ABSTRACT
Despite clear physiological roles, the ventromedial hypothalamus (VMH) developmental programs are poorly understood. Here, we asked whether the proneural gene achaete-scute homolog 1 (Ascl1) contributes to VMH development. Ascl1 transcripts were detected in embryonic day (E) 10.5 to postnatal day 0 VMH neural progenitors. The elimination of Ascl1 reduced the number of VMH neurons at E12.5 and E15.5, particularly within the VMH-central (VMHC) and -dorsomedial (VMHDM) subdomains, and resulted in a VMH cell fate change from glutamatergic to GABAergic. We observed a loss of Neurog3 expression in Ascl1−/− hypothalamic progenitors and an upregulation of Neurog3 when Ascl1 was overexpressed. We also demonstrated a glutamatergic to GABAergic fate switch in Neurog3-null mutant mice, suggesting that Ascl1 might act via Neurog3 to drive VMH cell fate decisions. We also showed a concomitant increase in expression of the central GABAergic fate determinant Dlx1/2 in the Ascl1-null hypothalamus. However, Ascl1 was not sufficient to induce an ectopic VMH fate when overexpressed outside the normal window of competency. Combined, Ascl1 is required but not sufficient to specify the neurotransmitter identity of VMH neurons, acting in a transcriptional cascade with Neurog3.
INTRODUCTION
The hypothalamus is a conserved brain region playing a crucial role in the maintenance of body homeostasis, including the regulation of energy balance, thirst, circadian rhythms, reproduction and social behaviors, such as anger and anxiety (Markakis, 2002; Settle, 2000; Stephan and Zucker, 1972; Swaab, 1995). The hypothalamus comprises 11 nuclei (Szarek et al., 2010), with each controlling various physiological processes (Machluf et al., 2011; Markakis, 2002; Szabo et al., 2009). Despite numerous studies exploring the physiological functions of the hypothalamus (Buckley and Schatzberg, 2005; Bose et al., 2009; Xu et al., 2011), the programs driving development of this brain region are poorly understood (Alvarez-Bolado et al., 2012; Bakos et al., 2016; Harder et al., 2018; Machluf et al., 2011; Nouri and Awatramani, 2017; Swaab, 1995).
The ventromedial hypothalamus (VMH) is located in the medial tuberal hypothalamus (King, 2006; Lu et al., 2013; McClellan et al., 2006; Van Houten and Brawer, 1978) and plays essential roles in regulating satiety (Bingham et al., 2008; Mobbs et al., 2013; Shimazu and Minokoshi, 2017; Scallet and Olney, 1986), aggression and anxiety (Cheung et al., 2015; Lin et al., 2011; Manoli et al., 2013; Yang et al., 2017; Zhao et al., 2008), and female sexual behavior (Georgescu et al., 2014; Sinchak et al., 1997; Krause and Ingraham, 2017). Structurally, the VMH can be roughly divided into three subdomains (Fig. S1): the dorsomedial VMH (VMHDM), central VMH (VMHC) and ventrolateral VMH (VMHVL) (Kurrasch et al., 2007; Millhouse, 1973). Functionally, VMHDM neurons express leptin receptors and are thought to mediate satiety signaling (Dhillon et al., 2006; Meister, 2000), whereas VMHVL neurons express estrogen and progesterone receptors, and play a role in female reproductive behaviors (Dugger et al., 2007; Grgurevic et al., 2012; Yang et al., 2017). To date, there is no known function associated with VMHC. The mechanisms governing the sorting of newborn VMH neurons into this tripartite structure remain unexplored.
Proneural genes encode basic-helix-loop-helix (bHLH) transcription factors that influence various developmental processes, such as neural cell fate specification, neuronal differentiation and migration (Huang et al., 2014; Imayoshi et al., 2013; Shimojo et al., 2008; Wilkinson et al., 2013). Originally identified in Drosophila as bHLH factors that confer a neural identity onto uncommitted ectodermal cells, proneural orthologs in vertebrates are instead expressed in progenitor cells that have already acquired a neural identity. In this study, we focused on the role of Ascl1 in VMH development. To date, proneural genes are well studied in the forebrain (Dixit et al., 2014; Parras et al., 2002; Schuurmans et al., 2004; Schuurmans and Guillemot, 2002; Fode et al., 2000), olfactory bulb (Cau et al., 1997; Zhang et al., 2007; Shaker et al., 2012), retina (Akagi et al., 2004; Brzezinski et al., 2011; Hufnagel et al., 2010; Pollak et al., 2013; Tomita et al., 1996) and spinal cord (Sugimori et al., 2007). In the ventral forebrain, Ascl1 is necessary and sufficient for conferring a GABAergic neuronal identity (Wilkinson et al., 2013; Fode et al., 2000; Casarosa et al., 1999; Horton et al., 1999), whereas in the spinal cord, Ascl1+ progenitors give rise to dorsal dl3-5 neurons (Müller et al., 2002; Helms and Johnson, 2003), as well as motor neurons and interneurons that reside ventrally (Li et al., 2005; Parras et al., 2002). Ascl1 also plays a role in gliogenesis, functioning in the telencephalon to promote oligodendrocyte differentiation (Nakatani et al., 2013; Wilkinson et al., 2013), and in the tuberal hypothalamus (Marsters et al., 2016) and spinal cord (Vue et al., 2014; Sugimori et al., 2007) to repress an oligodendrocyte fate. Finally, Ascl1 can also influence neuronal migration by regulating Rnd3 (Pacary et al., 2011).
Within the mediobasal hypothalamus, Ascl1 influences neurogenesis and neuronal subtype specification of GnRH- and NPY-expressing cells in the developing arcuate nucleus (ARC) (McNay et al., 2006). In addition, a decrease in expression of the pan-VMH marker steroidogenic factor 1 (SF1) was reported in the Ascl1 mutant hypothalamus (McNay et al., 2006); however, specific VMH neuronal populations were not examined. Furthermore, neurogenin 3 (Neurog3) is expressed in progenitors and postmitotic neurons in the developing ventral hypothalamus (Huang et al., 2014; Lee et al., 2003; Pelling et al., 2011), and is positively regulated by Ascl1 to promote pro-opiomelanocortin (POMC) and SF1 differentiation in ARC and VMH cells, respectively (Pelling et al., 2011). Neurog3 is also required for the specification of anorexigenic neurons in the arcuate nucleus (Anthwal et al., 2013) but its role in VMH subtype identification has not been explored. Here, we have focused on the VMH and examined whether the Ascl1-Neurog3 transcriptional circuit influences VMH subdomain-specific cell fate decisions during embryonic development.
RESULTS
Ascl1 is expressed in neural progenitors in the presumptive VMH
VMH development is divided into four main events: (1) regionalization of the VMH prosencephalic territory [at embryonic day (E) <10.5]; (2) VMH cell fate specification and differentiation (E10.5-15.5); (3) neuronal migration (E10.5-15.5); and (4) coalescing of VMH neurons into mature nuclei (E15.5-18.5) (McClellan et al., 2006; Nesan et al., 2018; Marsters et al., 2016). The cellular and molecular events that govern the patterning of VMH progenitors, as well as the cell fate decisions and migration of VMH neurons into distinct nuclear subdomains, are all poorly understood.
To determine whether the proneural gene Ascl1 might play a role in specifying the fates of VMH neurons, we first investigated the spatiotemporal expression pattern of Ascl1 in the developing murine tuberal hypothalamus (Fig. 1). Wild-type mouse brains were harvested at different embryonic (E10.5, E12.5, E15.5 and E17.5) and postnatal [postnatal day (P) 0] time points that collectively encompass the key stages of VMH development. We conducted in situ hybridization (ISH) with an Ascl1 riboprobe and assessed the distribution of Ascl1 transcripts within the ventricular zone (e.g. where progenitors reside) and mantle zones (e.g. where postmitotic neurons are positioned) of the VMH (Fig. 1A-E). Adjacent sections were immunostained for Nkx2.1 to confirm VMH location (Fig. 1A′-E′). Ascl1 transcripts were observed in neural progenitors that line the ventricular zone (VZ) starting at E10.5 and persisted through P0, with the highest expression levels evident in the early embryonic period (E10.5 and E12.5; Fig. 1A,A″,B,B″). Ascl1 transcript levels were reduced by E15.5, although expression was maintained through P0 (Fig. 1C-E,C″-E″). No Ascl1 transcripts were detected in the mantle zone where postmitotic neurons reside throughout the embryonic period.
Ascl1 is expressed within the VMH progenitor zone. (A-E) ISH results showing Ascl1 transcript distribution in mouse coronal sections at different VMH developmental time points: E10.5 (A), E12.5 (B), E15.5 (C), E17.5 (D) and P0 (E). Arrowheads point to the presumptive VMH progenitor zone. (A′-E′) Nkx2.1 staining used to define VMH boundaries across mouse embryonic development at E10.5 (A′), E12.5 (B′), E15.5 (C′), E17.5 (D′) and P0 (E′). (A″-E″) Schematic cartoon of the developing ventral hypothalamus with the emerging VMH highlighted at different developmental stages: E10.5 (A″), E12.5 (B″), E15.5 (C″), E17.5 (D″) and P0 (E″). n=3 embryos per group. 3V, third ventricle; VZ, ventricular zone; VMH, ventromedial hypothalamus. Red and yellow dashed lines represent rough boundaries of the VMH nucleus. White-dashed lines outline the third ventricle and pial surface. Scale bars: 100 µm (all except for A′ and B′); 50 µm (A′,B′).
Ascl1 is expressed within the VMH progenitor zone. (A-E) ISH results showing Ascl1 transcript distribution in mouse coronal sections at different VMH developmental time points: E10.5 (A), E12.5 (B), E15.5 (C), E17.5 (D) and P0 (E). Arrowheads point to the presumptive VMH progenitor zone. (A′-E′) Nkx2.1 staining used to define VMH boundaries across mouse embryonic development at E10.5 (A′), E12.5 (B′), E15.5 (C′), E17.5 (D′) and P0 (E′). (A″-E″) Schematic cartoon of the developing ventral hypothalamus with the emerging VMH highlighted at different developmental stages: E10.5 (A″), E12.5 (B″), E15.5 (C″), E17.5 (D″) and P0 (E″). n=3 embryos per group. 3V, third ventricle; VZ, ventricular zone; VMH, ventromedial hypothalamus. Red and yellow dashed lines represent rough boundaries of the VMH nucleus. White-dashed lines outline the third ventricle and pial surface. Scale bars: 100 µm (all except for A′ and B′); 50 µm (A′,B′).
Neurog3 transcription within VMH progenitors was reduced in the absence of Ascl1
There are several examples of transcriptional interactions between Ascl1 and members of the neurogenin (Neurog1, Neurog2 and Neurog3) gene family, including repressive interactions between Neurog1/Neurog2 and Ascl1 in the neocortex (Kovach et al., 2013) and retina (Hufnagel et al., 2010), and positive interactions in the arcuate nucleus, in which Ascl1 induces Neurog3 transcription (McNay et al., 2006). We thus asked whether the absence of Ascl1 affected the expression of Neurog1-3 within VMH progenitors using ISH and Neurog1, Neurog2 and Neurog3 riboprobes. Analysis of E12.5 wild-type and Ascl1GFPKI/GFPKI (e.g. Ascl1-null) brains (n=3) revealed that in the absence of Ascl1, Neurog1 (Fig. 2A,B) and Neurog2 transcription was not altered (Fig. 2C,D), whereas the normal expression of Neurog3 in the presumptive VMH was largely absent (Fig. 2E,F). In addition, to test whether Ascl1 was sufficient to induce Neurog3 transcription, we used in utero electroporation (IUE) to overexpress Ascl1 by injecting E12.5 wild-type embryos with pCIG2-Ascl1. We harvested the brains at E14.5 and conducted ISH using a Neurog3 riboprobe. Neurog3 transcripts were absent in control brains injected with pCIG2 at E14.5 (Fig. 2G,G′). In contrast, Neurog3 transcripts were observed in progenitors electroporated with pCIG2-Ascl1 (Fig. 2H,H′). Of note, the angle of the electroporation paddles in Fig. 2H,H′ caused bilateral electroporation of progenitors, consistent with the expression of both Neurog3 (Fig. 2H) and GFP (Fig. 2H′) on both sides of the third ventricle. These results suggest that Ascl1 does not regulate Neurog1 or Neurog2 transcription in the VMH, but positively regulates Neurog3 expression within the tuberal hypothalamus.
Ascl1 affects transcription of neurogenin 1, 2 and 3 within VMH progenitors. (A,B) ISH results showing Neurog1 transcript distribution in mouse coronal sections at E12.5 in control (A) and Ascl1−/− (B) backgrounds. (C,D) ISH results for Neurog2 transcript distribution in mouse coronal sections at E12.5 in control (C) and Ascl1−/− (D) backgrounds. (E,F) ISH results for Neurog3 transcript distribution in mouse coronal sections at E12.5 in control (E) and Ascl1−/− (F) backgrounds. (G,H) ISH results for Neurog3 transcript distribution in the C57BL/6 background injected with pCIG2 (G) and pCIG2-Ascl1 (H) constructs at E12.5 and brains harvested at E14.5. (G′,H′) Immunostaining for anti-GFP in the C57BL/6 background injected with pCIG2 (G′) and pCIG2-Ascl1 (H′) constructs at E12.5, and brains harvested at E14.5. n=3 embryos per group. Red arrowheads indicate electroporated VMH precursors. Scale bars: 100 µm.
Ascl1 affects transcription of neurogenin 1, 2 and 3 within VMH progenitors. (A,B) ISH results showing Neurog1 transcript distribution in mouse coronal sections at E12.5 in control (A) and Ascl1−/− (B) backgrounds. (C,D) ISH results for Neurog2 transcript distribution in mouse coronal sections at E12.5 in control (C) and Ascl1−/− (D) backgrounds. (E,F) ISH results for Neurog3 transcript distribution in mouse coronal sections at E12.5 in control (E) and Ascl1−/− (F) backgrounds. (G,H) ISH results for Neurog3 transcript distribution in the C57BL/6 background injected with pCIG2 (G) and pCIG2-Ascl1 (H) constructs at E12.5 and brains harvested at E14.5. (G′,H′) Immunostaining for anti-GFP in the C57BL/6 background injected with pCIG2 (G′) and pCIG2-Ascl1 (H′) constructs at E12.5, and brains harvested at E14.5. n=3 embryos per group. Red arrowheads indicate electroporated VMH precursors. Scale bars: 100 µm.
Ascl1 and Neurog2 give rise to distinct neuronal populations in the developing VMH
In many regions of the neural tube, Ascl1+ and Neurog1-3+ progenitors give rise to discrete cell lineages (Dennis et al., 2019). To assess whether these proneural genes are expressed in an overlapping or complementary manner in the VMH, we focused on Neurog2 as a representative member of the neurogenin gene family. E11.5 and E13.5 Neurog2GFP/+ sections were co-immunolabeled with anti-GFP and anti-Ascl1, using GFP as a short-term lineage trace to mark Neurog2+ progenitors and their progeny. At E11.5, GFP (i.e. Neurog2) and Ascl1 were both expressed in a highly overlapping fashion throughout the hypothalamic VZ, with some VMH progenitors co-expressing both proneural proteins (Fig. 3A-A‴), as observed in the neocortex (Britz et al., 2006), cerebellum (Chouchane and Costa, 2019; Zordan et al., 2008) and retina (Hufnagel et al., 2010). We also observed that GFP+ progenitors were located towards the outer edge of the VZ (e.g. closer to the pial surface; Fig. 3A′), whereas Ascl1+ progenitors were located closer to the third ventricle (Fig. 3A″), suggesting that Ascl1+ cells reside in a specific cell cycle phase and therefore occupy a distinct location of the pseudostratified neuroepithelium. By E13.5, Neurog2 expression was almost completely lost in the progenitors adjacent to the VMH, whereas Ascl1 expression was still at peak levels (Fig. 3B-B‴). Taken together, these data suggest that Neurog2 and Ascl1 are both expressed in presumptive VMH progenitors at E11.5, raising the question of whether these progenitors give rise to the same or different pools of VMH neurons.
Ascl1 and Neurog2 form distinct populations within VMH progenitors. (A-B‴) Co-expression of Ascl1 and Neurog2 in E11.5 (A-A‴) and E13.5 (B-B‴) Neurog2GFPKI/+ brains for anti-GFP (green), anti-Ascl1 (red) and Hoechst stain (blue) (n=3). White-dashed lines outline the third ventricle and pial surface. (C-J) RT-qPCR results for Ascl1 (C), Neurog2 (D), Vgll2 (E), Satb2 (F), Nkx2.1 (G), Rax (H), Shh (I) and Six3 (J) transcripts within the tuberal hypothalamus of Ascl1GFPKI; Neurog2GFPKI E12.5 embryos. Data are mean±s.e.m. (n=3 groups of pooled tuberal hypothalami from three embryos). *P<0.01; **P<0.005; ***P<0.0001; ns, not significant. Unpaired two-tailed Student's t-test. 3V, third ventricle. Scale bars: 50 µm.
Ascl1 and Neurog2 form distinct populations within VMH progenitors. (A-B‴) Co-expression of Ascl1 and Neurog2 in E11.5 (A-A‴) and E13.5 (B-B‴) Neurog2GFPKI/+ brains for anti-GFP (green), anti-Ascl1 (red) and Hoechst stain (blue) (n=3). White-dashed lines outline the third ventricle and pial surface. (C-J) RT-qPCR results for Ascl1 (C), Neurog2 (D), Vgll2 (E), Satb2 (F), Nkx2.1 (G), Rax (H), Shh (I) and Six3 (J) transcripts within the tuberal hypothalamus of Ascl1GFPKI; Neurog2GFPKI E12.5 embryos. Data are mean±s.e.m. (n=3 groups of pooled tuberal hypothalami from three embryos). *P<0.01; **P<0.005; ***P<0.0001; ns, not significant. Unpaired two-tailed Student's t-test. 3V, third ventricle. Scale bars: 50 µm.
To assess the relative contributions of Ascl1+ and Neurog2+ progenitors to distinct VMH lineages, we employed two knock-in lines for short-term lineage tracing, each expressing a distinct fluorescent reporter: Ascl1GFPKI (Leung et al., 2007) and Neurog2mCherryKI (C.S., unpublished) mice. The hypothalamus was dissected from E12.5 Neurog2mCherryKI/+ and Ascl1GFPKI/+ embryos and dissociated cells were fluorescence-activated cell sorted to isolate GFP+ (i.e. Ascl1+) and mCherry+ (i.e. Neurog2+) cells (Fig. S2A-C). Total RNA was extracted from the sorted cells and RT-qPCR was used to confirm the enrichment of Ascl1 transcripts in the GFP+ cells (Fig. 3C) and the enrichment of Neurog2 transcripts in mCherry+ cells (Fig. 3D). We then used these sorted cells to assess the transcription of lineage markers specific to VMH progenitors (Rax and Shh), VMH progenitor and pan-postmitotic neurons (Six3), and distinct VMH subdomains: VMHDM (Vgll2), VMHC (Satb2) and VMHVL (Nkx2.1). Satb2 and Rax were both expressed at similar levels in both GFP+- (i.e. Ascl1+) and mCherry+- (i.e. Neurog2+) sorted cells (Fig. 3F,H), suggesting that both Neurog2+ and Ascl1+ progenitors can give rise to VMHC lineages. In contrast, the pan-VMH markers Shh and Six3, and the VMHVL (Nkx2.1) and VMHDM (Vgll2) markers were more highly expressed in Ascl1+ compared with Neurog2+ cells (Fig. 3E,G,I,J). Combined, these results suggest that Ascl1+ and Neurog2+ progenitors give rise to distinct VMH populations, which we addressed further with long-term lineage tracing described below.
Ascl1+ progenitors preferentially give rise to VMHDM,C neurons
Given that discrete VMH neurons appeared to arise from Ascl1-expressing progenitors, we next conducted a long-term lineage trace to identify the subcompartmentalization (if any) of neurons derived from Ascl1+ progenitors within the mature VMH (i.e. VMHDM, VMHC and VMHVL). For long-term lineage tracing, we crossed Ascl1Cre:ERT2 (Kim et al., 2011) mice with a ROSACAG:tdTomato reporter line (Madisen et al., 2010) to label all Ascl1 progeny with tdTomato. Pregnant females were injected with 200 µl of 5 mg/ml 4-hydroxytamoxifen (OHT) at E10.5 and E11.5, and brains were harvested at E19.5. We first stained for Fezf1, a pan-VMH marker (Kurrasch et al., 2007), to ascertain the overall percentage of VMH neurons that arose from Ascl1+ progenitors. We showed that fewer than half of the Fezf1+ cells arose from Ascl1+ precursors (Fezf1+ cells: 383.1±14.01, n=3; Fezf1+/tdTomato+ cells: 153.4±14.66, n=3; Fig. 4A,C); however, these were only the cells exposed to 4-OHT at E10.5 and E11.5; it is possible that if we had injected animals at later embryonic time points we would have observed more Fezf1+ cells arising from an Ascl1+ lineage. In addition, Ascl1+ progeny were located primarily within the VMHDM and VMHC (Fig. 4A-A″). In contrast, Ascl1+ progenitors gave rise to fewer than 25% of Nkx2.1+ cells, which mostly reside within the VMHVL (Nkx2.1+ cells: 202.2±14.95, n=3; Nkx2.1+/tdTomato+ cells: 46.89±9.27, n=3; Fig. 4B-B″,D). These results suggest that Ascl1+ neuronal lineages occupy a large population of VMHDM and VMHC, and to a lesser extent VMHVL subdomains. Thus, despite expressing elevated levels of the VMHVL marker Nkx2.1 at early embryonic stages (E12.5; Fig. 3), Ascl1+ progenitors do not appear to give rise to the Nkx2.1+ VMHVL lineage, as expected.
Ascl1 gives rise to a subpopulation of VMH neurons. (A-B″) Co-expression of anti-Fezf1 (A,A′), anti-Nkx2.1 (B,B′) and anti-tdTomato (A,A″,B,B″) in the Ascl1Cre−ERT2; Rosa-tdTomato background. (C) Cell counts for overall number of cells expressing Fezf1 compared with the proportion of cells co-expressing Fezf1 and tdTomato (n=3). (D) Cell counts for overall number of cells expressing Nkx2.1 compared with the proportion of cells co-expressing Nkx2.1 and tdTomato+. Data are mean±s.e.m. (n=3 embryos per group; three brain sections per embryo). ***P<0.004. Unpaired two-tailed Student's t-test. Scale bar: 50 µm.
Ascl1 gives rise to a subpopulation of VMH neurons. (A-B″) Co-expression of anti-Fezf1 (A,A′), anti-Nkx2.1 (B,B′) and anti-tdTomato (A,A″,B,B″) in the Ascl1Cre−ERT2; Rosa-tdTomato background. (C) Cell counts for overall number of cells expressing Fezf1 compared with the proportion of cells co-expressing Fezf1 and tdTomato (n=3). (D) Cell counts for overall number of cells expressing Nkx2.1 compared with the proportion of cells co-expressing Nkx2.1 and tdTomato+. Data are mean±s.e.m. (n=3 embryos per group; three brain sections per embryo). ***P<0.004. Unpaired two-tailed Student's t-test. Scale bar: 50 µm.
Next, we investigated whether Ascl1 was necessary for early VMH development. Therefore, we collected Ascl1GFPKI/GFPKI embryos, which are Ascl1-null mutants, and their control siblings (Ascl1+/GFP). We harvested brains immediately after peak neurogenesis (E12.5) and quantified the number of VMH neurons generated in the absence of Ascl1. We observed a strong reduction in the number of Fezf1+ (pan-VMH marker) neurons across the VMH rostral-caudal plane at E12.5 (Fig. 5A-C; Ascl1GFPKI/GFPKI: 85.22±12.48, n=3; control: 218.7±4.62, n=3). We also observed a decrease in Nkx2.1+ cells (Nkx2.1 labels both VMH progenitors and VMH newborn neurons at this time point) in Ascl1GFPKI/GFPKI-null brains (127.2±26.02 cells) compared with the control (412±2.39 cells; Fig. 5D-F). These data suggest that Ascl1 is required for the generation of VMH neurons at this early developmental stage.
Ascl1 is required for the specification of VMH neurons. (A,B) Immunostaining for anti-Fezf1 on E12.5 coronal sections in control (A) and Ascl1−/− (B) mice. (C) Fezf1+ cell counts within the mantle zone (MZ) at E12.5 in both control and Ascl1−/− brains. (D,E) Immunostaining for anti-Nkx2.1 on E12.5 mouse coronal sections in control (D) and Ascl1−/− (E). (F) Nkx2.1+ cell counts within VZ and MZ at E12.5 in both control and Ascl1−/− brains. (G,H) Immunostaining results for anti-Fezf1 on E15.5 mouse coronal sections in control (G) and Ascl1−/− (H) hypothalami. (I,J) Immunostaining results for anti-Nkx2.1 on E15.5 mouse coronal sections in control (I) and Ascl1−/− (J) hypothalami. White arrowheads indicate the absence of Fezf1 staining in the VMHdm (H) and the presence of Nkx2.1 staining in the VMHvl in the Ascl1-null background (J). (K,L) ISH results for Vgll2 transcription in E15.5 mouse coronal brain sections in control (K) and Ascl1−/− (L) embryos. The black arrowhead in L indicates the reduction of Vgll2 transcripts in the Ascl1-null background. (M,N) ISH results for Sf1 transcription in E15.5 mouse coronal sections on control (M) and Ascl1−/− (N). (O) Fezf1+ cell counts for whole VMH at E15.5 in both control and Ascl1−/− brains. (P-R) Fezf1+ cell counts in both control and Ascl1−/− E15.5 VMHDM (P), VMHC (Q) and VMHVL (R). (S) Nkx2.1+ cell counts for whole VMH in both control and Ascl1−/− brains. (T-V) Nkx2.1+ cell counts in both control and Ascl1−/− E15.5 VMHDM (T), VMHC (U) and VMHVL (V). Dashed white lines represent approximate VMH subdomain boundaries used for counting cells. Data are mean±s.e.m. (n=3 embryos per group; three brain sections per embryo). *P<0.01, **P<0.001, ***P<0.005, ****P<0.0001. ns, not significant. Unpaired two-tailed Student's t-test. Scale bars: 50 µm (A,B,D,E); 100 µm (G-N).
Ascl1 is required for the specification of VMH neurons. (A,B) Immunostaining for anti-Fezf1 on E12.5 coronal sections in control (A) and Ascl1−/− (B) mice. (C) Fezf1+ cell counts within the mantle zone (MZ) at E12.5 in both control and Ascl1−/− brains. (D,E) Immunostaining for anti-Nkx2.1 on E12.5 mouse coronal sections in control (D) and Ascl1−/− (E). (F) Nkx2.1+ cell counts within VZ and MZ at E12.5 in both control and Ascl1−/− brains. (G,H) Immunostaining results for anti-Fezf1 on E15.5 mouse coronal sections in control (G) and Ascl1−/− (H) hypothalami. (I,J) Immunostaining results for anti-Nkx2.1 on E15.5 mouse coronal sections in control (I) and Ascl1−/− (J) hypothalami. White arrowheads indicate the absence of Fezf1 staining in the VMHdm (H) and the presence of Nkx2.1 staining in the VMHvl in the Ascl1-null background (J). (K,L) ISH results for Vgll2 transcription in E15.5 mouse coronal brain sections in control (K) and Ascl1−/− (L) embryos. The black arrowhead in L indicates the reduction of Vgll2 transcripts in the Ascl1-null background. (M,N) ISH results for Sf1 transcription in E15.5 mouse coronal sections on control (M) and Ascl1−/− (N). (O) Fezf1+ cell counts for whole VMH at E15.5 in both control and Ascl1−/− brains. (P-R) Fezf1+ cell counts in both control and Ascl1−/− E15.5 VMHDM (P), VMHC (Q) and VMHVL (R). (S) Nkx2.1+ cell counts for whole VMH in both control and Ascl1−/− brains. (T-V) Nkx2.1+ cell counts in both control and Ascl1−/− E15.5 VMHDM (T), VMHC (U) and VMHVL (V). Dashed white lines represent approximate VMH subdomain boundaries used for counting cells. Data are mean±s.e.m. (n=3 embryos per group; three brain sections per embryo). *P<0.01, **P<0.001, ***P<0.005, ****P<0.0001. ns, not significant. Unpaired two-tailed Student's t-test. Scale bars: 50 µm (A,B,D,E); 100 µm (G-N).
Given that our long-term lineage tracing results implied that Ascl1 might be required for the generation of VMHDM and VMHC but not VMHVL neurons, we next assayed whether VMHDM/C neuronal markers were specifically lost in Ascl1-null hypothalami at E15.5, a time point when the VMH subdomains become more organized. We chose pan-VMH, Fezf1 (Fig. S3A,B), the VMHDM-C marker Sf1 (Fig. S3C,D; Cheung et al., 2013) and subdomain-specific markers (Kurrasch et al., 2007), including Sox14 and Vgll2 (VMHDM; Fig. S3E-H), Satb2 (VMHC; Fig. S3I,J), and Nkx2.1 (VMHVL; Fig. S3K,L), and conducted ISH to investigate their expression profiles at E15.5. We labeled both control and Ascl1-null sections with the pan-VMH marker Fezf1 and observed an overall reduction in the number of Fezf1+ cells (control: 544±2.18 cells, n=3; Ascl1GFPKI/GFPKI: 226.2±13.66 cells, n=3; Fig. 5G,H,O). When we divided the VMH into the three subdomains (Fig. 5G,I), defined by the expression boundaries of region-specific markers (Fig. S3), we observed a spatially restricted loss of cells in the VMHDM and VMHC subcompartments (Fig. 5P,Q), and no change in VMHVL (Fig. 5R; control: VMHDM=160.3±1.54; VMHC=229.1±4; VMHVL=148.2±8.86, n=3; Ascl1GFPKI/GFPKI: VMHDM=16.11±1.55; VMHC=107.4±1.49; VMHVL=102.7±15.77, n=3.). As Fezf1 expression was not significantly reduced in the VMHVL, we next quantified the effect of Ascl1 elimination on the VMHVL marker Nkx2.1. We labeled wild-type and Ascl1-null E15.5 sections with Nkx2.1 (Fig. 5I,J) and observed a decrease in the overall number of Nkx2.1+ cells in Ascl1-null VMH compared with the control (control: 295±16.34 cells, n=3; Ascl1GFPKI/GFPKI: 183.1±21.16 cells, n=3; Fig. 5S). Furthermore, we quantified a distinct loss of Nkx2.1+ cells in the VMHVL and not in the VMHDM or VMHC (control: VMHDM=9.22±1.74; VMHC=110.3±9.26; VMHVL=175.4±10.73, n=3; Ascl1GFPKI/GFPKI: VMHDM=12.44±0.72; VMHC=75.78±11.64; VMHVL=94.89±9.72, n=3; Fig. 5T-V). These data suggest that Ascl1 is required for specification of VMH neurons throughout the nucleus, with a more prominent role in VMHDM/C neuronal differentiation.
To further explore whether Ascl1 was necessary to specify neurons within discrete VMH subdomains, we assayed the expression of subregion-specific genes. To start, we employed a Vgll2 riboprobe, which specifically marks the VMHDM (Fig. S3D), and observed that, in the absence of Ascl1, Vgll2 expression was largely reduced in the VMHDM (Fig. 5K,L). In addition, Sf1 starts to downregulate in VMHVL neurons at this time point (thereby labeling VMHDM,C), and we observed a severe reduction in Sf1 transcripts throughout both the VMHDM and VMHC in Ascl1-null brains compared with controls (Fig. 5M,N). Combined, these data suggest that Ascl1 is required to give rise to neurons that largely (but not exclusively) reside in VMHDM and VMHC subcompartments.
VMH neurons undergo a cell fate change from VMH-glutamatergic to VMH-GABAergic in the absence of Ascl1
There are two possible explanations (not mutually exclusive) to explain the loss of VMH neurons in Ascl1-null brains: (1) VMH neurons die; and (2) Ascl1+ progenitors switch programs and give rise to different cell fates, thus turning off VMH-specific genes. First, we tested whether VMH neurons undergo apoptosis in the absence of Ascl1. We found no increase in the apoptotic marker cleaved caspase 3 at either E12.5 or E15.5 in Ascl1-null and control brains (Fig. S4A-D). Second, we investigated whether a change in the density of neurons in the Ascl1-null VMH occurred and showed that there were no gross differences in NeuN+ staining, a pan-neuronal marker, between control and Ascl1-null E15.5 brains (Fig. S4E,F). Given that there is no global loss of neurons in the VMH in Ascl1-null embryos, we checked whether the neurons that were generated had acquired a new identity.
Embryonic VMH neurons adopt a glutamatergic fate, whereas neurons that surround the VMH are GABAergic (Cheung et al., 2015). To determine whether a loss of VMH glutamatergic neuronal identity occurred, we conducted ISH and immunohistochemistry for VGlut2, a marker of glutamatergic neurons, on control and Ascl1-null brains at E15.5. Our results revealed a striking loss of VGlut2 transcripts and protein within the VMH in Ascl1-null compared with control brains (Fig. 6A-B″; Fig. S5A-B″). Complementary to this, we employed a Gad67 riboprobe and antibody, which labels GABAergic neurons. In control brains at E15.5, we observed an exclusion of Gad67 transcripts and protein from the VMH (Fig. 6C-C″; Fig. S5C-C″). In striking contrast, in Ascl1 knockouts we detected Gad67 transcripts and protein within the VMH core, particularly in the VMHDM and part of the VMHC (Fig. 6D-D″; Fig. S5D-D″), consistent with a fate change from VMH-glutamatergic to VMH-GABAergic in the absence of Ascl1. To further confirm that VMH neurons underwent a fate change in the absence of Ascl1, we immunostained for the N-methyl-D-aspartate (NMDA) receptor, a marker of glutamatergic neurons. We showed a specific loss of NMDA receptors in Ascl1-null animals, especially within the VMHDM and VMHC (Fig. 6E-F″). Interestingly, NMDA receptor expression was maintained in the VMHVL (Fig. 6E-F″), suggesting that the loss of VMH-specific gene expression in the VMHDM/C in Ascl1 knockout brains occurs alongside an identity change to a GABAergic fate.
Cell fate change from VMH-glut+ to VMH-GABA+ in the absence of Ascl1. (A-B″) ISH results for the VGlut2 riboprobe in coronal mouse brain sections through the rostral to caudal plane at E15.5 in control (A-A″) and Ascl1−/− (B-B″) embryos. Red arrowheads denote the presence or absence of Vglut2 staining in the VMH in control (A-A″) and Ascl1-null (B-B″) backgrounds, respectively. (C-D″) ISH results for the GAD67 riboprobe in coronal mouse brain sections through the rostral to caudal plane at E15.5 in control (C-C″) and Ascl1−/− (D-D″) embryos. Red arrowheads highlight the presence or absence of Gad67 staining in the VMH in control (C-C″) and Ascl1-null (D-D″) backgrounds, respectively. The black-dashed lines outline the third ventricle and pial surface. (E-F″) Immunostaining results for the anti-NMDA receptor in coronal mouse brain sections through rostral to caudal plane at E15.5 in control (E-E″) and Ascl1−/− (F-F″) embryos. White arrowheads indicate the absence of NMDA receptor expression in VMHdm/c and yellow arrows denote the maintained expression of NMDA receptors in the VMHvl in Ascl1-null hypothalamus. The white-dashed lines represent the boundary of the brain. (G-I) RT-qPCR results for Ascl1 (G) Dlx1 (H) and Dlx2 (I) in both control and Ascl1-null backgrounds at E12.5. Data are mean±s.e.m. (n=3 groups of pools of three embryos tuberal hypothalamus). *P<0.026, **P<0.003. Unpaired two-tailed Student's t-test. Scale bars: 100 µm (A-D″); 50 µm (E-F″).
Cell fate change from VMH-glut+ to VMH-GABA+ in the absence of Ascl1. (A-B″) ISH results for the VGlut2 riboprobe in coronal mouse brain sections through the rostral to caudal plane at E15.5 in control (A-A″) and Ascl1−/− (B-B″) embryos. Red arrowheads denote the presence or absence of Vglut2 staining in the VMH in control (A-A″) and Ascl1-null (B-B″) backgrounds, respectively. (C-D″) ISH results for the GAD67 riboprobe in coronal mouse brain sections through the rostral to caudal plane at E15.5 in control (C-C″) and Ascl1−/− (D-D″) embryos. Red arrowheads highlight the presence or absence of Gad67 staining in the VMH in control (C-C″) and Ascl1-null (D-D″) backgrounds, respectively. The black-dashed lines outline the third ventricle and pial surface. (E-F″) Immunostaining results for the anti-NMDA receptor in coronal mouse brain sections through rostral to caudal plane at E15.5 in control (E-E″) and Ascl1−/− (F-F″) embryos. White arrowheads indicate the absence of NMDA receptor expression in VMHdm/c and yellow arrows denote the maintained expression of NMDA receptors in the VMHvl in Ascl1-null hypothalamus. The white-dashed lines represent the boundary of the brain. (G-I) RT-qPCR results for Ascl1 (G) Dlx1 (H) and Dlx2 (I) in both control and Ascl1-null backgrounds at E12.5. Data are mean±s.e.m. (n=3 groups of pools of three embryos tuberal hypothalamus). *P<0.026, **P<0.003. Unpaired two-tailed Student's t-test. Scale bars: 100 µm (A-D″); 50 µm (E-F″).
In the telencephalon, Ascl1 acts upstream of Dlx1 and Dlx2 to induce a GABAergic cell fate (Long et al., 2009; Poitras et al., 2007; Fode et al., 2000) and in the thalamus, Ascl1 downregulates Dlx1/Dlx2 (Song et al., 2015). Thus, we hypothesized that Dlx1 and Dlx2 might be mediating the glutamatergic to GABAergic fate change in the Ascl1 mutant VMH, and would hence be upregulated. To test this notion, we extracted RNA from the tuberal hypothalamus in both control and Ascl1-null embryos at E12.5. Using RT-qPCR, we revealed a significant increase in the transcription of both Dlx1 and Dlx2 in the absence of Ascl1 compared with control embryos (Fig. 6G-I). Mechanistically, these results suggest that Ascl1 functions to inhibit a GABAergic cell fate within the embryonic VMH via repressing Dlx1 and/or Dlx2, such that in the absence of Ascl1 these genes can mediate a GABAergic fate.
Neurog3 is required to specify a glutamatergic cell fate
Previous studies proposed that Neurog3 might be a downstream effector of Ascl1 within the tuberal hypothalamus (McNay et al., 2006; Pelling et al., 2011), which is supported by our finding that Neurog3 transcription is downregulated in the absence of Ascl1 (Fig. 2E,F). Considering the role of other neurogenin family members (e.g. Neurog1 and Neurog2) in specifying neurotransmitter cell fates in the neocortex (Dixit et al., 2014; Dennis et al., 2017; Fode et al., 2000; Schuurmans et al., 2004), we investigated whether Neurog3 might have a similar role in the hypothalamus. For this purpose, we examined Neurog3−/− embryos at E15.5 for neuronal defects in the developing VMH (Gradwohl et al., 2000). First, we observed a significant reduction in the number of Fezf1+ cells in Neurog3−/− compared with control VMH domains (Fig. 7A-C), phenocopying the decrease in Fezf1+ cells quantified in the Ascl1−/− VMH. In addition, in the absence of Neurog3, a VMH-specific glutamatergic to GABAergic cell fate change was observed (Fig. 7D,E,G,H), as seen in the Ascl1−/− VMH (Fig. 5G,H,O). As VGlut2 localization is cytoplasmic and makes cell counting unreliable, we instead quantified the area of VGlut2 staining and demonstrated a significant reduction in Neurog3−/− compared with control hypothalami (Fig. 7F; Fig. S6A,B). In addition, the linear plot (line was drawn from dorsomedial VMH to ventrolateral VMH) for VGlut2 signal showed an obvious shift in the distribution of VGlut2, such that, in control brains, VGlut2 expression was uniform across the entire VMH region, whereas in the Neurog3 brains, VGlut2 was excluded from most of the VMH nuclear region and was present only in the VMHVL region and the pial surface outside of the VMH zone (Fig. S6C,D). In contrast, the total area for the GABAergic marker GAD67 was significantly increased in Neurog3−/− compared with the control VMH domains (Fig. 7I, Fig. S6E,F), and the GAD67 signal plot was instead enriched within the VMH nucleus in Neurog3−/− VMH domains compared with the control (Fig. S6G,H). Together, these results suggest that Neurog3 is required to induce a glutamatergic cell fate and repress a GABAergic phenotype in the developing VMH.
Cell fate change from VMH-glut+ to VMH-GABA+ in the absence of Neurog3. (A,B) Immunostaining results for anti-Fezf1 in control (A) and Neurog3−/− (B) backgrounds. (C) Cell count for Fezf1+ cells in both control and Neurog3−/− backgrounds. (D,E) Immunostaining results for anti-VGlut2 in control (D) and Neurog3−/− (E) backgrounds. (F) Total area of VGlut2 staining in the control background compared with the Neurog3−/− background. (G,H) Immunostaining results for anti-GAD67 in control (G) and Neurog3−/− (H) backgrounds. (I) Graph on total area of GAD67 staining in control compared with Neurog3−/−. Data are mean±s.e.m. (n=3 embryos per group; three brain sections per embryo). *P=0.01, ***P=0.0003, ****P<0.0001. Unpaired two-tailed Student's t-test. White-dashed lines outline the third ventricle and pial surface. White arrows denote Vglut2 or Gad67 expression in the VMH in control (D,G) and Neurog3-null (E,H) backgrounds. Scale bars: 100 µm.
Cell fate change from VMH-glut+ to VMH-GABA+ in the absence of Neurog3. (A,B) Immunostaining results for anti-Fezf1 in control (A) and Neurog3−/− (B) backgrounds. (C) Cell count for Fezf1+ cells in both control and Neurog3−/− backgrounds. (D,E) Immunostaining results for anti-VGlut2 in control (D) and Neurog3−/− (E) backgrounds. (F) Total area of VGlut2 staining in the control background compared with the Neurog3−/− background. (G,H) Immunostaining results for anti-GAD67 in control (G) and Neurog3−/− (H) backgrounds. (I) Graph on total area of GAD67 staining in control compared with Neurog3−/−. Data are mean±s.e.m. (n=3 embryos per group; three brain sections per embryo). *P=0.01, ***P=0.0003, ****P<0.0001. Unpaired two-tailed Student's t-test. White-dashed lines outline the third ventricle and pial surface. White arrows denote Vglut2 or Gad67 expression in the VMH in control (D,G) and Neurog3-null (E,H) backgrounds. Scale bars: 100 µm.
Ascl1 and Neurog3 are not sufficient to give rise to VMH neurons
Given that Ascl1 seemed to be required for VMH neuronal specification, we next wanted to test whether Ascl1 was sufficient to induce a VMH fate. As Ascl1 was expressed as early as E10.5 and throughout the entire hypothalamic ventricular zone (Fig. 1A), it was not feasible to ectopically express Ascl1 at a developmental time point or in a hypothalamic domain that was void of Ascl1. Instead, we misexpressed Ascl1 outside the VMH competency window and assayed whether Ascl1 was sufficient to convert later progenitors to an earlier temporal VMH fate. Peak neurogenesis in the VMH occurs at ∼E11.5 (Aslanpour et al., 2020; Altman and Bayer, 1978), so we conducted IUE on wild-type animals at E12.5, injecting them with pCIG2-Ascl1 or pCIG2 control constructs (Dixit et al., 2014) into the third ventricle (n=3). All brains were harvested 2 days later at E14.5 and immunostained for co-expression of GFP and known VMH markers, including Fezf1, SF1, Sox14 and Nkx2.1 (Fig. 8A-D‴; Fig. S7A-D‴). Our results showed that Ascl1-GFP+ cells were largely excluded from the VMH proper, instead appearing to label at least some radial glia that resided dorsally and laterally to the VMH (Fig. 8A-D‴; Fig. S7A-D‴, arrowheads). Notably, there was a lack of co-labeled VMH markers and GFP within the VMH of embryos misexpressing Ascl1 (Fig. 8A-D‴; Fig. S7A-D‴). Electroporation of a pCIG2 control vector had no effect on the developing VMH (Fig. S7E-H‴). These data suggest that Ascl1 is not sufficient to induce a VMH-specific cell fate in the developing VMH.
Ascl1 is not sufficient to induce a VMH cell identity. (A-B‴) Immunostaining results for VMH-specific markers in the C57BL/6 background injected with the PCIG2-Ascl1 construct at E12.5 and brains harvested at E14.5. (A-A‴) Immunostaining results for anti-Fezf1 (A,A′), anti-GFP (A,A″) and Hoechst (A,A‴). (B-B‴) Immunostaining results for anti-SF1 (B,B′), anti-GFP (B,B″) and Hoechst (B,B‴). (C-C‴) Immunostaining results for anti-Sox14 (C,C′), anti-GFP (C,C″) and Hoechst (C,C‴). (D-D‴) Immunostaining results for anti-Nkx2.1 (D,D′), anti-GFP (D,D″) and Hoechst (D,D‴). n=3 embryos per group. White-dashed lines outline the third ventricle and pial surface. The white arrowheads in A″,B″,C″,D″ indicate the partial absence of GFP expression within the VMH nucleus. Scale bars: 100 µm.
Ascl1 is not sufficient to induce a VMH cell identity. (A-B‴) Immunostaining results for VMH-specific markers in the C57BL/6 background injected with the PCIG2-Ascl1 construct at E12.5 and brains harvested at E14.5. (A-A‴) Immunostaining results for anti-Fezf1 (A,A′), anti-GFP (A,A″) and Hoechst (A,A‴). (B-B‴) Immunostaining results for anti-SF1 (B,B′), anti-GFP (B,B″) and Hoechst (B,B‴). (C-C‴) Immunostaining results for anti-Sox14 (C,C′), anti-GFP (C,C″) and Hoechst (C,C‴). (D-D‴) Immunostaining results for anti-Nkx2.1 (D,D′), anti-GFP (D,D″) and Hoechst (D,D‴). n=3 embryos per group. White-dashed lines outline the third ventricle and pial surface. The white arrowheads in A″,B″,C″,D″ indicate the partial absence of GFP expression within the VMH nucleus. Scale bars: 100 µm.
Next, to examine whether Neurog3 is sufficient to give rise to Ascl1 transcription and/or VMH neuronal identities, we employed IUE and injected the pNAS-HA-Ngn3 construct into E12.5 wild-type embryos. Embryos were collected at E14.5 and ISH for Ascl1 transcripts, and immunostaining for Fezf1, VGlut2 and GAD67 proteins, was conducted. Overexpression of Neurog3 at E12.5 was not sufficient to give rise to Ascl1 transcription (Fig. S8A,B). Moreover, immunostaining showed an exclusion of Neurog3-GFP+ cells from the VMH and RFP+ cells were not co-labeled with Fezf1 (Fig. S8E,K). No change in expression of VGlut2 or GAD67 on the injected side compared with the non-injected side or controls was observed (Fig. S8C,D,I,J). Moreover, to test whether co-overexpression of Ascl1 and Neurog3 can induce a VMH cell fate outside of the window of the VMH competency, we injected E12.5 wild-type animals with both pCIG2-Ascl1 and pNAS-HA-Ngn3 constructs, collected embryos at E14.5 and immunostained for Fezf1, VGlut2 and GAD67. Co-overexpression of Ascl1 and Neurog3 was not sufficient to give rise to a VMH neuronal identity outside of the VMH window of competency (Fig. S8F-K).
Finally, to verify the efficiency of our construct and our approach, we injected the same Ascl1-GFP and control plasmids into the lateral ventricle of the cortex. We observed an increase in the number of Ctip2+ cells co-expressing GFP within embryos injected with pCIG2-Ascl1 IUE compared with embryos injected with a control construct (Fig. S9A-B‴, arrowheads), consistent with previous results (Dennis et al., 2017).
DISCUSSION
In this study, we demonstrated a role for Ascl1 in the development of the VMH. We determined that Ascl1 was expressed within VMH progenitors across different embryonic stages and that Ascl1+ progenitors gave rise to discrete populations of VMH neurons. Using Ascl1-null brains we showed subdomain-specific deficits, notably within the VMHDM and dorsal part of VMHC, uncovering a role for Ascl1 in the specification of neurons that reside in discrete VMH subcompartments. This intriguing result suggests that transcription factor programs in hypothalamic progenitors might underlie the suborganization of its nuclear structures. Furthermore, we revealed that Ascl1 is required for VMH neurons to acquire a glutamatergic cell fate, having observed a cell fate change from VMH-glutamatergic to VMH-GABAergic in the absence of Ascl1. Although previous reports showed that Neurog3 acts downstream of Ascl1 in the tuberal hypothalamus (McNay et al., 2006), the functional role of that interaction was unknown. We found a positive relationship between Ascl1 and Neurog3 in specifying a glutamatergic fate that has not been reported previously. Combined, we conclude that the Ascl1-Neurog3 transcriptional axis plays an important role in specifying VMH neuronal fate and identities, which are the first transcription factors revealed to drive VMH subcompartment-specific programs.
Ascl1 is the first determinant identified that is required to specify VMH subtype-specific identities
Ascl1 plays crucial roles in cell fate decisions and neuronal specification, and differentiation within other brain regions, such as the cerebrum, cerebellum and retina (Akagi et al., 2004; Bertrand et al., 2002; Brzezinski et al., 2011; Dixit et al., 2011b; Florio et al., 2012; Kim et al., 2008; Kovach et al., 2013; Leung et al., 2007; Ma and Wang, 2006). Previous studies in other brain regions suggested that Ascl1 and Neurog2 have distinct temporal and spatial expression patterns. For example, Neurog2 is expressed specifically in the dorsal telencephalon (Britz et al., 2006; Dixit et al., 2011b; Kovach et al., 2013; Parras et al., 2002; Fode et al., 2000; Wilkinson et al., 2013; Mattar et al., 2004), whereas Ascl1 is expressed in the ventral telencephalon (Berninger et al., 2007; Casarosa et al., 1999; Geoffroy et al., 2009; Schuurmans et al., 2004; Wilkinson et al., 2013). However, more recently it has begun to be appreciated that these distinct lineage determinants can in some instances be co-expressed in the same progenitors, including in the cerebral cortex (Britz et al., 2006), cerebellum (Chouchane and Costa, 2019; Zordan et al., 2008) and retina (Hufnagel et al., 2010).
Our study provides another example in which Neurog2 and Ascl1 are expressed in a shared progenitor zone, including in the same progenitors in some instances. However, despite this overlap, we found that Ascl1+ and Neurog2+ progenitors have different biases in their differentiation potential. The three recognizable subdomains of the VMH (Kurrasch et al., 2007; Millhouse, 1973) are populated by neurons with distinct phenotypes expressing different transcription factors and genes (Kurrasch et al., 2007). Using short-term lineage tracing at early embryonic stages (E12.5), Ascl1+ progenitors were shown to give rise to cells in the VMHDM, VMHC and VMHVL lineages, whereas Neurog2+ lineages did not express VMHDM neuronal markers. In contrast to our short-term tracing, long-term lineage tracing suggested that, at later stages (E19.5), cells in the more mature VMHVL are not derived from Ascl1+ progenitors. Accordingly, loss-of-function studies showed that Ascl1 is necessary for the proper formation of VMHDM,C but appears to play a minor role in the formation of VMHVL. There is, however, a reduction in the total number of Nkx2.1+ neurons (e.g. VMHVL) in E15.5 Ascl1-null mutants, showing that Ascl1 does play a role in the generation of these neurons. One possible reason why our Ascl1 long-term lineage trace failed to label VMHVL neurons is perhaps that the neurons derived from Ascl1+ progenitors are lost (either through death or cell migration) sometime between E15.5 and E19.5. Additionally, the timing of 4-OHT injections used to activate the Cre could also account for this absence of co-labeling. It is possible that the addition of more 4-OHT injection time points could capture additional Nkx2.1+ neurons co-labeled with GFP (e.g. Ascl1).
In addition, we employed Neurog2tm1(icre/ERT2*) (Florio et al., 2012) to study the lineage of Neurog2+ progenitors within the VMH but observed very poor recombination in the hypothalamus. Indeed, despite strong tdTomato+ labeling in the cortex of Neurog2tm1(icre/ERT2*);Rosa:tdTomato animals (4-OHT injected at E10.5 and E11.5; Fig. S10), we observed a complete absence of any tdTomato+ cells in the arcuate nucleus or the VMH, and only a few positive cells were detected in the DMH (Fig. S10). This is despite strong Neurog2 expression in the tuberal hypothalamic progenitors that give rise to these nuclei (Fig. 2). We tested different 4-OHT concentrations and injection times (E10.5–E13.5) but never observed strong Cre activity in the VMH. Thus, it was not possible to conduct a long-term lineage trace for Neurog2+ progenitors using this Cre line, which suggests that this knock-in line is not appropriate for studies in the hypothalamus.
Previous findings (Aslanpour et al., 2020; Shimada and Nakamura, 1973) reveal a lateral-to-medial neurogenic pattern to neuronal birth within the tuberal hypothalamus. In fact, VMH neurons born early (e.g. E10.5) migrate to the VMHVL and ventral part of VMHC, whereas neurons born later (e.g. E12.5-E13.5) occupy positions in the VMHDM and dorsal part of VMHC (Shimada and Nakamura, 1973; Marsters et al., 2016). Given that Ascl1 expression started around E10.5 and reached its peak at E12.5, and that VMH neurogenesis also initiates at E10.5 and peaks at E12.5 (Marsters et al., 2016), it is possible that the distinct role of Ascl1 in the development of VMHDM/C neurons, and not VMHVL neurons, is due to the timing of neurogenesis. In other words, Ascl1 transcription (e.g. E11.5-E13.5) occurs at the peak- and back-end of neurogenesis, suggesting that Ascl1+ lineages are born later and thus occupy the VMHDM and dorsal part of the VMHC.
One of the most important limiting factors in this study was the complexity of the VMH as a nuclear structure. Although previous studies show that the VMH comprises three recognizable subdomains (Van Houten and Brawer, 1978; Kurrasch et al., 2007; McClellan et al., 2006), there is a limited number of markers that differentially label distinct neuronal populations located within the VMH nucleus, thereby making it challenging to further identify the phenotypes of the affected cells in the Ascl1 loss-of-function model. Previous gene profiling studies in neonatal (P0) mice brains identified approximately 200 genes that are highly enriched within the VMH nucleus (Kurrasch et al., 2007); however, as our study was focused on the embryonic development of the VMH, the expression patterns of various subdomain-specific markers, such as Sox14, Nptx2, NPY, Erα and Satb2, were not as defined in the embryonic stages prior to when the subcompartments coalesce, as compared with the neonatal brain. Also, owing to the early postnatal lethality of Ascl1 knockout animals, we were not able to accurately define the neuronal populations potentially affected in the absence of Ascl1, or the lasting consequences of loss of Ascl1 during development, by studying behavioral changes. However, we were able to examine the expression of some VMH subdomain-specific markers, and ongoing studies seek to determine the cellular phenotype of the neurons affected in the absence of Ascl1 in the embryonic hypothalamus.
Ascl1 specifies a glutamatergic identity in the hypothalamus
A second key finding of our study was the dissection of the transcriptional relationship between Ascl1 and proneural genes in the neurogenin family in the VMH. The transcriptional interactions between these genes has been best studied in the cortex, in which Neurog1 and Neurog 2 repress Ascl1 transcription (Schuurmans et al., 2004; Fode et al., 2000), but a reciprocal relationship has not been reported. Here, we show that in the VMH, the absence of Ascl1 similarly does not affect transcription of Neurog1 and Neurog2. Instead, we confirmed a downstream regulatory effect of Ascl1 on Neurog3, with Neurog3 expression lost in the VMH progenitor pool in the absence of Ascl1, as was reported previously (McNay et al., 2006; Pelling et al., 2011).
Ascl1 is responsible for promoting a GABAergic cell fate within the ventral forebrain (Britz et al., 2006; Casarosa et al., 1999; Wilkinson et al., 2013), olfactory bulb (Kim et al., 2008; Long et al., 2007), cerebellum (Kim et al., 2008; Sudarov et al., 2011) and spinal cord (Mizuguchi et al., 2006; Li et al., 2005). In this study, we showed that in the VMH, Ascl1 is instead required to specify a glutamatergic cell fate, with the loss of Ascl1 resulting in the ectopic generation of GABAergic neurons within the VMH. In fact, Ascl1 can promote a glutamatergic fate in other brain regions as well, such as in the spinal cord (Helms et al., 2005) and hippocampus (Pleasure et al., 2000), and even during an early temporal window in the cortex (Dixit et al., 2011b). In the spinal cord specifically, Ascl1 is required to give rise to dl3 and dl5 excitatory neurons, and to inhibit a dl4 inhibitory neuronal fate (Helms et al., 2005). Here, we propose that Ascl1 acts via Neurog3 to specify the glutamatergic phenotype of VMH neurons, while repressing Dlx1/2 to prevent the acquisition of a GABAergic fate. Our loss-of-function studies for Neurog3 revealed similar cell fate change (VMH-glutamatergic to VMH-GABAergic) to that observed in the absence of Ascl1. These results support our hypothesis that Ascl1 drives VMH cell fate decisions by regulating Neurog3. Determining how the loss of Ascl1 and the concomitant downregulation of Neurog3 directly leads to the upregulation of Dlx1/2 to drive a GABAergic lineage [as occurs in other forebrain regions (Huang et al., 2014; Poitras et al., 2007)] will require further investigation. Regardless, the pronounced upregulation of Dlx1/2 in Ascl1-null brains was surprising given that Ascl1 directly and positively regulates Dlx1/2 expression in the forebrain (Poitras et al., 2007; Fode et al., 2000; Casarosa et al., 1999; Horton et al., 1999). Moreover, the ectopic expression of Ascl1 results in the upregulation of Dlx1/2 in neocortical neurons (Fode et al., 2000). Thus, how Ascl1 inhibits Dlx1/Dlx2 expression to repress a GABAergic cell fate in the VMH, thus enabling the acquisition of a glutamatergic fate, remains to be fully explored.
In conclusion, we found that Ascl1 is necessary but not sufficient for proper VMH development, and that Ascl1 preferentially gives rise to neurons located within VMHDM,C subdomains. Moreover, we demonstrated that Ascl1 is required for VMH-glutamatergic cell fate during mouse embryonic development, acting in part via its transcriptional regulation of Neurog3. When combined, these studies begin to shed light onto how subdomains and neurotransmitter phenotypes are specified within a developing nuclear structure in the brain.
MATERIALS AND METHODS
Animals and genotyping
Wild-type C57BL/6J and CD1, as well as Ascl1tm1.1(cre/ERT2) (Kim et al., 2011), B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J (Madisen et al., 2010), Ascl1GFPKI (Leung et al., 2007), Neurog3+/− (Gradwohl et al., 2000) and Neurog2IRESmCherryKI, in which IRES::mCherry is knocked-in after the Neurog2 stop codon (C.S., unpublished), were bred to obtain staged embryos. All animals were kept as heterozygotes except for B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J mice, which were maintained as homozygotes. To confirm mating, vaginal plugs were checked each morning and plug dates were considered as E0.5. To confirm genotype, PCR was applied to ear-notched samples using the Ascl1tm1.1(cre/ERT2) primers as follows: mutant forward, 5′-AACTTTCCTCCGGGGCTCGTTTC-3′; mutant reverse, 5′-CGCCTGGCGATCCCTGAACATG-3′; wild-type forward, 5′-TCCAACGACTTGAACTCTATGG-3′; and wild-type reverse, 5′-CCAGGACTCAATACGCAGGG-3′. RosatdTomato primers used were: mutant forward, 5′-GGCATTAAAGCAGCGTATCC-3′; mutant reverse, 5′-CTGTTCCTGTACGGCATGG-3′; wild-type forward, 5′-AAGGGAGCTGCAGTGGAGTA-3′; and wild-type reverse, 5′-CCGAAAATCTGTGGGAAGTC-3′. Ascl1GFPKI primers used were: mutant forward, 5′-AACTTTCCTCCGGGGCTCGTTTC-3′; mutant reverse, 5′-TGGCTGTTGTAGTTGTACTCCAGC-3′; wild-type forward, 5′-TCCAACGACTTGAACTCTATGG-3′; wild-type reverse, 5′-CCAGGACTCAATACGCAGGG-3′. Neurog3+/− primers used were: mutant forward, 5′-GCAGCGCATCGCCTTCTATC-3′; mutant reverse, 5′-CGGCAGATTTGAATGAGGGC-3′; wild-type forward, 5′-TCTCGCCTCTTCTGGCTTTC-3′; and wild-type reverse, 5′-CGGCAGATTTGAATGAGGGC-3′. Neurog2IRESmCherryKI primers used were: mutant forward, 5′-ACAAACAACGTCTGTCGCGACCCT-3′; and mutant reverse, 5′-CACCTTGAAGCGCATGAACTCTT-3′. Animal protocols were approved by the University of Calgary Animal Care Committee and follow the Guidelines of the Canadian Council of Animal Care.
Fluorescence-activated cell sorting
Hypothalamic tissue was dissected from E12.5 Ascl1GFPKI (Leung et al., 2007) and Neurog2IRESmCherryKI heterozygous embryos and cells were dissociated. Fluorescence-activated cell sorting for GFP+ and mCherry+ cells was performed using a BD Influx Cell Sorter (BD Biosciences).
RNA extraction and RTqPCR
E12.5 hypothalamus was dissected from total brain and RNA was extracted using a RNeasy Mini prep kit (Qiagen). Tissue was collected from wild-type and Ascl1GFPKI mutant embryos, as well as from Ascl1GFPKI and Neurog2IRESmCherryKI heterozygous embryos. We collected tissue from three embryos from each genotype (three biological replicates). RNA was reverse transcribed using the RT2 First Strand synthesis kit (Qiagen, 330001) and resultant cDNA was then amplified using the RT2 Primer Assay kit and SYBR Green (both from Qiagen). Primers used for gene expression analysis included: Gapdh (PPM02946E), B2m (PPM03562A) and Hrpt (PPM03559F) as housekeeping genes, as well as Nkx2.1 (PPM05535B), Nr5a1 (PPM27878C), Rax (PPM36905A), Satb2 (PPM31592A), Shh (PPM04516C), Six3 (PPM40710A), Vgll2 (PPM30231A), Fezf1 (PPM29520A), Ascl1 (PPM31367F), Dlx1 (PPM25127B) and Dlx2 (PPM04442A) as VMH markers. The Delta-Delta Ct method was used to calculate relative expression levels for each gene, with three housekeeping genes for normalization (Gapdh, B2M and Hrpt).
Tissue preparation
Embryos were collected after euthanizing the pregnant female via cervical dislocation. Whole heads were harvested at E10.5 and E12.5, whereas for later stages (E14.5-P0) brains were dissected out of the embryonic skull. Tissue fixation and preparation were conducted as described previously (Marsters et al., 2016; Rosin and Kurrasch, 2018). For lineage tracing, pregnant dams were intraperitoneally injected with 200 µl of 5 mg/ml 4-OHT once per day at E10.5 and E11.5, and resulting embryos were collected at E19.5.
RNA in situ hybridization
Brains were fixed in 4% paraformaldehyde, cryoprotected in 20% sucrose and embedded in Tissue-Tek OCT compound (Sakura Finetek, VWR). Brains were cryosectioned coronally (10 µm) and collected on Superfrost Plus slides (Fisher Scientific). RNA ISH was performed as described by Kurrasch et al. (2007). The riboprobes used in this study have been described previously: Neurog1 (Blader et al., 2004), Neurog2 (Gradwohl et al., 1996), Neurog3 (Gradwohl et al., 2000), Ascl1 (Cau et al., 1997), Vgll2 (Kurrasch et al., 2007), Satb2 (Kurrasch et al., 2007), Gad67 (Fode et al., 2000) and Vglut2 (Schuurmans et al., 2004).
Immunostaining
Cryosectioned (10 μm) brain samples were exposed to a primary antibody for 1 h at room temperature in 10% normal goat serum/PBS with 0.1% Tween-20 or Triton-X 100 followed by PBS washes and then incubation for 1 h at room temperature with the appropriate fluorescently conjugated secondary antibody. Primary antibodies employed were as follows: rabbit anti-Fezf1 (Fitzgerald; 1:100, 70R-7693); mouse anti-TTF-1 (also known as Nkx2.1; Millipore; 1:500, ab1333737); chicken anti-GFP (Abcam; 1:500, ab13970); mouse anti-Mash1 (also known as Ascl1; BD Biosciences; 1:500, 556604), mouse anti-NeuN (Millipore; 1:400, MAB377), mouse anti-GAD67 (Millipore; 1:500, MAB5406B); guinea pig anti-Vglut2 (Millipore; 1:1000, ab2251); rabbit anti-Sox14 (Thermo Fisher Scientific; 1:50, PA5-77173); rat anti-Nr5a1 (also known as Sf1; TransGenic; 1:250, KO610), rabbit anti-NMDA receptor (Abcam; 1:500, ab65783) and rabbit anti-cleaved caspase 3 (BD Pharmigen; 1:800, 559565). The secondary antibody used was goat anti-IgG and was either Alexa Fluor-488, -546 or -648-conjugated (Thermo Fisher Scientific; 1:200-1:500). All samples were counterstained with Hoechst nuclear stain (Thermo Fisher Scientific; 1:2000).
In utero electroporation
C57BL/6J pregnant female dams were subjected to IUE as conducted previously (Dixit et al., 2011a; Rosin and Kurrasch, 2018). The pCIG2 expression vector contains a β-actin promoter/CMV enhancer and an IRES–EGFP cassette, described previously (Dixit et al., 2014). Briefly, pCIG2, pCIG2-Ascl1 and/or pNAS-HA-Ngn3 were injected into the lateral ventricle of E12.5 brains. Following DNA injection, the embryonic brains were gently squeezed to allow the injected DNA to travel to the third ventricle adjacent to the hypothalamus. BTX platinum-plated electrodes (Harvard Apparatus) and a BTX ECM 830 Electro Square Porator (Harvard Apparatus) were used to pulse (45 V, 50 ms) embryonic brains five times, separated by intervals of 950 ms. The uterine horns containing the electroporated embryos then were returned to the abdominal cavity and the pregnant dam was sutured closed. Embryos were collected 2 days later at E14.5. The harvested embryonic pup brains were fixed and prepared as described above for immunostaining experiments.
Quantification and statistical analysis
Images were captured using a Zeiss Axioplan 2 manual compound microscope with a Zeiss Axiocam HRc camera and then analyzed for cell number quantification. Anti-Fezf1 and anti-Nkx2.1 (Kurrasch et al., 2007) immunolabeling of adjacent sections was used to mark the beginning and end of the VMH. Cells were counted in at least three brain sections (∼20 µm apart) for each embryo analyzed, focusing on the rostral to mid-caudal VMH. At least three embryonic samples from more than two pregnant dams were used for each experimental group. ImageJ software was used to produce binary images and plots. Using GraphPad Prism 7, an unpaired two-tailed Student's t-test was used to assess statistical differences between controls and mutants. Results are displayed as mean±s.e.m.
Acknowledgements
We thank our colleagues in the Kurrasch lab for critical discussion and their helpful insights throughout the course of this project.
Footnotes
Author contributions
Conceptualization: S.A., C.S., D.M.K.; Methodology: S.A., J.M.R., A.B., N.K., F.B.; Validation: S.A.; Formal analysis: S.A., J.M.R., A.B.; Investigation: S.A., J.M.R., A.B., N.K.; Resources: F.B., G.G., C.S., D.M.K.; Data curation: S.A.; Writing - original draft: S.A.; Writing - review & editing: S.A., G.G., C.S., D.M.K.; Visualization: S.A.; Supervision: C.S., D.M.K.; Project administration: D.M.K.; Funding acquisition: D.M.K.
Funding
This research was supported by the Canadian Institutes of Health Research (CIHR MOP-275053 to D.M.K.; PJT-406690 to C.S.; MEF-140891 to J.M.R.) and the Alberta Children's Hospital Foundation (fellowship award to S.A. and J.M.R.). C.S. also holds the Dixon Family Chair in Ophthalmology Research.
References
Competing interests
D.M.K. is co-founder of Path Therapeutics, a start-up focused on anti-seizure drug discovery. The other authors declare no competing or financial interests.