Histone Deacetylase 3 Governs Perinatal Cerebral Development via Neural Stem and Progenitor Cells.

We report that cerebrum-specific inactivation of the histone deacetylase 3 (HDAC3) gene causes striking developmental defects in the neocortex, hippocampus, and corpus callosum; post-weaning lethality; and abnormal behaviors, including hyperactivity and anxiety. The defects are due to rapid loss of embryonic neural stem and progenitor cells (NSPCs). Premature neurogenesis and abnormal neuronal migration in the mutant brain alter NSPC homeostasis. Mutant cerebral cortices also display augmented DNA damage responses, apoptosis, and histone hyperacetylation. Moreover, mutant NSPCs are impaired in forming neurospheres in vitro, and treatment with the HDAC3-specific inhibitor RGFP966 abolishes neurosphere formation. Transcriptomic analyses of neonatal cerebral cortices and cultured neurospheres support that HDAC3 regulates transcriptional programs through interaction with different transcription factors, including NFIB. These findings establish HDAC3 as a major deacetylase critical for perinatal development of the mouse cerebrum and NSPCs, thereby suggesting a direct link of this enzymatic epigenetic regulator to human cerebral and intellectual development.

Introduction

Histone acetylation is critical for establishing, maintaining, and regulating all eukaryotic epigenomes (Suganuma and Workman, 2008, Yang and Seto, 2008a). This modification is reversible and controlled by the opposing actions of acetyltransferases and deacetylases. There are 18 mammalian histone deacetylases (HDACs) (Haberland et al., 2009, Yang and Seto, 2008b). Different from the other 15 deacetylases, HDAC1, HDAC2, and HDAC3 are catalytic subunits of different stoichiometric multiprotein complexes, which are highly stable, even enough for purification via conventional chromatography (Yang and Seto, 2008b). These complexes serve as functional cores for additional interaction with other partners to exert gene-specific histone deacetylation and subsequent functional impact (Yang and Seto, 2008b). Although HDAC1 and HDAC2 are part of at least three different complexes containing >15 other proteins (Yang and Seto, 2008b), HDAC3 is the catalytic subunit of several paralogous tetrameric complexes, composed of the nuclear receptor corepressor NCoR (or its paralog SMRT), the WD40 protein TBL1 (a.k.a. TBL1X; or its paralogs TBL1XR1 and TBL1Y), and the signaling regulator GPS2 (no homology to GPS1 and no paralogs) (Karagianni and Wong, 2007, Oberoi et al., 2011). Thus, HDAC3 is unique among the HDAC superfamily. Within the complexes, NCoR and SMRT serve as scaffolds for complex assembly with their deacetylase-activating domains for HDAC3 interaction and other domains for association with TBL1 and GPS2 (Codina et al., 2005, Oberoi et al., 2011, Watson et al., 2012). Complex formation is crucial for the maximal activity of HDAC3 (Guenther et al., 2001, Guenther et al., 2000).

The mouse gene has been subject to various genetic studies. Germline deletion of Hdac3 leads to embryonic lethality at E9.5 (Knutson et al., 2008, Montgomery et al., 2008), indicating an essential role in embryo development. Conditional deletion of Hdac3 in the liver, heart, muscle, fat, or intestine causes abnormal development or metabolism (Alenghat et al., 2013, Emmett et al., 2017, Farooq et al., 2008, Feng et al., 2011, Knutson et al., 2008, Lu et al., 2018, Montgomery et al., 2008, Summers et al., 2013, Sun et al., 2012), indicating that Hdac3 is crucial for development and homeostasis of these tissues. Its role in the mouse brain also started to emerge (McQuown et al., 2011, Norwood et al., 2014, Nott et al., 2016, Zhang et al., 2016). These studies raise an interesting question about the roles in human development. Related to this, a child with a heterozygous HDAC3 mutation is associated with neonatal epileptic encephalopathy (Helbig et al., 2016). Another child with an HDAC3 mutation displays learning difficulties and brain abnormalities (Zhou et al., 2019a). Three individuals with mutations in the genes for NCoR and SMRT show learning difficulties, intellectual disability, or developmental delay (Zhou et al., 2019a). Multiple patients with TBL1XR1 mutations have intellectual disability (Heinen et al., 2016, Laskowski et al., 2016, Saitsu et al., 2014). These clinical features suggest that HDAC3 complexes may be critical for early cerebral development. We thus carried out cerebrum-specific deletion of mouse Hdac3. The mutant embryos and pups show cerebral defects, impaired NSPC development, altered histone modifications, and deregulated transcription.

Results

Hdac3 Is Highly Expressed in the Developing Cerebrum

To examine the role of HDAC3 in cerebral development, we initially determined the expression pattern of Hdac3 in the developing cerebral cortex. Indirect immunofluorescence microscopy using a specific anti-HDAC3 antibody indicated that Hdac3 is abundantly expressed in the developing ventricular zone (VZ), subventricular zone (SVZ), cortical plate (or preplate), and hippocampus at P0 (Figure S1A), E16.5 (Figure S1D), and E12.5 (Figure S1G). In addition, RNA sequencing (RNA-seq) showed that the FPKM value for Hdac3 mRNA is higher than 30 in the neonatal cerebral cortex and the neurospheres cultured from E16.5 embryonic cerebral cortex (Figure S1J), indicating high-level expression of this deacetylase in the developing cerebral cortex and embryonic NSPCs. The expression data suggest that HDAC3 may have an important role in regulating perinatal cerebral development.

Cerebrum-Specific Hdac3 Deletion Leads to Early Lethality and Abnormal Behaviors

To evaluate directly the function of HDAC3 in early cerebral development, we generated cerebrum-specific knockout mice by using the Emx1-Cre line, which expresses the Cre recombinase specifically in the cerebrum and its precursors as early as E10.5 (Chou et al., 2009, Gorski et al., 2002). Mating of this line with Hdac3 mice produced the Hdac3; Emx1-Cre knockout (or cKO) mice. Knockout pups were born at a normal Mendelian ratio (Table S1) and appeared grossly normal, with some displaying up-turned paws for a few days in the first week of life. Starting from week 2, all mutant pups failed to thrive and became significantly smaller than the wild-type (Figure 1A). In week 3, some mutant pups were runted and subsequently died, with a majority unable to survive in the week after (Figures 1A and 1B). The exact cause of lethality is unclear, but we have found similar lethality in another line with cerebral defects (You et al., 2015b).

Figure 1

Cerebrum-Specific Hdac3 Loss Causes Defects in the Neocortex, Hippocampus, and Corpus Callosum

(A) Growth curves for wild-type (WT), heterozygous (HET), and homozygous cerebrum-specific knockout (cKO) mice (n = 12, 8, and 7 for the WT, HET, and cKO groups, respectively). cKO was compared with WT for statistical analysis.

(B) Survival curves for control and homozygous mutant mice (n = 40 and 38 for the control and cKO groups, respectively).

(C and D) Brain weight and brain/body weight ratio at P3, P7, P17, and P23 (n ≥ 3 for all groups).

(E) Western blotting showed efficient deletion of Hdac3 in the cerebral cortex of mutant pups at P4 (n = 3). No other bands were detected.

(F) Representative brain photos of wild-type and cKO mice at P19. Red arrowheads mark cysts in the cKO brain. In the wild-type, an asterisk denotes damage caused during dissection.

(G) Representative images for Nissl staining of WT and cKO brain sections from pups at P19. A red arrowhead denotes the underdeveloped mutant hippocampus, and green asterisks mark the region corresponding to the cysts shown in (F).

(H) Enlarged images of hippocampal regions of the brain sections shown in (G).

(I and J) Same as (G and H) but at P7.

(K) Nissl staining of brain sections indicated that the ventricular/subventricular zone almost disappears in the P0 mutant cerebrum. The boxed areas in the left panels are enlarged at the right or in (L). A red asterisk denotes the ventricular/subventricular zone in the wild-type.

(L) Enlarged images of hippocampal regions of boxed in (K).

(M) Representative images showing agenesis of the mutant corpus callosum at P0. Red arrowheads denote the boundary of the corpus callosum in the WT cerebral cortex. For panels (F–M), representative images from serial sections are shown; at least three embryos or pups were analyzed for each group.

(N) Quantification of the cerebral neocortical thickness at the middle parts as marked with solid lines in (G), (I), and (K). n = 3 for all groups.

Abbreviations: CA1, Cornu Ammonis 1; CA2, Cornu Ammonis 2; CA3, Cornu Ammonis 3; Cb, cerebellum; CP, cortical plate; DG, dentate gyrus; HP, hippocampus; IZ, intermediate zone; LV, lateral ventricle; MB, middle brain; MZ, marginal zone; Nct, neocortex; SP, subplate; SVZ, subventricular zone; VZ, ventricular zone. ns, not statistically significant; **, p < 0.01; ***, p < 0.001. Scale bars, 2 mm in (F); 1 mm in (G–J), (M), and the left panels of (K); and 200 μm in (L) and right panels of (K). See also Table S1 and Figures S1 and S2.

Around P21, the mutant pups started to show obvious abnormalities in walking and were overactive. Moreover, 2- or 3-week-old mutant pups were also easily irritated when touched. We thus carried out open field tests to assess the anxious and hyperactive status of mutant mice (Li et al., 2019). Paired control and cKO mice at P22 were tested in an arena for 10 min and recorded by a video tracking system. The results showed that, compared with the wild-type counterparts, the mutant pups spent significantly more time at the border (Figures S2A and S2C) and traveled much more (Figures S2B and S2C). Thus, the homozygous mutant pups displayed hyperactivity and anxiety.

The brain weight of the mutant pups was comparable with that of the wild-type counterparts at P7 but showed significant reduction at P17 and P23 (Figure 1C). However, the brain/body weight ratio was comparable between the wild-type and mutant pups, even at P17 and P23 (Figure 1D). Thus, the mutant brain remained proportional to the body weight (Figures 1C and 1D), indicative of no microcephaly in mutant pups.

We assessed whether Hdac3 was efficiently and specifically deleted in the mutant cerebrum. For this, we carried out western blotting analysis of cerebral extracts from wild-type and mutant pups. As shown in Figure 1E, HDAC3 was specifically detected in the wild-type cerebral cortices at P4 (lanes 1–3). Importantly, the protein level was greatly reduced in the mutants (lanes 4–6). Quantification confirmed this assessment (Figure S1L). To investigate whether the deletion is also efficient at early developmental time points, we performed immunofluorescence microscopy to analyze expression of HDAC3 on brain or embryonic sections at P0, E16.5, and E12.5. As shown in Figures S1A–S1C, Hdac3 was efficiently deleted at a majority of cells in the neocortex or hippocampus but not the striatum at P0. A small number of cells remained positive in the cKO mutant neocortex (Figures S1A and S1B). Some of them may be interneurons that migrate in a parallel manner from the ganglionic eminence (Marin and Rubenstein, 2001). Another possibility is microglia, which are of hematopoietic origin. The Cre recombinase in the Emx1-Cre line is not expressed in interneurons or microglia (Chou et al., 2009, Gorski et al., 2002). The deletion was also efficient at E16.5 (Figures S1D and S1F). As for E12.5, HDAC3 was lost in the mutant cortical plate but not the ganglionic eminence (GE) (Figures S1G and S1I). Thus, the Emx1-Cre line induces efficient and specific Hdac3 deletion in the cerebrum and its embryonic precursors, at least starting at E12.5.

HDAC3 Is Crucial for Development of the Neocortex, Hippocampus, and Corpus Callosum

During brain dissection, we unexpectedly noticed prominent cysts at the caudal part of the mutant brain from 3-week-old pups (Figure 1F). Nissl staining of serial brain sections revealed thin neocortices and large lateral ventricles in the mutant forebrain (Figure 1G). Moreover, the mutant hippocampus was poorly developed, with the dentate gyrus barely visible (Figure 1H). Such neocortical and hippocampal defects were also present in the mutant pups at earlier time points, such as P7 (Figures 1I and 1J) and P0 (Figures 1K–1L). Moreover, Nissl staining of serial brain sections from P0 showed that the mutant corpus callosum was underdeveloped, with no clear boundaries between the corpus callosum and its adjacent tissues (Figure 1M). Quantification confirmed that the mutant cerebral cortex was significantly thinner starting at P0 (Figure 1N). Thus, Hdac3 deletion leads to cerebral defects starting at least around birth.

To substantiate these histological observations, we carried out immunostaining to examine the defects at the molecular and cellular levels. For this, we used various antibodies against makers (such as Cux1, Ctip2, and Tbr1) specific to different neocortical layers. Cux1 is specifically expressed in upper layer neurons, whereas Ctip2 (a.k.a. Bcl11b) and Tbr1 are cortical markers of deeper layer neurons (Molyneaux et al., 2007). As shown in Figure 2, neocortical lamination was altered in the mutant neonatal cerebrum. There was no clear boundary for the mutant Ctip2+ layer (Figures 2A and 2B, upper panels). Quantification confirmed abnormal distribution of such cells in the mutant neocortex (Figure 2C). This was also true for Tbr1+ neurons in the mutant neocortex (Figures 2D and 2E, upper panels and 2F). The Cux1+ and Cux2+ layers were obvious in the wild-type neocortices but missing in the mutant neocortices (Figures 2A, 2B, 2D, and 2E, lower panels). Such alterations were also evident at P7 (Figure S3) and E16.5 (Figure S4). Thus, cerebrum-specific Hdac3 deletion results in developmental defects in the neocortex at least starting at E16.5.

Figure 2

Abnormal Neocortical Lamination and Hippocampal Development in the Mutant Brain

(A and B) Co-immunostaining of wild-type and mutant neonatal brain sections with anti-Cux1 and -Ctip2 antibodies. The former labels the SVZ and layers II-IV of adult neocortex (Nieto et al., 2004), whereas the latter marks layers V–VI (Lai et al., 2008). A red asterisk denotes a disorganized Ctip2+ layer in the mutant.

(C) The mutant neocortical layers are not clearly separated, so the wild-type or mutant neocortex were divided equally into 10 zones for determination of the Ctip2+ cell distribution (n = 3 for each group).

(D and E) Co-immunostaining of wild-type and mutant P0 brain sections with anti-Tbr1 and -Cux2 antibodies. The former labels layer VI of the adult neocortex (Hevner et al., 2001), whereas the latter marks layers II–IV (Nieto et al., 2004). A red asterisk denotes ectopic distribution of Tbr1+ cells.

(F) The wild-type or mutant neocortex was divided equally into 10 zones for determination of the Tbr1+ cell distribution (n = 3 for each group).

Abbreviations are listed in the legend to Figure 1. Scale bars, 500 μm for the left panels and 100 μm for the rest of the panels of (A and B) and (D and E). ns, not statistically significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001. See also Figures S3–S5.

The above-mentioned immunostaining experiments also confirmed hippocampal abnormalities in the embryonic, neonatal, and postnatal mutant brains (Figures 2, S3, and S4). To substantiate this, we utilized two other markers, NeuroD and GFAP. The former is a neuron-specific transcription factor, and the latter is a marker of radial glia cells (precursors of neural stem and progenitor cells [NSPCs]) and activated astrocytes. Immunostaining with antibodies against both markers supported the abnormal morphology of the mutant hippocampus at P0, with fewer NeuroD+ cells but more GFAP+ cells than the wild-type (Figures S5A, S5C, S5D, and S5F). As shown in Figures S5B, S5C, S5E, and S5F, the abnormality was even more dramatic in the mutant hippocampus at P7. Thus, these results indicate that cerebrum-specific Hdac3 deletion leads to profound developmental defects in the hippocampus.

HDAC3 Is Required for Development of Neural Stem and Progenitor Cells

From DAPI and Nissl staining, we noticed that the ventricular/subventricular zone is thin in the mutant brain at E16.5 (Figure S1K) and almost missing at P0 (Figure 1K). The embryonic ventricular/subventricular zone is the area where radial glia cells reside (Noctor et al., 2001). To investigate how neural stem cells (NSCs) are affected, we carried out immunofluorescence staining of embryonic and brain sections using an antibody against Sox2, an NSC-specific marker (Ellis et al., 2004). At P0, the Sox2+ layer was dramatically reduced in the mutant (Figures 3A and 3B), indicating that Hdac3 loss exhausts the NSC compartment. Moreover, ectopic Sox2+ cells were found outside of the mutant ventricular/subventricular zone (Figures 3A and 3C). At E16.5, the Sox2+ layer was also dramatically reduced in the mutant (Figures 3D and 3E), and more ectopic Sox2+ cells were observed outside of the mutant ventricular/subventricular zone (Figures 3D and 3F). Thus, at both P0 and E16.5, HDAC3 is critical for proper development of the Sox2+ cell layer. By contrast, thickness of the Sox2+ cell layer was not significantly affected in the mutant at E12.5 (Figures 3G and 3H), although HDAC3 was efficiently deleted (Figures S1G–S1I), indicating that HDAC3 is dispensable for formation of the Sox2+ cell layer at this embryonic stage.

Figure 3

Cerebrum-Specific Hdac3 Deletion Compresses the NSC Compartment

(A) Immunostaining analysis with anti-Sox2 antibody showed almost complete disappearance of Sox2+ cells in the mutant VZ at P0.

(B) Sox2+ cells are dense in the VZ, so its thickness instead of the cell number was measured for the Sox2+ cell layer (n = 4).

(C) Sox2+ cells outside the VZ/SVZ were counted (n = 4).

(D) Immunostaining analysis revealed reduction of Sox2+ cells in the VZ of mutant cerebral cortex at E16.5. An asterisk in the middle bottom panel marks Sox2+ cells outside the VZ/SVZ. Two red arrowheads in the right bottom panel denote a thinner layer of the hippocampal neuroepithelium in the mutant.

(E and F) Similarly as in (B and C), the thickness of the Sox2+ cell layer in the VZ and the number of Sox2+ cells outside the VZ/SVZ were measured (n = 4).

(G) Immunostaining to compare wild-type and mutant Sox2+ cells at E12.5.

(H) Thickness of the Sox2+ layer at the VZ is comparable between the control and cKO cerebral cortices at E12.5 (n = 3).

Abbreviations: HF, hippocampal fissure; HNE, hippocampal neuroepithelium; Pri-HP, primordial hippocampus. Other abbreviations are listed in the legend to Figure 1. Scale bars, 500 μm for the left panels and 100 μm for the rest of the panels of (A), (D), and (G). ns, not statistically significant; **, p < 0.01; ***, p < 0.001.

We next analyzed different embryonic and brain sections using an antibody specific to the transcription factor Tbr2, a marker of neuronal progenitors critical for neocortex and hippocampus development (Arnold et al., 2008, Hodge et al., 2012). As shown in Figures 4A and 4B, the Tbr2+ progenitor number was reduced in the mutant ventricular/subventricular zone at P0. Reminiscent of what was observed with Sox2+ cells (Figure 3), more Tbr2+ cells were found outside the mutant ventricular/subventricular zone (Figures 4A and 4C). Similar to the mutant ventricular/subventricular zone, the Tbr2+ progenitor cell number was greatly reduced in the mutant hippocampus (Figure 4A). Consistent with this, few Tbr2+ progenitor cells were detected in the mutant hippocampus at P7 (Figure S6A). Moreover, at both E16.5 (Figures 4D and 4E) and E14.5 (Figures 4G and 4H), the Tbr2+ progenitor cell number was significantly decreased in the mutant neocortices. As with the neonatal mutant brain, more ectopic Tbr2+ cells were also observed outside the ventricular/subventricular zone at both E16.5 (Figure 4F) and E14.5 (Figure 4I), indicating impaired migration of the progenitors in the mutant brain. In contrast to what was observed at P0, E16.5, and E14.5, the Tbr2+ progenitor cell number remained comparable between the wild-type and mutant neocortical primordium at E12.5 (Figures S6B and S6C), indicating that HDAC3 is not essential for the formation of the Tbr2+ progenitor cell layer at this embryonic stage. Thus, Hdac3 deletion reduces the number of NSPCs required for neocortical development starting at E14.5.

Figure 4

Cerebral Hdac3 Deletion Depletes Tbr2+ Neuronal Progenitors

(A–C) (A) Immunostaining analysis with anti-Tbr2 antibody to detect Tbr2+ progenitors in brain sections from wild-type and mutant neonates. (B and C) Quantification of Tbr2+ progenitors from sections as shown in (A) (n = 4).

(D) Same as in (A) except E16.5 embryo sections were analyzed (n = 4). Dashed lines denote borders of the hippocampus and neocortex.

(E and F) Quantification of Tbr2+ progenitors from sections as shown in (C). n = 3.

(G) Same as (A) except that E14.5 embryonic sections were analyzed.

(H and I) Quantification of Tbr2+ progenitors from sections as shown in (G). n = 3. The number of Tbr2+ progenitor cells in the mutant is significantly reduced at E14.5 and E16.5 but comparable with the wild-type at E12.5 (Figures S6B and S6C). Red asterisks in (A), (D), and (G) mark ectopic distribution of Tbr2+ progenitors. Scale bars, 500 μm for the left panels and 100 μm for the rest of the panels of (A), (D), and (G). **, p < 0.01; ***, p < 0.001.

See also Figures S6A–S6C.

The hippocampal fissure is formed at E17.5, and its boundary within the CA1-CA3 regions also starts to emerge at this time point (Urban and Guillemot, 2014). As described earlier (Figure 3A, right panels), the upper blade of the dentate gyrus was almost missing and there were fewer Sox2+ cells in the dentate gyrus at P0. At E16.5, noticeably fewer Sox2+ cells were present in the hippocampal neuroepithelium area (Figure 3D, right panels). The hippocampus (or its primordium) in the mutant cerebral cortices at P0 and E16.5 were smaller when compared with the wild-type (Figures 3A and 3D). At P0, there were fewer Tbr2+ cells in the mutant hippocampus than the wild-type (Figure 4A). One week later, this population of Tbr2+ cells disappeared in the mutant hippocampus (Figure S6A). Thus, cerebrum-specific Hdac3 loss reduces the number of NSPCs required for hippocampus development.

Hdac3 Deletion Induces Premature Neurogenesis and Disrupts Neuronal Migration

Embryonic NSPCs either proliferate to replenish the population or differentiate to generate new neurons for proper brain development (Cattaneo and McKay, 1990). To investigate whether neurogenesis from NSPCs is affected, we performed immunostaining of embryo sections using Tuj1 antibody, which recognizes a neuron-specific β-tubulin (You et al., 2015a, You et al., 2015b). As shown in Figures S7A and S7B, thickness of the Tuj1+ cell layer was similar between the wild-type and mutant brains at E13.5. This became slightly different at E14.5. Although the cerebral cortex remained similar between wild-type and mutant brains, the Tuj1+ cell layer was thicker in the mutant (Figures S7C and S7D), indicating that in the absence of Hdac3, neurogenesis is enhanced at this developmental stage. At E15.5, the Tuj1+ cell layer was comparable between the wild-type and mutant brains but the cerebral cortex was thinner in the mutant (Figures S7E and S7F), so the ratio of the Tuj1+ cell layer to the neocortex was still higher in the mutant. These results indicate that the neurogenic zone expands at both E14.5 and E15.5. Related to this, at both developmental stages, the ventricular/subventricular zone was compressed in the mutant (Figures S7C and S7E, labeled with green asterisks). Moreover, ectopic distribution of Tuj1+ neurons was also observed in the mutant VZ/SVZ (Figures S7C and S7E), indicative of aberrant birth of Tuj1+ neurons from NSPCs. These results indicate that Hdac3 deletion leads to premature neurogenesis.

Embryonic NSPCs are subject to a delicate balance between self-renewal (proliferation) and differentiation (neurogenesis). To investigate whether NSPC proliferation is affected, we examined the S phase by carrying out BrdU labeling in vivo (You et al., 2015a, You et al., 2015b). For the assays, BrdU was injected into pregnant mice at E12.5 or E13.5 for embryo retrieval 1 h later. Indirect immunofluorescence microscopy was then performed with embryonic sections using anti-BrdU and -Ki67 antibodies. As shown in Figures S6D–S6G, no difference was observed between the mutant and wild-type at either developmental time point. Considering that Sox2+ NSCs or Tbr2+ progenitors are comparable between the wild-type and mutant neocortices at E12.5 (Figures 3G, 3H, S6B, and S6C), these results indicate that NSPC proliferation is not affected.

For neurogenesis, NSPCs exit the cell cycle for neuronal differentiation. To gain mechanistic insights into the observations about premature neurogenesis, we analyzed cell cycle exit (You et al., 2015a, You et al., 2015b). For this, BrdU was injected into pregnant mice at E12.5 and E14.5 for embryo retrieval 24 h later at E13.5 and E15.5, respectively. Indirect immunofluorescence microscopy was then performed with anti-BrdU and -Ki67 antibodies. As shown in Figures S8A–S8C, compared with the wild-type, there were significantly more BrdU+Ki67- cells (green) in the mutant ventricular zone at E13.5 and E15.5, indicating that more NSPCs exit the cell cycle at 24 h after BrdU injection. This supports enhanced neurogenesis in the mutant.

From immunostaining, we observed ectopic distribution of Sox2+ NSCs (Figure 3), Tbr2+ progenitors (Figure 4), Tuj1+ neurons (Figure S7), and cortical layer-specific neurons (Figures 2, S3, and S4). These observations suggest that neuronal migration might be affected. To evaluate this directly, we injected BrdU into pregnant mice at E12.5 or E14.5 for embryo retrieval 2 or 4 days later at E16.5. As shown in Figures S8D–S8E, neuronal migration was compromised in the mutant neocortex. Specifically, for BrdU tracing from E12.5 to E16.5, a majority of BrdU+ cells were observed in the cortical plate in the wild-type neocortex (Figure S8D, upper panel), whereas such cells were rather evenly distributed throughout the mutant neocortex (Figure S8D, lower panel). For BrdU tracing from E14.5 to E16.5, a majority of BrdU+ cells were localized to the wild-type ventricular/subventricular zone (Figure S8E, upper panel), whereas such cells were more evenly distributed throughout the mutant neocortex, with a lot of BrdU+ cells in the cortical plate (Figure S8E, lower panel). These results indicate impairment of neuronal migration in the mutant cerebral cortex.

Compared with the wild-type, there were more Ki67+ cells in the mutant immediate zone and cortical plate at E15.5 (Figure S8B) and E16.5 (Figures S8D–S8E). Quantification confirmed this assessment (Figure S8F). These results indicate that proliferating cells are ectopically located outside the ventricular/subventricular zones in the mutant at E15.5 and E16.5. Related to this, Sox2+ NSCs and Tbr2+ progenitors were ectopically located in the mutant (Figures 3 and 4). Thus, alteration of NSPC distribution is a prominent feature in the mutant.

Hdac3 Deletion Augments DNA Damage Response and Triggers Apoptosis

Global deletion of Hdac3 in embryonic fibroblasts induces DNA damage and triggers apoptosis (Bhaskara et al., 2008), so we investigated whether cerebrum-specific loss of Hdac3 causes similar defects. For this, immunostaining of embryonic and brain sections with an anti-γH2A.X antibody was used to detect DNA damage responses and TUNEL assays were employed to assess apoptosis. As shown in Figures 5A and 5B, the number of γH2A.X+ cells increased dramatically in the neonatal mutant neocortex. Such increase was found even at E12.5 (Figure 5C), when the NSPC number was normal (Figures 3G, 3H, S6B, and S6C). TUNEL assays revealed that apoptosis increased in the mutant at both E14.5 (Figure 5D and 5E) and P0 (Figures 5F and 5G), with a massive amount of apoptotic cells in the mutant VZ/SVZ at E14.5 (Figures 5D and 5E). Immunostaining with an antibody recognizing activated caspase 3 also detected an increase of apoptosis in the mutant at both E14.5 and P0 (Figures 5H and 5I). Together, these results indicate that DNA damage and apoptosis contribute to NSPC depletion in the mutant.

Figure 5

Cerebral Hdac3 Loss Leads to DNA Damage and Apoptosis

(A) Immunostaining of wild-type and mutant neonatal brain sections with anti-γH2A.X antibody.

(B) Quantification of sections as shown in (A) indicated more γH2A.X+ cells in the mutant cerebral cortex than in the wild-type (n = 3).

(C) Quantification of sections as in (B) except that E12.5 embryonic sections were analyzed (n = 3).

(D) TUNEL staining in control and cKO embryonic brain sections at E14.5.

(E) Quantification of TUNEL-positive cells in embryonic brain sections as shown in (D) (n = 3).

(F) TUNEL staining of control and cKO neonatal brain sections (n = 3).

(G) Quantification of TUNEL-positive cells in neonatal brain sections as shown in (F). n = 3.

(H) Immunostaining of control and cKO embryonic sections with an anti-activated caspase 3 antibody. Cerebral cortical areas are shown.

(I) Quantification of cells positive for activated caspase 3 in cerebral cortices at E14.5 and P0 (n = 3 for each group). Scale bars, 500 μm for the left panels and 100 μm for the rest of the panels of (A), (D), and (F); and 100 μm for (H). *, p < 0.05; **, p < 0.01; ***, p < 0.001.

See also Figures S6D–S6G, S7, and S8.

Mutant NSPCs Are Impaired in Forming Neurospheres In Vitro

To investigate whether the residual mutant NSPCs are still functional, we conducted neurosphere formation assays, which have been widely used to determine NSPC potential in vitro (Singec et al., 2006). For these assays, we collected cerebral cortices from wild-type and mutant embryos at E16.5 to prepare single cell suspension for subsequent neurosphere formation in an NSC culture medium in vitro. Like the wild-type, single cells from the E16.5 mutant cerebral cortex formed round neurospheres in the first 4 days (Figure 6A), but the mutant neurospheres became irregular at day 6 and afterward (Figure 6B). Even at day 4, more dead cells were visible in the medium from the mutant neurospheres. Quantification revealed that at this time point, there were significantly fewer mutant neurospheres than the wild-type (Figure 6C). Measurement of neurosphere sizes indicated that, at days 4 and 9, the mutant neurospheres were significantly smaller than the wild-type (Figure 6D). Although there was no difference between the wild-type and heterozygous groups in terms of the total cell number, the homozygous mutant neurospheres contained fewer cells than the wild-type (Figure 6E).

Figure 6

HDAC3 Is Essential for Neurosphere Formation In Vitro

(A) Normal round morphology of mutant neurospheres at day 4. In the first 4 days of culturing, the morphology of Hdac3-deficient neurospheres was similar to that of the wild-type counterparts.

(B) Photos of wild-type and mutant neurospheres at days 6 and 9 of culturing.

(C) Quantification of neurospheres formed from 2 × 105 single cells prepared from the E16.5 wild-type or mutant cerebral cortex (n = 3).

(D) Diameter of neurospheres grown for 4 days from the cerebral cortices of E16.5 control and mutant embryos (>50 neurospheres were measured for each group).

(E) Total cell number in neurospheres from wild-type, heterozygous (HET), and homozygous mutant (cKO) cerebral cortices. n = 5, 2, and 3 for the wild-type, heterozygous, and cKO groups, respectively.

(F) Morphology of E18.5 primary neurospheres at days 4 and 7 of culturing. Wild-type neurospheres were large and round, whereas the mutant ones were small and irregular.

(G) Microscopic photos of E18.5 neurospheres immunostained with anti-Sox2 antibody and counterstained with DAPI.

(H) Quantification of neurospheres formed from 10,000 single cells prepared from E18.5 wild-type or mutant primary neurospheres (n = 3).

(I) Treatment of E18.5 wild-type neurospheres using the HDAC3-specific inhibitor RGFP966 at 0, 0.08, and 0.8 μM for the indicated days. The IC50 value is 0.08 μM (Malvaez et al., 2013). The vehicle DMSO was maintained at the same amount for all conditions.

(J) Immunoblotting analysis of protein extracts from wild-type neurospheres treated with RGFP966 at 0, 0.08, and 0.8 μM. The inhibitor was added to the medium at ~12 h after neurosphere culturing started. After 24 h, cells were collected for immunoblotting with anti-acetylated histone H3 antibodies to detect changes of H3K9 and H3K27 acetylation.

Quantification of anti-H3K9 (K) and H3K27 (L) acetylation from two independent groups as shown in (J). For each sample, the acetylation levels were normalized to the total histone H3 level as the internal control. ns, not statistically significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001. Scale bars, 100 micrometers for panels (A), (B), (F), (G) and (I).

We also collected wild-type and mutant cerebral cortices at E18.5 to grow neurospheres. Strikingly, those from the E18.5 mutant embryos were irregular even at day 4 of culturing (Figure 6F). To examine the NSC nature of neurospheres, we immunostained them with the anti-Sox2 antibody. As shown in Figure 6G, there was a high ratio of Sox2+ cells in both wild-type and mutant neurospheres, confirming of their NSC nature. These results indicate that at E16.5 and E18.5, neurosphere formation from the mutant is impaired.

To investigate the impact of Hdac3 inactivation on NSC self-renewal, we analyzed formation of secondary neurospheres. For this, primary neurospheres were cultured as described earlier and used for preparation of single-cell suspension through pipetting; 10,000 cells were used for neurosphere formation in the NSC medium. As shown in Figure 6H, single cells prepared from the mutant neurospheres yielded a much smaller number of neurospheres, indicating impairment of self-renewal in the mutant NSCs.

HDAC3 is a transcriptional coregulator with intrinsic deacetylase activity (Emiliani et al., 1998, Yang et al., 1997), so we took a pharmacological approach to substantiate the above-mentioned conclusions drawn from genetically modified NSPCs. For this, we utilized an HDAC3-specific inhibitor, RGFP966 (Malvaez et al., 2013). Single cell suspension was prepared from E16.5 wild-type cerebral cortex for neurosphere formation in the presence or absence of this inhibitor. As shown in Figure 6I, treatment with this inhibitor even at the IC50 value of 0.08 μM was sufficient for interfering with neurosphere formation, indicating that HDAC3 inhibition abolishes neurosphere formation. These results support that HDAC3 is essential for NSPC homeostasis.

Hdac3 Deletion Alters Epigenetic Marks Differently in the Cerebrum and Neurospheres

To shed light on how Hdac3 deletion affects cerebral development and NSPC homeostasis, we asked how cerebrum-specific Hdac3 deletion may alter the epigenome. We thus determined levels of histone marks using western blotting analyses with various histone modification-specific antibodies. Protein extracts of cerebral cortices were prepared from wild-type and mutant pups at P4, because at this stage the mutant pups were still grossly normal (Figures 1A–1C). For histone H3, the levels of the marks H3K9ac, H3K18ac, and H3K27ac, but not H3K4ac, H3K14ac, and H3K23ac, increased in the mutant cerebral cortex (Figure 7A, left and middle panels). Interestingly, the H3K14pr level also increased. Consistent with the increased levels of H3K9ac and H3K27ac, trimethylation at both sites, but not H3K4, decreased (Figure 7A, middle panels). Notably, the decrease at the H3K27me3 level was much more dramatic than at H3K9me3 (Figure 7A, middle panels). Consistent with this, immunostaining of brain sections revealed that the H3K27me3 level decreased in the mutant cerebral cortex, whereas the change was not so obvious for H3K9me3 (Figure S9). As with histone H3, histone H4 was also hyperacetylated (Figure 7A, right top panel). In support of this, the H4K16ac level increased in the mutant cerebral cortex (Figure 7A, right bottom panel). Moreover, the H4K16pr level increased as well. These results indicate that cerebrum-specific Hdac3 deletion causes histone Hyperacetylation and hyperpropionylation and deregulates other histone modifications in the cerebrum.

Figure 7

Hdac3 Deletion Alters Epigenetic Marks in the Cerebral Cortex and Neurospheres

(A) Western blotting analysis of protein extracts from cerebral cortices of the control and mutant pups at P4. In agreement with Figures 1E and S1L, HDAC3 was efficiently deleted (compare the bottom two blots at the left). Modification-specific antibodies were used to detect changes of histone acetylation (ac), trimethylation (me3), propionylation (pr), and crotonylation (Cr). Notably, the deletion led to histone hyperacetylation, hyperpropionylation (e.g., H3K14pr and H4K16pr), and hypomethylation at multiple sites. Relative abundance of each modification was normalized to the corresponding histone H3 or H4 level; the ratio of the normalized modification level in the mutant was then calculated against the wild-type and shown at the right of each blot. The blots were repeated at least once, with similar results.

(B) Same as (A) except that protein extracts were from neurospheres cultured from control and mutant cerebral cortices of E16.5 embryos.

(C) Hdac3 deletion alters transcription of multiple genes related to NSPC and cell lineage development in neonatal cerebral cortex. Illustrated are FPKM values of genes related to NPSC and neural lineage development, as well as other genes important for brain development. The values were from RNA-seq of wild-type and mutant neonatal cerebral cortices (n = 2 for each group).

(D) Hdac3 loss deregulates transcription of multiple genes related to neurogenesis in neurospheres. Illustrated are FPKM values of pro- and anti-neurogenesis genes, as well as some other genes related to NSC function (based on RNA-seq of wild-type and mutant neurospheres cultured from E16.5 cerebral cortices for 4 days in vitro [n = 2 for each group]). Fragments Per Kilobase of transcript per Million mapped reads (FPKM) values in (C and D) are from Tophat analysis (complete tables available from GSE133195).

(E) A model on how specific DNA-binding transcription factors may recruit HDAC3 and its associated subunits for genomic locus-specific deacylation (including deacetylation and depropionylation) during NSPC and cerebral development. Other transcription factor partner candidates include c-Jun and HIF1, both of which are not only highly expressed in the neonatal cerebral cortex and E16.5 neurospheres (GSE133195) but also able to interact with the HDAC3 complexes via GPS2 in non-neural tissues (Drareni et al., 2018, Fan et al., 2016).

See also Tables S2 and S3 and Figures S9–S12.

We also performed western blotting with protein extracts from wild-type and mutant neurospheres. As shown in Figure 7B (bottom left two panels), Hdac3 was efficiently deleted in neurospheres. This is consistent with the results for the cerebral cortex (Figure 1E) and its embryonic precursors (Figures S1A–S1I). As expected, H3K4 acetylation slightly increased, whereas H3K9 or H3K27 acetylation increased in the mutant neurospheres (Figure 7B, top left panels). This increase was also evident in neurospheres by treatment with the HDAC3 inhibitor (Figures 6J–6L). Unexpectedly, instead of hyperacetylation observed with the mutant cerebral cortex (Figure 7A), H3K14, H3K18, and H3K23 acetylation decreased in the mutant neurospheres (Figure 7B, middle left panels), perhaps due to indirect effects. As for methylation, the H3K4me3 level decreased, the H3K9me3 level increased, and the K3K27me3 level remained unaltered in the mutant neurospheres (Figure 7B, bottom left panels). About histone H4, its total acetylation level decreased in the mutant neurospheres, whereas the H4K16ac level was not altered (Figure 7B, right panels). Thus, unlike what was observed with the cerebral cortex (Figure 7A), Hdac3 deletion causes histone hypoacetylation at some sites in neurospheres (Figure 7B). Differential impact in the cerebrum and neurospheres suggests cell type-specific effects.

HDAC3 Regulates Different Transcription Programs in the Cerebrum and Neurospheres

To identify molecular mechanisms by which Hdac3 deletion affects cerebral development and NSPC homeostasis, we performed transcriptomic analyses. For this, we extracted total RNA from cerebral cortices from wild-type and mutant newborns (P0). Poly(A) RNA profiling revealed altered transcription of multiple genes related to NSPC or lineage cell development (Figure 7C). With the Log2FC (fold change) value of 0.8 as the cutoff, we identified 58 upregulated genes and 58 downregulated genes (see analysis results at GEO: GSE133195). The top upregulated gene is Gfap, with Ngfr at the fifth and two non-coding RNA genes, Neat1 and Malat1, at the sixth and seventh positions, respectively. To evaluate the quality of RNA-seq datasets, we selected 19 significantly upregulated genes and 1 downregulated gene for validation by RT-qPCR. As shown in Figures S11A–S11D, the validation success rate was 75%, attesting to the quality of the datasets. In particular, RT-qPCR detected a dramatic increase in Gfap expression (Figure S11A), consistent with immunostaining results showing increased astrogliogenesis in the mutant brain sections at P0 (Figures S5D–S5F). This is in agreement with what was reported (Norwood et al., 2014). However, the pan-HDAC inhibitor trichostatin A reduces GFAP expression in primary human astrocytes and astrocytoma cells (Kanski et al., 2014), so other HDACs may be required for the expression. Among the other validated candidates are Ngfr, Tnc, Bcas1, Nefl, Hopx, Sgk1, Nr4a2, Id1, Atf5, and Sox1 (Figures S11A and S11B), most of which have known roles in the brain. The non-coding RNA genes Neat1 and Malat1 were also validated (Figures S11A–S11D), suggesting their possible roles in cerebral development.

To draw further insights from RNA-seq datasets, we carried gene ontology (GO) analysis. The Log2FC value of 0.8 was used as the cutoff to select top upregulated or downregulated genes for analyses via Enrichr (Kuleshov et al., 2016) and Metascape (Zhou et al., 2019b). Consistent results were obtained by both bioinformatic tools. The results indicated that neuropeptide signaling pathway (GO: 0007218), positive regulation of neuron differentiation (GO:0045666), negative regulation of neuron differentiation (GO:0045665), generation of neurons (GO:0048699), and regulation of neuron differentiation (GO:0045664) were among the top 15 GO categories affected in the mutant neonatal cerebral cortex, with p values ranging from 0.0015 to 5.8 × 10−5. Related to neuropeptide signaling pathway, Ngfr was one of the top upregulated gene, as supported by RT-qPCR results (Figures S11A).

To dissect out the molecular mechanisms further, we extracted RNA from E16.5 primary neurospheres cultured at day 4, when the morphology was similar between wild-type and mutant neurospheres (Figure 6A). RNA-seq and gene differential analysis identified 68 upregulated (fold change>2) and 75 downregulated (fold change<0.5) genes (Tables S2 and S3), including increased transcription of pro-neurogenesis genes such as the Dlx family genes and decreased transcription of anti-neurogenesis genes such as the Id and Hes family genes (Figure 7D). These changes were supported by RT-qPCR (Figures S11E–S11F). To evaluate the quality of RNA-seq datasets, we selected 11 upregulated genes and 5 downregulated genes for validation by RT-qPCR. As shown in Figures S11E and S11F, the success rate was 94%, attesting to the quality of the datasets. Transcript levels of many other neurogenesis-promoting genes (such as Sox1, Pbx, Lhx6, Gad1, Gad2, Cux2, Tbr1, Dcx, Erbb4, NeuroD1, and NeuroD2) increased, whereas the levels of some anti-neurogenesis genes (such as Notch3, Sox10, Ascl1, Tnc, and Nes) decreased (Figure 7D). These results indicate that Hdac3 deletion upregulates the expression of multiple pro-neurogenesis genes and inhibits the expression of some anti-neurogenesis genes. Extracellular matrix organization (GO:0030198), regulation of cell migration (GO:0030334), extracellular matrix disassembly (GO:0022617), central nervous system development (GO:0007417), and brain development (GO:0007420) were among the top 17 affected GO categories, with p values ranging from to 2.7 × 10−5 to 1.3 × 10−16.

Consistent with immunostaining with anti-BrdU and -Ki67 antibodies at E12.5 and E13.5 (Figures S6D–S6G), expression of cell cycle inhibitors such as p27, p57, p15, and p53 was not affected in the mutant P0 cerebral cortex or E16.5 neurospheres (Figures S10A and S10B). p21 (a.k.a. Cdkn1a) transcription increased in the Hdac3-deficient P0 cerebral cortex by ∼2-fold (Figures S10A and S10B), whereas p18 expression only decreased slightly in E16.5 mutant neurospheres (Figure S10B). Moreover, the p53 and Ki67 transcript levels were not significantly affected in the mutants (Figures S10A and S10B). These results indicate that overall proliferation is not dramatically affected in the mutant cerebral cortex or neurospheres. As the results are from bulk RNA-seq and RT-qPCR, effects at specific cerebral regions or cell types remain possible. Indeed, there were some effects on the Ki67+ cell number at E15.6 and E16.5 (Figure S8).

HDAC3 Binds Multiple Transcription Factors Abundant in the Cerebrum and Neurospheres

As an enzymatic transcriptional coregulator (Yoon et al., 2003, Zhang et al., 2002), HDAC3 is recruited to specific promotors for gene-specific regulation. To identify DNA-binding transcription factors that target it to specific genetic loci during cerebral development, we analyzed public ChIP-seq datasets on HDAC3 (Armour et al., 2017, Emmett et al., 2017, He et al., 2018, Hong et al., 2017, Nanou et al., 2017, Remsberg et al., 2017). DNA sequence motif analysis identified multiple potential candidates, including MEF2, COUP-TFI, nuclear factor-I (e.g., NFIB and NFIC), SOX4, and C/EBPγ. Important for brain development (Heng et al., 2015, Ohtaka-Maruyama et al., 2013), RP58 contains a POZ domain that may interact with the NCoR/SMRT corepressors (Huynh and Bardwell, 1998), so we included this repressor as a candidate. According to RNA-seq, Rp58, Nfib, Coup-tfi, Sox4, and Mef2c are well expressed in the cerebral cortex (Figure S12A) and neurospheres (Figure S12B). Immunostaining showed that NFIB is abundant in the cortical plate and SVZ/VZ zone at E16.5 but not in the immediate zone (Figure S12C, topper panel). Hdac3 deletion altered this distribution pattern and made it an even distribution in the cerebral cortex (Figure S12C, lower panel). These results indicate that HDAC3 affects NFIB expression in different cerebral areas during brain development.

Because MEF2C is a known binding partner of HDAC3 (Grégoire et al., 2007), we assessed the interaction of HDAC3 with NFIB and the other transcription factors mentioned earlier. FLAG-tagged HDAC3 was expressed with HA-tagged NFIB, NFIC, SOX4, C/EBPγ, and RP58 in HEK293 cells for co-immunoprecipitation. As shown in Figure S12D, HDAC3 co-immunoprecipitated NFIB, NFIC, and SOX4. Moreover, HDAC3 interacted with a truncated form of NFIB containing its N-terminal 200 residues (Figure S12D, lane 3). This region is conserved among NFIB, NFIC, and other NFI proteins (Figure S12E), suggesting that HDAC3 interacts with different NFI proteins. We also expressed FLAG-tagged HDAC3 alone in HEK293 cells to detect its interaction with endogenous COUP-TFI, NFIB, and RP58 proteins using antibodies against these proteins. As shown in Figures S12F–S12J, FLAG-tagged HDAC3 co-immunoprecipitated endogenous COUP-TFI and NFIB. The results suggest that several transcription factors, including NFI, COUP-TFI, and SOX4, recruit HDAC3 for gene-specific gene regulation during cerebral development.

Discussion

This study uncovers an important role of Hdac3 in governing perinatal cerebral development through NSPCs. Immunostaining revealed that HDAC3 is highly expressed in the developing mouse brain (Figure S1). Cerebrum-specific inactivation of the gene induces hyperactivity, anxiety, and early lethality in the pups, along with severe developmental defects such as neocortical disorganization, hippocampal hypoplasia, and callosal agenesis (Figure 1). As for the underlying cellular mechanisms, immunofluorescence microscopy revealed rapid loss of NSCs (Figure 3) and neuronal progenitors (Figure 4), resulting from premature neurogenesis (Figure S7) and enhanced DNA damage and apoptosis (Figure 5). Mutant NSPCs were also ectopically distributed to the cortical plate (Figures 3 and 4 and S8B–S8F). They were also defective in forming neurospheres in vitro (Figures 6A–6F), and an HDAC3-specific inhibitor abolished neurosphere formation by wild-type NSPCs (Figure 6I). Related to the molecular mechanisms (Figure 7), immunoblotting unveiled alteration of histone modifications in the mutant cerebral cortex and neurospheres (Figures 7A and 7B). Furthermore, RNA-seq not only revealed perturbation of transcription programs in the mutant cerebral cortex and neurospheres (Figures 7C and 7D) but also unmasked high expression of multiple HDAC3-interacting transcription factors in the wild-type cerebrum and neurospheres (Figures S12A and S12B). These results identify HDAC3 as a deacetylase critical for perinatal cerebral and NSPC development (Figure 7E).

These results complement and extend a recent report on the role of Hdac3 in brain development (Norwood et al., 2014). In the report, the Nestin-Cre strain was used and the resulting ablation induced neonatal lethality, thereby preventing analyses of neocortical development within the first 2 weeks after birth, when neocortical lamination occurs dynamically (Caviness et al., 2008). This Cre strain is also inefficient for inducing excision at E12.5–14.5 (Liang et al., 2012, Liang et al., 2013), whereas the Emx1-Cre line used herein confers efficient excision specifically in the neural stem/progenitor cells and cortical neurons starting at E12.5 (Figure S1F) (Gorski et al., 2002). Owing to cerebrum-specific expression, the resulting mutant pups survived for ∼4 weeks (Figure 1A), permitting systematic analysis of neocortical and hippocampal development at different time points before and after birth. Several reports have described Hdac3 deletion in the adult brain and showed that HDAC3 is important for learning and memory (McQuown et al., 2011, Nott et al., 2016), motor coordination (Norwood et al., 2014, Nott et al., 2016), social behavior (Nott et al., 2016), and drug addiction (Rogge et al., 2013). HDAC3 is also important for glial cell fate determination (Zhang et al., 2016) and myelin growth and regeneration (He et al., 2018). The present study highlights an important role of HDAC3 in prenatal neocortical and NSPC development. No reports have been made about the function of HDAC3 in the development of the hippocampus and corpus callosum, so our study fills the knowledge gap.

Our results showed that cerebrum-specific deletion of Hdac3 promotes premature neurogenesis (Figure S7). Moreover, in the mutant neurospheres, expression of pro-neurogenesis genes (such as Dlx family members) was upregulated, whereas expression of anti-neurogenesis genes was downregulated (such as Id and Hes family members) (Figures 7D, S11E, and S11F). These results are consistent with a recent report showing that Hdac3 knockdown initiates a neuronal differentiation program in vitro (Castelo-Branco et al., 2014). HDAC3 is the catalytic subunit of tetrameric complexes composed of NCoR (or SMRT), TBL1 (or its two paralogs), and GPS2 (Karagianni and Wong, 2007, Oberoi et al., 2011). These complexes function as transcriptional corepressors upon transcription factor-dependent recruitment to specific promotors. Consistent with the important role of HDAC3 in NSPCs (Figures 2, 3, and 4), Smrt−/− embryos display NSC defects (Jepsen et al., 2007) and Ncor−/− NSCs show enhanced astrogliogenesis in vitro (Hermanson et al., 2002). In agreement with premature astrogliogenesis upon Ncor deletion, Gfap transcription was elevated in Hdac3−/− neurospheres (1.5-fold) and cerebral cortices (4.3-fold). Moreover, immunostaining detected strong GFAP signals in Hdac3−/− cerebral cortices (Figures S5D–S5F). As shown in Figures S10C and S10D, both Ncor and Smrt are highly expressed in the wild-type cerebral cortex and neurospheres, suggesting redundant roles of these two paralogs. Together with the published reports on mouse Ncor (Hermanson et al., 2002) and Smrt (Jepsen et al., 2007), this study supports a model in which HDAC3 complexes govern pre- and postnatal NSPC and cerebral development (Figure 7E). Further genetic studies are needed to pinpoint the exact roles of different subunits in regulating cerebral development.

As for DNA-binding proteins that recruit these complexes for gene-specific action, RNA-seq uncovered high expression of multiple HDAC3-interacting transcription factors, including NFIB, COUP-TFI, and MEF2C, in the wild-type cerebrum and neurospheres (Figures S12A and S12B). These transcription factors are important in NSC maintenance and differentiation. For example, NFIB proteins are master regulators of neuronal differentiation (Becker-Santos et al., 2017). Disruption of Nr2f1 and Nr2f2 (encoding COUP-TFI and COUP-TFII, respectively) promotes neurogenesis and generation of early-born neurons in NSPCs and the developing forebrain (Naka et al., 2008). COUP-TFI is also important for oligodendrocyte differentiation (Yamaguchi et al., 2004). Co-immunoprecipitation confirmed the interaction of HDAC3 with NFIB and COUP-TFI (Figures S12D–S12I). MEF2C is important for NSPC development (Li et al., 2008) and interacts with HDAC3 (Grégoire et al., 2007). We thus propose that multiple DNA-binding proteins interact with and recruit HDAC3 complexes for epigenomic regulation during cerebral development (Figure 7E). Further studies with techniques such as bulk and single-cell ChIP-seq are needed to investigate how HDAC3 and its associated subunits are targeted to different genomic loci for governing development of the cerebrum and NSPCs.

HDAC3 complexes interact with MeCP2 (Kruusvee et al., 2017, Nott et al., 2016). RNA-seq revealed low Mecp2 expression in the developing cerebrum and cultured neurospheres (Figures S12A and S12B). Defects of Mecp2 mice are much milder (Chen et al., 2001, Guy et al., 2001) than those of the Hdac3 knockouts described here (Figure 1), suggesting that MeCP2 is not a major DNA-binding partner of HDAC3 during brain development. Rev-ErbA and Rev-ErbB are important transcription factors mediating functions of HDAC3 in controlling circadian clock and metabolism (Bugge et al., 2012). RNA-seq detected low expression of Rev-erba and Rev-erbb in the neonatal cerebrum and E16.5 neurospheres (Figures S12A and S12B), indicating that these two are not major transcription factors mediating HDAC3 functions during cerebral development.

The important role of HDAC3 in mouse cerebral development (Figure 1) begs the question whether this epigenetic regulator is essential for human cerebral and intellectual development. Consistent with the neocortical and hippocampal defects of Hdac3 knockout mice, five patients with HDAC3, NCOR1, or NCOR2 mutations are associated with cerebral anomalies and intellectual disability (Helbig et al., 2016, Zhou et al., 2019a). Moreover, multiple patients with TBL1XR1 mutations show intellectual disability (Heinen et al., 2016, Laskowski et al., 2016, Saitsu et al., 2014). Clinical features from these cohorts of patients shed light on potential problems of GPS2 mutations, even though such individuals have not been reported yet. The developmental defects described herein for the mutant mice serve a valuable guide for analysis of cerebrum-related clinical features in patients with mutations in the genes for HDAC3 and its associated subunits (Figure 7E). Such defects in mutant mice suggest a prenatal developmental origin for some clinical features in patients.

There are 18 human HDACs (Haberland et al., 2009, Yang and Seto, 2008b), so a key question is how each HDAC contributes to epigenomic regulation during cerebral development. The results described herein highlight the importance of HDAC3 in this regard. Published studies have identified important roles of HDAC1, HDAC2, HDAC4, and HDAC8 in the brain (Deardorff et al., 2012, Yang and Seto, 2008b). Among them, HDAC4 shows little detectable deacetylase activity and HDAC8 deacetylates cohesion (Deardorff et al., 2012, Yang and Seto, 2008b), so HDAC1 and HDAC2 are two deacetylases targeting histones in the brain. Mouse Hdac1 and Hdac2 have redundant roles in regulating brain development (Montgomery et al., 2009). Inactivation of both genes disrupts cell-cycle progression, leaving few S-phase cells (Montgomery et al., 2009). By contrast, BrdU incorporation was normal in Hdac3−/− neuroepithelia at E12.5 and E13.5 (Figures S6D–S6G). Although neuronal progenitors lacking Hdac1 and Hdac2 are defective in differentiation (Montgomery et al., 2009) and haploinsufficiency of SIN3A (encoding a subunit of an HDAC1/HDAC2 complex) inhibits cortical neurogenesis (Witteveen et al., 2016), Hdac3 deletion promotes premature neurogenesis and NSPC depletion (Figures 3 and 4). These studies support that HDAC3 is complementary to HDAC1 and HDAC2 during cerebral development. Despite minimal impact of Hdac3 deletion on expression of these two and other HDACs (Figures S10E and S10F), histones H3 and H4 are hyperacetylated in the mutant cerebral cortex (Figure 7A). In sum, this study identifies HDAC3 as a major deacetylase that regulates epigenetic and transcriptional programs important for perinatal cerebral and NSPC development in mice and perhaps also in humans.

Limitations of the Study

This study employs a mouse genetic approach that induces Hdac3 deletion at E12.5. Deletion at such an early embryonic stage complicates analysis of the roles at the late stages of development as it would be difficult to assess which stage an effect really occurs. We utilized bulk RNA-seq for transcriptomic analysis, from which we cannot distinguish between direct and indirect effects. Such bulk analysis also fails to differentiate effects from different cell types, so single-cell RNA-seq approaches should be advantageous in the regard.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.