Diversity of subpallial sources of tangential migrations and the associated terminology problem
In classical models, the subpallium was composed exclusively of striatal and pallidal parts. In recent times, the subpallium model was expanded to include the preoptic area (previously ascribed to the hypothalamus; see Shimogori et al.
2010; Puelles et al. 2010) and the hemispheric stalk (peduncular) region or diagonal area (Fig.
1a, b); the latter was known previously either as substantia innominata, or as anterior entopeduncular area; Bulfone et al.
1993,
1995; Puelles et al.
2000,
2004,
2013). Among the molecular characteristics that unify these subpallial regions is the overall early expression of
Dlx family,
Mash1 and
Arx genes (Puelles et al.
2004; Shimogori et al.
2010). Analysis of gene markers that label differentially a particular subpallial ventricular zone, such as
Nkx2.1 in the pallidum, stalk region and preoptic area, and
Shh in the POA1 part of the preoptic area (Fig.
1b), indicated that all major subpallial divisions extend from the septum to the amygdala, along the oblique septo-amygdaloid axis (Swanson and Petrovich,
1998; Puelles et al.
2000,
2013; Flames et al.
2007; Medina and Abellan
2012). The apparent caudal pole of this subpallial complex represents the caudal ganglionic eminence, ascribed to the amygdala, to which may be added the preopto-hypothalamic transition area (CGE, POH; Fig.
1a, b; Bulfone et al.
1993; Puelles and Rubenstein
1993; Xu et al.
2004,
2008; Butt et al.
2005; Fogarty et al.
2007; Sousa et al.
2009; Lee et al.
2010). Note the lateral ganglionic eminence (LGE) largely corresponds to the striatal domain, whereas the medial ganglionic eminence (MGE) encompasses the pallidal domain, the diagonal area, and part of the preoptic area (Fig.
1b).
Flames et al. (
2007) mapped at least 18 molecularly distinct progenitor domains at the subpallial ventricular/subventricular zone (Fig.
1c; note these are mostly aligned parallel to the septoamygdaloid axis). It was concluded that these domains might represent as many independent sources of specific neuronal types, which originate partly a diversity of tangentially migrating populations, and partly discrete radially stratified populations of the local mantle zone (i.e., a theoretical minimum of 18 × 3 = 54 cell populations, counting periventricular, intermediate and superficial strata at each subpallial domain; e.g., lateral bed nucleus striae terminalis, globus pallidus, and ventral pallidum plus pallidal olfactory tuberculum, within the pallidum). This conceptual background significantly qualifies the earlier simpler concepts of the lateral, medial, and caudal ganglionic eminences (LGE, MGE, and CGE), which were initially thought to represent homogeneous histogenetic entities. Indeed, classic embryologic studies had simplistically assimilated the striatum to the LGE, and the pallidum to the MGE.
The subpallial region that occupies the telencephalic stalk area, intercalated between the preoptic area and the pallidum, was recently renamed diagonal area (Dg) in the Allen Developing Mouse Brain Atlas (online since 2009) and Puelles et al. (
2013). This region is stretched like its neighbours along the septo-amygdaloid axis (Fig.
1b, c). Its intermediate stratum formed by dispersed neurons was classically identified as the ‘substantia innominata’ (which contains the cholinergic neurons of Meynert’s basal magnocellular nucleus, among other elements, and merges laterocaudally with the ‘sublenticular extended amygdala’). The Dg subpial stratum is occupied by the ‘diagonal band nuclei’ (horizontal and vertical), which lie interstitial to the diagonal (amygdaloseptal) tract and end at the medial septum; these nuclei also display in late embryos and the adult a prominent cholinergic population (Zaborszky et al.
2012; Allen Developing Mouse Brain Atlas). Various reports concluded that these cholinergic neurons, dependent on
Lhx7/8 and
Isl1 are not produced locally within the Dg domain (understood as medial extended amygdala, or MGEcv area; Zhao et al.
2003; Elshatory and Gan
2008; Medina and Abellan
2012); moreover, they do not share origins with other telencephalic cholinergic neurons presumedly produced at the preoptic area. Interestingly, our data reveal that ventricular and mantle expression of
Lhx7/8 is strictly restricted to the Pal and Dg areas, whereas the preoptic region only shows a few positive cells in its mantle, possibly migrated in from the Dg (Fig. S1). We believe this interpretation is consistent with corresponding data shown by García-López et al. (
2008; their Figures 1C, 4D; note also the neurons expressing ChAT illustrated therein only overlap the Dg region). In our opinion this weakens the hypothesis of a preopto-commissural origin of cholinergic neurons (since the differentiation of these cells needs
Lhx7/8 signal; Zhao et al.
2003). Here we merely wish to emphasize that whatever their origin may be (see also below our discussion of Pombero et al.
2011), these cells selectively mature within the diagonal band nuclei, as well as within the substantia innominata, both of which are components of the Dg. The diagonal histogenetic area also encompasses a periventricular stratum that includes the medial part of the stria terminalis complex (BSTM; note the adjoining BSTL part—which differentially expresses
Isl1 [Allen Developing Mouse Brain Atlas]—is instead pallidal; Fig.
1d). The Dg was characterized by Flames et al. (
2007) as a molecularly distinct areal component of the MGE, identified as the pMGE5 subdomain (see Fig.
1b, c). This domain was the only ‘pallidal’ (meaning MGE) area that expressed the transcription factor
Er81 (Flames et al.
2007; their Figure 2D); it also shows ventricular patches of
Shh signal which contrast with the massive POA expression (Flandin et al.
2010; present results). As mentioned above, the pallidal BSTL region expresses differentially
Isl1, compared to the diagonal BSTM.
The Dg was previously named ‘anterior entopeduncular area’ in early reports on the prosomeric model (AEP; e.g., Bulfone et al.
1993; Puelles and Rubenstein
1993; Rubenstein et al.
1994). The AEP name has since been used widely, though it proved to be imprecise (and thus inconvenient), since it suggests an isolated cell population interstitial to the medial or lateral forebrain bundles, rather than a complete radial (ventriculo-pial) histogenetic domain, as was intended. Puelles and collaborators thus came to regard their own term as obsolete. In the search for a better alternative name, the momentary absence of markers distinguishing this area from the pallidum led to the idea that this area could be seen as a part of the Pal, or, at least, of the MGE. Alternative names accordingly employed in recent literature include ‘caudal and medial MGE’ (Nery et al.
2002; Legaz et al.
2005), ‘central and ventral MGE’ (Fogarty et al.
2007), ‘anterior peduncular area’ (García-López et al.
2008), ‘ventral MGE’ (Flandin et al.
2010,
2011), and ‘caudoventral MGE’ (Bupesh et al.
2011a,
b; Medina and Abellan
2012). This excess of options generates by itself considerable semantic confusion, since none of these authors provided a subpallium map showing the precise location and extent of this area. We think that all these names are similarly inconvenient. First, because the axis of reference for the diverse positional descriptors used remains undefined and vague (probably the ‘central’ and ‘caudal’ terms allude to the arbitrary anteroposterior sequence of coronal section levels –the axis of the microtome- which is devoid of true morphologic value in the prosomeric model, but they might refer instead to the oblique, more realistic septoamygdaloid axis), and, secondly, because the precise position of the ‘medial’ or’ventral’ area along the dorsoventral dimension also seems unclear (no defining landmarks). Confusingly, the locus of this area is often implied to be circumscribed to a limited areal spot (most authors identify it only at a standard coronal section level), though the model shown in Fig.
1b, c suggests that the relevant distinct stalk area actually extends all the way from the septum into the amygdala (Puelles et al.
2013; see text of Figure 1d). Accordingly, the position within the MGE complex of this specific area and the developmental phenomena associated to it probably result imperfectly understood by most readers.
Curiously, Flames et al. (
2007) resolved minimally the molecular distinction of the stalk (Dg) histogenetic area relative to the pallidum, observing that the former selectively expresses
ER81 (Etv1) and lacks
Couptf1 signal, whereas the opposite is true for the pallidum. Present data illustrate a remarkably precise correlation of
Sst neurons with the mantle of this domain; García-López et al. (
2008) previously underlined the restricted presence at the same locus of calbindin-positive neurons, streaming tangentially into the pallial amygdala in a similar way as we observed
Sst cells. Therefore, it is both possible and convenient (and is supported by the present results) to have a distinctive name for the stalk area, leaving aside the old term AEP, as well as the vague descriptors of its position within the MGE. The recently updated prosomeric model (L.P., reference atlases developed for the Allen Developing Mouse Brain Atlas;
http://www.developingmouse.brain-map.org; Puelles et al.
2012,
2013) aimed to resolve this problem by proposing for the stalk area the alternative term
diagonal area (Dg); this conveys the correct topographic linear extent along the oblique hemispheric stalk and septoamygdaloid axis, and is based on a well-known classic concept and related identifiable surface landmark, the diagonal band (i.e., the subpial diagonal band formation generally forms a surface relief that separates the flat preoptic surface from the olfactory tuberculum, which represents the neighboring subpial portion of the Pal and St). This name refers to the adult region and corresponding characteristic cell populations, irrespective whether these are produced or not within the Dg proper (e.g., case of the cholinergic cells commented above).
In summary, the narrow diagonal radial histogenetic domain defined at the hemispheric stalk (Dg; Fig.
1d) is traversed orthogonally by the medial and lateral forebrain bundles, and encompasses the diagonal band nuclei superficially (jointly with the diagonal band tract), the innominate/magnocellular basal populations at the intermediate stratum (with the ventral amygdalofugal tract), and the medial bed nuclei of the stria terminalis complex (BSTM) periventricularly (with part of the stria terminalis tract). At its septal end, the Dg is continuous with the medial septum via the vertical limb nucleus of the diagonal band. At its opposite amygdalar end, our data suggest that the Dg finishes at the amygdalar BST nucleus (BSTA) and the associated lateral part of the central nucleus (CeL; see below).
Antecedents of the Sst cell migration
In this report, we study the apparent radial origin and subsequent wide tangential migratory distribution of
Sst cells in the mouse telencephalon between E10.5 and E16.5. In the final distribution such cells populate within the pallium the whole cerebral cortex (a well-studied point we do not need to examine in detail), the claustrum and the pallial amygdala, showing scarce contributions to the olfactory bulb.
Sst cells also invade the subpallium, though differentially. They are massively present in the lateral part of the central amygdaloid nucleus, as well as in the medial part of the bed nucleus stria terminalis complex, both of which are understood here as intrinsic (non tangentially migrated) derivatives of the diagonal area (see Suppl. Fig. 4; Puelles et al.
2013, and reference atlases of the Allen Developing Mouse Brain Atlas, online since 2009); there is also a labeled subpopulation of striatal interneurons, many labeled cells throughout the olfactory tuberculum, and some weakly labeled cells within the pallidal BSTL. In contrast,
Sst cells eschew rather selectively the globus pallidus and the preoptic area. Our results suggest that the earliest, and possibly the main, origin of telencephalic
Sst neurons is the diagonal area, which is a longitudinal hemispheric stalk subregion of the medial ganglionic eminence that appears intercalated between the pallidum and the preoptic area. The diagonal area (Dg) roughly corresponds to the pMGE5 ventricular progenitor domain of Flames et al. (
2007).
Somatostatin neurons were previously mapped immunocytochemically in rat and mouse embryos (e.g., Shiosaka
1992; García-López et al.
2008; Real et al.
2009). Compared to these earlier studies, our broader in situ hybridization analysis in different section planes and more embryonic stages provides much additional detail about their development, particularly because we explored in adjacent sections the relationships of
Sst cells with the subpallial expression domains of several transcription factors (
Gbx2,
Dlx5,
Lhx7-
8,
Nkx2.1,
Nkx5.1), as well as with transcripts of
Shh, the gene coding the morphogen SHH, significantly expressed in a large population of preoptic cells that invade the pallidum (Flandin et al.
2010). In agreement with Sousa et al. (
2009), we detected the first telencephalic
Sst cells at E10.5. In contrast, Shiosaka (
1992) first illustrated such cells in E16.5 rat embryos (equivalent to E14.4 mouse embryos; Clancy et al.
2001). Similarly, Garcia-López et al. (
2008) only reported data in E14.5 mice, while Real et al. (
2009) saw scattered cells in the mouse Dg at E11.5. These are all immunocytochemical studies.
Authors who examined the origin of cortical
Sst cells via transgenic and in vitro approaches generally concluded that these originate from a
dorsal pallidal part of the MGE (see specific citations below), whereas García-López et al. (
2008), Real et al. (
2009) and Medina and Abellán (
2012) associated the origin of
Sst cells that invade the amygdala to the anterior entopeduncular area or
MGEcv (i.e., the Dg), in agreement with our present conclusions. In the following sections, we will comment on the development of telencephalic
Sst cells, and discuss the conflicting views about the origin of these cells.
Early development of Sst cells
In the present report, we mapped descriptively the progressive changes observable in the topography of
Sst cells in the developing mouse telencephalon. Given the existence of several experimental transgenic tracings demonstrating that such cells migrate tangentially from a restricted (single) subpallial source into other telencephalic domains (see citations below), we have parsimoniously interpreted the observed absolute changes in topography as evidence of tangential migration, starting at the apparent source (i.e., where the earliest cells are found). Our results revealed that the earliest
Sst cells identified at E10.5 are not uniformly distributed within the MGE, being restricted to a thin radial domain of the MGE mantle zone that appears intercalated between the pallidal and preoptic regions, as identified by differential markers. This is precisely the estimated position of the Dg (Puelles et al.
2013; Fig.
1b) and roughly correlates with the ventricular pMGE5 domain of Flames et al. (
2007). Our correlative mappings of diverse subpallial markers at E10.5 and E11.5 showed that the early
Sst population does not overlap either with the massive subpopulation of
Gbx2 neurons that appears restricted to the pallidal area, nor with
Nkx5.1-expressing preoptic neurons. We detected also no significant overlap between early
Sst cells and the sizeable stream of
Shh-positive cells that migrates tangentially from the preoptic area into the pallidal mantle. Nevertheless, we systematically observed that the Dg ventricular zone itself shows patchy
Shh expression, a pattern that is distinct from the massive expression observed at the nearby preoptic POA1 ventricular zone; a small parallel contribution of Dg to the
Shh-positive population in the subpallial mantle seems thus possible, though such cells would not overlap the Dg-derived
Sst population (possible salt-and-pepper pattern, separately also implied in the hypothesis that some cholinergic poulations are locally produced). The subpallial nature of the diagonal
Sst cell population is corroborated by the shared expression of general subpallial markers such as
Dlx5,
Nkx2.1, and
Lhx7/8.
Already at E10.5, it can be seen that the Dg radial stream of
Sst cells is continuous superficially with an incipient subpial aggregate of similar cells spreading tangentially lateralwards; at this stage, these subpial cells partially cover marginally the pallidal mantle core occupied selectively by
Gbx2- and
Shh-positive cells. A few of these subpial
Sst elements even penetrate the striatal marginal stratum (identified as the subpallial mantle zone devoid of the
Nkx2.1 pallidal marker). This
superficial subpallial migration stream (abbreviated as SSpM) becomes much better developed at subsequent stages (E11.5-E14.5). It clearly represents the main pathway for the arrival of diagonal
Sst cells at the striatum (complemented by outside-in radial invasion from the SSpM) and the pallium (first traversing subpially the prospective layer III stratum of the olfactory cortex primordium, which transiently lies at the brain surface at these early stages; Valverde and Santacana
1994);
Sst cells then enter the marginal stratum of the insula and proceed into the isocortex in a gradiental pattern (the hippocampus and entrorhinal cortex seems to be invaded through an analogous caudal marginal stream across the amygdala). Our data indicate that the alternative intermediate and subventricular migration pathways so common for other subpallial migrating cortical interneurons (Anderson et al.
2001; Marín and Rubenstein
2003) are of minor importance in the case of the
Sst cells. In contrast, a subventricular tangential migration route seems relevant for the invasion of the pallial amygdala, in addition to the SSpM (compare Wang et al.
2010; their Fig.
2).
Though the SSpM first crosses the pallidal marginal stratum at the locus of the prospective pallidal olfactory tuberculum (where a subpopulation persists at later stages), the pallidal mantle core remains practically devoid of
Sst cells at all stages, as is true as well for the mature globus pallidus and, partly, for the lateral BST formation (also pallidal in nature). This apparently reveals a persistent non-permissive or repellent pallidal effect on the migrating
Sst cells; this phenomenon may explain as well the paucity of
Sst cells that migrate subventricularly into the striatum and the cortex, since they need to cross the pallidal territory. Moreover, no
Sst cells were ever observed within the preoptic area; this result indicates a clearcut spatial orientation of the SSpM in the opposite palliopetal direction, possibly influenced by repellent signals spreading out of the preoptic area. It may be speculated whether SHH highly present in the preoptic and pallidal environments is repulsive for the
Sst cell population (compare Xu et al.
2010).
Ulterior migration and development of definitive Sst cell populations
Independently of the
Sst cells that migrate tangentially away from the Dg via the SSpM and the deep and superficial amygdalar streams, the Dg histogenetic area also develops radial derivatives that mature locally; these appear thereafter clearly intercalated in the subpallial mantle between the
Sst-negative pallidal and preoptic areas. This locus corresponds to a radial domain placed obliquely along the telencephalic stalk, which has received little embryological attention
per se so far, but is known to exist since the nineties (the anterior entopeduncular area or AEP of Bulfone et al.
1993). This locus also shows characteristic structure in the adult. Previous morphologic analysis suggested that the Dg superficial stratum contains the bed nuclei of the diagonal band, extending obliquely (diagonally) from the amygdala to the medial septum. In its turn, the Dg intermediate stratum deep to the diagonal band forms the classic substantia innominata (which encompasses a major part of the cholinergic basal magnocellular nucleus of Meynert, but also contains other cell types). Finally, the corresponding Dg periventricular stratum contains what can be defined as the medial BST formation. The literature contains a clear idea of medial and lateral BST regions (e.g., in De Olmos et al.
1985), but confusingly also includes a diversity of classifications of individual BST nuclei into these main compartments, but we do not need to discuss these in detail here (compare, for instance, Ju and Swanson,
1989; Ju et al.
1989; Walter et al.
1991; De Olmos et al.
2004; Medina and Abellan
2012). Puelles et al. (
2013) suggested, and we now corroborate, that the lateral BST formation (BSTL) is pallidal, whereas the medial BST (BSTM) is diagonal, as noted in terms of relative abundance of
Sst neurons (see Walter et al.
1991 for the human brain) and differential expression of
Isl1 (only in BSTL). The BSTM also includes some supra- and subcapsular elements classified by some authors within the medial extended amygdala (e.g., Medina and Abellan
2012). Note Medina and collaborators already interpreted part of the extended amygdala as a radial derivative of the Dg, identifying it as ‘entopeduncular area’ in Garcia-Lopez et al. (
2008), or as ‘caudoventral MGE’ in Medina and Abellán (
2012). The ‘extended amygdala’ concept, which is supported by the present results, essentially underlines the developmental sharedness of molecular, histogenetic, and other (e.g., hodologic) properties along the septo-amygdaloid complex of subpallial areas (Fig.
1a–c), as was realized by Heimer himself (personal communication to LP).
We cannot postulate from our material that all Dg derivatives express
Sst, though we presently do not know of other selective markers of potential Dg mantle derivatives other than
Chat, which identifies cholinergic neurons. At the most advanced developmental stage examined by us (E16.5) we saw rather sparse
Sst cells at the level of the diagonal band nuclei, together with a rather dispersed
Sst population within the innominate area deep to them; in contrast, there appears a well-developed population of large and small
Sst cells centered within the periventricular BSTM formation. Importantly, the latter population was found to lie just outside the pallidal domain expressing
Nkx2.1 (or NKX2.1) from E13.5 onward (apparently due to down-regulation of initial
Nkx2.1 expression at the Dg), while it clearly lies within the
Dlx5-expressing subpallial mantle (Figs.
9i, l,
11 and S2); this observation was unexpected, since it is widely accepted in the field that the whole MGE (including Pal, Dg and POA) initially expresses
Nkx2.1, and it was assumed that this state was permanent throughout the MGE. There exists as well a parallel supracapsular migratory stream of hypothalamic
Otp-positive neurons that approach the medial amygdala (Wang and Lufkin
2000; Garcia-Moreno et al.
2010; Morales-Delgado et al.
2011). We checked whether this periventricular
Otp population coincides with the
Sst one at the telencephalic stalk However, the
Otp stream uses the thin
pallial
corridor that directly connects the alar hypothalamus with the pallial amygdala (topologically caudal to the subpallium as a whole), and therefore does not coincide with the
Sst-positive BSTM; it demonstrably passes just outside the
Dlx5-positive domain that contains the subpallial
Sst BSTM elements (Figs.
9i, l,
11b–f; S2a–c).
The cholinergic neurons of the diagonal band and the substantia innominata (basal magnocellular nucleus) probably represent a significant part of the
Sst-negative intermediate and superficial Dg mantle (as suggested by material shown at the Allen Adult Mouse and Developing Mouse Brain Atlases, or García-López et al.
2008). Given the recent unexpected notion that many of these cells (but not all) originate distantly in the ventral pallium and belong to a
Tbr1-expressing lineage that invades the Dg (Pombero et al.
2011), it is open to discussion how far the few
Sst cells present in these superficial Dg areas may represent the local intrinsic diagonal population, in contrast with local and immigrated cholinergic cells. On the other hand, Zhao et al. (
2003) reported a dependency of the subpallial cholinergic cell type on
Lhx8 gene expression, a pattern that is restricted to the Pal and Dg regions of the MGE (as was corroborated by present data; Fig. S1). These authors just assumed a local subpallial origin of the cholinergic cell type, without commenting on the potential role of diverse MGE progenitor subareas in their production, an issue subsequently underlined by Flames et al. (
2007). The demonstrated requirement of
Lhx8 signal in the MGE for the local differentiation of cholinergic neurons (though these potentially may originate elsewhere) may bespeak of an indirect non-cell-autonomous effect due to local MGE mantle conditions controlled by
Lhx8. According to the results of Pombero et al. (
2011) such an effect may act upon cells previously produced in the pallium under control of
Tbr1 and secondarily migrated tangentially into the diagonal part of MGE. If this were not so, we would expect cholinergic cells differentiating massively everywhere the
Lhx8 signal appears in the subpallium, that is, in the whole Pal and Dg domains, which is not the case. This hypothesis accordingly may conciliate the data of Zhao et al. (
2003) and Pombero et al. (
2011). The fact that most basal cholinergic neurons finally populate the Dg area (diagonal band nuclei and basal magnocellular population within the substantia innominata) suggests a differential local microenvironment (more specific than is allowed by the full domain of
Lhx8 expression), and further supports the notion that the Dg is a molecularly distinct from the Pal (Flames et al.
2007), as is also indicated by the restricted local downregulation of
Nkx2.1 and the selective production of
Sst cells. Interestingly, Zhao et al. (
2003) observed that
Sst cells were not affected by the null mutation of
Lhx8.
The supracapsular BSTM arch that forms at amygdaloid, central, and paraseptal periventricular regions of the Dg area (terminology of Puelles et al.
2013; Fig.
1b) appears well populated by
Sst cells from E14.5 onward; it is first visible as an emergent aggregate of
Sst cells at E13.5 (note Bayer,
1987 found in the rat birthdates between E15 and E17 for what Puelles et al.
2013 call the ‘paraseptal BSTM portion’; these rat stages correspond to E13.6–14.8 in the mouse; Clancy et al.
2001). The rather compact
Sst-positive BSTM population stands out from the largely
Sst-negative pallidal BSTL formation, which overlies the globus pallidus (incidentally, we think that our observation that external and internal globus pallidus parts can be distinguished according to their topography relative to the
Sst-positive BSTM arch—the latter covers directly the IGP (see Fig.
11)—may be aclaratory with respect to the traditional concept that the IGP originates in the hypothalamus). The BSTL locus only transiently shows a few
Sst cells that migrate tangentially through the pallidal subventricular stratum into the striatum (or the pallium). The BSTM supracapsular arch extends across the paraseptal Dg area (which neighbours the crossing fibers of the anterior commissure) into the diagonal part of the subpallial medial septum, where individual parts of the BSTM formation have been recognized in adult rodents (these are often described as ‘anterior BST’ in the relevant literature). The amygdaloid BST nucleus is wholly or in part a caudal prolongation of the supracapsular BSTM into the amygdaloid part of the subpallium (Puelles et al.
2013; Fig. S4).
The potential existence of intrinsic amygdaloid Dg derivatives—and
Sst cells—that are not tangentially migrated from outside the amygdala proper was apparently never considered previously. The existence of an amygdaloid diagonal subpallial subdomain first appeared defined in the Allen Developing Mouse Brain Atlas (online since 2009; see also Puelles et al.
2013). Our present results strongly suggest that at least a significant aggregate of
Sst neurons later found within the
lateral subregion of the central amygdaloid nucleus, which is widely accepted as a part of the subpallial amygdala, derives primarily via radial migration from the amygdaloid pole of the Dg progenitor area (CeL in Fig. S3A, B; Ce in Fig. S4; the medial and capsular parts of the central amygdalar nucleus may be pallidal or striatal in origin). Bupesh et al. (
2011a) reported a tangentially migrated contribution to Ce of more rostral parts of the Dg domain (their MGEcv area), but these experiments did not include labeling of the amygdalar end of the Dg complex, and were generally performed at E14.5, which we believe is too late to detect the radial migration we deduce forms the CeL; we start to observe this primordium already at E12.5. The presence of appropriately oriented radial glia fibres across the Ce is shown in our Fig. S4. The diagonal magnocellular nucleus (DgMC; known in the literature as ‘preoptic magnocellular nucleus’) seems also an intermediate stratum derivative of the amygdaloid Dg subdomain. In our identification of this cell group in a position lateral to the locus of the horizontal diagonal band nucleus we followed standard rodent atlas advice (Watson and Paxinos
2010; note Paxinos and Franklin
2013 identified this nucleus as ‘lateral nucleus of the diagonal band’, considering as we do that it is not preoptic in nature). As is shown in several of our Figures (Figs.
1b,
8,
10p, q,
12), our data indicate that the diagonal amygdala contacts directly the striatal amygdala behind the caudal end of the pallidal amygdala (the striatal amygdala is mainly a caudal part of the dorsal striatum, e.g., according to
ER81 and
Six3 expression). This arrangement had not been disclosed so far.
The lateral central amygdaloid subnucleus (CeL) is a compactly
Sst-positive population identifiable as an incipient radial migratory stream at the amygdaloid Dg subarea already at E12.5 (asterisk in Fig.
7f, g). This mass later appears systematically intercalated between the striato-pallidal complex and rostral parts of the pallial amygdala (L/BL primordia); it characteristically preconfigures the future CeL nucleus from E13.5 onward (Figs.
8e–g,
9a,
10g–j, p,
11a,
12c–f). We did not find any reference to this particular subpallial primordium in the embryologic literature, nor had we been aware ourselves of its existence previously. Representative sections shown in our Figs.
10 and
12 illustrate the histogenetic and genoarchitectonic continuity of the CeL primordium with the supracapsular BSTM arch, providing a novel insight into the development of this area. In contrast, Medina and Abellán (
2012) recently interpreted that
Sst cells reach the central amygdala via tangential migration from the ‘medioventral MGE’ (equivalent in their schemata to our paraseptal and/or central Dg; see comments above about similar conclusions of Bupesh et al.
2011a). The amygdaloid end of the BSTM component of Dg also apparently builds separately the well-known amygdaloid BST nucleus (BSTA). We thus believe this would represent likewise a radially migrated Dg-derived entity, rather than a result of tangential migration.
On the other hand, numerous tangentially migrating
Sst cells invade the
pallial amygdala, either via the subpial SSpM, or via a specific, well-developed, amygdaloid subventricular migratory pathway. We could not assess whether these cells originate specifically from paraseptal, central, or amygdaloid parts of the Dg area, or come from all of them. A feature suggesting a general Dg origin is that the earliest cells invading the pallial amygdala via the SSpM were observed at E11.5 (Fig.
5ae, af), and this pathway was still very distinct at E12.5 (superficial arrow; Fig.
7e, f). The subventricular subpallial migratory stream (SvSpM) that likewise targets the pallial amygdala appeared at E12.5 in a relatively more dorsal (perhaps more origin-selective) position (SvSpM; deep arrow; Fig.
7g–k). It is possible that some of these deep SvSpM cells reach the overlying cerebral cortex (entorhinal and hippocampal areas, and perhaps even occipitotemporal areas) via a transamygdalar route. Both streams are less distinct at E13.5 and subsequent stages; this later period is characterized by a substantial invasion by
Sst cells of the lateral (L), basomedial (BM) and amygdalohippocampal (AHi) amygdaloid nuclear primordia. There is the apparent exception of the basolateral, posteromedial cortical, and medial amygdaloid nuclei, which stand out as loci with few
Sst neurons (BL, PMCo, MePV, MA; Figs.
10,
11,
12).
At E16.5, the L primordium appears densely populated and even delineated by
Sst cells (Fig.
12f–h). Horizontal sections show that the rostral tip of L roughly coincides with the locus where the fibers of the anterior commissure reach the amygdala. Similar numbers of
Sst cells label the IPAC nucleus, a part of the extended amygdala (interstitial nucleus of the posterior limb of the anterior commissure; Fig.
12d–f). More rostrally, along the line where the anterior limb of the anterior commissure enters the external capsule (lateral to the striatum) a distinct laminar population of
Sst cells is visible which appears in a similar place as the L nucleus, but lies rostral to it, and is much thinner. We think that it corresponds to the primordium of the bed nucleus of the external capsule (BEC), a formation derived from the ventral pallium that was recently distinguished at this locus (Puelles
2014); this primordium probably also corresponds to the supposedly transient ‘reservoir’ of Bayer and Altman (
1991). The amygdalar L nucleus is held to be a ventropallial derivative on genoarchitectonic grounds, whereas the BL was ascribed to the lateral pallium (Medina et al.
2004). There accordingly appears to exist a preferential invasion of ventropallial cell masses by the tangentially migrated
Sst cells as they pass beyond the pallio-subpallial boundary (this includes the large population invading layer III of the piriform cortex, which also is an integral part of the ventral pallium (Puelles
2014). At postnatal stages, further development of the pallial amygdaloid nuclei and their neuropil leads to substantial dilution or decrease by cell death of the contained population of
Sst cells, though L still retains in the adult more
Sst cells than the BL nucleus (compare L and BL in Fig. S3b).
Our analysis of the invasion of the isocortical plate by
Sst cells was largely centered on the chronology of the arrival of these cells to the different areas. By E16.5 the whole cortex was covered superficially by
Sst cells. As mentioned above, we believe that the main pathway for this tangential migration is the SSpM. Following Puelles (
2014), we interprete that marginal
Sst cells observed over the cortical plate up to E14.5 result subsequently distributed to the subgranular cortical layers (since the cortical plate at E14.5 is largely formed by the prospective layer 5 and layer 6 pyramids). As was previously reported, the major adult population of cortical
Sst interneurons has a subgranular topography (see Fig. S3b). Presumably, labeled embryonic cells occupying a similar marginal position at E15.5 and E16.5 will be distributed to the sparser population later found in the supragranular layers.
Telencephalic subpallial domains and the origin of Sst cells
A subpallial origin of most inhibitory cortical interneurons in the mouse is supported by the fact that practically all cortical interneurons in mice derive from the
Dlx5/6-expressing subpallial lineage (Stuhmer et al.
2002), and such cells are not part of the massive complementary pallial
Emx1-expressing lineage (Iwasato et al.
2000; Gorski et al.
2002). Whereas some pallial explants do give rise to some cells producing GABA in vitro (Götz et al.
1995; He et al.
2001; Bellion et al.
2003; Nery et al.
2003), it is so far unclear whether these results are extrapolable to the in vivo situation (they may result alternatively from differences in genomic regulation created by in vitro conditions, or from a potential initial content of migrating cells of subpallial origin capable of spontaneous or stimulated proliferation in the explanted tissue). In humans, Letinic et al. (
2002) reported that more than half the population of cortical inhibitory interneurons derives from mitoses occurring within the pallial subventricular zone (but see Ma et al.
2013).
Our results corroborate previous data suggesting that SST-positive cells reaching the amygdala derive from the Dg domain (old AEP; García-López et al.
2008; Real et al.
2009; Bupesh et al.
2011a,
b). We provide here a more detailed description of the relationship of the Dg area with
Sst neuron production and migration, which leads us to suggest a Dg origin for most cells expressing
Sst in the telencephalon. This seems partly contradictory with some earlier experimental results, which suggested that the major origin of such cells is the pMGE1 sector of the pallidum, which clearly does not form part of the Dg (Fig.
1c). We think the analysis of this issue (and other analogous issues pertinent to the subpallium) improves by translating the diverse contributions into the comprehensive conceptual framework developed for the subpallium in the Allen Developing Mouse Brain Atlas (reference atlases, online since 2009;
www.developingmouse.brain-map.org); see also Puelles et al. (
2013). This model allows striatal, pallidal, diagonal, and preoptic alternative origins of the
Sst cells to be visualized (Fig.
1a, b), and diverse septal, paraseptal, central and amygdaloid sectors along these domains can be located rather precisely with reference to the septo-amygdaloid axis, as well as periventricular, intermediate and superficial strata with regard to the radial dimension.
Striatal origins
A striatal origin of
Sst cells can be dismissed straightforwardly, because the cerebral cortex of
Nkx2.1
−/− mutants, which essentially lose all MGE progenitor fates, and develop a larger striatum (Pal, Dg, and POA are repatterned as striatum-like domains; Sussel et al.
1999), contains practically no interneurons expressing
Sst at E18.5 (Anderson et al.
2001). Additionally, primary cultures testing postnatal development of interneuron subtypes in the
Nkx2.1
−/− mutant cortex showed absence of somatostatin and parvalbumin interneurons (Xu et al.
2004).
Pallidal versus diagonal origins
Transgenic mice lines expressing GFP or LacZ reporters in the
Nkx2.1Cre-labelled lineage were expected to show co-labeling of all SST-positive cortical interneurons. Surprisingly, many parvalbumin- and SST-positive interneurons did not show the reporter, particularly in superficial cortical layers of
Nkx2.1Cre:Z/EG and
Nkx2.1Cre:R26R-
LacZ mice (Xu et al.
2008). Our observation that
Nkx2.1 expression disappears at the Dg after E13.5—a period when many supragranular interneurons are produced—may be relevant to explain these data. Early Dg-derived cells might be represented instead among the reporter-colabeled infragranular population (see Sousa et al.
2009). Xu et al. (
2008) conjectured that the subpopulation of SST cortical interneurons that was not colabeled in their transgenic mice might derive from the dorsal portion of the MGE (corresponding to the pMGE1 domain; Fig.
1c), on the rationale that
Nkx2.1Cre activity was practically absent in the pMGE1 domain, as opposed to distinct
Nkx6.2 expression. It was accordingly suggested that SST cells lacking β-gal reaction derive from pMGE1, whereas those that coexpress the reporter would derive from a different domain (Xu et al.
2008). This conclusion seemed consistent with earlier in utero fate-mapping studies (Xu et al.
2004; Butt et al.
2005). In any case, these data are also consistent with an origin of both early and late SST neurons within the Dg area (pMGE5), due to the observed local downregulation of
Nkx2.1 expression there.
After initial experimental reports showed that the MGE gives rise to PV and SST cortical interneurons (Xu et al.
2004; Butt et al.
2005; Ghanem et al.
2007), Flames et al. (
2007) performed in utero transplantation of GFP-expressing MGE cells into isochronic host embryos at E13.5. They found that around 30 % of the GFP-positive cells coexpressed SST, as opposed to roughly 50 % of cells coexpressing PV. In the same report, a small cube of GFP-positive tissue obtained from the dorsal pMGE1 subdomain at E13.5 (the pallidal domain closest to the striatum; Fig.
1c) was dissociated, and the cells were grafted into the MGE of an isochronic host embryo. The distribution of GFP-expressing cells in the host mice at P14 revealed that over 60 % of transplanted cells were SST positive, against 7 % labeling obtained with similar grafts of pMGE4 cells; no experiment was performed with cells from our Dg, which corresponds to pMGE5 (Flames et al.
2007; compare Fig.
1c). The dorsal Pal domain was accordingly proposed as the main origin of the SST
+/CR
+ Martinotti cells, and this conclusion was corroborated by other authors (Flames et al.
2007; Fogarty et al.
2007; Wonders et al.
2008; see below). The concern raised by these apparently strong data, is that E13.5 may be a rather late stage for detecting the origin of
Sst cells, since a substantial number of the corticopetal Dg-originated elements have probably slipped beyond the MGE at that stage; moreover, many migrating Dg elements must be present at the transient SSpM stream found crossing the pallidal marginal stratum at E13.5. Accordingly, the cells dissociated at the pMGE1 locus at E13.5 may have included a significant number of Dg-derived
Sst neurons. The conclusion of Flames et al. (
2007) as regards the pMGE1 origin of
Sst cells thus appears weaker than expected, pending further experimental tests with younger pMGE1 cells (e.g., taken at E10.5), and including comparisons with the diagonal pMGE5 area. Nevertheless, though we saw little histologic evidence of a pMGE1 origin of
Sst cells, our results do not allow us to exclude categorically that some
Sst cells may arise at this pallidal subarea.
Wonders et al. (
2008) dissociated dorsal and ventral GFP-positive Pal tissue at E13.5 and injected the cells into newborn cortex, checking their differentiation into SST versus parvalbumin (PV) cells. They concluded that the dorsal Pal was mainly implicated in the production of SST cortical interneurons (63 % SST versus 30 % PV), whereas relatively more interneurons expressing parvalbumin apparently originated at the ventral MGE (31 % SST versus 59 % PV). The location given for the ventral MGE in their Figure 2A is compatible with the topography of the Dg (the depicted explants most probably contained a deep part of the mantle zone, in our opinion). In order to eliminate potential passing cells, Wonders et al. (
2008) explanted again dorsal Pal cells at E12.5 1 h after injecting BrdU to the dams and cultured them in vitro for 10 days. The rationale was to obtain in vitro-differentiated SST neurons double-labelled for GFP and BrdU, that is, derived exclusively from dorsal Pal progenitors They obtained over 60 % double-labeled SST cells for dorsal tissue versus 15 % for ventral tissue (their Fig. 3F). A concern that applies to these data is whether differentiation of dissociated immature cells injected directly into the newborn cortex, or in vitro differentiation of BrdU-labelled progenitors, reproduces faithfully enough the in vivo conditions for the relevant phenotypic decisions. Moreover, the first results may be contaminated by migrating Dg cells passing next to pMGE1 (as in Flames et al.
2007), and the in vitro results may result from re-specification of the explanted dorsal Pal progenitors (e.g., by downregulation of their
Nkx2.1 expression). Appropriate tests should be designed to check these possibilities. In any case, taken at face value, these results suggest that SST cortical interneurons are produced both at the dorsal and ventral Pal, that is at the pMGE1 and pMGE5 domains, thus providing partial support for our results. This suggests that our negative data about
Sst cells originating within the Pal, or their absence inside the early Pal mantle layer, might be caused by late transcription of the
Sst gene in the pMGE1-derived elements.
The homeodomain transcription factor
Nkx6.2 is expressed at the border between striatal and pallidal domains. It is specifically observed in a small subset of neural progenitors localized across the striatal progenitor domain pLGE4 and the pallidal progenitor domain pMGE1 (Stenman et al.
2003; Flames et al.
2007; Fogarty et al.
2007; Sousa et al.
2009). Genetic inducible fate-mapping using
Nkx6.2CreER/+ mice showed EGFP-colabeling in SST-positive cortical interneurons (Sousa et al.
2009). Interestingly, the
Nkx6.2-
Cre/R26R-
GFP transgenic mice line used by Fogarty et al. (
2007) also showed GFP expression in scattered neuroepithelial cells located more ventrally in the caudal ventral MGE (the Dg area), suggesting that the results of Sousa et al. (
2009) are not inconsistent with a double origin of SST cells. This work concluded that SST+/CR− cells generated at E10.5 selectively target deep cortical layers, whereas SST+/CR+ cells are generated at E12.5 and invade superficial cortical layers. The authors did not mention whether cells from this lineage differentiate within subpallial formations after radial migration.
Fogarty et al. (
2007) concluded that only small subsets of the interneurons produced in their
Nkx6.2-
Cre/R26R-
GFP transgenic mice line coexpress GFP and the studied interneuron markers (including SST; their Figure 3A–D). This generates the possibility that most SST cells are derived from other pallidal domains, including the Dg. When
Nkx6.2-
Cre/Nkx2.1-
Cre/R26R-
GFP double transgenic mice (lineage tracing targeting all MGE and POA derivatives) were studied, the majority of SST+ cells (70–80 % of cortical motor and somatosensory SST cells) were marked with GFP. Since
Nkx2.1Cre activity is practically absent at the pMGE1 domain (Fogarty et al.
2007; Xu et al.
2008). This result suggests that most SST-positive cells are generated in the alternative source, the ‘caudal ventral MGE’. i.e., the Dg (Fogarty et al.
2007). Approximately 35 % of
Nkx2.1Cre-
GFP-expressing cells were SST positive and they represent around 70–80 % of this type of interneuron in the motor and somatosensory cortex (Fogarty et al.
2007).
Poor recombination of R26R-GFP with
Lhx6-
Cre was also observed at the pMGE1 (Fogarty et al.
2007). Using both
Nkx2.1Cre:R26R-
GFP and
Lhx6Cre:R26R-
YFP, these authors thus essentially identified the origin of most SST interneurons as corresponding to the ‘caudal ventral MGE’, which represents the Dg domain.
Lhx6Cre:R26R-
YFP transgenic mice displayed in the motor and somatosensory cortex 100 % of SST-positive cells labeled with YFP, as well as nearly 100 % of the PV+ and CB+ interneurons. A high degree of SST and
Lhx6 coexpression was also observed using
Lhx6
+/LacZins mice, and
Sst-expressing cells were drastically reduced (around 93 %) in
Lhx6
−/− mutants (Liodis et al.
2007, Zhao et al.
2003). These experimental approaches suggest strongly that the ‘caudal ventral MGE’ (Dg) domain contains the main SST-cell source; early activity of the genes
Nkx2.1 and
Lhx6 seems required to generate SST-positive interneurons (Sussel et al.
1999; Liodis et al.
2007).
Carney et al. (
2010) analyzed the lineage origin of medial amygdala components, immunolabeling SST cells in combination with an anti-β-galactosidase (β-gal) antibody to visualize nuclear staining in recombined cells from
Nkx2.1
Cre
,
Shh
Cre
and
Gli1
CreER(T2)
:Tau
mGFP
brains. This analysis revealed that 10 and 23 % of
Nkx2.1
Cre
:Tau
mGFP
and 20 and 24 % of
Gli1
CreER(T2)
:Tau
mGFP
recombined cells coexpressed SST at the medial posterodorsal and medial posteroventral nuclei (MePD, MePV) of the amygdala, respectively. This proportion was reduced to 2 and 3 % of cells coexpressing SST and
Shh
Cre
:Tau
mGFP
in the MePD and MePV, respectively (Carney et al.
2010). These results led to the conclusion that SST cells that populate the medial amygdala derive from the ‘caudal ventral MGE’, that is, the Dg.
Flandin et al. (
2010) used
Shh-
Cre mice and a floxed
Nkx2.1 allele to selectively knock-out
Nkx2.1 function from the
Shh-expressing subpallium (POA1 ventricular zone and
Shh-positive migrated cells in the pallidal mantle; eventually, also some ventricular-cell patches in Dg). They observed that the globus pallidus cell population was substantially eliminated, whereas most cortical and striatal interneurons were generated, excepting the striatal cholinergic neurons. In their Discussion, the authors deduce that pMGE1-3 (perhaps also pMGE4) generate most parvalbumin+ and somatostatin+ neo-cortical and hippocampal interneurons, because these domains do not express
Shh. According to our interpretation, the
Sst cells originated from the Dg ventricular domain—the pMGE5 area—may have partly escaped the targetted knock-out, since
Shh expression only appears there in a patchy fashion (Figs.
4, 5, 6, S1; note Flandin et al.
2010 themselves observed reduced integration of
ShhCre activity in the Dg domain in
ROSA;Shh
Cre/+ mice). A 40 % of reduction in cortical SST+ cells was observed in the
ShhCre-
Nkx2.1-floxed mutant (Flandin et al.
2010), consistently with the possibility that this represents the product of a Dg progenitor fraction (40 %) that expresses
Shh, whereas the remaining SST+ elements may be originated by the
Shh-negative Dg progenitors, and are thus unaffected. On the other hand, a strong knock-out effect probably occurred at the septal Dg/POA sectors, where
Shh signal is intense in the ventricular zone (Figs.
4,
5,
6, S1); this is the place where striatal cholinergic neurons are perhaps produced (García-López et al.
2008; Hoch et al.
2015). The reported high activation of
ShhCre in several pallidal domains is difficult to understand, because
Shh is not expressed at all in the pallidal ventricular zone (present results); the missing or reduced globus pallidus phenotype may be due to a strong knock-out effect at the POA1 area, whose ventricular zone expresses strongly
Shh and produces
Shh-positive neurons that migrate into the pallidal mantle. These migrated cells may release SHH, which might be required for the normal development of the globus pallidus (e.g., maintenance of Nkx2.1 activity).
Most of these in vivo and in vitro fate-mapping and transgenic mice studies thus suggest a principal ‘ventral MGE’ main source of
Sst cells, consistent with our Dg area, highlighted in our in situ analysis as the main origin of this interneuronal type. Our study emphasizes the supra- and subcapsular histogenetic and topographic unity of the Dg domain along the septoamygdaloid axis, where various radially migrated derivatives are established (diagonal complex/medial septum, BSTM, amygdaloid BST and CeL nuclei), and from where laterally oriented tangential migrations proceed into the striatum and the whole pallium (sidestepping via the SSpM the apparently non-permissive or repelling pallidum). Several of the cited studies proposed that GABAergic interneurons from the MGE (including SST cells) first invade the pallium around E12.5 (e.g., Corbin et al.
2001; Marin and Rubenstein
2003); our sequential analysis suggests that this process already starts at E10.5 out of the Dg domain, and pioneering cells already reach the cortex around E12.5.
Preoptic origins
The POA region forms the ventromedial
Nkx2.1-positive component of the MGE. It shows three dorsoventrally superposed progenitor domains, identified as POA1 (next to Dg), POA2 (next to the terminal lamina) and POH (preopto-hypothalamic transition area, limiting with the paraventricular alar hypothalamus) (Flames et al.
2007; Bardet et al.
2010; Medina and Abellan
2012; Puelles et al.
2012,
2013). Lineage-tracing experiments using intra-utero electroporation of a
Nkx5.1Cre:R26R-YFP construct demonstrated labeled preoptic GABAergic cells that migrate tangentially into the cortex, septum, striatum, and amygdala (Gelman et al.
2009). These cells frequently coexpressed NPY, but never SST. Moreover, they derive from a
Nkx2.1+
/Lhx6- lineage, which confirms their POA identity (Gelman et al.
2009). Preoptic
Nkx5.1 signal present in postmitotic cells was initially held to be a general POA marker (Wang et al.
2000; Gelman et al.
2009), but Gelman et al. (
2011) later acknowledged that these cells originate specifically at the
Shh-positive POA1 domain, as is corroborated by our present data (see Fig.
6); indeed, we found that this restricted origin includes the median or acroterminal part of pPOA1, which encompasses the terminal lamina (Puelles et al.
2012). Gelman et al. (
2011) stated that the
Nkx5.1-positive elements that migrate out of the pPOA1 area rapidly lose this expression. Comparison of
Nkx5.1 expression with
Shh expression in the subpallial mantle indicated that preoptic
Shh-expressing cells are likewise produced selectively within the pPOA1 area, from where they selectively migrate tangentially into the pallidum (after crossing the Dg domain; note we never saw them entering the striatal mantle). In the chick and mouse, the pPOA2 area expresses
Nkx2.1, but not
Shh (Bardet et al.
2010; Flandin et al.
2010). Gelman et al. (
2011) investigated a subarea of pPOA2 that expresses
Dbx1, from where a separate population of derived GABAergic cells migrate into the cortex, invading predominantly its deep layers. Phenotypic analysis of these cells at P14 suggested that nearly 50 % of the cortical preoptic-derived
Dbx1 cells contain PV (the authors comment that the proportion may be larger, though, since this marker is a late differentiating one, and many of the observed GABAergic
Dbx1-derived cells did not express any peptidic markers) and approximately 25 % contained SST. They deduced that both PV and SST cells are produced at the pPOA2 area. This observation is contradictory with our observations, because we never saw
Sst cells arising within the preoptic area. We would thus predict that the SST elements observed by Gelman et al. (
2011) most probably originated likewise in the Dg area. Primary evidence that can be adduced in favor of this interpretation is the expression of
Dbx1 mapped at E11.5 in the Allen Developing Mouse Brain Atlas (
http://www.developingmouse.brain-map.org); the available sagittal and coronal sections both illustrate that at this stage the expression domain of
Dbx1 extends more importantly across Dg than across the preoptic area. Part of the
Dbx1-derived cells may be thus diagonal in character, rather than preoptic. Gelman et al. (2011) further acknowledged that permanent tracing of
Dbx1-derived cells in Dbx1
Cre; ROSA26
YFP embryos generated ‘a few clones of YFP-expressing cells… in the MGE’. The authors checked the relationship of
Dbx1-derived cells with
Lhx6-derived cells (
Lhx6 is a general MGE marker excluded from the POA; Liodis et al.
2007); the results indicated that some 36 % of the cortical
Dbx1-derived cells co-express
Lhx6, confirming that their origin can not be purely preoptic. On the other hand, Fogarty et al. (
2007) concluded that 100 % of cortical SST cells express
Lhx6. This suggests that the 25 % of SST neurons counted by Gelman et al. (
2011) among the cortical
Dbx1 progeny may originate as we suggest within the
Dbx1+/
Lhx6+ Dg sector of the MGE.
To summarize, in our opinion the published experimental evidence on lineage-tracing of SST-positive cells seems somewhat vague and inconclusive about their precise origin within the MGE, partly due to 1) confusion in the denomination and/or tracing of the relevant progenitor areas, 2) experimental study mainly of suboptimal stages (e.g., E13.5) which allow already migrated, or migrating neurons, to be ascribed to a wrong source, and 3) insufficient use (or dearth) of efficient molecular delimitation criteria for the Pal, Dg and POA subpallial domains and their diverse septoamygdaloid sectors. The present descriptive analysis cannot correct all these problems, but provides a solid basis of data that must be made consistent with any experimental analysis, as long as the existence of invisible Sst cells is not demonstrated beyond any doubt. They also illuminate in a novel way the development of the BSTM, CeL, and globus pallidus formations, as well as the intraamygdaloid relationships of the main subpallial domains.