|Home | About | Journals | Submit | Contact Us | Français|
During embryogenesis, the pallial-subpallial boundary (PSB) divides the two main progenitor domains in the telencephalon: the pallium, the major source of excitatory neurons, and the subpallium, the major source of inhibitory neurons. The PSB is formed at the molecular interface between the pallial (high Pax6+) and subpallial (high Gsx2+) ventricular zone (VZ) compartments. Initially, the PSB contains cells that express both Pax6 and Gsx2, but during later stages of development this boundary is largely refined into two separate compartments. In this study we examined the developmental mechanisms underlying PSB boundary formation and the post-natal consequences of conditional loss of Pax6 function at the PSB on neuronal fate in the amygdala and olfactory bulb, two targets of PSB-derived migratory populations. Our cell fate and time-lapse imaging analyses reveal that the sorting of Pax6+ and Gsx2+ progenitors during embryogenesis is the result of a combination of changes in gene expression and cell movements. Interestingly, we find that in addition to giving rise to inhibitory neurons in the amygdala and olfactory bulb, Gsx2+ progenitors generate a subpopulation of amygdala excitatory neurons. Consistent with this finding, targeted conditional ablation of Pax6 in Gsx2+ progenitors results in discrete local embryonic patterning defects that are linked to changes in the generation of subsets of post-natal excitatory and inhibitory neurons in the amygdala and inhibitory neurons in the olfactory bulb. Thus, in PSB progenitors, Pax6 plays an important role in the generation of multiple subtypes of neurons that contribute to the amygdala and olfactory bulb.
The amygdala and olfactory bulb (OB) are ancient structures that are contained within the relatively evolutionarily youthful mammalian forebrain. Both are important components of limbic system circuits that process memory and respond to fear and fear extinction; their dysfunction has also been implicated in several human disorders, including autism spectrum disorders (Rodrigues et al., 2004; Amaral et al., 2008; Herry et al., 2008; Markram et al., 2008; Monk, 2008). Unraveling the genetic and cellular mechanisms of development in these interconnected brain structures is therefore necessary to understand function and dysfunction of the adult forebrain.
Previous studies have revealed that major neuronal sub-populations in the amygdala and OB are derived from the PSB, a transient molecular border where telencephalic pallial and subpallial gene expression meet (Toresson et al., 2000; Yun et al., 2001; Corbin et al., 2003; Carney et al., 2006; Carney et al., 2009). PSB progenitors give rise to subsets of interneurons in the OB via the rostral migratory stream (RMS), and to amygdala baso-lateral complex (BLC) excitatory and intercalated cell mass (ICM) inhibitory neurons via the lateral cortical stream (LCS) (Stenman et al., 2003a; Carney et al., 2006; Likhtik et al., 2008; Lledo et al., 2008; Waclaw et al., 2009; Hirata et al., 2009; Kaoru et al., 2009; Waclaw et al., 2010).
The PSB is characterized by the convergence of expression of Pax6 in the ventral pallium (vP) and Gsx2 in the dorsal lateral ganglionic eminence (dLGE) (Toresson et al., 2000; Yun et al., 2001; Corbin et al., 2003; Carney et al., 2006; Carney et al., 2009). Analyses of mouse mutants have revealed that Pax6 and Gsx2 function in a cross-repressive and combinatorial manner for proper PSB patterning (Mastick et al., 1997; Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001; Nomura et al., 2006; Carney et al., 2009). During embryogenesis, Pax6 and Gsx2 expression patterns at the PSB are dynamic, initially overlapping at embryonic day (E) 10.5, with subsequent refinement by mid-neurogenesis into two largely separate compartments (Corbin et al., 2003). However, the cellular mechanisms that regulate the sorting of progenitors to their respective compartments and the link between the genetic regulation of PSB patterning and the generation of neuronal diversity in the amygdala and the OB remains unexplored.
In this study we utilized genetic fate mapping and conditional mutagenesis to address these questions. We found that the molecular refinement of the PSB is a dynamic process regulated by changes in gene expression and cell movements. Further, Gsx2-lineage cells generate multiple inhibitory interneuron subtypes in the OB and BLC, and, surprisingly, a subpopulation of BLC excitatory neurons. Moreover, conditional ablation of Pax6 at the PSB results in focal defects in embryonic patterning that correspond to alterations in the generation of multiple neuronal subpopulations in the post-natal OB and amygdala. Thus, the formation and maintenance of the PSB by Pax6 and Gsx2 is not only important for the correct patterning of progenitor domains during embryonic development; it is also critical for the establishment of specific excitatory and inhibitory neuronal subpopulations in the post-natal limbic system.
Swiss Webster (Taconic Farms, Albany, NY), ROSA-YFP (Jackson Laboratory, Bar Harbor, Maine, Gsx2Cre (N. Kessaris, University College London, (Kessaris et al., 2006)), Pax6-GFP (GENSAT, (Gong et al., 2003)) and Pax6 flox/flox mice (D. Price, University of Edinburgh (Simpson et al., 2009)) used in these studies were maintained according to the protocols approved by Children’s National Medical Center and the University of Edinburgh. Gsx2Cre and Pax6flox/flox mice were maintained on a mixed C57Bl/6 × SW background at Children’s National Medical Center; Pax6flox/flox and Gsx2Cre; RYFP; Pax6 flox/flox mice were maintained on C57/BL6 background at the University of Edinburgh. For staging of the embryos, midday of vaginal plug detection was considered as embryonic day 0.5 (E0.5). For post-natal animals, the day of birth was considered as post-natal day 0 (P0). The genotyping of animals was performed as described previously (Carney et al., 2009; Simpson et al., 2009). Gsx2Cre+/−; Pax6+/+; ROSA-YFP+/− were used as controls while Gsx2 Cre+/−; Pax6c/c; ROSA-YFP +/− (herein referred to as Pax6cKO) mice were used for analysis of conditional knockouts.
For immunofluorescence and in situ hybridization at embryonic ages, brains were fixed in 4% paraformaldehyde (PFA) for 2 hours or overnight, respectively. Brains were cryoprotected by sucrose immersion, embedded in Histoprep (Fisher Scientific, Pittsburgh, PA) and frozen. Serial coronal sections of embedded tissue were cut at 20–30 um thickness using a cryostat and mounted on glass slides. Post-natal animals were transcardially perfused at P21 with 4% PFA, post-fixed for 2–4 hours, and processed in the same manner as the embryonic tissue.
Cryostat mounted sections were air-dried and rinsed 3 times in PBS before blocking for 1 hour in 10% normal donkey serum diluted in PBS with 0.2% Triton to prevent non-specific binding. Primary antibodies were diluted in 1% serum diluted in PBS with 0.2% Triton; sections were incubated in primary antibody overnight at 4°C. The primary antibodies used were as follows: goat anti-Pax6 (1:200; Santa Cruz, Santa Cruz, CA), rabbit anti-Pax6 (gift of V. van Heyningen), mouse anti-NeuN (1:500; Covance, Princeton, NJ), rat anti-GFP (Nacala, Japan, 1:2000), rabbit anti-Gsx2 (1:1500, gift of K. Campbell), rabbit anti-Tbr1 (1:1000; gift of R. Hevner); guinea-pig anti-panDlx (1:1500; gift of K. Yoshikawa), rabbit anti-TH (1:1000; Chemicon), rabbit anti-Mef2c (1:1,000), rat anti-Somatostatin (1:1000; Millipore); rabbit anti-parvalbumin (1:1000; Swant, Switzerland), rabbit anti-calbindin (1:1000, Sigma), rabbit anti-calretinin (1:2000; Swant, Switzerland). To detect primary antibodies, secondary antibodies raised in mouse, rat, goat, guinea pig, and rabbit were used (Cy3 and Cy5 at 1:200, FITC at 1:50; all from Jackson Immunoresearch, West Grove, PA, USA). Sections were incubated for 2 hours in 1% serum in PBS with 0.2% Triton and were washed and cover slipped with gel mount (Sigma) or Vectashield with DAPI (Vector Labs, Burlingame, CA).
Tissue was prepared as described above. Non-radioactive dioxygenin-labeled RNA in situ hybridization was carried out as described previously (Carney et al., 2009). The probes used in this study were Islt1, Ngn2, Tsh1, Sp8, Sfrp2, Dbx1, Dlx1/2, and Er81.
Time lapse imaging experiments were carried out as described previously (Gal et al., 2006). Briefly, E11.5 and E13.5 brains were dissected in ice cold HBSS (Invitrogen), the skull and meninges were removed, and brains were placed in 3% low melt agarose (Fisher Scientific, Pittsburgh, PA, USA). 250–300 µm slices were cut using a vibratome (VT1000S; Leica, Nussloch, Germany), and sections were allowed to recover at 37°C in MEM supplemented with L-glutamine (1:100), penicillin/streptomycin (1:100) and 10% fetal bovine serum (all from Invitrogen). Slices were then mounted on an inverse Zeiss (Thornwood, NY) LSM 510 confocal microscope mounted with a custom-made incubation chamber and perfusion system. Slices were continuously perfused with Neurobasal medium supplemented with B27, N2, penicillin/streptomycin, l-glutamine, and CO2 and maintained at 370C. Z-stack images of GFP+ cells were taken using a Coherent Mira 900F multiphoton laser scanning at 790nm every 6 minutes for 16–21 hours. The time series Z stacks were compiled and processed using Velocity imaging software. For analysis of cell movements, Image J cell tracking software was used to calculate distance traveled and velocity of individual cells.
Electrophysiological recordings were performed from fluorescent Gsx2-lineage YFP+ cells located in the lateral nucleus of the amygdala. Briefly, animals were deeply anesthetized (with isoflurane) until non-responsive and then decapitated. Brains were removed and immediately immersed for 2–3 minutes in an ice-cold - oxygenated (95% O2 / 5% CO2) sucrose slicing solution (mM): 87 NaCl, 2.5 KCl, 1.25 NaH2PO4, 7 MgCl2, 0.5 CaCl2, 25 NaHCO3, 25 Glucose, 75 Sucrose (347 mOsmol) at pH 7.4. Coronal slices containing the amygdala were cut on a vibratome (VT1000S, Leica) at 250 µm. Slices were collected and placed in a clean chamber containing oxygenated sucrose slicing solution at room temperature and incubated for one hour, then transferred to a chamber containing oxygenated artificial cerebral spinal fluid (ACSF) (mM): 119 NaCl, 26 NaHCO3, 10 glucose, 2.5 KCl, 1.25 NaH2PO4*H2O, 1.3 MgSO4, and 2.5 CaCl2*2H2O; pH 7.4, room temperature. Neurons were located and visualized with a fixed staged, upright microscope (BX51WI, Olympus)) equipped with a 4× objective and a 63× insulated objective, infrared (IR) illumination, Nomarski optics, an IR-sensitive video camera (Oly-150) and fluorescent lamp. Glass pipettes (borosilicate glass with filament, WPI, FL) were pulled with an upright micropipette puller (PP 830, Narishige, Japan) to a resistance of 3–6 MOhms. For all recordings the intracellular pipette solution consisted of (in mM): 130 K-gluconate, 20 KCl, 10 HEPES, 0.1 EGTA, 2 MgCl2, 2 Na2ATP, 0.4 NaGTP and 3 mg/ml biocytin (pH adjusted to 7.2, 290 mOsmol). Recordings were performed at room temperature with continuous perfusion (2 ml/minute) of ACSF. Cells were recorded in current clamp mode using a multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA) and a Digidata 1322A acquisition system (Molecular Devices) connected to a PC running pClamp 9.2 (Molecular Devices). For all cells, membrane potential and input resistance values were recorded. Cells were then characterized on the response to depolarizing and hyperpolarizing current pulses for a duration of 1000 msec in 16 consecutive sweeps. Off-line analysis was performed using Clampfit 9.2 (Molecular Devices). In all experiments, data were filtered at 10 kHz during capacitance compensation and 5 kHz during subsequent data recording. The traces were digitized at 10 kHz. All voltage measurements and steps were corrected for a junction potential offset.
In situ hybridization photographs were taken using an Olympus (Center Valley, PA) BX51 microscope. Fluorescent photographs were taken using a Zeiss (Thornwood, NY) LSM 510 confocal microscope. For confocal image analysis, each fluorophore was scanned sequentially and Z-stacks of the images obtained were collapsed into a single projection image or presented as individual optical sections. Figures were prepared using Image J and Adobe Illustrator; brightness and contrast adjustments were applied equally across all images.
Post-natal sections from Gsx2Cre; Pax6+/+; ROSA-YFP and Gsx2Cre; Pax6flox/flox; ROSA-YFP brains were photographed as described above. For each immunohistochemical marker, coronal sections of the lateral and baso-lateral amygdala at Bregma levels −1.6 to −2.4 and of the olfactory bulb at Bregma levels 3.5 to 4.3 were examined from n=/>3 control and Pax6cKO brains. The following criteria were applied to determine co-localization of cell-subtype markers with YFP fluorescence: 1. Cells were counted from individual optical sections, not collapsed projection images. 2. Cells were counted as double positive if an immuno-positive YFP cell body was clearly co-localized with the fluorophore of interest and contained a nucleus that was also DAPI-positive. All statistical analyses of differences between groups were conducted using a two-tailed type two student’s t test, and the null hypothesis was rejected if p<0.05.
The PSB is a region in the ventricular zone of the developing telencephalon where high pallial-expressed Pax6 in the vP converges with high subpallial-expressed Gsx2 in the dLGE (Toresson et al., 2000; Yun et al., 2001; Corbin et al., 2003; Carney et al., 2006; Carney et al., 2009). In addition, our previous work has revealed that at the onset of Gsx2 expression (27 somites, ~E10.0), Gsx2 and Pax6 expression intermingles, and this overlap is largely resolved by 3–4 days later (Corbin et al., 2003). However, how this molecular border is refined remains unexplored. Therefore, we sought to further investigate the timing and the cellular mechanisms underlying PSB refinement. Previous studies of boundary refinement in other brain regions indicate that there are two main non-mutually exclusive mechanisms by which this may occur: 1. Cells expressing patterning genes not shared by their neighbor cells down-regulate the “incorrect” genes and up regulate the “correct” genes expressed by their neighbors 2. Cells expressing patterning genes not shared by neighboring cells move into the territory of like-expressing cells.
To explore whether PSB VZ progenitor cells change gene expression during the period of PSB refinement, we indelibly labeled Gsx2-expressing progenitors and their descendants in Gsx2Cre; ROSA-YFP embryos (Kessaris et al., 2006). Thus, by analysis of Gsx2+, Pax6+ and YFP+ (Gsx2-lineage) cells during the window of PSB refinement, we were able to determine if Gsx2+ cells changed their fate to express Pax6+. Consistent with work from our lab and others (Corbin et al., 2003; Yun, et al., 2001), we found that at E11.5, the PSB is a mixed boundary, with Pax6+ cells present in the Gsx2+ domain, and Gsx2+ cells present in the Pax6+ domain; a subpopulation of cells also express both Pax6 and Gsx2 (Figure 1A,B). At earlier embryonic ages, YFP+ (Gsx2-lineage) cells in the more ventral aspect of the PSB (dLGE) largely co-express Gsx2 protein (Figure 1C,D). Additionally, a sub-population of YFP+ cells also express Pax6 (Figure 1E,F). In contrast, by E15.5, gene expression at the PSB is largely refined, as the intermixing of these two progenitor domains is greatly reduced: fewer Gsx2+ cells are present in the Pax6+ domain and fewer Pax6+ cells are present in the Gsx2+ domain. In addition, with the exception of a few cell diameters of overlap at the border, Pax6 and Gsx2 protein co-expression in individual cells is also greatly reduced (Figure 2A,B). We found that there was a significant difference between the widths of the domain of overlap of Gsx2+/Pax6+ cells at the PSB at E11.5, compared to E15.5 (p=0.0006). Specifically, we found that the diameter of overlap of VZ cells expressing both Pax6 and Gsx2 was larger at E11.5, with an average width of 21.75 (SD=3.59, N=3) cells, while at E15.5, the diameter of overlap of VZ cells expressing Pax6 and Gsx2 was reduced to an average of 8.75 cells (SD=1.71, N=3). As YFP+, Gsx2−, Pax6+ cells are present beyond the refined Gsx2+ domain at E15.5, this reveals that they have down-regulated expression of Gsx2 and up-regulated expression of Pax6 (Figure 2 C–F, G). Thus, at least in case of Gsx2+ dLGE progenitors, changes in gene expression are a mechanism that contributes to the refinement of the PSB.
We next wanted to determine whether the cell movements of progenitors across the PSB during development also contributed to refinement of the PSB. To accomplish this, we carried out time-lapse imaging experiments using Pax6-GFP mice (Gong et al., 2003) to label the complementary dorsal side of the PSB (the vP), which is marked by high GFP expression, in comparison to the dLGE, which is marked by lower GFP expression (Supplementary Figures 1–3). Using two-photon microscopy to image E11.5 and E13.5 brain slices for at least 18 hours, we observed that cell movements across the PSB are much more frequent at E11.5, with numerous cells moving both dorsally and ventrally in the VZ and SVZ (subventricular zone). In contrast, at E13.5, cell movements are much less frequent and largely constricted to the SVZ (Supplemental Figures 1–3). These data indicate that as the border is refined and the number of overlapping cells decreases, both changes in transcription factor expression in cells locally at the PSB, and a concomitant change in cell movement between the pallial and subpallial domains occurs. Thus, our data suggest that both gene expression changes and cell movements contribute to refinement of the PSB.
Cell transplantation and genetic fate-mapping studies (Stenman et al., 2003a; Carney et al., 2006; Waclaw et al., 2006; Brill et al., 2008; Hirata et al., 2009; Waclaw et al., 2009; Waclaw et al., 2010) have shown that cells derived from the PSB contribute to the amygdala and olfactory bulb. We sought to determine the specific sub-types of neurons that dLGE Gsx2-lineage progenitors gave rise to in these two structures. Specifically, we examined the distribution of YFP+ cells in Gsx2Cre; ROSA-YFP mice at post-natal day 21 (P21). We found that in the lateral (LAT) and baso-lateral (BLA) amygdala the majority of Gsx2-lineage neurons expressed inhibitory markers such as calbindin (CB), calretinin (CR), parvalbumin (PV), and nitric oxide synthase (NOS) (Supplementary Figure 4). However, as revealed by expression of Tbr1, Gsx2-lineage cells also unexpectedly gave rise to a subpopulation of excitatory neurons (Figure 3A–C). This represented 35% (257/740, N=3) of the total Tbr1+ population in the LAT nucleus and 11% (80/756, N=3) of the total Tbr1+ population in the BLA nucleus. In addition, Mef2c, which is enriched in the LAT nucleus of the amygdala (Waclaw et al., 2010), also marked a subpopulation of Gsx2-lineage cells (22%, 186/835, N=3) (Figure 3D,E). Gsx2-lineage cells also comprised the lateral (LICM) and medial (MICM) intercalated cell masses (ICM) (Figure 3F), a specialized group of amygdala inhibitory neurons likely derived from the PSB (Waclaw et al., 2010). The ICM plays a central role in modulating inhibitory input between the cerebral cortex, the baso-lateral complex and the central nucleus, which together form an important circuit in mediating fear responses.
Previous studies have shown that LAT/BLA pyramidal-like excitatory neurons can be distinguished from interneurons based on their large soma and specific firing properties (Sah et al., 2003; Sah and Lopez De Armentia, 2003; Cocas et al., 2009). We therefore performed patch clamp recordings of Gsx2-lineage cells with excitatory neuronal morphologies in P21 brain slices to examine whether they would exhibit pyramidal-like functional properties. These large, pyramidal-shaped Gsx2-lineage neurons had an average membrane capacitance of 120 ± 44 pF and an average membrane resistance of 278 ± 89 Mohms (n = 18) (Supplemental Table 1). They all exhibited a regular but accommodating firing pattern (Frequency = 16.8 ± 3.8 Hz; Accommodation ratio = 0.29 ± 0.12; n = 18) and a large and slow after-hyperpolarization (AHP) (I AHP = −8.3 ± 2.0 mV; Half-width AHP = 132 ± 39 ms; n = 18), both features typical of pyramidal-like neurons in the lateral amygdala (Figure 3G, H, supplementary table 1) (Sah et al, 2003; Cocas et al, 2009). These neurons could easily be distinguished from the few Gsx2-lineage interneurons that were also recorded (n = 3), (Figure 3, I–K). These interneurons exhibited a significantly lower membrane capacitance (Cm = 40 ± 21), higher membrane resistance (Rm = 790 ± 290 Mohms), a higher firing frequency (Frequency = 36 ± 5 Hz), a biphasic AHP and a shorter action potential (Half-spike duration = 1.5 ± 0.3 ms).
We also examined the fate of Gsx2-lineage cells in the P21 olfactory bulb. Consistent with previous studies (Young et al., 2007), we found that this lineage gives rise to inhibitory neurons throughout the olfactory bulb, as indicated by co-labeling with the inhibitory neuronal marker pan-Dlx and Foxp2 (Figure 3L–N and Figure 7). Interestingly, a small (3%, 8/326, N=3) population of Gsx2-lineage cells in the periglomerular layer also expressed the excitatory neuronal marker Tbr1 (Figure 3L, M, N and Figure 7G). The small population of Gsx2-lineage neurons that co-expressed Tbr1 did not express pan-Dlx, further indicative that these cells are likely to be excitatory neurons. Together these data demonstrate that within two limbic structures, the amygdala and the olfactory bulb, Gsx2-lineage cells contribute to both excitatory and inhibitory neuronal populations.
Our previous work (Corbin et al., 2003) and this study (Figures 1 & 2) reveal that during a distinct window of development a subpopulation of cells at the PSB express both Pax6 and Gsx2, and that Gsx2-lineage cells give rise to both excitatory and inhibitory neurons in the amygdala and olfactory bulb (Figure 3). Homozygous Small eye (Sey/Sey) Pax6 mutants exhibit disruption of the PSB, with expansion of subpallial cells into the pallial domain, a loss of pallial-specific gene expression, and abnormal cell migration (Stoykova et al., 1996; Mastick et al., 1997; Dellovade et al., 1998; Toresson et al., 2000; Jimenez et al., 2002; Stenman et al., 2003b; Talamillo et al., 2003; Kroll and O'Leary, 2005; Nomura et al., 2006; Carney et al., 2009). However, Sey/Sey mutants are embryonic lethal, which has confounded analysis of the post-natal consequences of Pax6 loss of function. To examine the consequences of PSB mis-patterning on the generation of neuronal diversity in the post-natal brain, we sought to examine neuronal fate in Pax6 conditional mutants. To overcome embryonic lethality, we conditionally knocked out Pax6 in Gsx2-derived cells using Gsx2Cre; Pax6flox/flox mice.
We first examined putative embryonic patterning defects in Gsx2-driven Pax6 conditional mutants (Gsx2Cre; Pax6flox/flox; ROSA-YFP mice- hereafter referred to as Pax6cKO mice) compared to control (Gsx2Cre; Pax6+/+; ROSA-YFP) mice. Consistent with PSB changes previously observed in Sey/Sey mutants (Mastick et al., 1997; Yun et al., 2001; Carney et al., 2009), analysis at E12.5 and E15.5 revealed patterning defects specifically at the PSB. Expression of the vP secreted Wnt receptor Sfrp2 was substantially decreased in Pax6cKO mutants (Figure 4A,E). The vP-expressed homeobox transcription factor Dbx1 was also severely reduced in Pax6cKO embryos (Figure 4 B, F). In addition, expression of the dLGE marker Sp8 persisted and appeared slightly expanded (Figure 4C and G). Expression of the pan-border marker and Ets family transcription factor Er81 appeared unaffected at the PSB (Figure 4D,H). Additionally, expression of the dLGE and LCS Tea-shirt family transcription factor Tsh-1 was preserved, indicating that the LCS amygdala migratory stream is not lost in Pax6cKO embryos (Figure 4I, M). These changes at the border were accompanied by a retraction of expression of pallial marker Ngn2 at the PSB; however, its expression in the remainder of the pallium was normal (Figure 4J, N; Figure 5F, L). Further, the expression pattern of the subpallial marker Islt1, which marks the developing striatum, was unaffected, indicating that the patterning defects in these mutants do not extend into the LGE (Figure 4K, O). Finally, an examination of the VZ/SVZ expressed LGE patterning homeobox transcription factor Dlx1/2 revealed the presence of Dlx1/2+ cortical heterotopias in Pax6cKO embryos, a finding reminiscent of cortical malformations in Sey/Sey mutants. Moreover, LCS expression persists and may be expanded (Figure 4L, P), similar to Tsh1 expression (Figure 4I, M). Collectively, these data reveal local PSB patterning defects characterized by a decrease in vP markers (Sfrp2, Dbx1) and a persistence/possible expansion of dLGE/LCS markers (Tsh1, Dlx1/2).
We next analyzed whether Pax6cKO embryos displayed an alteration of pallial and subpallial marker expression locally at the PSB. In E15.5 Pax6cKO embryos, Pax6 expression was decreased at the PSB and vP compared to controls (Figure 5A, B, G, H). Coordinated with this, Pax6 expression was also decreased in Pax6cKO embryos in the developing amygdala (Figure 5C, I). In contrast, the Gsx2-lineage (YFP+) domain was expanded, encroaching closer to the sulcus (Figure 5D, J). In addition, the domain of Gsx2 expression was expanded, with more Gsx2+ cells scattered ectopically in the vP (Figure 5 E, K). Expression of Ngn2 was also decreased at the vP in E13.5 Pax6cKO mutants, further indicating that the vP domain was abnormal in these embryos (Figure 5F, L). Collectively, these results indicate that, rather than the global patterning defects previously observed in Sey/Sey mutants, Pax6cKO embryos have focal patterning defects that are largely restricted to the PSB, characterized by a decrease of the vP domain and an expansion of the dLGE domain.
We next sought to examine the long-term consequences of PSB disruption on neuronal diversity in two regions known to be derived from the PSB: the amygdala and the OB. In the amygdala, we found that the overall structure of the BLC appeared normal (Figure 6A,D). We also examined the number and distribution of inhibitory interneurons in the BLC, and quantified the number of CB+, CR+, NOS+, and PV+ inhibitory interneuron subpopulations in the LAT and BLA nuclei. We found that while the numbers of CR+, NOS+ and PV+ inhibitory neuron subpopulations were not significantly changed in the Pax6cKO BLC, there was a small, but significant, increase in the number of CB+ inhibitory interneurons in the LAT, but not BLA, nucleus of the amygdala (Supplementary Figure 4). This suggests a possible fate change between LAT excitatory/ICM neurons to CB+ inhibitory neurons. In contrast, there was a significant decrease in the numbers of Foxp2+ lateral and medial ICM neurons (Figure 6A,B, E, F, I). Additionally, the number of Gsx2-lineage Tbr1+ excitatory neurons was significantly decreased in Pax6cKO mice in the LAT nucleus (Figure 6A,C, E, G, I). Interestingly, however, there were no significant decreases in the number of Gsx2-lineage Tbr1+ neurons in the BLA nucleus (Figure 6A,D, E, H, I). We next wanted to examine whether Mef2c+ neurons, which likely overlap with Tbr1 to label excitatory neurons in the LAT nucleus (Waclaw, et al., 2010) were also reduced in Pax6cKO mice. Similar to Tbr1+ neurons, we found that the total number of Mef2c+ neurons in the LAT nucleus was decreased in Pax6cKO mice (data not shown). Analogous to the Tbr1+ population, this difference was not significant when compared to the total numbers of cells (742/1289, Mef2c+/DAPI+ N=3 control; vs. 763/1154, Mef2c+/DAPI+, N=3 Pax6cKO). However, when comparing the numbers of Mef2c+ Gsx2-lineage neurons in the LAT and BLA nuclei of Pax6cKO mice (data not shown), we found that the numbers of Gsx2-lineage Mef2c+ neurons in the LAT nucleus were also significantly decreased (p=0.0418) in Pax6cKO mice (36%, 192/537 Mef2c+/Gsx2-lineage cells in control compared to 18%, 81/447 Mef2c+/Gsx2-lineage; N=3 control and Pax6cKO). We also found that Mef2C+/Gsx2-lineage neurons were not significantly decreased in the BLA (data not shown). These data indicate that in LAT nucleus of the BLC, Pax6 is required in the Gsx2-lineage for the generation of a subpopulation of excitatory neurons, as well as inhibitory CB+ interneurons and ICM neurons.
Next, we examined the consequences of the conditional loss of Pax6 on the generation of neuronal diversity in the post-natal olfactory bulb. Previous work has shown post-natal olfactory bulb interneuron defects in Sey heterozygotes, but analysis of Sey homozygotes has not been possible due to embryonic lethality (Dellovade et al., 1998; Kohwi et al., 2005; Nomura et al., 2007; Brill et al., 2008; Haba et al., 2009). Despite the fact that many OB cells are derived from the Gsx2-lineage, the general structure of the OB appeared normal in P21 Pax6cKO mutants (Figure 7). However, the total percentage of Gsx2-lineage cells in the glomerular layer was decreased in Pax6cKO mice (44%, in control mice compared to 37%, in Pax6cKO mice, p=0.03, N=3). When we examined interneuron subtypes, we found significant decreases in the numbers of TH+, PV+, and CR+ interneuron subtypes (Figure 7C–F,I, J). In contrast, there were no significant differences in CB+ interneurons or Tbr1+ excitatory neurons (Figure 7A,B, G, H, I). These data reveal that Pax6 functions in Gsx2+ cells to generate the correct numbers and types of interneurons in the olfactory bulb.
In summary, the results presented here reveal that the pallial-subpallial boundary is a dynamic border, with a decrease in the number of cells co-expressing Pax6 and Gsx2 and a decrease in the intermixing of Pax6+ and Gsx2+ cells in their respective pallial and subpallial domains during development. Our Gsx2-fate-mapping data and time lapse imaging experiments suggest that this PSB refinement occurs via changes in gene expression (Figures 1 and and2)2) and complementary changes in cell motility, with fewer cell movements occurring in the VZ of the PSB as development proceeds (Supplementary Figures 1–3). Using genetic fate mapping to label the Gsx2+ dLGE, we found that cells from this domain gave rise to both excitatory and inhibitory lineages in the amygdala and olfactory bulb (Figure 3). The conditional ablation of Pax6 from cells that also express Gsx2 resulted in abnormal patterning at the PSB, specifically, reduction of the vP and an ectopic expansion of the dLGE (Figures 4 and and5).5). This PSB disruption resulted in differential alterations in neuronal cell subpopulations in the amygdala and olfactory bulb, regions that are seeded by cells from the PSB. Specifically, the numbers of inhibitory interneurons (but not excitatory neurons) were decreased in the olfactory bulb and the numbers of excitatory neurons and ICM interneurons were decreased in the LAT nucleus of the amygdala, while the numbers of CB+ inhibitory interneurons were increased in this same structure (Figures 6, ,7,7, and Supplemental Figure 4). These findings provide an important link between the essential function of Pax6 during the development and patterning of the embryonic brain and the generation of neurons that give rise to the post-natal limbic system.
While the establishment of other borders within the nervous system, for example, the mid-hindbrain boundary, has been extensively explored, development of the telencephalic PSB remains less well understood. Our current and previous work (Corbin et al., 2003) demonstrates that PSB establishment is highly dynamic, characterized by significant changes in gene expression during neurogenesis. In addition, previous studies from our laboratory and others have shown that PSB-derived cells contribute extensively to the limbic system, most prominently to the amygdala and the olfactory bulb (OB) (Stenman et al., 2003a; Carney et al., 2006; Lledo et al., 2008; Hirata et al., 2009; Waclaw et al., 2009). In this study, we found that the refinement of the PSB, as defined by separation of the majority of high Pax6+ and Gsx2+ progenitors, is generally complete by mid-neurogenesis (E15.5). We also reveal that a major mechanism of PSB refinement is via changes in gene expression, as many Gsx2+ cells change their fate to express Pax6. We further find that the permeability of the border, as evidenced by cell movements in the ventricular zone, decreases with development. Thus, refinement of the PSB is driven by both changes in gene expression in individual cells and by changes in the fluidity of cell movements at the PSB. These developmental mechanisms for refining the boundary are likely utilized to prevent pallial and subpallial compartments from mixing with each other later in development, and may occur via differential cell adhesion molecule expression (Stoykova et al., 1997; Nomura et al., 2006).
Although our findings show that the PSB is largely refined by E15.5, the question arose as to what role, if any, Pax6 plays in Gsx2+ cells at the PSB. To directly address this, we knocked out Pax6 only in cells that co-express Gsx2. During embyrogenesis, we observe an expansion of the dLGE at the expense of the vP, consistent with patterning defects observed in full Pax6 (Sey/Sey) mutants (Stoykova et al., 1996; Yun et al., 2001; Stenman et al., 2003b; Talamillo et al., 2003; Carney et al., 2009). As the PSB is the origin of cells that will populate the amygdala and OB, the well-characterized patterning defects in Sey/Sey mutants are predictive of limbic system disruptions. Consistent with this, we find that Pax6 is necessary in Gsx2+ cells for the correct formation of specific excitatory and inhibitory neuron populations in the post-natal amygdala as well as subsets of interneurons in the OB.
Previous gene expression and fate mapping studies have revealed that the amygdala is derived multiple progenitor pools, including the PSB (Nery et al., 2002; Tole et al., 2005; Carney et al., 2006; Cocas et al., 2009; Hirata et al., 2009; Waclaw et al., 2010). The vP and dLGE components of the PSB generate subpopulations of excitatory and inhibitory neurons that largely populate the amygdala baso-lateral complex (BLC). Earlier work from our laboratory and others has shown that both pallial- and subpallial-derived progenitor pools, as well as diencephalic sources, contribute to neuronal diversity the amygdala (Cocas et al., 2009; Hirata et al., 2009; Garcia-Moreno et al., 2010; Waclaw et al., 2010). With regard to the BLC, our previous genetic fate mapping studies have revealed that dorsal pallial-derived Emx1+ and vP-derived Dbx1+ populations generate excitatory neurons (Cocas et al., 2009; Hirata et al., 2009). In contrast, it appears that the Gsx2+ dLGE may be a major source of the specialized intercalated cell masses (ICM) (Waclaw et al., 2010).
It is of interest to note that while the generation of inhibitory neurons from the Gsx2 lineage is consistent with the subpallial expression of this gene, our finding that Gsx2-lineage neurons also generate a significant population of excitatory neurons was surprising. The most straightforward interpretation is that this population is derived from the transient PSB Gsx2+/Pax6+ progenitor pool. Thus, a broader picture is emerging in which amygdala excitatory neurons are generated from sets of genetically identifiable pallial-derived progenitor pools. Indeed, the broad reduction (but not complete loss) of Dbx1 expression in the vP in Pax6 conditional mutants indicates that Gsx2-lineage amygdala excitatory neurons may be a subset of Dbx1-lineage excitatory neurons. However, the full extent of overlap between Emx1+, Dbx1+ Pax6+ and Gsx2+ progenitors and how these genes act in concert to specify different functional excitatory neuronal subtypes in a combinatorial manner remains to be fully elucidated.
Regardless of how these genes act in combination to specify putative amygdala excitatory neuronal diversity, our Pax6 conditional mutagenesis data provide important and novel insight into the function of Pax6 in amygdala formation. The loss of Pax6 specifically in Gsx2+ cells results in a significant decrease in the number of BLC Gsx2-derived excitatory neurons. Interestingly, this effect is only observed in the lateral and not the baso-lateral nucleus, revealing that Pax6 expression in Gsx2-expressing progenitors is differentially required for generation of neuronal diversity in these two amygdala sub-domains. Complementing our findings is the previous analysis of conditional Gsx2 mutants which display the opposite result-- an increase in excitatory neurons in the amygdala, also specifically in the lateral and not the baso-lateral nucleus (Waclaw et al., 2010). Together, these results reveal that there is an absolute requirement for Pax6, but not Gsx2, in Gsx2-expressing progenitors, in the generation of lateral, but not baso-lateral, amygdala excitatory neurons.
Also interesting is the effect of the conditional loss of Pax6 on the generation of neurons in the intercalated cell masses (ICM). These neurons regulate feed-forward inhibition and are an essential component of fear circuitry. Recent studies have revealed that Gsx2 and Sp8 are required for the specification of these neurons (Waclaw et al., 2009; Waclaw et al., 2010). As we reveal that the conditional loss of Pax6 results in an expansion of Gsx2 at the PSB during embryogenesis, we would predict a complementary finding: an increase in ICM numbers. In contrast to this prediction, we find that the conditional loss of Pax6 in Gsx2+ cells results in a significant decrease in cells in the ICM. Therefore, it appears that Pax6 expression in Gsx2+ cells is also indispensible for the correct specification of ICM cells. However, in contrast to specification of lateral amygdala excitatory neurons, which can be specified in the absence of Gsx2, both Pax6 and Gsx2 are required for ICM generation.
In the olfactory bulb, our Pax6 conditional mutants show a loss of CR+, tyrosine hydroxylase (TH+) and parvalbumin (PV+) subclasses of interneurons. The decrease in TH+ and PV+ populations is consistent with previous studies of Pax6 function (Dellovade et al., 1998; Kohwi et al., 2005; Brill et al., 2008; Haba et al., 2009). Thus, the decrease in TH+ and PV+ interneurons in Pax6 conditional mutants is due to a direct loss of function of Pax6 in populations that express Gsx2. In contrast, the decrease in CR+ neurons, similar to changes in amygdala ICM neurons, is not predicted by previous analyses. Previous studies have shown that specification of CR+ neurons requires normal function of the zinc finger gene, Sp8 (Waclaw et al., 2006; Waclaw et al., 2009). Moreover, Sp8 is regulated by Gsx2, as Sp8 expression is lacking in Gsx2 mutants (Waclaw et al., 2009). However, our conditional Pax6 mutants display a maintained expression of Sp8 and expanded Gsx2 expression at the PSB. Thus, since disruption of either Gsx2 or Pax6 results in CR+ interneuron loss in the olfactory bulb it appears that similar to specification of ICM amygdala neurons, both Pax6 and Gsx2 are required to specify the olfactory bulb CR+ population, and may work in concert. Such a result is reminiscent of the combinatorial role of Pax6 and Dlx in the post-natal generation of dopaminergic (TH+) periglomerular neurons in the olfactory bulb (Brill et al., 2008). While we cannot rule out the possibility that some adult-born cells in the SVZ express Pax6 and Gsx2, and are therefore also affected in our Pax6cKO mice, TH+, PV+, and CB+ OB interneuron generation peaks during embryogenesis (Batista-Brito et al., 2008) at the time that Pax6 and Gsx2 are overlapping at the PSB. Collectivity, these studies suggest a model in which Pax6 and Gsx2 are differentially required for the specification of unique sub-populations of OB inhibitory neurons and of excitatory neurons and specialized ICM inhibitory neurons in the amygdala.
In summary, our studies provide novel insight into the relationship between embryonic telencephalic patterning and limbic system formation. As disruption of the development of the limbic system is a hallmark feature of numerous neuro-developmental disorders, most notably autism spectrum disorders (Rodrigues et al., 2004; Amaral et al., 2008; Herry et al., 2008; Markram et al., 2008; Monk, 2008), it will be interesting to determine if disrupted development of the telencephalic pallial-subpallial boundary is a common feature of such disorders. Indeed, Tlx mutant mice, which have a disrupted vP PSB domain and amygdala abnormalities, have increased aggression, suggesting that the correct development of this boundary is necessary for the formation of amygdala fear circuitry (Roy et al., 2002; Stenman et al., 2003b). Toward this end, understanding the long-term behavioral consequences of specific genetic disruptions of this embryonic boundary will prove highly informative.
We would like to thank the following people for sharing of reagents: K. Campbell for the Gsx2 antibody, R. Hevner for the Tbr1 antibody, V. van Heyningen for the Pax6 antibody, and K. Yoshikawa for the pan-Dlx antibody. We thank R. Carney for contribution of select in situ hybridization probes, and Martine Manuel for perfusing several Pax6cKO mice. We would also like to thank the following people for their critical comments and reading of the manuscript: Jose Luis Olmos-Serrano, Judy Liu, Molly Huntsman and Irene Zohn.
This work was supported by the following grants: R01 NIDA020140 (J.G.C.); NRSA predoctoral fellowship, NS564662 (L.A.C.), the Cellular Imaging Core of Mental Retardation and Developmental Disabilities Research Center (P30HD40677), The Wellcome Trust 08420 (D.J.P.), The Medical Research Council 0800429 (D.J.P.) and 75618 (N.K.), and the European Research Council 207807 (N.K.).
Author Contributions:L.A.C. and J.G.C. designed, planned and interpreted the experiments and wrote the manuscript. Conceptual input was also provided by D.J.P. L.A.C. carried out all of the experiments, analyses, and figure preparation except as follows: P.A.G., under the direction of D.J.P, completed the Ngn2 immunostaining, and in coordination with L.A.C, embryonic in situ experiments as well as provided genotyped conditional knockout tissues. J-M.M., under the direction of V.G, and L.A.C. carried out the patch-clamp experiments and analysis. J.M.C., under the direction of D.J.P, carried out the embryonic Pax6 DAB immunostaining and Dbx1 in situ hybridization. N. K. provided genotyped Gsx2Cre embryonic brains and Gsx2Cre mice. L.A.C. carried out the time-lapse experiments with training and guidance from T.F.H.