The Pit1-Oct1-Unc86 domain (POU domain) transcription factor Brn3a controls sensory neuron survival by regulating the expression of Trk receptors and members of the Bcl-2 family. Loss of Brn3a leads to a dramatic increase in apoptosis and severe loss of neurons in sensory ganglia. Although recent evidence suggests that Brn3a-mediated transcription can be modified by additional cofactors, the exact mechanisms are not known. Here, we report that homeodomain interacting protein kinase 2 (HIPK2) is a pro-apoptotic transcriptional cofactor that suppresses Brn3a-mediated gene expression. HIPK2 interacts with Brn3a, promotes Brn3a binding to DNA, but suppresses Brn3a-dependent transcription of brn3a, trkA, and bcl-xL. Overexpression of HIPK2 induces apoptosis in cultured sensory neurons. Conversely, targeted deletion of HIPK2 leads to increased expression of Brn3a, TrkA, and Bcl-xL, reduced apoptosis and increases in neuron numbers in the trigeminal ganglion. Together, these data indicate that HIPK2, through regulation of Brn3a-dependent gene expression, is a critical component in the transcriptional machinery that controls sensory neuron survival.
Programmed cell death (apoptosis) is controlled by highly conserved mechanisms and plays important roles in the nervous system during development and in pathological conditions (Yuan and Yankner, 2000; Yuan et al., 2003). Among many effectors that contribute to apoptosis, signaling mechanisms transmitted through neurotrophin receptors regulate the decision between survival and cell death during the development of sensory neurons (Kaplan and Miller, 2000; Chao, 2003; Huang and Reichardt, 2003). Failure to acquire an adequate amount of neurotrophins leads to elimination of neurons through apoptosis. Interestingly, the expression of neurotrophin receptors follows a dynamic pattern during sensory neurogenesis. For instance, p75NTR is expressed in the migrating neural crest cells, but its expression is rapidly down-regulated in differentiated sensory neurons. In contrast, expression of TrkA, TrkB, and TrkC sharply increases in dorsal root and trigeminal ganglia when active neurogenesis occurs (Farinas et al., 1998; Huang et al., 1999a; Rifkin et al., 2000). Loss of neurotrophins or Trk receptors leads to a dramatic increase in apoptosis in selective population of sensory neurons (Farinas et al., 1998; Huang et al., 1999a).
In contrast to the well-characterized role of Trk receptors during sensory neuron development, much less is known about the transcriptional mechanisms that control the expression of Trk receptors. Emerging evidence indicates that the cis-regulatory element in the trkA gene is enriched with highly conserved DNA binding sequences that can potentially recruit transcription factors to regulate the spatial and temporal expression of TrkA (Ma et al., 2000). Indeed, analyses of mice with a targeted deletion in Pit1-Oct1-Unc86 domain (POU domain) transcription factor Brn3a shows that Brn3a is essential for survival of sensory neurons (McEvilly et al., 1996; Huang et al., 1999b). Loss of Brn3a results in a progressive down-regulation of TrkA expression, leading to a dramatic increase in apoptotic cell death (Huang et al., 1999b). Consistent with this, Brn3a binding sites have been identified in the trkA enhancer that are required to drive transgene expression in sensory ganglia (Ma et al., 2003). Together, these data provide convincing evidence for a direct involvement of Brn3a in the regulation of TrkA expression.
In addition to regulating Trk expression, Brn3a has also been reported to regulate the expression of Bcl-2, Bcl-xL, and Bax in vitro, providing another possible mechanism through which Brn3a mediates neuronal survival (Smith et al., 1998, 2001; Budhram-Mahadeo et al., 1999; Sugars et al., 2001). Targeted disruption of bax eliminates programmed cell death in the developing nervous system, resulting in an increased number of neurons in selected populations (Deckwerth et al., 1996; White et al., 1998). In contrast, deletion of bcl-xL leads to global and excessive apoptotic cell death in the central and peripheral nervous systems (Motoyama et al., 1995). Despite the critical roles of Bcl-xL and Bax in neuronal survival, however, the mechanisms by which Brn3a regulates their expression under physiological conditions have not been fully characterized. Brn3a responsive elements have been identified in the promoter sequences of bcl-2, bcl-xL, and bax (Smith et al., 1998, 2001), yet it is unclear if expression of these Bcl-2 members is altered in the sensory ganglia of Brn3a−/− mutants.
Several lines of evidence indicate that additional cofactors may regulate Brn3a-dependent transcription. First, although Brn3a expression is detected in sensory ganglia as early as E9.5, cell death and down-regulation of Brn3a target genes are detected only after E13.5, suggesting that Brn3a may not be fully functional in early sensory ganglia (Fedtsova and Turner, 1995; McEvilly et al., 1996; Huang et al., 1999b). Second, brn3a enhancer element-regulated β-galactosidase (LacZ) expression is higher in the sensory ganglia of Brn3a−/− mutants than in wild-type mice, suggesting that Brn3a may negatively regulate its own expression directly or via recruitment of transcriptional corepressors (Trieu et al., 2003).
Here, we provide evidence that interaction of Brn3a and homeodomain interacting protein kinase 2 (HIPK2) regulates the transcription of a common set of prosurvival genes, including TrkA and Bcl-xL. Whereas HIPK2 interacts with the POU homeodomain of Brn3a and promotes Brn3a binding to DNA, HIPK2 suppresses Brn3a auto-regulation and Brn3a target genes. As a consequence, overexpression of HIPK2 leads to neuronal apoptosis in cultured sensory neurons. Conversely, HIPK2 null mutants show up-regulation of Brn3a and Brn3a target genes in the trigeminal ganglion, which leads to reduced apoptosis and increased neuron number in this sensory ganglion.
Identification of HIPK2 as an interacting partner for Brn3a
The important roles of Brn3a in regulating its own expression and the expression of Trk receptors and members of the Bcl-2 family motivated us to investigate additional components that regulate Brn3a-mediated gene expression. Using the yeast two-hybrid screen and a cDNA library prepared from mouse E10-11 embryos (Hollenberg et al., 1995), we isolated several independent clones that interacted with full-length Brn3a, one of which encoded protein sequence between amino acids 752 and 897 of HIPK2, a Ser/Thr kinase that interacts with homeodomain transcription factors of the Nkx family (Fig. 1; Kim et al., 1998). Further characterizations showed that full-length HIPK2 also interacted with Brn3a (not depicted) and that HIPK2 interacted with the POU homeodomain (POUHD) of Brn3a, but not POU specific (POUS) or non-POU domain (Fig. 1, A and B). Not surprisingly, given the high sequence homology in the POU domain among members of the Brn3 family (>95%; Ryan and Rosenfeld, 1997), HIPK2 also interacted with Brn3b and Brn3c (Fig. 1 A).
To further investigate the interaction between Brn3a and HIPK2 in vivo, we performed coimmunoprecipitation in COS cells that expressed Brn3a and EGFP-tagged HIPK2 (EGFP-HIPK2). Cell lysates from COS cells were immunoprecipitated with Brn3a antibody and probed with anti-GFP antibody in Western blots. Our data showed that HIPK2 was present in protein complexes immunoprecipitated by Brn3a antibody, but not by preimmune serum (Fig. 1 B, lanes 2 and 3). Interestingly, the stability of Brn3a and HIPK2 protein complex appeared to be influenced by increasing wash stringency, as the addition of 0.1% SDS significantly reduced protein complex formation between Brn3a and HIPK2 (Fig. 1 B, lanes 4–6).
The effect of wash stringency on the protein complex formation between Brn3a and HIPK2 suggested that the interaction of these two proteins might be transient or require additional components, similar to the interaction between HIPK2 and homeodomain transcription factors in the Nkx family (Kim et al., 1998; Choi et al., 1999). To further characterize this interaction, we determined the effects of HIPK2 on Brn3a binding to the consensus DNA sequence b3s1 (Trieu et al., 1999). Using in vitro translation products of full-length Brn3a and HIPK2 proteins, we showed that Brn3a formed a stable protein complex with this sequence (Fig. 1 C, lane 4). HIPK2 enhanced Brn3a binding, but caused no further shifting in the Brn3a–DNA complex (Fig. 1 C, lanes 5–7). Furthermore, whereas addition of a Brn3a antibody produced a supershift of the complex (Fig. 1 C, lanes 12 and 13), addition of a polyclonal antibody that recognized the COOH terminus of HIPK2 did not (Fig. 1 C, lane 14). These results were consistent with those from coimmunoprecipitation and again suggested that the interaction between Brn3a and HIPK2 may be transient or may require additional factors (Fig. 1 B). However, we cannot completely rule out the possibility that the HIPK2 antibody may not be fully capable of recognizing Brn3a–HIPK2 complex due to steric hindrance.
HIPK2 suppresses Brn3a auto-regulatory activity and expression of Brn3a downstream targets
The pro-survival effect of Brn3a depends on its auto-regulatory activity and the ability of Brn3a to up-regulate several prosurvival genes, including Trk receptors and members of the Bcl-2 family (Huang et al., 1999b; Smith et al., 2001; Ma et al., 2003; Trieu et al., 2003). Indeed, the mRNA levels of TrkA and Bcl-xL, as determined by quantitative RT-PCR (qRT-PCR) assays, showed progressive down-regulation in the trigeminal ganglia of Brn3a−/− mutants at E13.5 and E14.5, preceding the drastic increase in cell death at E15.5 to E17.5 (Fig. 2 A; Huang et al., 1999b). In contrast, expression of Bcl-2 and Bax in the trigeminal ganglion of Brn3a−/− mutants at the same stages remained unchanged (Fig. 2 A). Consistent with the qRT-PCR results, protein levels of TrkA and Bcl-xL showed significant reductions in the trigeminal ganglion of Brn3a−/− mutants at E15.5, whereas Bax remained unchanged (Fig. 2 B). To determine if the effect of HIPK2 on Brn3a-mediated gene expression was direct, we used luciferase constructs that contained three synthetic tandem repeats of Brn3a binding sites identified in the brn3a enhancer element (Prox3), which has been shown to be a reliable reporter for Brn3a activity. Consistent with previous data (Trieu et al., 1999), Brn3a activated the Prox3 reporter in a dose-dependent fashion, whereas HIPK2 by itself did not affect the baseline level of Prox3 luciferase activity (Fig. 2 C). Interestingly, addition of HIPK2 suppressed Brn3a-mediated activation of Prox3 (Fig. 2 C). To determine if expression of HIPK2 in sensory neurons leads to down-regulation of Brn3a in a more physiological setting, we used herpes virus vectors to introduce EGFP-HIPK2 into cultured sensory neurons (Coopersmith and Neve, 1999), achieving almost 100% infection efficiency at 50–100 multiplicity of infection (MOI). Subsequent to infection, total RNA was extracted and mRNA converted to cDNA. qRT-PCR assays were used to measure Brn3a mRNA levels, which were standardized to the endogenous GAPDH level. Our results indicated that, expression of EGFP-HIPK2 leads to a progressive down-regulation of Brn3a mRNA in sensory neurons. By 36 h after infection, Brn3a mRNA was reduced to <40% of that in control neurons infected with LacZ or EGFP virus (P < 0.01; Fig. 2 D).
The suppressive effect of HIPK2 on Brn3a suggests that HIPK2 may inhibit Brn3a auto-regulation, which may in turn lead to down-regulation of Brn3a target genes. However, it is equally possible that HIPK2 may directly suppress the ability of Brn3a to activate its targets. To test latter hypothesis, luciferase constructs containing trkA (two tandem repeats of 5′ Brn3a binding site) or bcl-xL (a single Brn3a binding site) enhancer sequences were used (Smith et al., 2001; Ma et al., 2003). Similar to the Prox3 reporter, Brn3a significantly up-regulated TrkA luciferase activity, whereas HIPK2 suppressed the ability of Brn3a to activate TrkA luciferase activity (Fig. 2 E). Brn3a also activated Bcl-xL luciferase activity and the addition of HIPK2 suppressed Brn3a-mediated activation of Bcl-xL luciferase (Fig. 2 G). Consistent with the luciferase data, addition of Brn3a to cultured sensory neurons up-regulated the expression of TrkA and Bcl-xL mRNA by >12- and 3-fold, respectively (P < 0.01, n = 3; Fig. 2, F and G). In contrast, expression of HIPK2 in sensory neurons reduced TrkA mRNA to ∼7% and Bcl-xL mRNA level to ∼20% of control EGFP-expressing neurons (Fig. 2, F and H). Together, these data support the notion that HIPK2 is a transcriptional corepressor that can suppress the expression of Brn3a and Brn3a downstream targets TrkA and Bcl-xL in sensory neurons.
HIPK2 induces apoptosis in sensory neurons
In previous work, we showed that loss of Brn3a led to a dramatic loss of trigeminal neurons due to increased apoptosis (Huang et al., 1999b). These results lead to the prediction that suppression of Brn3a by HIPK2 should mimic Brn3a loss-of-function phenotype and result in neuronal apoptosis. To test this hypothesis, a gain-of-function approach was used to determine the effect of HIPK2 on the survival of cultured sensory neurons. Although herpes simplex virus (HSV)–mediated expression of EGFP did not affect the survival of sensory neurons (Fig. 3, A–D), expression of EGFP-HIPK2 induced apoptosis, indicated by positive TUNEL staining, in a dose-dependent manner (Fig. 3, E–I). Ectopically expressed EGFP-HIPK2 could be detected either diffusely in the nucleus of trigeminal neurons or in structures resembling nuclear bodies (Fig. 3 E, inset). Neurons expressing EGFP-HIPK2 showed TUNEL-positive staining as early as 24 h after infection (Fig. 3, E–H, and J), by 96 h >80% of these neurons were dead as assessed by morphological criteria (Fig. 3 J). In contrast, expression of EGFP had no effect on survival of infected neurons (Fig. 3, I–J). To determine if the pro-apoptotic function of HIPK2 was solely dependent on its interaction with Brn3a, we examined the effect of HIPK2 expression in sensory neurons of Brn3a−/− mutants. One would predict that if HIPK2 function depends exclusively on its ability to suppress Brn3a, HIPK2-induced apoptosis should be attenuated in Brn3a−/− neurons. Alternatively, if HIPK2 continues to induce apoptosis in these neurons, it is likely that HIPK2 may also activate additional Brn3a-independent cell death mechanisms. Our results indicated that, similar to the data in J, ∼30% of wild-type and Brn3a+/− neurons survived at 72 h after infection with HSV-HIPK2 (Fig. 3 K). However, significantly fewer (∼10%) neurons from Brn3a−/− mutants survived under the same treatment (Fig. 3 K). These results support the notion that HIPK2 and Brn3a antagonize each other and the interaction between these two regulates a delicate balance of gene expression critical for neuronal survival. They also suggest that overexpression of HIPK2 in the absence of Brn3a can lead to significantly more apoptosis, probably due to activation of Brn3a-independent cell death pathway(s) (D'Orazi et al., 2002; Hofmann et al., 2002; Zhang et al., 2003).
Expression of HIPK2 during sensory neurogenesis
To begin to understand the in vivo function of HIPK2, especially in the context of sensory neuron survival during development, we investigated the expression pattern of HIPK2. In situ hybridization using an RNA probe derived from the 39 UTR indicated that HIPK2 mRNA was present in the developing sensory ganglia, including the trigeminal, vestibulocochlear, nodose-petrosal, and dorsal root ganglia (Fig. 4 A). HIPK2 mRNA was also detected in mid-brain/hindbrain regions and in nonneural tissues, such as liver, heart, and kidney (Fig. 4 A and not depicted). Using an antibody that recognized the COOH terminus of HIPK2, HIPK2 was detected in the developing trigeminal and dorsal root ganglia at E12.5 (Fig. 4, C, F, and G). In contrast to the more restricted expression of Brn3a, HIPK2 protein was also detected in the developing spinal cord (Fig. 4, B and C). Interestingly, whereas Brn3a was confined to the nuclei, the distribution of HIPK2 appeared to be more heterogeneous. At E12.5, some neurons showed dense nuclear staining, whereas the staining in others was more diffuse (Fig. 4, F and G, arrows). The presence of HIPK2 protein in the neuronal cytoplasm appeared to be more distinct at postnatal stages, suggesting that the subcellular localization of HIPK2 may change at different stages of neuronal development (not depicted).
Targeted deletion of HIPK2 gene
The suppressive effects of HIPK2 in vitro and the coexpression of Brn3a and HIPK2 in sensory ganglia suggest that both may be important components in the transcriptional mechanism that controls sensory neuron survival and cell death by regulating a common set of prosurvival genes. To test this hypothesis, we deleted HIPK2 in embryonic stem (ES) cells. The mouse HIPK2 locus encompasses ∼150 kb and contains 16 exons (unpublished data). We targeted exon 3, which encodes the first ATG and the entire HIPK2 kinase domain (from amino acids 192 to 520). The majority of exon 3 was replaced with the reporter gene LacZ, followed by a polyadenylation signal sequence and the selection marker neomycin gene. Four positive ES clones were identified by Southern blot analysis using specific 5′ and 3′ probes, three were injected into blastocysts (Fig. 5, A and B). Germline transmission of the targeted allele was confirmed by Southern blot analyses (Fig. 5 C). Northern blot analysis using a probe that hybridized with a sequence in the exons 3 and 4 confirmed that HIPK2 mRNA was undetectable in mouse embryonic fibroblasts (MEFs) derived from homozygous HIPK2 (HIPK2−/−) mutants (Fig. 5 D). The absence of HIPK2 transcripts in the trigeminal ganglion of HIPK2−/− mutants was further confirmed by RT-PCR using primers that detected sequences in exon 3 and in exons 15–16 of HIPK2 (Fig. 5 E). The insertion of LacZ in–frame with the coding sequence of HIPK2 (HIPK2LacZ) provided an excellent lineage tracer for cells expressing HIPK2. Consistent with the in situ hybridization and immunostaining results (Fig. 4), significant HIPK2LacZ staining intensity was detected in the developing sensory ganglia, including the trigeminal and dorsal root ganglia, diencephalon, and spinal cord at E13.5 (Fig. 5, F and G).
Loss of HIPK2 results in up-regulation of Brn3a and Brn3a target genes
The coexpression of Brn3a and HIPK2 in sensory ganglion and the suppressive effects of HIPK2 on Brn3a-mediated gene expression suggest that loss of HIPK2 could lead to increases in the expression of Brn3a and its downstream targets. To test this hypothesis, we used qRT-PCR assays to determine the mRNA levels of Brn3a, TrkA, Bcl-xL, and Bax in trigeminal ganglion from individual wild-type and HIPK2−/− embryos collected at E13.5, 14.5, and 15.5. Loss of HIPK2 resulted in a significant increase in Brn3a, TrkA, and Bcl-xL mRNA at E14.5 and E15.5, with no significant change at E13.5. Specifically, the level of Brn3a mRNA increased by 2.7-fold at E14.5 and twofold at E15.5 (Fig. 6 A; P < 0.01, t test). The levels of TrkA and Bcl-xL mRNA also showed significant, albeit smaller increases at E14.5 and E15.5 (Fig. 6 A). In contrast, the level of Bax mRNA remained unchanged from E13.5 to E15.5. Although Brn3a mRNA showed no significant increase at E13.5, Brn3a protein was modestly increased in some HIPK2−/− mutants at E13.5 (∼1.2-fold), suggesting that HIPK2 may also regulate Brn3a protein stability at this stage (Fig. 6 B). At E15.5, the protein levels of Brn3a, TrkA, and Bcl-xL proteins showed consistent increases by 2.6-, 1.6-, and 1.7-fold, respectively (Fig. 6 B). In contrast, levels of Bcl-2 and Bax proteins were not altered at any time point examined. Together, increases in the mRNA and proteins of Brn3a, TrkA, and Bcl-xL in the trigeminal ganglion of HIPK2−/− mutants suggest that HIPK2 negatively regulates Brn3a-mediated gene expression in vivo.
Reduced apoptosis and increased neuron number in trigeminal ganglion of HIPK2−/− mutants
Previous data have indicated that loss of Brn3a, TrkA, or Bcl-xL results in prominent cell death and dramatic loss of neurons in developing sensory ganglia (Motoyama et al., 1995; Huang et al., 1999a,b). Hence, the observed increase in the expression of Brn3a, TrkA, and Bcl-xL in the trigeminal ganglion of HIPK2−/− mutants should create a condition rendering HIPK2−/− sensory neurons more resistant to apoptosis during development (Ensor et al., 2001). To test this possibility, we determined the number of neurons undergoing apoptosis and the total number of neurons in the trigeminal ganglion of wild-type and HIPK2−/− mutants at different developmental stages. In the wild-type trigeminal ganglion, apoptotic profiles were detected as early as E10.5 and reached a maximum between E11.5 to E13.5, followed by a progressive decline at E15.5 and P0 (Fig. 7, A, C, and E; Huang et al., 1999b). Most of the apoptotic profiles stained positive for neurofilament, indicating that they are indeed neurons (Fig. 7 A, arrowheads). Whereas the number of apoptotic neurons was similar between wild type and HIPK2−/− at E10.5 and E11.5, there was significantly reduced apoptosis in the mutant ganglion at E13.5, E15.5, and P0 (Fig. 7, B, C, and E; 50–60% reduction, P < 0.01; t test). In agreement with this, there were significantly more neurons in the HIPK2−/− ganglion at E13.5 and P0 (0.01 < P < 0.025 at E13.5 and P < 0.01 at P0, n = 4, t test). Thus, these data indicate that loss of HIPK2 leads to up-regulation of Brn3a and its target genes and reduces apoptotic cell death during the development of sensory ganglia.
Maintenance of neuronal survival depends on a balanced expression of prosurvival and proapoptotic genes. Results from this work demonstrate that transcriptional corepressor HIPK2 negatively regulates Brn3a-mediated neuronal survival in the developing sensory ganglia. The mechanism by which HIPK2 regulates Brn3a is mediated through protein–protein interaction, which suppresses the Brn3a-dependent transcription of brn3a, trkA, and bcl-xL loci. As a consequence, overexpression of HIPK2 mimics the phenotype of the Brn3a−/− mutants, resulting in apoptotic cell death in cultured sensory neurons. Conversely, loss of HIPK2 leads to a stage-specific up-regulation of Brn3a and Brn3a target genes, reduced apoptosis, and increased neuron number in the trigeminal ganglion in vivo. Together, these data support the model that interaction between Brn3a and HIPK2 regulates gene expression, apoptosis, and neuronal survival during trigeminal ganglion development (Fig. 8 A). Due to their important roles in regulating the expression of TrkA and Bcl-xL, loss of Brn3a or HIPK2 leads, respectively, to either increased or reduced programmed cell death (Fig. 8 B).
It is well-established that a cascade of transcriptional factors controls the specification of cell fate and determination of cell survival and differentiation during the development of sensory ganglia (Anderson, 1999). For instance, expression of the basic helix-loop-helix factors Ngn1 and Ngn2 promotes neuronal fate in cranial and spinal sensory ganglia (Ma et al., 1998, 1999). Indeed, ectopic expression of Ngn1 results in ectopic Brn3a and TrkA expression in mesenchymal cells (Greenwood et al., 1999). In contrast, characterization of Brn3a−/− mutants indicates that Brn3a is essential for survival and differentiation of sensory neurons, but not neuronal cell fate specification (Huang et al., 1999b, 2001; McEvilly et al., 1996; Xiang et al., 1996; Eng et al., 2001). In the trigeminal ganglion of Brn3a−/− mutants, >70% of the neurons undergo apoptotic cell death between E15.5 and E17.5 (Fig. 8 B). Mechanisms of apoptotic cell death in Brn3a−/− mutant sensory ganglia involve disruption of neurotrophin signaling as a result of a progressive down-regulation of TrkA (Fig. 2; Huang et al., 1999b; Ma et al., 2003). In addition, the current work also demonstrates that Brn3a is required for the maintenance of Bcl-xL in vivo with a time course similar to that of TrkA (Fig. 2).
Recent analyses of cis-regulatory elements in trkA, brn3a, and bcl-xL extend these in vivo findings and provide convincing data for a direct involvement of Brn3a in transcriptional control of these target genes (Trieu et al., 1999; Ma et al., 2000, 2003; Smith et al., 2001). Mutations in Brn3a-binding sites abolish the ability of the trkA enhancer to drive transgene expression in sensory neurons (Ma et al., 2003). Similar to the trkA enhancer, multiple Brn3a binding sites have also been identified in the brn3a enhancer (Trieu et al., 1999). Brn3a activates luciferase constructs containing these binding sites, consistent with the notion that Brn3a positively regulates its own expression through a feed forward mechanism (Fig. 2 and Fig. 8 A; Trieu et al., 1999). Interestingly, all the Brn3a binding sites are located several kilobases upstream from the transcriptional initiation site, suggesting that a long-range control through the recruitment of multiple transcriptional cofactors may regulate Brn3a expression. Indeed, analyses of Brn3a autoregulation in transgenic mice expressing LacZ under the control of brn3a enhancer indicate that, rather than activating, Brn3a suppresses its own expression at the early stages of sensory ganglion development (Trieu et al., 2003). These observations have led to the hypotheses that either Brn3a suppresses its own expression or a yet unidentified transcriptional corepressor may inhibit Brn3a functions in a stage-dependent manner (Trieu et al., 2003). Data from this current work favor the latter hypothesis and indicate that HIPK2 is a candidate corepressor that suppresses Brn3a auto-regulation and Brn3a target genes. Consistent with this model, removal of HIPK2 results in up-regulation of Brn3a, TrkA, and Bcl-xL, a reduction in apoptosis during the period of programmed cell death, and a slight increase in neuron numbers during the development of trigeminal ganglion (Figs. 6 and 7).
A diversity of molecular mechanisms contributes to the execution of cell death through the canonical and noncanonical pathways (Yuan et al., 2003). The extensive cell death in the sensory ganglia of trkA−/− and bcl-xL−/− mutants coincides with the period of programmed cell death in these ganglia and underscores the important roles of these two molecules in the canonical cell death pathway (Motoyama et al., 1995; Huang et al., 1999a; Yuan et al., 2003). The fact that Brn3a regulates the expression of TrkA and Bcl-xL indicates that Brn3a-mediated transcriptional regulation is an integral regulatory component of the cell death process. Loss of Brn3a leads to down-regulation of TrkA and Bcl-xL, which precedes and most definitively contributes to the dramatic increase in cell death observed in the trigeminal ganglia of Brn3a−/− mutants (Huang et al., 1999b; Ma et al., 2003). The reduced apoptosis observed in the trigeminal ganglion of HIPK2−/− mutants provides supporting evidence that the interaction between Brn3a and HIPK2 regulates a delicate balance of gene expression that is critical for cell death. Interestingly, however, loss of HIPK2 appears to have no effect on apoptosis in the trigeminal ganglion at E10.5 or E11.5. Rather, the reduction in neuronal apoptosis in HIPK2−/− mutants is most significant at E13.5, E15.5, and P0, with a slight increase in neuron number at the same stages (Fig. 7, E and F; Fig. 8 B). One possible explanation for the lack of a detectable effect at the early stage of programmed cell death in HIPK2−/− mutant is that the efficiency of HIPK2-mediated suppression of Brn3a may require the formation of a larger and more effective corepressor complex. Indeed, recent data have indicated that HIPK2 can interact with additional corepressors, including Groucho, histone deacetylase (HDAC) and CtBP (Choi et al., 1999; Zhang et al., 2003). Thus, if a larger HIPK2-based corepressor complex does exist in vivo, removal of more than one component of this complex may be required to produce a more significant change in Brn3a-mediated gene expression.
Given the role of HIPK2 as a transcriptional repressor, how might the activity of HIPK2 be regulated during sensory neuron development? Recent evidence indicates that transcriptional corepressors of class II HDACs show a progressive switch from nucleus to the cytoplasm during muscle cell differentiation, whereas muscle specific transcription factor MEF remains constantly in the nucleus (McKinsey et al., 2000). Nuclear export of HDACs is regulated by Ca2+ signals at the cell surface and by the subsequent activation of calcium-calmodulin kinases (McKinsey et al., 2000). Due to the interaction between HIPK2 and HDAC (Choi et al., 1999), it is tempting to speculate that a similar mechanism may regulate HIPK2 functions during sensory neuron development. Future more in-depth analyses of HIPK2 protein distribution should help test this hypothesis. Alternatively, it is possible that other signaling pathways, such as those involving neurotrophins, TGFβ and Wnts, may also regulate the subcellular localization of HIPK2 or its interaction with other proteins (Harada et al., 2003; Kanei-Ishii et al., 2004).
In addition to its ability to interact with Brn3a, HIPK2 has also been shown to interact with several other factors that are important for cell survival and apoptosis, most notably, p53 and CtBP (D'Orazi et al., 2002; Hofmann et al., 2002; Zhang et al., 2003). Indeed, overexpression of HIPK2 in Brn3a−/− sensory neurons or sympathetic neurons, which do not depend on Brn3a for survival (Ensor et al., 2001), continues to cause apoptotic cell death (Fig. 3 K; unpublished data). These results support the notion that HIPK2 can activate additional Brn3a-independent cell death pathways and that such mechanisms might become more prominent in the absence of Brn3a (Fig. 3 K). Unlike Brn3a, TrkA, Bcl-xL, or HIPK2, however, the level of CtBP expression is extremely low in the sensory ganglia, making it a less likely contributor to sensory neuron survival (Hildebrand and Soriano, 2002; unpublished data). In contrast, the reported antagonistic roles of Brn3a and p53 in transcriptional regulation of Bcl-xL raises the interesting possibility that, in addition to suppressing Brn3a, HIPK2, and p53 may cooperatively suppress Bcl-xL expression (Sugars et al., 2001). Alternatively, HIPK2 could potentially up-regulate p53-dependent activation of Bax (Budram-Mahadeo et al., 2002). Our data indicate that the level of Bax is essentially unchanged in either Brn3a−/− or HIPK2−/− mutants (Figs. 2 and 7) and we did not detect any synergistic effect of HIPK2 and p53 on Bax or Bcl-xL luciferase activity (unpublished data). Future investigations using genome-wide screens for the expression of p53-dependent genes in HIPK2−/− mutants should provide more information regarding common target genes that are regulated by both HIPK2 and p53.
Materials And Methods
Yeast two-hybrid screen, coimmunoprecipitation, and Western blot analyses
To identify Brn3a cofactors, we screened a yeast two-hybrid cDNA library prepared from E10.5-11.5 mouse embryos (Hollenberg et al., 1995). Of the 107 transformants, 67 were His+ and LacZ+. After ruling out false-positive clones, nine clones showed specific interactions with full-length Brn3a. Among these, five interacted with the POU domain of Brn3a, but not the non-POU domain. Sequencing data showed that this clone encoded partial sequence (amino acids 782–897) of HIPK2 (Kim et al., 1998). To further define which region in Brn3a interacted with HIPK2, we prepared yeast constructs that contained only the POU-specific domain (POUS) or POU homeodomain (POUHD) and determined their ability to interact with HIPK2 (amino acids 752–897) in liquid X-gal assays. Full-length HIPK2 cDNA was provided by Y. Kim (National Institutes of Health, Bethesda, MD).
For coimmunoprecipitation, COS cells were transfected with EGFP-HIPK2 and Brn3a cDNA using LipofectAMINE (Invitrogen). Cell lysates were collected 48 h after transfection in RIPA buffer. Supernatant was incubated for 1–2 h at 4°C with Brn3a antibody and with protein A/G agarose beads (Santa Cruz Biotechnology, Inc.) overnight at 4°C. Beads were washed in buffers containing 1% NP-40 (wash 1), 1% NP-40/0.5% sodium deoxycholate (wash 2), and 1% NP-40/0.5% sodium deoxycholate/0.1% SDS (wash 3). EGFP-HIPK2 was detected with anti-GFP (BD Biosciences). Quantification of the protein levels in Western blots (Fig. 6 B) was performed using the NIH Image program (developed at the National Institutes of Health and available at http://rsb.info.nih.gov/nih-image/). A central lane that contained proteins from HIPK2+/− trigeminal ganglia was spliced out in Fig. 6 B according to the guidelines by JCB (Rossner and Yamada, 2004).
Luciferase assays and EMSA
The luciferase reporter constructs for Brn3a and Bcl-xL (provided by E. Turner, University of California, San Diego, San Diego, CA; and G. Packham, Imperial College, London, UK) have been characterized previously (Budhram-Mahadeo et al., 1999; Trieu et al., 1999; Sugars et al., 2001; Thornborrow et al., 2002). The TrkA luciferase construct was prepared by placing two tandem repeats of 5′-end Brn3a binding site in trkA promoter into pGL3-basic (Promega; Ma et al., 2003). To determine the effects of Brn3a and HIPK2 on the activity of these promoters, 293 cells (1.5 × 105) were transfected with Brn3a and EGFP-HIPK2 cDNA at designated concentrations. The total amount of DNA added was normalized using pcDNA3.1. Measurement of luciferase activity was done using the dual luciferase system using TD 20/20 luminometer (Turner Designs). The effects of HIPK2 in Brn3a binding to the consensus DNA sequence (b3s1) were determined using in vitro translated full-length Brn3a and HIPK2 protein (Coupled TNT System; Promega; Liu et al., 2000).
In situ hybridization and immunohistochemistry
Two relatives of HIPK2, HIPK1, and HIPK3, were reported to share high DNA sequence homology with HIPK2 (Kim et al., 1998). To circumvent the problem of cross reactivity, a 600-bp probe from the 3′ UTR of HIPK2 was isolated. Sequence homology of this probe to similar regions in HIPK1 and HIPK3 is <40%. Northern blot analyses using this probe showed a very different expression pattern from that of HIPK and HIPK3, and sense probes were used as negative controls and showed no detectable signals. A pAb recognizing NF150 was used (CHEMICON International, Inc.). Brn3a antibody was previously characterized (Huang et al., 1999a, 2001) and HIPK2 antibody was generated in rabbits using a maltose binding protein fused with COOH terminus of HIPK2 (MBP-HIPK2C) and purified using GST-HIPK2C affinity column.
Sensory neuron cultures
Trigeminal ganglia from P0-P2 mice were trypsinized and triturated in 1 ml of F-14 culturing medium with a fire-polished siliconized Pasteur pipette. The dissociated cell suspension was diluted and plated in F-14 culturing medium in tissue culture dishes or slide chambers. NGF (10 ng/ml) or glia-derived neurotrophic factor (50 ng/ml, for cultures of Brn3a−/− neurons; Huang et al., 1999b) was added to the medium at the time of plating. For real-time PCR experiments, cells were plated at a density of ∼2,500 cells per dish, whereas for cell counting and immunohistochemical analysis, ∼300 cells per dish were plated. After plating, cells were incubated at 37°C for 3 d before infection with HSV. The preparation of recombinant HSV carrying EGFP-HIPK2 or Brn3a was generated using the standardized methods and the titers of HSV were determined using PC12 cells (Coopersmith and Neve, 1999).
Assessment of cell survival and apoptosis
The number of living neurons in culture was determined by counting the neurons within a marked grid in the center of the culture dishes before HSV infection and at various times after infection. Neurons were observed by phase-contrast microscopy (model TE2000U; Nikon). Living neurons were identified as having a smooth, spherical, bright cell body and multiple neurites, whereas dead cells had a shrunken irregular cytoplasm and either fragmented or complete loss of neurites. Apoptosis was assessed using TUNEL with the in situ cell death detection kit, TMR red (Roche Molecular Biochemicals). Cells were analyzed using a three laser confocal microscope (model TCS SP; Leica) and were considered to be apoptotic by the presence of condensed, TUNEL positive nuclear material accompanied by shrinking of the cytoplasm and loss of axons. Laser intensity (measured as the PMT levels) for each fluorophor was kept within the linear range. Images were captured using confocal software (Leica) and imported into Photoshop 7.0. Countings of apoptotic neurons and total neuron number in the trigeminal ganglion were performed as described previously (Huang et al., 1999a,b).
A qRT-PCR technique was used to measure the levels of TrkA, Bcl-xL, Brn3a, and Bax mRNAs. Total RNAs were extracted using an RNeasy Kit (QIAGEN). Reverse transcription reactions were performed using random hexamer primers and the SuperScript first-strand synthesis system (Invitrogen). Quantification of cDNA was done on ABI Prism 7900HT detection system (PE Applied Biosystems). The levels of GAPDH were also quantified allowing TrkA, Bcl-xL, Brn3a, and Bax expression to be calculated relative to GAPDH mRNA.
Gene targeting of HIPK2 and generation of HIPK2 null mice
Murine HIPK2 genomic DNA was isolated from a mouse 129/SvJ λ library (Stratagene) using a cDNA probe containing the first coding exon (exon 3) of HIPK2. Recombination arms were generated by PCR and subcloned into the targeting vector pPNT, which contained neomycin (PGK-neo) and thymidine kinase (hsv-tk) genes (Tybulewicz et al., 1991). The reporter gene LacZ, followed by a poly(A) sequence, was inserted in frame with the first ATG of HIPK2. The targeting construct was electroporated into 129SvJ ES cells and selected with both G418 and gancyclovir. Genomic DNA from G418-resistant ES colonies was digested with Nhe1 and Xba1 and screened with 5′ and 3′ probes. 5′ probe detected a 8.8-kb Nhe1 fragment in wild-type allele and a 14-kb fragment in mutant allele, whereas 3′ probe detected a 7.6-kb Xba1 fragment in wild-type allele and a 12.8-kb fragment in mutant allele. Three independent ES lines were injected into C57Bl/6J blastocysts. Chimeras were mated to C57Bl/6J females to generate F1 heterozygotes. All mice were maintained in 129 and B6 mixed background. Animal care was performed according to the Institute of Animal Care and Use Committee guidelines. Genotypes were confirmed by Southern blot and PCR analysis. Northern blots using mRNA from MEFs and a probe containing exons 3 and 4 was performed to confirm the absence of HIPK2 expression. RT-PCR was also performed on mRNA from MEFs and trigeminal ganglia by using primers for the first coding exon (exon 3) of HIPK2 and the last two exons (exon 15 and 16).
Online supplemental material
The sequences of primers and probes used in qRT-PCR and genotyping of HIPK2 mutants and the experimental details are available in Table S1 (primers and probes for qRT-PCR for Brn3a and its target genes), Table S2 (primers for HIPK2 mutant mice genotyping), and Table S3 (primers for RT-PCR to detect HIPK2 cDNA). Online material is available.
We thank Dr. Y. Kim for full-length HIPK2 cDNA, Dr. D. Ginzinger for assistance with qRT-PCR primers, Drs. Z. Mo and M. Xiang for help with EMSA, Drs. E. Turner and G. Packham for luciferase constructs, and Drs. Ben Yen and Abul Abbas for support.
This work has been supported by grants from the UCSF Academic Senate and School of Medicine REAC, NCIRE Young Investigator Award, ARCD, PECASE and Merit Awards (VAMC), and the Whitehall Foundation (to E.J. Huang), and grants from the NIH (to E.J. Huang, L.F. Reichardt, and R.L. Neve). L.F. Reichardt is a Howard Hughes Medical Institute investigator. The authors have no commercial affiliations or conflict of interest.
A.K. Wiggins and G. Wei contributed equally to this work.
Abbreviations used in this paper: EGFP-HIPK2, EGFP-tagged HIPK2; ES, embryonic stem; HDAC, histone deacetylase; HIPK2, homeodomain interacting protein kinase 2; HSV, herpes simplex virus; LacZ, β-galactosidase; MEF, mouse embryonic fibroblast; MOI, multiplicity of infection; POU domain, Pit1-Oct1-Unc86 domain; qRT-PCR, quantitative RT-PCR.