Male SD rats, between 45 and 60 postnatal days, were provided by the animal facilities of the University of Murcia. SD rats were weighed and housed under the same conditions in standard size cages (50 cm × 35 cm × 35 cm with a 2–3 cm dry cork), shared by three animals, and replaced every 4–5 days. A 12:12 h light-dark cycle was set up and the brains were obtained during the active phase of the rats (dark cycle). The rooms were kept at a temperature of 22 to 25°C with a relative humidity of 45 to 60%. The animals had ad libitum access to a standard chow diet (ENVIGO, diet 2014, United States) and filtered water.

Sprague-Dawley brains were obtained and processed following the protocols of Ferran et al. (2015b,c). The brains were perfused with a saline solution followed by phosphate-buffered 4% paraformaldehyde (0.1 M PB; pH 7.4). Once extracted, the brains were maintained at 4°C for 24 h. Some brains were then washed in PBS, followed by 15 and 30% sucrose solutions in 0.1 M phosphate-buffed saline (PBS) solution (pH 7.4) and cut with a sliding microtome (Micron HM430, Thermo Scientific, United States) into sagittal, horizontal, and transversal sections (50 μm thick). Sections were collected as parallel or consecutive series on SuperFrost Plus slides (Menzel-Gläser, Braunschweig, Germany) and processed for hybridization and/or immunohistochemistry. A few other brains were washed with PBS and embedded in 4% agarose (low electroendosmosis-EEO agarose; catalog No. 8008; Pronadisa, Spain) to obtain vibratome sections (100 μm thickness) and processed as floating sections for immunohistochemistry (Ferran et al., 2015b,c).

Agrp, Cart, Mch, Npy, Oxt, Pomc, Sst, Th, and Trh cDNA fragments were obtained by RT-PCR and cloned into TA vectors to later synthesize the RNA probes. Fresh brain tissues from postnatal rats were sheared by 1 cycle of 20 s at 6500 RPM in 2 ml tubes (CK14) using the Precellys Evolution system (Bertin Technologies, France). The RNA was then extracted from the samples using the NZY total RNA isolation kit (Nzytech, MB13402, Portugal) and treated with DNaseI (Invitrogen, Cat. 18068-015, United States). Conversion to cDNA was performed with Superscript III reverse transcriptase (Invitrogen, Cat. 18080-044, Spain) and oligo dT anchored primers. The cDNA was used as a template for PCR using Taq polymerase (Promega, Cat. M8305, Spain) and specific primers (Supplementary Material 1). The resulting PCR products were then cloned into the pGEM-T Easy Vector (Promega, Cat. A1360, Spain) and sequenced by ACTI (University of Murcia, Spain).

Brain sections obtained from sliding microtome were collected in SuperFrost slides and processed according to in situ in cryosections as described in Ferran et al. (2015c). Digoxygenin-11-UTP (Roche, Lewes, United Kingdom) was used for the synthesis of labeled RNA sense and antisense riboprobes from cDNA samples of Agrp, Avp, Cart, Crh, Mch, Npy, Oxt, Pomc, Sst, Th, and Trh rat genes. Probe sequence information is provided in Supplementary Material 2. Linear cDNA templates were obtained by amplifying cloned fragments by PCR (Ferran et al., 2015b).

A detailed version of the immunohistochemical reaction has been described in previous work (Ferran et al., 2009, 2015c). Highlighting key steps, tissues were initially exposed to 0.3% hydrogen peroxide to inactivate endogenous peroxidases. The primary antibody was incubated overnight at 4°C (TH [NB300-109], Novusbio, 1:200 dilution, Bio-Techne R&D Systems, Spain). After the washes, the sections were incubated with secondary antibody for 2 h (Biotinylated anti-rabbit (H+L), Vector Laboratories, United States, 1:1 dilution). Next, a streptavidin-peroxidase complex (Vectastatin-ABC kit; Vector Laboratories, United States; 0.001% dilution) was applied for 1 h at room temperature. Finally, the sections were washed, and the peroxidase activity was developed with 0.03% 3,30-diaminobenzidine (Sigma, St. Louis, MO, United States), plus 0.003% hydrogen peroxidase. The specificity of the antibody was demonstrated by western blotting (see below) and by performing control experiments that omitted the primary antibody without observing residual immunostaining.

Fresh rat postnatal brain tissue from the mesencephalic-diencephalic region containing the tegmental ventral area (VTA) was sheared by 1 cycle of 20 s at 6500 RPM in 2 ml tubes (CK14) using the Precellys Evolution system (Bertin Technologies, France). The sheared tissue was resuspended in RIPA lysis buffer (Millipore, United States) and using protease inhibitors (Pierce Biotechnology Inc., United States) and PMSF 1M (Abcam, United Kingdom) for 30 min at 4°C. Protein concentration was initially determined by the Bradford protein system (Sigma-Aldrich, Germany), these were then separated in 10% SDS-polyacrylamide gel (PAGE-SDS, United Kingdom) to be transferred to a nitrocellulose membrane (Whatman, United Kingdom). As standard size in western-blot protein electrophoresis, PageRuler™ Prestained Protein Ladder (Thermo Scientific, United States) was used. After completion of the transfer, the protein bands were visualized by staining the nitrocellulose membranes with Ponceau S solution (Sigma-Aldrich, Germany). Blots were then incubated overnight at 4°C with rabbit antibodies against TH (1:200, Novusbio, Bio-Techne R&D Systems, Spain). The secondary antibody (anti-rabbit IgG (H+L), Vector Laboratories, United States) was used at 1:10.000 for HRP peroxidase (Thermo Fisher, United States). The enhanced chemiluminescence (ECL) Western blot detection system (Amersham Biosciences, United Kingdom) and LuminataTM Forte (Millipore Corporation, United States) was used to determine the immureativity of the TH antibody, using the ImageQuant LAS 500 Gel Documentation System (GE Healthcare, United States). The molecular weight obtained for TH was approximately 60 kDa (Supplementary Material 3).

Processed in situ and immunohistochemistry sections were digitalized with a ScanScope CS digital slide scanner (Aperio Technologies, Vista, CA, United States). Size, contrast, brightness, and focus in the images were adjusted by applying Adobe Photoshop CS3. Figures were produced using Adobe Illustrator CS2 (Adobe Systems Inc., San Jose, CA, United States).

Some tracts known to course along the dorsoventral or anteroposterior axes in the hypothalamic territory are visible in our material and thus help to recognize the hypothalamic structure. The main dorsoventral tracts connecting the telencephalon with the hypothalamus and other territories–the medial and lateral forebrain bundles–pass through the peduncular hypothalamus or PHy (Puelles et al., 2012a; Puelles and Rubenstein, 2015). The TH-negative fornix tract (fx) coming from the hippocampal region passes behind the anterior commissure (anterior wall of the interventricular foramen) and follows a strictly dorsoventral course in the PHy region just behind the intrahypothalamic boundary (radially its position approximates the limit between medial and lateral hypothalamic strata). The fornix ends contralaterally after giving collaterals to the mammillary body and thereafter crossing the retromamillary decussation (fx; Figures 1B,C, 3B and Supplementary Figure 6C). Dispersed throughout the peduncular lateral hypothalamic stratum there is the medial forebrain bundle (mfb); this complex tract contains many ascending or descending components, including the TH-positive axons from the ventral tegmental area (VTA; mesocortical pathway, mc) and substantia nigra pars compacta (SNC; nigrostriatal pathway, ns) (ns/mc; Figures 1B,C, 4–6, 8). The mfb tract is only present in the PHy, and thus helps to identify the latter’s curved rostral intrahypothalamic boundary with THy (ns/mc; Figure 4), as occurs likewise with the fornix. Superficial to the mfb and running dorsoventrally through the superficial TH-negative PHy stratum there is the cerebral peduncle (ped) or lateral forebrain bundle. This contains among other (basal ganglia) components corticothalamic, corticopontine, corticoreticular, corticonuclear, and corticospinal fibers (ped; Figures 1C, 6, 8).

Regarding the landmark longitudinal tracts, we can identify in our material the stria medullaris, the optic tract, and the mamillotegmental tract, as well as its collateral mamillothalamic tract. The stria medullaris tract (smed) arises mainly from a variety of septocommissural, diagonal, preoptic, and hypothalamic neurons (all possibly migrated from the prethalamic eminence; Alonso et al., 2020). Their axons fasciculate together in the dorsal PHy next to the prethalamic eminence, laterally to the fornix tract after the latter passes behind the anterior commissure. The compacted smed then runs anteroposteriorly through the whole prethalamic eminence and thereafter continues caudalwards next to the chorioidal taenia thalami until it incorporates into the habenular commissure, projecting all the while on the lateral and medial habenula (Figures 1B,D, 4A,B, 8; Puelles et al., 2012a). Several opportunistic TH-positive mfb fibers, either from the ns/mc tracts or from intrinsic hypothalamic TH neurons, send collaterals or swerve into the initial smed approximately at the preopto-hypothalamic boundary, thus causing the smed to appear TH-positive in our material (smed; black arrowheads; Figures 4A,B, 8). The TH-negative optic tract (ot) can be identified from suprachiasmatic levels (the acroterminal SPa alar region) running successively through the THy and PHy alar plate, to follow also longitudinally through the prethalamic, thalamic, pretectal, and mesencephalic alar plate to find its multiple targets (Figures 1B,D, 4, 5D, 6A, 7A,B, 11A; Supplementary Figures 8I–K; Puelles et al., 2012a; Puelles, 2022). The TH-negative mamillotegmental tract (mtg) is a longitudinal tract arising from the rostrobasal mamillary region (THy); it runs caudalwards across the retromamillary, diencephalic, midbrain, and isthmic tegmentum, and then ends primarily at the dorsal and ventral tegmental nuclei of rhombomere 1 (Figures 1B,D, 3A, 13, 14; Puelles et al., 2012a). The mtg produces collaterally the TH-negative mamillothalamic tract (mth) which runs ventrodorsally caudally to the zona limitans intrathalamica into the anterior nuclear complex of the thalamus (Figures 1B,D, 3A, 6B, 13, 14; Puelles et al., 2012a).

This region is constituted by two dorsoventrally arranged progenitor domains and adult subregions, the PA and SPa areas (Puelles et al., 2012a); both Pa and SPa stretch across PHy and THy (Figure 1B). The alar Pa domain limits dorsally with the telencephalic territory and ventrally with the SPa alar domain; caudally, both limit with the diencephalic prethalamus (Figure 1B); the SPa limits ventrally with the alar-basal boundary (Figure 1B). The PA hypothalamic nucleus that develops within Pa is subdivided dorsoventrally into dorsal, central, and ventral parts (DPa, CPa, VPa; Figure 1B). We will also distinguish in our description the periventricular stratum (pe), with cells adjacent to the ependym, from the conventional thicker medial hypothalamic stratum (mh), occupied by the Pa nuclear complex, found deep to the lateral and superficial hypothalamic strata (Puelles et al., 2012a; Ferran et al., 2015a; Puelles and Rubenstein, 2015; Díaz and Puelles, 2020). The SPa area is best developed within THy, where it forms the suprachiasmatic nucleus (held to lie within ATerm) as well as the anterior hypothalamic nucleus (SCH; AH; Figure 1B), whereas the reduced posterior part of the anterior hypothalamic nucleus (AHP) characterizes SPa within PHy (AHP; Figure 1B; Puelles et al., 2012a).

We will place here a few relevant data relative to the preoptic area, which we interpret not as hypothalamic, but as belonging within the hp2 (terminal) prosomere to the telencephalic subpallium territory (this consists of striatal, pallidal, diagonal, and preoptic domains, the preoptic area being the rostralmost of them; Puelles et al., 2012a,2013, 2016). The preoptic area is limited caudally from the evaginated hp1 telencephalic derivatives by a dorsal transversal extension of the intrahypothalamic boundary. This extends dorsalward into the rostral part of the roof plate, first coursing parallel and just rostrally to the fornix tract (Puelles and Rubenstein, 2015), then bending around the anterior commissure, and reaching the brain surface along with the limit between the septum (part of the evaginated telencephalon) and the preoptic area (unevaginated; Figures 2A,B, 3A,B). The dorsoventral (longitudinal) preopto-hypothalamic border is reported to correlate both with the change from subpallial Dlx/Arx preoptic signal versus Otp/Sim1 hypothalamic PA signal (Shimogori et al., 2010; Puelles et al., 2012a), and with telencephalic Foxg1 versus hypothalamic Foxg2 (Hatini et al., 1994). Our data reveal that some fasciculated TH-positive fibers ascending through the medial forebrain bundle diverge caudalwards (longitudinally) into the stria medullaris as they reach this boundary (smed; Figures 4A,B, 8A–D). The preoptic acroterminal domain is formed by the thin and median terminal lamina, which extends between the anterior commissure and the optic chiasma. The median preoptic nucleus is associated with this acroterminal domain; it extends subpially in front of the anterior commissure.

Tyrosine hydroxilase immunoreacted sagittal sections divide the preoptic area into three regions: there is a large central region next to the terminal Pa area, whose periventricular stratum is largely devoid of TH neurons (POce). Additionally, there is a septocommissural (SeCo) region that represents the preoptic paraseptal transition into the septum (Puelles et al., 2013, 2016); the SeCo shows strong ventricular Shh signal and is marked by a profuse TH-positive periventricular neuropil (SeCo; Figures 2A,B, 3A,B, 13A,B). Finally, we distinguish also a separate caudodorsal preoptic region found next to the hp1/hp2 boundary, just ventrally to the anterior commissure; this region may be named the preopto-strial nucleus (POSt), since its population apparently contributes to the complexity of the bed nuclei striae terminalis (POSt; Figures 2A,B,4A,B, 8A–D, 9A–E, 11F,H, 12A–E, 13A,B and Supplementary Figure 5A).

The POSt subdivision lies topologically ventral to the anterior commissure and is represented by a dense mass of periventricular TH-immunoreactive neurons that decreases laterally in a triangular shape. The axons from these cells seem to proceed laterocaudally into the supracapsular stria terminalis (Figures 2A,B, 8A–D). The ventral contour of POSt is lined (limited) by Avp/Oxt positive cells arranged in a small periventricular group and a larger comma-shaped lateral group forming a partial ventral shell to the POSt population. We named these peptidergic populations inner and outer preoptic magnocellular nuclei (iPOMc; oPOMc; Figures 9A,B,D,E, 10B,B″, 11F–H, 12A).

The hypothalamic basal plate corresponds to the classic regions of the tuberal, posterior and mamillary hypothalamus.

THy: however, according to the prosomeric thesis of dual peduncular and terminal hypothalamic prosomeres, the classic tuberal and mamillary areas strictly belong only to THy (hp2), jointly with a small perimamillary part of the classic posterior hypothalamus (Tu, PM, M; Figure 1B). The Tu is redefined as a longitudinal basal THy domain lying under the alar-basal boundary, and divided dorsoventrally into dorsal, intermediate, and ventral subdomains (Tu; DTu, ITu, VTu; Figure 1B; Puelles et al., 2012a; Ferran et al., 2015a; Puelles and Rubenstein, 2015). The Tu area ends rostrally in a distinct tuberal acroterminal subdomain with similar dorsoventral subdivisions and characteristic derivatives described below (ATerm; Figures 1B, 13A, 15A; Puelles et al., 2012a; Puelles and Rubenstein, 2015). Ventrally to the major Tu region there appear two other longitudinal basal plate components, the perimamillary area (PM) and the mamillary area (M) (PM; M; Figure 1B; see molecular delimitations in Puelles et al., 2012a and López-González et al., 2021). These areas also end rostrally in the corresponding ventral part of the basal ATerm (Figure 1B). M contacts ventrally the floorplate of hp2 (Figure 1B).

PHy: the PHy displays instead of a smaller, similarly, placed, and subdivided retrotuberal region (RTu; RTuD, RTuI, RTuV; Figure 1B), which was not clearly recognized classically. The intrahypothalamic boundary that separates transversally Tu from RTu roughly passes just rostral to the fornix tract (fx; Figures 1C, 3B; Puelles et al., 2012a; Puelles and Rubenstein, 2015). The RTu region contacts caudally the prethalamic tegmentum (basal plate) in the diencephalic prosomere 3 (Figure 1B; Puelles et al., 2012a; Puelles and Rubenstein, 2015). Ventral to RTu we distinguish periretromamillary and retromamillary areas (PRM; RM; Figure 1B); RM was classically known as ‘supramamillary area,’ whereas PRM represents another part of the classic posterior hypothalamus. Note that columnar authors added to posterior hypothalamus other more caudal basal plate regions now ascribed instead to the diencephalic tegmentum (basal p3; basal p2; Figure 1B).

Our description of basal hypothalamic TH cell populations will proceed successively through the Tu/RTu, PM/PRM, and M/RM longitudinal basal domains.

Most studies defining the hypothalamic region and its nuclei during the last century followed the columnar model (e.g., Herrick, 1910; Kuhlenbeck, 1927, 1973; Bleier, 1979; Swanson, 1987; Armstrong, 2004). Likewise, a large part of the research projects aimed at hypothalamic neurochemical mapping, connectivity, and physiology in rats were guided by the columnar paradigm (Swanson, 1987, 2003; Canteras et al., 1992, 2011; Elias et al., 1998, 1999; Torrealba et al., 2003; Armstrong, 2004, 2015; Choi et al., 2005; Yoshida et al., 2005; Markwell et al., 2010). In its origins, the columnar model proposed by Herrick (1910) aimed to extrapolate into the forebrain the theory of hindbrain columnar sensorimotor and viscerosomatic functional compartmentation developed after the discovery of longitudinal neuronal columns associated with specific cranial nerve fiber components (Gaskell, 1889; Johnston, 1902; Herrick, 1903). To this end, Herrick redefined de facto and without discussion the length axis of the forebrain. He departed from the earlier axial reference defined by His (the sulcus limitans separating neurogenetically heterochronic alar and basal plates; His, 1893, 1904). Herrick reinterpreted three other diencephalic ventricular sulci that intersected obliquely His’ longitudinal sulcus limitans as ‘longitudinal’ landmarks held to define four diencephalic columns analogous to the hindbrain ones (see Puelles and Rubenstein, 2015; their Figure 2). The consequently redefined hypothalamus (since that of His was restricted to the basal region ventral to sulcus limitans; see the historic account of the hypothalamus concept in Puelles et al., 2012a) was the ventralmost diencephalic column within the new “longitudinal four”: epithalamus, dorsal thalamus, ventral thalamus and hypothalamus (Herrick, 1910). Each of these diencephalic longitudinal columns supposedly ending in the telencephalon (soon shown to be untrue) was assumed to be structurally and functionally homogeneous, akin in this aspect to the hindbrain columns (Herrick, 1948; Kuhlenbeck, 1973).

However, 100 years later, the hypothalamic region turns out to be a structurally and functionally non-homogeneous region. The last three decades of progress in understanding the molecular map and fate of hypothalamic parts (Puelles and Rubenstein, 1993, 2003, 2015; Puelles et al., 2012a; Ferran et al., 2015a; Díaz and Puelles, 2020) increasingly suggest inconsistency of the molecular data with the columnar model, and agreement, instead, with the old axial notion of His (1893, 1904), as well as with earlier forgotten neuromeric studies (Puelles et al., 1987; Puelles, 2021). For instance, the Nkx2.2 gene is expressed in all vertebrates studied (both in embryos and adults) in a thin band along the alar-basal boundary all the way from the isthmus to the retrochiasmatic hypothalamic region, roughly as predicted by the sulcus limitans of His (see Puelles and Rubenstein, 2015; their Figure 7); this longitudinal band crosses all the diencephalic ‘longitudinal’ sulci of Herrick, and does not enter the telencephalon, contrary to what was predicted for the alar-basal boundary by the columnar model (discussed in Puelles et al., 2012a; Puelles and Rubenstein, 2015). Many detailed molecular hypothalamic subdivisions appeared which are not explained by the columnar model but can be understood easily as either neuromeric AP or dorsoventral (DV) compartmentation (Puelles and Rubenstein, 2003; Shimogori et al., 2010; Morales-Delgado et al., 2011, 2014; Puelles et al., 2012a; Ferran et al., 2015a; Puelles, 2017). This progress laid the foundation for the proposal of the alternative prosomeric brain model (see an account of its origin in Puelles, 2021). This model is based on the definition of transverse neuromeres as serial AP developmental units that display a shared organization into longitudinal DV floor, basal, alar, and roof zones or plates (Figures 1A,B).

The currently updated prosomeric model is consistent with masses of detailed molecular data obtained from different vertebrates (Puelles and Rubenstein, 1993, 2003, 2015; Martínez et al., 2012; Puelles et al., 2012a,c; Puelles and Ferran, 2012; Nieuwenhuys and Puelles, 2016; Moreno et al., 2017; Nieuwenhuys, 2017; Puelles, 2018, 2019, 2021). Studies of genetic patterning, regionalization, and gene expression profiles have led to a precise, causally underpinned definition of the prosomeric forebrain axis (thus defining causally, rather than arbitrarily or by convention, what is longitudinal in the brain) and several anatomical fiber tract landmarks that co-define the hypothalamic transversal boundaries (e.g., the supraopto/paraventriculo-hypophysial tract, the fornix tract and the medial and lateral forebrain bundles of the cerebral peduncle; Keyser, 1972; Puelles et al., 1987, 2012a; Bardet, 2007; Bardet et al., 2008; Puelles and Rubenstein, 2015). Gene expression patterns in mice likewise identified a variety of hypothalamic alar and basal plate subdivisions, which are not accounted for by the columnar model, not only in the mouse (Bardet et al., 2008; Shimogori et al., 2010; Morales-Delgado et al., 2011, 2014; Puelles et al., 2012a; Díaz et al., 2015; Ferran et al., 2015a; Puelles and Rubenstein, 2015), but also in chicken (Bardet, 2007; Bardet et al., 2008), frog (Bardet et al., 2008; Domínguez et al., 2013, 2014; González et al., 2020), reptiles (Moreno et al., 2012, 2017; Domínguez et al., 2015), catshark (Santos-Durán et al., 2015, 2016, 2018) and zebrafish (Lauter et al., 2011; Schredelseker and Driever, 2020). The adult structural and hodological organization of the hypothalamus of vertebrates thus appears to be the result of causal patterning mechanisms active during the development of this region with regard to the axial dimension first defined by His.

Most studies in the hypothalamic region in rats nevertheless followed the insufficient columnar model. In this model, the hypothalamic region is the ventralmost diencephalic longitudinal domain, separated from the ventral thalamus (prethalamus in the prosomeric model) by the ventral diencephalic sulcus (Herrick, 1910; see schemata of variant columnar models in Puelles et al., 2012a). A ventral (basal) longitudinal position of the hypothalamus is thus assumed, jointly with anteroposterior subdivisions into preoptic, anterior, tuberal, and mamillary/posterior hypothalamus regions (Figure 15A). In addition, some studies identified dorsal, intermediate, and ventral hypothalamic partitions, with the dorsal part supposedly bordering the ventral thalamus (Swanson et al., 1981; Chan-Palay et al., 1984; Ruggiero et al., 1984; Van den Pol et al., 1984; note Puelles et al., 2012a concluded that the classic dorsal hypothalamus structures actually fell within the prethalamus or ventral thalamus, and were thus not hypothalamic unless they migrated therefrom; see below some evidence that this may occur at the cellular level). The columnar intermediate and ventral ‘longitudinal’ hypothalamic parts roughly correspond to the prosomeric transverse Phy and THy parts, respectively.

According to the prosomeric model, the hypothalamus is instead, topologically, a rostral region of the brain bent morphogenetically into a ventral topography, due to the constant axial bend at the cephalic flexure. The hypothalamus is located primarily in the secondary prosencephalon, rostral to the diencephalon proper and ventral to the telencephalon, which is held to be a dorsal evagination from the alar hypothalamus (Figures 1A,A′; Puelles et al., 2012a; Puelles and Rubenstein, 2015; Ferran et al., 2015a; Nieuwenhuys and Puelles, 2016). The preoptic area is modernly not regarded as part of the hypothalamus but represents instead a non-evaginated subpallial telencephalic region (a.k.a. as telencephalon impar), as was thought originally in the time of His (POA in Figure 1A; Gurdjian, 1927; Crosby and Woodburne, 1940; Bardet et al., 2010; Puelles et al., 2012a,2013, 2016). The sub-telencephalic transversal hypothalamic region develops a large number of individual nuclei or dispersed cell populations derived from either the alar or basal plate domains of the hypothalamic prosomeres 1/2 (hp1, hp2). These define, respectively, a caudal peduncular hypothalamus that extends dorsally into the evaginated telencephalon (Phy; hp1; soft pink; Figures 1A,B) and a rostral terminal hypothalamus continuing into the non-evaginated preoptic area (THy; hp2; stronger pink; Figures 1A,B; Puelles et al., 2012a; Puelles and Rubenstein, 2015). The prefix tags ‘P’ and ‘T’ are used generally to refer to the fundamental Phy/THy division.

The prosomeric concept of the floor, basal, alar, and roof plates differs somewhat from that of His (1893, 1904). The bilateral ridge that delimits the open neural plate from head skin tissue is defined in accord with fate-mapping experiments as representing the prospective fused and thus median dorsal roof plate of the neural tube and brain. The roof plate of the closed neural tube ends rostrally at the septo-preoptic median bed of the anterior commissure (Figures 1A,A′; fate maps of Puelles et al., 1987; Cobos et al., 2001; Inoue et al., 2000; note His’s roofplate was shorter). The prosomeric ventral floorplate stops short of that neural plate ridge, due to its strict co-extensiveness with the underlying notochord, which is its inducer (the floor starts to form by nodal gastrulation of chordal nature caudal to the precocious prechordal plate; see Ferran et al., 2022), whereas the floorplate of His was merely assumed to end at the rostral ridge, similarly, as the basal plate, wrongly holding at the same time that the rostral ridge represents the prospective optic chiasma. The modernly defined anterior end of the floor plate accordingly stops at the mamillary body; see also Puelles et al. (2012a), Puelles and Rubenstein (2015), Puelles (2018).

The different neural plate topology of the prospective roof and floor plates implies that the alar and basal plates (separated by the alar-basal boundary) must reach in parallel a topologically dorsoventral lineal end of the rostral neural tube that is primarily closed and connects along the rostral midline of the neural plate the rostral ends of the floor and roof longitudinal zones. This unique rostromedian territory was defined recently as the ‘acroterminal’ forebrain domain (Aterm; Figure 1B; Puelles et al., 2012a; Puelles and Rubenstein, 2015; Puelles, 2018). The right and left basal and alar plates are thus continuous, respectively, from side to side across the acroterminal region of the terminal prosomere (ATerm; hp2; Figure 1B). The rostromedian derivatives found in the adult brain between the roof anterior commissure and the floor mamillary region include the thin terminal lamina of the preoptic area, the optic chiasma (plus suprachiasmatic nuclei and evaginated eyes), the retrochiasmatic or anterobasal tuberal area, the median eminence and infundibulum, the neurohypophysis, the arcuate nucleus, and the ventral tuberomamillary area, plus an acroterminal part of the mamillary body. All belong in the prosomeric model equally to the most rostral part of the brain, representing a median (virtually symmetric) acroterminal DV sequence (Puelles et al., 2012a; Puelles and Rubenstein, 2015; Puelles, 2018).

The molecular and structural patterns observed during the early stages of hypothalamic development indicate that after neural tube closure a fundamental AP regionalization of the hypothalamus takes place, resulting in at close range in the specification of two hypothalamic neuromeric units (hp1; hp2; Figure 1A′), separated by the intrahypothalamic boundary (IHB; Figure 1B). The caudal limit of the hypothalamus corresponds to the hypothalamo-prethalamic boundary (the thick black line separating pink from yellow territories; Figure 1B); the latter courses just caudal to the cerebral peduncle and in front of the prethalamic reticular nucleus and zona incerta derivatives (Puelles et al., 2012a; Ferran et al., 2015a; Puelles and Rubenstein, 2015).

Hypothalamic morphogenesis, neuronal differentiation, and axonal navigation patterns are held to be influenced causally by morphogens diffusing gradientally from the roof (DV), acroterminal (AP), and floor (DV) midlines; this is reflected in the observed courses of most axonal tracts (Figures 1C,D). Antagonistic signals diffusing from the roof plate (dorsalizing effects of WNTs/BMPs) and the floor plate (ventralizing effects of SHH) are involved early on in patterning the DV regionalization of the Phy, THy, and Aterm hypothalamic parts, creating their alar/basal regions and the related alar-basal boundary, as occurs elsewhere in the brain. In addition to the underlying alar hypothalamic PA and SPa domains (Tpa/Ppa; TSPa/PSPa), these opposed gradients also pattern the telencephalon as an exclusively alar outgrowth. On the other hand, molecularly distinct tuberal/retrotuberal (Tu/Rtu), perimamillary/periretromamillary (PM/PRM), and mamillary/retromamillary (M/RM) domains are patterned as DV parts of the hypothalamic basal plate, stacked dorsoventrally between the alar-basal boundary and the mamillary/retromamillary hypothalamic floor (Figure 1B; Puelles et al., 2012a; Puelles and Rubenstein, 2015; Nieuwenhuys and Puelles, 2016); each pair of denominations refers, respectively, to longitudinally corresponding domains of THy and Phy.

Note that the prosomeric alar Pa region used to be included classically in the preoptic area (as ‘preoptic magnocellular nuclei’), whereas the SPa alar region roughly corresponds to the columnar ‘anterior hypothalamus’ (though this concept is used modernly in a very vague way; comments in Puelles et al., 2012a). The Tu/Rtu basal complex roughly corresponds to the columnar ‘tuberal hypothalamus,’ and the PM/PRM plus M/RM complexes jointly form the columnar ‘posterior hypothalamus/mamillary’ region (compare Figures 1B,C). Obviously, the ontology of what exists in the hypothalamus is largely identical in both models, but individual elements are referred to descriptively with regard to different (actually orthogonal) length axes (e.g., confusingly causing the columnar ‘ventral premamillary nucleus’ to lie dorsally to the ‘dorsal premamillary nucleus’ in the prosomeric model, and both ‘premamillary entities’ are found dorsal to the mamillary body, rather than rostrally to it).

The prosomeric area map is on the whole more discriminative than the columnar one, since it adds AP and DV subdivisions, and refers to two neuromeres (hp1, hp2; or Phy, THy) plus the acroterminal domain (Aterm). As a result, there is a checkerboard pattern of subregions wherein specific cell populations can be pinpointed. The clearcut AP/DV pattern and attached morphogenetic signaling possibilities also introduce the possibility of causal explanations as a result of theoretically invoked DV and AP patterning mechanisms originated, respectively, from the roof and floor versus the ATerm (Puelles, 2017; Díaz and Puelles, 2020).

Theoretically, any spatial arrangement of neuronal populations in the brain may be explained causally by analysis of the data with regard to a morphologic model that (a) embodies a consistent set of assumptions about axial, dorsoventral, and anteroposterior dimensions, (b) maps consequently relevant regional subdivisions, and (c) postulates the relative positions of potential secondary organizers of the local pattern (i.e., relevant sources of patterning morphogens). We have attempted to perform this task for hypothalamic dopaminergic cell populations, examining their 3D positions relative to the dimensional assumptions, the hypothalamic subdivisions, and the potential secondary organizers postulated in the updated prosomeric forebrain model (Puelles et al., 2012a,c, 2013; Ferran et al., 2015a; Puelles and Rubenstein, 2015; Puelles, 2017, 2018, 2021; Díaz and Puelles, 2020; Amat et al., 2022). We have attempted to frame the histological data within the prosomeric model, and merely discuss briefly here some interpretive possibilities, believing that this approach may be of interest in the long run for those looking for causal explanations for related neuroanatomic structure. Obviously, further studies and experimental tests are needed to conclude scientifically about specific testable causal hypotheses for each of the dopaminergic cell groups.

In retrospect, it may be reflected that the alphanumeric classification of catecholaminergic cell groups advanced by Dahlström and Fuxe (1964) and further elaborated by various followers (see Introduction and Figure 15A) may have been the result of difficulties encountered in the classification of these cell groups within the then prevailing columnar forebrain model of Herrick (1910), Herrick (1933), Herrick (1948) and Kuhlenbeck (1927, 1973). The Nordic authors possibly discovered oddities or inconsistencies in their likely initial effort to ascribe their histochemically mapped cell groups to the postulated system of brain vesicles and columns of the columnar model then absolutely prevalent. For instance, several of the studied cell groups–including the A9, or A11 groups, for example–appear elongated along a dimension that is incoherent with the direction of columnar longitudinal sulci and columns. Remarkably, there exist no columnar descriptions or flat maps of the multiple brain CA cell groups (i.e., descriptions detailing column by column their content of catecholaminergic cells). Perhaps noticing the absence of a satisfactory model, Dahlström and Fuxe (1964) and followers turned to the pragmatic alphanumeric solution. During the last 60 years, this classification has been widely used in the field, unwittingly representing an aberrant result of anatomic analysis, insofar as it implicitly contradicts without saying so the then prevailing columnar brain model (incapable to account for the pattern). Unfortunately, this also tended to preempt efforts toward causal explanations, since the chosen arbitrary denominational solution is devoid of a contextual causal framework. In a hypercomplex system as is the brain, empiric data unrelated to a structural model or its developmental assumptions do not have explanations or morphologic meaning. Of course, a model-based interpretation is necessarily based on assumptions. A specific model provides an informed and useful interpretation coherent with various aspects of the data, but not the truth; the potential long-term validity of such interpretations depends on the historic development of the science (kinds of available data needing explanation) and how long the models’ interpretive usefulness prevails subjectively.

In this sense, our present effort thus attends to a probably still unrecognized old neuroanatomic problem: can we encompass the observed distribution of brain dopaminergic neurons (or just the hypothalamic ones) within a theoretic framework of spatial references and developmental phenomena – a model- that provides a credible explanatory causal background? The option to try the alternative and relatively novel prosomeric model toward achieving this generic aim was actually explored earlier with previous, less elaborate versions of this model. Indeed, L Puelles and collaborators attempted several times in the past to insert the conventional alphanumeric groups of catecholaminergic neurons within the prosomeric model in avian, mouse, and human embryonic brains. They had mixed success due to some problematic assumptions held at the time that needed to be corrected subsequently (Puelles and Medina, 1994; Puelles and Verney, 1998; Marín et al., 2005; Puelles et al., 2012a,c). Moreover, Puelles covered recently the scenario of known or possible secondary organizer sites existing in the forebrain (Puelles, 2017). Separately, Marín (Marín et al., 1998; their Figure 3) mapped schematically mesodiencephalic dopaminergic cell groups within the prosomeric model in tetrapods.

The main impact of this earlier prosomeric interpretive work so far was to convince several specialists on catecholaminergic neurons of the existence of a mesodiencephalic plurineuromeric constitution of the classic VTA and the SNC, a pattern already captured by the early prosomeric model (Puelles and Medina, 1994; their Figure 16.6B). In the columnar model, the same DA system is devoid of serial parts, and all the diencephalic components are ascribed rather arbitrarily to the midbrain. The alternative prosomeric interpretation visualizes a longitudinal tegmental DV differentiation pattern that is repeated with minor changes across several neuromeric units (midbrain m1–m2, and diencephalic p1–p3 in the updated model). The SNC/VTA dopaminergic phenotype seems to respond to potential close-range ventralizing floorplate/notochordal morphogenetic signals that are repeated along with these five modules. This mesodiencephalic neuromeric model background subsequently led to the identification of a floorplate-related gene cascade that operates serially to produce different features of these particular dopaminergic neurons (e.g., Smits et al., 2006). Not all forebrain dopaminergic neurons admit the same explanation, though, so we must conclude that other ranges of action, or other causal molecular DV and/or AP circumstances also deducible from the prosomeric model may lead to an analogous dopaminergic cell type elsewhere within the neural checkerboard pattern. An example of differences in dopaminergic identity is seen in meso-diencephalic dopaminergic cells, of midbrain and diencephalon (p1, p2, and p3) origin, which express dopamine active transporter (DAT), while this is absent in hypothalamic dopaminergic neurons (Sánchez-González et al., 2015). In the molecular era, one cannot expect the alar pretectal DA neurons to be patterned, similarly, to the midbrain VTA DA neurons. For this reason, careful topologic classification of hypothalamic TH cells is a prerequisite for the ulterior meaningful search of the probably combinatorial causes of each particular TH cell group that differs significantly in relative position.

The present analysis concentrated on the hypothalamic pattern of TH-positive neurons. This was studied previously with the same model, but in less detail (Puelles et al., 2012a). In their mapping of dopaminergic neurons in the mouse hypothalamus (based only on sagittal sections), Puelles et al. (2012a) reported basically widespread periventricular A14 TH-positive cells in the alar PA and SPa domains (with added subpial supraoptic cells), and four groups of basal plate dopaminergic neurons including the A13 group, the dorsal and ventral tuberal groups, and the arcuate A12 group. We examined the same territory in continuous series cut in three section planes (sagittal, horizontal, and transversal to the prosomeric axis through the hypothalamus). Note that our horizontal sections are parallel to the optic tract, being thus similar to conventional coronal atlas sections. They accordingly reveal a series of AP relationships. Despite our transversal hypothalamus sections being less commonly used, they otherwise illuminate DV relationships (they are essentially parallel to both the acroterminal domain and the cerebral peduncular and nigrostriatal tracts, and orthogonal to the optic tract). We further mapped by ISH the relative positions of several peptidergic cell populations found in the neighborhood of dopaminergic neurons.

An extensive alar periventricular stratum of TH-positive cells (possibly subdivided into Pa and SPa domains) was corroborated in the present analysis. We further distinguished several groups of TH neurons intermixed with peptidergic neurons in the neighboring medial hypothalamic stratum, both at the PA and SPa grisea. The PA complex shows more significant TH populations within PHy, whereas the TH cells of SPa mainly appear within the anterior hypothalamic nucleus within THy (Figure 15B). These differences may bespeak differential AP patterning (hp2 versus hp1), apart from the DV step existing between PA and SPa. Within the peduncular PA nucleus (subdivided between the model into dorsal, central, and ventral components; Puelles et al., 2012a) we noted inner and outer nuclear strata invaded by some TH cells (particularly at the Cpa), as well as similar partial TH subpopulations within the discontinuous inner, outer, mediolateral and laterolateral parts of the Vpa. Note standard columnar terminology identifies our Vpa as ‘caudal Pa’ (due to the orthogonal axis used). Our Dpa component (similar to columnar Apa) showed very sparse TH cells. The presence of TH neurons within the Pa did not seem to correlate specifically with any of the peptidergic cell types we mapped therein (see Results), though these did show themselves some already described specificity for given parts of the Dpa, Cpa, Vpa subnuclei. The conventional columnar ‘dorsal Pa’ subgroups (which also show some mixed TH cells), often ascribed to the ‘dorsal hypothalamus,’ were interpreted within our model as lying strictly within the deep strata of the prethalamus. We accordingly renamed them caudal Pa by reference to the prosomeric axis. It may be conjectured that these cPa neurons have migrated out of early PA hypothalamic origins into the neighboring compressed prethalamic domain (see the intercalated position of a thin wedge of GABAergic prethalamic territory between glutamatergic thalamic and PA territories in Puelles et al., 2012c; their Figures 9.4, 9.5).

The subpial position of the Avp-rich supraoptic nucleus and its accompanying TH cells along with the terminal Pa and just above the optic tract is a phenomenon unique in the hypothalamus and may require a singular explanation. It is to be noted that various neurogenetic data indicate that the hypothalamic mantle develops outside-in order. Curiously, SO cells do not exist at Phy levels occupied by the cerebral peduncle. Interestingly, there appears at Phy section levels corresponding to the Cpa and Vpa a small rounded mass of relatively lateralized Avp neurons adjacent medially to the ns/mc tract, which we named, respectively, laterolateral central and laterolateral ventral cell groups (llC; llV; Figures 9F,F′, 11C,E,E′). They probably correspond to what Peterson (1966) called the ‘accessory PA cell group.’ We now think it is possible that they may represent somewhat younger ectopic SO cells born after the peduncle (and the mfb) started to form superficially at the Phy surface, thus being secondarily separated from the subpial SO elements. Remarkably, Puelles et al. (2012a) alerted to transient data suggesting that some SO neurons appear to migrate ventralwards just deep to the optic tract, collecting thereafter subpially at the tuberal surface, ventrally to the optic tract, in the classic ‘tuberal supraoptic nucleus,’ which is obviously a partial topographic misnomer. It was suggested to identify them as a tuberal suboptic population. We did not notice significant numbers of TH cells at this secondary site.

In our study of the alar plate, we also noted TH cells in the preoptic area. This is interpreted in the prosomeric model as a telencephalic subpallial field included dorsally to THy within the hp2 prosomere. According to the historical summary offered in Puelles et al. (2012a), classic neuroanatomy first held this territory to be telencephalic and often included within it the subjacent PA area; later, after the nineteen forties, it became customary to include the preoptic area in the rostral part of the columnar hypothalamus. After the advent of molecular studies at the end of the twentieth century, it was discovered that the preoptic area shares various gene markers with the essentially GABAergic telencephalic subpallium, in contrast with the essentially glutamatergic PA phenotype, which led to the former’s renewed telencephalic adscription versus the latter’s alar hypothalamic adscription. We found scarce periventricular preoptic TH cells, with exception of a cell-dense TH-positive nucleus found just under the anterior commissure (the latter defines the preoptic- and THy-related roofplate); this cell group lies in the neighborhood of the intratelencephalic BST nuclear complex. We labeled it accordingly as strial preoptic nucleus (POSt). It seems to correspond to the dopaminergic cell group conventionally identified as ‘parvocellular anterior PA nucleus’ (pAPa; Swanson and Sawchenko, 1980), but we argue that no cell population contacting the roofplate (the anterior commissure) can belong to the PA area, that is, the alar hypothalamus, and must belong to the interposed telencephalic subpallium (hp2 preoptic area, in this case). Similarly, two constant medials (deep) hyperdorsal TH-positive cell groups with large neurons are observed above the Dpa nucleus (at the transition into the telencephalon of the Phy or hp1). These are conventionally identified as ‘magnocellular anterior PA formations.’ We likewise doubt the implied ascription of these neurons to the PA complex, considering them barely intratelencephalic. We accordingly have renamed them alternatively as inner and outer magnocellular preoptic nuclei (iPOMg; oPOMg). In this case, we think that their vicinity with the POSt nucleus, plus their topology altogether dorsal and intratelencephalic, compared with the PA complex, may justify the hypothesis that they mark an extension of the preoptic subpallial field into the hp1 prosomere (previously thought to be devoid of preoptic elements). This is a tentative proposal that has some advantages (e.g., all four subpallial domains –striatum, pallidum, diagonal area, and preoptic area- then show parallel topologies across hp2 and hp1, extending parallel to each other along the oblique septo-amygdalar axis). However, other conceptual options for ascribing these topographically hyperdorsal, thus non-hypothalamic, TH-positive cell groups have appeared recently. For instance, Morales et al. (2021) have defined recently a Foxg1-positive telencephalo-opto-hypothalamic domain (rich in Otp cells) that appears intercalated between the subpallium and the PA area; this might correspond to our presently postulated extended preoptic area domain and might contain both the POSt and the i/oPOMg formations. On the other hand, a recent study of Sim1-positive hypothalamic neurons that invade parts of the intratelencephalic pallial amygdala passing along the bottom of the interventricular foramen and the terminal sulcus suggested the postulate of a hypothalamo-amygdalar corridor (HyA), again at the interface of subpallium and the PA area (this HyA ‘corridor’ entity, however, is seen substantially as a dorsalward deformed extension of the PA area proper; see Garcia-Calero et al., 2021). Further studies will be needed to clarify this transitional region between telencephalon (including preoptic area) and alar hypothalamus. In the meantime, our conclusion is that there exist some TH-positive cell groups that clearly lie topologically dorsal to the PA area proper (i.e., closer to the roofplate); these groups probably correspond to one of the newly defined subpallial or PA subdomains (extended POA; TOH; HyA).

Regarding the basal hypothalamic TH populations recognized by Puelles et al. (2012a), the classic A13 group was then first ascribed to the hypothalamus (rather than to the prethalamus, as most earlier publications cited above had assumed; a notable exception was Tillet, 1994). The existence of an A13 commissure or decussation was also identified (Puelles et al., 2012a; their Figures 8.25A–D). Our present analysis suggests that the compact core of this distinct nucleus is indeed strictly basal and lies in the caudal end of the RtuD domain, whereas the dorsal part of the associated shell formation seems to lie within the alar PSPa (this may be primary or, more probably, the result of secondary migratory dispersion from below).

The dorsal tuberal group was originally defined as lying strictly dorsal to the VM nucleus within TuD, as a distinct component of the classic A12 group (A12dt; Puelles et al., 2012a; their Figures 8.25C–D). We have now observed that this group corresponds in fact to a linear (longitudinal) band of similar cells associated to the whole TuD/RtuD domain, that is, extending through most of THy and Phy, excepting the acroterminal rostral portion and the caudal locus where the A13 group is found (see section “Results”). That this configuration was not discovered before may relate to the higher sensitivity of the present immunocytochemical protocol. Nevertheless, there is at the middle of this band a slightly more radially extensive subarea that protrudes lateralwards in a triangular shape between the VM and the DM-P nuclei. It can be distinguished from the DM-P nucleus because the latter (a RtuI formation) has a characteristic diffuse TH-positive neuropil that is absent along the TuD/RtuD area occupied by the dorsal tuberal population.

The ventral tuberal TH-positive cell group is found essentially along the peripheral shell portion of the DM-P and DM-T formations. TH-positive cells are most abundant at the rostral end of DM-T next to the Arc, whereas the density of such cells decreases gradientally caudalwards into DM-P. This description basically agrees with that of Puelles et al. (2012a) but introduces the precision of a topography largely coinciding with the DM shell component, essentially formed by GABAergic neurons (Puelles et al., 2012a; their Figures 8.17, 8.18, 8.20A). The DM core cell mass, contrarily glutamatergic in nature, seems devoid of TH neurons, similarly, as the largely glutamatergic ventral premamillary nucleus (VPM) that occupies a rostrolateral part of DM-T after its tangential migration from the retromamillary area (López-González et al., 2021).

Our results relative to the acroterminal arcuate nucleus (Arc/ArcW) corroborate the results of Puelles et al. (2012a). We noted the variable topographic relationships of arcuate TH cells with diverse peptidergic cell populations along the radial dimension. The neighboring median eminence and infundibulum show no TH cells but have a dense plexus of TH-positive terminals and fibers. Not so, however, the TH-negative acroterminal tubero-mamillary area (identified as Vtu in Puelles et al., 2012a), representing the rostral end of the TuV/RtuV domain, which apparently coincides at ependymal level with the hypothalamic ventricular organ (HVC), associated with the origin of histaminergic neurons; the latter disperse by short-range tangential migrations in the tubero-mamillary and mamillary/retromamillary areas (Puelles et al., 2012a; their Figures 8.28, J–Y). Few TH cells were found within the mamillary body, whereas distinct medial and lateral TH elements (RMM, RML) were identified by us at the retromamillary area. A separate small group of large radially disposed of TH neurons was detected in the caudal part of the RML, close to the rostral end of the prethalamic component of the substantia nigra compacta. This element may represent a minute nigral population within the retromamillary area (SNC; Figure 15B). If they are truly hypothalamic rather than diencephalic, this would convert the mesodiencephalic SNC modular continuum into a meso-diencephalo-hypothalamic complex.

By reference to the prosomeric model of the hypothalamus, most of the studied TH populations represent longitudinally elongated dorsoventral tiers stretching more or less regularly across either the alar or basal plates of Phy and THy (prosomeres hp1, hp2), at different DV positions. In the alar plate, we have dorsalmost the hypothetically preoptic cell groups POSt and i/oPOMc, which occupy distinct hyperdorsal discontinuous positions within what could be an extended BST-related part of the preoptic area or a telencephalo-opto-hypothalamic domain (Morales et al., 2021) across hp2 and hp1. Complemented by additional periventricular elements found within the standard preoptic region, the mentioned cell populations would correspond to the classic A15 (Figures 15A,B). Underneath (ventral from) these preoptic elements there is the TH-poor Dpa subdomain of the Pa/SPa domain. This Dpa area would roughly correspond to the free space between A15 and A14 in the classic schema (Figure 15A). Ventral to that we have a variety of alar TH cell populations at various radial levels of the longitudinal Cpa (central) and Vpa (ventral) subdomains (across both THy and Phy). This includes dense periventricular elements, a variety of cells mixed with the medial subnuclei of the Pa nuclear complex (=medial or inner intermediate hypothalamic stratum), laterolateral displaced accessory elements (llC/llV = outer intermediate hypothalamic stratum), and the superficial subpial elements at the SO nucleus (jointly with the ectopic, tangentially ventrally migrated cells of the TuSbO nucleus). This radially varied system of cell groups associated with the Pa progenitor domain and nuclear complex probably should be distinguished from the SPa counterparts as dorsal A14 (A14d; Figures 15A,B). Apart from the likewise rich SPa periventricular TH cells forming a third longitudinal alar band across THy and Phy ventrally to the Pa complex, there are relatively fewer TH neurons at the medial or inner intermediate stratum of the SPa area (these are found predominantly at the anterior hypothalamic nucleus). This less important SPa system of TH cells would form the ventral A14 cell group (A14v; Figure 15B). Given that many embryonic neurons born at the SPa migrate tangentially into the underlying tuberal/retrotuberal basal hypothalamic domain (notably into the shell of the VM nucleus; Díaz et al., 2015), it is conceivable that some TH neurons of SPa origin may have reproduced such behavior and occupy basal locations in the adult brain. Corroboration of this possibility requires further research distinguishing theoretically alar-born from basal-born tuberal TH cells.

In general, the alar part of the acroterminal domain (including the SPa SCH nucleus) is free of TH cells in both the classic alphanumeric and prosomeric maps (Figures 15A,B). This alar Aterm represents as a whole the rostralmost transversal alar territory within the prosomeric model.

The basal hypothalamus also shows a distinct dorsoventral division of the longitudinally arranged tuberal TH cells of the classic A12 cell group into dorsal A12 and ventral A12 subgroups. These are aligned, respectively, with the dorsoventrally arranged TuD/RtuD and TuI/RtuI progenitor subdomains extending along THy and Phy: the corresponding TH cells appear associated (mixed) to other neuronal components of the Abas/Pbas and DM nuclei, respectively. In contrast to the uniformly poor alar acroterminal TH pattern, the related basal Aterm parts are different one from another. The AbasM Aterm formation of the TuD mantle is largely devoid of TH neurons, whereas the underlying TuI Aterm portion shows the densely TH-populated arcuate nucleus (Arc; Figure 15B), as well as the strongly innervated median eminences and hypophyseal infundibulum (Puelles et al., 2012a; present Results). However, the underlying tuberomammillary Aterm portion is devoid of TH elements similar to the underlying PM and M Aterm parts of THy.

The dorsal A12 extending along Abas/Pbas (TuD/RtuD) was thus already identified as ‘dorsal tuberal cell group’ (A12dt) by Puelles et al. (2012a); their Figure 8.25, though they did not observe the full longitudinal extent of this cell group, recognized in the present analysis apparently due to more sensitive antibody and/or protocol properties. These authors nevertheless did include the classic A13 group within the caudal PHy end of this specific longitudinal domain (A13 within Phy, caudal to Pbas; Figures 15A,B, rather than in the neighboring diencephalic prethalamus), a point we corroborated, particularly by noticing a direct relationship of A13 with the hypothalamic source of MCH neurons (see below). As far as we know, a partially hypothalamic nature of A13 was first suggested by Tillet (1994), p.211, who approximated our current interpretation by reporting that A13 neurons lie [partly] in the dorsomedial hypothalamus and [partly] in the [prethalamic] zona incerta. We think that any incertal TH elements would be alar prethalamic, and would probably correspond rather to the A11 group, though the latter was classically interpreted as tegmental (Figure 15A); the rostralmost prethalamic A11 cells, associated topographically to the zona incerta (Puelles et al., 2021), may easily be confused with those of A13. A remarkable peculiarity of the A13 group (particularly its distinct dense core portion) is that at its level there is no periventricular TH cell population, while there exists a localized decussation of TH-positive A13 fibers across the locally fused or absent ependym (Puelles et al., 2012a; their Figure 8.25B–D). In addition, the Phy RtuD progenitor domain corresponding to A13 shows in our material the apparent local origin of medial and lateral hypothalamic MCH neurons (Figures 14A–C). The latter does not extend significantly into the Pbas or Abas mantle domains. Therefore, it appears justified to maintain the distinction between A12d and A13, though their DV position is essentially the same; however, note they are distinct in terms of AP position.

The ventral tuberal cell group was defined as ‘ventral tuberal band’ (A12vt) by Puelles et al. (2012a). It runs basically along the periventricular and medial strata of the peduncular and terminal parts of the DM nucleus (TuI/RtuI). The TH-positive DM population is generally most abundant within the shell region of the DM nuclei at all levels (DM-T; DM-P) but changes caudalwards from one nucleus into the other in a decreasing gradient. These TH cells typically eschew populating the periventricular and medial strata of the VM nucleus (also located within TuI in late embryos and adults). This cell mass was interpreted by Puelles et al. (2012a; their Figure 8.26), as a conglomerate of diverse cell populations migrated tangentially out of the TuD, a point they demonstrated partially for its Nkx2.2-positive cell subpopulation, possibly related to prodynorphin neurons (ibid; same Figures 8.26; A-K and L-N, respectively). Interestingly, the glutamatergic TuI/RtuI populations associated with VM, DM, and VPM core nuclear parts seem completely devoid of TH cells.

The hypothalamic basal region underlying the described tuberal/retrotuberal region contains in the prosomeric model other three longitudinal basal bands, the TuV/RtuV, PM/PRM, and M/RM domains (Figure 1B). These were classically held to be largely devoid of TH cells, including their respective acroterminal subdomains (Figures 15A,B). We confirmed a lack of TH cells or TH innervation comparable to that of the ME for the tuberomamillary Aterm recess, as well as along the thin TuV/RtuV territory. The latter is known conventionally as the ‘tuberomamillary band’ (i.e., it forms the longitudinal boundary between the tuberal/retrotuberal and perimamillary/periretromamillary domains in our model; Figure 1B). This thin limiting domain was identified by Puelles et al. (2012a; their Figures 8.28J-Y), as the longitudinal embryonic site where hypothalamic histaminergic neurons are produced, with local selective expression of Arx, Lhx6, and Hdc (histidine decarboxylase); note the postmitotic histaminergic neurons thereafter spread variously into adjacent subregions, some of which may receive dopaminergic input, allowing in principle some dopaminergic modulation of the histaminergic system at its origin. The TuV/RtuV ependym also appears developmentally related to a possible secondary DV organizer, in so far as this band of the neuroepithelium, the so-called hypothalamic ventricular organ (HVO; Puelles et al., 2012a; Puelles, 2017; Díaz and Puelles, 2020) expresses Wnt8 and releases the WNT8 morphogen protein.

We found that the PM/PRM basal hypothalamic longitudinal band that surrounds the ventralmost M/RM territory shows a periventricular population of TH cells, which is absent at the corresponding Aterm area (Figure 15B).

In contrast, the underlying ventralmost M/RM longitudinal basal domain lying adjacent to the floorplate shows a fair population of TH cells, but mainly at periventricular, medial, and lateral radial levels of the RM area (periventricular, medial, and lateral RM nuclei; see comments above). The mamillary body and its Aterm subarea are largely devoid of TH neurons, though we saw some large positive cells ventrally next to the floorplate at the M/RM border.

It may be deduced that patterning phenomena acting along the hypothalamic DV dimension must be particularly relevant for explaining most of these patterns related to the differentiation of TH-positive presumptive dopaminergic neurons at different DV positions. On the other hand, AP patterning or AP attractor/repulsion signals may be required to generate the partial differential patterns displayed by the caudal Pa (cPa) elements interpreted by us as displaced into alar central prethalamus, the distinct basal peduncular A13 mass, the alar SPa AH cell population and the lack of TH cells at the SCH nucleus within the SPa Aterm area, plus the alar acroterminal SO/TuSbO and basal acroterminal Arc populations, which display distinct AP coordinates accompanied in every case by equally specific DV positions. Interestingly, the prospective locus of the infundibulum, median eminence, and Arc nucleus, which are distinguished from all other acroterminal areas by their intense TH innervation and differentiation of many small TH cells, corresponds to the rostromedian basal hypothalamic area where the endodermal prechordal plate attaches intimately during a transient early postgastrulation period up to the stage in which the adenohypophysis (Rathke’s pouch) starts to develop (Ferran et al., 2022, in press).

Dahlström and Fuxe (1964) analyzed catecholamine-containing cells in the brain of rats and classified them using an alphanumeric system that recognized initially twelve cell groups (A1–12), five of which were dopaminergic (A8-A12). This list was later expanded to seventeen components (A1–A17) (Hökfelt et al., 1984). The anatomical location of dopaminergic groups A8-A15 was defined pragmatically according to their rough topography within the midbrain and diencephalon (note the preoptic area was then ascribed to the diencephalic hypothalamus, following the columnar model). The hypothalamic cell groups included A11 in the posterior hypothalamus region (Post Hy; Figure 15A), A12 in the tuberal region (TUB; Figure 15A), A14 in or near the anterior hypothalamic region (ANT; Figure 15A), and A15 in the preoptic area (POA; Figure 15A) (e.g., Chan-Palay et al., 1984; Ruggiero et al., 1984; Van den Pol et al., 1984; Björklund and Dunnett, 2007). The A13 cell group was generally described as ventral thalamic (or subthalamic; see Smeets and Reiner, 1994), though Tillet (1994) thought that part of it was placed in the dorsomedial hypothalamic area, and Puelles et al. (2012a) included it wholly in the hypothalamus. Several rodent studies reached the conclusion that hypothalamic TH neurons are presumably dopaminergic (Swanson and Hartman, 1975; Swanson et al., 1981; Chan-Palay et al., 1984; Ruggiero et al., 1984; Van den Pol et al., 1984; Ugrumov et al., 1989; Björklund and Dunnett, 2007).

However, after mapping these conventional cell groups within the prosomeric model (implying a different forebrain axis and the hypothalamus held to lie rostral to the diencephalon, plus two neuromeric hypothalamic parts -THy, Phy- and floor, basal, alar, and roof plates according to His, 1893, 1904, and a new acroterminal area; Puelles et al., 2012a; Puelles and Rubenstein, 2015; Díaz and Puelles, 2020), we excluded group A11 from the hypothalamic region because it relates to diencephalic prosomeres 1, 2 and 3 (Skagerberg and Lindvall, 1985; Marín et al., 2005; Puelles et al., 2012a,2021; Ozawa et al., 2017). The dopaminergic group A15 is clearly preoptic, and thus belongs to the non-evaginated telencephalon within the prosomeric model (Figure 15B; Shimogori et al., 2010; Puelles et al., 2012a). Only the conventional groups A12, A13, A14 can be considered part of the hypothalamic region as defined in the prosomeric model (Puelles et al., 2012a,2015).

Our analysis further reveals that the alphanumeric classification is topographically imprecise in several aspects and thus has limited utility, particularly for causal understanding.

The A14 group, which was generally vaguely described, corresponds to TH-positive cells located in the periventricular stratum of the current PA and SPa regions of the terminal and peduncular prosomeres (Puelles et al., 2012a; Ferran et al., 2015a; Puelles and Rubenstein, 2015). The standard concept of group A14 thus does not differentiate between these two molecularly and functionally distinct areas (Figure 15A, compare with B; note that Pa largely contains glutamatergic neurons, whereas SPa forms mainly GABAergic neurons as a result of their different molecular profiles; Shimogori et al., 2010; Puelles et al., 2012a). We accordingly propose to subdivide A14 into A14d, corresponding to a complex, diversely stratified set of TH cells within Pa, and A14v, including the less important TH populations of the underlying SPa area.

In its turn, the conventional A12 or tubero-infundibular dopaminergic group was held to include both the acroterminal arcuate TH cells (Arc; Puelles and Rubenstein, 2015) and the non-acroterminal neighboring periventricular TH neurons of the tuberal/retrotuberal hypothalamus regions (Dahlström and Fuxe, 1964; Baker et al., 1983; Lookingland and Moore, 2005). Following the prosomeric model, the so-called “mediobasal hypothalamus” notion lumps the prosomeric acroterminal tuberal, terminal tuberal, and peduncular retrotuberal hypothalamic regions that encompass distinct basal subareas such as Abas, Pbas, Arc, VM, DM-T, DM-P, each of which has differential properties (Baker et al., 1983; Puelles et al., 2012a; Puelles and Rubenstein, 2015). Even the division by Puelles et al. (2012a) of A12 into a dorsal tuberal part (A12dt) mainly related to the pe stratum of the dorsal tuberal region, and a ventral tuberal part (A12vt) defined by TH-positive cells in the Arc nucleus and in the pe stratum of the intermediate tuberal region seems now insufficient. Leaving aside the TH-poor VM nuclear complex, we propose to distinguish as probably causally independent the acroterminal Arc group from the dorsal and ventral tuberal groups, associated, respectively, to the TuD/RtuD domain (A12dtu) or to the TuI/RtuI domain occupied largely by the DM nucleus (A12vtu; DM-T + DM-P; Figure 15B). Each of these components probably has a different AP or DV causal explanation, according to their different topologic positions, and it is possible that their respective periventricular TH cells also display corresponding differential properties.

The dopaminergic A13 group was sometimes defined as incerto-hypothalamic (IH) and described as located in the rostral portion of the zona incerta (Chan-Palay et al., 1984; Van den Pol et al., 1984; Puelles and Medina, 1994). The latter is now classified as a ventral alar part of the prethalamus within the prosomeric model (the prethalamic alar subcentral complex; Puelles et al., 2021). There are indeed several TH neurons just underneath this diencephalic area, extending back across a similar liminar alar thalamic band into the pretectum. However, we identify this longitudinal diencephalic band as corresponding to the conventional A11 cell group. We follow instead Puelles et al. (2012a) in identifying the A13 with a much denser core of TH neurons found mainly in the medial stratum of the caudal end of the RtuD domain of Phy (surrounded by a sparser shell population). Our demonstration that this cell group coincides topographically with hypothalamic MCH neurons of the medial and lateral strata of the Phy corroborates this interpretation. Baker et al. (1983) thought that the ‘mediodorsal’ hypothalamus contains TH-positive neurons from groups A13 and A14. Both groups are strictly outside the DM complex, which is populated instead by its own A12vtu cell group, usually present in the shell portion that surrounds the dense core portion. The doubtful description of these authors apparently confused the peduncular alar PA nucleus (Pa derivative) with the peduncular basal A13 and underlying DM nucleus (Figure 15B).

A, similarly, imprecise notion is that of the periventricular nucleus of Chan-Palay et al. (1984). According to these authors this population lying close to the third ventricle extends uninterruptedly from the PA and anterior hypothalamic nuclei (alar hypothalamus) to the suprachiasmatic, arcuate, and dorsomedial nuclei (basal hypothalamus), being absent only at the ventromedial nucleus (Chan-Palay et al., 1984). A causal explanation of its patterning and origin as a single entity would be difficult; we think that a periventricular position is not a sufficiently precise character for defining and naming a cell group (i.e., most parts of the brain have periventricular neurons) and use of such a concept merely hampers our analysis of more specific entities. It seems thus advisable to separate at least the alar and basal components (strictly different molecular environments), and our functional understanding might profit as well from distinguishing likewise DV portions ascribed developmentally to alar Pa and SPa and dorsal versus intermediate basal Tu/Rtu parts. Some functional and hodological studies suggest that periventricular cells of the hypothalamus do not function as a unit (Goudreau et al., 1995; Zhang and van den Pol, 2016). For instance, a group of paraventriculo-hypophysial neurons of our A14d cell group includes a subset of dopaminergic cells that project to the intermediate lobe of the posterior pituitary (Goudreau et al., 1995). Future studies should clarify the potential association of each distinct population of periventricular TH cells with differential connections and functions.