The strong expression of Dlx5, weak expression of Nkx2.1, Lhx6, Lhx7/8 and absence of pallial markers Tbr1 and Emx1 in the CA further indicates a striatal character (Puelles et al., 1999, 2000; Gorski et al., 2002; Medina et al., 2004; García-López et al., 2008). A schematic overview of the developmental origin of cells within the CA is shown in Figure 4A. Gene expression analysis and fate-mapping of Isl1-lineage cells proved that the principal cells in the CA originate from the ventral LGE (vLGE), and show medium spiny-like morphologies (Waclaw et al., 2010; Bupesh et al., 2011a). Via lineage tracing and cell tracking components, vLGE-derived cells were found to preferably inhabit the medial CA (CeM) and the centromedial part of the lateral CA (CeL), with an almost complete absence in the capsular CA (CeC) (Waclaw et al., 2010; Bupesh et al., 2011a). In contrast, Pax6-expressing cells from the dorsal LGE (dLGE) seem to inhabit complementary regions of the CA; the CeC contains abundant dLGE-derived cells, the lateral part of the CeL shows a moderate amount while the CeM completely lacks these cells (Puelles et al., 2000; Bupesh et al., 2011a). Surprisingly, while Gsh2 LOF severely affects the size of the vLGE and the development of the striatum, the size of the CA was mostly unaffected (Toresson and Campbell, 2001; Stenman et al., 2003; Waclaw et al., 2010). This finding prompted Waclaw et al. (2010) to hypothesize that vLGE-derived CA neurons arise from its most ventral portion, which is relatively unaffected in Gsh2 mutants.

The CeM and CeL also contain an important population of cvMGE-derived cells, generated starting from E10.5 (García-López et al., 2008; Bupesh et al., 2011a; Puelles et al., 2016b). An earlier study by Nery et al. (2002) concluded that a subset of CA cells is derived from the CGE, based on homotypic transplantation studies. However, some caution is advised since the transplanted cells could include the Dg/cvMGE. Another study that tried to decipher the origin of several amygdaloid nuclei by in utero gene transfer of eGFP described labeled cells that co-expressed Dlx5 in the CA when the CGE was targeted (Soma et al., 2009). As the cvMGE was separately investigated in this study, these cells must originate in the CGE “proper” (Soma et al., 2009). The authors themselves conclude that the CA might partially be derived from the dLGE or dCGE. Similar to the BM, a shell of hypothalamic-derived cells (Sim1+, Otp-) was reported to surround the CA (García-Moreno et al., 2010), however, this finding has not been replicated in any of the Otp/Sim1 reporter mice (Garcia-Calero et al., 2021; Morales et al., 2021).

The medial amygdala (MA) seems to have a different developmental origin, and should rather be considered a ventromedial extension of the pallidum as postulated by Puelles et al. (1999). The developmental origin of MA neurons is schematically presented in Figure 4B. Shh- and Dbx1-lineage tracing experiments revealed a high density of labeled cells in the anterior MA (MeA) and posteroventral MA (MePV), with a smaller amount located in the posterodorsal MA (MePD) (García-López et al., 2008; Hirata et al., 2009; Carney et al., 2010; Waclaw et al., 2010; Kanatani et al., 2015; Puelles et al., 2016a). In contrast to the principal projection neurons of the BLC, these Dbx1-expressing cells are generated in the POA at an earlier timepoint (E9.5-11.5) and are inhibitory (GABAergic) in nature (García-López et al., 2008; Hirata et al., 2009; Waclaw et al., 2010; Kanatani et al., 2015; Puelles et al., 2016a). A second subpopulation of POA-derived GABAergic neurons destined for the MA was recently identified (Lischinsky et al., 2017). These cells are characterized by their expression of Foxp2, and are complementary to the Dbx1-lineage cells within the MA (Lischinsky et al., 2017). Intriguingly, these two POA-derived subpopulations might contribute to the sexual dimorphism of the MA. Indeed, postnatally these populations differ in their expression of steroid pathway proteins between sexes and are activated in a sex-specific manner during mating (Lischinsky et al., 2017).

In addition to the POA, the cvMGE was identified as a potential source of interneurons in the MeA and MePD. However, the presence of cvMGE/Dg-derived neurons in the MA remains controversial, since only few Sst-expressing interneurons are shown to invade the MA (Real et al., 2009; Puelles et al., 2016b). Instead, expression of Calbindin was previously used to link these cells to the cvMGE (García-López et al., 2008; Bupesh et al., 2011b). Observations from homotopic transplantation experiments made by Nery et al. (2002) indicate that at least some MA neurons originate in the CGE, although these could also include the cvMGE. While Vucurovic et al. (2010) found no 5HT3aR:GFP+ grafted cells (E14-E14.5) from the CGE, Touzot et al. (2016) identified some 5HT3aR:GFP+ cells within the MA, of which 53% co-expressed Prox1 (CGE). As expected from the major subpallial contributions to this nucleus, it was not affected in Pax6^sey/sey mice (Tole et al., 2005).

Excitatory cells of the MA are similarly diverse, with origins ranging from telencephalic to hypothalamic progenitor pools. Fate-mapping of Emx1-lineage cells showed an absence in the MePD, although some Emx1-expressing cells were identified in the MePV (Gorski et al., 2002). A subset of Dbx1-lineage neurons in the MeA and MePV express Lhx9 and are hypothesized to be tangentially migrated cells from the anterior radial amygdala unit (García-López et al., 2008; Bupesh et al., 2011b; Puelles et al., 2016a; Garcia-Calero et al., 2020; Garcia-Calero and Puelles, 2021). Within the MeA, the Lhx9-expressing cells preferentially colonized the shell and avoided the core subdivision, while Lhx9-expressing cells in the MePV preferentially inhabited the core and avoided the shell (Garcia-Calero and Puelles, 2021). These neurons might overlap with a Tbr1-expressing population identified in the ventral-most portion of the MePV by Carney et al. (2010).

Ebf3/Tbr1 co-expressing glutamatergic cells originating in the CVP were reported to form a shell surrounding the MePV, thereby separating it from the BM laterally and from the MePD dorsally (Ruiz-Reig et al., 2018). These CVP-derived cells perfectly complement the abundant Lhx9 cells in the core of the MePV (Ruiz-Reig et al., 2018; Garcia-Calero and Puelles, 2021). While our current understanding is that the cortical and amygdalar pallial fields should be completely separated (Puelles et al., 2019), and thus the description of the CVP as a cortical pallial area is inherently wrong, these authors likely identified a Dbx1-negative pallial amygdalar progenitor zone that delivers cells to the MePV. Bupesh et al. (2011b) likely already identified these cells in 2011, as he observed labeled cells in the MePV when a cell tracker (CMFDA) was placed in the “caudal ventral pallium.” Targeting the more “caudal ventral pallium” resulted in labeled cells located superficially around the MePV, while placing CMFDA in more rostral parts of the “ventral pallium” resulted in labeled cells in the MeA, which supports that the cell tracker was placed in the more rostral progenitor zone of the anterior amygdala unit in the latter case (Bupesh et al., 2011b). As expected from its presumed pallial origin, Ebf3-expressing cells are strongly reduced in Pax6KO embryos (Ruiz-Reig et al., 2018).

The VZ of the SPV generates Otp-expressing cells starting from E11, which were found to cross the telencephalic-hypothalamic border toward the MA (García-Moreno et al., 2010; Bupesh et al., 2011b; Morales et al., 2021). Ninety-nine percent of all Otp-expressing cells in the medial amygdala co-expressed Foxg1 and were generated in the telencephalon-opto-hypothalamic domain (TOH), a dorsal subdomain of the SPV (Morales et al., 2021). Otp/Foxg1 co-expressing cells preferentially inhabited the anterodorsal part of the MeA, the medial part of the MePD and a ring surrounding the MePV, thereby agglomerating at its medial edge (Morales et al., 2021). An additional Otp+ cell population, located between the MePV and the superficial layer of the medial nucleus, was likely generated in the more central or ventral parts of the SPV based on their co-expression of Ebf3 and lack of Foxg1 expression, although confirmation of this birthplace is necessary (Ruiz-Reig et al., 2018). The significant contribution of Otp-derived cells to the MA was corroborated by the analysis of Otp^–/– mice, in which the size of the MA nucleus was severely reduced (García-Moreno et al., 2010). By analyzing the gene expression of Pax6, Bupesh et al. (2011a) identified an additional migratory stream that seemingly originated in the prethalamic eminence (PTE) and was bound for the MA. However, the hypothalamo-amygdalar corridor (HyA) lies rostrolaterally from the PTE, in close relation to it (Garcia-Calero et al., 2021). Since no other gene expression markers were used except for Pax6, we cannot exclude that the observed cells do not originate from other sources. Puelles et al. (2016a) similarly observed Dbx1-expressing cells originating from the SPV/PTE-region, but noted that distinguishing between both regions proved difficult. Lineage tracing and/or cell tracking experiments could help further prove this alternative hypothalamic origin.

The BST was originally subdivided into the BSTL and BSTM based on the similarity of its neurochemistry to other nuclei of the centromedial complex (Alheid and Heimer, 1988; Alheid et al., 1995; Cassell et al., 1999; De Olmos and Heimer, 1999; Alheid, 2003). The BSTia is often considered a caudal extension of the BSTM (García-López et al., 2008). The BST was later redivided into approximately 20 subgroups; an excellent overview scheme of different models of BST parcellation can be found in Dong et al. (2001). In general, the BST is marked by the expression of Nkx2.1 and Lhx6 (Puelles et al., 2000; García-López et al., 2008; Xu et al., 2008). The developmental origin of BST neurons is schematically shown in Figure 5.

Based on cell tracking (CMFDA) and gene expression studies, the BSTL was found to contain a mixture of vLGE-derived Isl1-expressing neurons and dLGE-derived Pax6-expressing neurons, although this latter cell population was less significant than in the CA (Bupesh et al., 2011a; Tinterri et al., 2018). In contrast to the vLGE-derived neurons of the CA, BSTL Isl1-expressing neurons are derivatives of corridor cells as evident from their migratory route and dependence on Ebf1 expression (Tinterri et al., 2018). A small amount of cvMGE-derived interneurons are also present in the BSTL, which could explain the expression of Nkx2.1 and Lhx6 in this region (García-López et al., 2008; Bupesh et al., 2011a; Puelles et al., 2016b). García-López et al. (2008) had previously already suggested a partial striatal origin for the BSTL, based on the weak expression of Nkx2.1 and strong expression of Dlx5 and Lmo4 at later developmental stages.

The BSTM can be further subdivided into an anterior (BTSMa), posteromedial (BSTMpm), and posterolateral (BSTMpl) part. In-depth analysis of Sst-expressing neurons in combination with cell tracking studies and gene expression analysis indicates that a significant portion of the BSTMa, BSTMpm, and BSTia is populated by cvMGE-derived neurons (García-López et al., 2008; Real et al., 2009; Bupesh et al., 2011b; Puelles et al., 2016b). In addition, transplantation experiments of presumed CGE-derived cells identified labeled cells within BST (Nery et al., 2002; Vucurovic et al., 2010). Intriguingly, the dLGE also contributed a marginal amount of cells to the BSTM (Bupesh et al., 2011a). Moreover, when a cell tracker was placed in the POA, labeled cells that co-expressed Shh were identified in the BSTM (García-López et al., 2008; Bupesh et al., 2011b). These cells are complemented with cells from the TOH, which preferentially colonize the BSTMpm, although scattered cells are also present in the BSTMpl while the BSTMa only contains a minority of TOH-derived cells (García-Moreno et al., 2010; Bupesh et al., 2011b; Morales et al., 2021).

The laminated structure of the nLOT resembles the pallium, yet its transcriptional signatures indicate that cells within the three layers originate from distinct embryonic progenitor regions. The developmental origin of the nLOT is represented in Figure 6A. In particular, only layer 2 (nLOT2) contains cells derived from the Emx1-lineage (Gorski et al., 2002), while lineage tracing studies of Dbx1-lineage mark cells in both nLOT1 and nLOT3 (Puelles et al., 2016a). Functional analysis clarified that both nLOT2/nLOT3 require Pax6 for correct development, as these layers are completely absent in Pax6^sey/sey mice, while nLOT1 was correctly formed (Tole et al., 2005). The dependence of nLOT2/3 development on Pax6 expression is reflected by expression of its downstream targets that mark this region such as Tbr1, Ngn2, and SCIP (Puelles et al., 2000; Medina et al., 2004; Remedios, 2004; Remedios et al., 2007; Soma et al., 2009). Cells destined for the nLOT1 are generated at an earlier timepoint (E10.5) as compared to nLOT2/3 (E11.5) (Remedios et al., 2007). While earlier publications described a strong expression of Lhx2 and Lhx9 in nLOT1 (Remedios, 2004; Tole et al., 2005; García-López et al., 2008), a more recent publication taking into account the novel radial dimension of the amygdala describes the whole nLOT as Lhx9-negative (Garcia-Calero and Puelles, 2021). Without the use of other markers, it is not clear whether the positive region below the nLOT belongs to the nLOT1 or the AAA. The lack of Lhx9-expressing cells in nLOT3 indicates a lateral or basal pallial amygdalar origin (García-López et al., 2008; Puelles et al., 2016a; Garcia-Calero and Puelles, 2021). Based on expression of Er81, a stream of cells seems to emerge from the horn of the medial BLA toward the nLOT, thereby surrounding the whole nLOT and specifically invading nLOT3, however, more research is required to confirm this observation (Garcia-Calero et al., 2020). The presence of Dg/cvMGE-derived interneurons in the nLOT3 is evident from analysis of Sst and Lhx6 (García-López et al., 2008; Real et al., 2009; Puelles et al., 2016b).

The developmental origin of nLOT2 is best characterized, and was described to reside in a unique division of the dorsal pallium around E11.5-12.5 (Remedios et al., 2007). This progenitor region, referred to as the caudoventral dorsal pallium (cvDP), was thought to be located in the caudal telencephalon, ventrally abutting the lateral ventricle in between the MP and VP/LP (Remedios et al., 2007). The DP character of this progenitor region was proposed based on its expression of SCIP, Emx1, and Tbr1, and absence of Cad8, Sfrp2, and Wnt2b, markers of the LP, VP, and MP, respectively (Medina et al., 2004; Tole et al., 2005; Remedios et al., 2007; Soma et al., 2009). The complete absence of nLOT2 in mice models that show extensive disruption of cortical development (Lhx2 and Tbr1 single LOF mutants, and Emx1/Emx2 double LOF mutants) further seemed to support a cortical origin (Remedios, 2004; Remedios et al., 2007). However, the reintroduction of the concentric ring theory of the pallium rendered the previous name (cvDP) unfit for an amygdalar progenitor region (Puelles et al., 2019). In a later publication these cells were shown to originate in the now called “posterior amygdalar radial unit,” or more specifically at the rostromedial subdivision of the amygdala-hippocampus interface (AHi) (Garcia-Calero et al., 2021). Of note, both sets of authors (Remedios et al., 2007; Garcia-Calero et al., 2021) seem to describe the same region, and merely disagree on the terminology. These telencephalic nLOT2 cells were found to be complemented by cells from the SPV, which enter the telencephalon via the hypothalamo-amygdalar corridor (HyA, see below), precisely at the posterior amygdalar progenitor region where the other nLOT2 cells are generated (Garcia-Calero et al., 2021). nLOT2 lacks expression of Lhx6 and Gad1, and seems devoid of Dbx1- and Dlx5/6-derived neurons (Remedios et al., 2007; Puelles et al., 2016a).

The developmental origin of the BAOT remains largely obscure to this day. The BAOT is characterized by expression of Lhx9, Tbr1, and contains fate-mapped Emx1-lineage cells (Gorski et al., 2002; Medina et al., 2004; García-López et al., 2008; Garcia-Calero and Puelles, 2021). While the BAOT was previously hypothesized to be of VP origin based on its expression of Lhx9, the now corresponding anterior radial unit does not match molecularly with the BAOT (Garcia-Calero et al., 2020; Garcia-Calero and Puelles, 2021). The BAOT is therefore currently hypothesized to be of posterior amygdalar or PTE origin, two regions that are also marked by Lhx9 and Tbr1 expression (Garcia-Calero and Puelles, 2021).

The cortical nuclei (CoA) of the amygdaloid complex comprise the ACo and the PCo, the latter of which contains a posterolateral (PLCo) and posteromedial (PMCo) subdivision. The developmental origin of the CoA is shown in Figures 6B,C. In general, the entire cluster is predominantly marked by pallial genes, including Tbr1 and Emx1 (Gorski et al., 2002; Medina et al., 2004; Tole et al., 2005; García-López et al., 2008; Hirata et al., 2009). Similar to other pallial tissue, scattered interneurons can be found throughout these nuclei (Nery et al., 2002; Real et al., 2009; Vucurovic et al., 2010; Puelles et al., 2016b).

Before the establishment of the radial unit model, the principal cells of the ACo were ascribed to the VP based on Dbx1-lineage tracing (non-overlapping with GABA) and its expression of Tbr1, Sema5A, Ngn2, and Lhx9 (Medina et al., 2004; García-López et al., 2008; Hirata et al., 2009; Bupesh et al., 2011b; Puelles et al., 2016a). ACo Dbx1+ cells are generated 1 day before Dbx1+ cells of the BLC complex (E10.5 vs. E11.5) and preferentially inhabit the posterior ACo (Hirata et al., 2009; Puelles et al., 2016a). A second population of Tbr1-expressing neurons is not derived from the Dbx1-lineage, and likely overlaps with Emx1-lineage cells (Gorski et al., 2002; Puelles et al., 2016a). The principal cells of the ACo are now understood to be of anterior pallial amygdalar origin, one of the regions within the amygdalar pallial tissue that shares some genetic markers with the ventral pallium (Garcia-Calero et al., 2020). A small subset of hypothalamic-derived TOH cells can also be found in the ACo (Morales et al., 2021). Comparable to the BLC, the ACo nucleus shows aberrant morphology and positioning in Pax6^sey/sey mice (Tole et al., 2005).

Both subdivisions of the PCo were also found to contain Emx1-lineage cells, and were suggested to be of VP, LP, DP, or MP origin (Puelles et al., 2000; Gorski et al., 2002; Medina et al., 2004; Cocas et al., 2009). Previous findings that the PLCo and PMCo only contained a marginal amount of Dbx1-lineage cells were later rectified, and both nuclei were reinterpreted to contain Dbx1-progeny (Puelles et al., 2016a; Garcia-Calero et al., 2021). The progenitor region of the PLCo is thus allocated to the basal amygdalar unit, while the progenitor zone of the PMCo is located within the posterior unit, as expected based on the strong expression of Lhx9 (Garcia-Calero et al., 2020; Garcia-Calero and Puelles, 2021). Some SPV-derived neurons (Otp+) were also identified in the PMCo, although it is currently unknown whether they comprise the population originating from the TOH (co-expressing Foxg1) or an Otp+ cell population from another SPV domain (García-Moreno et al., 2010). Since their presence within the PMCo was not mentioned in any of the more recent papers, it is possible that the ACo and PMCo population is the same, and the nuclei had been wrongly annotated.

Dlx5/6-expressing interneurons in the CoA are likely from both CGE and MGE origin, as homotopic transplantation studies found that CGE and MGE grafted cells overlapped in the CoA (Nery et al., 2002; Wang et al., 2011). A CGE “proper” origin was also corroborated by the transplantation of E14.5 CGE-derived 5HT3aR:GFP+ cells, which colonized the CoA (Vucurovic et al., 2010). The strong expression of CoupTf2 in the CoA at P0 supports a partial CGE origin (Tang et al., 2012). In addition, scattered Nkx2.1-lineage interneurons can be found throughout the CoA, a subset of these cells in the PLCo/PMCo are likely generated in the cvMGE, as evident from their early expression of Sst (Xu et al., 2008; Real et al., 2009; Puelles et al., 2016b).

The AHi was originally often classified within the “transition areas” of the amygdala, however, due to the presence of a distinct boundary between the AHi and hippocampus and its similarity to PMCo, we have opted to discuss this nucleus here instead. Cells within this caudal transition area are marked by the expression of Lhx9, Sema5A, and Ngn2 (Medina et al., 2004). Puelles et al. (2016a) and Garcia-Calero and Puelles (2021) recently remarked that an earlier observation that the AHi lacks Dbx1+ cells was faulty. Emx1-lineage cells can also be found within this area (Gorski et al., 2002). The AHi represents the periventricular part of the posterior radial amygdala unit, its cells therefore likely originate in the ventral ventricular zone at the caudal end of the hemisphere (see Figure 6D). Conditional deletion of CoupTf2 (Rx-Cre) results in the extensive loss of the AHi (Tang et al., 2012). Sst-expressing cells originating from the cvMGE are also present in this nucleus, and appear to be a caudoventral extension of the cvMGE-derived migratory stream toward the BSTia, LA, and BMA (see section Diagonal Area) (Puelles et al., 2016b).

The ITCs are located in paracapsular cell clusters (pcITCs) surrounding the BLC and in the more ventrally located main intercalated cell mass (IA). The ITCs are sometimes considered a part of the central extended amygdala, based on their shared molecular markers and probable developmental origin, as further explained below. However, ITCs are not as homogenous as previously assumed, explaining their classification within the “other” amygdalar nuclei.

A striatal character of the ITCs was suggested based on gene expression analysis (lack of Tbr1, weak Nkx2.1, Lhx6 expression and strong Dlx5 expression) (Medina et al., 2004; García-López et al., 2008; Kaoru et al., 2010). Moreover, the ITCs express Gsh2 and Pax6, indicating that they are derived from the LGE (Tole et al., 2005; Waclaw et al., 2010; Bupesh et al., 2011a; Cocas et al., 2011). Indeed, conditional Gsh2 LOF (Foxg1tTA/+) mice show reduced numbers of ITCs in both the IA and lateral pcITC (Waclaw et al., 2010; Kuerbitz et al., 2018). The dLGE identity was further proven by means of conditional Sp8 LOF (Dlx5/6-Cre) and lineage tracing of Isl1. As expected, the phenotype observed in the conditional Gsh2 LOF mouse was mimicked in its Sp8 counterpart, while no Isl1-lineage cells could be observed in the IA or pcITCs (Waclaw et al., 2010). Conditional LOF of Pax6 within the Gsh2-lineage resulted in a strong reduction of the ITCs, highlighting the necessity of Gsh2 and Pax6 co-expression for the correct specification of ITCs (Cocas et al., 2011).

Conditional LOF (Dlx1-Cre lineage) of Tshz1, a gene downstream from Sp8 that is expressed in migrating dLGE cells, exhibit a complete loss of ITCs in the lateral pcITC and in the IA at E18.5-P3, while cells of the medial pcITC were strongly reduced at P3, but not at E18.5 (Kuerbitz et al., 2018). Postnatal progeny of Tshz1KO cells were found to have switched cellular fates and upregulated Foxp1 (striatal projection neuron marker; Tamura et al., 2004; Precious et al., 2016) instead of Foxp2 (ITC marker; Kaoru et al., 2010; Waclaw et al., 2010; Kuerbitz et al., 2018). A strong co-labeling of cleaved caspase-3 and GFP was observed in the amygdala of conditional mutant animals (P0.5), indicating that loss of Tshz1 expression results in increased apoptosis (Kuerbitz et al., 2018). This effect was a direct result of loss of Foxp2 expression, as Foxp2^–/– mice recapitulated this phenotype (Kuerbitz et al., 2018).

A small subset of ITCs was found to strongly express Tbr1 and lacked expression of subpallial markers (Medina et al., 2004). These cells likely overlap with the scarce Emx1-lineage cells identified by Gorski et al. (2002). Remedios et al. (2007) hypothesized that a subpopulation of cells generated in the posterior amygdala radial unit (in the original paper called the cvDP) migrate toward the intercalated cell masses, although no other papers have confirmed the presence of nLOT2 cells within the ITCs.

The anterior amygdaloid area (AAA) consists of two subdivisions, termed the dorsal and ventral AAA (AAd and AAv, resp.). A schematic overview of its developmental origin is shown in Figure 6A. Both subdivisions were found to be mostly devoid of Emx1-expressing cells (Gorski et al., 2002). Gene expression analysis confirmed that the AAd is molecularly similar to the BSTL, and could be assigned a “striatal identity” based on the strong expression of Dlx5, Pax6, Lmo4, weak expression of Nkx2.1, Lhx6 and lack of Tbr1 expression (Stoykova et al., 2000; Medina et al., 2004; Tole et al., 2005; García-López et al., 2008). Cell tracking studies confirmed that the AAd contains a significant proportion of dLGE-derived Pax6-expressing cells, which explains its aberrant morphology in Pax6^sey/sey mice (Stoykova et al., 2000; Tole et al., 2005; Bupesh et al., 2011a).

García-López et al. (2008) postulated that the AAv is primarily derived from the cvMGE based on the expression pattern of Lhx6 and Npy. However, the AAv contains only a very limited amount of Sst-expressing cells (García-López et al., 2008; Puelles et al., 2016b). Moreover, the AAv seems developmentally related to the adjacent ACo based on its expression of Dbx1, Emx2, Lhx2, and Lhx9 (Tole et al., 2005; Puelles et al., 2016a; Garcia-Calero and Puelles, 2021). Indeed, while the AAA is generally considered a subpallial region, several Lhx9+ cells from the anterior pallial amygdala unit were found to invade the AAA (Garcia-Calero et al., 2020). Some Otp+/Foxg1+ cells from the hypothalamus (TOH) were also identified in the AAv (Morales et al., 2021). Similar to the AAd the AAv is malformed in Pax6^sey/sey mice (Stoykova et al., 2000; Tole et al., 2005).

Ex vivo migration assays with cell tracker components (CMFDA) and gene expression analysis (Isl1) at E13.5–E16.5 by Bupesh et al. (2011a) revealed that vLGE-derived cells tangentially invaded the BSTL medial to the internal capsule. This finding was corroborated by Tinterri et al. (2018), who investigated the tangential migration of corridor cells from the vLGE in detail (see Figure 9). Corridor cells tangentially migrated into the MGE territory around E13.5, after which a subset contributed to the BST (Tinterri et al., 2018). Conditional LOF of Ebf1 (Dlx5/6-Cre lineage), which show abnormal migration of corridor cells, reduced the amount of “corridor cells” within the BST and affected the total size of the BST at postnatal ages (P5) (Tinterri et al., 2018). Another tangential migratory stream that passes the striatum ventromedially and delivers cells to the interstitial nucleus of the posterior limb of the anterior commissure (Ipac), was identified (Tinterri et al., 2018). The anatomical location of the Ipac vs. the CA could imply that this migratory route also delivers cells to the CA. Data from Bupesh et al. (2011a) suggests that the vLGE-derived cells of the CA are either a ventrolateral extension of the radial vLGE-derived migratory stream toward the striatum, or tangential derivatives of the ventromedial migratory route described by Tinterri et al. (2018). However, more research is required to fully understand the migratory route of these cells.

The PSB separates the progenitor domains of the subpallial dLGE and the pallial VP primordia; the VP gives rise to the Ventropallial Migratory Stream (VMS), often referred to as the Lateral Cortical Stream (LCS) in earlier publications. The migration of these populations are temporally controlled, as the onset of migration of VMS Pax6+/Dlx2- cells (E11.5) occurs before dLGE Dlx2+ cells (E13.5) (Carney et al., 2006; Hirata et al., 2009).

In contrast to the pallial VMS, only few Dlx2-expressing, dLGE-derived cells appeared to contact the radial glia (Carney et al., 2006). Instead, the majority of the subpallial dLGE-derived population seems to migrate via chain-migration and exits the cell cycle only after the onset of their migration, similar to dLGE-derived cells of the rostral migratory stream (RMS) (Carney et al., 2006). Loss of Pax6 expression was found to severely affect the amount of migrating dLGE-derived cells, likely as a consequence of a migratory arrest (Tole et al., 2005). Based on data from Carney et al. (2006); Bupesh et al. (2011a), and Kuerbitz et al. (2018), the Dlx2+ and Tshz1+ cells derived from the dLGE seem to migrate around the developing BLC, thereby contributing to the intercalated cell masses that surround this complex (see Figure 9). Some cells also seem to invade the BLC. A significant part of the migration stream then continues their migration toward the AAd (see Figure 9; Bupesh et al., 2011a). How dLGE cells reach the CA and BSTL is currently not well-understood, but they are hypothesized to be tangential migrations from a stream splitting from the dLGE (BSTL), or the stream surrounding the BLC (CA) (Bupesh et al., 2011a).

Migration of dLGE-derived cells destined for the ITCs was found to be (partly) dependent on Tshz1 expression (Kuerbitz et al., 2018). While the main function of Tshz1 is regulation of ITC differentiation and survival, several genes involved in guidance and migration showed abnormal expression levels in the conditional LOF (Dlx1-Cre lineage), including ErbB4, Prokr2, and Dcc (Kuerbitz et al., 2018). Downregulation of the Neuregulin receptor ErbB4 was confirmed with immunostaining and is linked to the regulation of MGE-derived cortical interneuron and dLGE-derived olfactory bulb interneuron migration (Yau et al., 2003; Anton et al., 2004; Flames et al., 2004; Kuerbitz et al., 2018). Downregulation of ErbB4 might affect the migration of cells through the permissive Nrg1 corridor in the LGE, and therefore confine these cells to the migratory stream (Flames et al., 2004). Within the RMS, Tshz1 was previously shown to be critical for the transition from chain to radial migration (Ragancokova et al., 2014). In contrast to the ITCs, the CA seems devoid of Tshz1-derived cells (Kuerbitz et al., 2018).

The pallial amygdalar cells were originally believed to reach their amygdala destinations via the ventropallial migratory stream (VMS). It was recently discovered that the amygdalar pallial field is positionally separated from the cortical pallial field (Puelles et al., 2019; Garcia-Calero et al., 2020; Garcia-Calero and Puelles, 2021), complicating these observations. Moreover, while cortical pallial cells mainly utilize radial glia-guided migration, the amygdalar pallial nuclei are generated via outside-in cell aggregation (Molnár and Butler, 2002; Carney et al., 2006; Bai et al., 2008; Garcia-Calero et al., 2020). Below, we provide an overview of some molecules that might be relevant for the amygdala.

Similar to structures derived from the VP, no gross developmental abnormalities in amygdala structures have been described in Reeler or Cdk5^–/– mice, with the exception of the nLOT2 (see below) and a single cell population in the MePD, of which the origin has not been identified (Remedios et al., 2007; Boyle et al., 2011). The apparent migration of VP-derived cells of the VMS (Tbr1+) was found to be dependent on the expression of Dcx and Lis1, two genes that are known to be involved in migration (Bai et al., 2008). Transfer of Dcx shRNA to the VP at E13 caused cells to be stuck within the migratory stream, apparently unable to invade the surrounding amygdalar tissue (Bai et al., 2008). As we now understand that VP cells do not invade the amygdala and are restricted to the olfactory cortices (Garcia-Calero and Puelles, 2021), we can deduct that their in utero electroporation also included pallial amygdalar progenitors, which are therefore also dependent on Dcx expression for correct migration. Alternatively, non-amygdalar tissue might have been misinterpreted as amygdalar. Likewise, knockdown of Lis1 decreased the amount of migrating cells, and cells that did migrate got arrested in white matter tissue (Bai et al., 2008). The morphology of the cells was also severely affected; Dcx RNAi resulted in shorter leading processes, while those of cells treated with Lis1 RNAi were significantly longer (Bai et al., 2008). In addition, Lis1- and Dcx-RNAi increased the occurrence of multipolar cells by two and fourfold, respectively (Bai et al., 2008). Lis1 might also play a role in neuronal positioning and normal synaptogenesis in the amygdala, as evident from heterozygous Lis1 mutant mice (Toba et al., 2013).

Based on the expression of Lhx9, Garcia-Calero and Puelles (2021) reconstructed the gradual development of anterior pallial amygdalar derivatives into intermediate and superficial nuclei (see Figures 10A,B). At early stages of development (E12.5), Lhx9 labeled the entire radial unit, however, from E14.5 onward, Lhx9 expression completely disappeared at the periventricular stratum, while the intermediate (BMA) and superficial (ACo) nuclei grew (Garcia-Calero and Puelles, 2021). This finding highlights the unique development of the amygdalar pallial field, which develops in an outside-in instead of inside-out manner, unlike other mammalian pallial cortical regions (Garcia-Calero and Puelles, 2021). A superficial translocation completely depletes the periventricular stratum of cells, which agglomerate at the superficial and intermediate parts of the radial unit (Garcia-Calero and Puelles, 2021). A glial palisade separates the lateral/basal unit from the subpallium at the periventricular level. This massive translocation seems to only occur at the anterior radial unit, as it is the only amygdalar unit that lacks a periventricular part (Garcia-Calero et al., 2020). In addition to its direct pallial derivatives, Lhx9 positive cells were also found at the level of the AAA, MeA, and MePV, currently postulated to be the result of tangential cell migrations from the ACo (Garcia-Calero and Puelles, 2021). Bupesh et al. (2011b) had previously proposed a tangential migration of “VP-derived cells” toward the MeA and MePV, based on the orientation of the migrating cells and radial glial fibers.

Cellular migration to nLOT2 was initially characterized by Remedios et al. (2007) using in utero electroporation, gene expression analysis and LOF models. This migratory route (shown in Figures 10D,E) originates in the now identified posterior amygdala radial unit and was termed the “Caudal Amygdaloid Stream (CAS)” (Remedios et al., 2007). It was later renamed nLOT2 migratory stream (nLOT2ms) to avoid further confusion about the birthplace of the migrating neurons (Garcia-Calero et al., 2021).

Analysis of NeuroD expression, a marker that exclusively labels the nLOT2 and nLOT2ms, predicts the onset of migration around E14.5, and by E16.5 most neurons seem to have arrived in the nLOT2 (Remedios et al., 2007). At the rostromedial AHi, the locally generated nLOT2 cells are joined by hypothalamic Sim1+ cells at around E13.5 (see “Hypothalamus”) (Garcia-Calero et al., 2021). These cells then collectively migrate rostroventrally across the pallial amygdala (Garcia-Calero et al., 2021). At E15.5, when the stream is located at the medial side of the BLA and BMA, it crosses into the AAA and continues its migration toward the nLOT2 (Garcia-Calero et al., 2021). A small stream of cells diverge from the nLOT2 and seem to invade a medial shell surrounding the BMA (Garcia-Calero et al., 2021). The nLOT2ms can be subdivided into two phases, including a tangential phase during the initial rostral extension, followed by a radial phase with ventral extension toward the pial surface (Remedios et al., 2007). Garcia-Calero et al. (2021) later deducted that the migratory phases overlap with a switch from migration within pallial tissue to subpallial tissue, respectively.

Intriguingly, cellular migration through the nLOT2ms was found to be dependent on Reelin and Cdk5 signaling (Remedios et al., 2007). Cells migrating through the nLOT2ms express Dab1 and form a Reelin-negative corridor (Remedios et al., 2007). In Reeler mutant embryos a migratory stream was still visible, however, the route was severely malformed and a “nLOT2-like” structure with aberrant morphology emerged in a more medial and dorsal position along the route (Remedios et al., 2007). In addition, a migratory arrest was found to occur in Cdk5^–/– mice after the first phase of migration (Remedios et al., 2007). Zic2 is also expressed in migrating nLOT2 cells, and homozygous deletion of Zic2 resulted in a less populated and more dispersed migratory stream (Murillo et al., 2015).

Surprisingly (controversially), an older study by Tang et al. (2012) had identified a CGE-derived migratory stream that supplies glutamatergic neurons toward the BM. As mentioned before, we believe this population constitutes pallial derivatives instead of CGE-derived interneurons and have opted to discuss this migratory route here instead of in section “Caudal Ganglionic Eminence.”

At E13.5, two Pax6+ migratory streams were found to arise from the “CGE,” with cells within the medial stream migrating toward the BM (Tang et al., 2012). Similar to cells of the MMS, CoupTf2 seems to regulate the migration of this cell population, as the number of migratory Pax6 cells was severely diminished in conditional CoupTf2 LOF mice (Rx-Cre), without affecting apoptosis or proliferation in the “CGE” (Tang et al., 2012). Immunohistochemical analysis and qPCR assays revealed that the expression of Nrp1 and Nrp2 was significantly reduced in the conditional mutant (Tang et al., 2012). CoupTf2 was found to directly affect expression of Nrp1 via binding of a Sp1 site located within the 12th intron. Similarly, cell tracking experiments (CMFDA) in ex vivo slices identified a Tbr1+ migratory stream originating in the CVP, seemingly overlapping with the posterior amygdalar unit of Garcia-Calero et al. (2020), that splits into two streams surrounding the MePV (see Figure 10C; Ruiz-Reig et al., 2018). Whether CoupTf2-expression also regulates the migration of CVP-derived neurons should be further investigated.