|Home | About | Journals | Submit | Contact Us | Français|
Intermediate progenitors (IPs) amplify production of pyramidal neurons, but their role in selective genesis of cortical layers or neuronal subtypes remains unclear. Using genetic lineage tracing in mice, we find that IPs destined to produce upper cortical layers first appear early in corticogenesis, by embryonic day 11.5. During later corticogenesis, IP laminar fates are progressively limited to upper layers. We examined the role of Tbr2, an IP-specific transcription factor, in laminar fate regulation using Tbr2 conditional mutant mice. Upon Tbr2 inactivation fewer neurons were produced by immediate differentiation, and laminar fates were shifted upwards. Genesis of subventricular mitoses was, however, not reduced in the context of a Tbr2 null cortex. Instead neuronal and laminar differentiation were disrupted and delayed. Our findings indicate that upper layer genesis depends on IPs from many stages of corticogenesis, and that Tbr2 regulates the tempo of laminar fate implementation for all cortical layers.
Excitatory, pyramidal neurons of the cerebral cortex are generated during embryonic neurogenesis from radial glial progenitors (RGPs) both directly, and indirectly via transient-amplifying, committed neurogenic intermediate progenitors (IPs) (reviewed by Florio and Huttner, 2014; Sun and Hevner, 2014). Importantly, the generation of neurons of different cortical layers follows a general "inside-out" pattern as lower layer (LL) neurons are born first, and upper layers (ULs) last (Hevner et al., 2003).
IPs are distinguished from RGPs by short radial or multipolar morphology, mitotic division tending to occur away from the ventricular surface, and a unique molecular profile, including specific expression of the Tbr2 transcription factor (also known as Eomes; NCBI Gene Eomes) (Englund et al., 2005; Gal et al., 2006; Kawaguchi et al., 2008; Stancik et al., 2010). Given their distinct properties, IPs have been proposed to generate specific cortical neuron subtypes. However, the precise contribution of IPs to cortical neurogenesis and laminar fate specification remains poorly defined. Previously, IPs were suggested to produce mainly UL neurons (Tarabykin et al., 2001; Zimmer et al., 2004; Britanova et al., 2005), but further studies indicated that IPs produce all layers (Kowalczyk et al., 2009; Vasistha et al., 2014). More recently, the morphological and electrophysiological properties of UL neurons were reported to depend on their origins from Tbr2-negative RGPs, or from Tbr2-positive IPs (Tyler et al., 2015). Thus, one goal of the present study was to use a panel of molecular markers to more extensively define neuron subtypes produced from IPs.
The second goal of the present study was to determine if IPs generate cortical layers in a predetermined sequential order depending on the timepoint of their generation, or if individual IP cohorts may contribute to multiple layers. Previously, early IPs were observed to rapidly differentiate and produce LL neurons (Kowalczyk et al., 2009), but the possibility that some early IPs differentiate slowly and contribute to ULs has not been thoroughly investigated. The latter scenario could arise if some early IPs are restricted to UL fates, as reported for some early RGPs (Franco et al., 2012); or if some early IPs remain in the mitotic cycle for a protracted period before differentiating, as suggested for late IPs (Wu et al., 2005). We studied these possibilities by inducible genetic fate mapping using Tbr2CreER, which, together with the reporter gene Ai14, specifically labels IPs and their progeny (Pimeisl et al., 2013).
To investigate regulation of laminar fate in IPs, we studied mice lacking Tbr2. Previously, it was reported that Tbr2-deficient cortex had decreased thickness of ULs, suggesting that Tbr2 promotes UL neuron fates (Arnold et al., 2008; Sessa et al., 2008). Here, we characterized the effects of Tbr2 deficiency on laminar fates of early, middle, and late IP cohorts. We studied IP genesis, migration, differentiation, and fates using molecular markers, lineage tracing, and cell birthdating. Unexpectedly, we found that Tbr2 is not required for genesis of subventricular IPs from RGPs as reported previously (Sessa et al., 2008), but is required mainly for differentiation of IPs to projection neurons, including acquisition of LL and UL identities.
Our results show that sequential cohorts of IPs generate progressively limited sets of cortical layers, such that UL neurogenesis depends not only on late, but also on early generated IPs. Furthermore, the process of laminar neurogenesis from IPs is regulated by Tbr2, which controls the tempo of LL and UL differentiation. Thus, in addition to previously described roles to control cortical regional patterning (Elsen at al., 2013), IPs and Tbr2 play important roles in laminar differentiation of the cerebral cortex.
To label IP cohorts and their progeny, we used Tbr2CreER;Ai14 mice (Pimeisl et al., 2013), in which tamoxifen (Tam) administration induces permanent expression of tdTomato, a red fluorescent protein (RFP), in Tbr2-expressing cells and their progeny. Tam was administered on embryonic days 11.5 (E11.5) to E16.5, and brains were collected on postnatal day (P) 0.5 (Figure S1). To identify neurons born on the day of Tam administration, bromodeoxyuridine (BrdU) was given concurrently (Figure 1).
Lineage tracing revealed that IP cohorts contributed differentially to cortical layers (Figure 1A,B). Early IP cohorts (E11.5–E12.5) generated not only LL neurons as expected from BrdU birthdating (Figure 1B–D), but also substantial numbers of UL neurons. Indeed, neurons derived from E11.5–E12.5 IPs were distributed bimodally in LLs and ULs, and as many as 17% of the RFP+ neurons generated from E12.5 IPs settled in ULs (Figure 1Ba,b). While LL neurons derived from early IPs were early-born, the UL neurons derived from early IPs were not early-born, as they did not incorporate BrdU given concurrently with Tam (Figure 1Da,b).
IPs from mid-neurogenesis (E13.5–E14.5) contributed widely to LLs and ULs, while late IPs (E15.5–E16.5) produced ULs selectively (Figure 1A–D).
Since UL neurons are generally born later in neurogenesis (Hevner et al., 2003), we considered the possibility that some early IPs (E12.5) do not undergo final mitosis until much later in neurogenesis. To test this hypothesis, we administered Tam on E12.5 and BrdU on E14.5, followed by survival to P0.5. With this schema, many double-labeled (RFP+/BrdU+) neurons were detected in ULs (but not LLs), confirming that some early IPs persisted for at least two embryonic days before final mitosis (Figure 1E). However, we do not know if the early IPs divided only once, or repeatedly during this period.
Cortical projection neurons are defined by not only laminar position, but also molecular expression (Hevner et al., 2003; Molyneaux et al., 2007). To further identify the neuron subtypes produced from E12.5 IPs, we evaluated their expression of Reelin, a Cajal-Retzius (C-R) cell marker; Tbr1, expressed at high levels in layer 6 (L6); Ctip2 (L5); Satb2 (callosal projection neurons); Cux1 (L2–4); and Dlx (interneuron precursors) (Figure S2). Remarkably, E12.5 IPs produced all five subtypes of projection neurons (Figure S3). As expected, no IP-derived neurons expressed Dlx (data not shown; see also Figure 2L,L').
To molecularly characterize the subtypes of projection neurons produced by sequential IP cohorts, we extended our analysis to include E11.5–E16.5 IP cohorts (Figure 2). Cajal-Retzius cells (Reelin+) were produced from early IPs (E11.5–E12.5) only (Figure 2A,A’,F). LL neurons (Tbr1+ or Ctip2+) were produced from early to middle (E11.5–E14.5) IPs only. In contrast, UL neurons (Cux1+) and callosal projection neurons (Satb2+) were produced from all IP cohorts (Figure 2). Some RFP+ progeny of late IPs (E15.5–E16.5) were located in LLs on P0.5 (Figures 1Ae,f, 2E,L), but these appeared to be migrating neurons destined for ULs, as they had elongated morphologies and did not express LL markers (Figure S4). Thus, at the population level, early IPs exhibit diverse laminar and molecular fates, while later IP cohorts have progressively limited fates.
The interpretation of fates from E11.5–E12.5 IPs was complicated by the fact that at these ages, Tbr2 is expressed by not only IPs, but also postmitotic C-R and subplate (SP) neurons (Englund et al., 2005). Thus, C-R and SP neurons may be labeled by Tam treatment on E11.5 or E12.5, even if these neurons were not derived from IPs. To verify the origins of C-R and SP neurons from IPs, we injected Tam and BrdU on E11.5, and studied cortex on E15.5. Triple-label immunofluorescence (IF) demonstrated that many Reelin+ C-R cells, and Calretinin+ SP neurons, were labeled with RFP and BrdU, confirming their origins from proliferating Tbr2+ IPs (Figure 2M,N). Thus, IPs produce preplate (C-R and SP), as well as LL and UL projection neuron subtypes.
To test if Tbr2 regulates laminar fate in IPs cell-autonomously, we utilized Tbr2CreER to inactivate floxed Tbr2 (Tbr2FL; Intlekofer et al., 2008) in IP cohorts. Administration of Tam to Tbr2CreER/FL;Ai14 mice resulted in RFP labeling of Tbr2-deficient IPs and their progeny. (However, IPs presumably expressed functional Tbr2 transiently before Tbr2FL was recombined.) We labeled cohorts of Tbr2-deficient IPs on E12.5 or E14.5, and evaluated RFP+ progeny on P0.5. Control mice were treated identically, but lacked the Tbr2FL allele (i.e., Tbr2CreER/+;Ai14).
Tbr2-deficient E12.5 IPs produced significantly fewer total neurons than did control E12.5 IPs (control=49±4 cells/mm2, cKO=31±2 cells/mm2; p=0.0004; Figure 3A–D). Among Tbr2-deficient IPs, fewer became SP neurons (WT=16.4±2%, cKO=1.6±1%; p=7.1*10−7), while more became L5 neurons (WT=27±3%, cKO=39±4%; p=0.043) and L2–4 neurons (WT=9±2%, cKO=20±4%; p=0.0145; Figure 3A–C). The upward laminar shift of neurons from Tbr2-deficient IPs was matched by genesis of fewer Tbr1+ SP neurons (WT=38±5%, cKO=18±4%; p=0.0154), but increased numbers of Satb2+ callosal projection neurons (WT=15±2%, cKO=40±8%; p=0.0209), and Cux1+ UL neurons (WT=3±1%, cKO=40±11%; p=0.0104). Also, Tbr2-deficient E12.5 IPs were more likely to delay final division until E14.5 (WT=12±7%, cKO=30.7±6%; p=0.0493; Figure 3E). Thus, E12.5 IPs require Tbr2 mainly for genesis of early-born LL neuron subtypes, but not for later-born callosal and UL neuron subtypes.
The effects of Tbr2 deficiency on E14.5 IPs were broadly similar as on E12.5 IPs. Laminar fates were shifted upwards (Figure 3F–H), with more IP-derived neurons distributed in ULs 2–4 (WT=18±2%, cKO=45.4±2%; p=6.7*10−13), and fewer in L5 (WT=50.7±2%, cKO=30.6±1%; p=2.7*10−12) and L6 (WT=30.8±2%, cKO=17.9±1%; p=1.4*10−8). Also, Tbr2-deficient E14.5 IPs produced fewer Tbr1+ (SP/L6) neurons (WT=19.7±5%, cKO=5.9±1%; p=0.0144), and were less likely to differentiate rapidly on E14.5 (WT=25.2±9%, cKO=11.4±2%; p=0.0357; Figure 3I, J).
Together, these results suggest that Tbr2 is necessary for differentiation of IPs to produce rapidly-generated neuron types. The shift to genesis of later-generated neuron types from Tbr2-deficient IPs may result from delayed differentiation related to molecular dysregulation (see below).
We next investigated the effects of complete Tbr2 deficiency on cortical neurogenesis. To produce mice lacking Tbr2 in the nervous system, we recombined Tbr2FL using Nes11Cre (Tronche et al., 1999). For simplicity, we designated the Nes11Cre;Tbr2FL/FL mice as Tbr2 conditional knockout (cKO) mutants.
To profile the timing of neurogenesis in control and Tbr2 cKO cortex, we labeled early- (E12.5), middle- (E14.5), or late-born (E16.5) neurons with BrdU, and studied the distribution of BrdU+ cells after survival to P2. As expected from lineage tracing of Tbr2-deficient IPs (Figure 3), neurogenesis was moderately decreased in Tbr2 cKO cortex on E14.5 (79% of control; p=0.016) and E16.5 (71% of control; p=0.06), while the laminar fates of these cells were unchanged or shifted slightly upwards (Figure 4B,C). In contrast, the genesis of E12.5-born neurons was significantly increased by 2.3-fold (p=0.0001) in Tbr2 cKO cortex (Figure 4A).
The increased genesis of E12.5-born neurons in Tbr2 cKO cortex suggested that SP and L6 thickness might be increased postnatally, as SP and L6 are the major neuron types born on E12.5 in normal mice (Figure 1C). Instead, we found significant expansion of L5 (Ctip2+) neurons (1.7 fold; p=3.2*10−8), at the expense of SP/L6 (Tbr1+, 76%; p=0.0008) and L2–4 (Cux1+, 50%; p=0.0124) neurons (Figure 4D,E). Thus, approximately twice as many E12.5-born cells differentiated as Ctip2+ L5 neurons (p=0.0139) in Tbr2 cKO mutants as in controls (Figure S5B). These results indicated that the relation between cell birthday and laminar fate was perturbed in Tbr2 cKO cortex. Interestingly, one previous study also reported slight expansion of L5 in Tbr2 cKO cortex, although not statistically significant (Arnold et al., 2008).
Tbr2 cKO mice survived to adulthood, but had 20–30% reduced body and brain mass (Figure S5F–J). (The causes of reduced body mass in Tbr2 cKO mice remain uncertain, but could include reduced feeding, hyperactivity, or hormonal changes.)
To evaluate motor development, we used rotarod and balance beam tests. Paradoxically, Tbr2 cKO mice had enhanced performance on both tests, although the enhancement declined in older mutants (Figure S5C,E). Importantly, body size did not correlate with motor performance (R2<0.02, Figure S5D), so the enhancement in Tbr2 cKO mutants cannot be attributed to lower body mass.
Other brain abnormalities in Tbr2 cKO mice included severe olfactory bulb hypoplasia, and reduced cortical surface area (Figure S5H,J; see also Arnold et al., 2008). In contrast, cortical thickness was not significantly reduced (Figure 4). Also, the anterior commissure was absent, although the corpus callosum and hippocampal commissure showed no obvious defects (Figure S6G,H; see also Hodge et al., 2013).
To investigate molecular defects in Tbr2 cKO mutants, we analyzed microarray data comparing E14.5 WT and Tbr2 cKO neocortex from our previous study, which focused on rostrocaudal identity (Elsen et al., 2013). Here, we focused on critical genes in neurogenesis and laminar fate acquisition. Genes up- or down-regulated in Tbr2 cKO neocortex were identified by positive or negative log2FC (log2 of the fold change) values, indicating significant differences (p<0.05) between Tbr2 cKO and control cortex.
Interestingly, transcription factor genes "upstream" of Tbr2 were up-regulated in Tbr2 cKO cortex, including Pax6 (log2FC= +0.36; p=0.001) and Insm1 (log2FC= +0.46; p=0.001). In contrast, "downstream" markers of laminar differentiation were mixed, with down-regulation of Tbr1 (log2FC= −0.78; p=0.00004), Bcl11b/Ctip2 (log2FC= −0.38; p=0.009) and Satb2 (log2FC= -1.47; p=0.00008), but up-regulation of Zfpm2/FOG2 (log2FC= +0.63; p=0.0005) and Adcyap1/PACAP (log2FC= +1.49; p<10−7). Interestingly, both Tbr1 and FOG2 are L6 markers (Bedogni et al., 2010), but they were regulated in opposite directions on E14.5, as were L5 markers Ctip2 and PACAP. These data suggested that differentiation of postmitotic neurons was severely dysregulated in Tbr2 cKO cortex.
To further investigate the differentiation defects inferred from microarray analysis, we studied the expression of neuronal differentiation markers by immunofluorescence. This approach allowed us to define not only quantitative changes in gene expression, but also qualitative changes in zonal differentiation patterns (Bystron et al., 2008).
Patterns of neuron differentiation were profoundly disturbed in E14.5 Tbr2 cKO cortex (Figure 5). The cortical plate (CP) and intermediate zone (IZ) appeared thin, and fewer cells expressed markers of postmitotic LL (Tbr1+, 67%; p=0.0126) and callosal (Satb2+, 9%; p=5.9*10−6) differentiation (Figure 5A,B). In contrast, FOG2 (L6, 1.5 fold increase; p=0.0038) and Ctip2 (L5, 83% decrease; p=0.0441), which are restricted to the CP on E14.5 in WT embryos, showed ectopic expression in the IZ of Tbr2 cKO mutants (Figure 5C,D). PACAP+ cells, representing a subset of L5 neurons (Lodato et al., 2014), were more abundant and more immunoreactive in the Tbr2 cKO CP (Figure 5E). C-R neurons (Reelin+) appeared slightly increased in Tbr2 cKO mice (data not shown), similarly as in Pax6 mutants (Stoykova et al., 2003). Together, these results demonstrated that projection neuron differentiation was severely disorganized and dysregulated in E14.5 Tbr2 cKO mutants.
To further characterize the trajectory of cortical differentiation in Tbr2 cKO cortex, we studied E12.5 and E16.5 time points (Figure S6). In E12.5 mutants, the preplate appeared thicker than normal, due to an abundance of NeuroD+/Tbr1− immature neurons, along with approximately normal numbers of Tbr1+ neurons (Figure S6A,B). Interestingly, the boundary between VZ (Sox2+) and preplate (Tbr1+) appeared irregular due to the accumulation of immature neurons (Figure S6B). These results suggested that differentiation of preplate neurons was impaired in E12.5 Tbr2 cKO mutants, and early neurogenesis was increased (Figure 4A) in compensation.
In E16.5 Tbr2 mutants, the numbers of Tbr1+, Ctip2+, and PACAP+ LL neurons were strikingly increased over WT, and CP thickness was increased (Figure S6C,D,E). PACAP+ neurons also expressed Ctip2 (Figure S6E,F), supporting their identity as a subset of L5 neurons (Lodato et al., 2014). In contrast to increased LL thickness, UL thickness was decreased in E16.5 Tbr2 cKO cortex compared to WT (Figure S6C,D), and ULs remained thin on P2 (Figure 4D,E).
Together with BrdU birthdating (Figure 4A–C), these data indicated that early, middle, and late phases of cortical differentiation were severely abnormal in Tbr2 cKO cortex. The delayed differentiation of early-born neurons (Figure S6A,B) may account for the shift from L6 to L5 fates (Figure 4D,E), and for the compensatory burst of early neurogenesis (Figure 4A) in Tbr2 cKO mutants. In turn, excessive early neurogenesis may have depleted progenitors, leading to decreased neurogenesis by E14.5–E16.5 (Figure 4B,C) and consequent thinning of ULs (Figure 4D,E).
The defects of neuronal differentiation in Tbr2 cKO neocortex might be attributable to defective genesis and/or differentiation of IPs. To identify IPs and distinguish them from RGPs, we studied abventricular (basal) mitoses (AVMs) and ventricular (apical) mitoses (VMs) by phospho-histone H3 (pH3) IF (Kowalczyk et al., 2009).
In E12.5 Tbr2 cKO mutants, the number of AVMs was unchanged from controls, but ventricular mitoses (VMs) were increased to 1.7-fold (p=4.4*10−9), and S-phase cells (acute BrdU+) to 1.2-fold (p=0.0245) of control values (Figure 6A–E). By E14.5, VMs normalized, but AVMs increased to significantly exceed control numbers (1.4-fold; p=4.7*10−5; Figure 6F,G). Moreover, AVMs in Tbr2 cKO cortex were not confined to the VZ/SVZ as in controls, but were also found ectopically in the IZ and CP (Figure 6G,J). Furthermore, acute BrdU+ (S-phase) cells were also observed in the Tbr2 cKO IZ and CP, and BrdU+ cells in the VZ were disorganized (Figure 6G,H). Nevertheless, the total numbers of BrdU+ cells were similar in E14.5 mutants and controls. These results indicated that basal IPs were not diminished, but were actually increased in Tbr2 cKO mutants, contradicting previous reports (Arnold et al., 2008; Sessa et al., 2008).
We next studied expression of Pax6 and Insm1, transcription factors upstream of Tbr2 that promote genesis of Tbr2+ IPs (Quinn et al., 2007; Farkas et al., 2008; Sansom et al., 2009). Strikingly, Pax6+ and Insm1+ cells were not only increased 1.4-fold (p=0.0011) and 2.6-fold (p=4.3*10−10), respectively, but were also located ectopically in the IZ and CP of Tbr2 mutants (Figure 7A,B). Moreover, pH3+ AVMs were more likely to express Pax6 (3.6-fold; p=0.024) and Insm1 (24-fold; p=3*10−5) in mutant than in control cortex (Figure 7C,D), suggesting that Tbr2-deficient IPs failed to down-regulate these transcription factors (Englund et al., 2005; Farkas et al., 2008).
We next studied expression of NeuroD, a transcription factor expressed in basal IPs and newly generated neurons (Hevner et al., 2006). In E14.5 Tbr2 cKO cortex, NeuroD was expressed by increased numbers of cells (1.2-fold; p=0.033), including many in the IZ, SP and CP, demonstrating ectopic expression of NeuroD in neuronal maturation zones (Figures 7E and S7A–D). The fraction of AVMs that expressed NeuroD was also increased in Tbr2 cKO cortex relative to controls (1.7-fold; p=0.0105; Figure 7E), consistent with protracted IP differentiation despite active NeuroD expression. Many NeuroD+ cells aberrantly co-expressed Pax6 in Tbr2 mutant cortex (25-fold more than in WT; Figure S7D,E). Thus, the differentiation of IPs and new neurons was disorganized and prolonged in Tbr2 cKO cortex.
Together, these findings indicate that Tbr2 is not necessary for IP genesis, but is required to promote the transition from IPs to postmitotic neurons (Figure 7F). In the absence of Tbr2, IP genesis continues, AVMs accumulate, and differentiation of IPs to neurons is profoundly abnormal.
In the present study, we found that IP cohorts make complex contributions to cortical layers, including an unexpected contribution from early IPs to upper cortical layers. We also showed that Tbr2 regulates laminar organization of the cortex, by facilitating the transition from IP to neuron, and promoting the timely acquisition of laminar identity.
The finding that some early IPs produce UL neurons (Figure 1A) suggested two possible interpretations. First, if laminar fate is specified in RGPs and some early RGPs have restricted UL fates (Franco et al., 2012), then early IPs inherit UL fates from parent RGPs. Alternatively, if IPs are initially multipotent with regard to laminar identities, then daughter neuron fates may be determined by the timing of final mitosis, and limited by progressive fate restriction (Desai and McConnell, 2000). The latter possibility is favored by previous evidence that IPs can divide asymmetrically (with respect to laminar fate) to produce multiple layers (Wu et al., 2005). To resolve this issue, clonal analysis of IP lineages will be necessary.
Our findings challenge the previous conclusion that Tbr2 is required primarily for IP genesis (Sessa et al., 2008). Specifically, we found that basal mitoses were not reduced, but were actually increased in E14.5 Tbr2 cKO neocortex (Figure 6F,G,I). In contrast, previous studies reported that basal mitoses were significantly depleted in Tbr2 cKO neocortex (Arnold et al., 2008; Sessa et al., 2008). We attribute these discrepancies to different Cre drivers and floxed Tbr2 alleles. The previous study implicating Tbr2 in IP genesis (Sessa et al., 2008) used Foxg1Cre, a knockin allele that itself causes IP depletion (Siegenthaler et al., 2008). Another previous study (Arnold et al., 2008) used Sox1Cre, likewise a knockin allele (Takashima et al., 2007) that causes defects of brain development and function (Malas et al., 2003). In contrast, Nes11Cre, used in the present study, is a transgene that does not directly interfere with gene expression, or with brain development and function (Tronche et al., 1999).
The aberrant and protracted differentiation of neurons in embryonic Tbr2 cKO cortex can be traced to gene dysregulation in IPs. The ectopic expression of Pax6 and Insm1 in Tbr2 cKO IZ and CP (Figure 7) suggests that Tbr2 is required to downregulate these transcription factors, possibly by direct transcriptional repression in IPs. Indeed, Tbr2 binding sites are found near the Pax6 and Insm1 promoters (Teo et al., 2011). The persistent expression of Pax6 and Insm1 may interfere with neuronal differentiation in cortex, as demonstrated for ectopic Pax6 in the spinal cord (Bel-Vialar et al., 2007). Thus, Tbr2 appears to facilitate neuronal differentiation in part by repressing molecules that are normally expressed only in progenitor cells.
Conversely, Tbr2 may direct neuronal maturation by activating transcription of molecules expressed in differentiating neurons, such as Tbr1 and Satb2. Consistent with this possibility, both Satb2 and Tbr1 are initially detected in basal IPs, albeit at low levels (Britanova et al., 2005; Nelson et al., 2013). On the other hand, some important neuronal differentiation factors (such as NeuroD) are clearly not dependent on Tbr2. Also, despite abnormal gene expression in the Tbr2 cKO cortex, most projection neurons ultimately differentiated successfully. Indeed, the Tbr2 cKO cortex underwent a marked change between E14.5, when the CP was thin with a paucity of Tbr1+ neurons (Figure 5A), and E16.5, when the CP was thick with abundant Tbr1+ and Ctip2+ neurons (Figure S6C,D). The number of Satb2+ neurons likewise recovered substantially after E14.5 (Figure S6I). In the absence of Tbr2, upregulation of other molecules, such as NeuroD (Figure S7A–D), may compensate to ensure neuronal differentiation.
Tbr2 appears to regulate laminar fate by multiple mechanisms. Inactivation of Tbr2 in IP cohorts (E12.5 and E14.5) led to reduced genesis of rapidly-generated neuron subtypes (deeper in cortex), and relatively increased genesis of later-generated subtypes (more superficial) (Figure 3). Since Tbr2-deficient cohorts were sparse, these results indicated that Tbr2 is required cell-autonomously for rapid IP differentiation and neurogenesis. Extrapolating from these findings, Tbr2 cKO (throughout cortex) may have delayed the differentiation of all IP cohorts, causing an overall shift away from early-born neuron subtypes, towards increased genesis of later-born subtypes. Indeed, Tbr1+ neurons and L6 thickness were decreased in Tbr2 cKO mice (Figure 4E). However, Tbr2 may also regulate the balance of L6 and L5 fates directly: E12.5-born (BrdU+) cells were more likely to differentiate as Ctip2+ neurons in Tbr2 cKO cortex (Figure S5Bc), and L5 markers PACAP and Ctip2 were markedly increased by E16.5 in Tbr2 cKO cortex (Fig. S6D–F). Thus, Tbr2 appears to directly regulate both the rate of IP differentiation, and the balance of L6 and L5 fates during early neurogenesis.
Despite the delayed differentiation of IPs, neurogenesis was initially accelerated in E12.5 Tbr2 cKO cortex, but decreased subsequently on E14.5 and E16.5 (Figure 4A–C). The acceleration of early neurogenesis may represent a non-autonomous effect of altered IP differentiation. Such effects are anticipated because IPs interact with RGPs, for example, by Delta-Notch signaling (Nelson et al., 2013). We speculate that deficient Delta-Notch signaling in early Tbr2 cKO cortex caused RGPs to respond by increasing direct neurogenesis (Figure 4A) and overproducing preplate neurons (Figure S6A,B). In turn, the RGP pool may have been depleted prematurely in Tbr2 cKO mutants, thus accounting for reduced genesis of late-born UL neurons, and decreased UL thickness (Figure 4B–D). In sum, laminar defects in Tbr2 cKO cortex reflect a complex system of differentiation and feedback.
Interestingly, the reduction of late neurogenesis and UL thickness occurred despite ample production of basal progenitors in Tbr2 cKO cortex (Figure 6I). Previous studies have shown that IP genesis is driven by low Notch signaling in RGPs (Nelson et al., 2013), and by neurogenic transcription factors including Pax6, Insm1, and Neurog2 (Sun and Hevner, 2014). Those "upstream" mechanisms of IP specification occur in RGPs prior to the expression of Tbr2, so our finding that IP genesis was spared in Tbr2 cKO cortex is logical. Rather, Tbr2 deficiency perturbed gene expression in new IPs, and impaired their ability to differentiate as cortical projection neurons with well-defined laminar subtype identities. Our findings indicate that Tbr2 plays an important transitional role in neurogenesis, by both suppressing RGP identity and promoting specific features of cortical layers. Ultimately, neuronal differentiation and layer formation were delayed in Tbr2 cKO mice, but proceeded to completion due to compensatory mechanisms.
Remarkably, major motor skills were not impaired in Tbr2 cKO mice (Figure S5C,E), although previous studies detected hyperactivity and weakness (Arnold et al., 2008). The small olfactory bulb and rudimentary dentate gyrus in Tbr2 mutants (Hodge et al., 2013) presumably impair olfaction and memory, but those functions have not been tested. Interestingly, Nlgn3 mutant mice also show improved performance on repetitive motor tasks (Rothwell et al., 2014).
In sum, we have shown that IPs can persist in the cortex for prolonged periods, and that early IP cohorts contribute to multiple cortical layers. The pace of laminar neurogenesis, and the identities of projection neurons, are regulated by Tbr2 although the genesis of IPs is not. In future studies, it will be interesting to conduct clonal analysis of IP progeny, and determine if individual IPs contribute to multiple layers, as well as the size and distribution of IP-derived clones.
C57BL/6 mice used in this study were kept in a 12 hour light/dark cycle, with food and water ad-libitum, in Seattle Children’s Research Institute’ vivarium. All animal experimental procedures were performed with Institutional Animal Care and Use Committee approval. The following previously described mouse transgenic alleles were used: Ai14 reporter (Madisen et al., 2010), EomesCreER (Tbr2CreER) (Pimeisl et al., 2013), Nestin-Cre (Nes11Cre) (Tronche et al., 1999) (stock 003771, Jackson Labs), Tbr2-Flox (Tbr2FL) (Intlekofer et al., 2008). Also see Supplemental Experimental Procedures.
Pregnant dams were administered Tamoxifen (Sigma, T5648; 5mg/kg), Progesterone (Sigma, P3972; 2.5mg/kg) and BrdU (Sigma, B5002; 50mg/kg), by intraperitoneal injection, at the indicated embryonic ages. Acute BrdU treatment was done 30 minutes before brain collection.
Single plane optical sections and stacks were acquired with Zeiss LSM-710 confocal microscope. Cell counts were reported either as absolute number or density per area (mm2), or as distribution per bin/cortical zone. Also see Supplemental Experimental Procedures. Data was reported as mean ± SEM, from at least 3 sections from an animal, and 2–4 animals per condition/data point. Statistical analysis used 2-tailed, unpaired Student’s t test, and the confidence threshold chosen is p<0.05.
Male mice between 6 weeks and 6 months of age were tested on the Rotamex-5 (Columbus Instruments) for rotarod performance to assess their motor skills as described (Hsu et al., 2014). Also see Supplemental Experimental Procedures.
We thank E. Young and L. Honican for assistance with experiments and analysis. We thank Dr. C. Birchmeier (Max-Delbrück-Center for Molecular Medicine, Berlin) for Insm1 antibody, Dr. J. Kohtz (Northwestern Univ., Chicago) for Dlx antibody and Dr. J. Hannibal for PACAP antibody (University of Copenhagen). We thank Dr. T. Bammler, Dr. F. Farin, and Dr. D. Beyer (University of Washington Center for Ecogenetics and Environmental Health) for assistance with microarray experiments, and the UW Center on Human Development and Disability for partial support of microarray experiments (supported by grant U54 HD083091 from the National Institute of Child Health and Human Development to the University of Washington’s Center on Human Development and Disability). This study was supported by National Institutes of Health Grants R01 NS085081 and R01 NS092339 to RFH, German Research Foundation (DFG) Emmy Noether Programme (AR732/1-1) to SJA, Lejeune Foundation and ProRett Italia to FB.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Author ContributionsABM and RFH designed the study and wrote the manuscript. ABM, RAD, KAR, GEE and FB performed experiments. SJA provided research materials and manuscript editing/commenting. ABM and RFH analyzed data.