Fish were kept under standard conditions at a 14 hours light/10 hours dark cycle as previously described ^{44, 45} . All procedures were in accordance with the live animal handling and research regulations of the local Animal Care and Use Committee, the Regierungspräsidium Dresden (permit AZ 24D-9168.11-1/2008-1 and -4). Wildtype experimental animals (Biotechnology Center Dresden) were adult fish from the gol-b1 line in the AB genetic background ⁴⁶ . Adult fish were 6–8 months old and had a 24mm–32mm body length, zebrafish larvae were 7dpf old.

Brains (either dissected or within the skull) were fixed at 4°C overnight in 2–4% paraformaldehyde/0.1M phosphate buffer (PB), pH 7.5. They were washed 1 × 10 minutes and then up to 1h in 0.1M PB and subsequently transferred for decalcification and cryoprotection to 20% sucrose/20% EDTA in 0.1M PB, pH7.5. Brains were frozen in 7.5% gelatine/20% sucrose and sectioned into 14–16 µm cryosections. Sections were stored at -20°C.

Compared to tetrapods, teleost fish have undergone an additional whole genome duplication: the teleost genome duplication (TGD) (reviewed in ⁴⁷ ). Thus, there is the possibility of two co-orthologous genes in zebrafish compared to the single tetrapod gene.

We analyzed if there are two co-orthologous genes compared to the tetrapod gene using Ensembl73 gene trees ( http://www.ensembl.org) and synteny analysis with the Synteny Database ( http://syntenydb.uoregon.edu; ⁴⁸ ). This is clearly the case, e.g., for human EOMES with two TGD co-orthologs in zebrafish, eomesa and eomesb ( Figure S1A,B).

Two genes are currently termed ascl1 in zebrafish. Zebrafish ascl1a is clearly orthologous by phylogeny and conserved synteny to Ascl1 in lobefins (tetrapods and coelacanth). Zebrafish ascl1b not only has a separate ortholog in coelacanth but also shows conserved synteny to tetrapod Ascl2, suggesting that ascl1b is in fact the missing teleost ascl2 gene.

There are two prox1 genes described in zebrafish, yet while prox1a is clearly orthologous to tetrapod Prox1, teleost prox1b shows no conserved synteny to teleost prox1a or tetrapod Prox1. This suggests that teleost prox1b represent a more distant prox paralog and is not a TGD paralog of prox1a. The phylogeny of the emx genes in zebrafish has previously been determined in ⁴⁹ .

RNA in situ hybridization on sections and on whole-mount brains and RNA probe generation was essentially performed as previously described ^{7, 50} . Briefly, after defrosting at room temperature (RT), sections were rehydrated for 15 minutes in PBS with 0.3% TritonX (PBSTx) and incubated with the probe overnight at 62–65°C. Information on the antisense in situ riboprobes can be found in ⁷ for ascl1a (NM_131219), emx1 (NM_198144), emx2 (NM_131280), emx3 (NM_131279), eomesa (NM_131679). The in situ probe eomesb (NM_001083575) was cloned from zebrafish embryonic cDNA with the following primers ( eomesb-F, TTTCCAAAACGAAAAGCGTA, eomesb-R, GAGCCAGAACTGGATCCTTCT). The eomesb probe was tested on 3dpf embryos and showed specific staining only in the midbrain at 3dpf ( Figure S2, Dataset 1). The sections were washed at 60–65°C in washing solution (1 × SSC, 50% deionized formamide) for 1 × 15 minutes and 2 × 30 minutes followed by 2 × 30 minutes MAB with 0.1% Tween-20 (MABT) washes. Sections were incubated for 1h at RT in 2% DIG-blocking reagent (Roche) and incubated with anti-DIG antibody (Roche Diagnostics, sheep, polyclonal, Fab fragments conjugated to alkaline phosphatase, #11093274910) diluted 1:4000 in 2% DIG-blocking reagent overnight at 4°C. Subsequently, sections were washed 4 × 20 minutes in MABT, equilibrated with staining buffer and stained with the substrate NBT/BCIP. The staining was controlled using a stereomicroscope. Finally, sections were washed 2 × 5 minutes in PBS, postfixed with 4% PFA for 20–30 minutes, washed again 2 × 10 minutes in PBS and mounted with 70% glycerol in PBS. All washing steps were performed on a shaker, all incubation steps in a humid chamber. To test for nonspecific binding of the antibodies that detect digoxigenin, which are not endogenous to vertebrate tissue, we performed control experiments in which the labeled RNA was omitted from the hybridization mix. No signal was detected in the absence of the riboprobe, demonstrating that the antibody reacts specifically with the synthetic RNA ( Dataset 2).

Immunohistochemistry on cryosections was performed as previously described ⁵¹ . Briefly, to retrieve the antigens of Prox1, sections were pre-incubated in 50 mM Tris-buffer (pH 8.0) at 99°C for 5 minutes, cooled down to RT over 15 minutes and washed for 5 minutes in PBS and twice for 10 minutes in PBSTx. The sections were then incubated in primary and secondary antibodies in PBSTx. The primary antibody Prox1 (AB5475, 1:2000, Millipore) was incubated overnight at 4°C and secondary antibodies for 1h at room temperature. The slides were washed in PBSTx and mounted. The secondary antibody (dilution 1:750) was Alexa 488-Fluor conjugated (A-11034, Invitrogen, Karlsruhe).

Confocal images were acquired with Leica TCS-SP5 confocal microscope. Brightfield images were acquired with Zeiss Axio Imager Z1. The images were processed using ImageJ v. 1.4.3.67 and Adobe Photoshop CS2. Composites were assembled using Adobe Photoshop CS2 and Adobe Illustrator CS2.

We primarily followed the nomenclature proposed in ³² with modifications based on our combinatorial expression pattern analysis, which suggest a subdivision of Dl in a dorsal and ventral subdivision (which we have named Dld and Dlv, respectively). In the figures we employ a two-color code to separate subdivisions that can be made based on the current marker (red dashed line) from those that we make based on other markers (white dashed line). The black dashed line indicates the boundary between D and V. The distinction of ventricular zone versus neuronal layer was based on cellular morphology. The ventricular zone is the region where cells are directly facing the ventricle. The neuronal layer is characterized by cells with a round shape.

In tetrapod embryos, Eomes ( Tbr2) expression is found in the ventricular zone and mantle layer in all parts of the pallium at embryonic stages ^{9, 21, 23} . In the zebrafish embryo, eomesa expression is present throughout the dorsal telencephalon ^{52, 53} . Its TGD paralog, eomesb is not present in the embryonic or adult telencephalon ( Figure S2, Dataset 1) and thus was not further taken into account. In the adult zebrafish, eomesa positive cells are scattered in a salt-and-pepper pattern along the ventricular zone of Dm along the rostro-caudal axis ( Figure 1A–D, arrowheads). Further, eomesa expression is present in the ventricular zone and neuronal layer of Dc, Dlv and Dld in the rostral telencephalon and at mid-telencephalic levels and in the ventricular zone and neuronal layer of Dld and Dp at the anterior commissure (Cant; Figure 1A–C). Caudal to Cant, eomesa is expressed in the ventricular zone and neuronal layer of Dp, sporadically in Dld and in bed nucleus of the stria medullaris (BNSM, Figure 1D). In summary, eomesa is differentially expressed in the pallium along the rostro-caudal axis ( Figure 1E–H, Dataset 3; Table 1). The most abundant parenchymal eomesa expression is detected in parts of the central and lateral dorsal telencephalic nuclei.

In tetrapod embryos, the ventral pallial subdivision is characterized by absence of Emx1 and presence of Tbr1 ^{9, 10, 12, 13, 21} . In embryonic and adult zebrafish, tbr1 is expressed throughout the pallium ^{7, 53, 54} . Thus, we investigated the expression of emx1, emx2 and emx3 to identify a ventral pallial subdivision in the zebrafish pallium. In zebrafish, the three emx genes are expressed at 1 day post fertilization (dpf) in the embryo throughout the dorsal telencephalon ^{49, 55, 56} . In 7 day-old larva, the expression of emx1 and emx2 is restricted to a small caudo-lateral area in the dorsal telencephalon ( Figure S3A,B,D, Dataset 4). The expression of emx3 is found throughout the dorsal telencephalon ( Figure S3C,E, Dataset 4). In the adult zebrafish pallium, emx1 and emx2 are expressed in a very restricted fashion. The expression of emx1 only starts in the mid-telencephalon shortly rostral to Cant, where it is restricted to the ventricular zone and neuronal layer of Dp. This expression pattern continues to Cant ( Figure 2A). Furthermore, emx1 expression is present in Dp and in EN caudal to Cant ( Figure 2B). The expression of emx2 is only found caudal to Cant in an area in Dc lateral to Vp ( Figure 2C). Similar to the larvae, emx3 shows the broadest expression of the emx genes. In the rostral telencephalon, the expression of emx3 is found in the ventricular zone and neuronal layer of Dm and weakly in scattered cells in the neuronal layer of Dc ( Figure 2G). Additionally, the expression of emx3 is weakly present rostrally in the ventricular zone of Dlv ( Figure 2G, arrowheads). At mid-telencephalic levels, emx3 expression is found in the ventricular zone and neuronal layer of Dm, in scattered cells in Dc and in the ventricular zone and neuronal layer of Dlv ( Figure 2H). At Cant, expression of emx3 is found in the ventricular zone and neuronal layer of Dm, in scattered cells in Dc and in the ventricular zone and neuronal layer of Dp ( Figure 2I). Caudal to Cant, the expression gets weaker in Dm, Dc and Dp (data not shown). In summary, emx1 and emx2 show a very restricted expression pattern in the adult zebrafish pallium ( Figure 2D–F, Dataset 5; Table 1). Expression of emx3 is present in Dm, Dc and Dlv/Dp ( Figure 2J–L, Dataset 5; Table 1).

During mouse development, Prox1 expression is found in the amygdala, dentate gyrus and in the neocortex ⁵⁷ . At adult stages, however, strong Prox1 expression is restricted to the dentate gyrus of the hippocampus and is commonly used as a specific marker for granule cells of the hippocampus ^{57– 60} . In the adult zebrafish pallium, Prox1+ cells only are present in the midtelencephalon shortly rostral to Cant in the neuronal layer of Dld ( Figure 3A). This expression pattern continues to the anterior commissure ( Figure 3B). Shortly caudal to Cant, only scattered Prox1+ cells are present in Dld, more caudally no Prox1+ cells can be found (data not shown). In summary, Prox1 staining is present in Dld starting at mid-telencephalic levels until shortly caudal to Cant ( Figure 3A–D, Dataset 6; Table 1).

In tetrapod embryos, Ascl1 is expressed a subpopulation of progenitors in the dorsal telencephalon ^{61, 62} . In the zebrafish embryo, ascl1a expression is present in the caudomedial ventricular zone of the dorsal telencephalon ⁶³ . In the adult zebrafish pallium, ascl1a is expressed in scattered cells in the ventricular zone of Dm and in scattered cells in the ventricular zone of Dlv ( Figure 4A–D, arrowheads, Figure 4I,J, Dataset 7; Table 1). At Cant and posterior to Cant ascl1a is present in scattered cells in Dm and Dp ( Figure 4E–H, arrowheads, Figure 4K,L, Dataset 7; Table 1).

Due to the different development, the pallium of ray-finned fishes has a markedly different morphology compared to all other vertebrates, which makes the comparison between the areas of the pallium of ray-finned fishes to pallial nuclei of other vertebrates particularly challenging. Yet, the correct assignment of homologous pallial areas between teleosts and tetrapods is essential for usage of the teleost fish model in neurobiological research. We have analyzed several conserved molecular marker genes that are found in specific areas of the pallium in the domestic mouse, chicken and the African clawed frog. Based on the expression analysis we identify four main subdivisions of the pallium (Dm, Dl, Dc and Dp) and propose that Dl is subdivided in a dorsal (Dld) and ventral part (Dlv). Based on our data we also suggest putative homologies to pallial nuclei in tetrapods. We suggest that Dm is homologous to the ventral or ventral/lateral pallium, Dc to the dorsal pallium, Dl to the medial pallium, and suggest that Dp comprises a specialized part of Dl ( Figure 5). Additional marker analysis, lineage tracing experiments and functional analyses will be necessary to substantiate the proposed pallial subdivisions and their homology to pallial nuclei in tetrapods. As our study is based solely on comparative gene expression data, we have discussed our results in the framework of gene expression data of other organisms and subsequently compared them to connectional, neurochemical, and functional data in teleosts. For clarity, we discuss our results separately for each pallial subdivision.

In the African clawed frog, the chicken, and the domestic mouse, four pallial divisions have been identified in the embryo, the ventral pallium (VP), the lateral pallium (LP), the dorsal pallium (DP) and the medial pallium (MP) ^{9, 10, 13, 21} . In the African clawed frog, chicken, and domestic mouse, the ventral pallial subdivision is characterized in the embryo by the absence of Emx1 expression and presence of the pallial marker Tbr1 ^{9, 10, 13, 21} . In embryonic and adult zebrafish, tbr1 is expressed throughout the pallium ^{7, 53, 54} . In adult ray-finned fishes, Dm has been proposed to be homologous to the ventral pallium (pallial amygdala) based on topological, connectional and functional data ^{29– 31, 64– 66} . In contrast, Nieuwenhuys (2009) proposed that Dm is homologous to the lateral pallium based on topology and Yamamoto et al. (2007) suggest that Dm together with Dd and Dld is homologous to the dorsal pallium. Thus, we analyzed the expression of the emx genes in the larval and adult zebrafish to determine if the absence of these markers identifies a ventral pallial subdivision in the zebrafish pallium. In zebrafish, the emx genes show a dynamic expression pattern both in the embryo and the adult. At 1dpf, the three emx genes are expressed at throughout the dorsal telencephalon ^{49, 55, 56} . We found that expression of emx1 is restricted to a caudolateral expression domain in the zebrafish larvae at 7dpf and to Dp in the adult ( Figure 2D,E, Figure S3). Similarly, emx2 also shows a restricted expression pattern to a caudolateral domain both in the zebrafish larvae at 7dpf and in the adult zebrafish pallium ( Figure 2F, Figure S3). The expression of emx3 is present in the entire pallium in the zebrafish embryo and larvae ( ⁴⁹ , Figure S3C). In the adult, emx3 is present in the rostral telencephalon in Dm, Dc and in the ventricular zone of Dlv ( Figure 2J). Moving caudally, emx3 is present in Dm, in scattered cells in Dc and in the ventricular zone and neuronal layer of Dlv and Dp ( Figure 2K,L). As it has been previously suggested that zebrafish emx3 supplies the functions provided by Emx1 and Emx2 in mouse ⁴⁹ , we took the combined expression pattern of the three emx genes into account for our analysis. The lack of an area in Dm that is tbr1 positive and emx gene negative suggests that zebrafish might not have a distinct ventral pallial subdivision and that Dm is either homologous to the lateral pallium, or that Dm comprises a combined ventral and lateral pallium ( Figure 5A–D). Emx1 alone might not be an adequate marker for the ventral pallium in the domestic mouse and chicken, as they have lost the Emx3 gene ⁴⁹ .

Both the ventral pallium and the lateral pallium contribute together with subpallial derivatives to the amygdaloid complex in tetrapods ^{10, 67, 68} . In rodents, the Cannabinoid Receptor (CB1) is expressed in a subset of neurons in the basolateral amygdala ^{69, 70} . In the weakly electric fish Apteronotus leptorhynchus and zebrafish, cb1 is expressed in Dm ^{71– 73} , thus supporting the model that Dm is homologous to the pallial amygdala. Based on gene expression analysis, we have suggested previously that the supracommissural nucleus of the area ventralis telencephali (Vs) is homologous to the dorsal and ventral part of the central amygdala and the bed nucleus of the stria terminalis (BST) and that the postcommissural nucleus of the area ventralis telencephali (Vp) is homologous to the dorsal part of the central amygdala and the BST ⁷ . Thus, it is plausible that Vs and Vp (subpallial amygdala) and Dm (pallial amygdala) form the amygdaloid complex in the adult zebrafish ( Figure 5A–E).

Based on topological and gene expression data it has been proposed that the entire Dl ⁴ or Dl excluding Dlv or Dp is homologous to the medial pallium in other vertebrates ^{24, 30, 32, 64} . In contrast, Wullimann and Mueller (2004) considered only Dlv to be homologous to the medial pallium and Yamamoto et al. (2007) proposed that only the dorsal part of Dl is homologous to the medial pallium. Functional ablation experiments suggested that Dl is equivalent to the hippocampus of tetrapods ⁶⁵ . Our combinatorial expression data suggests that Dl is subdivided in Dld and Dlv rostrally and at Cant and posterior to Cant in Dld and Dp ( Figure 5A–D). During mouse development, Prox1 expression is found in the amygdala, dentate gyrus and in the neocortex ⁵⁷ . At adult stages, however, strong Prox1 expression is restricted to the dentate gyrus of the hippocampus and is commonly used as a specific marker for granule cells of the hippocampus ^{57– 59} . In the adult zebrafish, Prox1 positive cells are exclusively present in the caudal part of Dld. In summary, our data suggests that Dl (excluding Dp, see below) may be homologous to the medial pallium and hippocampus and the caudal part of Dld homologous to the dentate gyrus in the domestic mouse ( Figure 5A–E). We will discuss the possible homology of Dp in the subsequent part.

It has previously been shown that Dp receives olfactory input and has been on this basis homologized to the lateral pallium in amphibians and all other gnathostomes with evaginated forebrains ⁴ . Being homologous to the lateral pallium, it should be located next to Dm in the embryo, the presumptive ventral pallium. The discrepancy between the position of Dp in the adult and an expected location next to Dm has led to different models to explain the different position of Dp in the adult pallium: “the partial pallial eversion model” by Wullimann and Mueller ^{31, 32, 52, 74} , the “eversion-rearrangement theory” by Northcutt and Braford ^{33, 64, 75} , and the “new eversion model” by Yamamoto and colleagues ³⁷ . In the “partial pallial eversion model” ^{31, 32, 52, 74} , based on connectional and gene expression data, it has been proposed that the homolog of the lateral pallium does not participate in the eversion. Wullimann and Mueller put forward that neuroblasts generated by the ventricular zone of the uneverted, medially located lateral pallium homolog migrate laterally to give rise to the submeningeally located nucleus Dp ^{31, 74} . Our gene expression analysis does not support this part of their modified partial eversion model. Based on differential gene expression we can identify Dp, which is not located submeningeally, but has its own ventricular zone even in the adult ^{4, 64, 76} . New neurons are still generated in the zebrafish pallium in the adult ^{77, 78} . Thus, a small area of ventricular zone should be still present next to Dm to generate neurons for Dp. However, we have not observed neuroblasts migrating laterally across the pallium to reach Dp in the adult by BrdU pulse chase studies or genetic lineage tracing ^{51, 76, 77} . The “eversion-rearrangement theory” by Northcutt and Braford suggests that differential expansion of the ventricular surface of some pallial zones and differential proliferation and migration of neuroblasts from the different ventricular zones might result in displacement or shifting of the different pallial subdivisions ^{33, 64, 75} . Similar to the “partial pallial eversion model”, they propose that a small stretch of Dp is still located between Dm and Dl. Our gene expression data does not support this aspect of both models, as we do not identify a small area of ventricular zone and neuronal layer sandwiched between Dm and Dc rostrally and Dm and Dld more caudally that matches the gene expression profile found in Dp.

In the “new eversion model”, the eversion was suggested to occur in a caudolateral direction leading to a shift of the arrangement of the different pallial subdivisions ³⁷ . Yamamoto and colleagues propose that ventral pallial and lateral pallial homologs are not present in the rostral but only in the caudal pallium. In their model they suggest that Dp is homologous to the lateral pallium ³⁷ . In contrast, we find the putative ventral or ventral/lateral pallial homolog Dm contiguously from rostral to caudal. However, we also identify a separate Dp nucleus only in the posterior part of the pallium.

In contrast to the models discussed above, Nieuwenhuys has put forward that the lateral olfactory tract in vertebrates with everted telencephala is not homologous to the olfactory tract in vertebrates with evaginated telencephala ⁴ . He bases this on data that showed considerable variation in the pattern of secondary olfactory projections among different groups of ray-finned fishes and showed secondary olfactory projections both to Dm and Dp ^{4, 29, 75} . He suggests that the olfactory input has increased overall in Dl, with a subsequent confinement to Dp and has decreased in Dm during ray-fin fish evolution. He considers the lateral olfactory tract as an apomorphy of actinopterygian pallia ⁴ . Even though our gene expression data does not support the models that suggest a migration or displacement of Dp from a position equivalent to the lateral pallium to its caudolateral position, we can identify Dp as a subdivision of Dl separate from Dld. Furthermore, Neuropeptide Y and Parvalbumin immunohistochemistry clearly distinguishes Dl from Dp ⁷⁹ . Furthermore, as discussed above, our gene expression data suggest that Dm might be homologous to the lateral pallium or comprise a combined ventral and lateral pallium, which might suggest that Dp is not homologous to the lateral pallium. Our gene expression data is consistent with Nieuwenhuys’ model that Dp comprises a specialized part of Dl that receives olfactory input, even though the other parts of Dl might be homologous to the medial pallium and hippocampus in tetrapods ( Figure 5A–D).

In the “modified partial pallial eversion model” ³² , based on expression of Parvalbumin and nicotine adenine dinucleotide phosphate diphorase (NADPHd) in Dl, it was suggested that Dc is a true histogenetic unit that has its own germinative zone rostrally, but is caudally displaced to a more central location by differential growth of Dm and Dl. Mueller and colleagues suggest that Dc is homologous to the dorsal pallium in other vertebrates and that Dd is absent in zebrafish ³² . This model is consistent with connectional and gene expression data in Apteronotus leptorhynchus ²⁴ . In addition, it has been noted that Dc shows no significant immunoreactivity to Calretinin, Neuropeptide Y or Thyrosine hydroxylase separating it from the surrounding areas of the pallium ⁷⁹ . Other models based mainly on topology suggest that Dm, Dd and Dld are homologous to the dorsal pallium ³⁷ , whereas others have homologized Dd with the dorsal pallium ^{4, 29, 30, 33, 64} and have suggested that Dc is not a separate unit, but represents the deeper zone of the periventricular areas of Dm, Dd and Dl. Our gene expression data is consistent with the second part of the “modified pallial eversion model” of Mueller and colleagues. We have adapted the nomenclature of Mueller et al. (2011) to call the displaced nucleus Dc, even though our data do not rule out the possibility that Dc is rather a displaced Dd nucleus. Accordingly, Dc does not have its own germinative zone caudally, but only in the rostral part of the pallium. However, the morphogenesis of the telencephalon has been analyzed between 1dpf and 5dpf and no such displacement of Dc has been described ⁵ , so if it is the case, then it has to happen after 5dpf. To conclusively evaluate the displacement of Dc, detailed lineage analysis using Cre/loxP technology should be performed from embryonic stages till adulthood.

In the adult zebrafish, proliferating cells are found scattered in the ventricular zone of the pallium ^{76– 78} . In the adult zebrafish pallium, ascl1a is found in scattered cells in the ventricular zone of Dm and in Dlv/Dp. In addition, eomesa is present in scattered cells in the ventricular zone of Dm and in the ventricular zone of Dld and Dlv/Dp. Emx1 is also present in the ventricular zone of Dp and emx3 is found in the ventricular zone of Dm, Dlv and Dp ( Figure 5A–D).

In mouse, Ascl1 is expressed in a subpopulation of progenitors in the dorsal telencephalon ^{61, 62} . Eomes (Tbr2) is present in intermediate progenitors in the cortex in the embryo and in the dentate gyrus both in the embryo and the adult and is part of the glutamatergic differentiation cascade that also contains NeuroD and Tbr1 ⁸⁰ . In the adult zebrafish, neurod and tbr1 as well as the vesicular glutamate transporters vglut1/2.1/2.2 marking glutamatergic neurons are expressed throughout the pallium, suggesting that the glutamatergic differentiation cascade may still be present in the adult zebrafish telencephalon ^{7, 72} . The expression in scattered cells in the ventricular zone suggests that eomesa and ascl1a may label progenitor subpopulations in the pallium and eomesa could play a role in differentiation of glutamatergic neurons in the adult zebrafish brain. In addition, emx1 and emx3 might be involved in regulating neuronal differentiation processes in the pallium. It will be interesting to perform comprehensive lineage analysis using Cre/loxP technology to follow the progeny of these different markers in the adult zebrafish.