Single-stranded DNA binding proteins are required for LIM complexes to induce transcriptionally active chromatin and specify spinal neuronal identities

The LIM homeodomain (LIM-HD) family of transcription factors regulate diverse developmental programs in animals (Hobert and Westphal, 2000). All LIM-HD factors interact with nuclear adaptor protein NLI (nuclear LIM-interactor; also known as Ldb1, CLIM, Chip) via their LIM domains and bind DNA through their homeodomains (Agulnick et al., 1996; Bach et al., 1997; Breen et al., 1998; Jurata and Gill, 1997; Jurata et al., 1996,, 1998). Although LIM-HD factors appear to function via transcriptional activation, the mechanisms underlying this transcriptional activation during development have been poorly understood.

The widely expressed nuclear protein NLI also interacts with other types of transcription regulators, including GATA, basic helix-loop-helix (bHLH) factors, Ptx1 and bicoid (for a review, see Matthews and Visvader, 2003). NLI has three evolutionarily conserved domains: the N-terminal region that is capable of self-dimerization, the C-terminal region that interacts with all nuclear-class LIM factors and the Ldb1/Chip conserved domain (LCCD) that associates with single-stranded DNA-binding proteins (Ssdps or SSBPs) (Chen et al., 2002; Jurata and Gill, 1997; van Meyel et al., 2003). As a result of its interaction with a wide array of transcription factors and its ability to self-dimerize, NLI has been proposed as a structural linker to nucleate the formation of multi-protein transcription complexes involved in cell fate determination during development. Correspondingly, NLI null mouse embryos exhibit pleiotrophic abnormalities, including forebrain truncations, suggesting the broad roles of NLI in development (Mukhopadhyay et al., 2003). However, it remains unclear whether NLI contributes directly to the transcriptional activity of NLI-containing complexes.

Structural and functional analyses have revealed that LIM-HD factors form a higher-order complex with NLI. In vertebrates, LIM complexes have been best characterized in motor neuron (MN) and V2a-interneuron (V2a-IN) specification in the ventral spinal cord (Lee and Pfaff, 2001). LIM-HD factors Isl1 and Lhx3 direct differentiation of MNs and V2a-INs by forming two cell type-specific LIM complexes with NLI (Thaler et al., 2002). In V2a-INs expressing Lhx3, interactions of two Lhx3 proteins with NLI-homodimer lead to a tetrameric complex of 2NLI:2Lhx3, denoted as the V2-tetramer (Fig. 1A). In MNs that express both Isl1 and Lhx3, Isl1 uses its LIM domains and C-terminal domain to simultaneously bind NLI and Lhx3, respectively, generating a hexameric complex of 2NLI:2Isl1:2Lhx3, namely the MN-hexamer (Fig. 1A).

Ssdps have been reported to associate with NLI in biochemical and genetic studies (Chen et al., 2002; van Meyel et al., 2003). Whereas Drosophila and C. elegans have a single Ssdp gene, mammals have three genes: Ssdp1 (also known as Ssbp3), Ssdp2 and Ssdp4 (also known as Ssbp2 and Ssbp4). All Ssdps share an evolutionarily conserved forward/LUFS domain at the N-terminus, which interacts with the LCCD of NLI (van Meyel et al., 2003). Ssdp mutants display phenotypes similar to those observed with NLI mutants in mouse and Drosophila, and Ssdp genes and NLI interact genetically (Chen et al., 2002; Enkhmandakh et al., 2006; Nishioka et al., 2005; van Meyel et al., 2003). Ssdp2 facilitates the transcriptional activity of an erythroid DNA-binding complex containing NLI by stabilizing NLI and the LIM-only protein LMO2 (Xu et al., 2007). Ssdp1 also enhances transcriptional activation by LIM-HD factor Lim1 in mammalian cells (Nishioka et al., 2005). Although these data suggest a possible role for Ssdps as specific coactivators of NLI-containing LIM complexes, the mechanisms of action in transcriptional activation and the role of Ssdps in development of the vertebrate central nervous system (CNS), in which many LIM-HD factors control cell identities, remain largely unexplored.

Coactivators play essential roles in rendering transcription factors functional by affecting the transcriptional machinery in a variety of ways. These include their associated enzymatic activities, such as histone acetyltransferases and methyltransferases (Berger, 2007; Naar et al., 2001). Coactivators with histone acetyltransferase activity, including CBP and its paralog p300, acetylate lysine residues of histone tails and neutralize their positive charge, leading to chromatin decondensation (Marmorstein, 2001). Histone H3 lysine 4 trimethylation (H3K4me3) is an evolutionarily conserved mark linked to transcriptionally active chromatin. In higher eukaryotes, H3K4me3 is tightly associated with promoters and early transcribed regions of active genes (Schneider et al., 2004) and has been proposed to counter the generally repressive chromatin environment (Ruthenburg et al., 2007). Mammalian enzymes responsible for H3K4me3 include Set1α/β and mixed-lineage leukemia (MLL) 1-4 (Ruthenburg et al., 2007).

Here, we identify Ssdp1 and Ssdp2 (Ssdp1/2) as coactivators involved with neuronal subtype specification in the CNS development. We show that Ssdp1/2 enhance transcriptional activity of LIM complexes that specify neuronal subtypes, the MN-hexamer and V2-tetramer. Correspondingly, loss-of-function studies in mouse and chick embryos reveal that Ssdp1/2 are required for the differentiation of spinal MNs and V2a-INs. Furthermore, we demonstrate that Ssdp1/2 enable the MN-hexamer to induce histone H3 acetylation and H3K4me3 of its target genes by recruiting chromatin-remodeling coactivators. Our results suggest that Ssdp1/2 function as essential transcriptional coactivators of LIM complexes, enabling their ability to specify spinal neuronal identities during development.

To investigate the role of Ssdps in the developing spinal cord, we first examined their expression pattern in developing mouse and chick embryos using immunostaining and in situ hybridization. Overall, all three Ssdps are widely expressed, but relatively enriched in the spinal cord at embryonic day (E) 9.5 to E12.5, encompassing the period when MNs and V2a-INs are specified (Fig. 1B, data not shown). Notably, Ssdp1 is highly upregulated in differentiating neurons, similar to NLI (Fig. 1B) (Thaler et al., 2002). Ssdp2 expression is ubiquitous, but higher in the ventricular zone and MNs (Fig. 1B).

To monitor the autonomous transactivation function of Ssdp1 in the developing spinal cord, we electroporated UAS:GFP reporter with Ssdp1 fused to Gal4-DNA binding domain (Gal4DBD) into chick embryos. Whereas Gal4DBD alone showed little transcriptional activity, Gal4DBD-Ssdp1 activated the UAS:GFP reporter in the chick neural tube, triggering GFP expression in the entire electroporated side (Fig. 1C). The Ssdp1ΔN100 mutant lacking the NLI-interacting domain, residues 1-100 (van Meyel et al., 2003), exhibited a potent transactivation function (Fig. 1C), suggesting that the interaction with NLI is dispensable for the transactivation function of Ssdp1. Similarly, both Ssdp1 and Ssdp1ΔN100 strongly stimulated UAS:LUC reporter activity in postnatal day (P) 19 mouse embryonic cells, whereas Ssdp1-N150 encoding the N-terminal 150 amino acid region of Ssdp1 had much weaker activity (Fig. 1D). Ssdp2 also showed a significant level of autonomous transactivation function (Fig. 1D). These data indicate that Ssdp1/2 are expressed in the developing spinal cord and both Ssdp1 and Ssdp2 have autonomous transactivation function.

The expression and autonomous transactivation potential of Ssdp1 in the embryonic spinal cord raises the possibility that Ssdp1 serves as a coactivator for the MN-specifying MN-hexamer and the V2a-IN-specifying V2-tetramer (Fig. 1A). To test whether Ssdp1 associates with the MN-hexamer and V2-tetramer, we performed coimmunoprecipitation assays in P19 cells. Expression of Isl1 and Lhx3 leads to the assembly of the MN-hexamer with endogenous NLI and triggers differentiation of P19 cells to MNs (Lee and Pfaff, 2003). We purified a protein complex containing Ssdp1 using anti-HA antibody in P19 cells transfected with Isl1, Lhx3 and HA-tagged Ssdp1 and found that both Isl1 and Lhx3 associate with Ssdp1 in cells, demonstrating the association of Ssdp1 with the MN-hexamer (Fig. 2A). Likewise, Ssdp1 interacts with Lhx3, which forms the V2-tetramer with endogenous NLI, in P19 cells transfected with Lhx3 alone (Fig. 2A). We also performed coimmunoprecipitation assays with E13.5 mouse spinal cord extract and found association of endogenous Isl1 with Ssdp1 and Ssdp2 (Fig. 2B).

To test whether Ssdp1 is recruited to the target genomic element of the MN-hexamer, we performed chromatin immunoprecipitation (ChIP) assays. The MN-hexamer binds to the MN-specific enhancer of the Hb9 gene (Hb9-MNe) in the embryonic spinal cord (Lee et al., 2008; Lee and Pfaff, 2003). We expressed Flag-Ssdp1 without or with the MN-hexamer-mimicking chimera Isl1-Lhx3 in P19 cells and purified Ssdp1-associating chromatin using anti-Flag antibody (Fig. 2C). Interestingly, Ssdp1 occupied Hb9-MNe in chromatin only when it was coexpressed with Isl1-Lhx3 (Fig. 2C), indicating that Ssdp1 is recruited to target genomic elements of the MN-hexamer via the MN-hexamer. This prompted us to ask whether endogenous Ssdp1 and the related protein Ssdp2 occupy MN-hexamer and V2-tetramer target genomic elements in embryonic spinal cord cells. In our ChIP analysis with antibodies against Ssdp1 and Ssdp2 and E13.5 mouse embryonic spinal cord extracts, endogenous Ssdp1/2 bound not only Hb9-MNe, a MN-hexamer-binding enhancer, but also Chx10-TeRE, a recently defined target element of the V2-tetramer in a V2a-IN-specific gene Chx10 (also known as Vsx2; Fig. 2D) (Lee et al., 2008). By contrast, Isl1 and Lhx3, although they were robustly recruited to Hb9-MNe, failed to bind Untr6 (data not shown), a genomic region that recruits no transcription factors (Mali et al., 2008). These results support a role for Ssdp1/2 in the cell fate specifying activity of the MN-hexamer and V2-tetramer complexes.

Based on the direct interaction between Ssdp2 and NLI (van Meyel et al., 2003), Ssdp1/2 could be recruited to the MN-hexamer and the V2-tetramer via their shared component NLI. However, Ssdp1/2 might also associate with the MN-hexamer and V2-tetramer through their additional interactions with Isl1 or Lhx3. In glutathione S-transferase (GST) pull-down assays using bacterially purified GST-NLI and in vitro translated proteins, Ssdp1 showed no direct interactions with Isl1 and Lhx3, whereas the N-terminal 150 amino acid region of Ssdp1 efficiently bound NLI (Fig. 2E, data not shown). Thus, Ssdp1 is tethered to the MN-hexamer and the V2-tetramer through its direct interactions with NLI.

To precisely identify the residues within Ssdp1 important for NLI interaction, we tested interactions of deletion mutants of Ssdp1 with NLI in yeast two-hybrid assays. Ssdps have a conserved Forward/LUFS domain, comprising three α-helices at the N-terminal end, which interacts with NLI (van Meyel et al., 2003). We found that a region containing the Ssdp1 residues 49-100 within the Forward/LUFS domain binds to NLI (Fig. 2F). Thus, we generated a series of Ssdp1 mutants in which the evolutionarily conserved residues within 49-100 amino acid region are replaced with alanines (Fig. 2G) and tested their interaction with NLI in GST pull-down assays. Among five Ssdp1 mutants, m2 and m3, which contain alanines for Ssdp1 residues 58-63 and 65-70, respectively, failed to interact with NLI (Fig. 2G, data not shown).

To test whether Ssdp1/2 could function as coactivators for the MN-hexamer and V2-tetramer, we investigated the effect of Ssdp1 on the transcriptional activity of Gal4DBD-Isl1 and Gal4DBD-Lhx3 in the UAS:LUC reporter in P19 cells. Ssdp1 markedly stimulated transactivation by Isl1 and Lhx3 (Fig. 3A). Whereas NLI alone had little effect, NLI strongly synergized with Ssdp1 to increase Isl1- and Lhx3-mediated transactivation (Fig. 3A), indicating cooperation between NLI and Ssdp1 in transcriptional activation. To directly monitor the impact of Ssdp1 on transcriptional control by the MN-hexamer, we used HxRE:LUC reporter, which contains MN-hexamer binding sites and thus responds to the MN-hexamer (Lee et al., 2008). Ssdp1 augmented the MN-hexamer-mediated transactivation of HxRE:LUC in P19 cells, whereas Ssdp1-m2, which is defective in NLI interaction (Fig. 2G), failed to enhance transactivation by the MN-hexamer (Fig. 3B). Ssdp1ΔN100 repressed the activation of HxRE:LUC by Isl1 and Lhx3 (Fig. 3B), suggesting that Ssdp1ΔN100 antagonizes the function of endogenous Ssdp1/2 to activate HxRE:LUC. To examine whether endogenous Ssdp1 and Ssdp2 contribute to the MN-hexamer-mediated transactivation of HxRE:LUC, we used short-hairpin RNAs (shRNAs) targeted to Ssdp1 and Ssdp2 (Xu et al., 2007), which significantly compromised expression of endogenous Ssdp1 and Ssdp2, respectively, in P19 cells (Fig. S1). Reduction of Ssdp1 and Ssdp2 levels by cotransfecting sh-Ssdp1 and sh-Ssdp2 suppressed the transactivation by the MN-hexamer to ∼38% of control in HxRE:LUC (Fig. 3B). We next examined the action of Ssdp1 in transcriptional activation of V2a-IN genes by the V2-tetramer using TeRE:LUC, a reporter for V2-tetramer function (Lee et al., 2008). Ssdp1 markedly stimulated TeRE:LUC, whereas Ssdp1ΔN100 and sh-Ssdp1/2 inhibited it (Fig. 3C). These data demonstrate that Ssdp1/2 serve as crucial coactivators, which enhance transcriptional activation by the MN-hexamer and V2-tetramer, and that NLI interaction is necessary for this coactivator function.

To assess the effect of Ssdp1 in specification of MNs, we monitored the efficiency of ectopic MN formation in the chick spinal cord triggered by electroporation of Isl1 and Lhx3 without or with cotransfected Ssdp1. Electroporation of Isl1 and Lhx3, which form the MN-hexamer with endogenous NLI, generated ectopic MNs in ∼30% of Lhx3⁺ transfected cells (Fig. 3D). Cotransfection of Ssdp1 with Isl1 and Lhx3 markedly increased the induction rate of MNs to ∼58% (Fig. 3D). These results reveal that Ssdp1 promotes the function of the MN-hexamer in inducing MNs in the neural tube.

To investigate the role of Ssdp1 and Ssdp2 in the spinal neuronal specification during embryonic development, we examined mouse mutants in which Ssdp1 and/or Ssdp2 gene activity is impaired. We used the headshrinker (hsk) mice, which are Ssdp1 hypomorphic mutants that display severe defects in anterior head development (Nishioka et al., 2005). We generated an Ssdp2 mutant allele by replacing the first exon of Ssdp2 with a PGK-Neo cassette (Fig. 4A). The genotyping PCR and Southern blotting analyses confirmed a successful removal of the targeted exon of Ssdp2 (Fig. 4B, data not shown). In Ssdp2^−/− embryos, no Ssdp2 protein was detected by anti-Ssdp2 antibody in immunohistochemistry and western blot assays (Fig. 4C, data not shown). Although most Ssdp2^−/− embryos appeared morphologically normal, they occasionally displayed smaller heads (data not shown). The high homology of amino acid sequences between Ssdp1 and Ssdp2 and their shared property of autonomous transactivation and NLI interaction suggest functional redundancy between the two genes. To generate Ssdp1^hsk/hsk;Ssdp2^−/−, we produced Ssdp1^hsk/+;Ssdp2^+/− by mating Ssdp1^hsk/+ and Ssdp2^+/−. The number of Ssdp1^hsk/+;Ssdp2^+/− recovered by P28 was lower than that expected from the Mendelian ratio (P<0.05 in the chi-square test, Table S1), suggesting a reduction in survival of Ssdp1^hsk/+;Ssdp2^+/− animals. No Ssdp1^hsk/hsk;Ssdp2^−/− embryos were found from E10.5 onward, indicating their early embryonic lethality. The number of MNs was significantly reduced in Ssdp1^hsk/+;Ssdp2^−/− and most severely compromised in Ssdp1^hsk/hsk;Ssdp2^+/− mutants at E10.5, as determined by MN markers Hb9 and Isl2 (Fig. 4D,E, data not shown). Similarly, there were fewer V2a-INs in Ssdp1^hsk/hsk and Ssdp2^−/− at E11.5 compared with the wild-type control (Fig. 4F,G). Thus, the loss of Ssdp1 and Ssdp2 gene activity in mouse embryos leads to defects in the specification of MNs and V2a-INs during spinal cord development.

The presence of three highly related Ssdp genes in the mouse genome and the early embryonic lethality of Ssdp1/2 double homozygote mutants make more extensive loss-of-function analyses of Ssdp genes in mice difficult. Thus, we turned to chick embryos, which encode only two Ssdp genes, Ssdp1 and Ssdp2. To reduce the expression level of Ssdp1/2 at the stage when MNs are being specified, we electroporated duplex short interfering RNAs (siRNAs) against chick Ssdp1 and/or Ssdp2 into the chick neural tube at Hamburger–Hamilton stage (HH) 13 and assessed the development of MNs 1 day post-electroporation. si-Ssdp1 and si-Ssdp2 specifically downregulated chick Ssdp1 and Ssdp2, respectively, without cross-reaction (Fig. 5A,B). Knockdown of Ssdp1 or Ssdp2 drastically impaired MN differentiation to ∼50% compared with the unelectroporated control side, whereas electroporation of scrambled control RNA had no effect (Fig. 5A,B,F). Coelectroporation of both si-Ssdp1 and si-Ssdp2 further suppressed the generation of MNs to ∼30% of the control side (Fig. 5C,F), suggesting functional redundancy between Ssdp1 and Ssdp2 in MN specification and consistent with the recruitment of both Ssdp1 and Ssdp2 to the MN-hexamer target enhancer Hb9-MNe in mouse embryonic spinal cord (Fig. 2D). Knockdown of Ssdp1/2 did not decrease numbers of Sox2⁺ neural progenitors and BrdU⁺ proliferating cells (Fig. 5E, Fig. S2), indicating that the reduction in number of MNs caused by si-Ssdp1/2 was not due to the loss of neural progenitors. In support of this notion and the fact that Ssdp1/2 are not involved with patterning of progenitor domains, Pax6⁺ and Olig2⁺ progenitor domains remained intact in Ssdp1 and Ssdp2 mutant mouse embryos (Fig. S3A). Because Ssdp2 is known to stabilize NLI protein in erythroid cells (Xu et al., 2007), we examined whether the stability of NLI is affected upon reduction of Ssdp1/2 in the spinal cord. NLI expression was not changed by si-Ssdp2, but only slightly decreased by si-Ssdp1 alone or cotransfection of si-Ssdp1 and si-Ssdp2 (Fig. 5A-C). In addition, expression of NLI failed to recover the impaired MN development caused by knockdown of Ssdp1/2 (Fig. 5D,F). These results suggest that Ssdp1 is not likely to be involved in protecting NLI from proteasomal degradation in the developing spinal cord. Next, we monitored V2a-IN specification in the chick embryos 2 days post-electroporation of si-Ssdp1 or si-Ssdp2. Knockdown of Ssdp1 or Ssdp2 suppressed the number of Chx10⁺ V2a-INs to ∼50% (Fig. S4). Together, these results establish that Ssdp1 and Ssdp2 are required for the specification of spinal MNs and V2a-INs.

Based on our hypothesis that Ssdp1/2 play crucial roles in MN and V2a-IN specification by associating with NLI, we predict that mutant Ssdp1/2 (such as Ssdp1ΔN100, which lacks NLI-binding but retains the autonomous transactivation potential and interferes with transcriptional activation by the MN-hexamer and V2-tetramer; Fig. 3B,C) would inhibit MN and V2a-IN differentiation. Thus, we electroporated Gal4DBD-Ssdp1ΔN100 along with UAS:GFP into the chick neural tube to label cells in which Ssdp1ΔN100 is expressed and displays autonomous transactivation with GFP. Ssdp1ΔN100 profoundly reduced the formation of MNs and V2a-INs (Fig. 5G). Most Isl2⁺ MNs and Chx10⁺ V2a-INs in the Ssdp1ΔN100 electroporated sides appear to be non-transfected, as indicated by their lack of GFP expression (Fig. 5G), suggesting that Ssdp1ΔN100-expressing cells are resistant to differentiation to MNs and V2a-INs. To determine whether Ssdp1ΔN100 impairs MN differentiation in a MN cell-autonomous manner, we drove the expression of Ssdp1ΔN100 specifically in differentiating MNs by electroporating a Hb9-MNe:Gal4 driver with UAS:Ssdp1ΔN100 (Fig. 5H). Ssdp1ΔN100 expression in differentiating MNs greatly compromised the development of Isl2⁺ MNs, whereas it did not decrease the expression level of NLI (Fig. 5H). By contrast, GFP expression in MNs by electroporating Hb9-MNe:Gal4 driver and UAS:GFP constructs had little effect (Fig. S5). These results suggest that recruitment of Ssdp1/2 to LIM complexes is needed for the specification of MNs and V2a-INs during spinal cord development.

Our data implicate NLI as a docking site for Ssdp1/2, enabling transactivation by the MN-hexamer and V2-tetramer, in addition to its known role for the assembly of these LIM complexes. To test the specific role of NLI in recruiting Ssdps to LIM complexes for neuronal subtype specification, we sought NLI mutants that are defective in association with Ssdps but maintain the ability to dimerize and interact with LIM-HD factors. Evolutionarily conserved residues within the LCCD of NLI (the Ssdp-interacting domain) were mutated to alanines (Fig. 6A). The interaction between these NLI mutants and Ssdp1 was tested using GST pull-down assays. Among these mutants, NLIm (mutation of ILEPMQ residues to AAAAAA in NLI, Fig. 6A) exhibited severely impaired Ssdp1-binding, while it still interacted with NLI and the LIM domains of Isl1 and Lhx3 (Fig. 6B, Fig. S6A; data not shown).

Next, to create MN-hexamer and V2-tetramer analogs, which are defective specifically in their ability to associate with Ssdps and do not require the contribution of endogenous NLI in their formation, we made use of the previously defined chimeric molecules that mimic the structure and neuronal specification function of the MN-hexamer and V2-tetramer (Thaler et al., 2002). These were NLI-Lhx3, a fusion of NLI residues 1-298, containing both a self-dimerization domain and an Ssdp-interacting LCCD, to the DNA-binding homeodomain of Lhx3; and NLI-Isl1, a fusion of NLI residues 1-298 and the homeodomain of Isl1, which binds Lhx3 and assembles the MN-hexamer without incorporating endogenous NLI (Fig. 1A and Fig. 6C). We generated NLIm-Isl1 and NLIm-Lhx3, which carry mutations of ILEPMQ in the LCCD of NLI to alanines, as depicted in Fig. 6C. We then assessed their interaction with Ssdp1 using coimmunoprecipitation analyses in P19 cells (Fig. 6D). NLI-Isl1 and NLIm-Isl1 interacted with Lhx3 (data not shown), suggesting that both are capable of forming the MN-hexamer (Fig. 1A). As expected, NLIm-Isl1 and NLIm-Lhx3 failed to associate with Ssdp1 in cells, unlike NLI-Isl1 and NLI-Lhx3, which readily bind Ssdp1 (Fig. 6D). These data establish that NLIm-Isl1 and NLIm-Lhx3 are able to form the MN-hexamer and V2-tetramer, respectively, but are incapable of recruiting Ssdp1/2.

To monitor the transactivation potential of NLI-Lhx3 and NLI-Isl1 to activate the V2-tetramer and MN-hexamer target genes, respectively, we performed luciferase reporter assays using TeRE:LUC and HxRE:LUC in P19 cells. NLI-Lhx3 potently activated TeRE:LUC, which is further enhanced by coexpressed Ssdp1 or Ssdp2 (Fig. 6E). By contrast, NLIm-Lhx3 was ineffective in stimulating TeRE:LUC and did not efficiently respond to Ssdp1/2 (Fig. 6E). The combined expression of NLI-Isl1 and Lhx3 activated HxRE:LUC (Fig. 6F), as predicted from their assembly of the MN-hexamer complex. This is augmented by Ssdp1 and Ssdp2 but not by Ssdp1-m2 mutant, which is impaired in NLI interaction (Fig. 6F and Fig. 2G). NLIm-Isl1 was a much weaker activator compared with NLI-Isl1, when coexpressed with Lhx3 (Fig. 6F). These results demonstrate that the recruitment of Ssdp1/2 via NLI is necessary for the V2-tetramer and MN-hexamer to activate their target elements.

To investigate requirement for association of Ssdp1/2 with the MN-hexamer and V2-tetramer in MN and V2a-IN differentiation, we examined whether Ssdp1/2 binding is a prerequisite for recognition of the cognate genomic target by the MN-hexamer and V2-tetramer. ChIP assays in P19 cells revealed that both NLI-Isl1 and NLIm-Isl1, when coexpressed with Lhx3, were recruited to Hb9-MNe (Fig. 7A), establishing that their binding to MN-hexamer-response elements is intact. Likewise, both NLI-Lhx3 and NLIm-Lhx3 occupied Chx10-TeRE, a V2-tetramer-binding genomic element, in P19 cells (Fig. 7B). Thus, the recruitment of Ssdp1/2 is not needed for the MN-hexamer and V2-tetramer to bind their response elements in chromatin.

Next, we monitored MN differentiation triggered by the MN-hexamer in multipotent P19 mouse embryonic cells, which acquire MN phenotypes when transfected with Isl1, Lhx3 and Ngn2 (Lee and Pfaff, 2003). Whereas cotransfection of NLI-Isl1, Lhx3 and Ngn2 triggered expression of the MN marker Hb9, expression of NLIm-Isl1, Lhx3 and Ngn2 failed to induce MN differentiation (Fig. 7C). NLI-Isl1 and NLIm-Isl1 were expressed at comparable levels (Fig. S6B). These results indicate that the recruitment of Ssdp1/2 to the MN-hexamer is essential for MN differentiation mediated by the MN-hexamer. Consistently, coexpression of Ssdp1 with NLI-Isl1, Lhx3 and Ngn2 increased the efficiency of MN induction in P19 cells by ∼4.8-fold (Fig. 7C,D). Reduction of Ssdp1 and Ssdp2 expression with sh-Ssdp1 and sh-Ssdp2, or blocking the function of endogenous Ssdp1/2 with Ssdp1ΔN100 significantly compromised MN differentiation (Fig. 7C,D). To further evaluate the importance of the interactions of NLI and Ssdp1/2 for MN and V2a-IN differentiation triggered by the MN-hexamer and V2-tetramer, respectively, in developing embryos, we assessed the potential of NLI-Isl1, NLIm-Isl1, NLI-Lhx3 and NLIm-Lhx3 to direct neuronal subtype specification in the neural tube using chick embryo electroporation. Whereas coexpression of NLI-Isl1 and Lhx3 induced formation of ectopic MNs in the dorsal spinal cord, coinjection of NLIm-Isl1 and Lhx3 did not (Fig. 7E). Similarly, NLI-Lhx3, but not NLIm-Lhx3, readily triggered ectopic V2a-interneuron formation (Fig. 7F). These results demonstrate that NLI-mediated recruitment of Ssdp1/2 is essential for the MN-hexamer and V2-tetramer to trigger the generation of MNs and V2a-INs during spinal cord development.

We considered the possibility that the impact on dynamic chromatin architecture of mobilization of chromatin remodeling enzymes may underlie the potent coactivator function of Ssdps. To investigate whether the MN-hexamer induces transcriptionally active chromatin in its target genes, we assessed two prominent chromatin marks associated with active transcription, acetylation of histone H3 and H3K4me3. We examined the chromatin modifications in Hb9-MNe, a direct binding site for the MN-hexamer, as well as the early transcribed region of Hb9 gene (5′-Hb9) upon expression of the MN-hexamer. Chromatin fragments were immunopurified using antibodies against acetylated-H3 and H3K4me3 in P19 cells transfected with empty vector, NLI-Isl1 plus Lhx3 or NLIm-Isl1 plus Lhx3, followed by PCR analyses to monitor the presence of Hb9-MNe and 5′-Hb9. Intriguingly, expression of NLI-Isl1 and Lhx3, which assemble the MN-hexamer and trigger MN differentiation, further induced H3 acetylation and H3K4me3 in both Hb9-MNe and 5′-Hb9 regions, whereas expression of NLIm-Isl1 and Lhx3, despite their ability to bind Hb9-MNe (Fig. 7A), was ineffective (Fig. 8A). Also, interference of endogenous Ssdp1/2 function by Ssdp1ΔN100 severely attenuated H3 acetylation and H3K4me3 triggered by the MN-hexamer at Hb9-MNe and 5′-Hb9 (Fig. 8B). These data demonstrate that recruitment of Ssdp1/2 is required for the MN-hexamer to trigger a transcriptionally active chromatin state in MN genes, revealing a novel mechanistic basis for the coactivator function of Ssdp1/2 in neuronal subtype specification.

Next, we examined whether Ssdp1/2 recruit chromatin-modifying enzymes for H3 acetylation and H3K4me3 to the MN-hexamer target MN genes. CREB-binding protein (CBP) is a histone acetyltransferase enzyme and ASC-2 (also known as Ncoa6) is a component of histone methyltransferase complexes containing MLL3/4 (Goo et al., 2003; Lee et al., 2006; Marmorstein, 2001). ChIP assays with anti-CBP or anti-ASC-2 antibodies revealed that CBP and ASC-2 are recruited to Hb9-MNe when Isl1 and Lhx3 are expressed in P19 cells (Fig. 8C), concomitant with the induction of H3 acetylation and H3K4me3 (Fig. 8A). Remarkably, downregulation of Ssdp1/2 by sh-Ssdp1/2 or inhibition of Ssdp1/2 function by Ssdp1ΔN100 compromised the binding of CBP and ASC-2 to Hb9-MNe, suggesting that Ssdp1/2 mediates the recruitment of histone modifying complexes containing CBP and ASC-2 to Hb9-MNe. To further test a role of Ssdp1/2 in establishing transcriptionally active chromatin, we examined the status of H3 acetylation in the early transcribed region of Hb9 and Chx10 genes in Ssdp1/2 mutant embryos using ChIP analyses. The level of H3 acetylation was significantly lowered in embryos defective in Ssdp1 and Ssdp2 activity (Fig. 8D). These data suggest that Ssdp1/2 play an important role in establishing histone H3 acetylation and H3K4me3 at target genes of the MN-hexamer and V2-tetramer by recruiting histone-modifying activities.

Despite overwhelming evidence that LIM-HD factors play crucial roles in development of invertebrates and vertebrates, the molecular mechanisms by which they activate gene transcription during development remain poorly defined. To address this issue, we focused on how Ssdp1/2 regulate gene expression for the differentiation of MNs and V2a-INs as transcriptional coactivators of two LIM complexes, the MN-hexamer and V2-tetramer, respectively. Our study reveals an Ssdp1/2-mediated mechanistic link between LIM complexes and chromatin remodeling that operates during development.

While various structure and functional analyses demonstrated the essential role of NLI as a key nucleating molecule to assemble LIM-HD complexes, it has been unclear whether NLI directly contributes to gene transcription. We found that NLI is required for the transactivation function of the MN-hexamer and V2-tetramer by serving as a docking site for Ssdp1/2, which subsequently induce a transcriptionally active chromatin state. It is noteworthy that Drosophila NLI was genetically identified as a chromosomal factor that promotes the communication between remote enhancers and promoters, presumably by facilitating formation of a chromatin structure that brings them closer (Morcillo et al., 1997). Similarly, NLI appears to promote chromatin loop formation during mammalian erythroid differentiation, facilitating long-range gene activation (Song et al., 2007). Thus, it will be interesting to test whether the function of NLI as a facilitator of interaction between enhancer and promoter involves the chromatin remodeling activity mobilized by Ssdp1/2. NLI has also been suggested to recruit Osa, a component of Brahma chromatin remodeling complexes, to the Pan/NLI complex in Drosophila (Heitzler et al., 2003). In this context, Osa negatively regulates Pan/NLI-mediated proneural achaete/scute expression and neural development. Thus, NLI may have an ability to integrate the action of multiple chromatin remodelers to NLI complexes during development.

Ssdps were isolated as evolutionarily conserved NLI-interacting partners in yeast two-hybrid screening and purification of NLI complexes (Chen et al., 2002; Meier et al., 2006; van Meyel et al., 2003). The genetic interactions between Ssdp and NLI in Drosophila, Xenopus and mouse (Chen et al., 2002; Enkhmandakh et al., 2006; Nishioka et al., 2005; van Meyel et al., 2003) indicate that Ssdp and NLI coregulate expression of shared target genes during development. Ssdps also interact genetically with LIM-HD genes Apterous and Lim1 (Chen et al., 2002; Enkhmandakh et al., 2006; Nishioka et al., 2005; van Meyel et al., 2003). In addition, Ssdp1/2 associate with a transcription complex consisting of NLI, Tal1/SCL, E47, GATA1 and LMO2, which regulates erythroid gene expression and differentiation (Xu et al., 2007). These suggest that Ssdps mediate the developmental action of NLI-mediated LIM complexes.

Biochemical, phylogenetic and genetic studies suggest that NLI and Ssdps belong to a tightly associated functional module that is conserved among multiple species from C. elegans to human (Chen et al., 2002; Enkhmandakh et al., 2006; Nishioka et al., 2005; van Meyel et al., 2003). Thus, it is tempting to speculate that NLI and Ssdps have evolved to take discrete functions within this module; NLI assembles multi-protein complexes via its ability to interact with multiple transcription factors, whereas Ssdps recruit chromatin remodelers to the complex, enabling the transcriptional activity of the NLI complexes. This model predicts that disrupting the interaction between NLI and Ssdps affects a variety of NLI-dependent developmental systems.

Our studies reveal a novel mechanism for the developmental function of Ssdps. Ssdp1/2 associate with the MN-hexamer and V2-tetramer and elicit their transcriptional activity by inducing an active chromatin state, thus upregulating MN and V2a-IN genes. Considering that Ssdp1/2 lack any obvious histone-modifying enzymatic activity, H3 acetylation and H3K4me3 mediated by Ssdp1/2 is likely to occur though the association between Ssdp1/2 and other histone modifiers. ChIP assays reveal that Ssdp1/2 are involved in recruiting CBP and ASC-2, components of histone-modifying complexes, to the MN-hexamer target MN genes (Fig. 8C). The dominant negative function of Ssdp1ΔN100 in histone modification of Hb9 gene as well as MN differentiation might involve its ability to titrate out one or more of these chromatin modifiers from the MN-hexamer target genes. Future studies are needed to understand how Ssdp1/2 bring histone modifiers to target genes of NLI/LIM complexes.

Finally, it should be noted that V0-INs marked by Evx1 (Fig. S2) and dorsal INs marked by Isl1 (data not shown) are also decreased by si-Ssdp1/2 in chick neural tube, and that the number of dorsal INs marked by Isl1 is reduced in Ssdp1/2 mutant mice (Fig. S3B,C). These results raise an interesting possibility that Ssdp1/2 also function as essential transcriptional coactivators of other LIM complexes and non-LIM-HD transcription factor(s) involved with fate specification of other neuronal cell types in the developing spinal cord.

In summary, our results demonstrate that Ssdps play crucial roles in MN and V2a-IN specification by enabling the transcriptional activity of LIM complexes through their ability to induce an active chromatin state, a strategy that is likely to be widely used by other NLI-containing complexes that are involved with various cell fate determinations.

Rat Isl1; mouse Lhx3, Ngn2, NLI, Ssdp1 and Ssdp2; human SSDP2; and fusions Isl1-Lhx3, NLI-Isl1 and NLI-Lhx3 were cloned into pCS2, pcDNA3 (Invitrogen) containing HA, Flag or Myc-epitope tags, pGEX4T-1 (Amersham), pCMX-Gal4 or pM (Clontech), pLexA and/or pJG4-5-B42 vectors. Ssdp1 deletion vectors encode partial sequences of mouse Ssdp1 as follows: Ssdp1ΔN100 (aa 101-361), Ssdp1-N150 (aa 1-150), Ssdp1-N100 (aa 1-100), Ssdp1-1-50 (aa 1-50), Ssdp1-49-100 (aa 49-100). Isl1-Lhx3 was generated by fusing full-length Isl1 and Lhx3 via a flexible linker (Lee et al., 2008). NLI-Isl1 and NLI-Lhx3 were generated by fusing the N-terminus of NLI (aa 1-298) with the homeodomain of Isl1 (aa 111-349) or Lhx3 (aa 152-401) (Fig. 6C, Thaler et al., 2002). Ssdp1 and NLI mutants were generated using PCR-based mutagenesis method and verified by sequencing. Hb9p-MNe:Gal4 was generated by subcloning 1.2 kb mouse genomic sequence containing Hb9-MNe (Lee et al., 2004) into a reporter in which the thymidine kinase minimal promoter drives the expression of full-length Gal4 transcription factor. UAS:GFP and UAS:HA-Ssdp1ΔN100 vectors encode GFP and HA-tagged Ssdp1ΔN100 driven by Gal4-binding enhancer UAS. MNe:LUC, HxRE:LUC, TeRE:LUC and pSilencer-Ssdp2 (sh-Ssdp2), pSilencer-Ssdp3 (sh-Ssdp1) and pSilencer-EGFP were as described (Lee et al., 2008,, 2004; Lee and Pfaff, 2003; Xu et al., 2007).

The antibodies used for immunohistochemistry, immunoprecipitation and immunoblotting assays are: rabbit anti-Ssdp1 and anti-Ssdp2 (1:1000, gift from Dr. L. Nagarajan; Xu et al., 2007); guinea pig and rabbit anti-Lhx3 (1:2500, Sharma et al., 1998); guinea pig anti-Chx10 (1:2500); rabbit anti-Hb9 (1:5000, Thaler et al., 1999); rabbit anti-NLI (1:1000, Thaler et al., 2002); rabbit anti-Isl1 (1:2500, Tsuchida et al., 1994); mouse anti-Mnr2/Hb9 (1:200, 5C10, DSHB); mouse anti-Isl2 (1:50, 4H9, DSHB); goat anti-Sox2 (1:500, Santa Cruz, sc17320); rabbit anti-CBP (1:500, Santa Cruz, sc-369); rabbit anti-ASC-2 (1:1000, Bethyl, A300-410A); mouse anti-HA (1:2500, Covance, MMS-101R); mouse anti-Flag (1:2500, Sigma, F3165); mouse anti-myc (1:2000, Upstate, DSHB); rabbit anti-H3Ac (1:2000, Upstate, 06-599); rabbit anti-H3K4me3 (1:2000, Abcam, 8580); and mouse, rabbit IgG (Santa Cruz, sc-2025, sc-2027). mouse, rabbit IgG (Santa Cruz, sc-2025, sc-2027).

All mouse works were performed under an approved protocol by the Institutional Animal Care and Use Committee of the Oregon Health and Science University. The generation of headshrinker (hsk) disrupting Ssdp1 has been described previously (Nishioka et al., 2005). Only Ssdp1^hsk/hsk embryos whose body size were comparable to wild-type littermates were used for analyses. To generate Ssdp2 null allele, a neomycin cassette was cloned into the 7 kb Ssdp2 genomic clone encompassing 3.5 kb upstream and downstream of the exon of Ssdp2, replacing the first coding exon of Ssdp2. The targeting construct was electroporated into mouse embryonic stem cells and homologous recombinants were selected for generating chimeric animals using standard techniques. The genotype of animals was monitored using the following PCR primers: 5′-GCT TGG GTG GAG AGG CTA TT and 5′-ATA CTT TCT CGG CAG GAG CA to detect a 300 nucleotide mutant allele of Ssdp2; and 5′-GGA GGA GAC TGC CGC TTA G and 5′-CAC CTG TCA ACC CAT CAC AG to detect a 200 nucleotide product for the wild-type allele of Ssdp2.

These experiments were carried out as previously described (Lee et al., 1998).

ChIP was performed as described (Cho et al., 2014). PCR primers for mouse Hb9-MNe and Chx10-TeRE were as described previously (Lee et al., 2008; Lee and Pfaff, 2003). Primers for mouse 5′-Hb9 and mouse 5′-Chx10 are: 5′-Hb9 forward, 5′-ATC GCA GTA ACA ATA CCG GC; 5′-Hb9 reverse, 5′-CGG GAC CAG ATA CTG TAG TC; 5′-Chx10 forward, 5′-GGC TCC AGA GCA TTA GAC AC; and 5′-Chx10 reverse, 5′-TGT TCA AGC CTA GGA TCT CC.

These experiments were performed as described (Yeo et al., 2005).

In ovo electroporations, immunohistochemistry and in situ hybridization were performed as described (Lee and Pfaff, 2003; Thaler et al., 2002). Co-electroporation of two plasmids typically resulted in greater than 90% of cells coexpressing both plasmids. More than 15 embryos were analyzed for each electroporation experiment and over 90% of embryos produced the identical/similar results along the rostral-caudal spinal cord. The images in figures are representative of more than 15 embryos. Sense strand sequences of duplex siRNA for chick Ssdp1 (two distinct sets) or Ssdp2 are: 5′-GUG ACA ACA UCU ACA CAA U for Ssdp1, 5′-AGA ACA UCA CGC UGG GAG A for Ssdp1, and 5′-CUA UGG AGG UGC AAU GAG A for Ssdp2. For ISH, digoxigenin-labeled riboprobes complementary to mouse Ssdp1, Ssdp2 and Ssdp4 were synthesized according to the supplier's protocol (Roche).

P19 and HEK293 cells were used in this study, as previously described (Lee and Pfaff, 2003). Transient transfections were performed using Lipofectamine 2000 (Invitrogen) or Superfect (Qiagen). Luciferase reporter data are shown in relative activation fold (mean±s.d. of at least three experiments). For cell differentiation assays, P19 cells were seeded onto chamber slides (Falcon) and fixed and immunostained 3 days after transfection.

Statistical differences were determined by two-tailed Student’s t-test. Statistical significance is displayed as *P<0.05, **P<0.01, ***P<0.001 or ****P<0.0001.

We are grateful to Samuel Pfaff and Karen Lettieri for help to generate Ssdp2 mutant mice; Lalitha Nagarajan for Ssdp1 and Ssdp2 antibodies; Stephen Brandt for sh-Ssdp1/2 vectors; Hiroshi Sasaki for Ssdp1^hsk mice; Hilda Puente for her excellent technical assistance; Ming Tsai and Kaumudi Joshi for comments on the manuscript.