Characterization of Hypothalamic MCH Neuron Development in a 3D Differentiation System of Mouse Embryonic Stem Cells

Abstract Hypothalamic melanin-concentrating hormone (MCH) neurons are important regulators of multiple physiological processes, such as sleep, feeding, and memory. Despite the increasing interest in their neuronal functions, the molecular mechanism underlying MCH neuron development remains poorly understood. We report that a three-dimensional culture of mouse embryonic stem cells (mESCs) can generate hypothalamic-like tissues containing MCH-positive neurons, which reproduce morphologic maturation, neuronal connectivity, and neuropeptide/neurotransmitter phenotype of native MCH neurons. Using this in vitro system, we demonstrate that Hedgehog (Hh) signaling serves to produce major neurochemical subtypes of MCH neurons characterized by the presence or absence of cocaine- and amphetamine-regulated transcript (CART). Without exogenous Hh signals, mESCs initially differentiated into dorsal hypothalamic/prethalamic progenitors and finally into MCH+CART+ neurons through a specific intermediate progenitor state. Conversely, activation of the Hh pathway specified ventral hypothalamic progenitors that generate both MCH+CART− and MCH+CART+ neurons. These results suggest that in vivo MCH neurons may originate from multiple cell lineages that arise through early dorsoventral patterning of the hypothalamus. Additionally, we found that Hh signaling supports the differentiation of mESCs into orexin/hypocretin neurons, a well-defined cell group intermingled with MCH neurons in the lateral hypothalamic area (LHA). The present study highlights and improves the utility of mESC culture in the analysis of the developmental programs of specific hypothalamic cell types.

Despite the growing body of literature on the physiology of MCH neurons, the number of studies focusing on their developmental process is limited. In rodents, the birth of MCH neurons occurs in mid-to-late gestation [embryonic day (E)10-E16 in rats and E9-E14 in mice] and the expression of MCH becomes evident in the prenatal period (Brischoux et al., 2001;Steininger et al., 2004;Croizier et al., 2010;Díaz et al., 2014). MCH neurons have at least two subpopulations, which are characterized by their neurochemical phenotypes, birthdates, and projection patterns (Brischoux et al., 2002;Cvetkovic et al., 2004;Croizier et al., 2010). One population co-expresses cocaine-and amphetamine-regulated transcript (CART) and neurokinin-3 receptor (NK3R), the other expresses neither of these molecules. The latter (i.e., MCH 1 CART À NK3R À group) corresponds to initially generated MCH neurons, localized in the lateral hypothalamic area (LHA), and is the main source of descending MCH fibers to the spinal cord. The later born MCH 1 CART 1 NK3R 1 population is distributed in the LHA and more medial hypothalamic areas and constitutes the major fraction of cortically projecting MCH neurons. The existence of these two subpopulations was recently corroborated by a single-cell transcriptomic analysis of MCH 1 cells in the mouse LHA (Mickelsen et al., 2019). The heterogeneous nature of MCH neurons may contribute to parallel regulation of memory processes and muscle tone during REM sleep (Hanriot et al., 2007;Hassani et al., 2009;Izawa et al., 2019). Although previous studies have identified some transcription factors and morphogens critical for MCH neuron development (Szabó et al., 2009;Sokolowski et al., 2015;Seifinejad et al., 2019), little is known about how the MCH neuronal subtypes are produced in the developing hypothalamus.
In vitro generation of hypothalamic neurons from embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs) can provide insights into the specification programs of those neurons. Such a differentiation system was first established for mouse ESCs (mESCs) using three-dimensional (3D) culture named SFEBq (serum-free floating culture of embryoid body-like aggregates with quick reaggregation; Wataya et al., 2008). Since then, hypothalamic neurons have been made from human ESCs and iPSCs (hESCs/hiPSCs) in SFEBq-based 3D culture Ogawa et al., 2018;Kasai et al., 2020) or two-dimensional culture Wang et al., 2015;Rajamani et al., 2018). These hESC/hiPSC cultures would be useful for basic and clinical research on the human hypothalamus, but there are two advantages of using mESCs over hESCs/hiPSCs for general purposes: a shorter differentiation period and direct comparability with in vivo data from animal experiments. However, to our knowledge, no previous study has focused on MCH neuron development in mESC-based differentiation systems.
For the dissociation culture of mESC-derived neurons, cells were dissociated from aggregates on days 19-20 or 30-33 using Neuron Dissociation Solutions (Wako, #297-78101) and plated onto poly-D-lysine/laminin-coated coverslips or 24-well plates at a density of 5.0-8.5 Â 10 4 cells/cm 2 in NDM. After 3 d of adherent culture, cells were subjected to an immunocytochemical analysis or calcium imaging experiment.

Quantification of Rax::GFP 1 cells
To examine the time-dependent change of the Rax:: GFP expression, differentiating mESC aggregates (eight aggregates per day) from a single experimental batch were dissociated with Accumax (Innovative cell technologies, #AM105) to single cells, and their fluorescence images were acquired with a DMI6000B microscope (Leica MicroSystems). Rax::GFP 1 cells were manually quantified using the cell counter plugin for Fiji (a distribution of ImageJ, RRID: SCR_002285; Schindelin et al., 2012). To verify the reproducible induction of Rax::GFP 1 cells, their percentage was measured by flow cytometry in multiple experimental batches on day 7.

Flow cytometry and cell sorting
To measure the percentage of Rax::GFP 1 cells on day 7, cells were dissociated from at least 32 aggregates per batch with Accumax and suspended in Cell Staining Buffer (BioLegend, #420201). Dead cells were stained with 7-AAD (Tonbo Biosciences, #13-6993), and the 7-AAD À live cell fraction (.10,000 events) was analyzed using a BD FACS Calibur (BD Biosciences) and the BD CellQuest Pro software program (BD Biosciences, RRID: SCR_014489). WT mESC-derived cells were used as a negative control for setting the Rax::GFP 1 gate.

Calcium imaging
Dissociation cultures prepared in 24-well plates were used for calcium imaging on day 36. Before imaging, cells were subjected to a loading of the Ca 21 indicator Calbryte-520 AM (AAT Bioquest, #20650) and nuclear staining with Hoechst 33 342 (Dojindo, #H342) as follows. 2Â loading solution containing 10 mM Calbryte-520 AM, 5 mg/ml Hoechst 33342, and 0.04% Pluronic F-127 (Biotium, #59004) was prepared in artificial CSF (ACSF). The composition of ACSF was (in mM) 132 NaCl, 3 KCl, 1.3 MgCl 2 , 2.4 CaCl 2 , 20 NaHCO 3 , 1.2 KH 2 PO 4 , 3 HEPES, and 10 glucose. Half the volume of the culture medium was changed with 2Â loading solution, and cells were cultured for 60-90 min. Cells were then washed and maintained in ACSF at room temperature (RT) before and during imaging. To measure Ca 21 responses, time-lapse images of cells were captured at six time points (0, 60, 120, 250, 310, and 370 s) using the Opera Phenix highcontent imaging system (PerkinElmer). Cells were stimulated with KCl, glutamate, or GABA at 210 s. For KCl stimulation, extracellular fluid (ECF) was manually changed from ACSF to high-K 1 ACSF containing 50 mM KCl and 85 mM NaCl. Glutamate and GABA were applied by replacing half the volume of ECF with ACSF containing 200 mM glutamate or GABA. After calcium imaging, cells were fixed with 4% PFA in PBS for 10 min and subjected to immunocytochemistry for MCH to identify MCH-expressing cells. Calbryte-520 fluorescence from the identified MCH 1 cells was quantified using the Harmony software program (PerkinElmer) by measuring mean pixel intensity in somatic areas, which were automatically detected based on nuclear staining with Hoechst. The fluorescence intensity was graphed as DF/F 0 , where DF is the intensity difference compared with background fluorescence (F 0 ). The intensity value at 120 s was used as F 0 . Responding cells were defined as the cells with DF/F 0 . 0.5 at 370 s because spontaneous fluctuations in DF/F 0 were limited in 60.5.

Immunofluorescence staining
After more than two weeks of differentiation, mESC aggregates were analyzed by whole-mount immunohistochemistry as follows. First, aggregates were fixed with 4% PFA in PBS for 20 min at RT. Fixed aggregates were blocked and permeabilized with 5% normal donkey serum (NDS) and 0.5% Triton X-100 in PBS for 1 h at RT and then incubated with primary antibodies overnight at 4°C, except that incubation with an anti-orexin-A antibody was performed for 36-48 h. After washing in PBS supplemented with 0.05% Triton X-100 (PBSX), aggregates were incubated with appropriate secondary antibodies for 1 h at RT. In this step, nuclear staining with DAPI (Dojindo, #D523) was performed simultaneously. Antibodies and DAPI were diluted with PBS containing 5% NDS and 0.1% Triton X-100 just before use. Finally, aggregates were washed in PBSX and mounted on glass slides using Fluoromount (Diagnostic BioSystems, #K024). The antibodies used in this study are listed in Table 1.
For the immunohistochemical analysis of aggregates cultured for 7-13 d, they were fixed with 4% PFA for 1 h at RT and cryoprotected in sucrose solutions of increasing concentrations (10, 20, and 30%) for 30 min each, followed by embedding in O.C.T. compound (Sakura, #4583). Serial 10-mmthick cryosections were prepared with a CM3050S cryostat (Leica Biosystems) and mounted on PLATINUM PRO adhesive slides (Matsunami, #SPRO-02). The sections were blocked and permeabilized with 5% NDS and 0.1% Triton X-100 in PBS for 30 min at RT and then immunostained as described above.
For the immunocytochemical analysis of dissociation cultures, cells were fixed with 4% PFA for 10 min at RT and then processed in the same way as cryosections. Fluorescence images of whole-mount aggregates were obtained using a LSM710 confocal microscope (Carl Zeiss), and those of cryosections and dissociation cultures were obtained using a DMI6000B microscope, LSM710 confocal microscope, or BZ-9000 microscope (Keyence).

Birth-dating analysis of postmitotic neurons
During days 9-19, mESC aggregates were cultured for 24 h in the differentiation medium containing 10 mM 5ethynyl-2'-deoxyuridine (EdU) by the day. After the EdU treatment, aggregates were washed and cultured in differentiation medium free of EdU until day 28 and fixed with 4% PFA for 20 min at RT. Aggregates were then subjected to visualization of EdU, followed by immunofluorescence staining for the neuronal marker HuC/D. The EdU labeling and staining were performed using a Click-iT EdU Alexa Fluor 647 Imaging kit (Invitrogen, #C10340). Fluorescence images were captured with a LSM710 confocal microscope, and EdU 1 HuC/D 1 double-positive cells were considered to be postmitotic neurons born during EdU treatment. Cell counts were performed as described below.

The imaging analysis of immunofluorescence data
The long and short diameters of MCH 1 cell bodies were measured using the straight line and measure tools in Fiji. In the birth-dating analysis of postmitotic neurons, the percentage of EdU 1 cells in the HuC/D 1 neuronal population was determined as follows. First, in the fluorescence images acquired from each aggregate, HuC/D 1 cells were manually quantified using the cell counter plugin for Fiji (200-500 positive cells per aggregate). Next, EdU 1 cells were counted in the prequantified HuC/D 1 population to calculate the percentage of its EdU 1 fraction. Although a part of HuC/D 1 cells exhibited diffuse EdU staining, they were also considered as EdU 1 cells. This analysis protocol was also applied to determine the MCH 1 percentage in the HuC/D 1 population, the HuC/D 1 percentage in the MCH 1 population, and the positive rates of neurochemical markers in the MCH 1 population. For the analysis of cryosections, fluorescence images were acquired from three non-adjacent sections per aggregate. Several images from day-7 aggregates were analyzed using the Columbus software program (PerkinElmer) to calculate the Pax6 1 or Nkx2.1 1 percentage in the Rax 1 cell population semi-automatically (200-400 Rax 1 cells were detected per aggregate).

Statistical analysis
Statistical analyses were performed using the R software program (version 4.0.5, The R Foundation, RRID: SCR_001905). Results of grouped data were presented as the mean 6 SEM in the text or figures (bar plots). Some data were presented with box plots showing the median (bold line), interquartile range (IQR; box), and 61.5 IQR (whisker) with outliers ( Fig. 2C) or all data points (Fig. 5H). Statistical significance between two groups was assessed by Welch's t test, except for Figure 5H, in which two groups were compared using the Brunner-Munzel test, which does not require assumptions of normality and homoscedasticity. For time series data, statistical significance between the earliest time point and other timepoints was assessed by Dunnett test. For statistical analysis of qRT-PCR data, a log10 transformation was performed to reduce variability of data. The results of the statistical analyses are summarized in Table 2.

ES-Hypo recapitulates the temporal pattern of hypothalamic development
The SFEBq method, using gfCDM, efficiently induces Rax 1 hypothalamic progenitors from mESCs after 7 d of differentiation (Wataya et al., 2008). Rax is a transcription factor specifically expressed in the hypothalamic and retinal neuroepithelium (Furukawa et al., 1997;Mathers et al., 1997;Shimogori et al., 2010;Lu et al., 2013). The Rax 1 progenitors subsequently generate their neuronal derivatives, such as vasopressin cells, after 20 or more days (Wataya et al., 2008). In this culture system, however, the time course of neuronal differentiation has not been fully characterized. We therefore started our study by systematically tracing Rax 1 cell induction and subsequent neurogenesis during ES-Hypo development (Fig. 1A).
To monitor Rax 1 cell generation in real time, a Rax:: GFP reporter mESC line (Wataya et al., 2008) was subjected to SFEBq culture. The expression of Rax::GFP was sharply increased on day 5 and was maintained until day 7 in a large proportion of aggregated cells, followed by a gradual decrease (Fig. 1B,C). A flow cytometric analysis on day 7 confirmed that Rax::GFP 1 progenitors were reproducibly induced in gfCDM (;64% of total cells) but greatly decreased when the medium was supplemented with KSR (;2% of total cells; Fig. 1D). KSR is a widely used serum replacement with minimal growth factors, but it is reported to severely inhibit hypothalamic differentiation of mESCs via activating the insulin/PI3K/Akt pathway and rather induce telencephalic differentiation (Wataya et al., 2008). Since the expression of Rax::GFP was downregulated from days 7-10 ( Fig. 1C), we expected that neurogenesis starts in this period. To evaluate the time course of neurogenesis, we conducted a birth-dating analysis using the thymidine analog EdU during days 9-19. Newly born postmitotic neurons were labeled with EdU by the day, and they were quantified on day 28 as EdU and HuC/ D (a neuronal marker) double-positive cells (Fig. 1E,F). The results of the analysis were summarized in Figure 1G. The birth of postmitotic neurons has already occurred on day 9, and it reached a peak during days 11-13 (roughly half of HuC/D 1 neurons were born in this period) and persisted at lower levels thereafter. Eventually, 85% of HuC/ D 1 neurons were generated throughout the labeling period. Taken together, hypothalamic progenitor specification occurs within the first week of mESC differentiation and after neurogenesis, mostly within the second and third weeks. Considering that mESCs are derived from blastocysts (;E3.5 embryos), this time course resembles the temporal pattern of embryonic development of the mouse hypothalamus (Shimada and Nakamura, 1973;Ishii and Bouret, 2012;Lu et al., 2013).

Development of MCH neurons in ES-Hypo
Based on the determined time course of neurogenesis, we explored MCH neuron development in ES-Hypo after two weeks of differentiation. qRT-PCR showed that MCH mRNA levels increased exponentially during weeks 3-5 ( Fig. 2A). MCH-immunoreactivity was faintly observed on day 15, but clear MCH-immunoreactive (ir) cell bodies and fibers appeared on day 22 (Fig. 2B). These MCH-ir cell bodies were characterized by round shapes and small diameters (,10 mm), both of which are features of immature MCH neurons (Steininger et al., 2004;Li et al., 2018). After 36 d of culture, MCH-ir cells were often angularshaped and larger in comparison to MCH-ir cells on day 22 (Fig. 2B,C), which is reminiscent of neuronal maturation in vivo (Steininger et al., 2004;Li et al., 2018). Another hallmark of neuronal maturation is neurite growth; while young MCH neurons have only one or two neurites (i.e., unipolar or bipolar cells), a major fraction of adult MCH neurons are multipolar cells with an axon and 2-5 primary dendrites (Steininger et al., 2004;Diniz et al., 2019). To assess morphologic changes in mESC-derived MCH 1 cells in terms of neurite number, we prepared dissociation cultures from ES-Hypo on different days. Immunocytochemistry showed that 77% of MCHir cells were classified into unipolar or bipolar cells on days 22-23, but 59% of MCH-ir cells exhibited a multipolar morphology on days 33-36 (Fig. 2D,E). Some of the multipolar cells displayed an extended and branched axonal process (Fig. 2F). We also performed a calcium imaging experiment in the dissociation culture on day 36 and found that most MCH-ir cells were activated by KCl or glutamate but not by GABA (Extended Data Fig. 2-1). These morphologic and functional properties strongly support the neuronal identity of MCH-ir cells, and we further confirmed that the neuronal marker HuC/D was expressed in 93.2 6 1.2% of MCH-ir cells (n = 8 aggregates on day 36; Fig. 2G, left). The MCH 1 percentage of total HuC/D 1 neurons varied among individual aggregates (37-72%) but became much lower (2-13%) when SFEBq culturing had been performed in KSR-containing medium (Fig. 2G,H). These data indicate that ES-Hypo robustly generates MCH neurons, which reproduce the morphologic development of native MCH neurons.

Putative synaptic connections between MCH and orexin neurons in ES-Hypo
The axon growth observed for MCH neurons in ES-Hypo suggests that they may form synaptic connections with other differentiated neurons. There is some histologic and electrophysiological evidence to support reciprocal connections between MCH neurons and neighboring orexin neurons in the LHA (Bayer et al., 2002;Guan et al., 2002;van den Pol et al., 2004;Torterolo et al., 2006;Rao et al., 2008;Apergis-Schoute et al., 2015), which may be implicated in sleep/wake control (Konadhode et al., 2014;Hung et al., 2020). To test the possibility that such local wiring occurs in ES-Hypo, we first analyzed the differentiation of orexin neurons. Although orexin-ir cells were rarely observed before day 25, we encountered clusters of orexin-ir cells, showing neuronal morphologies, in a fraction of mESC aggregates cultured for four to five weeks (;25% of the aggregates tested; Fig. 3A). Consistent with this immunoreactivity, orexin mRNA levels were significantly higher on day 31 than on day 14 (Fig. 3B). Based on these results, we next performed double immunofluorescence staining for MCH and orexin in ES-Hypo cultured for more than one month. Confocal microscopy of orexin-rich aggregates showed that orexin-ir cells were largely immunonegative for MCH and ,10% of them were detected as double-positive cells (Fig. 3C). However, single-positive orexin-ir perikarya were often apposed by MCH-ir fibers or boutons (Fig. 3D). Similarly, we identified MCH-ir cells in contact with orexin-ir boutons (Fig. 3E). These findings support the hypothesis that ES-Hypo can produce some MCH-orexin neuronal networks that possibly resemble those in the LHA.
Neuropeptide/neurotransmitter phenotype of MCH neurons in ES-Hypo MCH neurons are known to express other neuropeptides, including CART and nesfatin-1, and GABAergic/ glutamatergic markers such as GAD67 (a GABA-synthesizing enzyme) and VGLUT2 (a machinery for glutamate uptake into synaptic vesicles; Broberger, 1999;Vrang et al., 1999;Brailoiu et al., 2007;Sapin et   examine whether mESC-derived MCH neurons have a similar neurochemical profile, we performed double immunofluorescence staining for MCH and the above markers in ES-Hypo on day 36 (Fig. 4A). Nesfatin-1 and VGLUT2 were expressed in 95.3% and 96.8% of MCH neurons in ES-Hypo, respectively (Fig. 4B), which is comparable to the values for MCH neurons in vivo (Foo et al., 2008;Fort et al., 2008;Vas et al., 2013;Chee et al., 2015;Mickelsen et al., 2017;Schneeberger et al., 2018;Naganuma et al., 2019;Hung et al., 2020). The expression of GAD67 was previously detected in 85% of native MCH neurons by an in situ hybridization (ISH) analysis (Sapin et al., 2010;Jego et al., 2013;Mickelsen et al., 2017), but the GAD67-ir fraction of MCH neurons in ES-Hypo was somewhat smaller (64.7%; Fig. 4B). Because immunostaining methods are less sensitive than ISH in the detection of cell bodies of neurons that express low levels of GAD mRNAs (Esclapez et al., 1994), it is likely that the GAD67 1 ratio of mESC-derived MCH neurons is similar to that of native MCH neurons. In contrast, this similarity in the population ratio was not observed for CART; while MCH 1 CART 1 neurons constitute only half of the whole MCH 1 population in the mouse hypothalamus (Croizier et al., 2010;Mickelsen et al., 2017), 95.9% of MCH-ir cells in ES-Hypo were labeled by an anti-CART antibody (Fig. 4B). Collectively, mESC-derived MCH neurons share a series of neuropeptide/neurotransmitter markers with native MCH neurons and mostly represent the CART 1 subpopulation.
Subsequent culture with the continuous addition of SAG still produced many MCH-ir neurons (Extended Data Fig.  5-2), but they contained a larger proportion of CART-negative cells in comparison to MCH neurons without SAG (Fig.  5F). MCH 1 CART À neurons were generally found in clusters and not co-distributed with MCH 1 CART 1 neurons (Fig.  5E), implying separate developmental pathways for both populations. Since CART 1 but not CART À MCH neurons specifically express NK3R in vivo (Croizier et al., 2010), we further tested NK3R-immunoreactivity in SAG-treated samples. Because the expression of CART was limited to the MCH 1 cell population (Fig. 5E, left), we simply performed double immunostaining for CART/NK3R and found over 95% co-localization of these markers (Extended Data Fig. 5-3). This result indicates that NK3R is expressed in the MCH 1 CART 1 but not MCH 1 CART À cell group in ES-Hypo, as in the native hypothalamus. The SAG treatment also influenced the differentiation of orexin neurons. Orexin-ir cells were nearly absent in SAG-free ES-Hypo on day 30, but several orexin-ir cells were found in the majority of SAG-treated ES-Hypo on the same day (Fig. 5G,H). Together, our findings suggest that the activation of the Hh pathway is necessary for the differentiation of mESCs into certain types of LHA cells, including CART-negative MCH neurons and orexin neurons, and that such LHA cells are derived from ventral hypothalamic progenitors.
Since both CART 1 and CART À MCH neurons as well as orexin neurons were simultaneously produced in SAGtreated ES-Hypo, we reevaluated whether the MCH-orexin neuronal connectivity is specific to either subtype of MCH neurons. We performed triple immunostaining for MCH/orexin/CART but could not determine which subtypes of MCH neurons send fibers to orexin neurons, because CART-immunoreactivity was weak or undetectable in the fiber compartment (Fig. 5E). On the other hand, we observed both MCH 1 CART 1 and MCH 1 CART À cell bodies in contact with orexin-ir boutons (Fig. 5I). Thus, both subtypes of MCH neurons seem to receive orexinergic inputs in ES-Hypo.

Progenitor origin of MCH neurons in ES-Hypo
Our results suggest that MCH neurons develop from multiple progenitor lineages which are specified through early dorsoventral patterning of the hypothalamus. To better understand the differentiation process of MCH neurons, we examined the expression patterns of regional markers as well as differentiation markers in ES-Hypo from days 7-22.

Discussion
MCH neurons are preferentially located in the LHA, but they have no well-defined territory and are intermingled with other cell groups, such as orexin neurons (Hahn, 2010). This organization makes it hard to predict the embryonic progenitor domain as well as the developmental signals (e.g., morphogens and transcriptional factors) allocated for MCH neurons. ES-Hypo can partially resolve this difficulty; it is usable for screening culture conditions to induce specific hypothalamic lineages in a systematic, stepwise manner, and such trials should provide crucial information on the lineage specification and maturation. In the present study, we used this approach to gain insight into the mechanism of MCH neuron development. We found that ES-Hypo can produce several MCH-ir neurons that reproduce well-known features of native MCH neurons. Most importantly, we uncovered that the major neurochemical subpopulations of MCH neurons, the CART 1 and CART À MCH neurons, are differentiated from mESCs under the influence of Hh signaling.
The original report on ES-Hypo was primarily focused on the culture conditions to specify Rax 1 hypothalamic progenitors and did not fully examine the neuronal fates of the progenitors (Wataya et al., 2008). The authors described that the Rax 1 progenitors bearing a dorsal or ventral hypothalamic identity could generate vasopressin neurons or neurons observed in the ventromedial and arcuate nuclei of the hypothalamus, respectively. Subsequent reports have documented the occurrence of orexin neurons in ES-Hypo  and its modified culture (Hayakawa et al., 2013); however, the present study provides the first characterization of MCH neurons in ES-Hypo. We found that a large fraction (37-81%) of neurons in ES-Hypo were immunopositive for MCH, regardless of whether ES-Hypo was derived from dorsal or ventral hypothalamic progenitors. This proportion seems to be much higher in comparison to previous reports on vasopressin neurons (6% of day-20 cells differentiated from Rax::GFP 1 progenitors; Wataya et al., 2008). Indeed, we observed less copeptin-ir cells (copeptin is a peptide derived from the vasopressin precursor) in comparison to MCH-ir cells in ES-Hypo after four to five weeks of differentiation (data not shown). These data suggest that ES-Hypo generates a wide range of hypothalamic neurons, among which MCH neurons are a predominant population.
We examined the time course of neural progenitor specification and subsequent neurogenesis in ES-Hypo and showed that it follows the temporal pattern of fetal hypothalamic development. In addition to this overall similarity, MCH cells born in ES-Hypo reproduced morphologic changes observed in vivo for developing MCH neurons and acquired neurochemical phenotypes common to mature hypothalamic MCH neurons (the latter feature is discussed in detail below). Moreover, we found that orexin neurons developed later than MCH neurons in ES-Hypo and these cell groups appeared to form reciprocal connections. Both findings are reminiscent of native MCH and orexin neurons in the rodent hypothalamus (Bayer et al., 2002;Guan et al., 2002;Steininger et al., 2004;van den Pol et al., 2004;Díaz et al., 2014). Although it is unknown whether the MCH-orexin neuronal connections are specific to either of the subpopulations of MCH neurons in vivo, our results suggest that both CART 1 and CART À MCH neurons receive inputs from orexin neurons in ES-Hypo. A recent study succeeded in generating MCH and orexin neurons from mouse iPSCs using a culture protocol based on that for ES-Hypo (Seifinejad et al., 2019). The authors achieved neuronal differentiation within two weeks of culture by applying inductive factors, such as BMP7 and vitamin C, whereas we confirmed the generation of MCH and orexin neurons after longer culture of ES-Hypo without any inductive factors. Taken together, our data support the idea that ES-Hypo can recapitulate the developmental programs of MCH neurons and probably orexin neurons in an autonomous manner.
One of the most remarkable properties of native MCH neurons is their co-expression of multiple neurochemical molecules, including neuropeptides (CART and nesfatin-1) and markers for GABAergic and glutamatergic cells (GAD67 and VGLUT2). The expression of CART, nesfatin-1, and GAD67 in MCH neurons was previously detected at both mRNA and protein levels (Broberger, 1999;Vrang et al., 1999;Brailoiu et al., 2007;Sapin et al., 2010;Mickelsen et al., 2017;Noble et al., 2018). The expression of VGLUT2 was found in nearly all MCH neurons using VGLUT2-GFP reporter mice, single-cell transcriptomics, and ISH (Chee et al., 2015;Mickelsen et al., 2017Mickelsen et al., , 2019Schneeberger et al., 2018;Naganuma et al., 2019;Hung et al., 2020). It has been reported that VGLUT2 deletion in MCH neurons affects glucose metabolism, food reward, and REM sleep in mice (Schneeberger et al., 2018;Naganuma et al., 2019), suggesting functional glutamate signaling by MCH neurons. On the other hand, another study using VGLUT2-tdTomato mice (generated by mating VGLUT2-Cre mice to Ai14 Cre reporter mice) failed to detect tdTomato signals in MCH neurons (Blanco-Centurion et al., 2018). Our results based on double immunofluorescence staining indicated that mESC-derived MCH neurons largely co-express CART, nesfatin-1, GAD67, and VGLUT2. The co-expression of these markers has been confirmed in native MCH neurons at mRNA levels (Mickelsen et al., 2017); however, to our knowledge, this is the first report to demonstrate the simultaneous presence of endogenous GAD67 and VGLUT2 proteins in MCH neurons. Although the functional consequences of this neurochemical multiplicity remain to be elucidated, it has been shown that native MCH neurons can release GABA (Jego et al., 2013) and glutamate (Chee et al., 2015), as well as MCH (Noble et al., 2018). Thus, the present study not only supports the neurochemical similarity between mESC-derived and native MCH neurons, but also provides additional evidence of the dual GABAergic-glutamatergic phenotype of MCH neurons.
The most striking finding of our study is that CART-negative MCH neurons substantially developed in ES-Hypo only when the Hh agonist SAG was added to the differentiation media. In the embryonic mouse brain, most MCH 1 CART À neurons are born at the initial phase of neurogenesis (E9-E10) and followed by the generation of MCH 1 CART 1 neurons (E11-E12; Croizier et al., 2010). This birth order is reflected in the final localization of those neurons in the adult brain. Earlier born MCH 1 CART À neurons are predominantly located in the dorsal-lateral part of the caudal LHA close to the cerebellar peduncle, although a minor fraction of them are found in the perifornical region of the LHA at more caudal levels (Croizier et al., 2010). Later born MCH 1 CART 1 neurons are mainly distributed in the rostral and medial hypothalamic portions, but also in additional locations, including the rostral zona incerta, the dorsal-lateral part of the caudal LHA, and the ventral capsule of the posterior hypothalamic nucleus (Croizier et al., 2010). It remains unknown whether these subtypes of MCH neurons originate from a common or distinct progenitors. On the other hand, a previous study has shown that early-born MCH 1 cells in the E13-E15 rat brain appear to come from the Nkx2.1/Nkx2.2 coexpressing region in the caudal hypothalamic neuroepithelium (Croizier et al., 2011;Fig. 5A). The corresponding region is also proposed as an origin of early MCH cells in the fetal mouse brain (Díaz et al., 2014). Consistent with these in vivo data, SFEBq-cultured mESCs differentiated into Nkx2.1 1 Nkx2.2 1 hypothalamic progenitors and then generated MCH neurons, including both CART 1 and CART À cells, in the presence of SAG. Under this condition, most mESCs were specified into Nkx2.1 1 ventral hypothalamic progenitors by Hh signaling, like the patterning event in the hypothalamic anlage . Only half of the Nkx2.1 1 progenitors coexpressed Nkx2.2, but we observed co-localization of Nkx2.2 with both subtypes of MCH neurons after terminal differentiation. Based on this result, we concluded that Nkx2.1 1 Nkx2.2 1 progenitors are an origin of both CART 1 and CART À MCH neurons in SAG-treated ES-Hypo. We cannot rule out the possibility that other types of neural progenitors also contribute to the production of MCH neurons, as the expression of Nkx2.1/Nkx2.2 was limited to 30-40% of MCH neurons (Fig. 6I). However, the expression of the Nkx genes seems downregulated on terminal differentiation of MCH neurons (Croizier et al., 2011), suggesting that most MCH neurons are produced by Nkx2.1 1 Nkx2.2 1 progenitors in the presence of SAG. It is currently unclear how Nkx2.1 1 Nkx2.2 1 progenitors generate the two subtypes of MCH neurons. A plausible hypothesis is that CART À and CART 1 MCH neurons are derived from early and intermediate stages of Nkx2.1 1 Nkx2.2 1 progenitors, respectively, according to the birth order for in vivo MCH neurons (Fig. 8). This idea is consistent with our observation that both subpopulations of MCH neurons were generally found in discrete clusters and not intermingled with each other in ES-Hypo (Fig. 5E). In this case, a stage-specific phenotype of Nkx2.1 1 Nkx2.2 1 progenitors may determine the phenotype of their descendant MCH neurons. Alternatively, extracellular factors may influence late generated MCH neurons to express CART, as suggested previously (Cvetkovic et al., 2004).
In the present study, we only performed SAG treatment in a continuous manner (from day 4 onward) and therefore could not clarify whether Hh signaling has an effect not only on ventralizing hypothalamic progenitors but also on their subsequent terminal differentiation. In this respect, a genetic study has revealed that the elimination of the functional Shh expression in the developing diencephalon severely impairs the production of MCH neurons, especially in the LHA (Szabó et al., 2009), wherein MCH 1 CART À neurons are predominantly localized. It has also been reported that the MCH mRNA level was decreased by 90% in E13 mouse embryos after pregnant mice were injected with an Hh inhibitor on E11 (Croizier et al., 2011), when most presumptive MCH 1 CART À neurons have already been born in embryos (Croizier et al., 2010). These previous findings provide evidence that Hh signaling is required for terminal differentiation and/or survival of MCH 1 CART À neurons.
Without SAG treatment, SFEBq-cultured mESCs differentiated into Rax 1 Pax6 1 dorsal hypothalamic progenitors or Rax À Pax6 1 dorsal hypothalamic/prethalamic progenitors within the first week of culture. Under this condition, terminally differentiated MCH neurons were almost completely CART 1 . Through FACS sorting and subsequent differentiation analysis, we found that Rax À Pax6 1 progenitors are an early origin of MCH 1 CART 1 neurons in SAG-free ES-Hypo. These early progenitors next generate the second form of neural progenitors (Nkx2.1 1 Nkx2.2 À ), and this group of intermediate progenitors directly produces MCH neurons (Fig. 8). Although the regional identity of these intermediate progenitors remains unclear, a possible candidate is the perimamillary/periretromamillary (PM/PRM) region in the prosomeric hypothalamic model (Díaz et al., 2014). This region was identified as the second site of occurrence of MCH 1 cells, more ventral to the first site (i.e., the Nkx2.1/Nkx2.2-positive region), by using the Allen Developmental Mouse Brain Atlas (Díaz et al., 2014). The PRM/PM area expresses Nkx2.1 but not Nkx2.2 (Díaz et al., 2014), consistent with the phenotype of the intermediate progenitors identified in SAG-free ES-Hypo.
In summary, we propose three different progenitor origins of MCH neurons in ES-Hypo ( Fig. 8): (1) Nkx2.1 1 Nkx2.2 1 early progenitors, (2) Nkx2.1 1 Nkx2.2 1 intermediate progenitors, and (3) Nkx2.1 1 Nkx2.2 À intermediate progenitors. The first group selectively produces CART-negative MCH neurons, and the others CART-positive MCH neurons. The multiple progenitor origins of MCH neurons can be supported by recent transcriptome studies. Kim et al. (2020) performed clustering of embryonic mouse LHA cells (at E11-E13) using single-cell RNA-sequencing and identified 13 subclusters of young postmitotic neurons. Among them, five subclusters expressed both MCH and CART mRNAs and one subcluster MCH mRNA alone, implying that MCH neurons develop through multiple pathways. A similar transcriptional diversity was also found in hypothalamic POMC precursors (immature POMC 1 neurons), and it was shown that distinct subclusters of POMC precursors finally develop into different phenotypes of arcuate neurons (Yu et al., 2022). In light of these findings, it is plausible that CART 1 and CART À MCH neurons are developed from distinct progenitors. Our results also suggested that CART-positive MCH neurons originate from two distinct progenitor lineages, which are specified in the presence or absence of Hh signaling (Fig. 8). This may account for the relatively wide distribution of MCH 1 CART 1 neurons compared with MCH 1 CART À neurons in the adult brain. Further studies, using a lineage tracing approach are needed to better understand the ontogeny of MCH neurons.
Another notable effect of SAG on ES-Hypo was the increased production of orexin neurons. Orexin-ir cells were observed only occasionally in the absence of SAG but were more reproducibly observed in the SAG-treated samples. Although the ontogeny of orexin neurons remains largely unknown, the LIM homeobox transcription factor Lhx9 has been identified as a critical inducer of the orexin neuron specification in zebrafish and developing mice (Dalal et al., 2013;Liu et al., 2015). In the latter species, the expression of Lhx9 occurs in the mantle layer of the tuberal hypothalamus during the neurogenic period and then colocalizes with a subset of orexin neurons in the LHA (Rétaux et al., 1999;Shimogori et al., 2010;Sokolowski et al., 2015). It has also been revealed that Shh signaling is indispensable for the development of orexin neurons as well as Lhx9 1 cells in the mouse hypothalamus (Szabó et al., 2009;Shimogori et al., 2010). These previous data and our current findings strongly suggest that orexin neurons primarily arise from ventral hypothalamic progenitors under the control of Hh signaling. In this view, the stochastic induction of orexin cells without SAG is attributable to a small number of Nkx2.1 1 progenitors (Fig. 5B,C). In comparison to the abundant production of MCH neurons in ES-Hypo, the number of co-existing orexin neurons was much smaller, even in the presence of SAG (,15 cells on day 30). In rodents, most MCH neurons appear before birth; however, the number of orexin neurons as well as orexin immunoreactivity increase in the postnatal period (Yamamoto et al., 2000;Steininger et al., 2004;Ogawa et al., 2017). These observations suggest that the generation of orexin neurons in ES-Hypo may be promoted by extending the culture period in addition to SAG treatment. However, with more than five weeks of mESC differentiation, there was a massive expansion of residual undifferentiated cells, which Figure 8. Schematic diagram of the different progenitor origins of MCH neurons suggested in the current study. SFEBq culture of mESCs in gfCDM with or without exogenous Hh signals can generate Nkx2.1 1 Nkx2.2 1 ventral hypothalamic progenitors or Rax À Pax6 1 dorsal hypothalamic/prethalamic progenitors, respectively, within a week. The Nkx2.1 1 Nkx2.2 1 early progenitors directly produce MCH 1 CART À neurons and Nkx2.1 1 Nkx2.2 1 intermediate progenitors, the latter of which generate MCH 1 CART 1 neurons. The Rax À Pax6 1 early progenitors also differentiate into MCH 1 CART 1 neurons through Nkx2.1 1 Nkx2.2 À intermediate progenitors.
negatively affected the survival of differentiated neurons (data not shown).
In conclusion, ES-Hypo can recapitulate the developmental process of MCH neurons in vitro and therefore be a powerful tool to dissect the molecular mechanisms underlying this process. We also demonstrated that the Hh signaling pathway is a key organizer of the CART 1 and CART À subpopulations of MCH neurons. These neurochemical subpopulations have also been found in the human hypothalamus (Menyhért et al., 2007). Hence, future comparative studies on MCH neurons derived from rodent and human stem cell lines may help clarify whether Hh signaling plays a conserved role in establishing the heterogeneity of MCH neurons in mammals.