Recombineering Hunchback identifies two conserved domains required to maintain neuroblast competence and specify early-born neuronal identity Free

Small pools of multipotent neural progenitors give rise to a large number of neurons and glia to allow proper assembly of a functional nervous system (Cepko, 1999; Doe and Skeath, 1996; Rapaport et al., 2001; Walsh and Reid, 1995). However, as progenitors change over time to accommodate the production of different tissues, they also undergo a progressive restriction and lose their competence to produce the full assortment of cell types (Desai and McConnell, 2000; Rapaport et al., 2001). The ability to maintain progenitors in their early competent states yields the capacity to generate any desired tissue for use in future cell therapy applications. In recent years, substantial progress has been made in the identification of the factors involved in regulating progenitor competence, bringing with it the need to understand the molecular basis of their function.

The identification of genes that regulate neural progenitor competence in vertebrates and insects has provided an entry point for investigating the molecular mechanism of progenitor competence (Elliott et al., 2008; Hanashima et al., 2004; Isshiki et al., 2001; Novotny et al., 2002). The zinc-finger transcription factor Ikaros (Ik) is both necessary and sufficient to specify early progenitor competence, leading to the production of early-born cell types in the mouse retina (Elliott et al., 2008). This function of Ikaros mimics that of its Drosophila ortholog, the zinc-finger transcription factor Hunchback (Hb), which has also been shown to promote early progenitor competence during Drosophila neurogenesis (Isshiki et al., 2001; Novotny et al., 2002). Although it is known that Hb and Ik can maintain progenitor competence in the nervous system, the molecular mechanism by which they act remains unknown.

The Drosophila embryonic CNS is an excellent system in which to investigate Hb-dependent regulation of neural progenitor competence. Individual neural progenitors (neuroblasts) can be uniquely identified based on their position and pattern of gene expression (Broadus et al., 1995). Each of the 30 neuroblasts in a hemisegment divides to produce an invariant order of neurons and/or glia, the identity of which is determined by the sequential expression of transcription factors: Hb → Krupple (Kr) → Pdm1 (Nubbin – FlyBase)/Pdm2 (henceforth Pdm) → Castor (Cas) → Grainy head (Grh) (Baumgardt et al., 2009; Bossing et al., 1996; Isshiki et al., 2001; Maurange et al., 2008). Hb is expressed early on in many neuroblasts and is required for the specification of the first-born cell identity, or first temporal identity, in those lineages (Isshiki et al., 2001; Novotny et al., 2002). In addition, Hb can also confer the early competent state to many neuroblast lineages (Cleary and Doe, 2006; Pearson and Doe, 2003). This is achieved by maintaining Hb expression in neuroblasts throughout neurogenesis, or by reintroducing Hb into neuroblasts after its normal expression window. The ectopic Hb expression results in the specification of extra early-born progeny.

Interestingly, the ability of Hb to specify and extend the early competence window declines over time (Cleary and Doe, 2006; Pearson and Doe, 2003). When Hb is reintroduced into NB7-1 at progressively later time points, its ability to specify ectopic U1/U2 neurons is greatly reduced. Eventually, Hb is unable to specify early-born cells after the fifth neuroblast division. This raises two interesting questions regarding neuroblast competence. First, how does Hb regulate gene expression to maintain early neuroblast competence? Second, why does this ability decline over time?

Hb regulates gene expression via multiple well-characterized modes during the formation of the Drosophila body plan; however, little is known about its modes of function in the CNS. In the cellular blastoderm, the Hb protein gradient initiates and establishes the spatial expression domains of the gap genes Kr, knirps (kni) and giant (gt). Rigorous genetic and molecular analyses have shown that Hb acts as a concentration-dependent transcriptional activator and repressor of gene expression during embryonic segmentation (Berman et al., 2002; Hoch et al., 1991; Pankratz et al., 1992; Rivera-Pomar et al., 1995; Schulz and Tautz, 1994; Struhl et al., 1992). Furthermore, Hb can induce permanent repression of target genes through its interaction with Mi2 and the recruitment of Polycomb complex proteins (Kehle et al., 1998). As a result, the multifunctional Hb protein is a potent regulator of gene expression in the early embryo. Here, we extend the analysis of Hb-mediated gene regulation to include its role in maintaining neuroblast competence during nervous system development.

We generated the VP16::Hb chimera by PCR amplifying the VP16 activation domain (Lai and Lee, 2006) using primers with a 3′ tail that contained hb 5′ sequence, and ligating to PCR-amplified full-length hb coding sequence. For primer sequences, see Table S1 in the supplementary material. The chimeric gene was verified by sequencing, cloned into the pUAST vector (Brand and Perrimon, 1993) and transgenic flies produced (GenetiVision, Houston, TX, USA).

We generated hb genes deleted for the six previously described conserved domains (R. Sommer, PhD thesis, University of Munich, 1992) (Tautz et al., 1987), as well as for two additional domains (B′ and E) that we identified as conserved in at least eight sequenced Drosophila species using EvoPrinter (Odenwald et al., 2005). Each hb deletion construct (except the D domain deletion) was generated using recombineering by targeted insertion and replacement of the galK expression cassette (Warming et al., 2005). galK targeting cassettes were prepared by PCR amplification of the galK expression cassette using primers with homology to hb. To insert galK, SW102 cells containing BAC clone BACR01F13 were electroporated with the appropriate targeting cassette and plated on minimal medium with galactose and chloramphenicol. To replace galK, SW102 cells were electroporated with the appropriate replacement cassette and plated on minimal medium with glycerol, 2-deoxy-galactose and chloramphenicol. The replacement cassette for epitope tagging (3×FLAG::3×HA) was prepared by PCR amplification using homology primers. For primer sequences, see Table S1 in the supplementary material. The D domain deletion was generated by two-step PCR. Each construct was sequenced to confirm that Hb was modified correctly. Deletions were cloned into a pUAST(attB) vector (Bischof et al., 2007) and sent to GenetiVision for injections into flies carrying the attP40 docking site on chromosome 2 (Markstein et al., 2008). In addition to the deletions, we also generated flies carrying the same epitope-tagged wild-type Hb in the attP40 locus as a standard control. The fly stocks generated are described below.

The following pre-existing fly stocks were used: yw (wild type); v32a-gal4 for ubiquitous embryonic expression (Siegrist and Doe, 2005); UAS-hb (Wimmer et al., 2000); engrailed-gal4 for expression in the posterior compartment of each segment (Harrison et al., 1995; Isshiki et al., 2001; Pearson and Doe, 2003); worniu-gal4 for expression in neuroblasts (Albertson et al., 2004); UAS-HA is UAS-HA::UPRT (Miller et al., 2009), and this transgene was used as a UAS control so that each misexpression experiment had two UAS transgenes; it does not change the number of U neurons when expressed alone.

The following fly stocks were generated in this work:

UAS-VP16::hb on chromosomes 2 and 3;

UAS-VP16::hb; hb^FB hb^P1/TM3 ftz-lacZ [hb^FB is a null mutant, hb^P1 is a segmentation rescue construct (Hulskamp et al., 1994; Isshiki et al., 2001)];

UAS-VP16::hb; Df(2L)ED773/CyO ftz-lacZ [Df(2L)ED773 removes both pdm1 and pdm2 (Grosskortenhaus et al., 2006)];

UAS-hb^wild-type/CyO and UAS-hb^wild-type/CyO; hb^FB hb^P1/TM3 Ubx-lacZ;

UAS-hb^ΔAB/CyO and UAS-hb^ΔAB/CyO; hb^FB hb^P1/TM3 Ubx-lacZ;

UAS-hb^ΔB′/CyO and UAS-hb^ΔB′/CyO; hb^FB hb^P1/TM3 Ubx-lacZ;

UAS-hb^ΔDBD/CyO and UAS-hb^ΔDBD/CyO; hb^FB hb^P1/TM3 Ubx-lacZ;

UAS-hb^ΔC/CyO and UAS-hb^ΔC/CyO; hb^FB hb^P1/TM3 Ubx-lacZ;

UAS-hb^ΔD/CyO and UAS-hb^ΔD/CyO; hb^FB hb^P1/TM3 Ubx-lacZ;

UAS-hb^ΔE/CyO and UAS-hb^ΔE/CyO; hb^FB hb^P1/TM3 Ubx-lacZ;

UAS-hb^ΔDMZ/CyO and UAS-hb^ΔDMZ/CyO; hb^FB hb^P1/TM3 Ubx-lacZ; and engrailed-gal4/CyO; hb^FB hb^P1/TM3 Ubx-lacZ.

Antibody staining was performed according to standard methods. Polyclonal antiserum against Zfh2 was made by expressing a GST fusion protein containing amino acids 1641-2149 of the Zfh2 protein [pZFH-2c (Lai et al., 1991)], and purified proteins were sent to Alpha Diagnostic International (San Antonia, TX, USA) for injection into rats. Primary antibodies used were: rat Zfh2, 1:400 (this work; M. Lundell, UT San Antonio, USA); mouse Eve 2B8, 1:10 (Patel et al., 1994); guinea pig Eve, 1:1000, rat Eve, 1:1000 and guinea pig Kr, 1:800 (Kosman et al., 1998); rabbit Hb, 1:200 (Tran and Doe, 2008); rabbit Kr, 1:500 (Gaul et al., 1987); rat Pdm2, 1:10 (Grosskortenhaus et al., 2006); rabbit Cas, 1:2000 (Kambadur et al., 1998); mouse En 4D9, 1:5 (Patel et al., 1989); mouse β-galactosidase, 1:500 (Promega, Madison, WI, USA); rabbit β-galactosidase, 1:1000 (MP Biomedicals, Solon, OH, USA); mouse HA, 1:500 (Roche, Palo Alto, CA, USA); and rabbit HA, 1:1000 (Sigma, St Louis, MO, USA). Secondary antibodies were purchased conjugated to Alexa 488, Rhodamine RedX, Cy5 (Jackson ImmunoResearch, West Grove, PA, USA), biotin (Vector Labs, Burlingame, CA, USA) or alkaline phosphatase (Southern Biotechnology, Birmingham, AL, USA) and used at 1:400.

Confocal image stacks were collected using a Leica SP2 or Bio-Rad Radiance 2000 confocal microscope and displayed as two-dimensional projections. Images of histochemical preparations were acquired using a Zeiss AxioCam HRc camera on an Axioplan microscope. Two-tailed t-tests were used to determine significance in the differences in cell number.

To address the molecular mechanism underlying Hb maintenance of neuroblast competence, we began by examining whether Hb achieves this role through the transcriptional activation or repression of target genes. Because no transcriptional activation or repression domains have been identified within Hb, we began by converting Hb to act solely as a transcriptional activator by fusing the VP16 transactivation domain (Triezenberg et al., 1988) to the N-terminus of full-length Hb (Fig. 1A). We then tested whether VP16::Hb could induce competence or first-born temporal identity.

Before assaying the effects of VP16::Hb in the CNS, we first examined the ability of VP16::Hb to recognize and activate known Hb direct target genes in the early embryo. In the embryonic blastoderm, Hb is a direct activator of the gap gene Kr (Hoch et al., 1991). Expression of Hb throughout the embryo using a heat-shock promoter can activate Kr and expand the Kr central gap domain to the posterior of the embryo (Hoch et al., 1991; Hulskamp et al., 1990; Struhl et al., 1992). We expressed wild-type Hb throughout the embryo using the Gal4/UAS expression system (Brand and Perrimon, 1993) with a maternal Gal4 driver (v32a-gal4 UAS-hb) and were able to reproduce this expansion of the Kr central gap domain (Fig. 1B). Next, we examined v32a-gal4 UAS-VP16::hb embryos and found a dramatic increase in Kr expression throughout the embryo (Fig. 1B), indicating that VP16::Hb is a potent activator of Kr gene expression.

We further tested the capacity of VP16::Hb as an activator by examining its effect on genes that are normally directly repressed by Hb. Hb expression in the posterior regions of the embryonic blastoderm directly represses the gap genes kni and gt (Fig. 1C,D) (Berman et al., 2002; Berman et al., 2004; Pankratz et al., 1992; Pelegri and Lehmann, 1994). However, VP16::Hb overexpression activated both kni and gt, expanding their domains anteriorly and posteriorly into the domains of their respective repressors (Fig. 1C,D). We conclude that VP16::Hb is also sufficient to activate genes that are normally repressed by Hb (Fig. 1E). The strong activation and expansion of the Kr, kni and gt gap domains indicates that VP16::Hb acts as a potent transcriptional activator that overcomes or eliminates the normal transcriptional repression function of Hb. Taken together, we conclude that VP16::Hb has the capacity to recognize and activate Hb target genes whether they are normally activated or repressed by Hb in the early embryo.

Because VP16::Hb acts as a strong activator of Hb direct target genes in the early embryo, we next examined the expression of Hb targets in the CNS in response to VP16::Hb expression. Previous studies suggest that Hb regulates its own transcription in the blastoderm, but not in the CNS (Grosskortenhaus et al., 2005; Treisman and Desplan, 1989). However, it is possible that the VP16::Hb chimera could activate endogenous hb transcription via the VP16 activation domain, leading to cells that contain both a transcription-activating Hb protein (VP16::Hb) and a potential transcription-repressing Hb protein (endogenous Hb). To test whether VP16::Hb can activate endogenous hb transcription, we performed in situ hybridizations against the 3′ UTR of endogenous hb mRNA. In engrailed-gal4 UAS-hb embryos, we found that Hb could not induce its own transcription in neuroblasts (Fig. 2A,B), consistent with previous findings (Grosskortenhaus et al., 2005). By contrast, engrailed-gal4 UAS-VP16::hb embryos showed strong activation of endogenous hb transcription in neuroblasts (Fig. 2C). We conclude that VP16::Hb can activate endogenous hb transcription in the CNS.

We next examined whether VP16::Hb can activate pdm (pdm1 and pdm2), a gene normally repressed by Hb through well-characterized Hb binding sites in its CNS enhancer element (Kambadur et al., 1998). In stage 15 embryos, only a few neuroblasts expressed Pdm (Fig. 2D), and there was minimal change in Pdm following the expression of wild-type Hb protein (Fig. 2E), probably because most neuroblasts can no longer respond to Hb at this stage (Cleary and Doe, 2006). However, the overexpression of VP16::Hb in neuroblasts resulted in the upregulation of Pdm (Fig. 2F). We conclude that VP16::Hb can activate the direct target pdm in the CNS.

Having shown that VP16::Hb can activate Hb direct target genes, we extended our analyses of VP16::Hb activity to all other known Hb targets in the CNS (Fig. 2G). Hb is known to activate Kr and repress zfh2, cut, runt and cas (Fig. 2H) (Grosskortenhaus et al., 2006; Isshiki et al., 2001; Kambadur et al., 1998), although it is not known whether Hb acts directly or indirectly to regulate the expression of these genes. We overexpressed VP16::Hb in a hb mutant background and found that it can activate Kr in neurons, similar to wild-type Hb (Fig. 2I). Additionally, we found that VP16::Hb expression also results in the activation of the normally repressed target genes zfh2, cut, runt and cas in neurons (Fig. 2I). We conclude that VP16::Hb can activate all known Hb CNS targets, whether they are normally activated or repressed by Hb. Because Hb normally acts as a transcriptional repressor of most CNS targets (Fig. 2J), we suggest that this repression might play an essential role in maintaining neuroblast competence, a prediction that we test below.

Because VP16::Hb is a potent transcriptional activator with little ability to repress gene expression, we can use it to test whether Hb-mediated transcriptional activation of target genes is sufficient for maintaining neuroblast competence. Overexpression of wild-type Hb can extend neuroblast competence and the production of early-born neuronal cell types (Isshiki et al., 2001; Novotny et al., 2002; Tran and Doe, 2008). In NB7-1, Hb misexpression produces ~18-20 Eve⁺ U motoneurons. If Hb maintains early competence by acting solely as a transcriptional activator, then VP16::Hb should mimic Hb function and specify the same, or more, U neurons. In wild-type embryos, NB7-1 generates five Eve⁺ U motoneurons (Fig. 3A,A‴) (Pearson and Doe, 2003; Schmid et al., 1999). Overexpression of wild-type Hb in a hb mutant NB7-1 produced an average of 12 U neurons (range 6-18, n=100; Fig. 3A′). However, the overexpression of VP16::Hb in a hb mutant NB7-1 only generated an average of six U neurons (range 2-12, n=88; Fig. 3A″). We conclude that the constitutive activator VP16::Hb is not as good as wild-type Hb at maintaining the early competence window necessary for Eve⁺ U neuron production, and suggest that the repression of Hb downstream targets in neuroblasts might be essential to maintain early competence.

Because Hb normally represses downstream targets such as pdm and cas, and VP16::Hb can activate these targets, we tested whether the repression of these downstream genes is required to maintain neuroblast competence. First, we examined engrailed-gal4 UAS-hb, hb mutant embryos at early stage 12 and found that neuroblasts in the engrailed-gal4 domain expressed Kr but not pdm and cas (Fig. 3B) (Isshiki et al., 2001). By contrast, the majority of neuroblasts in the engrailed-gal4 domain of engrailed-gal4 UAS-VP16::hb, hb mutant embryos expressed Kr, pdm and cas (Fig. 3C). We conclude that VP16::Hb can ectopically induce pdm and cas expression in neuroblasts.

Next, we tested whether the co-expression of wild-type Hb plus Pdm, or Hb plus Cas, could lead to the same reduction in ectopic cells as seen in the VP16::Hb-overexpression experiments. In control engrailed-gal4 UAS-hb UAS-HA embryos, NB7-1 generated ~17 Eve⁺ U neurons (range 13-22, n=87; Fig. 3D). We next examined engrailed-gal4 UAS-hb UAS-pdm2 embryos for the total number of U neurons generated and found a large decrease in the number of ectopic U neurons compared with our control embryos (average of 9, range 3-14, n=70; Fig. 3D). engrailed-gal4 UAS-hb UAS-cas embryos showed a slight decrease in the number of U neurons generated (average of 15, range 8-20, n=118; Fig. 3D). We conclude that Hb normally represses downstream targets, such as pdm and cas, to maintain neuroblast competence.

Because Pdm2 is sufficient to block Hb-induced neuroblast competence, we tested whether Pdm2 was also necessary to terminate neuroblast competence. If pdm2 is the only factor activated by VP16::Hb that limits neuroblast competence, then the overexpression of VP16::Hb in a pdm mutant background should extend neuroblast competence and result in the generation of many Eve⁺ U neurons. By contrast, if VP16::Hb activates multiple factors, then VP16::Hb might not be able to extend competence or make large numbers of U neurons even in the absence of Pdm. Control prospero-gal4 UAS-hb embryos generated an average of nine Eve⁺ U neurons per hemisegment (range 5-16, n=100; Fig. 3E). By contrast, prospero-gal4 UAS-VP16::hb embryos generated on average only 5.6 Eve⁺ U neurons (range 4-9, n=100, P⪡0.001; Fig. 3E), presumably owing to the transcriptional activation of factors that limit competence. Strikingly, performing the same experiments in a pdm mutant embryo (lacking both pdm1 and pdm2) did not increase the number of Eve⁺ U neurons (average of 4.3, range 2-8, n=90, P⪡0.001; Fig. 3E). We conclude that Hb must normally repress multiple factors, in addition to pdm1/pdm2, to maintain early neuroblast competence.

Because Hb repression of target genes is essential for the maintenance of neuroblast competence, we next sought to identify the Hb protein domain(s) required for transcriptional repression. We generated a series of hb transgenes, each deleted for one or more of the eight evolutionarily conserved domains: the first group of zinc-finger domains bind DNA (DNA-binding domain, DBD); the second group of zinc-finger domains allow for Hb dimerization (dimerization domain, DMZ); four previously identified conserved domains (A, B, C and D) (Tautz et al., 1987); and two additional domains (B′ and E) that contain short DNA sequences conserved in at least eight sequenced Drosophila species (Fig. 4A,B). We deleted each of these domains in a series of UAS-HA::hb transgenes, which we placed in the same attP site on chromosome 2 so that the results of each transgene could be directly compared.

We tested each protein for its ability to activate the direct target Kr (Hoch et al., 1991) or repress the direct target pdm (Kambadur et al., 1998) within the CNS. Wild-type embryos at stage 11 are Kr⁻Pdm⁺ in the neuroblast layer (Grosskortenhaus et al., 2005; Isshiki et al., 2001; Tran and Doe, 2008). Overexpression of the full-length wild-type Hb protein resulted in the activation of Kr and repression of pdm (Fig. 4B,C), consistent with previous findings (Isshiki et al., 2001; Kambadur et al., 1998). Similarly, overexpression of Hb proteins lacking either the A+B domains, the B′ domain, or the E domain also activated Kr and repressed pdm (Fig. 4B,C), showing that none of these domains is required for transcriptional activation or repression. Overexpression of Hb proteins lacking the DBD or the C domain did not activate Kr or repress pdm (Fig. 4B,C), suggesting that they are non-functional, although these proteins were nuclear localized and exhibited similar stability to wild-type Hb protein in neuroblasts (Fig. 4C). By contrast, overexpression of Hb proteins lacking the D domain (which includes the Mi2 binding sites) failed to repress pdm, but still weakly activated Kr; an identical result was observed for Hb protein lacking the DMZ domain (Fig. 4B,C). The sparse and intermittent activation of Kr might be due to Pdm repression of Kr (Grosskortenhaus et al., 2006; Tran and Doe, 2008), which would be expected to counteract Hb-induced Kr activation.

We next tested whether the Hb^ΔD and Hb^ΔDMZ proteins fail to repress other known Hb target genes. Whereas wild-type Hb protein efficiently repressed zfh2, cut, runt and cas (Fig. 3C), the Hb^ΔD and Hb^ΔDMZ proteins failed to repress any of these genes (see Fig. S1 in the supplementary material). We conclude that both the D and DMZ domains are required for Hb-mediated transcriptional repression.

We showed above that overexpression of the constitutive transcriptional activator VP16::Hb is not sufficient to extend neuroblast competence, suggesting that this function might require Hb-mediated transcriptional repression. We therefore tested whether the D or DMZ repression domains of Hb are required for extending competence. We assay neuroblast competence by measuring the number of Eve⁺ U neurons that can be induced by overexpression of Hb within the NB7-1 lineage using the engrailed-gal4 driver (Isshiki et al., 2001; Pearson and Doe, 2003). Wild-type embryos have five Eve⁺ U neurons per hemisegment (Fig. 3A) (Isshiki et al., 2001), whereas overexpression of wild-type Hb can extend neuroblast competence to allow the formation of ~14 Eve⁺ U neurons (n=100; Fig. 4D; Table 1). Similarly, overexpression of Hb proteins lacking either the A+B domains, the B′ domain, or the E domain generated ~16 Eve⁺ U neurons (n>100 each; Fig. 4D; Table 1), showing that none of these domains is required for Hb-mediated extension of neuroblast competence. As expected, overexpression of the non-functional Hb proteins lacking the DBD or the C domain neither specified ectopic U neurons nor altered the identity of the existing neurons (Fig. 4D; see Fig. S2 in the supplementary material; Table 1). Interestingly, overexpression of the Hb proteins that lacked transcriptional repressor activity, i.e. those lacking the D or DMZ domain, failed to extend neuroblast competence, generating only five or six Eve⁺ U neurons (n>200; Fig. 4D; Table 1). The identity of the Eve⁺ U neurons is addressed below, but based simply on the change in the number of Eve⁺ U neurons, we conclude that Hb-mediated transcriptional repression using the D and DMZ domains is required to extend neuroblast competence.

In addition to its role in regulating neuroblast competence, Hb has an essential role in the specification of early-born neuronal identity in multiple neuroblast lineages (Isshiki et al., 2001; Novotny et al., 2002; Pearson and Doe, 2003; Tran and Doe, 2008). To determine whether Hb transcriptional repression is required to specify early-born neuronal identity, we expressed Hb domain deletion proteins in the NB7-1 lineage, either in wild-type or hb mutant embryos. Wild-type embryos have five U neurons per hemisegment: the first-born U1 neuron is Hb⁺ Kr⁺, the second-born U2 neuron is Hb⁺ Kr⁺ Zfh2⁺, and the later-born neurons are Zfh2⁺ Kr⁺ Cut⁺ (U3), Zfh2⁺ Cut⁺ Runt⁺ (U4), or Zfh2⁺ Cut⁺ Runt⁺ Cas⁺ (U5) (Fig. 5A).

Overexpression of wild-type Hb generated ~14 early-born U1/U2 neurons when endogenous Hb was present (Table 1), and ~12 U1/U2 neurons in a hb mutant background (Fig. 5B; Table 1), consistent with previous reports (Isshiki et al., 2001; Novotny et al., 2002; Pearson and Doe, 2003). Similarly, overexpression of Hb proteins lacking either the A+B domains, the B′ domain, or the E domain also generated ~12 Eve⁺ U1/U2 neurons in a hb mutant background (Table 1), showing that none of these domains is required for Hb-mediated specification of early-born neuronal identity.

Overexpression of the Hb deletion proteins that lack transcriptional repression activity (Hb^ΔD and Hb^ΔDMZ) generated five to six Eve⁺ U neurons when endogenous Hb was present (Fig. 5C,E; Table 1). The ectopic neuron was typically the Kr⁺ U3 neuron, which is consistent with the Hb^ΔD and Hb^ΔDMZ proteins having the ability to transcriptionally activate Kr (Fig. 5C) and thus specify U3 identity (Cleary and Doe, 2006; Isshiki et al., 2001). Consistent with this result, overexpression of the Hb^ΔD and Hb^ΔDMZ proteins in a hb mutant background resulted in the loss of the Hb-dependent early-born U1/U2 neurons, while still generating an ectopic U3 neuron (Fig. 5D,F). The only difference we have observed between the Hb^ΔD and Hb^ΔDMZ proteins is that Hb^ΔD, but not Hb^ΔDMZ, frequently generated an ectopic U5 neuron (Fig. 5C,D; Table 1; see Discussion). Because both Hb^ΔD and Hb^ΔDMZ retain the ability to activate Kr expression, we conclude that Hb-mediated transcriptional repression through the D and DMZ domains is required for the specification of first-born neuronal identity.

We have shown that Hb acts as an activator and repressor of gene expression in the CNS, but only its transcriptional repressor function is essential for maintaining neuroblast competence and specifying early-born neuronal identity. We have identified two repression domains within the Hb protein: the Mi2-binding D domain and the DMZ domain.

How do the D and DMZ domains repress gene expression? It is interesting to note that the D and DMZ domains are not dedicated repression domains, such as the one found in Engrailed (Han and Manley, 1993; Jaynes and O'Farrell, 1991). Instead, both are known to mediate protein-protein interactions. The DMZ allows Hb dimerization, leading to the proposal that high Hb levels promote dimerization and thus transcriptional repression (Papatsenko and Levine, 2008). For example, at cellular blastoderm stages, high levels of Hb in the anterior of the embryo are required to repress Kr, whereas low Hb levels activate Kr (Hulskamp et al., 1990; Schulz and Tautz, 1994; Struhl et al., 1992), and mutations in the DMZ lead to an anterior expansion of the Kr expression domain (Hulskamp et al., 1994). Yet it remains unknown how Hb dimerization leads to gene repression. The D domain is also involved in protein-protein interactions. The region of Hb containing the D domain is known to bind the chromatin regulator Mi2, and this interaction promotes epigenetic silencing of the Hb target gene Ubx during early embryonic patterning (Kehle et al., 1998). Our results suggest that the D and DMZ domains could act in distinct processes that are both required for transcriptional repression (Fig. 6B,C), or that they could act in a common pathway such as dimerization-dependent recruitment of Mi2 and/or other repressor proteins to the D domain (Fig. 6D).

Hb proteins lacking the D or DMZ domain have very similar phenotypes in the CNS (this study). Although both the D and DMZ domains appear to be required for Hb-mediated transcriptional repression, they do not have identical functions. Overexpression of Hb^ΔD leads to the specification of two U5 neurons at the expense of the U4 cell identity, whereas overexpression of Hb^ΔDMZ results in normal U4 and U5 identities (Fig. 5). Perhaps Hb^ΔDMZ retains some ability to repress cas expression, allowing the production of the Cas⁻ U4 identity. Alternatively, Hb might use the D and DMZ domains to repress different target genes. Currently, we cannot distinguish between these models owing to the limited number of known Hb direct target genes.

Both Hb and the related mammalian protein Ik have major roles as transcriptional repressors, but are also weak transcriptional activators. How does Hb activate gene expression within the CNS? We were unable to identify a discrete activation domain despite the fact that our systematic deletion series covered the entire protein (see Fig. 4). We can rule out the possibility that the activation domain maps to the D region, as it does in the closely related Ik protein (Sun et al., 1996), because the Hb^ΔD protein has no effect on Kr transcriptional activation or the specification of U3 neuronal identity (Fig. 4). We can also rule out the presence of a single activation domain within the A, B, B′, E or DMZ domains for the same reason. Mechanisms for Hb-mediated transcriptional activation consistent with our data are: (1) Hb activates transcription indirectly by blocking DNA binding of a repressor (Fig. 6A); (2) Hb has multiple activation domains; or (3) the Hb activation domain is tightly linked to an essential domain, such as the DBD. In any case, our VP16::Hb experiments, together with our repression domain deletion experiments, show that Hb-mediated transcriptional repression, not transcriptional activation, is essential for maintaining neuroblast competence and specifying early-born neuronal identity.

What are the Hb-repressed target genes that are involved in extending neuroblast competence? One negatively regulated target is pdm, as co-expression of Pdm with wild-type Hb failed to extend neuroblast competence. However, overexpression of VP16::Hb in a pdm mutant background (lacking both pdm1 and pdm2) was incapable of extending neuroblast competence, showing that Hb must repress multiple genes to extend competence. In the future, further characterization of Hb function in the CNS will require genomic analyses, such as chromatin immunoprecipitation to identify Hb binding sites within the genome, or TU-tagging (Miller et al., 2009) experiments to identify all the genes regulated by Hb within the CNS. Such comparative analyses might help to elucidate the complex gene interactions involved in regulating neuroblast competence.

We thank Drs C. Chabu, M. Kohwi, S. Lai, M. Cleary and J. Hanawalt for helpful discussions and/or comments on the manuscript and Drs T. Lee, L. Tessarollo, D. Court, M. Lundell, J. Renitz, W. Odenwald and H. Jackle for their generosity with reagents. This work was supported by an American Heart Association pre-doctoral fellowship (K.D.T.), an NSF pre-doctoral fellowship (M.R.M.) and NIH R01 HD27056 to C.Q.D., who is an HHMI Investigator. Deposited in PMC for release after 6 months.