Sim1-expressing cells illuminate the origin and course of migration of the nucleus of the lateral olfactory tract in the mouse amygdala

The mammalian amygdalar pallial complex consists of a heterogeneous group of nuclei located in the telencephalic temporal pole, rostrally to the caudo ventral hippocampus. They are implicated in reward evaluation of stimuli and emotional learning (Burdach 1819–1822; Johnston 1923; Loo 1930; Weiskrantz 1956; Price et al. 1987; De Olmos et al. 1985, 2004; Alheid et al. 1995; Gloor 1997; Swanson and Petrovich 1998; Amaral et al. 2003; Sah et al. 2003; Phelps and Ledoux 2005; Ledoux 2007; Whalen and Phelps 2009; Rolls 2014, 2015; Olucha-Bordonau et al. 2015; Medina et al. 2017). The nucleus of the lateral olfactory tract (NLOT), the bed nucleus of the accessory olfactory tract (BAOT) and the posteromedial cortical nucleus receive olfactory bulb input (Scalia and Winans 1975; Martinez-Marcos and Halpern 2006; Pro-Sistiaga et al. 2007; Igarashi et al. 2012).

We examine in this work the developing Sim1- and Tbr1-expressing populations of NLOT layer 2, whose postulated neocortical origin (Remedios et al. 2007) seems inconsistent with our results and standard rodent brain atlas data (Puelles et al. 2019a; see “Discussion”). The NLOT is an isolated tri-laminar ovoid cell mass, which lies embedded (after its migration) within the subpallial anterior amygdala, just rostrally to the anterior cortical nucleus or ACo (Krettek and Price 1978; De Olmos et al. 1985, 2004; Martinez-Garcia et al. 2012; Igarashi et al. 2012; Olucha-Bordonau et al. 2015). Nissl staining subdivides the NLOT nucleus in three cell layers. Layer 1 (NLOT1) is a subpial molecular zone with scattered neurons, which receives mitral cell input from the main olfactory bulb (Igarashi et al. 2012). Layer 2 (NLOT2) is a thick and dense corticoid aggregate of pyramidal neurons of medium size with apical dendrites entering NLOT1. Layer 3 (NLOT3) is a multiform layer disposed deep to layer 2. It contains a mixture of small and large neurons (De Olmos et al. 1985, 2004), some of them possibly representing inhibitory interneurons of subpallial origin (i.e., they express Dlx5 and Lhx6; Marín and Rubenstein 2001; García-López et al. 2008). Recently Garcia-Calero et al. (2020) described in addition a cell-sparse NLOT shell formation of genoarchitecturally distinct neurons (Azin2-, Er81, and Cyp26-positive) contributing likewise to layer 3. This shell is continuous with a tail of similar neurons leading backwards along the former migration trail of the NLOT2 up to the medial horn of the basolateral nucleus (BLA). We named the new entity the ‘amygdalo-olfactory stream’ (Garcia-Calero et al. 2020).

The adult NLOT is located as a whole superficially within the subpallial anterior amygdala (AA) domain. Its heterogeneous neuronal composition suggests different neuroepithelial origins of the cells that populate its three layers (Garcia-Lopez et al. 2008; Medina et al. 2017). NLOT2 was hypothesized to represent dorsopallial (neocortical) cells migrated via a characteristic caudal amygdalar migratory stream (CAS), forming ‘a link between neocortex and the amygdala’ (Remedios et al. 2007; Deussing and Wurst 2007; Murillo et al. 2015; Ruiz-Reig et al. 2017). Puelles et al. (2019a) recently expressed doubt about this interpretation of NLOT2, due to the distant position of the molecularly defined dorsal pallium (neocortex) from the amygdala, with large interposed portions of mesocortex and allocortex. The majority of NLOT2 cells express the pallial mantle marker Tbr1, consistently with a pallial origin, whereas other cells express Sim1, a marker otherwise mainly present in the hypothalamus (Fan et al. 1996; Michaud et al. 1998; Balthasar et al. 2005). The transient NLOT2 migratory stream (NLOT2ms), or CAS, appears labelled by the markers Neurod1, Neurod2, Tbr1, Math2 (Neurod6), SCIP and Zic2 (Remedios et al. 2007; Murillo et al. 2015; see also Dach1 signal at the NLOT2ms in the Allen Developing Mouse Brain Atlas). This stream appears in sagittal sections of E14.5–E17.5 mouse embryos, and the definitive nucleus forms within the subpallial anterior amygdala (AA) between E17.5 and E18.5 (Remedios et al. 2007).

Mutant mice lacking Emx1/Emx2, Lhx2, Pax6 or Zic2 functions do not develop the NLOT2 (Remedios et al. 2004; Tole et al. 2005; Murillo et al. 2015). There is so far no rationale indicating how these diverse determinants interact to produce the definitive NLOT structure. We will present a tentative synthesis on how these diverse genes control NLOT2 development, including also the presently studied Sim1 case, since we observed that loss of Sim1 signal likewise impedes the final formation of the NLOT.

The transcription factor Sim1 is expressed in several separate regions of the central nervous system apart the NLOT2, e.g., in the basomedial amygdala, as well as in alar and basal hypothalamic regions (Fan et al. 1996; Wang and Lufkin 2000; Balthasar et al. 2005; Puelles et al. 2012). This gene is necessary for normal development and terminal differentiation of diverse cell types within the paraventricular and supraoptic nuclei in the alar hypothalamic region (Fan et al. 1996; Michaud et al. 1998). Sim1-positive neurons derived from the paraventricular area (Pa) reportedly regulate food intake and energy expenditure, and mice heterozygous for Sim1 are obese (Michaud et al. 2001; Holder et al. 2004; Tolson et al. 2010). The melanocortin-4-receptor pathway (MC4R) apparently mediates the function of Sim1-expressing cells within hypothalamic Pa and amygdalar NLOT (Balthasar et al. 2005). MC4R expression appears only after E18.5 at the post-migratory NLOT2, in parallel with some mesocortical areas and the subiculum (Allen Developing Mouse Brain Atlas). In addition, the Rett syndrome, a pathology belonging to autism spectrum disorders, relates to loss of Mecp2 signal in Sim1-expressing cells at the medial amygdala and the NLOT (Fyffe et al. 2008).

The most probable origin of NLOT2 Sim1-expressing cells is the hypothalamic paraventricular area or Pa (Fan et al. 1996; Puelles et al. 2012). It is already known that Otp-expressing cells of Pa origin migrate into the medial amygdala (Wang and Lufkin 2000; Garcia-Moreno et al. 2010; Morales-Delgado et al. 2011). However, disruption of the Otp gene in mice does not affect the expression of Sim1 in the normally migrated NLOT2 (Wang and Lufkin 2000).

In the present report, we first examined the presumptive extratelencephalic Pa source of Sim1-positive cells, and their migration path across the hemispheric stalk, until they reach the caudomedial pallial amygdala. To this end, we investigated in detail the developmental progression of the Sim1 in situ expression pattern and Sim1-tauLacZ-labeled progeny in correlation with other regional markers. We identified the pathway followed by the Sim1 cells as the hypothalamo-amygdalar corridor (HyA). Next, we explored how such cells, once arrived at the amygdala, incorporate into the NLOT2 migration stream to reach the NLOT2 target.

In the second phase of this study, we related the amygdalar population of Sim1-expressing cells to the rostrally migrating stream of Tbr1-expressing pallial cells which constitute what we call the NLOT2 migratory stream (NLOT2ms), also known as caudal amygdalar stream, or CAS (first described by Remedios et al. 2007). We found that this pallial stream starts at the posterior pallial amygdalar unit we recently defined (Garcia-Calero et al. 2020), which is the anlage of the amygdalo-hippocampal area and the posteromedial cortical nucleus (AHi/PMCo complex). This locus is precisely where the migrating hypothalamic Sim1 cells arrive via the HyA. The NLOT2ms thus carries from its origin pallial Tbr1-expressing and hypothalamic Sim1-expressing cells (or Sim1-tauLacZ labeled ones). This mixed cellular stream advances first within pallial amygdala next to the pallio-subpallial limit, until it reaches the medial aspect of the BLA and BMA nuclei. The stream then crosses the pallio-subpallial border roughly at E15.5–E16.5, and the migrating mass approaches radially its superficial target site within the subpallial anterior amygdala area (AA), rounding up thereafter as the NLOT2. This mechanism thus translocates Sim1- and Tbr1-positive cells from the posterior pallial amygdalar unit (AHi/PMCo primordium) into the superficial NLOT2.

While our results on the NLOT2 migration itself corroborate the relevant data of Remedios et al. (2007), we disagree regarding their characterization of the amygdalar locus of origin as a dorsopallial extension, as well as on their implicit parallel conclusion that neocortex reaches caudally the posterior amygdala (being ‘linked to it’, as is affirmed in their title). We consider in the Discussion the doubtful assumptions held for years in this field of research that possibly caused the cited disagreements. We conclude that hypothalamic Sim1 cells incorporate into the NLOT2 pallial migratory stream at its true amygdalar origin, i.e., the posterior pallial amygdalar radial unit, after arriving there at E12.5–E13.5 via the hypothalamo-amygdalar corridor (HyA). We also observed that homozygous Sim1 loss-of-function (obtained from a Sim1-tauLacZ mouse line) does not alter the initial phase of migration of Sim1-expressing cells into the NLOT2ms, but these cells stop their advance at E16.5 and many of them presumably die without advancing into the final subpallial phase of migration. This also impedes the final steps of migration of the accompanying pallial Tbr1-positive components of the NLOT2, revealing a functional interaction between the mixed Tbr1 and Sim1 cell populations. This pattern (cell death at the time of final differentiation) already emerged at the hypothalamic derivatives of the paraventricular area, using the same mouse line (Michaud et al. 1998).

The first goal of this study was to examine the origin of the Sim1-expressing young neurons that later populate the migrated NLOT2 nucleus. The second goal was to determine how the Sim1 population relates to the stream of migrating pallial NTOL2 cells that occurs in the mouse roughly between E14.5 and E17.5 (Remedios et al. 2007). Our results unexpectedly illuminated the parallel issue of the ambiguous statements made by the cited authors on pallial origin and migration course of the NLOT layer2 cell population. We found that the Tbr1-positive neurons of the NLOT2ms (or CAS) originate from the posterior amygdalar radial unit (Garcia-Calero et al. 2020). Remedios et al. (2007) previously deduced that this component arises from a caudal extension of the dorsal pallium. Their origin and our origin seem topographically identical (present data), confirming an amygdalar and non-cortical source of the phenomenon.

Our recent analysis of the cortical and amygdalar pallium fields (Puelles et al. 2019a; Garcia-Calero et al. 2020; Garcia-Calero and Puelles 2020, 2021) leads us to doubt that it is possible to extrapolate cortical pallium sectors into the histogenetically separate amygdalar pallial field. This now obsolete analytic approach was actually initiated by our group (Puelles et al. 2000, and Medina et al. 2004), and its use is still common (e.g., Desfilis et al. 2018; Ruiz-Reig et al. 2018), but should be discontinued. Along the Discussion, we will argue that all the data considered by Remedios et al (2007) are coherently reinterpretable according to a simpler amygdalar origin hypothesis, without involving wider cortical relationships other than the hippocampal/entorhinal close neighbors. This option is advantageous at least in offering higher consistency with the known anatomy of developing and adult rodent brains, by recognizing that the primordium of the neocortex is always distant from the amygdalar field, due to their absolute separation by interposed mesocortical and allocortical pallial domains (Puelles et al. 2019a; Garcia-Cabezas et al. 2019; Pattabiraman et al. 2014; Bayer and Altman 1991; Swanson 1987).

With regard to Sim1 cells, we described a Sim1-expressing hypothalamo-amygdalar corridor (HyA). Fan et al. (1996) first mapped it at E10.5 and E12.5, but left it unnamed (see their Figs. 1n, 15a). In agreement with tentative schemata of Puelles and Rubenstein (2003, 2015) on this point, we think that the HyA is a dorsal prolongation of the hypothalamic paraventricular area (Pa) that results co-evaginated into the early telencephalic vesicle jointly with the neighboring rostrodorsal ‘telencephalic’ part of the diencephalic prethalamic eminence (PThEt; Puelles 2019). This implies a novel concept which may be of interest in comparative neuroanatomy, namely the existence of an alar hypothalamic subdomain that stretches into the telencephalic roof without losing its original molecular character (Fig. 15a). This hypothesis explains the course of HyA through the floor of the interventricular foramen and of the terminal sulcus (Fig. 15a). The HyA always lies next to neighboring subpallial formations (e.g., main BST nuclei, supracapsular BST and MeA) and finally reaches the posterior pallial amygdala at the end of the chorioidal fissure (Fig. 15a).

The Pa/HyA progenitor domain represents the apparent origin of all forebrain alar Sim1-expressing cells (there are separate basal ones; Fig. 1n). This implies that the derivatives of this area must include the population that reaches the amygdala and eventually enters the NLOT2. The HyA can be understood as the migratory pathway for the arrival of paraventricular Sim1 and Otp cells to pallial or subpallial parts of the amygdala, as was already demonstrated for Otp cells (Wang and Lufkin 2000; Garcia-Moreno et al. 2010; Morales-Delgado et al. 2011; Morales et al. 2021; present results; see also Bardet et al. 2008 for a comparative perspective). Considering that various names applied previously to this pathway were conceptually inappropriate (see below), we renamed it ‘hypothalamo-amygdalar corridor’ (HyA), emphasizing its hypothalamic origin and molecular profile, as well as its amygdalar ending.

The topographically caudal end of the HyA corridor (which topologically is actually its dorsal end, where the hypothalamus reaches the chorioidal roof plate; Fan et al. 1996) allows the access of Sim1 cells to the extreme caudomedial part of the pallial amygdala (namely its posterior radial unit, or prospective AHi/PMCo complex). The invasion occurs specifically at its rostromedial subdivision (AHiRM; Garcia-Calero et al. 2020), which characteristically protrudes into the medial brain surface with the underlying PMCo nucleus, behind the subpallial MeA (p; MeA; Fig. 15a; De Olmos 2004).

Analysis of the posterior amygdalar radial unit in horizontal sections at E13.5 and E14.5 revealed that the migrated Sim1-expressing cells first shift laterally within AHi, passing behind the MeA, and adopting a new position within the rostrolateral AHi subdivision found lateral to the MeA (Fig. 15b). Here they join the Tbr1-positive local elements that start to migrate into the arc-shaped NLOT2 migration stream. The latter, aptly (though somewhat ambiguously) named the ‘caudal amygdaloid stream’ (CAS), arrives at its target between E16.5 and E17.5 (Remedios et al. 2007). Our NLOT2ms results corroborate absolutely these earlier data, adding some points of interest.

We examined in more detail the course of the arc-shaped Sim1-expressing NLOT2ms/CAS using Tbr1 immunoreaction and Dlx5 ISH, which respectively mark the pallium versus subpallium domains. We thus were able to divide the migration into successive pallial and subpallial phases. The first phase traverses several parts of the pallial amygdala, always next to the pallio-subpallial boundary, proceeding orthogonally to the local radial glia between E14.5 and E15.5 (as already noted by Remedios et al. 2007, and corroborated by us; Fig. 15b). Once the stream reaches at about E15.5 the medial side of the BLA and BMA nuclei (basolateral and anterior radial amygdalar units; b, a; in Fig. 15b; Garcia-Calero et al. 2020), the NLOT2ms/CAS proceeds into its second subpallial phase. To this end it crosses the pallio-subpallial border into the subpallial anterior amygdala (AA), and advances therein in a radial course into its target locus (Fig. 15b, d), surrounded by dispersed Dlx5-, Six3-, Pax6-, calbindin, and Lhx9/Lhx2- expressing cells. Here the migration ends, and the NLOT nucleus forms. Remedios et al. (2007) observed these two phases as regards the changing relationship with radial glial processes (tangential to radial), but they apparently did not notice that the change also coincides with the pallial versus subpallial character of the tissue surrounding the migrating cells. We think this result has relevance towards understanding the roles of diverse genes known to control this migration (see below). The decision point where the cells change into the second phase lies next to the anterior radial unit of the pallial amygdala (prospective BMA/ACo nuclei), and in the medial vicinity of the BLA nucleus (intermediate mass of the basolateral unit; Garcia-Calero et al. 2020).

The convergence of the hypothalamic Sim1-expressing HyA pathway with the posterior amygdalar source of migrating Tbr1-positive pallial cells (Fig. 15b) clearly identifies the origin of the NLOT2ms/CAS migratory process as the posterior unit of the pallial amygdala. The latter contacts caudally the hippocampus and entorhinal cortex, both of them allocortical (De Olmos 2004; Franklin and Paxinos 2013; Puelles et al. 2019a; Garcia-Calero et al. 2020). In contrast, Remedios et al. (2007), referring to our amygdalar AHi/PMCo complex, reported that the migration originates from a caudal part of the dorsal pallium, thus pretending to establish a ‘link between the amygdala and neocortex’. This conclusion is clearly contradictory with the easily observable caudal direct boundary of the AHi posterior unit of the pallial amygdala with the hippocampus (as appears reflected in its conventional name, ‘amygdalo-hippocampal transition area’). Several other commentaries, reviews or reports accepted the Remedios et al. (2007) interpretation of the CAS migration as a cortical dorsal pallium phenomenon without raising doubts or objections (Deussing and Wurst 2007; Subramanian et al. 2009; Murillo et al. 2015; Ruiz-Reig et al. 2017; Chou and Tole 2019). In a recent review of cortical models featuring classically defined concentric ring-shaped domains (Puelles et al. 2019a) we did express doubts about this hypothesis of Remedios et al. (2007), since all these cortex models showed the neocortex to be totally separated from the amygdala by the massive outer allocortical ring and the thinner mesocortical ring.

We studied carefully the Remedios et al. (2007) paper, exploring various ways to understand what seemed a remarkable error of interpretation coming from first-rate researchers. We reached the conclusion that the paper includes various sorts of doubtful assumptions and interpretive errors that were actually widely shared (including by us) in the field of pallial studies back in the first years of 2000. We will divide our resulting interpretation into three levels of analysis.

First, we will examine whether there is discrepancy between Remedios et al. (2007) and present results about the observed embryonic location and adult identity of the origin of the NTOL2ms/CAS phenomenon. The possible use of ambiguous terms might have approximated at least implicitly our respective interpretations. It seems clear that we both see the origin of the NTOL2ms at exactly the same place of the embryonic telencephalon, but for various reasons we classify this locus as posterior pallial amygdala, whereas Remedios et al. (2007) classified it as dorsal pallium. We cannot be both right.

In a second, more semantic level of analysis, we consider whether there existed in 2007, and still exist today, solid grounds to classify the particular telencephalic locus where the CAS originates as ‘dorsal pallium’. This question possibly needs a ‘yes and no’ answer, because such a notion was indeed possible, and even conventional, during a number of years, but this has changed now, turning it into a risky idea. We also must address the scientific meaning of ascribing an embryonic brain pallial locus to a given pallial cortical sector, such as the ‘dorsal pallium’. The molecularly distinct pallial sectors with which Remedios et al. (2007) dealt were partly corroborated postulates within a conceptual pallium model that is now 20 years old (Puelles et al. 2000; Medina et al. 2004; Tole et al. 2005, and has recently evolved for good reasons to different postulates. The relevant assumptions have indeed changed significantly in recent years (Puelles 2014, 2017; Puelles et al. 2016b, c, 2017, 2019a).

Finally, we will discuss the notion of Remedios et al. (2007) that the dorsal pallium, the primordium of neocortex, actually extends caudalwards between the caudal ends of hem and antihem to establish as its caudalmost subregion the locus of CAS origin. This idea seems inconsistent with available developmental and adult rodent brain neuroanatomic knowledge. Tradition in this field does not detect any direct contiguity whatsoever between neocortex and pallial amygdala (Puelles et al. 2019a; Garcia-Cabezas et al. 2019; Pattabiraman et al. 2014; Bayer and Altman 1991; Swanson 1987).

Proceeding now to our first topic, we inquire whether we really disagree about the embryonic location of the source of NLOT2ms or CAS. As stated above, interpretation of our results within our recent radial amygdala model (Garcia-Calero et al. 2020) allows little doubt that this origin lies in the posterior pallial amygdalar unit (AHi/PMCo complex). On the other hand, Remedios et al. (2007) vaguely described this locus as the ‘caudal telencephalic neuroepithelium’ or ‘the caudal extreme of the telencephalon’ (for instance, in the legend to their Fig. 2, while locating their electroporation experiment). They probably referred to the local end of the lateral ventricle. According to us, the neuroepithelium at the front of the end of the ventricle is in large part pallial amygdalar (note subpallial amygdala also steps in somewhere), whereas the neuroepithelium that lines caudally the end of the ventricle is hippocampal and entorhinal (check the plates in Garcia-Calero et al. 2020). Remedios et al. (2007) seem to have lacked these anatomic references. The description of their important electroporation experiment did not include any suggestion that the experimental site lies rostral to the lateral ventricle, as is clearly visible (see their Fig. 2). This anatomic locus is systematically ascribed to the amygdala in all developing rodent brain atlases (e.g., Altman and Bayer 1995; Alvarez-Bolado and Swanson 1996; Jacobowitz and Abbott 1997; Foster 1998; Paxinos et al. 2007; Ashwell and Paxinos 2008).

After proceeding to an analysis of ventral, lateral, medial and dorsal pallium markers which was concluded to exclude the first three options, Remedios et al. (2007) jumped to the conclusion that the caudal locus of CAS origin had to be an integral part of the remaining option, the dorsal pallium, adducing some molecular evidence (considered below) and the existence of a gap between the hem and antihem organizer territories at the periphery of the pallium, which might allow the dorsal pallium to extend into the area of CAS origin. The ambiguous term ‘caudal amygdaloid stream’ (CAS) they used does not refer to an amygdalar origin of the NLOT2, but to a caudal dorsopallial (neocortical) origin of this migrated, finally ‘amygdalar’ structure. The authors apparently did not conceive any primary amygdalar non-cortical pallial neuroepithelial region at the ‘caudal telencephalic neuroepithelium’, though their own radial glia preparations clearly suggested it (their Figs. 7a–c, 8h). They apparently believed, probably inspired by our earlier study of ventral and lateral pallium parts of the amygdala (Medina et al. 2004), or by other contemporaneous sources, that all parts of the pallial amygdala originate in one of the four postulated cortical pallial sectors, and secondarily migrate to distinct final amygdalar sites. To illustrate this notion, Remedios et al. (2007) drew dashed lateral and ventral pallium arrows in their Figs. 1 and 8, complemented by their dorsopallial CAS, jointly depicting the hypothetic paths of amygdalopetal migrating cells originated in various parts of the cortex. Similar arrows were reproduced by Deussing and Wurst (2007; their Fig. 1), reflecting editorial approval of this notion. The whole Discussion of Remedios et al. (2007) developed thereafter out of the initial ‘dorsal pallium’ diagnosis. We did not find any place where Remedios et al. (2007) considered even hypothetically that the CAS locus of origin might be intrinsically pallial amygdalar (this option might have been qualified to represent an analog of cortical dorsal pallium in terms of its molecular profile). We understand that Murillo et al. (2015) and Ruiz-Reig et al. (2017) later simply accepted the scientific authority of the respected research group and Nature Neuroscience.

We initially also thought in much the same way, since we also shared the assumption of varied cortical origins of distinct amygdalar components, all the way from Puelles et al. (2000) to Puelles et al. (2016b). Subsequently, we started to doubt such interpretations, as we gradually assimilated various contradictory data accruing from the work of Gorski et al. (2002), Puelles (2014) and Puelles et al. (2016c), as we will comment below. We have since realized that this conventional viewpoint on the mode of pallial amygdala formation is unfortunately erroneous. This happens, partly, but importantly, because all studies performed in this period confided on coronal sections oblique to radial amygdalar structure. In such sections, we never see the true amygdalar ventricular zone, because the latter appears in other coronal sections caudal to the amygdala, precisely at the ‘caudal telencephalic neuroepithelium’ of Remedios et al. (2007). That is the simplest reason why these authors missed the true amygdalar identity of the CAS origin.

We thus conclude that Remedios et al. (2007) did not discover or deduce that the CAS/NLOT2 origin is at the posterior part of the pallial amygdala because they ignored (as did everybody else at the time) that the latter, equivalent to their ‘caudal neuroepithelium’, is a local pallial amygdalar progenitor domain, that does not derive from the cortical dorsal pallium, or from any other cortical subregion. The cause of this error was the widely shared assumption that cortical pallial sectors produce claustral and amygdalar nuclear populations that migrate long distances into extracortical adult positions, such as the amygdala. This false assumption, tied methodologically to the use of coronal sections, misled Remedios et al. (2007) into the simplistic and, as it turns out, false conclusion involving the actually distant dorsal pallium as the origin of the pallial cells of the CAS/NLOT2.

The implications of such biased thinking, given the real distance of the amygdalar target from the molecularly delimited dorsal pallium (Puelles et al. 2019a), are reduced to the absurd in our Fig. 15c, d (red dots and lines). The data produced by Remedios et al. (2007), and notably their excellent electroporation experiment (whose position we represent as a black dot within the posterior amygdala; Fig. 15c, d), do not support long-distance migration of the CAS. Such translocation would be needed if these cells really came from the dorsal pallium (red dot). The site labeled by electroporation clearly corresponds to a posterior amygdalar pallial site, not to any distant dorsal pallium cortical sector. We will come back to this conclusion under the third point of discussion.

As a second topic of clarification, we examine now how solid were the grounds used by Remedios et al. (2007) to classify as ‘dorsal pallium’ the caudal telencephalic locus where the CAS/NLOT2ms originates, and whether their argument remains valid today. We think that they reasonably excluded the ventral and lateral pallial sectors, but not so the medial pallium, as we will see below. We also address the meaning and interest of ascribing an embryonic pallial locus to a given pallial sector postulated within a model (Puelles et al. 2000).

In the molecular era of pallium developmental studies, started roughly between 1998 and 2000, ascriptions of pallial elements to the postulated four sectors of the Puelles et al. (2000) pallium model proceeded by demonstrating a particular molecular marker profile shared by the corresponding territory. This was the case even when the pallial portion of interest was not cortical in structure (there are pallial parts of the septum and of the amygdala which essentially lack cortical structure; see also more general comments on comparative studies of non-corticoid pallium regions of non-mammals in Puelles 2017 and Puelles et al. 2017).

Remedios et al. (2007) employed the standard approach in their Results and Discussion, but did not consider first whether their area of interest—the ‘caudal telencephalic neuroepithelium’—was cortical or nuclear (i.e., amygdalar) in nature. This possibly distracted their attention from the theoretical possibility of a primary amygdalar classification of the CAS origin. Eventually, after thinking they had excluded the molecular profiles of the ventral, lateral and medial cortical sectors, they concluded that the data available pointed clearly enough to the single remaining option, the ‘dorsal pallium’. We think in retrospect that the markers they considered specific of the dorsal pallium, mainly Emx1 and Lhx2 (as well as Neurod1/2 and Tbr1) were not really ‘dorsal pallium’ specific. These signals also extend importantly at appropriate early stages within the mediopallial primordia of hippocampal and entorhinal cortical areas, even though the hippocampal hem is indeed negative for Lhx2, as was pointed out by these authors (see our Figs. 9d, e, l, 10, 11a–j, 12). These mediopallial or allocortical areas are close neighbours of the amygdala (see Hi, ERh, MP; Fig. 12a, b, d, j–p), much closer than the dorsal pallium or neocortex (DP). This close relationship was also evident in some of the illustrations of Remedios et al. (2007), such as their Figs. 1b, c, 2a, 3b, i–m, 4t, 8a, as well as in various other reports of the same group illustrating Lhx2 expression. The lapse in recognizing a lack of specificity of these markers is difficult to understand. Probably other less relevant comparative considerations, like the sharing of a reelin/Dab1/Cdk5 neuronal migration control mechanism at both the CAS and the dorsal pallium, or their new data on a gap between the hem and antihem organizers, added salience to the ‘dorsal pallium’ hypothesis. The interesting comparison made in their Fig. 4w between Wnt2b (a hem marker) and sFrp2 (an antihem marker) against the pallial Emx1 signal should have been accompanied by a similarly oriented image comparing the same landmarks with Lhx2. Note the Lhx2 pattern shown in their Fig. 4t is quite different from that of sFrp2 in Fig. 4r. To identify the hem versus antihem is not the same as identifying the medial pallium versus the dorsal pallium, because these specializations lie topologically outside the medial and ventral pallium allocortical regions, and do not contact the dorsal pallium at all (Fig. 15c). We thus believe Remedios et al. (2007) did not attend to this diagnostic task optimally, according to available evidence. In our opinion, the authors should have concluded that the area of interest could as likely be either medial pallium or dorsal pallium. Between these two possibilities, the most parsimonious option was to ascribe the caudal area of interest to cortical medial pallium, rather than to dorsal pallium, due to the observable vicinity of mediopallial cortical areas to the area of interest. More indirectly, the then known joint requirement of hem-related Emx1/Emx2 function for NLOT development also bespoke of a mediopallial relationship (Tole et al. 2005; Shinozaki et al. 2004; Suda et al. 2010). This viewpoint was indeed reached 7 years later by Abellán et al. (2014), who emphasized a number of shared LIM-homeobox genes between the medial pallium and the AHi/PMCo amygdalar complex, that is, the amygdalar source of the CAS.

We next consider critically the rationale of ascribing any amygdalar pallial parts to one of the four pallial cortical sectors defined by Puelles et al. (2000). We presently hold that, in any case, the seemingly reasonable pallial structure assumptions that Remedios et al. (2007) employed back in 2007 are now outdated (followed also, implicitly, by Deussing and Wurst 2007; Subramanian et al. 2009; Murillo et al. 2015; Puelles et al. 2016c; Ruiz-Reig et al. 2017; Chou and Tole 2019).

Historically, Puelles et al. (2000) and Medina et al. (2004) used a peculiarity in Emx1 expression (its absence in a previously non distinguished portion of cortical pallial mantle lying next to the subpallium) to develop a tetrapartite mouse/chick pallium model that left behind the classic tripartite pallium model (the latter only contemplated lateral, dorsal and medial constituents; see Striedter 1997). Four (ventral, lateral, dorsal and medial) pallium sectors were distinguished at middle levels of the hemisphere. Though it was based on rather sketchy molecular data, this model apparently was valid ab initio for tetrapods (Smith-Fernández et al. 1998; Puelles et al. 2000). Corroborations of the new tetrapartite model (demonstrating presence of the Emx1-negative VPall) accrued thereafter in various anamniote species, as well as in man, giving ample credibility to the model. It was initially unclear whether the four pallial sectors extend throughout the whole length of the telencephalic pallium, that is, e.g., whether they reach the caudal amygdalar pole in a parallel arrangement of longitudinal partitions, as had been theorized previously by Kuhlenbeck (1973), Holmgren (1925), and other classic authors. Recently we have learned that the four cortical pallial sectors do not reach the pallial amygdala, due to the emerging alternative concentric ring-shaped arrangement of allocortical and mesocortical areas around a central iso/neocortical island (Garcia-Cabezas et al. 2019; Puelles et al. 2019a; see other references therein, an idea ranging back at least to Swanson 1987). Medina et al. (2004) first addressed the partial demonstration of parallel longitudinal pallial cortical sectors expected to reach the amygdala, using chosen markers thought to define the novel ventral and lateral cortical pallium sectors. The conclusion was that, according to the distribution of such markers, both ventropallial and lateropallial cortical sectors seemed to extend caudalwards into the pallial amygdala, each having specific radially migrated amygdalar nuclear derivatives. This tendentious, rather preconceived conclusion inspired in what one sees in coronal sections lamentably soon became a sort of established fact, or strong assumption, in the field, leading to the non-fundamented stream arrows drawn by Remedios et al. (2007) and Deussing and Wurst (2007).

Indeed, the illustrations of the Medina et al. (2004) report, based on standard coronal sections and artist drawings, arbitrarily suggested that the sources of any claustral or amygdalar nuclear structures lay at the ventropallial or lateropallial cortical ventricular zones appearing in the same sections, next to the pallio-subpallial boundary. This was an error inconsistent with the glial structure already known at the time, because no radial glia processes extend from the cortical ventral pallium into amygdalar pallial territory (the ventropallial glial processes rather plunge straightforwardly into the local olfactory cortex). Nobody recognized this conceptual error at that moment, due to the conventional massive use of coronal sections and the mentioned assumptions. These induced subliminally the belief that the nuclei you see in one coronal section come from the ventricular zone you have in that same section (disregarding other possible origins more rostrally or caudally in the brain).

It so happens that telencephalic coronal sections are oblique by some 45 degrees to amygdalar radial glial structure, because the ventricular zone where amygdalar nuclei actually arise lies behind the amygdalar nuclei, precisely at the ‘caudal telencephalic neuroepithelium’ whose glial structure was examined by Remedios et al. (2007), whereas the corresponding pial surface lies ventrally to the amygdala (Garcia-Calero et al. 2020; Garcia-Calero and Puelles 2020). Many authors accepted at face value the fore mentioned arbitrary conclusion about cortical origins of amygdalar cell masses held to migrate via ‘ventropallial and lateropallial migration streams’ into secondary amygdalar positions. The list of reports incurring in this inherited error includes Remedios et al. (2004, 2007), Tole et al. (2005), Subramanian et al. (2009), Martínez-Garcia et al. (2007, 2012) and Olucha-Bordonau et al. (2015), and there surely are other cases, including Puelles et al. (2016c).

A lateropallial migration stream that ends superficially in a dorsal part of the olfactory cortex, as proposed by Puelles et al. (2000) and Medina et al. (2004), was never demonstrated in terms of radial glia or experimental analysis. A later re-examination of this conundrum led LP to redefine the concept of lateral pallium as a radially organized ‘claustro-insular’ complex, now believed to represent the true lateral pallium, whereas the whole olfactory cortex resulted ascribed to the updated ventral pallium (Puelles 2014). Remarkably, the new lateropallial cortico-nuclear unit also appears with corresponding topology and selective gene markers in chick and reptiles, where nobody had previously expected to find claustrum and insula homologs (Puelles et al. 2016b, 2017, Puelles 2017; these surprising results were recently corroborated with a transcriptomic approach in reptiles by Tosches et al. 2018 and Norimoto et al. 2020). However, as concluded in the cited reports, as well as in Puelles et al. (2019a), this updated lateral pallium has no molecularly identifiable derivative in the mouse amygdala.

Contrarily, the postulated ventropallial migration stream is easily observable with radial glia stains (Puelles 2014; Garcia-Calero et al. 2020), but throughout its whole length it leads superficially into the olfactory cortex, not into the amygdala (see experiments labeling radial glia in Garcia-Calero et al. 2020). Another conflict with the supposed amygdalar ventropallial derivatives had appeared previously in the work of Gorski et al. (2002), who found that all parts of the pallial amygdala contained Emx1-derived progeny (labeled with Emx1-LacZ). According to the pallial model in vogue this marker supposedly had to be absent in true ventral pallium derivatives, held to include most amygdalar nuclei. Medina et al. (2017) and Desfilis et al. (2018) reproduced the Gorski et al. (2002) result in a lizard, and proposed a possible distinct ‘ventrocaudal pallial sector’, restricted to the amygdalar domain, as the origen of amygdalar Emx1 cells (such solution was already mentioned as a possibility in mouse in Puelles et al. 2016c; however, this hypothesis has not yet been correlated with the recent radial model of the mouse amygdala which we presently use; Garcia-Calero et al. 2020). At the time of the Gorski et al. (2002) paper, we did not understand what this sharp contradiction might mean, but we were motivated to accommodate these and other discrepant data in a better model. Unfortunately, we also remained dominated by coronal section-driven assumptions for years.

Our last coronal section-based effort to define the ventropallial contribution to the amygdala used analysis of the theoretically conclusive Dbx1-derived progeny, since Dbx1 expression was held to be strictly restricted to cortical ventral pallium progenitors (Medina et al. 2004; Puelles et al. 2016c). What we did not consider significant was that Dbx1 also appears expressed primarily, and, actually, more importantly, at the ‘caudal telencephalic neuroepithelium’ (Bielle et al. 2005; Teissier et al. 2010), probably because we thought we already knew where the amygdalar nuclei came from. Our standard coronal analysis surprisingly produced various results that seemed contradictory, inconclusive, or difficult to explain. This included parts of olfactory cortex (and the whole olfactory bulb) which lacked Dbx1 derivatives largely, or altogether, or only halves of some amygdalar nuclei appearing to be positive. It was also unclear that the observed Dbx1-LacZ labeled amygdalar elements had actually migrated through the distinctly labelled ventropallial migration stream seen reaching olfactory cortex just outside of the pallio-subpallial boundary and of the pallial amygdala. Importantly, the amygdalar periventricular masses at the back of the amygdala oddly belonged likewise to Dbx1 progeny (partly seen, for instance, in Figs. 3c–e and 4a–e in Puelles et al. 2016c). These data were unexplainable by the standard model we were using. This led to offering two alternative interpretations (so far both unverified, and maybe both wrong), proposed respectively by Medina and Puelles, since we could not convince each other about either interpretation.

Eventually, we realized that the source of all these incoherencies of the tetrapartite pallial model as regards derivatives in the pallial amygdala were two false assumptions. (1) The assumption that coronal sections of the hemisphere identify the real neuroepithelial origin of amygdalar pallial formations, and, (2) the assumption that a ‘molecularly similar’ amygdalar nucleus should derive from the comparable cortical sector, without showing anything of its primary cortical nature. It was simply wrong and misdirected to expect cortical pallial sectors to produce parts of the nuclear amygdalar domain. The latter clearly represents a different sort of pallium adjacent to the cortex, which has its own progenitors, as happens likewise at the pallial septum. The issue of shared gene markers was just a distracting circumstance, whose significance seemed more important when we only were able to study a few markers. Nowadays we can inspect thousands of gene markers in databases, and know that distinct cortical or nuclear fates result from complex combinations of hundreds of gene functions. The existence of a few shared markers is not important, though some interesting correlations may be drawn out of them (Nieuwenhuys and Puelles 2016). A case in point is the fact that olfactory bulb efferents innervate both the olfactory cortex and several superficial amygdalar nuclei (although these do not derive from the former’s progenitor domain). This might be due to effects of shared genes guiding these fibres. Our point is that there is no predictive value of interest in explanations of amygdalar nuclear structure that simply are based on a presumed cortical origin.

This novel analytic light put us in the search of credible origins of the amygdalar nuclei, and we were helped by the idea to use as guides the radial glia processes (first investigated in the area of interest by Remedios et al. 2007, and later also by Bupesh et al. 2011). Many hours of work afterwards, and having checked carefully over 80 promising amygdalar gene markers, and similar numbers of cortical markers, we produced first our updated version of the previously existent concentric ring model of cortical pallium (see references reviewed in Puelles et al. 2019a, as well as Barbas 2015; Garcia-Cabezas et al. 2019). This new cortical model (initiated in Pattabiraman et al. 2014) fills holes identified in our previous updated model (Puelles 2014). The concentric cortical map already showed that the pallial amygdala is a different neuroepithelial field that is adjacent, but separate, from the cortex. The two adjacent fields nevertheless do clearly share general pallial molecular characteristics, and thus are both pallial (Puelles et al. 2000).

This first step allowed us conceptually to investigate next what happens inside the isolated amygdalar pallial field. We eventually produced a radial model of the pallial amygdala (Garcia-Calero et al. 2020), which postulates that all amygdalar nuclei originate in different subareas within the ‘caudal telencephalic neuroepithelium’ locus identified by Remedios et al. (2007). The model derives all amygdalar nuclei from no less than 9 molecularly different amygdalar radial units, subsumed under 5 macrounits (Fig. 15a), each of which develops molecularly distinct periventricular, intermediate and superficial strata. This new, much more complex amygdalar model now explains easily the previously contradictory data of Gorski et al. (2002), Puelles (2014) and Puelles et al. (2016c). Garcia-Calero and Puelles (2021) reexamined and reinterpreted the Dbx1 data of Puelles et al. (2016c) within the novel radial amygdalar model. They reached the conclusion, consistent with both radial glia pattern and Bielle et al. (2005), that amygdalar Dbx1 derivatives are intrinsic to the amygdalar field, and cannot be assimilated to cortical ventral pallium Dbx1 derivatives. This is precisely the conclusion that also satisfactorily explains the Emx1-LacZ data in Gorski et al. (2002).

In the wake of these advances, the effort made by Remedios et al. (2007) to trace the amygdalar CAS migration to the ‘dorsal pallium’ cortical sector seems outdated and devoid of explanatory meaning, when reexamined under the light of the radial model (and the abandonment of coronal sections). There is simply no interest in extracting amygdalar pallial nuclei from parts of the cortical pallium, because no part of pallial amygdala (including the CAS/NLOT2ms) comes out of any cortical domain. Moreover, the concentric ring cortical model provides wider significance and novel lines of analysis for the hem and antihem functions, admitting other possible organizer sites as well. Note our Fig. 15c illustrates that the dorsal pallium does not even come close to the hem and antihem organizers, because the inner mesocortical and outer allocortical cortical rings are interposed (i.e., these organizers lie topologically outside the entire concentrically organized cortex; interestingly, the gap illustrated by Remedios et al. 2007 that separates ‘longitudinally’ the hem from the antihem relates to the new amygdalo-allocortical or amygdalo-entorhinal border proposed by us). This new topologic concept of the cortical field and its organizers recovers the classic notions of allocortex and mesocortex (the two concentric peripheral cortical rings) as the natural evolutionary environment of the central neocortex/isocortex island (Puelles 2001, 2011; Puelles et al. 2019a).

It was thus a commonly held, but not wholly supported, or solidly evidence-based, assumption of many workers in the pallial developmental field at the time of the Remedios et al. (2007) study that cortical pallial sectors might or must have derivatives in the amygdala. It is important to realize that all the referred amygdalar molecular mappings aiming to establish cortical origins of amygdalar parts stood systematically on coronal sections, which are oblique to amygdalar radial glia.

The new amygdala model agrees with the facts that the respective cortical and amygdalar structures are different (layers versus nuclei). Note that amygdalar neurogenetic stratification is outside in, whereas the cortical one is inside out (Garcia-Calero and Puelles 2020), not to speak of differential connections and functions. Consequently, pallium models postulating a number of parallel longitudinal pallial sectors extending throughout the whole telencephalon are presumably on the wane. We should study the cortical pallium and the amygdalar pallium as causally separate complex fields, each with quite diverse molecular profiles, and probably with correspondingly diverse causal mechanisms, irrespective of the known partially shared gene patterns.

As regards our third discussion point sketched above, we refer to atlases of developing rodent brains (e.g., Altman and Bayer 1995; Alvarez-Bolado and Swanson 1996; Jacobowitz and Abbott 1997; Foster 1998; Paxinos et al. 2007; Ashwell and Paxinos 2008), which unanimously identify the posterior pallial region (AHi) where the NLOT2ms originates as an integral part of the amygdala. This clearly includes the caudal pallial area where Remedios et al. (2007) electroporated at E11.5 a GFP-fluorescent label, which later nicely concentrated in the CAS migration stream (their Fig. 2). The AHi name (‘amygdalo-hippocampal transition area’) given conventionally in atlases to this posterior amygdalar area actually alludes to its easily visible extensive caudal neighborhood with the hippocampus (and entorhinal cortex), rather than with the distant neocortex. This fact obviously already stood ab initio against the conclusions of Remedios et al. (2007). These authors explicitly identified the ‘dorsal pallium’ as the primordium of the neocortex, and even announced a ‘link between the amygdala and neocortex’ in their title. However, no part of the amygdala is histologically neocortical or eulaminate (i.e., six-layered; Garcia-Cabezas et al. 2019), or even contacts neocortical structures (recent reviews in Barbas 2015; Garcia-Cabezas et al. 2019; Puelles et al. 2019a; Garcia-Calero et al. 2020; see also Paxinos and Franklin 2013, and other rodent brain atlases; see also our Fig. 15c, d). This means that there is no solid anatomic or developmental evidence for the neocortex-amygdala link, irrespective that the posterior amygdala seems to share given molecular markers with both dorsal and medial pallium (as was argued above).

Neuroanatomists know for a long time that the neocortex lies far away from the amygdala. If neocortex really produced the CAS cells, these would have to migrate across the interposed mesocortex and allocortex rings that surround the neocortex, possibly having to pass across the entorhinal schizocortex, in order to reach the amygdala beyond the caudal end of the lateral ventricle (see red dot and red arrow schematics in our Fig. 15c, d). Actually, Remedios et al. (2007) did not really think, or wanted to imply, that their postulated ‘neocortical origin’ was distant, since they did not explore that possibility. After all, they knew from their crucial electroporation experiment that the migration originates entirely in front of the caudal end of the lateral ventricle (compare Fig. 15c, d with their Fig. 2).

These authors accordingly just played with the hypothesis that the relatively unexplored caudal part of the dorsal pallium might extend caudalwards through the caudal gap they had discovered between hem and antihem, passing beyond the caudal end of the ventricle into a similarly unexplored ‘caudal telencephalic neuroepithelium’. This implies an aberrant model of cortical development which we need to criticise, because patterning studies are brought into significant confusion by arbitrary modifications of the ‘area map’ that needs to be explained (see comments in Puelles et al. 2019a).

In holding this hypothesis, Remedios et al. (2007) implied (probably unwittingly) that the limits of neocortex are wrong in all precedent publications, atlases and specialized book chapters on the cortex since Brodmann (1909) and von Economo (1927). These sources do not show that the neocortex extends backwards to contact the posterior pallial amygdala (Garcia-Cabezas et al. 2019; Puelles et al. 2019a). It is a sign of a certain dismissive attitude to conventional neuroanatomy in the molecular era that Remedios et al. (2007) actually got this conclusion published in Nature Neuroscience! However, these authors did not even explore in their Discussion the conventional interpretation suggesting that any cortex immediately caudal to the amygdala most probably had to be of the hippocampal allocortical sort, that is, mediopallial. This involves either entorhinal cortex (a variant non-eulaminate sort of allocortex, also termed ‘schizocortex’; Puelles et al. 2019a), or hippocampal subicular or Ammon’s horn areas. We represented these diverse amygdalar caudal vicinity relationships in optimal horizontal sections in Garcia-Calero et al. (2020; e.g., our Fig. 5). Various cortex models supporting this notion were recorded since 1987 (e.g., Swanson 1987; Bayer and Altman 1991; Pattabiraman et al. 2014; Garcia-Cabezas et al. 2019). See also Witter (2012; his Figs.5.1 and 5.2) and Martínez-García et al. (2012; their Fig. 6.3).

Alternatively, the conclusion of Remedios et al. (2007) might imply that the postrhinal, entorhinal and hippocampal cortex domains are actually developmentally and molecularly neocortical, against the majority of existing opinions in the field, and a host of inconsistent molecular data (Thompson et al. 2014; Allen Mouse Brain Atlas, and other analogous repositories). We thus think that Remedios et al. (2007) were not aware of this anatomic difficulty, and did not realize that the hypothetic existence of dorsal pallium passing through the hem-antihem gap to contact the amygdala implied a highly improbable rupture of the allocortical ring present around the whole mesocortex and neocortex, as was already known in 2007 (Swanson 1987; Bayer and Altman 1991; Fig. 15c).

All this implies that it is possible in 2020 to reinterpret coherently the objectively beautiful and rich results of Remedios et al. (2007) as we presently did here, by substituting a primary posterior amygdalar origin of the CAS/NLOT2ms migration, and leaving the dorsal pallium apart in its central place within the cortical ‘area map’. We eliminated doubtful assumptions based on the use of coronal sections, and other conceptual errors of the near past. We refrained from deciding the issue at hand with hardly specific gene expression patterns (Lhx2, Emx1), or from believing that selective cortical gene patterns inform us about the origin of differently fated amygdalar structures, or intrinsic amygdalar migratory phenomena. We will see below that we can insert the duly corroborated gene requirements observed by Remedios et al. (2007), into a different and more complete explanatory rationale that does not involve the dorsal pallium, and does not disrupt on the sly the anatomic traditions preserved by rodent brain atlases and other relevant literature.

The real CAS origin at the amygdalar AHi/PMCo primordium (posterior amygdalar unit) is consistent, moreover, with our present results showing that this amygdalar unit is also the target of the HyA corridor and its migrated hypothalamic Sim1-positive cells. These are fated to help the CAS reach the NLOT2 locus by their apparently required presence in this cell stream. Divergent evolution of an equivalent NLOT2 homolog in other vertebrate groups thus may be also linked to co-evolution of the Sim1-expressing population and its timely arrival at the pallial amygdala, in order to participate in the NLOT migration stream.