Coordinated regulation of gene expression by Brn3a in developing sensory ganglia

Studies of the developing vertebrate nervous system have revealed a large number of transcription factors that are expressed in specific populations of neurons or their precursors. In dividing neuroepithelial cells, transcription factors of the bHLH, homeodomain and other classes characterize regions of the neural tube with specific developmental potentials. Later in neurogenesis,other transcription factors are expressed in specific populations of neurons,and may persist in the mature nervous system. Naturally occurring and induced mutations of both the early and late transcription factors have been shown to exert profound effects on neural development.

Transcription factors reside permanently or conditionally in the nucleus,and are presumed to work by interacting with specific cis-acting binding sites in the vicinity of the transcription units they regulate. These `target genes'in turn mediate the effects of the transcription factor on developmental fate decisions, neuronal phenotype and cell survival. However, the downstream targets of these factors cannot necessarily be inferred from their expression patterns, because they are usually not congruent with those of other classes of neural genes, such as neurotransmitters or their receptors. In a few cases,plausible regulatory relationships have been established between neural transcription factors and their targets, but for the vast majority, no clear pathways are known. Using conventional methods applied to individual genes,establishing these transcription factor-target relationships is quite inefficient.

In principle, comparing the transcript pool of neural tissue from a wild type animal to that of an animal under- or overexpressing a given factor should yield a complete set of genes regulated in that cell type. However,because of the tremendous cellular diversity present in most regions of the nervous system, the resulting changes in gene expression in a specific cell type may be obscured by the heterogeneity of the sample. Furthermore, the changes in target gene expression may be regulated indirectly, either as downstream or compensatory effects.

We have been engaged in studies of Brn3a (Pou4f1 – Mouse Genome Informatics), a transcription factor of the POU-domain family which is expressed in terminally differentiating neurons of the sensory peripheral nervous system and caudal CNS. Targeted mutations in mice have shown that Brn3a is necessary for the correct development and/or survival of neurons in the sensory ganglia and some CNS nuclei(McEvilly et al., 1996; Xiang et al., 1996). Sensory neuron death in Brn3a knockout mice is preceded by loss of neurotrophin receptor expression (Huang et al., 1999; Ma et al.,2003), and by markedly defective axonal growth(Eng et al., 2001). Despite the success of the knockout approach in demonstrating the importance of Brn3a and related POU factors in neural development, these experiments have yielded little information about what genes these factors regulate, and why they are essential for normal axon growth or neuronal survival.

In the present study, we have used microarrays to compare the patterns of gene expression in the trigeminal ganglia of Brn3a knockout and wild-type mice. To maximize the homogeneity of the samples and to minimize secondary effects on gene expression, we have analyzed embryonic ganglia. At the stage chosen for analysis, embryonic day 13.5 (E13.5), major defects in sensory axon growth are observed in the mutant mice(Eng et al., 2001), but the phase of marked sensory neuron death has not yet commenced(Huang et al., 1999).

Our results demonstrate that Brn3a regulates a coordinated program of gene expression that defines the phenotype of developing trigeminal neurons,including the regulation of neurotransmitters, receptors, ion channels,mediators of axon growth, and other transcription factors. Many of these target genes have known roles in sensory neurons and are strong candidates for mediating the observed effects of Brn3a on axon growth and cell survival. Some of the genes regulated by Brn3a in the trigeminal ganglion are also changed in other sensory ganglia in Brn3a knockout mice, but do not appear to be altered in Brn3a-expressing CNS neurons, suggesting that the roles of Brn3a in the sensory system and CNS may be distinct.

To generate tissue for microarray analysis, timed matings of Brn3a heterozygote animals were performed, and the embryos harvested at embryonic day (E)13.5. Only embryos at E13.5±0.5 days, based on the staging system of Theiler (Theiler,1972), were pooled for microarray analysis. Trigeminal ganglia were removed by blunt dissection and carefully freed of adherent non-neural tissue with fine forceps. Only complete ganglia were used for analysis. Dissected ganglia were placed in RNAse inhibitor solution (RNAlater, Ambion)and frozen until enough tissue was harvested to be pooled. Embryos were genotyped from a sample of tail or hindlimb tissue harvested at the time of ganglion dissection. Genotypes were determined for the native Brn3aallele and the neomycin resistance cassette included in the null allele as previously described (Eng et al.,2001). Approximately 10-12 genotyped ganglia were sufficient to provide 5 μg of total RNA for a single analysis, which was extracted using the RNeasy system (Qiagen). The generation of cDNA, production of labeled cRNA, and hybridization to GeneChip arrays were all performed according to standard protocols provided by the manufacturer (Affymetrix).

Non-isotopic in situ hybridization was performed as previously described(Birren et al., 1993). A table of probes used and their sources appears in the Supplemental Data (Table S1). Immunofluorescence for Brn3a was performed with rabbit polyclonal antisera as previously described (Fedtsova and Turner,1995). Immunofluorescence for other antigens was performed with commercially available antibodies, including rabbit anti-calretinin (Swant),rabbit anti-galanin (Peninsula Laboratories), rabbit anti-somatostatin-14(Peninsula Laboratories), and rabbit anti-tyrosine hydroxylase (Chemicon).

The primary analysis of microarray data, including determination of the absence/presence of the assayed transcripts, transcript expression levels, and the probability of change in transcript expression between genotypes(`change-p') was performed with Microarray Suite 5.0 (Affymetrix). Two proprietary databases were used to relate microarray results for ESTs to the identity of the expressed transcripts, NetAffx (Affymetrix) and GeneSpring(Silicon Genetics). The results for those transcripts identified in both databases were discordant in less than 1% of cases.

To begin to address the complement of genes regulated by Brn3a, we chose to analyze gene expression in the trigeminal ganglia of Brn3a wild-type,heterozygous and knockout mice at E13.5. These ganglia do not represent a homogeneous population with respect to their eventual sensory subtype, but at this stage most of the cells in the trigeminal ganglion exhibit Brn3a immunoreactivity (Trieu et al.,2003) and thus it is likely that Brn3a regulates genes common to a majority of the neurons present in these ganglia. E13.5 was chosen as the time point for analysis because in mice lacking Brn3a, a clear defect in the growth of axons from the trigeminal ganglion is evident by this stage, with aberrant innervation of peripheral and central targets. Furthermore, the extensive sensory cell death observed in Brn3a knockout mice does not take place until E14.5-E16.5 (Huang et al.,1999), so any effects from cell loss should be minimal at this stage.

Although no differences in the phenotype of Brn3a wild-type and heterozygous embryos have been identified(Eng et al., 2001), tissue samples from these genotypes were analyzed separately. The comparison of gene expression across all three genotypes was performed to look for subtle differences in heterozygotes, and to provide a partial replication of the results within each experiment. All three genotypes were analyzed in two completely independent experiments.

Trigeminal RNA were analyzed using the commercial oligonucleotide-based U74Av2 and U74Bv2 microarrays (Affymetrix). The U74Av2 array represents 12,422 transcripts, including 5993 known genes and 6429 ESTs, and the U74Bv2 array includes an additional 12,411 EST sequences. A significant number of the EST sequences present on both arrays have subsequently been related to identified genes in public and proprietary databases. Of all the transcripts represented on the U74Av2 array, 4885 were detected as `present' in both experiments in at least one of the three Brn3a genotypes, using the manufacturer's standard criteria for array analysis. The transcripts that were reproducibly present on the U74Av2 and U74Bv2 arrays were then further analyzed with respect to their relative expression in the three genotypes.

Two measures were used to compare transcript levels between samples from different genotypes, the change-p value and the fold change in the intensity of the hybridization signal. The change-p value is calculated by proprietary data analysis software (Affymetrix) using the Wilcoxon's signed rank test applied to the hybridization signals for the 16 matched and mismatched oligonucleotide probe pairs representing each transcript in the array. Change-p values <0.003 (increased expression in the arbitrarily designated`experimental' sample) or >0.997 (decreased in the experimental sample) are considered highly significant. For abundant transcripts, change-p values may be significant even when the fold change in expression is small, because for strong signals even minor relative differences may achieve statistical significance by this method. Because small relative changes in expression are not easily verified, and have uncertain biological significance, a minimum twofold increase or decrease in expression was used as an additional criterion for determining the changed transcripts of interest. More detailed information on the transcripts included and excluded by these criteria are given in Fig. S1.

Figure 1A compares the relative expression of all present transcripts in heterozygote and knockout ganglia in one analysis using the U74Av2 array. The vast majority of the expressed transcripts fall between parallel lines designating less than a twofold change in expression. The expression values for significantly changed transcripts are located off the central axis, and the positions of selected mRNAs encoding proteins with known roles in sensory development or function are indicated.

A complete list of the known transcripts that met the criteria for significantly increased expression appears in Table 1, and the significantly decreased genes are shown in Table 2. Most of the known transcripts that were significantly changed in Brn3a knockout ganglia encode proteins with established roles in neural function or neural development, including neurotransmitters and their receptors, enzymes of neurotransmitter synthesis, ion channels, specialized components of axons and synapses, mediators of intracellular signal transduction, and transcription factors. The specificity of these changes is underscored by the fact that we observed very few significant changes in transcripts associated with processes other than neurogenesis and neural function, such as factors regulating metabolic pathways, the cell cycle, or apoptosis. A more extensive list of 271 known neural genes, which were expressed in E13.5 trigeminal ganglia but were either unchanged in the Brn3a knockout or changed but did not meet the inclusion criteria,appears in Table S2.

In addition to the previously known transcripts represented on the U74Av2 array, the U74Av2 and U74Bv2 arrays include oligonucleotide probes for nearly 19,000 ESTs, some of which have subsequently been linked to known genes. Although many significantly changed ESTs on these arrays could not be identified, searches of public and proprietary databases allowed a significant number to be assigned with confidence, and these are summarized in Table 3. Several of the ESTs confirmed results obtained for the known transcripts, including GATA3,AP2β, NeuroD1, Scn9a and Runx1. In addition, analysis of the ESTs contributed a number of novel changed genes, particularly transcription factors and mediators of intracellular transduction pathways.

Brn3a heterozygous mice are viable and do not exhibit the defects in sensory axon growth or neuronal survival observed in knockouts. In previous studies we have shown that Brn3a attenuates its own expression via an autoregulatory enhancer (Trieu et al.,2003; Trieu et al.,1999). In heterozygotes, increased Brn3a expression from the intact allele restores Brn3a transcript levels to approximately 90% of the wild-type level, effectively compensating for reduced gene dosage. If this mechanism in fact normalizes Brn3a regulatory activity in heterozygotes, one would predict that all downstream targets of Brn3a, whether directly or indirectly regulated, would have nearly normal expression levels in heterozygous mice.

To test this hypothesis, we compared the expression levels of 41 increased,62 decreased and 160 unchanged transcripts in ganglia from the three Brn3a genotypes. For both the increased and decreased transcripts,the target gene expression levels in wild-type and heterozygous ganglia showed similar differences from those of the knockout(Fig. 1B). In contrast, the wild-type and heterozygote ganglia did not significantly differ from each other for either class of target genes. These results confirm the complete suppression of a heterozygous phenotype at this stage in the trigeminal ganglia of Brn3a knockout mice.

Numerous genes have been previously reported to be transcriptionally activated by transfected Brn3a in cell culture models of sensory ganglia. These putative Brn3a targets include structural components of axons and synapses, neurotransmitter receptors and oncogenes(Table 4). Almost all of these proposed downstream genes are represented on the U74A and B arrays, and were detected (present call) in E13.5 trigeminal ganglia. However, aside from a modest but statistically significant decrease in the neurofilament NF-H, none of these genes were markedly affected by the loss of Brn3a expression in vivo.

Work in cell line models has also led to the hypothesis that decreased expression of the anti-apoptotic gene Bcl2 may contribute to the extensive death of sensory neurons in Brn3a knockout mice(Smith et al., 1998). In recent work, we have shown that Bcl2 mRNA levels are unaltered in Brn3a knockout mice just prior to the onset of cell death(Eng et al., 2003), suggesting that loss of Bcl2 expression is not a primary defect in Brn3a knockout mice. Consistent with this result, two probe sets on the array gave unchanged signals for Bcl2. Overall, the results reported here do not support the in vivo regulation of the putative Brn3a targets from previous over-expression studies in transfected cell lines, and illustrate the difficulty of identifying physiological regulatory pathways in these model systems.

Previous studies of Brn3a knockout mice have also revealed changes in the expression of several genes in the sensory ganglia. A major focus of these studies has been the neurotrophins and their receptors. We have previously reported microarray assays showing a reduction in TrkA transcripts in Brn3a knockout trigeminal ganglia at E13.5(Ma et al., 2003), and our array analysis is consistent with previous reports that the TrkA neurotrophin receptor is decreased in these mice (Huang et al., 1999; McEvilly et al.,1996). Transcripts for the p75 NGF receptor have been reported to be significantly decreased in mice lacking Brn3a(McEvilly et al., 1996),whilst immunohistochemistry for p75 protein has been reported as unchanged in mid-gestation knockout ganglia (Huang et al., 1999). In the present study, knockout levels of p75 mRNA were approximately 40% of wild type. Transcript levels for BDNF, previously reported to be reduced to undetectable levels at E12.5 in the trigeminal ganglia of Brn3a null mice(McEvilly et al., 1996), were found to be unchanged from controls in the present analysis. Loss of expression of the TrkB and TrkC neurotrophin receptors has also been reported in Brn3a null mice, but transcripts for the TrkB and TrkC receptors were not detected (absent call) in any genotype by the probe sets designed for these genes on the U74A array. This is an inconclusive result, which may reflect a problem in array design.

In addition to the trigeminal ganglion, Brn3a is expressed in neurons of the vestibulocochlear (VIII) ganglion complex, IX/X ganglion complex, and in the dorsal root ganglia (Fig. 2B). In order to verify the gene expression changes noted in the trigeminal array analysis, and to determine whether the trigeminal target genes are regulated elsewhere in the nervous system, we examined the expression of several Brn3a regulatory targets by in situ hybridization and immunohistochemistry in E13.5 embryos.

Among the Brn3a-regulated gene products related to neurotransmitter systems, examination of the 5HT3 receptor mRNA by in situ hybridization confirmed markedly increased expression in the trigeminal ganglion, the IX/X ganglion (Fig. 2C), and the cervical dorsal root ganglion (not shown). Conversely, the regulator of G-protein signaling RGS10 exhibited strong expression in the trigeminal ganglion, IX/X ganglion (Fig. 2D), and dorsal root ganglion (not shown) of control mice, which fell to background levels in Brn3a knockout embryos. However, the VIII ganglion did not show increased expression of 5HT3R in Brn3aknockout embryos, or endogenous expression of RGS10 in controls.

The expression patterns of the mediator of Ca²⁺ signaling,calretinin, the neuropeptides somatostatin and galanin, and the enzyme of catecholamine synthesis, tyrosine hydroxylase were examined in the trigeminal ganglion by immunohistochemistry. Consistent with cell-autonomous regulation by Brn3a, galanin and tyrosine hydroxylase co-localized with Brn3a protein in the trigeminal neurons of control ganglia(Fig. 2E,F), and the direction and approximate extent of the expression changes in each of these proteins was entirely consistent with the array results(Fig. 2G). We also examined the DRG and spinal cord for changes in the expression of these four proteins (data not shown). Galanin immunoreactivity was markedly decreased in the DRG, but no significant changes in calretinin or tyrosine hydroxylase were evident. Somatostatin immunoreactivity accumulated abnormally in the dorsal root entry zone of Brn3a knockout mice, a finding which may reflect either increased expression, or the failure of sensory axons to appropriately enter the CNS in these mutants (Eng et al.,2001), or both. No changes were noted in any of these markers in the Brn3a-expressing neurons of the dorsal spinal cord.

Several transcription factors were also prominent among the most changed transcripts in the array analysis. To verify the array results for the trigeminal ganglion, and examine the expression of these factors in other cranial sensory ganglia and the caudal CNS, we performed in situ hybridization for the increased transcripts GATA3, Irx1, Irx2, AP2b, MyoR, Math3(Fig. 3A), and NeuroD1 (not shown), and for the decreased transcripts HoxD1 and Runx1(Fig. 3B), in E13.5 wild-type and Brn3a knockout embryos. In each case the direction and magnitude of change in the in situ hybridization signal in the trigeminal ganglion correlated well with the array results.

Further examination of these transcripts in the cranial sensory ganglia clearly indicate a role for Brn3a in the coordinated regulation of gene expression in the sensory system. GATA3, Irx1, Irx2, MyoR and NeuroD1 were all expressed in the vestibulocochlear ganglion (VIII) complex in control mice,but were weakly expressed to undetectable in the trigeminal, IX and dorsal root ganglia. In mice lacking Brn3a, the expression of these factors was markedly increased in the trigeminal ganglion and IX/X complex(Fig. 3A). GATA3, MyoR and NeuroD1, but not the Irx transcripts, were also increased in the dorsal root ganglion (not shown). AP2β and Math3 were not detectable in the VIII ganglion of control mice, but showed a similar coordinated increase in expression in the trigeminal and IX ganglia in embryos lacking Brn3a.

The transcription factors HoxD1 and Runx1 showed decreased expression in the array analysis of Brn3a knockout mice. In situ hybridization for these transcripts confirmed markedly decreased expression in the trigeminal and IX ganglion. Endogenous expression of Runx1 in the VIII ganglion appeared to be less affected.

In addition to the sensory ganglia, Brn3a is expressed in specific neurons of the CNS, residing in the habenula, midbrain tectum and tegmentum,hindbrain, dorsal spinal cord and retina. The examination of the hindbrain region and spinal cord by in situ hybridization (Figs 2 and 3) did not indicate any obvious changes in the expression of neurotransmitters or transcription factors in the CNS of Brn3a knockout mice. However, in most areas of the CNS,Brn3a-expressing neurons have a scattered distribution, requiring methods of detection with cellular resolution to identify changes in target gene expression. For this reason we examined the CNS of embryos in more detail by immunohistochemistry for the increased gene products calretinin and somatostatin, and for the decreased gene products galanin and tyrosine hydroxylase.

In the midbrain and hindbrain, calretinin and Brn3a are expressed in adjacent but non-overlapping cell populations(Fig. 4A,B), while in the retina, a subset of neurons cells co-express these antigens(Fig. 4D,E). The expression of calretinin was not altered in either of these regions in the absence of Brn3a(Fig. 4C,F). Similarly,somatostatin was not ectopically expressed in the CNS of mice lacking Brn3a(Fig. 4G-K). Galanin and tyrosine hydroxylase were not co-expressed with Brn3a in the CNS as they are in the sensory system (Fig. 4L,M and data not shown), and thus could not be the targets of cell-autonomous regulation by Brn3a. Taken together, the in situ hybridization and immunohistochemical data for the targets of Brn3a regulation in the trigeminal ganglia demonstrate considerable conservation of the regulatory role of Brn3a in sensory neurons at different levels of the neural axis, but suggest a distinct role for Brn3a in the CNS.

In this study, we have presented a systematic examination of the changes in gene expression resulting from the loss of a key regulator of sensory neurogenesis. Although the list of significantly changed genes is fairly extensive, the specificity of the developmental effects of Brn3a are underscored by several findings: a majority of the changed transcripts encode proteins with known or hypothesized roles in sensory neuron development or function, a far greater number of neuron-specific genes did not significantly change (see Supplemental Data), and there were almost no significant changes in the expression of ubiquitously expressed or `housekeeping' genes.

Most of the genes with profoundly changed expression can be divided into three functional categories: neurotransmitter systems and ion channels,mediators of axonogenesis/synaptogenesis, and transcriptional regulators. Each of these classes of regulated transcripts may be related to the defects in axon growth and cell survival seen in Brn3a mutant mice, but it is likely that these changes in gene expression synergise to produce the Brn3a knockout phenotype, and that no single target gene is sufficient to account for the observed defects.

Beyond explaining the sensory phenotype of Brn3a knockout mice,two interesting generalizations may be made which encompass many of the genes with altered expression. First, in the absence of Brn3a, trigeminal development is retarded, in the sense that the expression of numerous markers of a mature sensory phenotype are reduced, and the developmental expression of factors that play a transient role in the early phases of differentiation is abnormally prolonged. Second, several transcription factors are expressed outside their normal axial level in the sensory ganglia, suggesting that Brn3a acts to spatially restrict their expression.

The array results clearly demonstrate that Brn3a has a major role in determining the neurotransmitter phenotype of the developing trigeminal ganglia. Expression of the neuropeptides PACAP and galanin and the NPY1 receptor are highly dependent on Brn3a, and the rate-limiting enzyme of catecholamine synthesis, tyrosine hydroxylase, is also significantly reduced in Brn3a knockouts. In contrast, the expression of somatostatin and the 5HT3A receptor are markedly increased. Studies in the developing rat have shown that somatostatin is strongly expressed throughout the sensory ganglia soon after neurogenesis, but by mid-gestation its expression is restricted to a relatively small subset of sensory neurons(Katz et al., 1992). Thus the increased expression of somatostatin at E13.5 is very likely to represent a failure in the normal developmental attenuation of this gene, consistent with the idea that Brn3a knockout ganglia exhibit a pervasive maturation defect.

Also notable are changes in the expression of sodium channels, including Scn6 and Scn9, which are markedly decreased in Brn3a knockout ganglia, and Scn10, which is moderately decreased (Table S2). Remarkably, these changes affect only those sodium channels that appear to have specific expression in the sensory nervous system(Goldin, 1999; Waxman et al., 1999),suggesting that Brn3a directly or indirectly coordinates expression of these channels. In contrast, expression levels of most neurotransmitter receptors,such as the GABA and glutamate receptors, and several classes of ion channels with wide expression in the CNS and PNS, are unchanged. Two other markedly changed genes, calretinin and the regulator of G-protein signaling RGS10, have putative roles in the modulation of neurotransmitter signals mediated by Ca²⁺-dependent and G-protein pathways, respectively. Altered expression of these genes may represent primary changes, or they may occur in an attempt to compensate homeostatically for other changes in neurotransmitter systems.

Mice lacking Brn3a have marked defects in sensory axon growth, including defasciculation of axon bundles and failure to innervate peripheral and central targets (Eng et al.,2001; Trieu et al.,2003). The transcripts for several proteins known to be involved in axon growth and synaptogenesis were significantly decreased in Brn3a null mice. Among the proteins in this category is advillin(pervin), an actin-binding protein with specific expression in sensory and sympathetic ganglia, which increases neurite outgrowth in cultured dorsal root ganglia (Ravenall et al.,2002). Apolipoprotein E knockout mice exhibit anatomical and functional defects in unmyelinated nerve fibers(Fullerton et al., 1998). Although this has been attributed to loss of ApoE expressed in associated glia, our results suggest that the defect may be intrinsic to sensory neurons.

Also decreased in Brn3a knockout ganglia were transcripts for the functionally interrelated proteins insulin-like growth factor 1 (IGF1) and insulin-like growth factor binding protein 5 (IGFBP5). Mice lacking IGF1 have abnormalities in sensory neurons (Gao et al., 1999), and show defective cortical dendritic growth(Cheng et al., 2003). IGFBP5 is a widely expressed protein whose role in vivo has not been clearly defined. However, it is highly expressed in the axon terminals of developing sensory neurons (Cheng et al., 1996),where it is frequently co-localized with IGF1, suggesting that it also has a role in axon growth. Because these proteins are known to interact, relative deficiencies in their expression may have a synergistic effect.

Another group of Brn3a-regulated proteins likely to have a role in axon growth are those involved in cell signaling and intracellular signal transduction. Transcripts with significantly changed expression include N-chimaerin, downstream of tyrosine kinase 4 (Dok4), the low affinity neurotrophin receptor p75, the small GTPases RAP (Ras family) and WRCH1 (Rho family), and Dusp6/MKP3. The expression and potential role of some of these factors in the sensory nervous system has been described; in other cases, the function of related proteins suggest that they may have significant and synergistic effects on axon growth.

Loss of Brn3a results in profound changes in the expression of several transcriptional regulators of various types, suggesting a web of cross-regulation between genes involved in sensory neurogenesis. The expression of a few transcription factors expressed late in sensory development, such as Runx1, were decreased in the absence of Brn3a, but the majority of the changes were increases, suggesting that Brn3a functions as a repressor of transcription factors that would be temporally or spatially inappropriate in the maturing trigeminal ganglion.

The clearest example of the role of Brn3a in restricting the spatial expression of other transcription factors is the ectopic expression of GATA3,Irx1, Irx2, NeuroD1 and MyoR/musculin in Brn3a knockout mice. These factors are all expressed in the developing vestibulocochlear ganglion in control embryos, and in the absence of Brn3a are markedly increased in the trigeminal and IX/X ganglia, demonstrating an expansion of the expression domain of these genes in both directions of the rostrocaudal axis. It is likely that some of the downstream changes in gene expression in Brn3a knockout ganglia are mediated by these factors, but current knowledge of their role in neural development is not sufficient to predict the effect of their mis-expression in the trigeminal ganglion. GATA3 has a known role in the development of motor neurons originating in rhombomere 4, which innervate the inner ear, and the inner ear itself(Karis et al., 2001). NeuroD1 is also required for normal development of the sensory neurons of the inner ear (Liu et al., 2000), and may have a cross-regulatory relationship with GATA3(Lawoko-Kerali et al., 2004). Although the role of the Irx genes in sensory development has not been described in mice, the zebrafish protein iro7, a possible paralogue of Irx1,is required for trigeminal placode development in fish(Itoh et al., 2002). The bHLH factor MyoR (musculin) is normally expressed in the developing facial muscles of the first branchial arch, which are innervated by trigeminal neurons, but not in the trigeminal ganglion itself (Lu et al., 2002). Our observation that MyoR is expressed in the developing auditory system is the first report of the sensory expression of this gene, and its role in neurogenesis is unknown.

Although it was not detected in the vestibulocochlear ganglion at this stage, AP2β showed a similar pattern of ectopic expression in the trigeminal and IX cranial ganglia in E13.5 Brn3a knockout embryos. AP2β is normally expressed in the embryonic hindbrain and spinal cord,but little is known about its role in neural development. The nervous system of AP2β mutant mice, which die from polycystic kidney disease,has no obvious abnormalities (Moser et al., 1997). However, mice lacking the related factor AP2α,which is highly expressed in migrating neural crest and in the developing sensory ganglia, exhibit extensive cranial abnormalities and dysgenesis of the cranial ganglia (Schorle et al.,1996). There is some evidence that AP2β is a weak transcriptional activator, and may oppose gene activation by AP2α(Bosher et al., 1996). Thus the increased expression of AP2β observed here may mimic some aspects of the loss of AP2α.

The increased expression of Math3 and NeuroD1 in Brn3a knockout trigeminal ganglia, together with decreased expression of the inhibitor of bHLH function Id1, suggest a marked increase in bHLH activity in the absence of Brn3a. Math3 and NeuroD1 have been characterized in the early development of the trigeminal ganglion (E9.0), where both factors appear to be downstream of the neurogenic HLH factor Ngn1 (Ma et al., 1998). Thus the increased expression of bHLH factors in Brn3a knockout mice may reflect the abnormal persistence of genes normally down-regulated as sensory development progresses. Although the loss of NeuroD1 or Math3 alone does not have an obvious effect on neurogenesis in the trigeminal (Tomita et al.,2000), the increased expression of multiple bHLH genes may have a synergistic effect in Brn3a knockout mice.

Embryonic day 13.5 was chosen for gene expression analysis because it precedes the extensive loss of sensory neurons observed at later stages in Brn3a knockout mice, and consistent with this, we did not observe altered expression of genes usually associated with cell death pathways, such as caspases or bcl2-family genes. Sensory cell death in mice lacking Brn3a occurs after these neurons normally become neurotrophin dependent, and the decreased expression of neurotrophins and their receptors in Brn3aknockout mice has been suggested as a cause of this mortality(Huang et al., 1999). We have previously reported that the expression of the TrkA neurotrophin receptor mRNA is moderately decreased in Brn3a knockout mice(Ma et al., 2003). This observation, and the decreased expression of the p75 low affinity NGF receptor shown here, are consistent with previous results(Huang et al., 1999; McEvilly et al., 1996). However, because the TrkA receptor is generally regarded as anti-apoptotic,and the p75 receptor as pro-apoptotic in sensory neurons(Huang and Reichardt, 2001),it is not obvious what net effect a moderate decrease in both receptors would have on cell survival. Given the severity of the axon growth defects in Brn3a knockout mice, another possibility is that excessive neural death occurs because of a failure to obtain target-derived neurotrophins, but this hypothesis has not been tested directly.

In the present study we have defined a set of genes regulated by Brn3a in sensory ganglia. This represents one of the first comprehensive descriptions of the in vivo regulatory targets for any factor regulating vertebrate neurogenesis. Like many developmental regulators, Brn3a is expressed in a highly specific, yet diverse set of neurons, including those of the retina,diencephalon, midbrain, spinal cord and sensory system, leading to the question of whether Brn3a regulates a common set of targets in these distinct locations. In the present study we have found little evidence that the targets of Brn3a regulation in the trigeminal ganglia are also regulated in the CNS or in the retina. A recent analysis of the regulatory targets of the closely related POU-factor Brn3b in the retina revealed few changed transcripts in common with the present study, despite the fact that the retinal ganglion cells in Brn3b knockout mice show a secondary loss of Brn3a(Mu et al., 2004); it also did not detect changes in the retinal target genes in sensory ganglia.

Even within the peripheral sensory system, Brn3a targets appear to be distinctly regulated in the vestibulocochlear ganglion when compared to the coordinated changes in expression in the trigeminal, IX, and dorsal root ganglia. The lack of change in trigeminal target genes in the vestibulocochlear ganglion cannot be attributed to functional redundancy of Brn3 genes. Although Brn3b is also expressed in the vestibulocochlear system,the loss of Brn3a expression in the vestibulocochlear ganglion also leads to diminished expression of Brn3b, and results in significant defects in cochlear innervation (Huang et al.,2001). Thus it appears probable that Brn3a will have at least a partially distinct set of regulatory targets in the auditory system.

The genes downstream from Brn3a in the sensory ganglia are very likely to include targets that are regulated directly, and regulated indirectly by the several other transcription factors that change expression in the absence of Brn3a. One of the surprising features of the current study is the large number of markedly increased transcripts in the knockout ganglia, implying direct or indirect transcriptional repression by Brn3a. Although nearly all prior studies of the transcriptional activity of Brn3a have proceeded from the assumption that it is a positive regulator of gene expression, we have recently shown that Brn3a is a direct repressor of its own expression in the trigeminal ganglion in vivo (Trieu et al.,2003). The recent study of the target genes of Brn3b in the retina showed mainly decreased expression of downstream transcripts(Mu et al., 2004), but this study was conducted with a retina-specific cDNA array, which would be unlikely to include strongly increased transcripts which have low levels of expression in the normal retina. Thus it is plausible that Brn3a, and perhaps other factors in this class, exert their direct effects by transcriptional repression.

Identifying the regulatory targets of neural transcription factors is an essential component of understanding developmental pathways in the nervous system. Here we have demonstrated an extensive program of gene regulation mediated by one such factor. Future studies of this kind will be greatly facilitated by the availability of more complete gene expression arrays based on genomic sequences rather than cDNA libraries. Additional data about the location of the transcription units in the mouse genome, and better information about the DNA recognition properties of the various transcription factor classes, will help to distinguish direct from secondary targets. In addition, the confirmation of direct regulation by chromatin immunoprecipitation may be facilitated by combining this method with array technology or other high throughput methods(Ren et al., 2002). These anticipated technical advances should in principle allow the identification of a complete set of regulatory targets for any transcription factor in any tissue.

We would like to acknowledge Bob Stuart, Bill Wachsman and Lutfunnessa Shireen of the San Diego VA microarray core facility for assistance with microarray technology. We would also like to thank the following investigators and their associates for in situ hybridization probes: Reinhard Buettner,Peter Gruss, Alar Karis, Allan Basbaum, Quifu Ma, Ryoichiro Kageyama, Denis Duboule, Stephen Gold and Eric Olson. In addition, we thank John Kelsoe for collaborative use of the Celera Genomics database, and Ying Liu, Jane Johnson and William Snider for helpful comments on the manuscript. Supported in part by Department of Veterans Affairs MERIT funding and VISN 22 MIRECC (E.E.T.),and NIH awards HD33442 and MH065496. E.E.T. and N.F. are a NARSAD Investigators.