Mutual regulation between Satb2 and Fezf2 promotes subcerebral projection neuron identity in the developing cerebral cortex

Significance

Mutations in special AT-rich sequence-binding protein 2 (Satb2) cause severe intellectual deficiency in humans. However, its function in brain development is not completely understood. Our study focuses on the function of Satb2 in specifying cortical projection neuron fates. We find that, although Satb2 activates the expression of some subcerebral neuronal genes, it also inhibits the expression of other genes that are expressed in subcerebral neurons. We report that Satb2 promotes Fezf2 and Sox5 expression in subcerebral neurons, and that Fezf2 in turn inhibits high-level Satb2 expression. We show that the mutual regulation between Satb2 and Fezf2 is essential for Satb2 to promote subcerebral neuron fate.

Abstract

Generation of distinct cortical projection neuron subtypes during development relies in part on repression of alternative neuron identities. It was reported that the special AT-rich sequence-binding protein 2 (Satb2) is required for proper development of callosal neuron identity and represses expression of genes that are essential for subcerebral axon development. Surprisingly, Satb2 has recently been shown to be necessary for subcerebral axon development. Here, we unravel a previously unidentified mechanism underlying this paradox. We show that SATB2 directly activates transcription of forebrain embryonic zinc finger 2 (Fezf2) and SRY-box 5 (Sox5), genes essential for subcerebral neuron development. We find that the mutual regulation between Satb2 and Fezf2 enables Satb2 to promote subcerebral neuron identity in layer 5 neurons, and to repress subcerebral characters in callosal neurons. Thus, Satb2 promotes the development of callosal and subcerebral neurons in a cell context-dependent manner.

Projection neurons in the six-layered neocortex can be classified based on axonal projections. The corticocortical neurons send axons to other areas in the ipsilateral cortex or through the corpus callosum and into the contralateral cortex (i.e., callosal projection neurons). The callosal neurons are distributed throughout layers 2–6, but are most abundant in layers 2 and 3. The corticofugal neurons consist of corticothalamic and subcerebral neurons. The corticothalamic neurons are most abundant in layer 6, and send axons into the thalamus. The subcerebral neurons are located in layer 5, and project axons into the spinal cord, superior colliculus, pons, and other brain areas (1–4).

Much progress has been made toward understanding how corticofugal neurons are specified during development (1–4). The zinc-finger transcription factor Fezf2 (also known as Fezl and Zfp312) is essential for specifying subcerebral neuron identity by repressing alternate corticothalamic and callosal fates (5–9). In Fezf2^−/− mice, subcerebral neurons were lost; instead, mutant layer 5 neurons differentiated into corticothalamic or callosal neurons. Another zinc-finger transcription factor, Ctip2 (also known as Bcl11b), is important for the development of subcerebral axons. The subcerebral axons defasciculated in Ctip2^−/− mice and failed to reach the pyramidal decussation (8, 10). The T-box–containing transcription factor Tbr1 promotes corticothalamic neuron identity by repressing subcerebral fate in layer 6 (9, 11–13). TBR1 binds to a 3′ region of the Fezf2 gene and directly represses high-level Fezf2 expression in layer 6 neurons (9, 11). Thus, the expression level of Fezf2 is precisely regulated in the developing cortex for proper specification of subcerebral and corticothalamic neuronal fates (14, 15).

The special AT-rich sequence-binding protein 2 (SATB2) is involved in transcriptional regulation and chromatin remodeling (16–19). Human mutations in Satb2 cause severe intellectual deficiency, language impairment, and behavioral defects (20–23). However, the function of Satb2 in cortical neuron development remains incompletely understood. Previous studies using mouse models revealed that SATB2 was essential for specifying callosal neuronal fate by repressing subcerebral identity in these cells (17–19, 24). Intriguingly, a recent paper reported that SATB2 is also required for the development of subcerebral axons (25). However, it remains unknown how SATB2 performs two seemingly opposing functions. In this study, we investigate how SATB2 promotes subcerebral neuron identity despite repressing the expression of some subcerebral neuronal genes.

Results

SATB2 Is Expressed in Subcerebral and Corticothalamic Projection Neurons During Development.

We determined whether SATB2 is expressed in corticofugal neurons by examining axons from SATB2-expressing neurons using the Satb2^lacZ allele, in which a beta-galactosidase gene (lacZ) was inserted in the Satb2 locus (16–18). In the postnatal day (P) 0 Satb2^lacZ/+ mice, LacZ⁺ axons were observed in the internal capsule, cerebral peduncle, and thalamus; however, compared with the callosal axons, labeling of the corticofugal axons was less intense (Fig. S1 A–C).

Fig. S1.

(A–C) LacZ staining shows that Satb2-LacZ⁺ axons are present in the corpus callosum (red star, A), internal capsule (red arrow, A), cerebral peduncle (red arrow, C), and thalamus (red arrow, B). LacZ staining in the cerebral peduncle and thalamus is less intense than in the corpus callosum. (D–K) Immunostaining shows that the rabbit SATB2 antibody (Abcam) is specific to the SATB2 protein. (D) SATB2 immunostaining of a P0 WT cortex. (E) No SATB2 antibody staining was detected in the Satb2^lacZ/lacZ mice at P0. (F) Low-magnification DAPI image of the control (Satb2^flox/+) brain section shown in G and H. (G) SATB2 immunostaining of the section in F. (H) Higher-magnification image of the boxed area in F and G shows SATB2 staining. (I) DAPI staining of a Satb2 cKO (Satb2^flox/flox; Emx1-Cre) brain section. (J) SATB2 staining of the same brain section shown in I. (K) Higher-magnification image of the boxed area in I and J shows SATB2 staining. (Scale bars: A–C, 500 µm; D and E, 100 µm; F and I, 500 µm; G and J, 500 µm; H and K, 100 µm.)

To confirm that SATB2 is expressed in the corticofugal neurons, we combined cholera toxin β (CTβ) retrograde tracing from the cerebral peduncle or thalamus with SATB2 immunohistochemistry (Fig. 1). The specificity of the SATB2 antibody was confirmed by lack of staining in the neocortices of the Satb2^lacZ/lacZ and Satb2 conditional KO (Satb2 cKO) mice (Fig. S1 D–K). At P0, in general, SATB2 was expressed at higher levels in the upper layers and at lower levels in the deep layers (Fig. S1D). Low- or medium-level SATB2 expression was observed in 67% and 89% of traced subcerebral and corticothalamic neurons, respectively (Fig. 1 I–J). At P8, 45% and 59% of the labeled subcerebral and corticothalamic neurons, respectively, expressed low- or medium-level SATB2 (Fig. 1 I–J). Thus, many corticofugal neurons express SATB2 at birth and early postnatal stages.

Fig. 1.

SATB2 is expressed in subcerebral and corticothalamic neurons. Cortical neurons projecting to the cerebral peduncle (A–D) or the thalamus (E–H) were labeled by CTβ-Alexa Fluor retrograde tracing. The enlarged images of the boxed area in A are shown in B (CTβ), C (SATB2), and D (merge). The enlarged images of the boxed area in E are shown in F (CTβ), G (SATB2), and H (merge). The yellow arrows point to traced cells that expressed SATB2. The purple arrows point to traced cells not expressing detectable SATB2. (I and J) Percentages of traced subcerebral (I) and corticothalamic (J) neurons that expressed SATB2 at P0 and P8. Error bars indicate SD. (Scale bars: A and E, 100 µm; B–D, 50 µm; F–H, 50 µm.)

Misregulated Gene Expression in Satb2^lacZ/lacZ Cortices.

Previous studies showed that callosal and subcerebral axons failed to reach their targets in the Satb2 mutant mice (16–18, 25). To investigate the molecular underpinning of these defects, we performed gene expression analysis of Satb2^lacZ/lacZ cortices by using RNA sequencing (RNA-seq). Compared with Satb2^+/+ mice, expression of 1,277 genes was increased to at least 1.2 fold and expression of 1,852 genes was decreased by more than 20% in P0 Satb2^lacZ/lacZ cortices (adjusted P < 0.1; Fig. S2A). We examined the expression patterns of some of the misregulated genes at neonatal stages [embryonic day (E) 18.5 and P4] by using the online in situ hybridization database generated by Allen Brain Institute (Allen Brain Atlas). Consistent with previous reports that callosal neurons in Satb2 mutant mice showed molecular characteristics of subcerebral neurons (16–18), expression of many genes that are normally expressed in the deep layers was increased in the Satb2^lacZ/lacZ cortices (Fig. S2A and Table S1). In addition, we found that genes that are specifically expressed in upper-layer neurons in normal mice showed reduced expression in the mutant mice (Fig. S2A and Table S2). CUX1 (26) was expressed in layers 2–4 in control mice (Fig. 2G). In the Satb2 cKO mice, CUX1 expression was significantly reduced (Fig. 2J). Thus, in addition to repressing deep-layer gene expression in upper-layer neurons, Satb2 promotes callosal neuron identity through activating upper-layer gene expression.

Fig. S2.

Expressions of upper-layer neuronal genes and deep-layer neuronal genes were misregulated in Satb2 mutant mice. (A) RNA-seq analysis showed that 3,129 genes were misregulated (adjusted P < 0.1) in the P0 Satb2^lacZ/lacZ cortices. Compared with Satb2^+/+ cortices, at least 15 upper-layer genes (red tick marks) showed reduced expression, and at least 74 deep-layer genes (blue tick marks) showed increased or decreased expression in the Satb2^lacZ/lacZ cortices. (B–E) TLE4 expression is reduced in the P0 Satb2^lacZ/lacZ cortices. (B and C) Satb2^+/+ brains. (D and E) Satb2^lacZ/lacZ brains. (F–I) NFIA expression is reduced in the P0 Satb2^lacZ/lacZ brains. (F and G) Satb2^+/+ brains. (H and I) Satb2^lacZ/lacZ brains. (J and K) DARPP32 expression was reduced in the Satb2^lacZ/lacZ brains. (J) DARPP32 staining in Satb2^lacZ/+ cortices. (K) Satb2^lacZ/lacZ cortices. (Scale bars: B–E, 100 μm; F–I, 100 μm; J and K, 500 μm.)

Table S1.

Deep-layer genes based on in situ hybridization data in Allen Brain Atlas that show misregulated expression in the Satb2^lacZ/lacZ cortices

Gene name	Changes in Satb2lacZ/lacZ cortices	P value
Grm2	0.20529751	6.61E-75
Cdh6	0.27675038	1.91E-24
Npy	0.16025512	3.66E-84
Dkk3	0.30507669	3.79E-32
Tgfbr1	0.30654068	6.12E-27
Myc	0.35832214	7.73E-33
Sstr2	0.37881725	2.49E-12
Otx1	0.39990856	1.54E-15
Scg2	0.39992902	5.15E-08
Foxp2	0.40370881	6.72E-08
Nfe2l3	0.43725345	2.54E-13
Pdzd2	0.46468576	1.71E-16
Kcnab1	0.47601412	1.13E-05
Cyp26b1	0.47704768	2.34E-08
Nr4a3	0.48352674	1.15E-11
Ppp1r1b	0.48854776	4.84E-14
Lrrtm1	0.50806766	2.01E-17
Lypd1	0.5307642	6.19E-06
Grm3	0.53273906	8.76E-10
Fosl2	0.54704973	1.65E-10
Kit	0.58440677	2.78E-06
Nfia	0.584763	6.28E-07
Sfrp2	0.6428176	3.61E-04
Hivep2	0.64363187	3.77E-08
Sox5	0.65492649	1.29E-05
Fezf2	0.65668887	9.96E-07
Tle4	0.66497135	6.81E-06
Chrna4	0.70498442	1.32E-03
Tox	0.72285593	2.51E-03
Ptpru	0.73107426	4.51E-03
Cdh11	0.74388182	1.14E-02
Adnp2	0.74487771	4.70E-03
Atf4	0.7648128	2.12E-03
Robo1	0.7853079	3.78E-03
Kctd12	0.7905371	2.32E-03
Wnt7b	0.79813825	1.50E-03
Apba1	1.23865013	8.80E-03
Ldb2	1.25899226	1.27E-02
Atp1a1	1.29657591	4.21E-03
Efnb3	1.30906908	7.00E-04
Plxnd1	1.31274827	6.25E-04
S100a10	1.31547043	1.39E-02
Slc32a1	1.31927601	5.05E-03
Grm5	1.37255398	2.18E-04
Tmed3	1.38418955	1.32E-02
Reln	1.39426911	3.93E-05
Clu	1.40642198	1.64E-03
Dner	1.41103694	4.54E-05
Cpne2	1.43475201	6.17E-03
Vat1l	1.47598159	1.94E-03
Grin3a	1.53615672	1.13E-03
Doc2b	1.54698299	7.32E-05
Gria3	1.58524317	2.99E-03
Mgst3	1.62468235	2.25E-05
Zdhhc2	1.63901717	2.97E-03
Gls	1.74580402	9.65E-06
Ephb1	1.755501	2.53E-10
Cadm1	2.27607807	2.16E-16
Grik2	2.28630348	2.17E-08
Glra2	2.28872138	7.77E-11
Sema3c	2.3761516	6.57E-13
Bcl11b	2.53050009	1.31E-30
Pex5l	2.55438504	1.48E-07
Tmem163	2.76022443	2.19E-29
Kitl	2.90448579	3.75E-20
Slc16a2	2.92383611	1.58E-26
Cntn6	3.02923239	1.82E-08
Crtac1	3.3299394	5.61E-19
Cdh7	3.56685025	2.33E-04
Gda	3.59780471	1.64E-16
Ntng1	3.75020557	1.09E-10
Nov	4.54107996	1.89E-27
Crym	5.2450782	3.93E-41
Alcam	5.26687303	2.41E-39

Table S2.

Upper-layer genes based on Allen Brain Atlas that show decreased expression in the Satb2^lacZ/lacZ cortices

Gene name	Changes in Satb2lacZ/lacZ cortices	P value
Dtx4	0.27289644	3.25E-60
Ephb6	0.45689849	6.69E-23
Sema7a	0.47282016	8.17E-23
Cacna2d1	0.48087673	2.56E-09
Zfp462	0.48536269	9.02E-18
Bhlhe40	0.49070703	1.92E-07
Cux1	0.4971219	1.19E-14
Odz1	0.53998315	5.06E-05
Trpc4	0.57444866	9.36E-06
Pvrl3	0.58411498	1.10E-03
Lhx2	0.63835929	3.90E-09
Cux2	0.72810086	1.32E-05
Mef2c	0.73833696	5.83E-05
Ntng2	0.75951106	1.10E-03
Mdga1	0.8267283	1.45E-02

Fig. 2.

SATB2 regulates Fezf2 expression. (A and B) In situ hybridization showed that Fezf2 mRNA expression was reduced in P0 Satb2^lacZ/lacZ cortices. (A′ and B′) Enlarged views of the cerebral cortex showing Fezf2 mRNA expression in Satb2^+/+ (A′) and Satb2^lacZ/lacZ (B′) cortices. (C) The normalized expression level of Fezf2 mRNA in P0 Satb2^+/+ and Satb2^lacZ/lacZ cortices revealed by RNA-seq (*P < 0.001). (D) ChIP-seq revealed that SATB2 bound to the enhancer 434 (red) of the Fezf2 gene (blue). (E) Luciferase assay showed that cotransfection of Satb2 cDNA increased the activities of luciferase regulated by enhancer 434 and an HSV-TK promoter in Neuro-2A cells (*P < 0.001). (F–H) Immunohistochemistry for β-gal (green) and CUX1 (red) in P4 Emx1-Cre; Satb2^+/+; enhancer 434-lacZ mice. (I–K) The number of β-gal⁺ neurons was reduced in P4 Emx1-Cre; Satb2^flox/flox; enhancer 434-lacZ mice. (F–H and I–K) Enlarged images of the boxed areas in Fig. S3 M and N, respectively. They were generated by merging three panels each. (L) Quantification of β-gal⁺ neurons in P4 control and Satb2 cKO cortices (*P = 0.013; n = 4 mice per genotype from three litters, unpaired t test). Error bars in C, E, and L indicate SD. (Scale bars: A and B, 500 μm; A′ and B′, 500 μm; F–K, 100 μm.)

Surprisingly, whereas some deep layer-specific genes showed increased expression, expression of other deep-layer genes was reduced in Satb2^lacZ/lacZ brains (Fig. S2A and Table S1). The reduced expression of these genes was confirmed by immunohistochemistry (Fig. S2 B–K), suggesting that SATB2 is required for the development of deep-layer neurons.

SATB2 Promotes the Expression of Fezf2, a Subcerebral Cell-Fate Determinant.

Among the genes that showed reduced expression in the Satb2^lacZ/lacZ cortices was Fezf2 (Fig. 2C and Table S1), a cell fate-determining gene of subcerebral neurons (5–8). In situ hybridization confirmed that Fezf2 mRNA was severely reduced at E16.5 (Fig. S3 A–F) and P0 (Fig. 2 A–B′ and Fig. S3 G–J) in the mutant mice. Fezf2 expression was largely absent from layer 5 across the cortex except for the most medial regions. Its expression was detected in layer 6 but reduced in caudolateral areas.

Fig. S3.

SATB2 regulates Fezf2 expression. (A–J) In situ hybridization showed that Fezf2 mRNA expression was reduced in the E16.5 (A–F) and P0 (G–J) Satb2^lacZ/lacZ cortices. (A–C, G, and H) Satb2^+/+ brains. (D–F, I, and J) Satb2^lacZ/lacZ brains. (K) Immunohistochemistry revealed that β-gal was expressed in cortical neurons in the P0 control Emx1-Cre; Satb2^+/+; enhancer 434-lacZ mice. (L) The number of β-gal⁺ neurons was reduced in P0 Emx1-Cre; Satb2^flox/flox; enhancer 434-lacZ mice. (M) Immunohistochemistry for β-gal in the P4 control Emx1-Cre; Satb2^+/+; enhancer 434-lacZ mice. (N) The number of β-gal⁺ neurons was reduced in P4 Emx1-Cre; Satb2^flox/flox; enhancer 434-lacZ mice. The enlarged images of the boxed areas in M and N are shown in Fig. 2 F–H, I, and K, respectively. (Scale bars: A–F, 500 μm; G–J, 500 μm; K and L, 100 μm; M and N, 500 μm.)

To determine whether SATB2 directly regulates Fezf2 expression, we performed ChIP and high-throughput DNA sequencing (ChIP-seq) by using SATB2 antibodies and dissected E15.5 cortices. We observed strong and specific binding of SATB2 to the highly conserved enhancer 434 at the 3′ end of the Fezf2 gene (Fig. 2D) (27, 28). To test whether SATB2 regulates the enhancer 434, we cloned enhancer 434 in the luciferase reporter plasmid pGL4CP-TK (29) and performed the luciferase assay in Neuro-2A cells. Cotransfection with Satb2 cDNA significantly decreased the activity of the luciferase in empty pGL4CP-TK plasmid (Fig. S4A). However, when enhancer 434 was cloned in pGL4CP-TK plasmid in either orientation, cotransfection with Satb2 cDNA significantly increased the luciferase activity (Fig. 2E and Fig. S4A).

Fig. S4.

Luciferase assays suggest that SATB2 likely directly regulates Fezf2 and Sox5 expression. (A) Cotransfection with pCAG-Satb2 reduced the luciferase activity of the pGL4CP-TK empty plasmid (control). However, cotransfection with pCAG-Satb2 increased the luciferase activity of pGL4CP-TK containing the enhancer 434 sequence in either orientation. (B) Cotransfection with pCAG-Satb2 led to reduction of the luciferase activity of the pGL4CP-TK empty plasmid (control). When the SATB2 binding sequence in the Sox5 intron was cloned into the pGL4CP-TK in either direction, the reduction of the luciferase activity was partially mitigated. The relative luciferase activity for each reporter plasmid in the presence of pCAG-Satb2 was normalized to the activity of the same reporter plasmid in the presence of pCAG plasmid. Error bars indicate SD.

We have generated stable transgenic mouse lines that expressed a lacZ gene under the control of enhancer 434 and an hsp68 minimal promoter (enhancer 434-lacZ mice) (15). To test whether SATB2 regulates the activity of enhancer 434 in vivo, we generated Emx1-Cre; Satb2^+/+; enhancer 434-lacZ (control) and Emx1-Cre; Satb2^flox/flox; enhancer 434-lacZ (Satb2 cKO; enhancer 434-lacZ) mice. In P0 (Fig. S3K) and P4 (Fig. 2 F and H and Fig. S3M) control brains, although the strongest LacZ activity was present in the ventricular zone, many enhancer 434-LacZ⁺ cells were detected in the cortical plate. The LacZ⁺ cells consisted of projection neurons and few astrocytes, and were located throughout all cortical layers (Fig. 2 F–H and Fig. S3K). In the Satb2 cKO; enhancer 434-lacZ mice, the number of LacZ⁺ cortical projection neurons was significantly reduced (Fig. 2 I–L and Fig. S3 L and N). These data indicate that SATB2 positively regulates the enhancer 434, and that SATB2 likely directly regulates Fezf2 expression in cortical neurons.

SATB2 Positively Regulates Sox5 Expression.

Mutations in multiple genes affect the development of subcerebral neurons and axons (27, 28, 30, 31). Sox5 is expressed at a high level in layer 6 corticothalamic neurons, and at a lower level in layer 5 subcerebral neurons (27, 30). However, its function in regulating the development of deep-layer cortical projection neurons has been controversial (27, 30). By using Sox5^lacZ and Sox5^flox alleles (27, 30), we confirmed that Sox5 is essential for the development and likely fate specification of subcerebral neurons. Subcerebral neurons, marked by high-level expression of CTIP2 (9), were missing in layer 5 and the lateral cortical regions in the Emx1-Cre; Sox5^flox/flox mice (Fig. S5). Accordingly, the corticospinal tract was present in the control Emx1-Cre; Sox5^lacZ/+; RCE-GFP mice, but severely reduced in the Emx1-Cre; Sox5^lacZ/flox; RCE-GFP brains (Fig. 3 A and B).

Fig. S5.

Subcerebral neurons, labeled by high-level CTIP2 expression, were missing in layer 5 and the lateral regions of the cerebral cortex in Sox5 cKO mice. CTIP2 staining in P4 brains is shown. (A, C, E, G, and I) Sox5^+/+ mice. (B, D, F, H, and J) Emx1-Cre; Sox5^flox/lacZ mice. (Scale bar: 200 μm.)

Fig. 3.

SATB2 regulates expression of Sox5, a gene required for the development of subcerebral axons. (A and B) Corticospinal tract axons are missing in the Sox5 mutant mice (B). Arrows in A point to the GFP⁺ corticospinal axons in Emx1-Cre; Sox5^+/lacZ; RCE-GFP mice, which are missing in Sox5 cKO (Emx1-Cre; Sox5^lacZ/flox; RCE-GFP) mice. (C–F) SOX5 staining (red) is reduced in layer 5 neurons in Satb2^lacZ/lacZ mice. CTIP2 staining is shown in green. (G) The normalized expression level of Sox5 mRNA in P0 Satb2^+/+ and Satb2^lacZ/lacZ cortices revealed by RNA-seq (*P < 0.001). (H) SATB2 binds to conserved intronic sequences of Sox5 gene (red line, Top). (Bottom) Close-up view of the boxed SATB2 binding peak. (I) Luciferase assay showed that, compared with the empty luciferase reporter plasmid, cotransfection of Satb2 cDNA increased the activities of luciferase in the reporter plasmids containing the SATB2 binding sequence in the Sox5 intron in either orientation (*P < 0.001). Error bars in G and I indicate SD. (Scale bars: A and B, 500 μm; C–F, 100 μm.)

RNA-seq showed that expression of Sox5 was reduced to 65% in the Satb2^lacZ/lacZ cortices (Fig. 3G and Table S1). Immunohistochemistry confirmed that, in the Satb2^lacZ/lacZ cortices, the number of SOX5⁺ neurons was reduced, and SOX5 was no longer expressed in layer 5 (Fig. 3 C–F). ChIP-seq analysis showed that SATB2 bound specifically to a highly conserved region in an intron of Sox5 gene (Fig. 3H). We cloned the SATB2 binding sequence into the pGL4CP-TK plasmid (29) and performed luciferase assays (Fig. 3I and Fig. S4B). Cotransfection with Satb2 cDNA reduced the luciferase activity in the pGL4CP-TK plasmid (Fig. S4B). When the SATB2 binding sequence was cloned into the pGL4CP-TK plasmid, luciferase activity was still reduced when Satb2 cDNA was cotransfected (Fig. S4B). However, compared with the empty pGL4CP-TK plasmid, contransfection of Satb2 cDNA increased the activities of the luciferase when the SATB2 binding sequence was cloned into the pGL4CP-TK plasmid in either direction (Fig. 3I). These results suggest that, in addition to regulating Fezf2, SATB2 may promote the development of subcerebral neurons through the regulation of Sox5 expression.

Mutual Regulations Between Fezf2 and Satb2 in Deep-Layer Neurons.

Although the aforementioned data reveal that SATB2 is required to promote Fezf2 expression in layer 5 neurons, we previously reported that Fezf2 represses high-level SATB2 expression: SATB2 expression was significantly increased in the deep layers of Fezf2^−/− cortices (8). These results implicate a mutual regulatory relationship between Satb2 and Fezf2. To further investigate this mutual regulation, we analyzed Satb2-LacZ and Fezf2- placental alkaline phosphatase (PLAP) expressions in Satb2 cKO mice (Fig. 4). As the LacZ and PLAP reporters were knocked into endogenous Satb2 and Fezf2 genes, respectively (6, 16), the expression of Satb2-LacZ and Fezf2-PLAP likely reflects that of Satb2 and Fezf2 loci. In the Satb2^lacZ/+; Fezf2^PLAP/+ brains, Fezf2-PLAP was expressed in layers 5 and 6. Consistent with Fezf2 inhibiting high-level Satb2 expression, the Satb2-LacZ⁺ domain consisted of a Satb2-LacZ^high domain in layers 2–4 and a Satb2-LacZ^low domain in deep layers (Fig. 4 A–C). The Satb2-LacZ expression was lowest in layer 5, where high-level Fezf2 expression was located. In the Emx1-Cre; Satb2^lacZ/flox; Fezf2^PLAP/+ mice, Fezf2-PLAP expression was absent in layer 5, and the Satb2-LacZ^high domain expanded from upper layers into layer 5, and stopped where Fezf2 expression persisted (Fig. 4 D–F). This result confirms that, although Satb2 is required for Fezf2 expression in layer 5 neurons, Fezf2 in turn represses Satb2 expression and reduces its expression level in deep-layer neurons.

Fig. 4.

Satb2 and Fezf2 mutually regulate the expression of each other. Immunohistochemistry using β–gal (red) and PLAP (green) antibodies of P7 brains. (A–C) A section of a Satb2^lacZ/+; Fezf2^PLAP/+ brain. (D–F) A section of an Emx1-Cre; Satb2^lacZ/flox; Fezf2^PLAP/+ brain. Dotted lines indicate the boundary between Satb2-LacZ^high and Satb2-LacZ^low domains. (Scale bar: 500 μm.)

Discussion

In this study, we have identified a mechanism by which Satb2 promotes the development of subcerebral and callosal neurons in a cell context-dependent manner (Fig. 5).

Fig. 5.

SATB2 regulates cortical projection neuron subtype identities in a cell context-dependent manner. (A–C) Summary of cortical projection neuron subtypes and projections in Satb2^+/+ mice. (A) Distribution of SATB2⁺callosal (blue circle), subcerebral (Fezf2-PLAP⁺; pink triangle), and corticothalamic (Golli-GFP⁺; green triangle) neurons in the Satb2^+/+ cortices. (B) Callosal neurons project axons through corpus callosum and into the targets in contralateral hemisphere. (C) Layer 6 corticothalamic neurons project axons into the thalamus, and layer 5 Fezf2-PLAP⁺ neurons project axons into cerebral peduncle, pyramidal tract, and thalamic nuclei. (D and E) Summary of cortical projection neuron subtypes and projections in the Satb2 cKO mice. (D) Fewer Satb2-LacZ⁺ callosal neurons were present in the mutant mice. Most upper-layer neurons show mixed identity and do not project axons into the corpus callosum and into the contralateral hemisphere. Subcerebral neurons were absent. Instead, Satb2-LacZ expression is increased in layer 5 neurons. There is a mild decrease of layer 6 Golli-GFP⁺ corticothalamic neurons. (E) Fewer Satb2-LacZ⁺ neurons extend axons through the corpus callosum and into the contralateral hemisphere. (F) Golli-GFP⁺ axons from layer 6 neurons project into the thalamus, and Fezf2-PLAP⁺ axons fail to project into the cerebral peduncle, pyramidal tract, or the thalamic nuclei innervated by layer 5 neurons. Instead, Satb2-LacZ⁺ axons innervated the thalamic nuclei that normally receive inputs from layer 5 neurons. Satb2-LacZ⁺ axons do not project into the pyramidal tract. (G) SATB2 regulates cortical projection neuron fates in a cell context-dependent manner.

SATB2 Promotes Callosal Neuron Identity by Repressing Expression of Subcerebral Neuronal Genes and Promoting Expression of Upper-Layer Neuronal Genes.

Satb2 was reported to promote the development of callosal neurons through repressing expression of subcerebral neuron genes such as CTIP2 (17–19, 24). Our study confirmed this, and identified many more deep-layer genes that were up-regulated in the Satb2 mutant mice (Fig. S2A and Table S1). Further, we found that expression of upper-layer neuronal genes was reduced in the Satb2^lacZ/lacZ cortices (Table S2), suggesting that Satb2 is required to promote their expression. Thus, our study demonstrates that Satb2 promotes callosal neuron development not only through repressing expression of deep-layer genes in upper cortical layers, but also by promoting expression of upper-layer neuronal genes (Fig. 5G).

Corticothalamic Neurons and Axons Are Present in the Satb2 Mutant Mice.

The expression of Satb2 in the corticothalamic neurons (Fig. 1 E–H and J) suggests that it may regulate the development of these neurons. Indeed, a recent study reported increased Satb2-LacZ⁺ axons in the thalamus of the Satb2 cKO mice (25). We used multiple methods to investigate the corticothalamic axons, and discovered that the innervation pattern of the thalamus by corticothalamic axons was mostly normal in the Satb2 mutant brains (Fig. 5F, SI Results, and Fig. S6). The increased Satb2-LacZ⁺ axons observed in the Satb2 cKO mice (Fig. S6 A and B) (25) were a result of increased expression of Satb2-LacZ in the layer 5 and upper layer 6 neurons that normally express low-level SATB2 and project axons into the thalamus (Figs. 1 E–H and 5 A–F and Fig. S6). Thus, SATB2 appears to have a limited role in the initial fate specification and axon projection of corticothalamic neurons.

Fig. S6.

Innervation pattern of the thalamus by corticothalamic axons appears not to be significantly changed in the Satb2 mutant mice. (A and B) LacZ staining revealed more Satb2-LacZ⁺ axons were present in the thalamus of P4 Emx1-Cre; Satb2^lacZ/flox mice (Satb2 cKO, B) than in the Satb2^lacZ/+ mice (A). (C and D) PLAP staining revealed that fewer Fezf2-PLAP⁺ axons were present in the thalamus of P4 Emx1-Cre; Satb2^lacZ/flox; Fezf2^PALP/+ mice (Satb2 cKO, D) than in the Satb2^lacZ/+ Fezf2^PALP/+ mice (C). (E–P) Complementary changes in the innervation patterns of Golli-GFP and Satb2-LacZ–labeled corticothalamic axons in the P28 Satb2 cKO mice. Golli-GFP labeling is in green, and Satb2-LacZ staining is in red. (E–G and K–M) Sabt2^lacZ/+; Golli-GFP mice. (H–J and N–P) Emx1-Cre; Satb2^lacZ/flox; Golli-GFP mice. Note the reduced corticothalamic innervation by the Golli-GFP⁺ axons in the Satb2 cKO mice (compare arrows in H to E and in N to K), and the increased innervation by the Satb2-LacZ⁺ axons (compare arrows in I to F and in O to L). The arrows in E–J point to the laterodorsal (LD) thalamic nuclei, and the arrows in K–P point to the posterior (Po) thalamic nuclei. Both are targets of layer 5 corticothalamic axons in WT mice. (Scale bars: 500 μm.)

Mutual Regulation Between Satb2 and Fezf2 Enables Satb2 to Promote the Development of Subcerebral Neurons.

RNA-seq analysis indicated that SATB2 activates the expression of deep-layer neuronal genes that are essential for specifying subcerebral neuronal fate (e.g., Fezf2 and Sox5) and represses the expression of other deep-layer neuronal genes that are necessary for the development of subcerebral axons (e.g., Ctip2; Fig. S2A and Table S1). Then how does Satb2 promote the development of subcerebral neurons?

Although SATB2 expression has been generally associated with callosal neurons (8, 17, 18), our retrograde tracing experiment showed that most corticofugal neurons express low- or medium-level SATB2 at birth (Fig. 1). The dynamic expression of SATB2 in corticofugal neurons is further supported by a recent gene-expression study of different cortical projection neuron subtypes (32). Although the authors used high SATB2 expression as a criteria to sort callosal neurons, SATB2 was not among the top 25 most specific markers to distinguish callosal neurons from corticofugal neurons. During cortical neurogenesis, SATB2 is expressed in the newly generated projection neurons in the intermediate zone (33). This expression of SATB2 ensures the expression of Fezf2 and Sox5 in newly generated subcerebral neurons. Indeed, reduced Fezf2 mRNA was already evident in the E16.5 Satb2^lacZ/lacZ mice (Fig. S3). However, SATB2 also represses the expression of Ctip2 (16–18), which is necessary for the development of subcerebral axons (10). How do the developing subcerebral neurons overcome the inhibition of Ctip2 by SATB2? The repression of high-level SATB2 by Fezf2 is likely the key. Once Fezf2 is expressed in subcerebral neurons, it inhibits high-level SATB2 expression (8), so that high-level CTIP2 is maintained in these cells. Indeed, although most newly generated CTIP2⁺ deep-layer neurons express SATB2 at E13.5 (18), by E15.5, high-SATB2⁺ and high-CTIP2⁺ neurons start to segregate into distinct populations (18, 33).

We propose that Satb2 promotes the identities of subcerebral and callosal neurons in a cell context-dependent manner (Fig. 5G). In the presumptive subcerebral neurons, SATB2 activates the expression of Fezf2, Sox5, and other subcerebral neuronal genes. The high-level Fezf2 in subcerebral neurons in turn represses Satb2 expression. Reduced SATB2 ensures the expression levels of subcerebral neuronal genes such as Ctip2 remain high in these cells, which further promotes a subcerebral neuron identity. In addition, reduced SATB2 in subcerebral neurons prevents expression of callosal neuron genes that are positively regulated by SATB2. In the developing callosal neurons, SATB2 does not activate Fezf2, and SATB2 level remains high. High SATB2 promotes the expression of upper-layer genes that are required for callosal neuron development and prevents expression of deep-layer neuron genes such as CTIP2, which may lead to corticofugal axon development (8).

Although this model can explain most of the molecular and axonal defects of the Satb2 mutant mice, some important questions await further investigation. For example, what genes turn on Satb2 expression in the newly generated projection neurons? Artegiani et al. recently showed that HMG group transcription factor Tox binds to Satb2 gene and thus potentially regulates its expression (34). Dominguez et al. showed that electroporation of Brn1/2 at E12.5 led to precocious expression of Satb2 in the cortical neurons (35). However, whether these genes directly activate Satb2 transcription awaits further investigation. In addition, if SATB2 is expressed in presumptive subcerebral neurons and callosal neurons, why does it not activate Fezf2 expression in callosal or upper-layer neurons? It is possible that other genes that are required to activate Fezf2 expression are present only in corticofugal neurons, or that there are repressors in the callosal neurons that inhibits Fezf2 expression. We emphasize that, although we have identified Fezf2 and Sox5 as likely direct targets of SATB2 in promoting subcerebral neuron differentiation, additional downstream targets of SATB2 may also be essential for the development of subcerebral axons. Further delineation of the functions of these genes may reveal novel regulatory mechanisms for subcerebral neuron fate specification.

SI Results

Corticothalamic Axons Are Present in the Satb2 cKO Mice.

The expression of SATB2 in corticothalamic neurons (Fig. 1 E–H and J) suggests that it may regulate the development of corticothalamic neurons and axons. To examine the corticofugal axons in Satb2 cKO mice, we used two genetic markers to specifically label these axons. The Fezf2^PLAP allele contains a human PLAP cDNA that was inserted in the Fezf2 gene, and labels corticothalamic and subcerebral axons with PLAP (6). The Golli-GFP transgene specifically labels corticothalamic and some subcerebral axons (37).

Compared with control Satb2^lacZ/+; Fezf2^PLAP/+ mice, we found that more Satb2-LacZ–labeled axons were present in the thalamus of the P4 Emx1-Cre; Satb2^lacZ/flox; Fezf2^PLAP/+ (Satb2 cKO) brains (Fig. S6 A and B). This result is consistent with a recent study (25). Intriguingly, fewer Fezf2-PLAP labeled corticothalamic axons were present in the P4 Emx1-Cre; Satb2^lacZ/flox; Fezf2^PLAP/+ brains (Fig. S6 C and D). Similarly, reduction of Golli-GFP⁺ corticothalamic axons was observed in P4 Emx1-Cre; Satb2^lacZ/flox; Golli-GFP brains.

To further investigate the corticothalamic axons, we compared the GFP⁺ axons in Emx1-Cre; Satb2^+/+; RCE-GFP (control) and Emx1-Cre; Satb2^lacZ/flox; RCE-GFP (Satb2 cKO) brains. Although GFP⁺ corticospinal axons were absent in the Satb2 cKO mice, a careful comparison of GFP⁺ axons between Emx1-Cre; Satb2^+/+; RCE-GFP and Emx1-Cre; Satb2^lacZ/flox; RCE-GFP brains at P24 did not show a significant difference in the corticothalamic innervation patterns.

Puzzled by this, we investigated Golli-GFP (Fig. S6 E–P) and Fezf2-PLAP (Fig. 4) expression in the cortex, and discovered that these genetic markers were no longer expressed in layer 5 and upper layer 6 neurons in the Satb2 mutant mice. In the mean time, we observed complementary expanded expression domain of the Satb2-LacZ reporter in the deep layer neurons (Fig. 4 and Fig. S6). A comparison of Golli-GFP⁺ and Satb2-LacZ⁺ axons between P28 Satb2^lacZ/+; Golli-GFP and Emx1-Cre; Satb2^lacZ/flox; Golli-GFP mice showed that, although some thalamic nuclei received less innervation from Golli-GFP⁺ axons, more Satb2-LacZ⁺ axons were observed in these nuclei, so that the total innervation pattern by the Golli-GFP⁺ and Satb2-LacZ⁺ axons was not significantly affected in the Satb2 cKO mice (Fig. S6 E–P). The reduced innervation by Golli-GFP⁺ axons was most apparent in the thalamic nuclei targeted by layer 5 corticothalamic axons (Fig. S6 E–P) (46). More Satb2-LacZ⁺ corticothalamic axons were observed in these same nuclei (Fig. S6 E–P).

Thus, the corticothalamic axonal defects observed in the Satb2 mutant brains using Golli-GFP, Fezf2-PLAP and Satb2-LacZ axonal reporters were at least partially attributable to the changed reporter expressions. Overall, our axonal analysis suggested that Satb2 has a more limited role in regulating the development of corticothalamic axons.

Methods

All experiments on mice were carried out in accordance with the protocols approved by the institutional animal care and use committee at the University of California, Santa Cruz (UCSC), and institutional and federal guidelines. The day of vaginal plug detection was designated as E0.5. The day of birth was designated as P0.

The Satb2^lacZ/+, Fezf2^PLAP/+, enhancer 434-lacZ, Emx1-Cre, Golli-GFP, and RCE-GFPCre reporter mice were described previously (6, 15, 16, 36–38). The Satb2^{tm1a(KOMP)Wtsi/+} mice were obtained from the Knockout Mouse Project (KOMP) repository and bred with Rosa26Sor^tm2(FLP*^)Sor mice from the Jackson Laboratory (JAX number 007844) to generate the Satb2^flox(KOMP)/+ mice (referred as Satb2^flox/+ mice in this study).

Details of retrograde tracing, immunohistochemistry (9, 39), cell counting, and luciferase assays are described in SI Methods.

Gene Expression Analysis.

RNA-seq was performed from P0 cortices of Satb2^+/+ and Satb2^lacZ/lacZ mice (n = 3 for each genotype). The sequencing libraries were prepared following the TruSeq RNA sample preparation protocol (Illumina), and pair-end (50 bp per end) sequenced on an Illumina Genome Analyzer II platform. Approximately 40 million reads were obtained for each library. The reads were first mapped with Bowtie (40) against the set of RepeatMasker (www.repeatmasker.org, 1996–2010) elements in mouse to remove repetitive elements from further analysis. The remaining reads were mapped with TopHat (41) against the mouse (NCBI37/mm9) assembly. Only uniquely mapped and properly paired reads were kept. Potential PCR duplicates were removed with the SAMtools (42) “rmdup” command. The net count of remaining mappings for the six samples ranged from 27.8 million to 33.0 million. From these TopHat mappings, coverage of each gene in the set of UCSC Known Genes (43) for the mm9 assembly was determined. The read counts for each gene in each sample were used as input to DESEq. (44) for differential expression analysis. The resulting DESeq “sizeFactors” were used to normalize the read counts of individual replicates. Differentially expressed genes were defined as the ones that were decreased by 20% or increased by 20% in the Satb2^lacZ/lacZ cortices, compared with the controls with DEseq, with a false discovery rate < 0.1.

ChIP-Seq Analysis.

ChIP was performed on dissected E15.5 WT mouse cortices as previously described (9) by using an SATB2 antibody (ab34735; Abcam). The input DNA and SATB2 ChIPed DNA were pair-end (100 bp per end) sequenced on an Illumina Genome Analyzer II platform. Approximately 35 million and 33 million reads were obtained for the SATB2 ChIPed DNA and input DNA, respectively. The raw 100 × 100 bp paired-end reads were trimmed to 75 × 75 bp by discarding 5 bp from the 5′ end and 20 bp from the 3′ end of read1 and read2. The trimmed reads were first mapped with Bowtie (40) against the set of RepeatMasker elements to remove repetitive elements from further analysis. The remaining reads were mapped with Bowtie (40) to the mouse (NCBI37/mm9) assembly. Only uniquely mapped and properly paired reads were kept. Fewer than 2% of the mappings were found to be potential PCR duplicates, so they were not removed. The net count of remaining mappings was approximately 21 million for both the SATB2 ChIPed DNA and input DNA. Peak calling was done with the MACS 1.4 tool (45). For input to MACS, only read1 of the mapped reads was used, and the MACS “shiftsize” parameter was determined from the genomic length of the paired-end Bowtie mappings of the ChIPed sample.

SI Methods

Retrograde Tracing.

The fluorescent tracer CTβ conjugated with Alexa 488 or Alexa 594 was injected into the thalamus or pyramidal decussation by using Vevo770 ultrasound-guided injection systems on E17.5 or P2. The brains were collected on P2 or P8 and processed for immunohistochemistry.

Histochemistry.

Immunohistochemistry and PLAP staining for mouse tissues were performed as previously described (9, 39). LacZ staining was performed as previously described (15).

Primary antibodies used in this study were as follows: chicken anti–β-gal (ab9361; Abcam), 1:1,000; rat anti-CTIP2 (ab18465; Abcam), 1:200; rabbit anti-CUX1 (sc-13024; Santa Cruz Biotechnology), 1:500; rabbit anti-DARPP32 (ab40801; Abcam), 1:200; goat anti-BHLHB5 (sc-6045; Santa Cruz Biotechnology), 1:100; rabbit anti-FOG2 (ZFMP2; sc-10755; Santa Cruz Biotechnology), FOXP2 (ab16046; Abcam), 1:1,000; rabbit anti-GFP (A11122; Invitrogen), 1:1,000; rabbit anti-NFIA (39397; Active Motif), 1:500; rabbit anti-PLAP (AHP537; Accurate Chemical), 1:100; rabbit anti-SATB2 (ab34735; Abcam), 1:1,000; mouse anti-SATB2 (sc-81376; Santa Cruz Biotechnology), 1:1,000; and rabbit anti-TBR1 (ab31940; Abcam), 1:500.

For counting of LacZ⁺ cells in the Emx1-Cre; Satb2^+/+; enhancer 434-lacZ (control) and Emx1-Cre; Satb2^flox/flox; enhancer-434 (Satb2 cKO) cortices (n = 4 for each genotype, collected from three litters), three sections stained with X-Gal were counted from each animal. The quantification for the control genotype was 59 ± 8.2 LacZ⁺ neurons per square millimeter (mean ± SD), and for the Satb2 cKO was 23 ± 9.0 LacZ⁺ neurons per square millimeter. To determine statistical significance, GraphPad was used to preform an unpaired t test between experimental groups.

Microscopy.

Bright-field and epifluorescence images were taken on an Olympus BX51 microscope by using a Q Imaging Retiga EXj camera or a Keyence BZ-9000 microscope. Confocal images were obtained on a Leica TCS SP5 or a Zeiss LSM5 confocal microscope. Images were processed by using Adobe Photoshop CS5 to adjust brightness and contrast.

Luciferase Assay.

Enhancer 434 was amplified from mouse genomic DNA using primers 5′-ATCGCTCGAGCAGGCTGTAGGATGGGCAGCAGGAGTTTC-3′ and 5′-ATCGAAGCTTGTAACAAGTCAGGTGAGCAGGCGGTA-3′. The amplified DNA was inserted into the SalI site of the luciferase reporter plasmid pGL4CP-TK (29) in both orientations. Thus, the enhancer 434 was inserted downstream of the polyadenylation site of the luciferase gene.

The SATB2 binding site in the intron of Sox5 was amplified by using primers 5′-ATCCACAATGGTCCGCGGGTAACCG-3′ and 5′-TGTAGGAGGTGCCACAGTGCTGGG-3′. The amplified sequence was inserted into the BsrG1 sites of the pGL4CP-TK plasmid in both orientations, upstream of the HSV thymidine kinase gene promoter, which drives luciferase gene expression.

Neuro-2A cells (American Type Culture Collection) were transfected with pCAG-Satb2 (gift from Nenad Sestan, Yale University, New Haven, CT) or empty pCAG vector, Renilla luciferase (pRL-SV40) plasmids (internal control), and empty pGL4CP-TK firefly luciferase reporter plasmids, or pGL4CP-TK reporter plasmids containing enhancer 434 or SATB2 binding site in the Sox5 gene, using Lipofectamine 2000 (Invitrogen). Luciferase activities were assayed 48 h later using the Dual-Luciferase System (Promega). Three technical replicates were measured for each biological replicate. The ratio of firefly luciferase and Renilla luciferase for each replicate was calculated. In Figs. 2E and 3I, relative luciferase activities for enhancer 434 (Fig. 2E) and SATB2 binding site in Sox5 (Fig. 3I) in the presence of Satb2 cDNA are shown relative to the activity in the empty pGL4CP-TK plasmid measured in the presence of Satb2 cDNA. Error bars represent SD. All experiments were repeated three times.