Control of tuberal hypothalamic development and its implications in metabolic disorders

Abstract

The tuberal hypothalamus regulates a range of crucial physiological processes, including energy homeostasis and metabolism. In this Review, we explore the intricate molecular mechanisms and signalling pathways that control the development of the tuberal hypothalamus, focusing on aspects that shape metabolic outcomes. Major developmental events are discussed in the context of their effect on the establishment of both functional hypothalamic neuronal circuits and brain–body interfaces that are pivotal to the control of metabolism. Emerging evidence indicates that aberrations in molecular pathways during tuberal hypothalamic development contribute to metabolic dysregulation. Understanding the molecular underpinnings of tuberal hypothalamic development provides a comprehensive view of neurodevelopmental processes and offers a promising avenue for future targeted interventions to prevent and treat metabolic disorders.

Introduction

The hypothalamus is an evolutionarily ancient part of the forebrain. Its core cell types have been functionally conserved owing to their role in the central regulation of life-supporting homeostatic, adaptive and reproductive physiological processes^1,2. Hypothalamic cells and circuits integrate inputs from the external and internal environment, compare these to internal set points, and then initiate feedback responses that act via the nervous, endocrine and immune systems to restore or promote optimal physiological responses. Within this broad system, the tuberal hypothalamus regulates appetite, energy balance, growth, and reproductive behaviours, controls emotional behaviours, and enables adaptation to changing physiological situations across the lifespan such as puberty and pregnancy.

Anatomically, the tuberal hypothalamus sits in the anterior-ventral-most hypothalamus, encircling the third ventricle of the brain. Tuberal neurons are situated in the neuron-rich retrochiasmatic nucleus, arcuate nucleus (ARCN), ventromedial hypothalamic nucleus (VMH), dorsomedial hypothalamic nucleus (DMH) and premammillary hypothalamic nucleus (PMH) (Fig. 1a), and in the neuron-scarce lateral hypothalamic area. The tuberal hypothalamus is bordered anteriorly by the anterior hypothalamic nucleus, supraoptic nucleus, suprachiasmatic nucleus and optic chiasm, dorsally by the paraventricular nucleus (PVN) and zona incerta, and posteriorly by the mammillary body (Fig. 1a). The posterior-ventral tuberal region is devoid of nuclei and neuronal cell bodies but instead is composed of highly specialized glial cells, termed tanyctes and pituicytes. Tanyctes are radial glial-like cells that form key components of the median eminence, a circumventricular organ lying outside the blood–brain barrier, whereas pituicytes are astrocytic-like cells found in the pituitary stalk and posterior pituitary (Fig. 1b). Astrocytes are found throughout the tuberal hypothalamus.

Fig. 1 |. Anatomy of the adult and developing tuberal hypothalamus.

a, The position of the adult tuberal hypothalamus and key resident nuclei (pink), relative to the mamillary body (MB), paraventricular nucleus (PVN), zona incerta (ZI), and the anterior pituitary and the posterior pituitary. b, Adult tuberal hypothalamic neurons and glia discussed in this Review. An important group of pro-opiomelanocortin (POMC) and neuropeptide Y (NPY) neurons project from the arcuate nucleus (ARCN) to the PVN. A subset of neurons from the PVN project to the ARCN; other PVN subsets project to the median eminence and posterior pituitary. Tanycytes line the base of the third ventricle and interface with portal capillaries in the median eminence. Pituicytes interface with portal capillaries in the posterior pituitary. c, Sagittal views of the developing brain in an embryonic chick at early stages. d, Sagittal views of the developing brain in an embryonic chick at late neurogenic stages. Boxed regions (enlarged views) focus on regions harbouring tuberal neurogenic (pink), tuberal gliogenic (yellow) and mammillary (purple) progenitors. The Rathke pouch gives rise to the anterior pituitary (blue). The infundibulum (yellow) develops in mid-embryogenesis, extending ventrally from posterior-most tuberal progenitors, to abut the forming anterior pituitary. A, anterior; D, dorsal; DMH, dorsomedial hypothalamic nucleus; P, posterior; PMH, premammillary hypothalamic nucleus; RC, retrochiasmatic nucleus; V, ventral; VMH, ventromedial nucleus; SCN, suprachiasmatic nucleus.

In this Review, we summarize studies in model organisms that have identified mechanisms by which embryonic developmental programmes construct the tuberal hypothalamus, with a particular focus on the formation of hypothalamic neurons and tanycytes that regulate metabolic processes, and discuss how defects in their development can lead to metabolic disease.

Hypothalamic control of energy balance

Research into the hypothalamic control of energy balance focuses primarily on the ARCN, which harbours orexigenic neuropeptide Y–agouti-related protein (NPY–AgRP)-expressing neurons (NPY–AgRP neurons) and anorexigenic pro-opiomelanocortin (POMC)-expressing neurons (POMC neurons). Activation of POMC neurons suppresses feeding, whereas activation of NPY–AgRP neurons stimulates feeding³. Both NPY–AgRP and POMC neurons respond to leptin (a circulating adiposity signal) and ghrelin (released under conditions of negative energy balance). POMC and NPY–AgRP neurons project to numerous hypothalamic nuclei, in particular to the PVN (Fig. 1b), where α-melanocyte-stimulating hormone (a proteolytic cleavage product of the POMC proprotein) activates and NPY–AgRP antagonizes melanocortin 3 and melanocortin 4 receptors. AgRP neurons in particular also have widespread extrahypothalamic projections to feeding-regulating structures in the brainstem, such as the parabrachial nucleus of the pons, and also form transient projections to the midbrain ventral tegmental area^4,5.

Increasingly, studies are pointing additionally to a role for tanycytes in the regulation of energy homeostasis, in which they might act through multiple mechanisms. First, tanycytes regulate brain–body interactions by controlling neurohormone release from the endfeet of neurons that project to the median eminence, including those of the PVN and ARCN^6,7 (Fig. 1b). The PVN has a principal role in metabolic regulation, acting as a crucial integrating centre to coordinate physiological responses to energy challenges⁸. Second, tanycytes sense circulating peripheral signals such as glucose and thyroid hormone and have been reported to actively transport leptin and ghrelin across the blood–brain barrier^9,10. However, the physiological relevance of this reported function remains controversial owing to the relatively mild phenotype observed following selective tanycyte ablation¹¹ and leptin receptor disruption¹². Third, some tanycytes also serve as hypothalamic stem or progenitor cells^13,14, a role in which they have been implicated in the de novo neurogenesis of ARCN neurons and the adaptation to changing physiological situations across the lifespan. In the pituitary stalk and posterior pituitary, glial pituicytes similarly influence neurosecretion. Neurons from the PVN and supraoptic nucleus, producing oxytocin and arginine vasopressin, project axons to the posterior pituitary (Fig. 1b), where pituicytes regulate the release of these neurohormones into the portal system^15,16.

Classical studies suggested the existence of relatively few distinct hypothalamic neuronal subtypes within each nucleus, considered hypothalamic astroglia to be homogeneous, and classified only four tanycyte subsets. However, more recent large-scale single-cell RNA sequencing (scRNA-seq) analyses of the adult mouse hypothalamus and specific tuberal nuclei have uncovered considerable heterogeneity among neurons that co-express neurotransmitters and/or neuropeptides^17–22 and have identified new tuberal neuron subsets that are potentially crucial for metabolic functions²³. Such studies also show that some tuberal hypothalamic nuclei contain neurons transcriptionally akin to those in neighbouring brain areas. For instance, the DMH and lateral hypothalamic area contain GABAergic neurons that molecularly resemble cells found in the zona incerta and reticular nucleus²⁴. Likewise, scRNA-seq analyses of the adult mouse suggest that both astroglia²⁵ and tanycytes²⁶ are also far more heterogeneous than previously assumed. Development studies have begun to provide a better understanding of how such heterogeneity might arise.

Developmental origins of the hypothalamus

Hypothalamic floorplate-like cells

The hypothalamus is one of the earliest parts of the central nervous system (CNS) to be specified, with its development beginning in the neural plate-stage embryo. Decades of study have shown that the hypothalamus develops from a subpopulation of cells in the midline of the neural plate. The vast majority of neural plate midline cells differentiate into floorplate cells that occupy the ventral midline of the neural tube and are instrumental in patterning and axon guidance²⁷. However, scRNA-seq and multiplex in situ hybridization studies in the embryonic chick have revealed two distinct populations of midline cells at the neural plate stage: anterior and posterior subtypes^28,29. Both populations express characteristic floorplate markers, namely, SHH, FOXA1 and FOXA2. The anterior subtype, however, additionally expresses NKX2.1, RAX and SIX3, which are indicative of the future tuberal hypothalamus, and DBX1, indicative of the future mammillary or supramammillary hypothalamus. The co-expression of floorplate and hypothalamic markers means that anterior midline cells of the neural plate are termed hypothalamic floorplate-like (HypFP) cells. As embryogenesis proceeds and the neural plate forms the neural tube, HypFP markers expand laterally^28,29, in part through the recruitment of neighbouring cells by way of growth factor-mediated induction³⁰. Lineage-tracing studies and tests of specification show that midline cells in the posterior neural plate give rise to floorplate cells that extend from the midbrain to the hindbrain^27,29. By contrast, midline cells in the anterior neural plate give rise to ventral midline cells of the diencephalon at neural tube stages, which in turn generate ventral-most tuberal and mammillary or supramammillary progenitors, whereas more lateral (later-recruited) HypFP cells generate more dorsal hypothalamic progenitors, including dorsal tuberal progenitors^{28,29,31–33}. In summary, the hypothalamus, including the tuberal region, develops from anterior midline HypFP cells and their immediate neighbours, reflected in its anterior-ventral position in the forebrain at neurogenic stages (Fig. 1c,d). Although these developmental stages are challenging to study in mice, expression profiling and lineage-tracing studies indicate that, as in chicks, HypFP cells form the anterior neural plate midline and give rise to cells that include ventral tuberal hypothalamic cells^34–36.

HypFP cells originate from diencephalic prethalamic-like progenitors

Multiple lines of evidence suggest that HypFP cells are specified from prethalamic-like diencephalic progenitors. First, midline HypFP cells, and then their recruited neighbours, are largely surrounded by diencephalic prethalamic-like cells, and indeed briefly co-express prethalamic markers such as PAX6, ARX and SP8; only a small fraction of the anterior-most HypFP cells directly contact telencephalic progenitors (expressing FOXG1)^28,29. Second, analysis of chick scRNA-seq using RNA velocity has identified a pseudotime cell lineage trajectory from diencephalic prethalamic-like progenitor cells to HypFP cells²⁸. Finally, cell lineage-tracing studies demonstrate that, although HypFP cells can emerge from relatively posteriorly located diencephalic cells, no common lineage relationship exists between hypothalamic and telencephalic cells^29,33. In summary, the weight of evidence indicates that HypFP cells predominantly originate from diencephalic prethalamic-like progenitors. This evidence contradicts the prosomeric model, which posits hypothalamic development from non-diencephalic tissue within the anterior-most neural plate^37,38.

Developing HypFP cells are underlain by prechordal mesendoderm, the anterior-most population of axial mesendoderm cells^39,40. Genetic and pharmacological experiments show that HypFP cells are specified under the influence of Sonic hedgehog (SHH) and WNT antagonists^{27,39,41–44}. These factors are expressed in numerous head tissues, but a wealth of evidence indicates that prechordal mesendoderm provides a critical source of both SHH and WNT antagonists to promote the induction of HypFP cells from prethalamic-like cells^39,40. Classically, SHH is considered a ventralizing factor⁴¹ and WNT antagonists are thought of as anteriorizing factors^45–47. In the context of HypFP cell induction, however, SHH and WNT antagonists integrate through a gene regulatory network (GRN) to promote the development of anterior-ventral HypFP from posterior-dorsal prethalamic-like cells (Fig. 2a). Central to the GRN are activated Gli genes (GliA; activators of the canonical SHH signalling pathway)⁴¹ and the transcription factors SIX3, RAX and NKX2.1^47–53. Interdependent positive regulatory loops maintain SHH, NKX2.1, SIX3 and RAX in HypFP cells. Additional signals, including members of the TGFβ superfamily, integrate into the GRN. Best known for its role in specifying the prechordal mesendoderm, the TGFβ family member Nodal has a direct role in HypFP specification (at least in chick and zebrafish) and upregulates SHH expression^27,54–56. BMP ligands, such as BMP2 and BMP7, also work together with SHH to promote the expression of NKX2.1 (refs. 29,31,32). Research in chick suggests a potential mechanism for this action: the BMP antagonist follistatin (FST) is expressed in prethalamic-like cells, where it inhibits hypothalamic development; FST is best known as a potent antagonist of BMP signals but, in the hypothalamus, BMPs might also suppress FST expression, thus promoting hypothalamic specification²⁸. Genetic analysis in mice further shows that Nkx2.1 promotes hypothalamic identity while simultaneously repressing prethalamic identity^19,52. These studies suggest a model in which BMPs inhibit FST, thereby derepressing NKX2.1 and enabling HypFP cell specification. Together, the GRN integrates signals from both SHH and WNT antagonists to promote HypFP cell identity. As discussed below, adaptations to the GRN then promote particular hypothalamic progenitor programmes, including those of tuberal progenitors.

Fig. 2 |. Gene regulatory networks underlying HypFP and tuberal progenitor specification.

The key interdependent regulatory loops that specify and maintain hypothalamic floorplate-like (HypFP) cells (part a) and tuberal progenitors (part b). BMP, bone morphogenetic protein; FST, follistatin; GliA, activated Gli genes; SHH, Sonic hedgehog. Pale grey arrows in b denote components of the gene regulatory network that may primarily have a role in HypFP specification.

Tuberal progenitor development

SHH and BMP signalling

Studies in the chick indicate that HypFP cells generate tuberal progenitors through a spatiotemporal patterning mechanism involving SHH and BMP signalling. Simultaneous activation of SHH and BMP signalling (via phosphorylation of SMAD1–SMAD5–SMAD8) in anteriormost HypFP cells initiates the tuberal programme²⁹. The phosphorylation of SMAD1–SMAD5–SMAD8 then spreads sequentially through HypFP cells from anterior to posterior, discernible as a travelling wave of BMP signalling. The wave ends within HypFP cells, and posteriormost HypFP cells without phosphorylated SMAD1–SMAD5–SMAD8 become mammillary or supramammillary progenitors²⁹. Therefore, a stream of tuberal progenitors develops over time in the anterior-ventral neural tube, in a conveyor-belt-like mechanism, as the wave of BMP signalling sweeps through HypFP cells. The tuberal progenitors generated first are distinguished by expression of the transcription factors SIX6 and TBX2 and the signalling factor FGF10, expressed alongside a subset of HypFP markers (SHH, NKX2.1, SIX3 and RAX)^28–30,33 (Fig. 2b). Better known for their role in eye development⁵⁷, the expression of SIX6 and TBX2, together with genetic loss-of-function studies indicate strong similarities between the genetic programme governing tuberal hypothalamus and ventral eye development²⁸. Tuberal progenitors expand markedly as they are generated, creating a substantial separation between the telencephalon and the mammillary or supramammillary hypothalamus.

Neurogenic versus gliogenic fates

Lineage tracing reveals a precise correlation between the position of a HypFP cell along the anteroposterior axis and the location of neurogenic and gliogenic tuberal progenitors. HypFP cells that are relatively anterior and are the first to activate BMP signalling give rise to tuberal neurogenic progenitors. Pseudotime analysis of scRNA-seq datasets reveals a trajectory from SIX6–SHH–FGF10⁺ progenitors to SIX6–ATOH7–ISL1–ASCL1⁺ progenitors, and eventually to ISL1–POMC–NPY neurons, characteristic of the ARCN²⁸. In tuberal neurogenic progenitors, BMP signalling is required only transiently: loss of function of BMP or BMP signalling eliminates almost all SIX6–ISL1⁺ neurogenic progenitors^29,30 but ectopic BMP or BMP signalling also antagonizes their development²⁹. This seeming contradiction is because BMPs suppress SHH^29,30, and pharmacological studies show that sustained SHH signalling is required for tuberal neurogenic progenitor development³³. Together, these studies indicate that, in the chick, the vast majority of tuberal neurogenic progenitors require the transient activation of canonical BMP signalling and the sustained activation of SHH signalling. Notch signalling is also transiently upregulated and required in nascent tuberal neurogenic cells (see below).

HypFP cells that are relatively posterior, and thus activate BMP signalling at relatively later stages, give rise to gliogenic progenitors²⁹. This programme reflects a switch in the balance of SHH and BMP signalling: in posterior HypFP cells, phosphorylation of SMAD1–SMAD5–SMAD8 is sustained, and promotes the sustained expression of TBX2, TBX3 and FGF10, while repressing SHH and leading to relatively weak expression of RAX, NKX2.1 and SIX6 (refs. 28–30). Over time, the expression of TBX2, TBX3 and FGF10 becomes confined to the infundibulum — an embryonic structure that forms the posterior pituitary and contributes to the median eminence. FGF family members are essential for infundibular cell expansion⁵⁸. A scRNA-seq analysis revealed that, as posterior tuberal progenitors develop into the infundibulum, they upregulate Notch signalling components and genes associated with radial glia and gliogenic progenitors, including genes characteristic of multipotent neural stem cells²⁹. The molecular changes in posterior tuberal progenitor cells probably orchestrate the transition from neuroepithelial cells to neurogenic radial glia, which subsequently generate quiescent radial stem-like cells, tanycytes and pituicytes.

In summary, studies in chicks show that the developmental stage at which HypFP cells are exposed to BMP signalling dictates their future identity. Early-specified HypFP cells that only briefly activate phosphorylated SMAD1–SMAD5–SMAD8 transition into SIX6–ASCL1–ISL1⁺ anterior tuberal neurogenic progenitors, whereas late-specified HypFP cells that show more sustained activation of phosphorylated SMAD1–SMAD5–SMAD8 become posterior tuberal gliogenic progenitors. Thus, the tuberal hypothalamus, along with its neurons and glia, is constructed spatially from anterior to posterior, reflecting the stage and duration of BMP signalling exposure.

Whether this mechanism is conserved across species remains unknown as the fate-mapping studies and temporal interference assays required to illustrate this dynamic process pose greater challenges in species other than chick. Most studies in mice and zebrafish emphasize the critical role of SHH signalling in specifying tuberal neurogenic progenitors^{35,36,50,51,59,60}. Furthermore, studies in mice and zebrafish show that many components of the HypFP GRN, including Shh, Nkx2.1, Rax and active SHH signalling, continue to be expressed in newly generated tuberal progenitors (Fig. 2b), their expression being critical to the tuberal neurogenic differentiation programme^{19,35,50,51,60}. Importantly, though, genetic analyses confirm that a fine balance between SHH and BMP signalling guides the development of tuberal neurogenic progenitors in mice⁵⁹. Further research is necessary to clarify whether a conserved GRN drives infundibular development and the differentiation of various tanycyte and pituicyte subsets. As in chick, an antagonism between BMP and SHH has a role in infundibulum specification in mouse⁶¹, as does Notch signalling⁶². In mice, the transcription factor LHX2 is also necessary for infundibulum formation during embryonic development⁶³. However, the precise mechanism by which these diverse signalling pathways and transcription factors regulate infundibular specification is not fully understood.

Transcription factor expression

Through the combination of positional cues and temporal growth patterns, spatial domains of transcription factor expression begin to delineate broad progenitor domains within the tuberal hypothalamus along the anterior–posterior and dorsal–ventral axes by phylogenetic stages (Fig. 3a). NKX2.1, NKX2.4 and RAX are broadly expressed at high levels in neural progenitor cells that later give rise to the ARCN as well as to the VMH, PMH, and parts of both the DMH and lateral hypothalamic area; weaker levels are detected in the future mammillary body^19,28,35. Directly dorsal to this domain, transcription factors, such as DLX1, DLX2, NKX2.2, ARX and OLIG2, shared with prethalamic-like neural progenitor cells and with prethalamic-derived domains such as the intrahypothalamic diagonal (destined to give rise to GABAergic neurons), are expressed (Fig. 3b). SIX3 and SIX6 show more restricted expression within the nascent tuberal hypothalamus, with notably high expression in anterior-ventral tuberal progenitors and then specific expression in the ARCN^35,64,65. These broad spatial domains further subdivide based on transcription factor expression, demarcating neural progenitor cells that will give rise to specific tuberal nuclei. Examples include NR5A1 and PTF1A, which mark the VMH^66,67; TBX3, HMX2 and HMX3, which mark the ARCN and much of the PMH^68,69; and PRDM13, which marks a domain that gives rise to the ventral DMH⁷⁰ (Fig. 3b). Although the loss of function of most of these genes does not disrupt regional identity, TBX3 mutants not only lose expression of ARCN and PMH markers but also upregulate markers of the PVN and mammillary body⁷¹.

Fig. 3 |. Topology of the hypothalamus at neurogenic stages.

a, Side view of the major tuberal hypothalamic progenitor regions during neurogenic stages (embryonic day 12.5 mouse; Hamburger–Hamiliton stage (HH)18–HH20 chick) and their positions relative to adjacent structures. b, Expression of key transcription factors during neurogenesis is indicated. A, anterior; ARCN, arcuate nucleus; D, dorsal; FP, floorplate; ID, intrahypothalamic diagonal; MMN, mammillary nuclei; P, posterior; PMH, premammillary hypothalamus; PTh, prethalamus; PVN paraventricular nucleus; SMN, supramammillary nuclei; TT, tuberomammillary terminal; V, ventral; vAH, ventral anterior hypothalamus; VMH, ventromedial hypothalamic nucleus; ZLI, zona limitans intrathalamica.

Control of neurogenesis

Role of signalling factors

Tuberal neurogenesis has been studied most extensively in mice, where it begins around embryonic day 10 (E10)⁷² and broadly proceeds in a sequential outside–inside pattern, peaking around E12–E13 and declining dramatically by E15. The most laterally located neurons are generated first, and neurons in nuclei close to the third ventricle, such as the ARCN, are generated last^22,35,72. Multiple extrinsic signals modulate this process. Canonical WNT signalling stimulates neural progenitor proliferation at the expense of neurogenesis⁴⁸ but is essential for neurogenesis in the PMH⁷³. Sustained SHH signalling also enhances both regionalization and neurogenesis in the VMH and ARCN⁵⁹. As is the case elsewhere in the CNS, however, Notch signalling has perhaps the most pervasive and complex role in regulating tuberal neurogenesis. Active Notch signalling is first detected in the tuberal neuroepithelium around E9.5 (ref. 35). Selective disruption of Notch signalling in the Nkx2.1-expressing ventral tuberal hypothalamus, by deleting the Notch effector transcription factor Rbpj, leads to the early and increased generation of POMC, NPY and GNRH-expressing ARCN neurons, while depleting proliferative neural progenitors⁷⁴. Rbpj loss of function in Nkx2.1-expressing progenitors also leads to a selective loss of kisspeptin neurons⁷⁵. Conversely, constitutive activation of Notch signalling in these same cells broadly inhibits neurogenesis, yet maintains neural progenitors through E18.5, long after developmental neurogenesis has essentially ceased^74,75. This regulatory role of Notch in balancing proliferative tuberal progenitors and differentiating tuberal neurons is conserved in chick^76–78.

Notch signalling is thus crucial for initiating neurogenesis, and subsequently has a vital and dynamic role in determining whether a progenitor will continue to divide symmetrically or generate neurons. The principles are the same as in other CNS regions, in which high levels of Notch signalling inhibit neurogenesis by induction of HES family transcription factors, whereas low levels promote neurogenesis by induction of proneural transcription factors^79,80. A scRNA-seq analysis revealed that the onset of tuberal neurogenesis coincides with a separation of two distinct progenitor subtypes¹⁹. Primary progenitors express relatively high levels of Notch pathway genes and genes characteristic of radial glial-like neural progenitor cells such as Vim and Slc1a. By contrast, neurogenic progenitors are distinguished by relatively lower expression of Notch pathway genes and higher expression of proneural basic helix–loop–helix factor genes, including Hes6, Ascl1, Neurog2 and Neurog3 (ref. 19). Ascl1 is expressed in most neurogenic tuberal progenitors, whereas Neurog2 is predominantly expressed in mammillary and anterior hypothalamic domains^19,35. Neurogenic VMH progenitors are an exception, with Neurog2 required for neurogenesis in this region⁸¹. Neurog3 is also expressed in a subset of ARCN neurogenic progenitors and is required for the development of AgRP and POMC neurons⁸². Ascl1 and Neurog3 act in parallel in the ARCN to specify glutamatergic and repress GABAergic cell identity⁸¹.

Generation of specific cell types

Once neurogenesis is initiated, the combinatorial interactions among transcription factors in both primary and neurogenic progenitors control the generation of a vast diversity of hypothalamic cell subtypes. Estimates suggest that there are ~1,000 molecularly distinct cell types in the adult mouse hypothalamus, most of which are present in the tuberal hypothalamus²⁴. Combinatorial patterns of transcription factors are the most reliable method for distinguishing each cell type²⁴. However, deciphering the role of specific transcription factors in specifying these cells is inherently limited without the ability to simultaneously profile the markers of all relevant tuberal cell types. Studies examining changes in neurotransmitter and neuropeptide expression in individual transcription factor mutants have provided some insight into their roles in tuberal hypothalamic cell fate specification, particularly in the ARCN^83–85. This process is part of identifying a hierarchical transcriptional network that both positively and negatively specifies ARCN cell identity (Fig. 4a).

Fig. 4 |. Molecular pathways underlying ARCN neuronal and tuberal tanycyte differentiation.

a, The sequential activity of transcription factors that direct tuberal progenitors to arcuate nucleus (ARCN) neurogenic precursors, to immature ARCN neurons, and then to different sets of mature ARCN neurons. b, The sequential activity of transcription factors that direct tuberal progenitors to radial glial cells and subsequently to tanycytes. AgRP, agouti-related protein; DA, dopaminergic; GHRH, growth hormone-releasing hormone; KISS, kisspeptin; NPY, neuropeptide Y; POMC, pro-opiomelanocortin.

Neural subtype specification.

In the developing ARCN, the neuropeptide POMC is initially expressed broadly in nearly all immature neurogenic precursors, then progressively repressed in all but a subset of primarily glutamatergic neurons. These neurons continue to express POMC into adulthood and have a central role in repressing feeding behaviours^3,86,87. The initiation, and possibly also maintenance, of Pomc expression requires expression of Nkx2.1 and Tbx3, which are broadly expressed in primary neural progenitors in the ARCN and most other tuberal regions^88,89. Additionally, Isl1, expressed in tuberal neurogenic progenitors and postmitotic neurogenic precursors, also has a role in regulating Pomc expression^85,90. However, sustained expression of POMC requires Prdm12, selectively expressed in a subset of postmitotic neurogenic precursors^83,87,91, and to a lesser extent, Nhlh2, which is more broadly expressed in ARCN neurogenic precursors^92,93. Micro-RNAs also promote terminal differentiation of POMC neurons as loss of function of both the miRNA processing enzyme Dicer and miR107 leads to a reduction in the number of POMC neurons and an increase in NPY–AgRP neurons⁹⁴.

Differentiation of various other ARCN neuronal subtypes involves the upregulation of specific transcription factors in POMC-expressing cells, generally dependent on Nkx2.1, Isl1 and Tbx3 (refs. 71,85,88,89). The specification of orexigenic NPY–AgRP-expressing GABAergic neurons is mediated by Dbx1 and Bsx^95,96, whereas somatotrophic GHRH neurons require Gsx1, Foxp2, Hmx2 and Hmx3 (refs. 68,97,98). Furthermore, Otp simultaneously promotes the specification of NPY–AgRP neurons while repressing GHRH identity, contrasting with Dlx1 and Dlx2, which have the opposite function⁸⁴. Dlx1 and Satb2 in turn are required for the generation of dopaminergic ARCN neurons^99,100. The specification of KISS⁺ neurons requires Sox14, Ptf1a, Nr5a2 and Prdm13 (refs. 67,98,101,102), with the maintenance of KISS expression relying on the sustained activity of Tbx3 (ref. 71). Although all major ARCN neuronal subtypes seem to be generated simultaneously²², the mechanisms underlying the diversification from an initially homogenous population of POMC-expressing neuronal precursors to distinct subtypes remain unclear.

Tanycyte specification.

Beginning around E13, RAX-expressing radial glial cells along the ventral third of the third ventricle, close to the developing ARCN, start to exit the cell cycle and begin differentiating into hypothalamic tanycytes (Fig. 4b). The majority of tanycytes are generated by E15.5, before the end of developmental neurogenesis^2,103,104. Mature tanycytes robustly express a range of hypothalamic progenitor markers, including Rax, Lhx2, Nfia, Nfiab, Nfiax, Sox2, Sox8 and Sox9, and exhibit high levels of Notch signalling¹³. Their transcriptional profile closely resembles that of retinal Muller glia, which also retains radial morphology¹⁰⁵. Loss of function of both Rax and Lhx2 leads to a loss of expression of many tanycyte-enriched genes and the ectopic expression of markers of multi-ciliated ependymal cells (which normally line more dorsal parts of the third ventricle) such as Foxj1 and Rarres^63,106. Lhx2-deficient tanycytes become multi-ciliated yet retain their radial morphology⁶³. Furthermore, loss of function in Tbx3 and Nfia disrupts tanycyte specification, though the fate of mutant tanycytes is unclear^20,71. Additionally, loss of function in NrCAM, selectively expressed by differentiating tanycytes, leads to defects in mature tanycyte morphology¹⁰⁷, and loss of function in Sema7a results in impairments in physiological state-dependent tanycyte plasticity¹⁰⁸.

Tanycytes retain a limited ability for proliferation and can generate both neurons and glia at least until postnatal day 14 (P14) in unstimulated mice¹⁰⁹. Importantly, feeding a high-fat diet to juvenile or young adult female mice stimulates tanycyte-derived neurogenesis, with tanycyte-derived cells modulating neural circuitry controlling body weight^13,110. β-Tanycytes, located ventrally, seem to be relatively more proliferative than other tanycyte subtypes, while the more dorsally located α2 tanycytes have greater neurogenic competence^13,14. Wild-type tanycytes primarily generate astrocytes postnatally, along with limited numbers of oligodendrocyte progenitors and ependymal cells¹⁰⁹. FGF2 enhances tanycyte proliferation, while FGF10 represses it^14,111,112. The NFI factors Nfia, Nfib and Nfix actively repress both proliferation and neurogenesis in postnatal tanycytes and are essential for tanycyte-derived gliogenesis¹⁰⁹. In mice deficient in Nfia, Nfiab and Nfiax, hypothalamic neurons increase more than 20-fold in number, yet both these and wild-type mice generate essentially identical neuronal subtypes. These neuronal subtypes are somewhat biased towards GABAergic neurons but include a broad array of molecularly distinct neuronal subtypes found in the ARCN, VMH and DMH. Tanycyte-derived neurons mature, integrate into hypothalamic circuitry and respond appropriately to changes in internal physiological states, with specific subsets responding to either leptin or heat stress¹⁰⁹, thus potentially allowing the organism to adapt to changing physiological metabolic demands over its lifespan.

Finally, a subpopulation of Rax⁻Irx3⁺Irx5⁺ neurogenic tanycytes has been identified in the postnatal hypothalamus^113,114, though it remains unclear whether they generate different progeny from the more abundant Rax⁺ tanycytes.

Development of neuronal circuitry in tuberal hypothalamus

Hypothalamic axonal outgrowth begins within a few days after neurons exit the cell cycle, with projections to the infundibulum (which goes on to form the median eminence and posterior pituitary) first detectable in mid–late embryogenesis in both rodents and chick^115,116. Arginine vasopressin axons that project to the infundibulum are guided there by a chemotropic action of FGFs and a chemorepulsive action of SHH^116,117. Intrahypothalamic connectivity progresses through the middle of the third postnatal week, with most projections first clearly observed during the second postnatal week¹¹⁸. The best studied of all intrahypothalamic circuits is the axonal projection from POMC neurons to the PVN, which is crucial for the regulation of feeding and metabolism¹¹⁹. This projection is modulated by reciprocal excitatory projections from the PVN to POMC neurons¹²⁰. POMC axons, which innervate the PVN between P6 and P12 in mice¹²¹, are attracted by SEMA3A released from target cells, detected by neuropilin 1 and neuropilin 2 (ref. 122). Contactin-dependent cell adhesion is necessary to stabilize these nascent connections, as are both autophagy¹²³ and the presence of primary cilia¹²⁴. The formation of POMC projections to the PVN coincides with a surge in circulating leptin levels and the transition from nursing to independent feeding¹²⁵. Disruption of leptin signalling — whether induced genetically, by neonatal overfeeding or maternal diabetes — impairs the formation of this circuit^121,126,127. Administration of exogenous leptin can rescue these defects if administered before P28 to leptin receptor-deficient ob/ob mice¹²⁸. After P28, however, the effects of leptin might be blunted partly by the formation of perineuronal nets, which are specialized chondroitin sulfate proteoglycan-rich extracellular matrix structures that form at the border of the ARCN and median eminence and might constrain binding by circulating leptin¹²⁹. Axonal projections of AgRP neurons to the PVN develop during this same period and are also leptin dependent^130,131. The molecular mechanisms regulating the formation of extrahypothalamic projections of POMC and AgRP neurons remain uncharacterized.

The development of synaptic connectivity among tuberal hypothalamic neurons is poorly studied compared to structures such as the cerebral cortex and hippocampus. Generally, fast synaptic connections are considerably less common in almost all hypothalamic regions studied thus far than in the cortex or hippocampus^132–135. Instead, neuropeptide signalling assumes a more significant role in regulating neuronal activity in most adult hypothalamic regions^132–135. A notable exception is the extensive inhibitory fast synaptic connections between NPY–AgRP neurons and POMC in the ARCN as well as the reciprocal connections between POMC neurons and the PVN. There is a progressive increase in PVN-derived glutamatergic input onto POMC neurons from P6 to P22, which is dependent on Efnb1 and Efnb2 (ref. 120). Additionally, connections from NPY–AgRP to POMC neurons increase as animals progress to middle age; these connections are also enhanced by high-fat diets, although the functional significance of this change remains unclear¹³⁶. Finally, a progressive increase in GABAergic input onto NPY neurons is observed from the early postnatal period through adulthood, whereas GABAergic input onto POMC neurons is mature by P13 (ref. 137). Despite these observations, the precise origins of these synaptic inputs are still not well understood.

Sex steroids and tuberal hypothalamic development

Neurons of the tuberal hypothalamus — particularly the VMH — regulate sexually dimorphic patterns of behaviour, including mating, aggression and overall activity levels^138–140. Gonadal steroids are essential for the development of sexually dimorphic hypothalamic circuitry, with sex-specific differences first emerging in the first two postnatal weeks in mice following a neonatal surge in gonadal steroid production¹⁴¹ and fully maturing after puberty, which typically begins around P26 in females and P30 in males¹⁴². Sexually dimorphic tuberal neurons selectively express the androgen (Ar) and oestrogen (Esr1 and Esr2) receptors along with the aromatase Cyp19a1, which synthesizes oestrogen from testosterone^143,144. A smaller subset of neurons in the ventrolateral VMH express the progesterone receptor (Pr)¹³⁸. Marked sex-specific differences are also observed in the number of cells that express gonadal steroid receptors. A scRNA-seq analysis has identified several neuronal subtypes in the VMH, ARCN and ventral premammillary hypothalamus that are strongly sexually dimorphic in their distribution²⁴. Although the specific function of most of these cell types has yet to be investigated, ESR1–CCKAR-expressing cells in the ventrolateral VMH have a critical role in driving sex-specific patterns of mating and sex recognition, and project to postsynaptic targets that are distinct from other VMH neurons¹⁴⁵. Although the precise mechanism of action of sex steroid receptors within the tuberal hypothalamus has yet to be investigated, an analysis of direct genomic targets of ESR1 in the nearby bed nucleus of the stria terminalis identified >2,000 target sites in neonatal mice and 18,000 sites in adults that were specific to females, almost all of which were oestradiol dependent¹⁴⁶. Sex-specific neuronal subtypes were typically distinguished by selective expression of specific transcription factors, with Esr1–Nfix⁺ cells being twice as abundant in males compared to females; ESR1 and NFIX showed direct transcriptional cooperation in these cells¹⁴⁶. Although gonadal steroids stimulate axonogenesis and induce synaptic remodelling within the VMH^147,148, sex-specific differences in response to oestradiol seem to be directly determined, at least in part, by sex chromosomal genes¹⁴⁹.

Tuberal development and disease

Genetic factors can disrupt any stage of tuberal hypothalamic development, with the most severe congenital defects affecting early hypothalamic patterning, specification and growth. These defects can lead to holoprosencephaly, marked by a severe reduction in hypothalamic size and often resulting in fetal or neonatal lethality^42,150,151 (Fig. 5). The genes mutated in this condition fall into two functional categories. The first includes prechordal mesendoderm and HypFP-derived signals and their direct effectors, such as NODAL, SHH and PTCH1, which promote and reinforce HypFP identity. Some of these factors also later regulate the axon guidance of tuberal neurons (Fig. 5). The second consists of transcription factors that initially specify anterior and tuberal hypothalamic identity, including SIX3, ZIC2 and TGIF^42,150,151. Prader–Willi syndrome, caused by the loss of the paternally expressed copy of a cluster of maternally imprinted genes located at chromosome 15q11.2-q13, shows features consistent with hypothalamic disruption, including excessive appetite, severe early-onset obesity, type 2 diabetes, sleep disorders and congenital hypogonadism¹⁵². The Prader–Willi syndrome locus encodes multiple small nucleolar RNAs regulating ribosomal RNA processing and five protein-coding genes, including the transcriptional coregulator MAGEL2, which is broadly expressed in the developing hypothalamus. Selective disruption of MAGEL2 leads to a reduction in POMC–PVN axonal connections (AgRP projections are unaffected)¹⁵³, while also disrupting the response of POMC neurons to leptin¹⁵⁴.

Fig. 5 |. Stage-specific insults to the developing hypothalamus and their role in metabolic disorders.

Genetic mutations and environmental insults at different stages of hypothalamic development contribute to an array of defects and metabolic disorders. ARCN, arcuate nucleus; FP, floorplate; HypFP, hypothalamic floorplate-like cell; PWS, Prader–Willi syndrome; SHH, Sonic hedgehog.

Loss of function of multiple transcription factors that mediate the specification of POMC and NPY–AgRP neurons in mice, including those encoded by Neurog3, Isl1, Prdm12 and Otp, each leads to early-onset obesity in mice^{83,90,155,156}. Mutations in several of these same genes are also linked to early-onset monogenic obesity in humans^156–159. Furthermore, mutations in genes that are known or suspected to control the formation of POMC–PVN connections, including SEMA3A, PLXNA1, PLXNA2, PLXNA3, PLXNA4, NRP1 and NRP2, are all also associated with early-onset obesity, exhibiting similar phenotypes in both mice and humans¹²². Similar phenotypes result from mutations in BDNF and TRK2, which disrupt neurotrophic signalling¹⁶⁰, although it remains unclear whether this results from defective expression in the tuberal hypothalamus. Mutations in genes controlling primary cilia formation frequently disrupt hypothalamic regulation of body weight and lead to early-onset obesity in disorders such as Bardet–Beidel and Alström syndromes¹⁶¹. Loss of function of Bbs1 in POMC and AgRP neurons is sufficient to induce obesity¹⁶² and disrupts periciliary targeting of the leptin receptor¹⁶³. However, it is not known whether this receptor mistargeting leads to obesity or whether there are other causes.

Epidemiological studies suggest that both restricted and excessive caloric intake during pregnancy can lead to long-term metabolic disorders in children^164,165. Evidence is also increasing that intrauterine and early-life experiences might disrupt multiple aspects of tuberal hypothalamic development, potentially leading to the development of metabolic disorders¹⁶⁶. In mice, maternal obesity leads to reduced levels of tuberal neurogenesis at both E13 and P0 (refs. 167,168), and induces complex changes in gene expression by P21, including reduced levels of neurotrophin signalling and increased expression of immature neuronal markers such as Dcx¹⁶⁹. Likewise, disruption of neonatal leptin signalling, induced by maternal obesity, diabetes or overfeeding, disrupts the formation of POMC–PVN projections in neonatal mice^121,126,127. Other developmental effects of overnutrition and undernutrition include changes in hypothalamic neuropeptide expression, insulin resistance and defective regulation of glucocorticoid signalling^163,170. Although direct evidence for similar effects in humans is generally lacking, studies have suggested that fetuses of mothers with obesity might become insulin-resistant in utero¹⁷¹ and that maternal obesity can trigger inflammation and associated reactive gliosis in the fetal mediobasal hypothalamus¹⁷².

In vitro approaches to study human hypothalamic development

Marked advances have been made in the development of in vitro approaches to study hypothalamic patterning and cell fate specification, utilizing directed differentiation of embryonic stem cells and induced pluripotent stem (iPS) cells. Fortuitously, minimal culture conditions of both mouse and human embryonic stem cells induce basal forebrain identity, characterized by the expression of early hypothalamic markers such as RAX and SIX3 (ref. 173). Enhanced SHH signalling and extended culture duration further promote tuberal hypothalamic identity, inducing the generation of both POMC neurons^174–178 and tanycyte-like cells¹⁷⁷. This has permitted analysis of POMC processing in human embryonic stem cell-derived neurons¹⁷⁵ and the generation of functional arcuate and pituitary assembloids¹⁷⁹, providing new ways to model human diseases that result from disrupted development. This approach has been applied to iPS cells derived from patients with obesity, which show altered expression of metabolic genes relative to iPS cells from healthy individuals as controls¹⁸⁰. Similarly, ARCN-like organoids derived from iPS cells of patients with Prader–Willi syndrome recapitulate some cellular defects seen in vivo, including reduced numbers of POMC neurons and increased levels of gliogenesis¹⁸¹.

Despite the promise of direct differentiation-based analysis, several technical hurdles remain. Although protocols for generating ARCN neurons are relatively efficient, the derivation of other cell types poses significant challenges. Precise titration of ventralizing signals, such as SHH, might be necessary to generate more dorsally located cell types. Additionally, the carefully timed application of both BMPs and Notch inhibitors might be required to produce different progenitor subsets and early-born cell types, respectively^29,74,75. Finally, a detailed scRNA-seq-based characterization of embryonic stem-derived or iPS cell-derived tuberal hypothalamic cell types has yet to be performed, and it remains unclear whether cells from these preparations closely resemble native hypothalamic cells or whether they contain cells resembling those of other forebrain regions.

Conclusions

The past decade has seen considerable progress in our understanding of the molecular mechanisms that control tuberal hypothalamic development and their relevance to metabolic disorders. Despite these advances, many fundamental questions regarding the regulatory mechanisms that control tuberal hypothalamic patterning and neurogenesis remain unresolved. Notably, how the patterning of the tuberal hypothalamus is coordinated with other forebrain structures (most notably the telencephalon and prethalamus) remains unclear. Increasing evidence challenges the accuracy of the widely accepted prosomere model in describing hypothalamic organization, calling for further experiments to definitively resolve this issue. Despite the availability of scRNA-seq data from the developing hypothalamus of multiple species, the extent to which similar molecular mechanisms drive temporal changes in proliferation and neurogenesis in hypothalamic progenitors — as seen in better-characterized populations such as those of the retina and cerebral cortex — remains unclear. Likewise, although it is well-documented that immature interneuron precursors undergo long-distance tangential migration and disperse widely throughout the telencephalon, very little is known about the migration of tuberal hypothalamic neuronal precursors. Finally, while the overall structure and morphology of the hypothalamus are relatively well-conserved across vertebrate model organisms, it is unclear whether changes in the patterning or neurogenesis of the tuberal hypothalamus could directly contribute to some of the dramatic metabolic and behavioural differences observed between species, such as between surface and cave tetra¹⁸² or solitary and social mole rats¹⁸³, to name just two examples.

The broader effects of changes in the intrauterine and early-life environment on tuberal hypothalamic development have been characterized to only a limited extent, and the application of advanced multi-omic technologies will undoubtedly help to determine exactly how physiological stressors, including both overnutrition and undernutrition, influence tuberal hypothalamic neurogenesis and neural circuit formation, and how these factors affect health later in life. These questions have only been studied in depth in POMC and AgRP neurons of the ARCN, and the topic is ripe for further exploration using single-cell and spatial transcriptomic analysis. Similar questions apply to the regulation of neurogenic competence in hypothalamic tanycytes. It remains unclear what restricts their ability to generate neurons in adults, which subtypes of tuberal hypothalamic neurons can be generated from tanycytes, and the extent to which changes in internal physiological states — particularly in conditions such as obesity and diabetes — might modulate tanycyte-derived neurogenesis. Insights gained from studying developmental processes might be applied to the generation of therapeutically relevant neuronal subtypes from tanycytes.

Finally, we are only beginning to uncover how structural and regulatory variation in genes that control tuberal hypothalamic development mediate both normal and pathological variation in the immense range of physiological processes and behaviours that are controlled by this critical brain region. A small number of rare and severe mutations have been identified but many more probably remain undiscovered, and the impact of more common genetic variants has not yet been explored. A critical part of these efforts will be the development of efficient and reproducible protocols for the directed differentiation of both tuberal hypothalamic organoids and specific cell types from human embryonic stem cells and iPS cells, in which the functional effects of these genetic variants can be studied.

Key points.

• Energy balance is regulated by pro-opiomelanocortin and neuropeptide Y neurons within the arcuate nucleus and by tanycytes within the median eminence.

• Sequential signalling events govern the progression of diencephalic prethalamic-like cells to generate regionally distinct populations of hypothalamic progenitors; sustained signalling events and hierarchical transcription factor networks mediate tuberal neurogenesis and the specification of tuberal neuronal subtypes and tanycytes.

• Leptin and a high-fat diet regulate diverse aspects of tuberal cell specification, including neurogenesis, axon guidance and synaptic connectivity.

• Neurons of the tuberal hypothalamus are sexually dimorphic in their distribution and regulate sexually dimorphic patterns of behaviour.

• Genetic and environmental factors disrupt tuberal hypothalamic development and lead to lifelong metabolic defects.

• Directed differentiation of human induced pluripotent stem cells towards hypothalamic identities is in its infancy but holds the promise of generating therapeutically important tuberal hypothalamic cell types.