SP8 and SP9 coordinately promote D2-type medium spiny neuron production by activating Six3 expression

The striatal subdivisions – the caudate nucleus, putamen (caudate-putamen), nucleus accumbens and olfactory tubercle – are key components of distributed circuits integral to cognitive, motor and reward processes. Their dysfunction contributes to brain disorders, such as Huntington's and Parkinson's disease, and addiction (Kreitzer and Malenka, 2008; Maia and Frank, 2011). There are two major cell types of medium spiny neurons (MSNs), the principal projection neurons of the striatum: dopamine receptor DRD1-expressing MSNs (D1 MSNs) and dopamine receptor DRD2-expressing MSNs (D2 MSNs). Although these GABAergic neurons share many properties, they have some distinct molecular properties and differentially contribute to distinct circuits (e.g. direct and indirect pathway) (Lobo et al., 2006; Heiman et al., 2008; Gerfen and Surmeier, 2011; Ena et al., 2013; Gokce et al., 2016; Merchan-Sala et al., 2017).

Transcriptional control of MSN development occurs in the lateral ganglionic eminence (LGE), and involves a cascade of transcription factors (TFs), including GSX1/2, DLX1/2/5/6, ASCL1, EBF1 and ISL1 (Rubenstein and Campbell, 2013). Previously, we reported that the zinc finger TF gene Sp9 is expressed in subventricular zone (SVZ) progenitors of both D1 and D2 MSNs, and its expression persists in the postmitotic D2 MSNs but not in D1 MSNs (Zhang et al., 2016). We demonstrated that Sp9 is required for the generation, differentiation and survival of D2 MSNs, as ∼95% of them are lost in the caudate-putamen of Sp9 null mutants (Zhang et al., 2016). We also showed that SP9 promotes the expression of Drd2, Adora2a, P2ry1, Gpr6 and Grik3 genes, which are specific molecular markers for D2 MSNs (Zhang et al., 2016). Currently, SP9 is the only TF that has been shown to control the expression of D2 MSN terminal differentiation genes.

The LGE progenitor zone is proposed to be organized into four domains along the dorsoventral axis (pLGE1, 2, 3 and 4) (Flames et al., 2007). Whereas Sp9 is broadly expressed in all pLGE domains in the SVZ, its paralog, Sp8, is most prominently expressed dorsally (dorsal LGE, dLGE), in the SVZ of pLGE1 and pLGE2. Sp8 and Sp9 are co-expressed prenatally and postnatally in the SVZ, rostral migratory stream (RMS) and olfactory bulb (OB) interneurons (Long et al., 2007; Li et al., 2017). Together, Sp8 and Sp9 regulate OB interneuron differentiation, tangential and radial migration, and survival by promoting Prokr2 and Tshz1 expression (Li et al., 2017). In the absence of Sp8 and Sp9, immature OB interneurons fail to express Prokr2 and Tshz1, genes that are required for OB interneuron development and are involved in human Kallmann syndrome (Li et al., 2017).

The ventral LGE (vLGE, i.e. pLGE3 and pLGE4) domains generate MSNs (Toresson and Campbell, 2001; Stenman et al., 2003; Silberberg et al., 2016). Sp9 is broadly expressed in the SVZ of these domains, whereas Sp8 appears to be weakly expressed in the SVZ of pLGE3 (Waclaw et al., 2006; Ma et al., 2012). The function and targets of SP8 in the vLGE are unknown.

Here, we analyzed the generation of D1 and D2 MSNs in the caudate-putamen, nucleus accumbens and olfactory tubercle in Sp8, Sp9 and Sp8/Sp9 mutant mice [Dlx5/6-CIE; Sp8^F/F (hereafter referred to as Sp8-CKO), Sp9^lacZ/lacZ null mutant (Sp9-KO) and Dlx5/6-CIE; Sp8^F/F; Sp9^F/F double conditional knockout (Sp8/9-DCKO)]. Sp8/9-DCKO mice, compared with Sp8-CKO and Sp9-KO, have a more severe reduction of D2 MSNs. This is mainly due to severely reduced LGE neurogenesis. Crucially, Sp8 and Sp9 are required for expression of the TF Six3 in the SVZ of pLGE3. Six3 belongs to the Six/sine oculis family of homeobox TFs; it promotes cell proliferation and differentiation (Oliver et al., 1995; Kobayashi et al., 1998; Loosli et al., 1999; Del Bene et al., 2004; Del Bene and Wittbrodt, 2005; Appolloni et al., 2008). We show that Six3-CKO mice (Dlx5/6-CIE; Six3^F/F) lose the majority of D2 MSNs, phenocopying Sp8/9-DCKO mice. Finally, chromatin co-immunoprecipitation combined with followed by high-throughput DNA sequencing (ChIP-Seq) reveals that SP9 directly binds to the promoter and a putative enhancer of Six3, providing evidence that SP9 regulates D2 MSN development by direct regulation of Six3 expression. Thus, we provide evidence for components of a transcription pathway, in a regionally restricted LGE domain, that selectively drives the generation of D2 MSNs.

SP8 protein is strongly expressed in the dLGE SVZ and weakly expressed in the vLGE SVZ (Fig. 1A-H) (Waclaw et al., 2006; Ma et al., 2012). Most SP8⁺ cells (>91%) in the LGE SVZ expressed SP9 at embryonic day (E) 14.5 and E16.5 (Fig. 1D,H), whereas only a few SP8⁺ cells expressed ASCL1, a marker for neural progenitors (Fig. S1A-D). SP8 expression in MSNs is at background levels (Fig. 2A,B). In contrast, SP9 is strongly expressed in the SVZ of both the dLGE and the vLGE, and its expression is maintained in the D2 MSNs but not in D1 MSNs (Fig. 1A-H, Figs S1E-H, S2A) (Zhang et al., 2016).

In the Sp9-KO embryos, SP8 expression is upregulated in the SVZ of both the dLGE and vLGE at E14.5 (Fig. S2B,C) and E16.5 (Fig. 1I,J, arrows). In the postnatal striatum of Drd2-GFP transgenic mice, only a small number of DRD2-GFP⁺ cells expressed SP8 (Fig. 2A,B,E-H). Although D2 MSNs were severely reduced in the striatum of Sp9-KO; Drd2-GFP mice, the number of SP8⁺ and DRD2-GFP⁺/SP8⁺ neurons showed a ∼4-fold increase (Fig. 2C,D,I-N). The DRD2-GFP⁺ cells in the caudate-putamen of Sp9-KO mice that do not express SP8 are choline acetyltransferase (ChAT)⁺ interneurons (Zhang et al., 2016). The fact that the majority of the D2 MSNs that remained in the Sp9-KO striatum upregulated SP8 expression suggests that SP8 could be compensating for SP9 function. Thus, given their sequence homology, and LGE expression, Sp8 and Sp9 may share functions in regulating MSN development. To test this hypothesis, we compared MSN development in control (Dlx5/6-CIE), Sp8-CKO, Sp9-KO and Sp8/9-DCKO mice.

We first examined the D2 MSNs in the control and mutant striatum at postnatal day (P) 11, as most Sp8/9-DCKO mice die at P12. Drd2, Adora2a, Penk and Gpr6 are expressed by D2 MSNs (Fig. 3A-J, Fig. S3A-L). In situ RNA hybridization of Drd2 and Adora2a revealed that only ∼5% of D2 MSNs were lost in the Sp8-CKO striatum (Fig. 3A-J). In the Sp9-KO striatum, in contrast, ∼85% D2 MSNs were lost; the remaining D2 MSNs were mainly located in the ventral and lateral striatum (Fig. 3C,H) (Zhang et al., 2016). In Sp8/9-DCKO mice, we did not observe Adora2a⁺ or Gpr6⁺ cells in the striatum, suggesting a lack of D2 MSNs (Fig. 3I, Fig. S3L). Some neurons in the Sp8/9-DCKO striatum retained Drd2 expression but they were ChAT⁺ interneurons (Fig. S4A-H) (Durieux et al., 2009; Zhang et al., 2016). At P0, Sp8/9-DCKO mice was also lacked molecular markers of D2 MSNs (Fig. S5A-P).

We next examined the D1 MSNs in the mutant striatum. In situ hybridization of Drd1 and Tac1 (tachykinin precursor 1) revealed that, relative to controls, striatal D1 MSNs were largely unaffected in Sp8-CKO, Sp9-KO and Sp8/9-DCKO mice at P11 (Fig. 3K-T). Similar results were also observed at P0 (Fig. S5Q-X). Taken together, our results support the hypothesis that Sp8 and Sp9 have redundant and essential functions in generating D2 MSNs, whereas they only have minor roles in the development of D1 MSNs.

To define the transcriptional changes in the LGE of Sp8/9-DCKO in order to elucidate the mechanisms underlying the preferential loss of D2 MSNs, we compared control and Sp8/9-DCKO transcriptomes by RNA sequencing (RNA-Seq). Gene expression profiles from the E16.5 LGE [including the VZ (ventricular zone), SVZ and MZ (mantle zone)] of Sp8/9-DCKO mice and littermate controls (Sp8/9 floxed mice without Dlx5/6-CIE allele) (n=3 biological replicas) were analyzed. Gene expression levels are reported in fragments per kilobase of exon per million fragments mapped (FPKM) (Trapnell et al., 2012; Li et al., 2017) (Table S1). For example, the expression levels (FPKM) of Sp8 and Sp9 genes in the control mouse LGE at E16.5 were 7.7 and 137.8, respectively, whereas their expression levels in the Sp8/9-DCKO LGE were reduced to 0.7 and 1.4 (Table S1). Thus, as expected, the mutation greatly reduced Sp8 and Sp9 LGE expression. About 80 genes were downregulated and 120 genes were upregulated (P<0.05, Student's t-test).

Next, we analyzed the expression of D2 MSN-enriched genes, and found that Adora2a, Drd2, Gpr6, Penk and Ptprm RNA was near background levels in the mutants (Table S1), supporting the evidence that D2 MSNs were not generated in the Sp8/9-DCKO mice. In contrast, the expression of D1 MSN-enriched genes, such as Drd1, Pdyn, Isl1, Ebf1 and Sox8, were largely unaffected (Table S1).

RNA levels of pan-neuronal genes (Gad1, Dcx, Bcl11a, Bcl11b, Foxp1, Meis2, Rxrg and Tubb3) were also reduced (Table S1), further supporting the suggestion that striatal MSN neurogenesis is comprised in Sp8/9-DCKO mice. In contrast, expression levels of genes that are highly enriched in LGE neural stem/progenitor cells, such as Gsx1, Gsx2, Ascl1 (also known as Mash1), Dbi, Dlx2, Fabp7 (also known as Blbp), Lhx2 and Vim, were increased (Table S1). The increased expression of progenitor genes and reduced expression of pan-neuronal and D2 MSN genes suggest that lack of Sp8/9 impedes the ability of neural progenitors to generate D2 MSNs.

In Sp8-CKO embryos, in situ RNA hybridization showed that in the E14.5 and E16.5 LGE (VZ-SVZ) Gsx2, Dlx1, Dlx2 and Ascl1 expression were subtly increased (Fig. 4A-D, Fig. S6A-D). ASCL1 has roles in both cell cycle promotion and cell cycle termination through direct activation of positive and negative cell cycle regulators (Castro et al., 2011). Consistent with this, we observed upregulation of Ascl1 and its target genes (E2f1, Cdca7, Cdk1 and Hes5) that promote cell cycle and maintenance of the progenitor state, and its target genes (Gadd45g and Btg2) that inhibit cell cycle, using in situ hybridization on E14.5 Sp8-CKO embryos (Fig. S6E-J) (Castro et al., 2011). Thus, upregulation of Ascl1 in the Sp8-CKO mice may contribute to the blockage of neuronal differentiation in the LGE SVZ. Importantly, misregulation of this Ascl1 pathway in neural progenitors was not apparent in Sp9-KO mice, but was more even severe in Sp8/9-DCKO than the Sp8-CKO (Fig. 4A-D, Fig. S6A-J, Table S1). Indeed, quantification of GSX2⁺ and ASCL1⁺ cells in the E16.5 LGE VZ and SVZ further demonstrated that there were more neural progenitors in Sp8-CKO and Sp8/9-DCKO than control and Sp9-KO mice (Fig. 4E-P). Together, these results provide evidence that Sp8 and Sp9 have different and redundant roles in the generation of D2 MSNs.

We next examined neurogenesis in the caudate-putamen of wild-type control, Sp8-CKO, Sp9-KO and Sp8/9-DCKO mice. The SVZ of the mouse LGE is larger than the SVZ of the cortex (Smart, 1976; Sheth and Bhide, 1997). At early developmental stages, the LGE SVZ has a high density of cycling progenitors, whereas the late SVZ contains relatively more differentiated cells, such as postmitotic FOXP1⁺ and EBF1⁺ immature neurons at E16.5 (Fig. 5A,F). Sp8/9-DCKO lose most of these FOXP1⁺ and EBF1⁺ cells; the single mutants have more subtle reductions (Fig. 5A-J). Expression of DCX and TUBB3 (general markers of immature neurons) was also reduced in the SVZ of Sp8/9-DCKO (Fig. 5K-N), supporting the suggestion of a reduction of neurogenesis in the LGE at E16.5.

We determined whether cell cycle exit was affected by Sp8 and Sp9 mutations by examining the state of differentiation of cells following a pulse-chase of 5-bromo-2′-deoxyuridine (BrdU) (Fig. 5O-T). In control and Sp8-CKO mice at E16.5 (24 h after BrdU injection), most BrdU⁺ cells were still in the VZ/SVZ, and very few BrdU⁺ cells were found in the striatum (Fig. 5P,Q). By contrast, in Sp9-CKO and Sp8/9-DCKO mice, we observed an increase in the number of BrdU⁺ cells in the striatum (Fig. 5P-T), suggesting a premature differentiation of SVZ progenitors at this age. In contrast, Sp8-CKO and Sp8/9-DCKO mice had increased numbers of BrdU⁺ cells in the LGE VZ/SVZ (Fig. 5O). This is consistent with the above observations that more progenitors accumulated in the VZ/SVZ of these mice (Fig. 4, Fig. S6).

To examine further the reduction in striatal neurogenesis, we performed a BrdU birthdating analysis. BrdU was given at E13.5, a time point in the middle of the striatal neurogenic period. We then counted the number of BrdU⁺ cells in the caudate-putamen in P0 tissue sections. Sp8-CKO, Sp9-KO and Sp8/9-DCKO mutants each had reduced BrdU⁺ cells with Sp8/9-DCKO showing the greatest reduction (Fig. 5U-Y). Thus, Sp8 and Sp9 both promote MSN production.

We recently observed that loss of Sp9 function induced Bax-dependent apoptosis, as cleaved caspase 3 cells were not observed in the postnatal caudate-putamen of Sp9-KO; Bax^−/− double mutant mice (Fig. S7Q) (Zhang et al., 2016). We next examined apoptotic cell death in the Sp8-CKO, Sp9-KO and Sp8/9-DCKO caudate-putamen. Compared with control (Dlx5/6-CIE) and Sp8-CKO mice, there were more caspase 3⁺ cells in the caudate-putamen of Sp9-KO (∼3- to 6-fold increase) and Sp8/9-DCKO (∼3 fold increase) mice at P2 and P3 (Fig. S7A-T). Notably, the number of caspase 3⁺ cells in the Sp9-KO caudate-putamen was larger than that in Sp8-CKO and Sp8/9-DCKOs at P2 and P3 (Fig. S7R-T). This could be because Sp9-KO mice generated more mutant striatal cells relative to Sp8/9-DCKO mice (based on our BrdU birthdating analysis, Fig. 5U-Y) and a large number of these mutant cells die without Sp9 function (Zhang et al., 2016). Therefore, we conclude that reduction of striatal MSNs in Sp8/9-DCKO mice results from a combination of reduced neurogenesis and apoptosis.

Accumulation of neural progenitors and upregulation of cell cycle-regulated genes were observed in the LGE of the Sp8-CKO and Sp8/9-DCKO, but these were not apparent in Sp9-KO mutants (Fig. 4, Fig. S6). However, Sp9-KO mice generated many fewer MSNs (particularly D2 MSNs) than did Sp8-CKO mice (Figs 3A-J, 5U-Y, Fig. S3A-L). This implies that there are additional mechanisms that regulate the generation of D2 MSNs in Sp9-KO mice, one of which was suggested by the LGE RNA-Seq analysis. We identified that expression of the TF Six3 was decreased ∼5-fold in E16.5 Sp8/9-DCKO mice compared with control mice (control versus Sp8/9-DCKO: FPKM, 64.2 versus 12.7; P=0.000000000857, FDR=0.00000113) (Table S1). Six3 promotes proliferation and differentiation through both transcription-dependent and -independent mechanisms (Oliver et al., 1995; Kobayashi et al., 1998; Loosli et al., 1999; Del Bene et al., 2004; Del Bene and Wittbrodt, 2005; Appolloni et al., 2008). Furthermore, single-cell RNA-Seq has identified that Six3 is preferentially expressed in D2 MSNs (Gokce et al., 2016). Therefore, we hypothesized that Six3 could be a key LGE target of SP8 and SP9 in driving D2 MSN development.

Next, we analyzed Six3 expression in the LGE of wild-type mice. In the E12.5-E14.5 LGE, Six3 is weakly expressed in the VZ and strongly expressed in the SVZ (preferentially in pLGE3, the dorsal part of the vLGE), but not in the dLGE (pLGE1 and 2) (Fig. 6A-C, Fig. S8A-D). Scattered Six3⁺ cells were also observed in the immature striatum (Fig. 6A-C, Fig. S8A-E) (Oliver et al., 1995). Six3 expression was also observed in the VZ of the medial ganglionic eminence (MGE) and preoptic area (POA) (Fig. 6A-C) (Xu et al., 2010; Sandberg et al., 2016).

At E16.5, SIX3 continued to be expressed in the vLGE SVZ, but not in the dLGE (Fig. 6D, Fig. S8F-I). BrdU 30 min pulse-labeling demonstrated that about 9% of SIX3⁺ cells in the LGE SVZ were in S phase (Fig. 6D-F,K). We found that, in the LGE SVZ, ∼90% SIX3⁺ cells co-expressed SP9 and ∼80% DRD2-GFP⁺ cells co-expressed SIX3 (Fig. 6G-M). Furthermore, SP9 expression in the LGE SVZ began before SIX3 (based on the position of the SP9 single-positive cells closer to the VZ; Fig. 6J) and SIX3 expression began before DRD2-GFP (Fig. 6G-I). As SIX3⁺ cells migrated from the SVZ into the MZ, most of them expressed lower SIX3 levels (Fig. 6A-D,H). SIX3/SP8 double immunostaining revealed that a subset of SP8⁺ cells in the E13.5 and E16.5 LGE SVZ and MZ also expressed SIX3 (Fig. S8A-J). In contrast, SIX3⁺ cells in the LGE did not express D1 MSN markers ISL1 or EBF1 (Fig. S9A-F).

Most Sp8, Sp9 and Sp8/9 mutant cells appear to be present in the mutant LGE SVZ based on Sp9-lacZ expression (in Sp9-KO), and Dlx5/6-GFP expression (in Sp8-CKO and Sp8/9-DCKO) at E16.5 (Fig. 7A-D) (Zhang et al., 2016). However, Six3 expression was reduced in the LGE SVZ of Sp9-KO mice and was almost undetectable in of Sp8/9-DCKO mice, whereas its expression appeared normal in Sp8-CKO mice (Fig. 7E-L; Fig. S10A-D). Six3 expression in the LGE SVZ at E16.5 was also almost undetectable when we used a Nestin-Cre transgenic allele to knock out Sp8 and Sp9 (Fig. 7M-P). Together, these data show that Sp9, together with Sp8, promotes Six3 expression in the SVZ.

To examine Six3 function in MSN development, we used Dlx5/6-CIE to delete exon 1, which encodes the Six domain and homeodomain of the Six3 gene (Liu et al., 2006). We analyzed D1 and D2 MSNs in the mutant striatum at P11. More than 90% of Drd2⁺, Adora2a⁺, Penk⁺ and Gpr6⁺ D2 MSNs were lost in the Six3-CKO striatum (Fig. 8A-L), whereas the majority of Drd1⁺ and Tac1⁺ D1 MSNs remained (Fig. 8M-R). Thus, the preferential reduction of D2 MSNs in the Six3-CKO closely phenocopied the Sp8/9-DCKO.

To test the hypothesis that SP9 directly promotes Six3 expression in the LGE SVZ, we performed ChIP-Seq using E13.5 and E16.5 wild-type ganglionic eminences (LGE+MGE) (McKenna et al., 2011; Sandberg et al., 2016) and an SP9 rabbit polyclonal antibody (Zhang et al., 2016). The presence of an SP9 ChIP-seq peak, combined with RNA-Seq data, is presumptive evidence for SP9 in vivo binding and possible transcriptional regulation of a locus. E13.5 ChIP-Seq had 6272 peaks and E16.5 ChIP-Seq had 6561 peaks (Fig. 9A). Intersection of the two ChIP results yielded 3976 binding peaks (Fig. 9A, ∼60% overlap). SP9 binding was located in intergenic (∼36%), intronic (∼30%), promoter [±1 kb of the TSS (transcription start site)] (27%), and other genomic regions (exon, 5′ UTR and 3′ UTR) (Fig. 9B). De novo motif discovery using genomic sequences associated with SP9 binding was performed using HOMER (Heinz et al., 2010). The most frequent de novo motif, likely representing the primary sequence motif, enriched to the center of SP9-bound DNA fragments (Fig. 9D). According to the de novo motif analyses, (G|T)(G|C)TAATTA is the primary motif of SP9 in the developing ganglionic eminences (Fig. 9D), and ∼40% binding peaks had this motif. In addition to SP9 motifs, frequent motifs included in DLX1, YY2 and MEIS1 were also present in the SP9 ChIP-Seq peaks (Fig. 9E). The TF genes Dlx1 and Meis1 are highly expressed in the LGE (Long et al., 2009), suggesting that they may function as co-regulators with SP9 to promote striatal development.

We found an SP9 ChIP-Seq peak at the Six3 promoter region (−5 to +411 of the TSS) (Fig. 9C), and several additional peaks upstream of the Six3 gene (regions 1, 2 and 3) (Fig. 9C); each of these loci are evolutionarily conserved (see mammal conservation track in Fig. 9C). Thus, these noncoding domains are strong candidates for the Six3 promoter and three Six3 enhancers.

Next, to test the ability of SP9 to regulate these candidate regulatory elements, we performed a dual-luciferase transcription activation assay using P19 embryonal carcinoma cells. SP9 robustly activated transcription from the putative promoter of Six3 and to a lesser extent from Six3 region 1, whereas it did not activate Six3 region 2 and 3 (Fig. 9F). Together, these results provide evidence that SP9 promotes Six3 expression in the LGE SVZ through direct binding to the promoter and possibly through binding to regulatory elements within the Six3 locus.

The dLGE mainly generates OB interneurons, whereas the vLGE mainly generates striatal MSNs. Sp8-CKO and Sp9-KO single mutants differentially alter LGE development. The salient phenotype of Sp8-CKO mice is disruption of OB interneuron development (Waclaw et al., 2006; Li et al., 2011, 2017), whereas in Sp9-KO mice D2-MSN development is disrupted (Zhang et al., 2016). Here, by comparing the phenotype of Sp8/9-DCKO with that of the single mutants, we provide evidence for distinct and redundant functions of Sp8 and Sp9. Our RNA-Seq and SP9 ChIP-Seq results provide strong evidence that Sp8 and Sp9 coordinately promote D2 MSN production by directly activating Six3 transcription.

Sp8 and Sp9 expression are regionally distinct within the LGE. Strong Sp8 expression is largely restricted to the SVZ of the dLGE (pLGE1 and 2), and weak Sp8 expression is largely restricted to the SVZ of the vLGE (pLGE3). By contrast, Sp9 is expressed uniformly throughout the SVZ of the LGE (pLGE1-4). The strongest effect of Sp8 knockout is in the dLGE, whereas the strongest effect of Sp9 knockout is in the vLGE. Although Sp8 and Sp9 have redundant functions in promoting neurogenesis in the OB (Li et al., 2017) and striatum (this study), they have different leading roles in these processes.

During OB interneuron development in the dLGE and postnatal SVZ-RMS-OB, Sp8 has the central role, based on three lines of evidence. First, Sp8 expression is upregulated in the Sp9-KO OB, whereas Sp9 expression is not upregulated in the Sp8-CKO OB (Li et al., 2017). Second, a large number OB interneurons are depleted in Sp8-CKO mice (Waclaw et al., 2006; Li et al., 2011, 2017), but not in Sp9-KO mice (Li et al., 2017). Third, Prokr2 and Tshz1 expression levels are reduced in the dLGE and postnatal SVZ of Sp8-CKO mice, but not in Sp9-KO mice (Li et al., 2017). Prokr2 and Tshz1 are known to promote OB neuronal differentiation and migration and are involved in human Kallmann syndrome (Ng et al., 2005; Matsumoto et al., 2006; Ragancokova et al., 2014).

Sp8 is required to promote neuronal differentiation (transition of neural progenitors to neurons) in the SVZ of the LGE. Loss of Sp8 function results in subtle accumulation of neural progenitors in both the dLGE and the vLGE (Fig. 4, Fig. S6). In contrast, accumulation of neural progenitors is not observed in Sp9 mutant LGE SVZ (Fig. 4, Fig. S6).

During the generation of D2 MSNs in the vLGE, Sp9 has the central role, based on four lines of evidence. First, in the SVZ of the vLGE, Sp9 is strongly expressed, whereas Sp8 is weakly expressed (Fig. 1, Figs S1 and S2). Second, unlike Sp8, which is downregulated in most D2 MSNs, Sp9 is expressed in nearly all postmitotic D2 MSNs and promotes their differentiation and survival (Fig. S1E-H) (Zhang et al., 2016). Third, there are many more MSNs, especially D2 MSNs, lost in Sp9-KO mice than in Sp8-CKO mice. Fourth, Six3 expression in the LGE SVZ is downregulated in Sp9-KO, but not in Sp8-CKO, mice (Fig. 7, Fig. S10).

In Sp9 mutants, Sp8 expression is more broadly expressed in the SVZ (Fig. 1I,J and Fig. S2), suggesting that this may enable to Sp8 compensate for the loss of Sp9. Indeed, this hypothesis is borne out. In the Sp8/9-DCKO, development of both OB interneurons and striatal D2 MSNs are much more severely disrupted than in single mutants (Li et al., 2017) (Fig. 3, this study). Sp8 also compensates in striatal development of Sp9-KO mice, in which the few D2 MSNs that are produced have increased SP8 expression (Fig. 2C-N).

As noted above, Sp8-CKO mice have more progenitors in the LGE than do controls and Sp9-CKO mice (Fig. 4, Fig. S6). Furthermore, in the Sp8/9-DCKO mice there are more VZ-SVZ progenitors than in Sp8-CKO mice (Fig. 4, Fig. S6). Thus, together Sp8 and Sp9 promote the progression of progenitors out of the cell cycle and towards a neural fate.

In the Sp8/9-DCKO mice, newly born neuroblasts in the dLGE and postnatal SVZ-RMS-OB fail to express Prokr2 and Tshz1 lack virtually all OB mature interneurons. We suggest that the lack of Prokr2 and Tshz1 RNA in large part accounts for the defects in differentiation and migration (tangential and radial) of OB immature interneurons, as similar phenotypes are observed in Prokr2 (and Tshz1) mutant mice (Ng et al., 2005; Matsumoto et al., 2006; Ragancokova et al., 2014; Li et al., 2017). Thus, in the dLGE and postnatal SVZ-RMS, Sp8 and Sp9 coordinately regulate OB interneuron development mainly by promoting Prokr2 and Tshz1 expression; however, here Sp8 has the leading role. This contrasts with the vLGE, where Sp9 has the leading role in driving Six3 expression and the generation of D2 MSNs (Fig. 7) (Zhang et al., 2016). Nonetheless, Sp8 can compensate in this process as well.

In Sp8/9-DCKO mutants, neural progenitors accumulate in the SVZ (Fig. 4, Fig. S6) (Li et al., 2017). These cells have increased Dlx1/2 TF expression (Fig. 4, Fig. S6), which provides evidence that they are arrested as immature neural progenitors (Yun et al., 2002). A subset of mutant SVZ cells have also increased ASCL1⁺ proneural TF expression. ASCL1, by driving Dll1 (delta-like 1) expression, can increase Notch signaling in adjacent cells and thereby promote the maintenance of their progenitor state (Fig. 4, Fig. S6, Table S1) (Castro et al., 2011). In addition, ASCL1 promotes expression of genes that inhibit cell cycle progression and/or promote cell cycle arrest (Fig. 4, Fig. S6) (Castro et al., 2011).

In the striatum, we observed that D1 MSNs were largely unaffected in Sp8-CKO, Sp9-KO and Sp8/9-DCKO mice (Fig. 3K-T, Fig. S5Q-X). In addition, we observed a reduced number of EBF1⁺ cells in the LGE SVZ of Sp8/9-DCKO mice at E16.5 (Fig. 5F-J). The TF Ebf1 is specifically expressed in young D1 MSNs, and is essential for the differentiation of D1 MSNs (Lobo et al., 2006, 2008). Thus, reduced EBF1⁺ cells in the LGE SVZ provides evidence that SVZ cells in the D1 MSN developmental pathway may not be normal. However, this could be a result of a general SVZ phenotype in the mutants as the increased ASCL1 expression is observed in the presumptive both D1 and D2 MSN progenitors. Therefore, EBF1 reduction in the E16.5 LGE SVZ does not directly demonstrate a defect in the development of D1 MSNs.

Six3 has multiple roles in embryonic development. In the forebrain, Six3 is required to activate SHH signaling and repress WNT signaling to provide key patterning instructions (Lagutin et al., 2003; Geng et al., 2008). In retinal development, Six3 is crucial for the activation of Pax6, a key regulator of eye development (Liu et al., 2006). Six3 is also required for ependymal cell maturation; Six3 mutants develop hydrocephaly (Lavado and Oliver, 2011).

Six3 is expressed in the developing basal ganglia. Mice haploinsufficient for Six3 (Six3^+/ki; Shh^+/− mouse) fail to activate Shh and Nkx2-1 expression in the MGE (Geng et al., 2008; Geng and Oliver, 2009), whereas Ebf1 expression in the LGE is largely unaffected. A recent study (Gokce et al., 2016) and our analysis indicate that during striatal development Six3 is preferentially expressed in and functions in the D2 MSN lineage (Figs 6 and 8), but not in the D1 MSN lineage (Fig. S9). Furthermore, Sp9 and Sp8/9 mutants have greatly reduced Six3 expression (Fig. 7, Fig. S10), and Six3-CKO mutants fail to generate most D2 MSNs, which closely phenocopies the Sp8/9-DCKO mutants (Fig. 3, Fig. S3, Fig. 8).

Six3 is not expressed in the dLGE (Fig. 6, Fig. S8) and, accordingly, the Six3-CKO mutants did not show altered Tshz1 expression in the dLGE (Fig. S11). Six3 is also not expressed in the postnatal SVZ-RMS-OB (data not shown and Allen Developing Mouse Brain Atlas: developingmouse.brain-map.org/gene/show/20235). Thus, it is not surprising that we did not find major olfactory bulb interneuron defects in the Six3-CKO mice (data not shown). Six3 LGE expression is restricted to the vLGE SVZ, and is most prominent in the pLGE3 (Flames et al., 2007). This suggests that D2 MSNs are largely generated by pLGE3. We hypothesize that D1 MSNs are mainly generated by a more ventral domain (pLGE4). Future studies should address this hypothesis and also search for additional transcription mechanisms that generate D1 MSNs (Lobo et al., 2006; Shirasaki et al., 2012; Ehrman et al., 2013; Lu et al., 2014; Merchan-Sala et al., 2017).

To investigate the genomic loci regulated by SP9, we performed SP9 ChIP-seq using E13.5 and E16.5 ganglionic eminences. We found that SP9 binds to the Six3 promoter and several putative Six3 enhancers. The Six3 promoter was strongly activated by SP9 in luciferase reporter assays. In addition, one putative enhancer (Six3 region 1, Fig. 9C) was moderately activated. Thus, we propose that SP9 directly promotes Six3 expression in the pLGE3 SVZ.

In summary, this study provides evidence that Sp8/9 and Six3 are crucial components in the SVZ of the pLGE3, where they selectively drive the generation of D2 MSNs. Thus, these findings broaden the foundation for elucidating the transcriptional mechanisms underlying striatal MSN development and survival. This information has ramifications for understanding how to engineer MSNs in vitro, and for understanding disorders of striatal development and degeneration.

All experiments were performed in accordance with institutional guidelines at Fudan University. Sp9^lacZ/+ (Zhang et al., 2016), Sp9^F/F (Zhang et al., 2016), Sp8^F/F (Bell et al., 2003), Six3^F/F (Liu et al., 2006), Bax^−/+ (Knudson et al., 1995), Dlx5/6-CIE (Stenman et al., 2003), Nestin-Cre (Tronche et al., 1999; Graus-Porta et al., 2001) and Drd2-GFP (from MMRRC) (Gong et al., 2007) mice were previously described. Wild-type, Dlx5/6-CIE or Sp8/9 floxed littermate mice without Cre allele were used as controls. Mice were group-housed with approximately four to a cage on a 12-h light, 12-h dark cycle, with ad libitum chow and water. These mice were maintained in a mixed genetic background of C57BL/6J, 129S6 and CD1. The day of vaginal plug detection was calculated as E0.5, and the day of birth was considered as P0. Males and females were both used in all experiments.

A single intraperitoneal injection of BrdU (50 mg/kg, body weight) was given to pregnant dams at E13.5, and the pups were sacrificed at P0, or a single intraperitoneal injection of BrdU was given to pregnant dams at E15.5, and the mice were sacrificed at E16.5.

Postnatal mice were deeply anesthetized and perfused intracardially with 4% paraformaldehyde (PFA) in 1×PBS (pH 7.4); embryonic brains were fixed by immersion in 4% PFA. All brains were fixed overnight in 4% PFA, cryoprotected in 30% sucrose for at least 24 h, frozen in embedding medium and cryosectioned.

Immunohistochemistry was performed on 12-μm-thick cryostat sections on Fisherbrand Superfrost Plus microscope slides or 30- to 50-μm-thick free-floating sections (Ma et al., 2013; Zhang et al., 2016). For SP9, EBF and SIX3 immunohistochemistry, sections were boiled in 10 mM sodium citrate briefly for antigen retrieval (Ma et al., 2013). For BrdU staining, sections were incubated in 2 N HCl for 1 h at room temperature, and then rinsed in 0.1 M borate buffer twice (Liu et al., 2009).

The following primary antibodies were used: rabbit anti-SP9 (1:500) (Zhang et al., 2016), goat anti-SP8 (1:3000; Santa Cruz Biotechnology, sc-104661); chicken anti-GFP (1:3000; Aves Labs, GFP-1020); rabbit anti-ASCL1 (1:2000; Cosmo Bio, SK-T01-003); rat anti-BCL11b (1:2000; Abcam, ab18465); rat anti-BrdU (1:200, Accurate Chemical, OBT0030s); chicken anti-β-gal (1:1000; Abcam, ab9361); rabbit anti-cleaved caspase 3 (1:300; Cell Signaling, 9661); rabbit anti-FOXP1 (1:2000; Abcam, ab16645); goat anti-DCX (1:1000; Santa Cruz Biotechnology, sc-8066); mouse anti-TUBB3 (1:500; Covance, MMS-435P); goat anti-EBF (1:200; Santa Cruz Biotechnology, sc-15888); rabbit anti-EBF1 (1:2000; Merck, AB10523); mouse anti-SIX3 (1:500; Santa Cruz Biotechnology, sc-398797); rabbit anti-SIX3 (1:2000; Rockland Immunochemicals, 600-401-A26).

All in situ RNA hybridization experiments were performed using digoxigenin-labeled riboprobes on 20-μm-thick cryostat sections (Long et al., 2009; Zhang et al., 2016). Riboprobes were amplified by PCR using the following primers: Gsx2 Forward: CCTTTGTTCGAGTCCCAGACACC; Gsx2 Reverse: AAAGGGACTTCCAGAGACCTGGAT; Dlx1 Forward: ATGACCATGACCACCATGCCAG; Dlx1 Reverse: TCACATCAGTTGAGGCTGCTGC; Dlx2 Forward: TTTCCTGTCCCGGGTCAGGAT; Dlx2 Reverse: AAGTCTCAGACGCTGTCCACTCGA; Ascl1 Forward: CTAACAGGCAGGGCTGGA; Ascl1 Reverse: TAAGGGGTGGGTGTGAGG; Hes5 Forward: TAATCGCCTCCAGAGCTCCAGGC; Hes5 Reverse: ACACAAAACAACCCCACGGGGTC; E2f1 Forward: TGGTAAGCGGCTTGAAGGCCTG; E2f1 Reverse: AATGTGGCAGCAACCAAACCCC; Cdca7 Forward: AGACTCTCGAGACGTTTGCTA; Cdca7 Reverse: AGCTGCAGTTGCAAATTCCTC; Cdk1 Forward: GATGTAGCCCTCTGGATGGA; Cdk1 Reverse: GCCCCTGATCTCTAGCTGTG; Gadd45g Forward: CCCTCCGCACTCTTTTGGATAACT; Gadd45g Reverse: ATTCAAAGCTTCCACGATAGCGTC; Btg2 Forward: GCCAGACCGTCATCATCGTTCTAAT; Btg2 Reverse: CCTAGGCAAACACTGGCTCACAGA; Six3 Forward: CTCTATTCCTCCCACTTCTTGTTGC; Six3 Reverse: CATCACATTCCGAGTCGCTGG; Chat Forward: GATGCCTATCCTGGAAAAGGTCC; Chat Reverse: CGTTGGACGCCATTTTGACTATC.

RNA-Seq analysis was performed as previously described (Li et al., 2017). The LGE (including VZ, SVZ and MZ) from E16.5 Sp8/9-DCKO mice and littermate controls (Sp8/9 floxed mice without Dlx5/6-CIE) were dissected (n=3 mice per group). RNA quality was assessed using a bioanalyzer (Agilent Technologies). RNA-Seq libraries were prepared according to the Illumina TruSeq protocol. Levels of gene expression were reported in FPKM (Trapnell et al., 2012; Li et al., 2017). A gene was considered to be expressed if it had an FPKM>1. Genes were considered differentially expressed if P<0.05.

SP9 ChIP-Seq was performed using dissected E13.5 and E16.5 ganglionic eminences (LGE+MGE) of wild-type CD1 mice, and an SP9 rabbit polyclonal antibody as described previously (McKenna et al., 2011; Sandberg et al., 2016; Zhang et al., 2016). Briefly, the E13.5 or E16.5 ganglionic eminences (including VZ, SVZ and MZ) (two litters for one independent ChIP experiment) were dissected, and fixed in 10 ml 1% formaldehyde for 10 min at room temperature. The formaldehyde was quenched by adding glycine to a final concentration of 0.125 M, and the samples were incubated for 5 min at room temperature. Tissues were then washed twice in cold PBS, pelleted, re-suspended in 1% SDS lysis buffer (with 1 mM PMSF/cocktail), and incubated on ice for 10 min. Sonication was performed using Bioruptor Pico (15-20 cycles) to keep most DNA fragments in the range of 200-500 bp. Eight micrograms anti-SP9 rabbit polyclonal antibody were incubated with the cleared chromatin (SDS concentration diluted to 0.1%) at 4°C overnight. Protein A+G beads were added to the samples and incubated for 4 h at 4°C. The samples were washed with low salt [150 mM NaCl, 0.1% w/v SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.0)], high salt [500 mM NaCl, 0.1% w/v SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.0)] and LiCl solutions [250 mM LiCl, 1% NP-40, 1% w/v sodium deoxycholate, 10 mM Tris-HCl (pH 8.0), 1 mM EDTA], and the precipitated DNA was eluted in elution buffer (0.1 M NaHCO₃ and 1% w/v SDS). The eluted DNA samples were reverse cross-linked overnight in the presence of 250 mM NaCl, treated with RNase and proteinase K and purified using the Zymo Research D5205 Kit. ChIP-seq sequencing libraries were prepared from both the input DNA and the precipitated DNA using the NEBNext DNA Library Prep Kit and Illumina standard adaptors, and were sequenced using a 150 bp pair-end strategy using the Illumina Hiseq 4000 platform.

Reads from the ChIP and input libraries were mapped to the mouse genome (mm9) using Bowtie with default parameters. Peak calling between the replicate ChIP samples and the input control sample was carried out using model-based analysis for ChIP-Seq (MACS). Peaks were called considering input as the control sample with filtering to remove peaks in repeat regions. Motif analysis of ChIP-Seq peak DNA sequences was performed using HOMER. ChIP-Seq peaks were visualized using the UCSC genome browser.

The DNA fragments of Six3 promoter, Six3 putative enhancers (region 1, region 2 and region 3, Fig. 9C) were amplified by PCR and subsequent cloned into pGL4.10 and pGL4.23 firefly luciferase vector (Promega) upstream of the Luc2 gene, respectively (McKinsey et al., 2013).

The primers used for amplifying the Six3 promoter were: Fwd: 5′-CCTCCGGCTAAGTGGTAAAACCGTC-3′; Rev: 5′-GGAGGAGGAAGGACGTAAGGGACAC-3′. The primers used for amplifying the Six3 putative enhancers were: Region 1 Fwd: 5′-TATTGTTCGGGATTTGCCGAAATC-3′; Region 1 Rev: 5′-GAGGAGGCTGCCACTGCTTAACAG-3′; Region 2 Fwd: 5′-TAGGCGATTCAATAGTTAATGGAGG-3′; Region 2 Rev: 5′-ATGACAAACTCATCTCCCAGCATT-3′; Region 3 Fwd: 5′-CCAGGCCTCGCACAGTCTTTCTT-3′; Region 3 Rev: 5′-GAGCAGGGCAAACAATGAGCAAAC-3′. The coding sequence (CDS) of Sp9 was amplified by PCR from mouse genomic DNA and cloned into pCS2 (+) expression vector, using the following primers: Sp9 CDS Fwd: 5′-GACCTGAATCGTGATTCCCAGC-3′; Sp9 CDS Rev: 5′-TGCAACCCACATAAACTTCATTGC-3′.

Mouse embryonal carcinoma P19 cells were grown in MEMα (Gibco, 12571-063) supplemented with 10% fetal bovine serum (Gibco 10099-141). For the luciferase assay, P19 cell transfections were performed in triplicate in 24-well plates using Fugene HD transfection reagent according to the manufacturer's protocol (Promega, E2311). The amounts of plasmids for each transfection were: 40 ng pGL4.73 (Promega, Renilla luciferase vector), 240 ng pCS2-empty or pCS2-Sp9, 240 ng pGL4.23-empty (pGL4.10-empty) (Promega), or pGL4.23-Element (pGL4.10-Promoter). The dual luciferase assay was performed with a GloMax 20/20 Luminometer (Promega). Statistical significance was determined by Student's t-test.

Bright-field images (in situ hybridization results) and some fluorescent images were imaged with Olympus BX 51 microscope using a 4× or 10× objective. Other fluorescent images were taken with the Olympus FV1000 confocal microscope system using 10×, 20×, 40× or 60× objectives. z-stack confocal images were reconstructed using FV10-ASW software. All images were merged, cropped and optimized in Photoshop CS5 without distortion of the original information.

Cell quantification was made in the SVZ of LGE or in the striatum using three randomly chosen sections from mutant and their respective control mice (n=3 mice per genotype per age).

The numbers of DRD2-GFP⁺/SP8⁺ cells in the striatum of Drd2-GFP control and Drd2-GFP; Sp9-KO mice at P9 were quantified in three randomly chosen 30-μm-thick sections from each mouse, and three mice each group were analyzed.

The numbers of Drd2⁺, Adora2a⁺, Drd1⁺, Tac1⁺ and ChAT⁺ cells in the striatum of Dlx5/6-CIE control, Sp8-CKO, Sp9-KO, Sp8/9-DCKO and Six3-CKO mice at P11 were quantified in three randomly chosen 20-μm-thick sections from each mouse, and three mice per group were analyzed.

The LGE VZ and SVZ were defined based on a BrdU pulse-labeling experiment (see Fig. S1E-H). The numbers of FOXP1⁺ and EBF1⁺ cells in the E16.5 LGE SVZ, the numbers of BrdU⁺ cells in the LGE VZ/SVZ and MZ, and the numbers of BrdU⁺ cells in the P0 striatum were quantified in three randomly chosen 12-μm-thick sections from each mouse, and three mice per group were analyzed.

The numbers of cleaved caspase 3⁺ cells in the P0, P2 and P3 striatum were quantified in three randomly chosen 20-μm-thick sections from each mouse, and three mice per group were analyzed.

The numbers of SIX3⁺, SIX3⁺/BrdU⁺, DRD2-GFP⁺ and SIX3⁺/DRD2-GFP⁺ cells in the E16.5 LGE SVZ were quantified in three randomly chosen 12-μm-thick sections from each mouse, and three mice per group were analyzed.

Statistical significance was assessed using unpaired Student's t-test or one-way ANOVA followed by the Tukey–Kramer post-hoc test. All quantification results were presented as the mean+s.e.m. P-values less than 0.05 were considered significant.

We are grateful to Guillermo Oliver at Northwestern University for the generous gift of Six3^F/F mice, and Kenneth Campbell at University of Cincinnati College of Medicine for the Sp8^F/F and Dlx5/6-CIE mice.