Development of the basal hypothalamus through anisotropic growth

Abstract The adult hypothalamus is subdivided into distinct domains: pre‐optic, anterior, tuberal and mammillary. Each domain harbours an array of neurones that act together to regulate homeostasis. The embryonic origins and the development of hypothalamic neurones, however, remain enigmatic. Here, we summarise recent studies in model organisms that challenge current views of hypothalamic development, which traditionally have attempted to map adult domains to correspondingly located embryonic domains. Instead, new studies indicate that hypothalamic neurones arise from progenitor cells that undergo anisotropic growth, expanding to a greater extent than other progenitors, and grow in different dimensions. We describe in particular how a multipotent Shh / Fgf10‐expressing progenitor population gives rise to progenitors throughout the basal hypothalamus that grow anisotropically and sequentially: first, a subset displaced rostrally give rise to anterior‐ventral/tuberal neuronal progenitors; then a subset displaced caudally give rise to mammillary neuronal progenitors; and, finally, a subset(s) displaced ventrally give rise to tuberal infundibular glial progenitors. As this occurs, stable populations of Shh +ive and Fgf10 +ive progenitors form. We describe current understanding of the mechanisms that induce Shh +ive /Fgf10 +ive progenitors and begin to direct their differentiation to anterior‐ventral/tuberal neuronal progenitors, mammillary neuronal progenitors and tuberal infundibular progenitors. Taken together, these studies suggest a new model for hypothalamic development that we term the “anisotropic growth model”. We discuss the implications of the model for understanding the origins of adult hypothalamic neurones.

model organisms that challenge current views of hypothalamic development, which traditionally have attempted to map adult domains to correspondingly located embryonic domains. Instead, new studies indicate that hypothalamic neurones arise from progenitor cells that undergo anisotropic growth, expanding to a greater extent than other progenitors, and grow in different dimensions. We describe in particular how a multipotent Shh / Fgf10-expressing progenitor population gives rise to progenitors throughout the basal hypothalamus that grow anisotropically and sequentially: first, a subset displaced rostrally give rise to anterior-ventral/tuberal neuronal progenitors; then a subset displaced caudally give rise to mammillary neuronal progenitors; and, finally, a subset(s) displaced ventrally give rise to tuberal infundibular glial progenitors. As this occurs, stable populations of Shh +ive and Fgf10 +ive progenitors form. We describe current understanding of the mechanisms that induce Shh +ive /Fgf10 +ive progenitors and begin to direct their differentiation to anterior-ventral/tuberal neuronal progenitors, mammillary neuronal progenitors and tuberal infundibular progenitors.
Taken together, these studies suggest a new model for hypothalamic development that we term the "anisotropic growth model". We discuss the implications of the model for understanding the origins of adult hypothalamic neurones.

K E Y W O R D S
anisotropic growth, development, Fgf10, hypothalamus, prechordal mesendoderm, progenitor, sonic hedgehog compares these with ideal basic set-points for features such as hormone and metabolite levels, temperature and electrolyte balance, and then initiates feedback systems to restore optimal physiology. In addition, the hypothalamus mediates allostasis, that is, the ability to re-evaluate optimal set-points to anticipate the organism's changing environment. The adaptive responses of homeostasis and allostasis operate through autonomic, endocrine and behavioural systems and over different durations of time to maximise the chance of individual and species survival. In this way, hypothalamic cells enable the body to respond, anticipate and adapt to changing physiological conditions over life.
Classically, the adult hypothalamus is divided into four domains: pre-optic, anterior, tuberal and mammillary. Each domain harbours cell clusters, termed nuclei, and less well-defined territories, all arranged in a patchwork manner. Early reports, based on lesion studies, led to the idea that a particular nucleus, or territory, might centrally control a particular behaviour; however, sophisticated new approaches, including cell-specific and conditional knockouts, chemogenetic and optogenetic studies, now suggest that activation of a particular neurone increases the likelihood of an event or a behaviour, such that homeostasis of a particular physiological state is mediated by complex interactions of multiple nuclei/neurones. 2

| E ARLY MODEL S OF HYP OTHAL AMIC DE VELOPMENT
There is a pressing need to determine how particular hypothalamic neurones arise in life, to provide insight into the ability of the body to anticipate and adapt robustly, to provide insight into pathological conditions/dysfunctional behaviours such as chronic stress, reproductive and eating disorders, and to inform efforts to direct the differentiation of human pluripotent cells to hypothalamic neuronal fates, all studies with enormous potential for the evaluation of future novel therapies for conditions such as obesity. Many previous models of hypothalamic development have been proposed and two in particular have received much attention: the columnar model and the prosomeric/revised prosomeric model. Each suggests that adult domains, and their resident nuclei/territories/neurones, arise from correspondingly located embryonic domains that expand isotropically (ie, to the same extent). The columnar model suggests that the hypothalamus is a diencephalic-derived structure, with preoptic, anterior, tuberal and mammillary progenitor subsets arrayed in columns along the anterior-posterior (future rostro-caudal) axis, reflecting an early anterior-posterior regionalisation of the neural tube. 3 The prosomeric/revised prosomeric model 4 are derived from basal progenitors. Importantly, proponents of each model point out that these provide a useful starting point for probing the origins of hypothalamic neurones, but acknowledge the difficulties in ascribing adult neuronal populations to progenitor domains, not least because differentiating neurones may undergo complex migrations. 6,7 Each model was proposed before the advent of conditional knockout approaches, or sophisticated lineage-tracing studies, and so neither takes account of the extensive growth of the hypothalamus in the earliest stages of its development, or of the possibility that progenitor cells might migrate as they are specified.
Recent work in the embryonic chick, which examines the growth of a previously-undefined progenitor population, now suggests that progenitor displacement/migration is key to hypothalamic development, and suggests a fundamentally different model of hypothalamic development to those previously suggested. Here, we summarise these studies 8 and describe an "anisotropic growth model" of hypothalamic development.  [13][14][15][16][17] , that extend to the boundary with Foxg1, 8 (ie, the telencephalic boundary) 18 ( Figure 1A). As discussed below, RDVM cells play a critical role in subsequent steps in hypothalamic development.

| Induction of Shh +ive ventral midline cells
The PM expresses the secreted glycoprotein, Shh, and studies of isolated chick tissue explants reveal that Shh is required to induce Shh +ive RDVM cells. 13,15 Other factors, however, synergise with Shh to mediate this event, including the transforming growth factor β signalling ligand, Nodal, deriving from the PM 15,19,20 and the transcription factor (TF), Six3. 16 In Shh-null embryos, embryos haploinsufficient for Six3, or with dysfunctional Nodal signalling, RDVM cells are not induced and embryos develop holoprosencephalic phenotypes 16,21 .
The precise regulation and duration of Shh expression in the PM is crucial for RDVM cell induction. Loss of a single copy of Shh, or mutations that lead to reduced expression of Shh in the PM, result in holoprosencephaly. 22 The temporal perturbation of Shh signalling correlates with the severity of holoprosencephalic phenotypes: the earlier the alteration, the more severe the phenotype. 23,24 The tight temporal control of Shh in the PM is regulated by Nodal, which acts in a juxtacrine manner to control the duration of Shh expression. 25 Elegant analyses in mouse show that Shh expression in RDVM cells is regulated by a unique enhancer, SBE2 (Shh brain enhancer 2). 26 Once induced, Shh diffuses out of RDVM cells to establish a morphogen gradient in adjacent diencephalic cells that is translated into a cell-intrinsic GliA-GliR gradient, 27-29 similar to that found in the spinal cord. 30 The predicted GliA-GliR gradient is considered to set up

| bHyp cells are proliferating progenitors that give rise to the basal hypothalamus through anisotropic growth
As RDVM cells transit to bHyp cells, they undergo pronounced changes in cell cycle: first they undergo a transient arrest, then they become highly proliferative. thus, when first induced, bHyp progenitors directly abut Foxg1 + Foxg1 +ive telencephalic progenitors but then become displaced by their anterior-daughters and so, ultimately, Fgf10 +ive progenitors come to be located in the ventral tuberal hypothalamus (Figure 2).

| Molecular mechanisms of basal hypothalamic anisotropic growth
One outstanding question is whether the basal hypothalamus is gen-  neurones of the tract of the post-optic commissure. 1,8,18,[27][28][29]36,44,45 However, future lineage-tracing studies that build on previous/recent studies 8,27,29,51 are needed to tease out which neurones require Shh non-autonomously vs autonomously; namely, to distinguish between neurones born from progenitors that respond to Shh (but do not express it) vs neurones that differentiate from Shh +ive progenitor populations.

| Maintenance of a ventro-tuberal Fgf10 +ive progenitor pool
Throughout the generation of anterior neuroepithelial Shh +ive progenitors, a pool of Fgf10 +ive progenitors is established and maintained. 8 The mechanisms that select a steady supply of anterior neuroepithelial Shh +ive progenitors simultaneously form part of the mechanism that ensures a stable-size pool of Fgf10 +ive progenitors in a mechanism that may be conserved across species. Thus, genetic and pharmacological interventions suggest that Shh, deriving from anterior neuroepithelial progenitors, feeds back to regulate the size of the Fgf10 +ive pool 8,36 (note that Wnt may also be involved 39,40 ). Certainly, in other parts of the brain, differentiating cells feedback to progenitor cells to maintain their appropriate numbers and behaviours. 52 In chick, mouse and zebrafish, progenitor cells within the

| Mammillary progenitor generation and differentiation
In the short-term, the maintenance of pools of undifferentiated

| Infundibular progenitor generation and differentiation
Finally, having generated anterior and mammillary progenitors that extend in opposite directions, the Fgf10 +ive progenitor pool gives rise to another set(s) of progenitors: infundibular progenitors that grow ventrally. 41 The SoxB1 HMG-box transcription factor, Sox3, is likely to interact with Rx/Lhx2: in humans, either reduced or elevated dosage of SOX3 leads to infundibular hypoplasia. 57 Taken together, then, this sequence of growth leads to progenitor cells of basal anterior, tuberal and mammillary neurones arrayed around the ventro-tuberal infundibulum. The sequential anisotropic growth in three-dimensions from bHyp progenitor cells is peculiar and unprecedented within CNS development.

| ANISOTROPI C G ROW TH MODEL AND HYP OTHAL AMIC ORG ANISATION
The "anisotropic growth model" of hypothalamic development shows that, in the chick, different rudiments of the adult hypothalamus are established at different times, suggesting sequential progenitor programmes: an early programme that arises as progenitor cells are born in response to an early GliA-GliR gradient, followed by a later programme that arises as bHyp progenitor cells develop, and then itself has three temporally-sequential components: anterior, mammillary and then infundibular. Additionally, the model emphasises the importance of progenitor migration/displacement in establishing different hypothalamic domains, and shows that, in the chick, at least some neurones of the anterior basal hypothalamus are likely to be generated from Shh-expressing progenitor cells (anterior RDVM cells/anterior Shh +ive neuroepithelial progenitors). The movement of progenitor cells, whether passively or actively, is likely to occur to a significant extent as the hypothalamus develops over time. Although not examined in detail, alar progenitor cells also undergo extensive migration. 35 The migration of bHyp and alar progenitor cells explains the difficulty in matching early progenitors to adult neurones and nuclei.

ACK N OWLED G EM ENTS
We thank the anonymous reviewers for their helpful comments, as well as support provided by the Wellcome Trust [212247/Z/18/Z to MP].