures were generated from E12.5 control and Eif4a3 cHET cortices, and electroporated with one of the following plasmid combinations: (1) GFP, (2) GFP and 3x-FLAG-EIF4A3WT, or (3) GFP and 3x-FLAG-EIF4A3T163D (Fig. 8A,B). The GFP plasmid served as a readout for electroporated cells, and expression of the 3x-FLAG-EIF4A3 constructs was validated by FLAG immunostaining (Fig. S4A-F). After 48 h, the primary cultures were fixed and stained for TUJ1 to mark neurons. For control cultures, there was no significant difference in the proportion of neurons between the GFP alone, EIF4A3WT and EIF4A3T163D conditions (31%, 31% and 36% TUJ1+GFP+/GFP+, respectively) (Fig. 8C-E,I). Thus, with the amounts used, neither EIF4A3WT nor EIF4A3T163D alone caused an overexpression phenotype (see Materials and Methods). For Eif4a3 cHET cultures, there was a significant 1.8-fold increase in the proportion of GFP+ neurons (56% compared with 31% for control) (Fig. 8F-I). These proportions are consistent with increased neurons seen in Eif4a3 cHET fixed brains (Mao et al., 2016) and in live imaging (Figs 2,3). As predicted, expression of EIF4A3WT rescued premature neurogenesis of Eif4a3 cHET progenitors (Fig. 8G,I). However, expression of the EJC mutant, EIF4A3T163D, did not rescue these defects in Eif4a3 cHET progenitors (59% TUJ1+GFP+/GFP+) (Fig. 8H,I). Altogether, these rescue experiments suggest that the core EJC components EIF4A3, MAGOH and RBM8A are likely working in a complex to control neuronal generation (Fig. 9).
EIF4A3 acts via the EJC to control neuron generation. (A) Schematic depicting EIF4A3WT (blue) bound to both mRNA and the MAGOH-RBM8A heterodimer (orange and purple) (left) and the point mutant EIF4A3T163D (right), which cannot bind RNA and MAGOH-RBM8A. Adapted, with permission, from Alsina et al. (2022 preprint). (B-H) Schematic of rescue experiments (B) in which E12.5 control or cHET primary cultures were electroporated with GFP and Eif4a3 plasmids, cultured for 48 h and then fixed and stained for (C-H) GFP (green), TUJ1 (magenta) and Hoechst (white) (representative images are shown). (I) Quantification of the proportion of electroporated neurons (GFP+, TUJ1+; green, magenta) relative to all electroporated cells (GFP; green). Control cells (orange) and cHET (light blue) plotted with electroporation conditions on the x-axis. n=1135 ctrl cells; n=503 cHET cells; four embryos from two independent litters. ANOVA with Tukey post-hoc: **P<0.01; ns, not significant. Individual dots indicate results from individual biological replicates. Data are mean±s.d. Scale bars: 20 µm.
EIF4A3 acts via the EJC to control neuron generation. (A) Schematic depicting EIF4A3WT (blue) bound to both mRNA and the MAGOH-RBM8A heterodimer (orange and purple) (left) and the point mutant EIF4A3T163D (right), which cannot bind RNA and MAGOH-RBM8A. Adapted, with permission, from Alsina et al. (2022 preprint). (B-H) Schematic of rescue experiments (B) in which E12.5 control or cHET primary cultures were electroporated with GFP and Eif4a3 plasmids, cultured for 48 h and then fixed and stained for (C-H) GFP (green), TUJ1 (magenta) and Hoechst (white) (representative images are shown). (I) Quantification of the proportion of electroporated neurons (GFP+, TUJ1+; green, magenta) relative to all electroporated cells (GFP; green). Control cells (orange) and cHET (light blue) plotted with electroporation conditions on the x-axis. n=1135 ctrl cells; n=503 cHET cells; four embryos from two independent litters. ANOVA with Tukey post-hoc: **P<0.01; ns, not significant. Individual dots indicate results from individual biological replicates. Data are mean±s.d. Scale bars: 20 µm.
Model for Eif4a3-mediated control of neurogenesis. Neurogenesis in wild type (left) or in a Eif4a3 haploinsufficient mouse or RPCS brain (right). Eif4a3 haploinsufficient progenitors delayed in mitosis produce fewer progenitors, more neurons and more apoptotic progeny. Our data suggest that EIF4A3 controls neuron neurogenesis through both p53-dependent and -independent mechanisms that diverge over the course of development.
Model for Eif4a3-mediated control of neurogenesis. Neurogenesis in wild type (left) or in a Eif4a3 haploinsufficient mouse or RPCS brain (right). Eif4a3 haploinsufficient progenitors delayed in mitosis produce fewer progenitors, more neurons and more apoptotic progeny. Our data suggest that EIF4A3 controls neuron neurogenesis through both p53-dependent and -independent mechanisms that diverge over the course of development.
DISCUSSION
The exon junction complex is essential for virtually every step of RNA metabolism across eukaryotes. Moreover, EJC components are crucial for cortical development and underlie neurological disease. Using mouse models and human cortical organoids, we define key cellular and molecular mechanisms by which the RNA helicase EIF4A3 functions in cortical development and disease. EIF4A3 controls mitosis duration of both mouse and human progenitors to influence neuronal generation and survival. We further parse out developmental mechanisms of neurogenesis that are both p53-dependent and independent. Finally, we show that control of neuronal generation by EIF4A3 relies upon an intact EJC. Taken together, our study furthers our understanding of how EIF4A3 and the EJC control cortical development and suggests possible mechanisms by which they cause neurodevelopmental disorders.
EIF4A3 controls mitosis duration to influence cell fate
We demonstrate that Eif4a3 haploinsufficiency in mouse and human neural progenitors impairs neuronal generation and cell survival, and, strikingly, that these defects are associated with a prolonged mitosis. This is reflected not only by live imaging of individual progenitors of both species, but also by increased PH3+ progenitors and decreased cell cycle exit of RCPS organoids. Eif4a3-depleted progenitors delayed in mitosis undergo fewer successful divisions and produce fewer progenitors, more neurons, and less viable progeny. These data corroborate previous findings from our group showing that mitotically delayed Magoh haploinsufficient RGCs or interneuron progenitors also exhibit altered neurogenesis (Pilaz et al., 2016; Sheehan et al., 2020). Our previous experiments support a causal role for mitosis in cell fate and microcephaly. Indeed using pharmacology to transiently prolong mitosis in vitro, ex vivo or in vivo, we demonstrated prolonged mitosis of mouse and human neural progenitors directly alters progeny cell fate (Mitchell-Dick et al., 2019; Pilaz et al., 2016). Thus, the current data for Eif4a3 reinforce the relevance of mitosis duration for neural fate decisions.
A key question is how EIF4A3 controls mitosis duration. One possibility is that EIF4A3, through the EJC, post-transcriptionally controls genes necessary for mitosis progression. For example, in mouse embryonic stem cells (ESCs), EIF4A3 post-transcriptionally controls cell cycle regulators that are important for pluripotency maintenance (Li et al., 2022). Additionally, EIF4A3 directly binds transcripts encoding the cell cycle regulator Ccnb1 to mediate its nuclear export in ESCs. In support of this possible mechanism, all three core components are essential for mitosis (Pilaz et al., 2016; Sheehan et al., 2020; Silver et al., 2010). A second intriguing possibility is that EIF4A3 directly influences mitosis by manipulating the cytoskeleton. Indeed, EIF4A3 can directly bind to microtubules and also localizes to the mitotic spindle (Alsina et al., 2022 preprint). Likewise, RBM8A and MAGOH localize to centrosomes (Ishigaki et al., 2014). Thus, it is possible the EJC could influence mitosis duration by directly modulating mitotic spindle integrity (Silver et al., 2010). Future detailed studies will be invaluable for teasing apart these potential canonical and non-canonical mechanisms.
More broadly, our findings support the notion that prolonged mitosis of progenitors is a causal mechanism for some microcephaly cases. To date, more than 25 genes encoding mitotic regulators have been implicated in primary microcephaly (Degrassi et al., 2019). Mitotic defects are present in many experimental microcephaly models, including organoids and mice (Jayaraman et al., 2018). For example, a study of Miller-Dieker syndrome (MDS) used patient iPSC-derived cortical organoids to show that delayed mitosis of outer radial glia is associated with aberrant neurogenesis (Bershteyn et al., 2017). Future studies are crucial to understanding how mitosis duration influences fate and viability to shape development and influence disease.
Disentangling apoptosis and cell fate in EIF4A3-mediated microcephaly
Eif4a3 haploinsufficiency in progenitors leads to striking microcephaly, which is associated with altered cell composition and extensive apoptosis. Thus, a key question is the extent to which these cell composition changes and microcephaly are explained by apoptosis. By compound ablation of p53, we show that microcephaly is due to both p53-dependent apoptosis and p53-independent mechanisms. We discover that p53 ablation is sufficient to rescue deep-layer (layers V VI) but not upper-layer (IV-II/III) neuron numbers. This finding is consistent with previous analyses of E18.5 Rbm8a;p53 compound mutant brains, which showed that p53 loss rescued layer VI but not layer II/III neuron number (Mao et al., 2016). Our analysis of RORβ compared with LHX2 suggests that layer IV neurons may be particularly sensitive to Eif4a3 loss, an observation that may be investigated in the future. Probing different embryonic stages, we discover that p53 ablation significantly rescues deep layer-generating progenitors but is insufficient to fully rescue upper layer-generating progenitors. These developmental differences suggest that perhaps progenitors have differential sensitivity to apoptosis over the course of development. It further suggests that EIF4A3-mediated control of neurogenesis relies on several downstream pathways, with p53 signaling being especially important for deep-layer neurogenesis, but additional yet undiscovered pathways relevant for genesis of upper-layer neurons.
Previous work from our lab and others indicates that loss of P53, especially in mitotic mutants can be accompanied by, or result in, DNA damage and ploidy changes (Ganem and Pellman, 2012; Mitchell-Dick et al., 2019; Pilaz et al., 2016; Shi et al., 2019). Indeed, we observe increased DNA damage in the Eif4a3 haploinsufficient brains. Given the temporal differences in p53-dependent progenitor number, it is intriguing to consider whether late-stage progenitors are more prone to accumulate DNA damage as they have undergone additional rounds of defective mitosis relative to early-stage progenitors. Although we did not assess the ploidy of the surviving cells in compound mutants, Magoh+/− brains, which are also mitotically delayed and apoptotic, did not show evidence of aneuploidy in vivo (Pilaz et al., 2016), in contrast to that of some other microcephaly models (Marthiens et al., 2013). Given the other shared phenotypes of Magoh and Eif4a3 mutants, we predict that Eif4a3 cHET cells will also not show aneuploidy. However, this remains to be formally tested.
Interestingly, our p53 findings resemble that of some microcephaly mutants in which p53 partially but incompletely rescues brain size (Little et al., 2021; Shi et al., 2019), whereas other mutants are largely p53 dependent (Insolera et al., 2014). This further highlights p53 as a key downstream pathway involved in microcephaly. Future studies, including transcriptomics, epigenomics and proteomics, will be valuable to tease out how EIF4A3 differentially controls neurogenesis of upper- and deep-layer neurons via p53. Furthermore, although we have shown that mitosis is a key feature affected by Eif4a3 mutation, the extent to which p53 signaling acts upstream or downstream of mitosis is a question for future studies.
Unique requirements for EIF4A3 across cell types, developmental time and disease
As a component of the EJC, EIF4A3 canonically controls RNA metabolism. Our rescue experiments argue that EIF4A3 acts in an EJC-dependent manner to control neuron number. This fits with the observation that Eif4a3, Rbm8a and Magoh haploinsufficiency each have similar cortical development phenotypes, including microcephaly, precocious neurogenesis, apoptosis and mitotic defects. Furthermore, all three mutants exhibit converging dysregulation of common transcripts (Mao et al., 2016). Thus, we postulate that EJC components work together for neurogenesis. In contrast, recent findings from our lab indicate there are EJC-independent requirements for EIF4A3 in neurons; Eif4a3, but not Magoh or Rbm8a, is required for axonal outgrowth (Alsina et al., 2022 preprint). EIF4A3 is competent to bind to microtubules independently of the EJC and RNA, and directly promotes microtubule polymerization and stability (Alsina et al., 2022 preprint). This raises the intriguing possibility that EIF4A3 could also have some EJC-independent roles in neurogenesis, e.g. at later stages of development not examined in this study.
Although Eif4a3 has conserved roles in the production of projection neurons in both mouse and human neural progenitors, it is intriguing to consider whether this holds true in other brain regions. In addition to excitatory neurons in inhibitory progenitors, Magoh and Rbm8a are required for proliferation and neurogenesis, illustrating parallel functions are at play (McSweeney et al., 2020; Sheehan et al., 2020). In future studies it will be interesting to assess whether EIF4A3 similarly controls inhibitory neurogenesis and whether it does so via canonical mechanisms.
Mutations in EIF4A3 and RBM8A are linked to neurodevelopmental disorders, including microcephaly, intellectual disability and RCPS (Asthana et al., 2022; Bertola et al., 2018; Favaro et al., 2014; Hsia et al., 2018; Nguyen et al., 2013). Our analysis of cortical organoids and dissociated neural progenitors reveal that neurogenesis is a key process impacted in RCPS. In particular, we note conserved roles for EIF4A3 in mitotic duration, progeny fate and apoptosis. Our findings complement previous observations that axonal maturation is impaired in RCPS cortical organoids (Alsina et al., 2022 preprint). Continued investigation of RCPS organoids can provide important insights into the etiology of this rare disorder. In summary, we propose a model in which Eif4a3 haploinsufficiency causes microcephaly through both apoptosis and altered cell fate due to mitotic delay. Altogether, our study provides new insights into the basis for RCPS and other EJC-dependent disorders.
MATERIALS AND METHODS
Mouse husbandry and genetics
All animal procedures were approved by the Duke Institutional Animal Care and Use Committee (IACUC) and performed in agreement with the ethical guidelines of the Division of Laboratory Animal Resources (DLAR) from Duke University. We used the previously described mouse lines: Eif4a3loxP and Emx1-Cre (B6.129S2-Emx1tm1(cre)Krj/J) (Gorski et al., 2002). The following mouse strains were obtained from Jackson Laboratories: C57BL/6J (wild type) and B6.129S2-Trp53tm1Tyj/J (p53LoxP). For embryo staging, plug dates were defined as embryonic day (E) 0.5 on the morning the plug was identified.
Immunofluorescence
Embryonic brains
Embryonic brains were fixed and sectioned as previously described (Mao et al., 2015). Coronal 20 µm sections from the somatosensory cortex were permeabilized with 1×PBS/0.25% TritonX-100 and blocked with 5% NGS/PBS for 1 h at room temperature. Sections were incubated with primary antibodies overnight at 4°C and secondary antibodies at room temperature for 2 h (Alexa Fluor-conjugated, Thermo Fisher Scientific; anti-rabbit Alexa Fluor 488, A-11034; anti-mouse Alexa Fluor 594, A-11005; and anti-rat Alexa Fluor 647, A-21247, Invitrogen, 1:500). The following primary antibodies were used: SOX2 (Thermo Fisher Scientific, 14-9811-82, 1:1000), CTIP2 (Abcam, c8035, 1:500), TBR2 (Abcam, AB23345, 1:1000), CC3 (Cell Signaling Technology, 9661, 1:250), PH3 (Millipore, 06-570, 1:500), TBR1 (Cell Signaling Technology, 49,661 S, 1:1000) and LHX2 (Millipore, ABE1402, 1:500). Slides were mounted with Vectashield (Vector Labs, H-1000-10). For antigen retrieval of RORβ (R&D systems, N7927, Lot A-2, 1:100) and γH2AX (Millipore, 05-636, 1:100), coronal 20 µm sections were boiled for 15 min in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween-20 at pH 6.0) in a histology slide container. Slides were taken off the heat but kept in sodium citrate buffer for an additional 10 min to cool. Slides were then removed from the container and allowed to cool for an additional 10 min before being rinsed three times in 1×PBS. Permeabilization and antibody incubations were then followed as described above.
Cortical organoids
Organoids were fixed in ice-cold 4% PFA/PBS for 2 h, rinsed three times in ice-cold 1×PBS, and submerged in 30% sucrose/PBS overnight. Organoids were embedded, sectioned and stained as described above for embryonic brain tissue. SOX2 and PH3 antibodies were used as described above, as well as Ki67 antibodies (Cell Signaling Technology, 12202, 1:250).
Imaging and analysis
Images were captured using a Zeiss Axio Observer Z.1 equipped with an Apotome for optical sectioning, at 5×, 10× and/or 20×. Two or three sections were imaged per embryo or organoid, and all images for a given experiment were captured with identical exposures. Images were cropped to 300 µm radial columns and the brightness was equivalently adjusted across all images in Fiji (Schindelin et al., 2012). Cells were either manually (Fiji cell counter) or automatically (QuPath; Bankhead et al., 2017) counted. QuPath parameters were used as previously described (Hoye et al., 2022). For binning analysis, 300 µm wide radial columns were divided into five evenly spaced bins spanning from the ventricular (bin 1) to the pial (bin 5) surface, as previously described (Hoye et al., 2022). Each cell was assigned to a bin to calculate the distribution.
Primary cultures and live imaging
Primary cortical cultures were derived from E12.5 embryonic dorsal cortices, as previously described (Mitchell-Dick et al., 2019), but with minor modifications: (1) cortices were trypsinized for 5 min and (2) 150,000 cells were plated on poly-D-lysine coated glass-bottomed 24-well culture plates (MatTek). Images were captured every 10 min as previously described (Pilaz et al., 2016). Mitosis duration and cell division were identified by morphology (rounding and condensation of chromatin). Fate determination was performed post-imaging by immunostaining for TUJI, SOX2 and TBR2, as previously described (Mitchell-Dick et al., 2019). Live imaging of dissociated human progenitors was performed similarly, except fate determination was performed post-imaging by immunostaining for anti-KI67 (Cell Signaling Technology, 12202, 1:1000) and TUJI (TUBB3, Biolegend, 801202, 1:2000).
Generation of brain cortical organoids
Work with H9 cells and iPSCs was approved by the Duke University Institutional Research Board. The H9 (WA09) human ES cell line was purchased from WiCell (hPSCReg ID: WAe009-A) and authenticated (Thomson et al., 1998). The control and RCPS patient iPSC lines used in this study have been previously characterized and genotyped (Alsina et al., 2022 preprint; Miller et al., 2017; Yoon et al., 2019), and the authors were unaware of the identity of the individuals. The isogenic iPSC line was previously generated and characterized (Alsina et al., 2022 preprint). All cell lines were periodically tested for contamination. Cells were maintained in Essential 8 medium (Gibco) supplemented with 100 µg/ml normocin (Invivogen) on Matrigel (Corning)-coated plates, and then seeded on vitronectin (Gibco) for two passages before generating the 3D cultures. Cortical organoids were generated and maintained as previously described (Alsina et al., 2022 preprint; Yoon et al., 2019). Images were taken at 4×magnification using the EVOS XL Core (Thermo Fisher Scientific) microscope (scale: 2.145 pixels/µm).
Dissociation and electroporation of organoids
Dissociation of H9 organoids into 2D cultures was performed as previously described with minor modifications (Miura et al., 2020). To electroporate the single cell suspension, the standard protocol for the Lonza P3 primary cell 4D Nucleofector X Kit S was followed using pulse code CB-150. For the electroporation, 1×106 cells and 1.5 μl of 10 µM siRNA mix were used for each well in the 16-well cuvette. The following siRNAs were used from Qiagen: scrambled siRNA, Hs_EIF4A3_2 (SI00107828), Hs_EIF4A3_3 (SI00107835), Hs_EIF4A3_5 (SI02663794) and Hs_EIF4A3_6 (SI03049676). 25 nM of each EIF4A3 siRNA were mixed to create a 100 nM stock solution. The stock was diluted 1:100 to 10 µM before electroporation. 5×104 cells were seeded per well in 24-well MatTek glass-bottomed plates (P24G-1.5-10-F) coated with 0.1 mg/ml poly-D-lysine (Sigma-Aldrich; 30,000-70,000 MW). Cultures were maintained in organoid base medium without small molecules. After 48 h of knockdown, live imaging was started. Knockdown of EIF4A3 was confirmed by RT-qPCR (Fig. 5B).
EdU in organoids
Organoids were incubated in media containing 10 mM EdU for 48 h (Paşca et al., 2015). Organoids were fixed as described previously (Mitchell-Dick et al., 2019). Click-It reaction was performed using Life Technologies Click-It Plus Kit (C10749) following the manufacturer's instructions. Antibody staining was then performed as described above.
Electroporation of primary progenitor cultures
Primary cortical cultures were derived from E12.5 embryonic dorsal cortices as described above. The standard protocol for the Lonza P3 primary cell 4D Nucleofector X Kit S was followed using pulse code CM-150 and 0.5×106 cells per electroporation in the 16-well cuvette. Three different electroporation conditions were used per sample: 1 µg pCAG-GFP only, 1 µg pCAG-GFP+0.25ug pCAG-3xFLAG-Eif4a3 (Mus musculus), and 1 µg pCAG-GFP+0.25 µg pCAG-3xFLAG-Eif4a3 T163D (Mus musculus) (Alsina et al., 2022 preprint). Concentrations of pCAG constructs were empirically determined to avoid overexpression phenotypes. The primary cultures were incubated for 48 h before fixation and immunostaining for Tuj1 as already described.
RT-qPCR and primers
RNA was extracted in RLT buffer plus 0.01% β-mercaptoethanol following RNeasy kit (Qiagen) and cDNA was prepared using the iScript kit (Bio-Rad) following the manufacturer's protocol. qPCR was performed using SYBR Green iTaq (Bio-Rad) in at least three independent biological samples (three technical replicates per sample) in a QuantStudio 3 machine (Applied Biosystem). Values were normalized to TBP as loading control. The following primers were used: Homo sapiens TBP (forward 5′-GTGACCCAGCATCACTGTTTC-3′ and reverse 5′-GCAAACCAGAAACCCTTGCG-3′) and Homo sapiens EIF4A3 (forward 5′-GGAGATCAGGTCGATACGGC-3′ and reverse 5′-GATCAGCAACGTTCATCGGC-3′).
Statistical methods and rigor
Exact statistical tests, P-values and n for each analysis are reported in the figure legends. For each experiment, both male and female mice were used and littermates were used when possible. All analyses were performed by one or more investigators who were unaware of the genotype and/or condition. We have performed normality tests on those data that use t-tests. In Figs 2B and 6C, the larger datasets do not pass normality tests and as such we additionally used Mann–Whitney non-parametric tests to compare the unmatched groups of data represented in these graphs.
Acknowledgements
We thank the members of the Silver Lab for helpful discussions and careful reading of the manuscript. We thank the Duke Mouse Transgenic Facility for providing daily maintenance, housing and veterinary care for our mouse colony, and the Duke light and microscopy core for use of shared equipment. We are grateful to Maria Rita Passos-Bueno, Dr Roseli M. Zechi Ceide and Gershon Kobayashi for the RCPS iPSCs.
Footnotes
Author contributions
Conceptualization: B.M.L., D.L.S.; Methodology: B.M.L., C.M.M., F.C.A., D.L.S.; Validation: B.M.L., R.A.S., C.M.M.; Formal analysis: B.M.L., R.A.S.; Investigation: B.M.L., R.A.S., C.M.M., F.C.A.; Writing - original draft: B.M.L., D.L.S.; Writing - review & editing: B.M.L., R.A.S., C.M.M., F.C.A., D.L.S.; Visualization: B.M.L., D.L.S.; Supervision: D.L.S.; Project administration: D.L.S.; Funding acquisition: D.L.S.
Funding
This work was supported by the National Institutes of Health (R01NS083897, R01NS110388 and R01NS120667 to D.L.S.) and by a School of Medicine, Duke University Genetic Discovery in Rare Diseases Pilot Grant to D.L.S. Deposited in PMC for release after 12 months.
Data availability
All relevant data can be found within the article and its supplementary information.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201619.reviewer-comments.pdf.
References
Competing interests
The authors declare no competing or financial interests.