The chromatin-remodeling protein Satb2 plays a role in the generation of distinct subtypes of neocortical pyramidal neurons. Previous studies have shown that Satb2 is required for normal development of callosal projection neurons (CPNs), which fail to extend axons callosally in the absence of Satb2 and instead project subcortically. Here we conditionally delete Satb2 from the developing neocortex and find that neurons in the upper layers adopt some electrophysiological properties characteristic of deep layer neurons, but projections from the superficial layers do not contribute to the aberrant subcortical projections seen in Satb2 mutants. Instead, axons from deep layer CPNs descend subcortically in the absence of Satb2. These data demonstrate distinct developmental roles of Satb2 in regulating the fates of upper and deep layer neurons. Unexpectedly, Satb2 mutant brains also display changes in gene expression by subcerebral projection neurons (SCPNs), accompanied by a failure of corticospinal tract (CST) formation. Altering the timing of Satb2 ablation reveals that SCPNs require an early expression of Satb2 for differentiation and extension of the CST, suggesting that early transient expression of Satb2 in these cells plays an essential role in development. Collectively these data show that Satb2 is required by both CPNs and SCPNs for proper differentiation and axon pathfinding.
While the mammalian cerebral cortex contains a multitude of distinct cell types, neurons can be broadly classified into 2 major groups: excitatory glutamatergic projection neurons and GABAergic inhibitory interneurons. The projection neurons can, in turn, be subdivided into 3 major subclasses (reviewed in Leone et al. 2008; Greig et al. 2013): corticofugal projection neurons, commissural projection neurons, and ipsilateral local circuit connection neurons (associative projection neurons). Corticofugal projection neurons (also called subcortical projection neurons) are located in the deep Layers 5 and 6 and send their axons to subcortical targets such as the thalamus and subcerebral targets such as the midbrain and spinal cord (subcerebral projection neurons, SCPNs). Commissural projection neurons send interhemispheric axons across the midline either along the corpus callosum (callosal projection neurons; CPNs; reviewed in Fame et al. 2011), or through the anterior commissure. The majority of CPNs are found in superficial layers, but they also populate the deep Layers 5 and 6.
The production of these distinct subtypes requires a number of key transcription factors including Fezf2 (also known as Fezl or Zfp312), which regulates the specification of SCPNs (Molyneaux et al. 2005; Chen et al. 2005,, 2008; Shim et al. 2012), Ctip2 (also known as Bcl11b), which is necessary for axonal outgrowth of SCPNs (Arlotta et al. 2005), Tbr1, which is required for Layer 6 corticothalamic neurons (Hevner et al. 2001; Han et al. 2011; McKenna et al. 2011), and Satb2, which is essential for the specification of CPNs (Alcamo et al. 2008; Britanova et al. 2008; Zhang et al. 2012). These factors interact with additional transcription factors that further refine fate acquisition. For example, Sox5 controls the timing of corticofugal neuron generation (Lai et al. 2008) and is also required postmitotically for the repression of Fezf2, ultimately regulating the generation of distinct subtype identities in Layers 5, 6, and the subplate (Kwan et al. 2008). In addition, the transcription factor Bhlhb5 is required during the postmitotic acquisition of area identity and differentiation of SCPNs (Joshi et al. 2008). Taken together, these data suggest that initial specification of pyramidal neurons occurs in early postmitotic, immature neurons, but definitive fates are achieved during differentiation.
The chromatin-remodeling protein Satb2 was first identified as a matrix attachment region (MAR)-binding protein (Dobreva et al. 2003), and we and others subsequently showed that Satb2 is a major regulator of CPNs in the developing cerebral cortex (Alcamo et al. 2008; Britanova et al. 2008). The axons of Satb2-deficient pyramidal neurons fail to cross the corpus callosum and instead project subcortically. In Satb2 mutants, expression of Ctip2, a transcription factor normally expressed by Layer 5 neurons (Arlotta et al. 2005; Molyneaux et al. 2005), is expanded into the superficial layers, suggesting that CPNs in the absence of Satb2 adopt characteristics of subcortical projection neurons. Biochemical studies revealed that Satb2 binds to regulatory regions (MAR sequences) in the Ctip2 locus and thereby represses Ctip2 expression directly (Alcamo et al. 2008). However, effective repression of Ctip2 requires the protooncogene Ski, which is necessary for the formation of a multi-protein complex consisting of Ski, Satb2, and the histone deacetylases HDAC1 and MTA2 that enables Satb2 to act as a repressor (Baranek et al. 2012).
The fates of cortical projection neurons are defined by a large number of factors besides their long-distance projections, since cells in different layers develop distinct electrophysiological signatures, local axonal connections, dendritic morphologies, and patterns of gene expression (reviewed in Molyneaux et al. 2007; Leone et al. 2008; Greig et al. 2013). In our previous studies, we were unable to examine many of these features that emerge postnatally, since mice bearing a null mutation in Satb2 die at birth. Here we analyze conditional Satb2 mice using different Cre driver lines that allowed us to bypass the early lethality of a null mutation and to ablate Satb2 at different time points during development. These studies reveal distinct roles of Satb2 in regulating the electrophysiological characteristics of upper layer neurons and the long-distance connections of CPNs in the upper versus deep layers. In addition, we have discovered an unexpected role for Satb2 in the differentiation of Fezf2-expressing SCPNs and the corticospinal tract (CST).
Materials and Methods
Conditional mutants were generated by breeding Emx1-Cre;Satb2+/lacZ, Rbp4-Cre;Satb2+/lacZ or Nestin-CreERT2;Satb2+/lacZ mice with Satb2lox/lox animals to obtain mutant (Cre+;Satb2lox/lacZ) and control (Cre+;Satb2+/lox) littermates. For some experiments, Cre+;Satb2+/lacZ controls were used. Finally, for some experiments, animals were also heterozygous for the Fezf2AP allele (Chen et al. 2005) or the reporter allele Ai9 (Madisen et al. 2010). See Supplementary Methods for genotyping. The morning of the vaginal plug was defined as E0.5. For Nestin-CreERT2 experiments, pregnant females were injected with a single dose of 0.2–0.3 mg Tamoxifen (Sigma) dissolved in Dimethyl sulfoxide (Sigma).
Histology, Immunocytochemistry, Antibodies, and In Situ Hybridization
Standard protocols for immunohistochemistry were used. Perinatal animals were perfused on postnatal day P4 with 4% paraformaldehyde (PFA) supplemented with 0.1% saponin in Dulbecco’s phosphate buffered saline (PBS), and all other postnatal animals were perfused with 4% paraformaldehyde in PBS. Embryonic brains were dissected and drop-fixed in 4% paraformaldehyde at least overnight. Brains were cryoprotected in 30% sucrose in PBS and either embedded in OCT tissue tek for cryosectioning (20 µm) or directly cut on a sliding microtome (70 µm). Radioactive in situ hybridization was carried out as previously described (Frantz et al. 1994); see Supplementary Methods for list of probes.
We used the following primary antibodies: rabbit anti-Satb2 (Abcam), mouse anti-Satb2 (Abcam), rat anti-Ctip2 (Abcam), rabbit anti-red fluorescent protein (RFP) (Clontech), rabbit anti-protein kinase Cγ (PKCγ) (Santa Cruz Biotechnology), chicken anti-β-galactosidase (Abcam), rabbit anti-Tbr1 (Abcam), rabbit anti-Fog2 (Abcam), and rat anti-L1 (Millipore). Alexa Fluor-coupled secondary antibodies were used to detect primary antibodies.
For alkaline phosphatase (AP) staining, fixed brains were incubated in PBS at 65°C for 25 min to inactivate endogenous phosphatases, stained with NBT/BCIP (Roche) in 100 mM Tris–HCl, 50 mM MgCl2, pH 9.5 at 37°C and postfixed with PFA. For X-gal histochemistry, brains were fixed in 2% formaldehyde/0.2% glutaraldehyde in PBS for 10 min., stained with 1 mg/mL X-gal for ∼2 h, and postfixed with PFA.
Confocal images were acquired on a Zeiss LSM 510 meta microscope, and epifluorescent pictures were taken on a Nikon 80i with a Hamamatsu Orca ER camera. Images were postprocessed using Adobe Photoshop CS3 and ImageJ.
For retrograde tracing, P15 animals were anesthetized with ketamine and xylazine and their heads were stabilized in a stereotaxic frame. The brain was exposed and 0.1 µL of fluorescent labeled latex beads (Lumafluor Inc.) were injected into either the thalamus, the anterior pretectal nucleus, or the cerebral peduncle using a stereotaxic injector (Stoelting). Animals were sacrificed 48–60 h after surgery to allow for transport of the beads.
Anterograde tracings were performed according to instructions (Vector Laboratories; Gerfen and Sawchenko 1984). Briefly, P15 animals were anesthetized and PHA-L was infused using glass capillaries with an interior diameter of 15 µm. Iontophoresis was performed using 5 µA pulses for 10–15 min using 7 s intervals (7 s current, 7 s pause). After allowing 48 h for PHA-L transport, animals were perfused and processed using the Vectastain ABC kit (Vector Laboratories).
Acute coronal slices (300 µm) were prepared from P21–P25 mutant and control littermates. Slices were incubated at 37°C for 1 h in artificial cerebrospinal fluid (ACSF) containing 126 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 26 mM NaHCO3, 1 mM MgSO4·7H2O, 2 mM CaCl2·H2O, and 10 mM glucose, equilibrated with 95% O2 and 5% CO2 prior to being transferred to a submerged recording chamber perfused with warmed (30–34°C), oxygenated ACSF at 1–2 mL/min. Whole-cell patch-clamp recordings in current-clamp mode were made from superficial layer pyramidal neurons in presence of synaptic blockers (50 µM picrotoxin and 1 mM kynurenic acid). Recording electrodes made of borosilicate glass were pulled to a final tip resistance of 3–6 MΩ and filled with internal solution as per Chen et al. (2008). Intracellular recordings and analysis were performed using pClamp (Molecular Devices). Mean resting membrane potentials were −71.9 and −75.4 mV for mutants and controls, respectively, and input resistance were 178 and 133 MΩ for mutants and controls, respectively, not significantly different between the 2 groups (n = 17 and 19 for mutant and control group, respectively).
Altered Projections in Satb2-deficient Mice
A lacZ gene inserted into the Satb2 null allele (Supplementary Fig. 1) previously allowed us to visualize changes in the projections of Satb2-expressing neurons (Alcamo et al. 2008; Srinivasan et al. 2012), but perinatal lethality precluded a detailed postnatal analysis of these tracts. To circumvent early lethality, we employed a conditional Satb2 allele (Satb2lox, Supplementary Fig. 1) in tandem with the Emx1-Cre driver line, which limits recombination to dorsal neocortex (Gorski et al. 2002). We assessed the efficiency of recombination by using the Ai9 reporter allele (Madisen et al. 2010), in which a CAG-driven RFP cDNA, preceded by a floxed-stop cassette, is inserted into the Rosa26 locus (Soriano 1999), leading to RFP expression in recombined cells. The Ai9 reporter construct carries at its 3′ end a woodchuck hepatitis virus post-transcriptional regulatory element that enhances mRNA stability (Madisen et al. 2010), enabling visualization of axons and dendritic arbors in RFP+ neurons. Emx1-Cre conferred efficient recombination of both the Satb2lox (Fig. 1) and Ai9 reporter alleles. In control Emx1-Cre;Satb2lox/+;Ai9 mice at P4, Satb2 protein was readily detected in neocortex (Fig. 1A), and the vast majority of Satb2+ neurons in both superficial (Fig. 1C,C′) and deep layers (Fig. 1E) colabeled with RFP. In Emx1-Cre;Satb2 mutants, Satb2 expression was lost in all layers (Fig. 1B,D,D′,F), suggesting a high recombination efficiency for Emx1-Cre. Conditional mutant mice survived the juvenile period, but died between P25–30 for unknown reasons. In both null and conditional mutants, Satb2-expressing neurons and their axons are marked by β-galactosidase expression.
In previous studies of Satb2−/− mice, the axons of CPNs failed to extend across the corpus callosum and instead formed aberrant subcortical projections; in addition, the deep layer markers Ctip2, Ldb2, Grb14, and Kitl were upregulated in superficial layer neurons, suggesting that the fates of these cells had been altered (Alcamo et al. 2008; Britanova et al. 2008). Conditional mutants recapitulate many but not all of these phenotypes. Interestingly, whereas Satb2−/− mice at E18.5 lacked labeled axons in the corpus callosum (Alcamo et al. 2008), examination of postnatal Emx1-Cre;Satb2 mutants revealed some β-galactosidase+ axons in the corpus callosum and others following an aberrant pathway into the septum (Supplementary Fig. 2B). It is conceivable that Satb2 is required for early-born pioneer axons to cross the corpus callosum, but that some axons at later stages can reach the contralateral hemisphere in the absence of Satb2. Consistent with this possibility, the expression of Crystallin µ (CRYM) expanded into the superficial layers in cingulate cortex (Fig. 2B), which contains neurons that pioneer the corpus callosum (Koester and O'Leary 1994; Rash and Richards 2001; Fothergill et al. 2013). This region normally expresses the zinc-finger protein Fog-2 (also known as Zfpm2; Fig. 2C), but in Emx1-Cre;Satb2 mutants, Fog-2 expression in superficial layers of cingulate cortex is abrogated (Fig. 2D). These changes may relate to the failure of callosal development in Satb2 null mutants at E18.5 and a defect in the ability of the cingulate pioneers and early-generated neurons to extend callosal axons. In the neocortex, CRYM expression, which is normally present in differentiating SCPNs (Fig. 2A,A′; Arlotta et al. 2005), was markedly expanded across the entire cerebral wall in Emx1-Cre;Satb2 mutants (Fig. 2B,B′), adding to the list of deep layer markers that are upregulated in upper layer neurons in the absence of Satb2. The SCPN markers Diap3 and Crim1 showed no obvious changes in conditional mutants (data not shown).
Consistent with previous reports, control animals showed no β-galactosidase labeling of subcortical axon tracts in the thalamus or cerebral peduncle, although weak β-galactosidase activity was present in the internal capsule (Supplementary Fig. 2C) likely resulting from callosal neurons that extend collateral axons to the striatum (Mitchell and Macklis 2005; reviewed in Fame et al. 2011). Both null and conditional mutants showed a dramatic increase in β-galactosidase+ axons extending subcortically within the internal capsule (Supplementary Fig. 2B), thalamus and cerebral peduncle (Supplementary Fig. 2D). In postnatal Emx1-Cre;Satb2 mutants we also observed β-galactosidase+ axons within the CST (Supplementary Fig. 2F), providing new evidence that the identity of at least some Satb2-expressing neurons is altered toward a SCPN fate in Satb2 mutants. However, these studies do not address whether the β-galactosidase+ CST axons arise from upper layer neurons or from the smaller population of Satb2+ cells that reside in the deep layers.
Superficial layer neurons in Satb2 mutants adopt some electrophysiological properties that are characteristic of deep layer neurons.
To further assess whether upper layer neurons adopt characteristics typical of deep layer cells in the absence of Satb2, we examined their electrophysiological properties. Normal Layer 5 SCPNs fire distinct bursts with fixed, short interspike intervals in response to a current injection (Kasper et al. 1994), and are characterized by voltage responses that include an early peak and then decay, the so-called sag response (Stafstrom et al. 1984). Superficial layer CPNs, on the other hand, have been described as non-bursting (Larkman and Mason 1990; Mason and Larkman 1990; Kasper et al. 1994) and they lack a voltage sag response.
To investigate whether changes in gene expression by upper layer neurons in Satb2 mutants correlate with changes in layer-specific electrophysiological properties, we recorded from superficial layers of Emx1-Cre;Satb2 mutants and littermate controls. Acute coronal slices were prepared from P25 animals and whole-cell patch-clamp recordings in current-clamp mode were made from Layer 2/3 pyramidal neurons in primary somatosensory cortex. We did not detect significant differences in action potential threshold, firing frequency, burst firing or adaptation ratio between mutant and control neurons (Supplementary Table 1). We did, however, detect differences in these cells' responses to injection of a 100 pA hyperpolarizing current step, and representative voltage responses from a control and mutant neuron are shown in Figure 3A. Emx1-Cre;Satb2 mutant neurons showed a significantly increased sag response (which immediately follows current injection; Fig. 3A,C) and a doubling of the after-depolarization in response to current injection (Fig. 3A,C). Full families of voltage responses to current steps from −100 to +100 pA are shown in Supplementary Figure 3. Quantification revealed statistically significant increases in both the sag response (mutants: 2.17 ± 0.39 mV; controls: 0.42 ± 0.1 mV; P < 0.01) and after depolarization (mutants: 2.06 ± 0.30 mV; controls: 0.75 ± 0.16 mV; P < 0.01).
The dramatic increase in sag response in Emx1-Cre;Satb2 mutants (Fig. 3C) suggests a prominent hyperpolarization-activated current Ih, an inward, excitatory current that is normally high in SCPNs, where it plays important roles in rhythmic firing (resonance) and production of sag currents; Ih is low in corticostriatal and corticocortical projection neurons (Dembrow et al. 2010; Sheets et al. 2011). Analysis of peak versus end-of-pulse voltage responses in Satb2 mutants and controls (Fig. 3B) revealed a statistically significant interaction (2-way analysis of variance [ANOVA]; F1,25 = 13.12, P = 0.0013) after correction for multiple comparisons (Holm–Sidak method). Ih currents are mediated by hyperpolarization-activated cyclic nucleotide-gated (HCN) cation channels, which are encoded by the HCN1–4 gene family. In a microarray screen for genes that showed differential expression in the Satb2−/− mutant cortex, we found that HCN1 is upregulated 3.8-fold in Satb2 mutants compared with controls (data not shown). To validate the array findings, we performed in situ hybridization for HCN1 at E18.5. In control mice, HCN1 is expressed at high levels in the deep layers (Fig. 3D), in agreement with previous reports (Nolan et al. 2003; Sheets et al. 2011; Stoenica et al. 2013). However, in Satb2−/− mutants, HCN1 expression was elevated in the superficial layers (as well as in Layer 6; Fig. 3E). Surprisingly, we measured a statistically significantly higher membrane resistance in Emx1-Cre;Satb2 mutants compared with controls (0.178 ± 0.014 GΩ vs. 0.133 ± 0.016 GΩ; Supplementary Table 1). The mechanism for the increased input resistance remains unknown, but did not affect our ability to detect the sag response induced by current injections that produced similar levels of hyperpolarization. Taken together, our data show that Satb2-deficient superficial layer neurons display a strong sag response, a high Ih current and a striking increase in HCN1 expression, all of which are normally characteristic of Layer 5 SCPNs.
Aberrant Subcortical Projections in Satb2 Mutants Arise Exclusively From the Deep Layers
The anatomical, electrophysiological, and molecular changes observed in Emx1-Cre;Satb2 mutants raise the possibility that superficial layer neurons contribute to the aberrant extension of subcortical projections by β-galactosidase+ axons. To address this hypothesis, we performed a more detailed analysis of subcortical projections and their laminar origins in mutant and control animals.
We first examined corticothalamic projections, which can be visualized using the golli-τ-GFP transgene (golli-GFP; Jacobs et al. 2007) in which 1.3 kb of the golli promotor of the myelin basic protein gene directs the expression of a τ-EGFP cassette to deep layer pyramidal neurons, including Layer 6 corticothalamic neurons (Supplementary Fig. 4) and CPNs in Layers 5 and 6 that project across the corpus callosum (data not shown). In sagittal sections of P4 control animals expressing golli-GFP, robust GFP expression marked Layers 5 and 6 (Supplementary Fig. 4A) as well as axons in the internal capsule and thalamus (Supplementary Fig. 4A), as expected. In Emx1-Cre;Satb2 mutant brains, golli-GFP reactivity remained confined to the deep layers, although it appears increased in Layer 6 (Supplementary Fig. 4B). While thalamic GFP expression appeared similar to controls, axonal GFP+ staining in the mutant internal capsule (Supplementary Fig. 4B) appeared more prominent. The failure of upper layer neurons to acquire golli-GFP expression, together with enhanced staining of axons in the internal capsule, raised an alternative to the hypothesis that upper layer neurons redirect their axons in Satb2 mutants. These observations suggest that the source of subcortically extending β-galactosidase+ axons might actually be Satb2-expressing CPNs that reside in the deep layers.
To directly assess the laminar origins of subcortical axons in the mutants, we first injected fluorescent labeled latex beads into the thalamus of P15 animals for retrograde labeling and analyzed laminar positions of retrogradely labeled neurons 2 days after surgery. Labeled neurons were detected exclusively in Layer 6 of both control mice (Fig. 4A) and Emx1-Cre;Satb2 mutants (Fig. 4B). We then injected fluorescent microspheres into the cerebral peduncle at the midbrain/pons junction (to target SCPNs in primary motor cortex M1) and into the anterior pretectal nucleus (to target corticotectal projection neurons) of P15 animals and similarly analyzed the positions of retrogradely labeled cells 2 days later. Strikingly, we did not detect any retrogradely labeled neurons in the superficial layers of Emx1-Cre;Satb2 mutants; instead we observed labeled cells in Layer 5 of both controls and mutants following injections into the cerebral peduncle (Fig. 4C,D) or into the pretectal nucleus (Fig. 4E,F). Thus, despite the extensive number of β-galactosidase+ axons that extend to subcortical targets in Satb2 mutants (Supplementary Fig. 2D and F), these retrograde labeling experiments suggest that Satb2-expressing neurons in the upper layers do not contribute to these novel projections. Instead, the axons must arise from deep layer Satb2-expressing cells, which normally form callosal and corticocortical projections.
Our data suggest that neurons that normally express Satb2 show a differential ability to form subcerebral projections, depending on their laminar position. Deep layer CPNs can form subcerebral projections in the absence of Satb2, but superficial layer CPNs appear unable to do so. To further explore this hypothesis, we utilized a Tamoxifen-inducible Nestin-CreERT2 line (Imayoshi et al. 2006) that targets neural progenitors in a Tamoxifen-dependent manner: early Tamoxifen administration at E11.5 induces recombination in progenitors that will sequentially generate both deep layer neurons and, via later subventricular zone progenitors, neurons of the upper layers (Noctor et al. 2001; Miyata et al. 2001). Late Tamoxifen administration, on the other hand, leads to recombination in the late progenitor pool that gives rise to superficial layer neurons, bypassing already-born deep layer neurons. This enabled us to target Cre-mediated recombination of Satb2 exclusively to upper layer neurons.
As a control, we first administered 0.3 mg Tamoxifen to pregnant Nestin-CreERT2;Ai9+ females at E11.5 and analyzed their offspring at P4. RFP expression was detected in a large fraction of Satb2-expressing neurons of the superficial layers (Supplementary Fig. 5A) and in deep layer pyramidal neurons (Supplementary Fig. 5B), indicating widespread recombination across the cortical wall after a single dose of Tamoxifen. As predicted, a single Tamoxifen administration at E15.5 led to RFP expression predominantly in superficial layers (Supplementary Fig. 5C and E), but, as expected, few if any RFP+ neurons were observed in deep layers (Supplementary Fig. 5F).
To compare the effects of ablating Satb2 in neurons of all layers versus exclusively in upper layer cells, we administered 0.3 mg Tamoxifen to pregnant mice at E11.5 or E15.5 and analyzed Nestin-CreERT2+;Satb2lox/lacZ offspring at P4. These brains showed mosaicism for Satb2: non-recombined neurons (Satb2lox/lacZ) have one functional copy of Satb2 and should extend β-galactosidase+ projections across the corpus callosum. Satb2Δ/lacZ recombined neurons (lox allele recombined; Δ), on the other hand, should serve as the origin of β-galactosidase+ subcortical projections. In sagittal sections of P4 animals treated with a single Tamoxifen dose at E11.5, which enabled recombination in all layers, β-galactosidase+ axons were present in the internal capsule and thalamus (Fig. 4G), and in the CST along the brainstem (Fig. 4I). However, in animals treated with Tamoxifen at E15.5 to target recombination exclusively to the upper layers, no β-galactosidase+ axons were detected in the internal capsule or thalamus (Fig. 4H), nor within the brainstem (Fig. 4J). These data suggest strongly that Satb2-deficient superficial layer neurons lack the capability to form subcortical projections. Collectively our retrograde labeling experiments and timed Tamoxifen Nestin-CreERT2 experiments suggest that the loss of Satb2 causes deep layer CPNs to project subcortically, but reveal that superficial layer callosal neurons fail to contribute to subcortical β-galactosidase+ projections.
To characterize the projection patterns of Satb2-deficient CPNs in more detail, we placed injections of an anterograde tracer in the somatosensory cortex of Emx1-Cre;Satb2 control and mutant animals. Phaseolus vulgaris leucoagglutinin (PHA-L) was administered by iontophoretic injection (Gerfen and Sawchenko 1984) at P15, and animals were analyzed 48 h later (Fig. 4K,L). Although we attempted to selectively label the upper layers, we found that PHA-L also labeled deep layer neurons, perhaps via their dendrites in the superficial layers. PHA-L injections in somatosensory cortex of controls resulted in labeling of both ipsilateral and callosally projecting axons, as well as axons that extended subcortically (Fig. 4K). In Emx1-Cre;Satb2 mutants, however, very few PHA-L labeled axons were visible in the corpus callosum or contralateral hemisphere; instead we observed a qualitative increase in ipsilateral projections. Many labeled axons in the region of the midline projected ventrally toward the septum, taking a pathway similar to that of β-galactosidase+ fibers (Supplementary Fig. 2B), suggesting that some of the Satb2-deficient pyramidal neurons aberrantly project into the septum.
A Surprising Role for Satb2 in CST Formation by Layer 5 SCPNs
During our analysis of subcortical connectivity in Emx1-Cre;Satb2 mutants, we decided to investigate CST formation by taking advantage of the AP marker utilized to generate a null allele of Fezf2 (Fezf2AP), which labels the axonal projections of Fezf2+ neurons (Chen et al. 2005). We showed previously that Fezf2 expression is not altered in Satb2−/− animals (Alcamo et al. 2008), thus Fezf2AP can be used to mark the axons of deep layer SCPNs.
In sagittal sections of P4 brains, Fezf2AP labels the CST along the cerebral peduncle, the brainstem, through the pyramidal decussation into the spinal cord (Fig. 5A). To our surprise, in Emx1-Cre;Satb2 mutants carrying the Fezf2AP allele, AP activity was not detected caudal of the cerebral peduncle (Fig. 5B). Cross sections through the brainstem at 2 separate levels confirmed these findings: in controls, strong AP reactivity was observed at the ventral surface where the pyramids are located (Fig. 5C,C′), but in Emx1-Cre;Satb2 mutants, no AP reactivity was found (Fig. 5D, D′). Ventral views of whole-mount brain preparations further supported the loss of the AP+ CST in Emx1-Cre;Satb2 mutants: controls revealed a robust AP signal in the cerebral peduncle, pyramids and pyramidal decussation (Fig. 5E). In Emx1-Cre;Satb2 mutants, however, AP marked the cerebral peduncle but no signal was detectable in the brainstem (Fig. 5F). To determine whether the lack of AP+ CST axons in Emx1-Cre;Satb2 mutants is due to a developmental delay, we analyzed cross sections of the cervical spinal cord at P15. Controls showed strong AP labeling of the CST within the ventral dorsal funiculus (Fig. 5G), whereas no signal was detected in Emx1-Cre;Satb2 mutants (Fig. 5H).
To preclude the possibility that Fezf2AP might simply be downregulated in the mutants, we labeled the CST at P15 using an antibody against PKCγ, a CST-specific marker (Mori et al. 1990). In control sagittal sections, PKCγ strongly labeled the CST at the cerebral peduncle, through the pyramidal decussation and in the spinal cord (Fig. 5I). In Emx1-Cre;Satb2 mutants, however, strong PKCγ labeling was only found at the cerebral peduncle; weak labeling was found at the level of the pons and very few PKCγ+ axons were visible in the brainstem (Fig. 5J), suggesting a failure of CST formation at the cerebral peduncle. Immunolabeling of P15 cross sections through the cervical spinal cord revealed robust PKCγ staining in the control CST at the ventral dorsal funiculus, as well as in spinal interneurons (Fig. 5K). In Emx1-Cre;Satb2 mutants, PKCγ expression by interneurons was present but CST labeling was absent (Fig. 5L). Taken together, our findings indicate that the Emx1-Cre-induced loss of Satb2 leads to the unexpected failure of Fezf2-expressing Layer 5 neurons to extend CST axons past the cerebral peduncle.
Transient Expression of Satb2 in Ctip2+ Layer 5 Neurons During Cortical Development
Satb2 function in the developing brain has been associated with the specification of CPNs (Alcamo et al. 2008; Britanova et al. 2008; Baranek et al. 2012), but not with that of SCPNs. The failure of Layer 5 neurons to form a normal CST in Emx1-Cre;Satb2 mutants suggests either that Satb2 is playing a non-cell-autonomous role in their development, or that an early, transient, expression of Satb2 in SCPNs plays a critical role in their differentiation. Our previous studies suggested that as many 40% of Layer 5 neurons coexpress Ctip2 and Satb2 during embryonic development (Alcamo et al. 2008). We reexamined this issue by performing a systematic analysis of Ctip2/Satb2 coexpression between E13.5 and P4. We note that the primary antibodies used here differ from those used previously (Alcamo et al. 2008), which may account for quantitative differences between these 2 studies.
From E13.5, the peak time of Layer 5 neurogenesis, through E15.5, when Satb2 expression is becoming more robust, we find that few Ctip2+ cells in the developing cortical plate coexpressed Satb2 (Fig. 6A–F). Between E16.5 and P4, roughly 20% of Ctip2+ neurons also expressed Satb2 (Fig. 6G–J). After P7, Satb2 expression in the cortex gradually decreases, and by P15 we were unable to detect either Satb2 protein or β-galactosidase in Satb2+/lacZ animals (data not shown). These data suggest that Satb2 and Ctip2 are coexpressed in a substantial fraction (20% to 40%, depending on the antibodies employed) of Layer 5 neurons between E16 and P4; these numbers would underestimate the fraction if coexpression is asynchronous within Layer 5. More importantly, these observations raise the possibility that the transient expression of Satb2 in SCPNs might play a role in their normal differentiation.
In support of this hypothesis, we observed alterations in the expression of 2 genes that are normally enriched in SCPNs, the Forkhead box transcription factors Foxo1 and Foxp2, in Emx1-Cre;Satb2 mutants. Foxo1, which is expressed in Layer 5 of control brains (Fig. 7A,A′) at P1, is lost in Emx1-Cre;Satb2 mutants (Fig. 7B,B′). Similarly, Foxp2, which is normally expressed in Layers 5 and 6 (Fig. 7C,C′), is lost in Layer 5 of Emx1-Cre;Satb2 mutants but its expression is not obviously affected in Layer 6 (Fig. 7D,D′). The layer 5-specific loss of both transcription factors in Emx1-Cre;Satb2 mutants indicate that differentiation of Layer 5 SCPNs is incomplete, supporting a role for Satb2 in the postmitotic differentiation of SCPNs.
Early Transient Expression of Satb2 is Required for SCPN Development
To test its role in SCPN differentiation, we attempted to ablate Satb2 specifically in Layer 5 neurons at a stage later than that conferred by Emx1-Cre, which mediates recombination as early as E10.5 (Gorski et al. 2002). The BAC transgenic line Rbp4-Cre (Gong et al. 2007) targets Layer 5 neurons (Kozorovitskiy et al. 2012; Glickfeld et al. 2013); however, a detailed characterization of the line and the onset of Cre activity have not yet been described. We first investigated the specificity and timing of Rbp4-Cre-mediated recombination using the Ai9 reporter. In Rbp4-Cre+;Ai9+ animals, RFP+ neurons were detected as early as E16.5 (Fig. 8A,B); however, at E16.5 and E17.5, only a small fraction of Ctip2+ neurons expressed RFP (Fig. 8C). By P4, RFP expression in Layer 5 was robust and widespread, and the majority of Ctip2+ neurons were marked with RFP (Fig. 8D–F). The axonal transport of RFP labeled the internal capsule (Fig. 8D), cerebral peduncle (Fig. 8H), pyramidal decussation, and CST (Fig. 8I); RFP was also present in the anterior commissure (Fig. 8G), indicating that Rbp4-Cre recombines both SCPNs and commissural neurons of Layer 5. These results demonstrate that Rbp4-Cre shows a high specificity for Layer 5 neurons. The earliest activity on the reporter allele was detected at E16.5, several days after the birth of Layer 5 neurons at around E13.5, suggesting that Rbp4-Cre ablates target genes during Layer 5 neuron differentiation and maturation.
To assess whether Rbp4-Cre effectively ablates Satb2 in Layer 5, we analyzed Rbp4-Cre;Satb2;Ai9 animals for Satb2 and RFP at P4. In controls, RFP was readily observed in Layer 5 (Fig. 9A), where some cells colocalized RFP and Satb2 (Fig. 9A′, similar to the coexpression of Ctip2 and Satb2 in wildtype mice in Fig. 6J). In Rbp4-Cre;Satb2 mutants, however, the fraction of RFP+ neurons that expressed Satb2 was dramatically reduced (Fig. 9B′), indicating that Rbp4-Cre efficiently ablates Satb2 in the RFP+ population.
To investigate whether late Rbp4-Cre-mediated recombination phenocopies the failure of CST development observed in Emx1-Cre;Satb2 mutants, we analyzed both lines at P4 using the Ai9 reporter to specifically label the axons of recombined neurons. In control Emx1-Cre mice that were heterozygous for Satb2, RFP marked subcortical axons of the CST in the brainstem and pyramidal decussation (Fig. 9C). In Emx1-Cre;Satb2 mutant littermates, however, robust RFP expression terminated at the cerebral peduncle; only very weak signal was visible in the brainstem (Fig. 9D; compare with Fig. 5B,F,J). In the Rbp4-Cre;Satb2 strain, on the other hand, RFP robustly marked CST axons extending through the brainstem, pyramidal decussation and into the spinal cord of both controls and mutants (Fig. 9E,F).
To investigate CST development more caudally in both strains, we immunolabeled sections through the spinal cord of P15 animals for RFP and PKCγ. In Emx1-Cre;Satb2 heterozygous controls, the ventral dorsal funiculus showed strong labeling with both markers (Fig. 9G). In Emx1-Cre;Satb2 mutants, however, neither RFP nor PKCγ was detectable (Fig. 9H), confirming the failure of the CST rostrally. In Rbp4-Cre;Satb2 mice, RFP and PKCγ labeled the CST in both controls and mutants (Fig. 9I,J), suggesting that Satb2 is dispensable during the late maturation (from ∼E17.5 to P4) of Layer 5 SCPNs. Since results from Emx1-Cre;Satb2 mutants demonstrate that Satb2 is required for normal CST formation, our data are consistent with the possibility that the transient expression of Satb2 in layer 5 SCPNs plays an unexpected, cell-autonomous role early in their differentiation, before E17.5. Additional studies will be required to rule out non-cell-autonomous functions.
Previous studies of the chromatin-remodeling protein Satb2 have focused on its role in regulating the specification and differentiation of callosal projection neurons in the developing neocortex (Alcamo et al. 2008; Britanova et al. 2008; Gyorgy et al. 2008; Srinivasan et al. 2012). We have broadened these studies and extended them into novel realms. Using a conditional allele of Satb2 in combination with the cortex-specific Emx1-Cre driver, we first found that the β-galactosidase+ axons of Satb2-expressing neurons are capable of extending into the CST during postnatal life, further demonstrating that CPNs can adopt an SCPN fate in the absence of Satb2. Surprisingly, though, retrograde labeling studies and the use of different Cre driver lines revealed that these novel projections arise exclusively from the deep layers of cortex, and not from the upper layers. Thus, although the loss of Satb2 leads to electrophysiological changes in superficial layer neurons and upregulation of several genes normally expressed in Layer 5, our studies reveal that upper layer neurons are unable to convert entirely to an SCPN fate—their axons do not extend subcortically.
Distinct Developmental Potentials of Upper and Deep Layer Cortical Neurons
Our results suggest that the populations of Satb2-expressing neurons in the upper and deep layers are fundamentally different, despite their molecular similarities. Perhaps this ought not to have been surprising since superficial layer CPNs are a recent acquisition during evolution, first observed in placental mammals (reviewed in Aboitiz and Montiel 2003 and Fame et al. 2011). In addition, it has long been appreciated that deep layer CPNs are generated side by side with SCPNs from Fezf2-expressing progenitor cells; it seems reasonable that deep layer CPNs retain enough of their genetic heritage to project subcortically in the absence of Satb2. Superficial layer neurons, on the other hand, are derived from the evolutionary recent cortical expansion involving the emergence of the subventricular zone (SVZ). A number of genes and molecular markers are uniquely associated with SVZ progenitors and their superficial layer progeny, including Tbr2 (Baala et al. 2007; Arnold et al. 2008; Sessa et al. 2008), Svet1 (Tarabykin et al. 2001), and Cux1 and Cux2 (Zimmer et al. 2004; Cubelos et al. 2010). While SVZ and upper layer cells might retain or reuse genes like Satb2 that are part of an evolutionarily more ancient genetic program, the interpretation of this program may differ profoundly in the context of a distinct transcriptional and layer-specific environment.
The concept that Satb2 has been “redeployed” by neurons in the evolutionary recent superficial layers helps to reconcile previous models of inhibitory interactions between Fezf2, Satb2, and Ctip2 (Alcamo et al. 2008; Britanova et al. 2008; Gyorgy et al. 2008; Chen et al. 2008; Srinivasan et al. 2012) with the current data. Our findings suggest that these genetic interactions are cell-type and context dependent—indeed, one key difference between deep and upper layer neurons is that the protooncogene Ski is coexpressed with Satb2 in superficial layers, and Ski plays a critical role in enabling Satb2 to form a repressor complex in situ (Baranek et al. 2012). Satb2 binds to MAR in regulatory sequences of Ctip2 and represses Ctip2 by assembling 2 members of the nucleosome remodeling and deacetylase (NuRD) complex, the histone deacetylases HDAC1 and MTA2 (Alcamo et al. 2008; Britanova et al. 2008). Ski is not required for the interaction between Satb2 and the MAR sequences, however, it is required for the recruitment of HDAC1, thereby allowing the NuRD complex to assemble, which then leads to the repression of Ctip2.
These studies suggest that the presence of Ski in superficial layer neurons enables the formation of a Satb2/Ski/NuRD complex and the efficient repression of Ctip2. In addition, upper layer neurons fail to express critical deep layer determinants such as Fezf2, even in the absence of Satb2 function. The presence of Ski and absence of Fezf2 expression may account for the inability of upper layer neurons to completely adopt an SCPN fate in the absence of Satb2.
In the deep layers, on the other hand, the percentage of Ski+;Satb2+ neurons is much lower than in the superficial layers, suggesting that Satb2 may function independently of Ski (Baranek et al. 2012). Indeed, the substantial coexpression of Satb2 and Ctip2 in Layer 5 neurons during development suggests that Satb2 is less effective in inhibiting Ctip2 expression compared with cells in the upper layers. Satb proteins can regulate gene expression by activating or repressing transcription and by altering chromatin structure (Cai et al. 2003,, 2006; Dobreva et al. 2006), and we speculate that Satb2 may function in deep layer neurons by recruiting as yet unidentified partners in the regulation of target gene expression.
An Unexpected Role for Satb2 in the Development of Layer 5 SCPNs
Our experiments unveiled an unexpected role for Satb2 in the differentiation of Layer 5 SCPNs. In Satb2 mutant animals, the expression of several Layer 5 differentiation markers was reduced compared with wildtype controls, indicating an incomplete differentiation of Satb2-deficient SCPNs. These molecular changes were accompanied by a failure of the CST to extend past the cerebral peduncle and into the spinal cord. The use of a Layer 5-specific Cre driver showed that the ablation of Satb2 at relatively late stages of Layer 5 differentiation failed to disrupt CST formation, whereas recombination at early times resulted in a CST phenotype. These data are consistent with the possibility that the transient expression of Satb2 in differentiating SCPNs may actually be required for their normal development. However, further experiments will be needed to rule out a non-cell-autonomous role.
We previously demonstrated that several axon guidance ligands and receptors show altered expression in Satb2-deficient brains, and in particular we observed a dramatic reduction in the expression of Unc5h3 and EphA4 (Alcamo et al. 2008). Both receptors have been implicated in CST development and pathfinding (reviewed in Harel and Strittmatter 2006): in both Unc5h3 and EphA4 mutants, most CST axons terminate along the medulla, rostral to the pyramidal decussation (Finger et al. 2002; Dottori et al. 1998), a phenotype quite similar to that observed here. It is therefore tempting to speculate that the loss of Unc5h3 and EphA4 expression in Satb2 mutants (Alcamo et al. 2008) leads to defects in axon guidance by SCPNs and a failure of normal CST formation. Whether Satb2 directly regulates the expression of Unc5h3 and EphA4 in Layer 5 neurons remains to be determined.
This work was supported by grants from the National Institutes of Health (EY08411 and MH051864)
We are grateful to Dr James Weimann for technical help with retrograde tracing and Dr Chris Kaznowski for technical help. We would also like to thank Dr Elizabeth Alcamo for previous experiments. Conflict of Interest: None declared.