Identification of Pax2-regulated genes by expression profiling of the mid-hindbrain organizer region

The midbrain and cerebellum develop from an organizing center that is formed at the junction between the embryonic midbrain and hindbrain, known as the isthmus. This isthmic organizer (IsO) was discovered because of its property of inducing an ectopic midbrain or cerebellum, when transplanted into the chick diencephalon or hindbrain, respectively (reviewed by Liu and Joyner, 2001a; Wurst and Bally-Cuif, 2001). The IsO activity recruits the surrounding tissue into either a midbrain or cerebellum fate by controlling cell survival, proliferation and differentiation along the anteroposterior axis of the mid-hindbrain region. The formation of the IsO is the result of complex cross-regulatory interactions between transcription factors (Otx, Gbx, Pax and En) and secreted proteins (Wnts and Fgfs), culminating in the expression of the signaling molecule Fgf8 at the mid-hindbrain boundary(Liu and Joyner, 2001a; Wurst and Bally-Cuif, 2001; Ye et al., 2001). Fgf8 is the central mediator of IsO activity, as it is both necessary and sufficient for inducing midbrain and cerebellum development(Crossley et al., 1996; Chi et al., 2003). Once formed,the IsO is maintained by a positive feedback loop involving multiple mid-hindbrain-specific regulators. Consequently, the IsO is lost upon individual mutation of these regulators, whereas ectopic expression of a single factor activates most other components of the regulatory cascade(Nakamura, 2001). Owing to this interdependence, the hierarchical relationship among the different regulators remains largely elusive during the maintenance phase of IsO activity (Liu and Joyner,2001a; Wurst and Bally-Cuif,2001).

The initiation of IsO development crucially depends on the transcription factor Pax2 (Favor et al.,1996; Brand et al.,1996), which shares similar DNA-binding and transactivation functions with Pax5 and Pax8 of the same paired domain protein subfamily(Kozmik et al., 1993; Dörfler and Busslinger,1996). Pax2 is the earliest known gene to be expressed throughout the prospective mid-hindbrain region in late gastrula embryos(Rowitch and McMahon, 1995). The initially broad expression pattern of Pax2 is progressively refined to a narrow ring centered at the mid-hindbrain boundary by embryonic day 9.5, while the related Pax5 and Pax8 genes are activated in the same region at 3-4 and 6-7 somites, respectively(Urbánek et al., 1994; Rowitch and McMahon, 1995; Pfeffer et al., 1998). Consistent with this sequential gene induction, mutation of the Pax2gene leads to the loss of the midbrain and cerebellum in mouse and zebrafish embryos (Favor et al., 1996; Brand et al., 1996; Bouchard et al., 2000), whereas the inactivation of Pax5 or Pax8 results in a mild cerebellar midline defect or no brain phenotype at all(Urbánek et al., 1994; Mansouri et al., 1998). The severe mid-hindbrain deletion is, however, only observed in Pax2^–/– mouse embryos on the C3H/He genetic background (Bouchard et al.,2000), where the compensating Pax5 and Pax8genes fail to be activated at the mid-hindbrain boundary(Pfeffer et al., 2000; Ye et al., 2001) similar to the Pax2.1 (noi) mutant embryos of the zebrafish(Pfeffer et al., 1998). In the absence of Pax2, Otx2, Gbx2 and Wnt1 are normally transcribed at early somite stages, while the expression of En1 is reduced in the developing mid-hindbrain region(Ye et al., 2001). Importantly, Fgf8 expression is never initiated at the mid-hindbrain boundary of Pax2^–/– C3H/He embryos(Ye et al., 2001), resulting in the complete absence of IsO activity and subsequent apoptotic loss of the mid-hindbrain tissue starting at the 12-somite stage(Pfeffer et al., 2000; Chi et al., 2003).

To further investigate the role of Pax2 at the onset of mid-hindbrain development, we searched for novel Pax2-regulated genes by gene expression profiling of mid-hindbrain cells isolated by FACS sorting from wild-type and Pax2^–/– E8.5 embryos. This unbiased approach identified the En2, Brn1 (Pou3f3 – Mouse Genome Informatics), Sef (Il17rd – Mouse Genome Informatics), Tapp1 (Plekha1 – Mouse Genome Informatics) and non-coding Ncrms genes as genetic Pax2 targets that are totally dependent on Pax2 function for their expression in the mid-hindbrain region. The transcription factors En2 and Brn1, as well as the signaling modifiers Sef and Tapp1, implicate Pax2 in the establishment of distinct transcriptional programs and the control of intracellular signaling during mid-hindbrain development. Biochemical and transgenic analyses demonstrated that Pax2 directly activates the mid-hindbrain-specific expression of Brn1 by interacting with two functional Pax2/5/8-binding sites in the promoter and an upstream regulatory element of the Brn1 gene. Moreover, ectopic expression of a dominant-negative Brn1 protein in chick embryos implicated Brn1 as a novel regulator of Fgf8 expression. The identification of new Pax2-regulated genes has thus provided important insight into the role of Pax2 in mid-hindbrain development.

Pax2^+/– and Pax5^+/–mice as well as the Pax2^GFP mice carrying the BAC transgene 30 were maintained on the C3H/He background and genotyped as described (Urbánek et al.,1994; Bouchard et al.,2000; Pfeffer et al.,2002).

The mid-hindbrain region of GFP⁺ E8.5 embryos from Pax2^+/– Pax2^GFP intercrosses was dissected with 26-gauge needles and dissociated into single cells at 37°C for 15 minutes in 24-well plates containing 500 μl of 1% trypsin in PBS. The reaction was stopped by transferring the single-cell suspension into 4 ml of cold DMEM containing 10% fetal calf serum followed by centrifugation and resuspension in phenol red-free DMEM containing 10% fetal calf serum and 1μg/ml propidium iodide (PI). Live PI^– GFP⁺cells of individual embryos were sorted with a FACSVantage TSO flow-cytometer(Becton-Dickinson) directly into the Trizol Reagent (Gibco-BRL), vortexed for 1 minute and then stored in liquid nitrogen. This sorting protocol yielded 5000-10,000 GFP⁺ cells per embryo. Total RNA from selected samples was submitted to linear amplification as described(Hoffmann et al., 2003) with some modifications. Briefly, the total RNA from a minimum of 5000 cells was reverse-transcribed with an oligonucleotide consisting of d(T)₁₅linked to a T7 RNA polymerase recognition site. Following second-strand synthesis, the samples were amplified by T7 polymerase-mediated in vitro transcription. The resulting aRNA was reverse-transcribed using random nonamer oligonucleotides [pd(N)₉] and used for a second round of cDNA synthesis and in vitro transcription. Two rounds of amplification from 5000 cells typically yielded 30 to 80 μg of aRNA.

The cDNA microarray screening was essentially performed as described(Cheung et al., 1999). A detailed description of the method used can be found as supplementary information. Briefly, aRNA was reverse-transcribed into cDNA in the presence of Cy3-dUTP or Cy5-dUTP using the Gibco-RT kit. The Cy3- and Cy5-labeled cDNA probes were pooled and ethanol-precipitated together with poly-dA, tRNA and mouse Cot.1 DNA. The precipitated cDNA probes were washed, prehybridized at 50°C for 1 hour in a solution containing 35% formamide, 4 ×SSPE,0.5% SDS, 5 × Denhardt's solution and 10 μg/ml denatured salmon sperm DNA and then added to microarray slides for overnight hybridization at 50°C. Post-hybridization washes were carried out for 10 minutes in 0.2×SSC, 0.1% SDS and 10 min in 0.2 ×SSC. The slides were dried and scanned using an Axon GenePix 4000 scanner. The hybridization results were normalized using the marray-package of Bioconductor(http://www.bioconductor.org)and the `Print Tip Loess' algorithm (Yang et al., 2002). The cDNA microarrays contained 26,000 spotted EST clones (11,000 BMAP clones from Research Genetics and 15,000 NIA clones from the National Institute of Aging), which corresponded to 17,000 UniGene clusters.

Embryos were dissected and processed for in situ hybridization with digoxigenin-UTP-labeled RNA probes as described(Henrique et al., 1995). The En2 probe was previously described(Davis et al., 1988). The Sef probe contained a 750 bp cDNA sequence extending from the PCR oligonucleotide 5′-GGAGCCTGACTGGTTTGAGAA-3′ to the NdeI site. The Tapp1 and Ncrms probes were derived from the identified ESTs (Tapp1, BC020017; Ncrms, BE655589). The mouse Brn1 probe (643 bp) was cloned into the pGEM-Teasy plasmid(Promega) following RT-PCR from E10.5 head cDNA using the primers 5′-GGGCAGAAGTCAAGGGAAGTG-3′ and 5′-TGGCGTCGTCGGTGGAGAACA-3′ and the chick Brn1 probe (429 bp) following RT-PCR amplification from chick embryo RNA with the primers 5′-ATGGT(G/C)CAGAG(C/T)GACTTCATGCAGGG-3′ and 5′-GCT(C/T/G)AGCAT(G/A/T)CCGTT(C/T)AC(C/A)GTGAA-3′. The partial chick Brn1 cDNA sequence was submitted to GenBank (Accession Number DQ002393).

The transcriptional start sites of the Brn1 gene were identified by 5′-RACE, using the SMART RACE cDNA amplification kit (BD Bioscience)according to the manufacturer's instructions. RNA isolated from the head of E10.5 embryos was reverse-transcribed into cDNA with the Brn1-specific primer 5′-GCTTCCACGGCAGCGGCGGCGGCAGCAG-3′followed by PCR amplification with the oligonucleotides 5′-ACGGGAGACAACAAAGGACGAAGCGGTTCC-3′ (outer) and 5′-GGAAGAAGAGTGCATTGGTGGAGGTGGAGA-3′ (inner) in combination with the primers provided with the RACE kit.

The Pax2 protein was synthesized by coupled in vitro transcription/translation and used for EMSA analysis with published CD19 and Blnk oligonucleotide probes as described(Kozmik et al., 1992; Schebesta et al., 2002). The competitor fragments C, D and P were cloned by PCR with the following primers:

• C, 5′-CAGACAAAACAATCACACTCC-3′ and 5′-GGGGAGGATAGGAACAGAGCC-3′

• D, 5′-GGCTGCGAGGCTGCTGCTGAG-3′ and 5′-GTTTTGCGAGGTGGCTGTGAC-3′

• P, 5′-ACAACAGATTTCCAGCTTCTA-3′ and 5′-CTCTCCCTCTCTCCTTCTCTC-3′.

The following double-stranded Brn1 oligonucleotides were used as competitor DNA:

• Da, 5′-tcagATTCGGAGCACACCGACCGCCGGGTTACGTTCTCGGCTGCTGCTT-3′ and 5′-tcagAAGCAGCAGCCGAGAACGTAACCCGGCGGTCGGTGTGCTCCGAAT-3′;

• Db, 5′-tcagTGCTTGGACTAGAAACTGCAGATTGCGGTCCGGTGCCCTGCAGCT-3′ and 5′-tcagAGCTGCAGGGCACCGGACCGCAATCTGCAGTTTCTAGTCCAAGCA-3′;

• Dc, 5′-tcagTTCTCTTTTTTCTTGGTTCGCTGAGGTTCCTCTGTatcGCGTTTC-3′ and 5′-tcagGAAACGCgatACAGAGGAACCTCAGCGAACCAAGAAAAAAGAGAA-3′;

• Dd, 5′-tcagTCCTCTGTatcGCGTTTCCGCTTGGCCGCGTCGTCCCCCCCCCCC-3′ and 5′-tcagGGGGGGGGGGGACGACGCGGCCAAGCGGAAACGCgatACAGAGGA-3′;

• Ddm, 5′-tcagTCCTCTGTatcGCGTTTCCatTTGGCCGCGTCGTCCCCCCCCCCC-3′and 5′-tcagGGGGGGGGGGGACGACGCGGCCAAatGGAAACGCgatACAGAGGA-3′;

• Pa, 5′-tcagGGGGGACAACAGATTTCCAGCTTCTACGACGCTCTGtCaAAATTA-3′ and 5′-tcagTAATTTtGaCAGAGCGTCGTAGAAGCTGGAAATCTGTTGTCCCCC-3′;

• Pb, 5′-tcagGATTTCCAGtcTCTACGACGCTCTGCCTAAATTAAAAAGCAACCA-3′ and 5′-tcagTGGTTGCTTTTTAATTTAGGCAGAGCGTCGTAGAgaCTGGAAATC-3′;

• Pc, 5′-tcagCTGtCaAAATTAAAAAGCAACCAATCGGAACGGCCGGAAGGGGGG-3′ and 5′-tcagCCCCCCTTCCGGCCGTTCCGATTGGTTGCTTTTTAATTTtGaCAG-3′;

• Pcm, 5′-tcagCTGtCaAAATTAAAAAGCAACCAATCaGtACGGCCGGAAGGGGGG-3′and 5′-tcagCCCCCCTTCCGGCCGTaCtGATTGGTTGCTTTTTAATTTtGaCAG-3′.

lacZ transgenes were generated by insertion of a 3.2-kb BglII-NotI fragment or a 6.2-kb AflII-NotI fragment from the 5′ flanking region of Brn1 into the BglII-NotI sites of pTRAP-PL, which is a modified version of pTRAP (Pfeffer et al., 2000)lacking the minimal promoter. The mutant lacZ transgenes were obtained by site-directed mutagenesis, using the QuikChange kit (Stratagene)together with the following oligonucleotides:

• Dd, 5′-CTCTGTGCGGCGTTTCCatTTGGCCGCGTCGTCCCCCCCC-3′ and 5′-GGGGGGGGACGACGCGGCCAAatGGAAACGCCGCACAGAG-3′;

• Pc, 5′-CCTAAATTAAAAAGCAACCAATCaGtACGGCCGGAAGGG-3′ and 5′-CCCTTCCGGCCGTaCtGATTGGTTGCTTTTTAATTTAGGC-3′.

Plasmid-free DNA of the transgene was injected into pronuclei followed by the transfer of zygotes into pseudopregnant females. Transgenic embryos were stained for β-galactosidase activity as described(Pfeffer et al., 2000).

cDNAs for electroporation were cloned into the expression vector pCIE containing a chick β-actin promoter, a polylinker and an internal ribosomal entry sequence (IRES) linked to a GFP gene. Full-length mouse Otx2, chick Gbx2 and chick Pax2 constructs have been described(Ye et al., 2001). VP16-Brn1 and EnR-Brn1 constructs were generated by fusing the POU homeodomain of rat Brn1 (amino acids 301-497) in-frame C-terminal to the transcriptional activator domain of VP16 or the repressor domain of DrosophilaEngrailed (amino acids 1-298), respectively. These constructs were unilaterally electroporated into chick embryos at HH stage 8-10 as described(Hynes et al., 2000). Briefly,DNA at 3-6 mg/ml was microinjected into the central canal of the neural tube,and platinum electrodes flanking the neural tube delivered six square pulses of 28 V with a duration of 40 mseconds and an interpulse interval of 45 mseconds. Two days later, chick embryos were analyzed by in situ hybridization.

As Pax2 is an essential regulator of midbrain and cerebellum development(Favor et al., 1996; Bouchard et al., 2000; Ye et al., 2001), we wanted to further investigate the function of Pax2 in mid-hindbrain formation by identifying novel target genes by cDNA microarray screening. This approach relies on the detection of gene expression difference in the mid-hindbrain region of wild-type and Pax2^–/– embryos, but required three technical difficulties to be addressed. First, the prospective mid-hindbrain tissue, though initially formed in Pax2^–/– embryos on the C3H/He strain background, rapidly degenerates starting at the 12-somite stage(Pfeffer et al., 2000) owing to failed formation of the IsO (Ye et al.,2001). We therefore dissected the mid-hindbrain region at the four- and six-somite stage before the onset of tissue degeneration in Pax2^–/– embryos. Second, to enrich for Pax2-expressing cells in the microdissected mid-hindbrain tissue, we took advantage of the Pax2^GFP BAC transgene(Pfeffer et al., 2002), which expresses a GFP gene under the control of the Pax2 locus equally well in Pax2^–/– and wild-type embryos(Fig. 1A-C; data not shown). Fluorescence-activated cell sorting (FACS) was thus used to isolate 5000-10,000 live GFP⁺ cells from the mid-hindbrain region of individual Pax2^GFP embryos(Fig. 1C). Last, because the total RNA isolated from the sorted GFP⁺ cells of an individual embryo was very low (5-10 ng), we linearly amplified the poly(A)⁺RNA by two consecutive cycles of cDNA synthesis and in vitro transcription(Wang et al., 2000), which resulted in 30-80 μg of aRNA corresponding to an estimated 10⁶-fold mRNA amplification.

As the expression of mid-hindbrain-specific genes is upregulated during somitogenesis (Wurst and Bally-Cuif,2001), we determined the gene expression increase during mid-hindbrain development and used this information, in addition to the genotype comparison, as a second criterion for the identification of Pax2-regulated genes. To this end, we prepared aRNA from Pax2^GFP-expressing mid-hindbrain cells of control(wild-type or Pax2^+/–) embryos at the 0-2 somite and 8-9 somite stages. The different aRNAs were reverse-transcribed in the presence of Cy3-dUTP or Cy5-dUTP, and the labeled cDNA probes were hybridized to microarrays, which contain 26,000 cDNA clones corresponding to 17,000 UniGene clusters. Only genes with expression levels that were more than fourfold above background were chosen for further analysis. Among the selected 13,200 ESTs, putative Pax2-regulated genes were identified (1) by an expression ratio of more than 1.7 in at least one genotype comparison of six-somite-stage embryos (103 ESTs) and (2) by an expression difference of more than 2.0 between the 0-2 and 8-9 somite stages (168 ESTs). Selection according to both criteria resulted in 12 candidate genes, including the known Pax2 target gene Pax5 (Pfeffer et al., 2000). Six of these genes (shown in Table 1) could subsequently be validated as Pax2-activated genes by in situ hybridization analysis (see below).

As an initial step of validating the microarray results, we investigated whether the candidate genes are expressed in a similar pattern as Pax2 in the developing mid-hindbrain region at embryonic (E) day 8.5 and 9.5. In situ hybridization analysis of the 11 new candidate genes revealed that the expression of six of them could either not be detected or was not specifically localized to the developing mid-hindbrain region (data not shown). By contrast, the remaining five genes were expressed in the mid-hindbrain domain of wild-type embryos at E8.5 and E9.5(Fig. 2). Two of these genes, En2 and Brn1 (Pou3f3), code for homeodomain and POU domain transcription factors, respectively(Joyner and Martin, 1987; Hara et al., 1992). Sef and Tapp1 code for intracellular modulators of Fgf and phosphatidylinositol signaling, respectively(Dowler et al., 2000; Fürthauer et al., 2002; Tsang et al., 2002), and Ncrms corresponds to a non-coding RNA gene that is highly expressed in alveolar rhabdomyosarcoma (Chan et al.,2002).

The five genes En2, Brn1, Sef, Tapp1 and Ncrms are expressed at E8.5 in a broad domain corresponding to the prospective midbrain and anterior hindbrain (Fig. 2C,E,G,I,K), which reflects Pax2 expression at this developmental stage (Fig. 2A). At E9.5, the expression patterns of the putative Pax2-regulated genes have started to diverge in the mid-hindbrain region, while novel expression domains have emerged in other parts of the embryo. At this stage, only Ncrmsis expressed like Pax2 in a narrow stripe centered at the mid-hindbrain boundary (Fig. 2B,L). En2 and Sef are broadly expressed in the posterior midbrain and anterior hindbrain(Fig. 2D,H), as previously published (Davis et al., 1988; Fürthauer et al., 2002). Tapp1 expression is observed in the anterior midbrain(Fig. 2J), whereas Brn1 is strongly expressed from the forebrain throughout the entire midbrain to the hindbrain (Fig. 2F). In summary, these expression data suggest that Pax2 may control the initiation, but not the maintenance of expression of the candidate genes in the mid-hindbrain boundary region.

We directly tested this hypothesis by comparing the expression pattern of the five putative Pax2 target genes in wild-type and Pax2^–/– embryo at the five- to eight-somite stage (E8.5). All five genes failed to be expressed in the developing mid-hindbrain region of Pax2^–/– embryos(Fig. 3B,D,F,H,J), in contrast to stage-matched control embryos (Fig. 3A,C,E,G,I), while other expression domains remained unaffected by the Pax2 mutation. These results unequivocally demonstrate that Pax2 controls the mid-hindbrain-specific expression of En2, Brn1, Sef,Tapp1 and Ncrms during early somitogenesis.

The transcription factor Brn1 is a member of the class III POU protein family (Ryan and Rosenfeld,1997) and has been implicated in patterning of the cerebellum(Sugitani et al., 2002). As the cerebellum develops from the Pax2-dependent expression domain of Brn1, we selected this gene for biochemical and transgenic analysis of its regulatory elements to determine whether Brn1 is a direct transcriptional target of Pax2. We first searched for conserved sequences within the human and mouse genomic sequences that extend from –20 kb upstream to +15 kb downstream of the Brn1 translation start site. The highest sequence conservation outside of the coding region was found in a 6 kb sequence located immediately upstream of the Brn1 start codon. This region contains five elements (A, B, C, D and P) with an evolutionary conservation of more than 93% (Fig. 4A; see Fig. S1 in the supplementary material). Four of these elements (B, C, D and P) are even conserved between the Brn1 genes of mammals and the pufferfish Fugu rubripes(Fig. 4A).

Gene promoters are often located within CpG islands(Antequera and Bird, 1999),which are also present in the first 4 kb upstream of the Brn1 start codon (Fig. 4A). The PromoterInspector program (Scherf et al.,2000) identified elements D and P as potential promoter regions. To test this possibility, we determined the start sites of Brn1transcription by using primers located in the two putative 5′untranslated regions for 5′-RACE analysis of E10.5 head RNA. PCR fragments could readily be amplified with an element P primer(Fig. 4B) in contrast to element D sequences (data not shown). Cloning and sequence of 63 RACE products demonstrated that the 5′ ends of 41 clones clustered at five sites spanning a 17 bp region within element P(Fig. 4C). As expected for heterogeneous transcription initiation, no TATA box could be found upstream of these start sites, which are located between positions –1400 and–1383 relative to the initiation codon(Fig. 6D). Taken together,these results identified a single Brn1 promoter, giving rise to heterogeneous transcription initiation in embryonic brain cells.

The binding of Pax2 to the different conserved regions of Brn1 was next assessed by electrophoretic mobility shift assay (EMSA). To this end, we cloned the conserved Brn1 sequences and used them as competitor DNA to prevent the binding of in vitro translated Pax2 protein to a labeled oligonucleotide containing the high-affinity Pax2/5/8-binding site of the CD19 promoter (Kozmik et al.,1992). In addition, we used the high-affinity Pax2/5/8-binding site of the Blnk promoter(Schebesta et al., 2002) as a reference sequence for comparing the competition strength of the different conserved Brn1 elements. This competition assay revealed the presence of Pax2-binding sites in elements P and D, but not in the conserved regions A,B or C (Fig. 5A; data not shown). Several potential Pax2-binding sites were identified within elements P and D (Fig. 5B,C) by comparison with the consensus Pax2/5/8 recognition sequences(Fig. 5D)(Czerny and Busslinger, 1995). Competition analysis with oligonucleotides of individual candidate sites revealed a single Pax2-binding sequence (Dd) within element D(Fig. 5B). The interaction of Pax2 with promoter element P was further investigated by comparing three overlapping subfragments (P1-P3) in the competition assay, which mapped the Pax2-binding activity to the 5′ region of fragment P1(Fig. 5C). This region contains three candidate Pax2-binding sequence, only one of which (site Pc) was able to interact with Pax2 (Fig. 5C). Importantly, the recognition sequences Dd and Pc bound Pax2 with similar efficiency as the high-affinity binding site of Blnk(Fig. 5B,C) consistent with the fact that the Dd and Pc sites match the consensus recognition sequence at 13 out of 15 positions (Fig. 5D). Moreover, Pax2 binding to these two sites was completely abolished by mutating two consensus nucleotides in each binding site (Ddm, Pcm; Fig. 5B,C). In summary, these experiments identified two high-affinity Pax2-binding sites, which are located 134 (Pc) and 1413 (Dd) bp upstream of the major transcription initiation site of Brn1.

We next investigated by transgenic analysis whether the Pax2-binding sites Dd and Pc are important for Brn1 expression in the developing mid-hindbrain region. For this, we generated the transgenes 3.2wt-lacZ and 6.2wt-lacZ by inserting 3.2 kb and 6.2 kb 5′ flanking sequences of Brn1 (starting at position –59 relative to the ATG codon) upstream of a lacZ reporter gene. Transgenic embryos were generated by pronuclear DNA injection and analyzed by X-gal staining for lacZ expression at E9.5. The shorter 3.2wt-lacZ transgene was unable to drive reporter gene expression,indicating that the promoter (P) and upstream element D with their Pax2-binding sites are not sufficient for activating Brn1 expression in the embryo (not shown). By contrast, the 6.2wt-lacZ transgene containing the promoter (P) and all four conserved upstream elements (A-D)gave rise to localized lacZ expression in the posterior forebrain(diencephalon), mid-hindbrain boundary region and spinal cord(Fig. 6A). This expression pattern differs from that of the endogenous Brn1 gene at E9.5, as the 6.2wt-lacZ transgene failed to be expressed in the mesonephros and throughout the entire forebrain-hindbrain region (compare Fig. 2F with Fig. 6A). Hence, the 6.2wt-lacZ transgene contains the control elements for initiating Brn1 expression in the mid-hindbrain boundary region, while lacking regulatory sequences for maintaining Brn1 expression throughout the entire midbrain and hindbrain. To further study the role of the Pax2-binding sites Dd and Pc in the initiation of mid-hindbrain-specific Brn1expression, we mutated each Pax2 recognition sequence individually in the 6.2wt-lacZ transgene by introducing the two-nucleotide substitutions that abrogate Pax2 binding (Fig. 5B-D). Both mutant transgenes, 6.2Ddm-lacZ and 6.2Pcm-lacZ, failed to be expressed above the detection limit in the mid-hindbrain boundary region, although they gave rise to strongβ-galactosidase staining in the diencephalon of the forebrain and along the entire spinal cord. The two high-affinity Pax2-binding sites in the promoter and upstream element D are therefore essential for initiating the mid-hindbrain-specific expression of Brn1 during early somitogenesis. These data unequivocally identify the Brn1 gene as a direct target of Pax2 during mid-hindbrain development.

In addition to Brn1 (Pou3f3), the class III POU protein family consists of Brn2 (Pou3f2 – Mouse Genome Informatics), Brn4 (Pou3f4 – Mouse Genome Informatics) and Tst1 (Pou3f1 – Mouse Genome Informatics). In situ hybridization analyses revealed strong Brn2 and weak Brn4 expression in the mid-hindbrain domain of E8.5 embryos (see Fig. S2 in the supplementary material), whereas Tst1 was not expressed in this CNS region (data not shown). Moreover, Brn2 expression was downregulated but not lost in the mid-hindbrain domain of Pax2^–/– embryos (see Fig. S2 in the supplementary material). Consistent with these overlapping expression patterns, brain development is largely normal in Brn1 or Brn2 single-mutant mice(Schonemann et al., 1995; Nakai et al., 1995; Nakai et al., 2003), whereas double-mutant mice show a severe defect in cerebellum patterning(Sugitani et al., 2002). To specifically analyze the role of Brn1 in mid-hindbrain development, we next performed ectopic expression experiments in chick embryos by in ovo electroporation. To investigate the expression pattern of the endogenous Brn1 gene, we cloned a partial chick Brn1 cDNA(Fig. 7A) and used it as a probe for in situ hybridization of chick embryos at HH stages 14/15 and 19/20(Fig. 7B,C). These embryos expressed Brn1 throughout the midbrain, hindbrain and spinal cord as well as in the mesonephros (Fig. 7B,C), which resembles the corresponding expression pattern in the mouse embryo (Fig. 2F). As Fgf8 is the central mediator of IsO activity(Crossley et al., 1996), we used Fgf8 expression as a read-out for the chick embryo electroporation experiments. Previous studies have demonstrated that ectopic Pax2 expression could efficiently induce endogenous Fgf8transcription only in Gbx2⁺ hindbrain cells adjacent to ectopic Otx2-expressing cells (Ye et al.,2001) (Fig. 7F). However, combined expression of rat Brn1 and mouse Otx2 failed to induced Fgf8 transcription (Fig. 7D), possibly because endogenous Brn1 was already expressed in the hindbrain of electroporated chick embryos(Fig. 7B,C). Next, we fused the VP16 transactivation or Engrailed repression (EnR) sequences to the DNA-binding POU domain of rat Brn1 in an attempt to generate dominant-active or -negative Brn1 proteins, respectively. Whereas electroporation of VP16-Brn1 together with Otx2 had no effect on Fgf8 expression(Fig. 7E), ectopic expression of EnR-Brn1 and Otx2 strongly induced Fgf8 transcription in the chick hindbrain. Hence, these data implicate Brn1 in the regulation of Fgf8expression during mid-hindbrain development.

Pax2 plays a key role in the formation of the isthmic organizer (IsO) that controls midbrain and cerebellum development(Favor et al., 1996; Bouchard et al., 2000; Ye et al., 2001). Despite the importance of Pax2 for mid-hindbrain patterning, relatively little is known about target genes that are activated by Pax2 at the onset of mid-hindbrain development. We have previously shown that Pax2 is essential for initiating expression of the closely related Pax5 and Pax8 genes,indicating that the inactivation of Pax2 is equivalent to mutation of all three Pax2/5/8 family members at the mid-hindbrain boundary(Pfeffer et al., 1998; Pfeffer et al., 2000; Ye et al., 2001). Pax2 is furthermore necessary and sufficient for inducing the expression of the IsO signal Fgf8 (Ye et al., 2001). Here, we have used gene expression profiling of mid-hindbrain cells from wild-type and Pax2^–/– embryos as a more systematic strategy for identifying Pax2-regulated genes. This approach relies on transgenic GFP labeling and FACS sorting of Pax2-expressing cells followed by linear RNA amplification, probe preparation and cDNA microarray screening. In this way, we identified five genes, En2, Brn1, Sef, Tapp1 and Ncrms, which are expressed in the developing mid-hindbrain region under the control of Pax2. The molecular nature of these new target genes implicates Pax2 in the control of intracellular signaling and the establishment of transcription factor networks in the mid-hindbrain region.

Fgf receptor stimulation by the IsO signal Fgf8 activates the Ras/mitogen-activated protein kinase (MAPK) pathway(Kouhara et al., 1997; Corson et al., 2003), which controls cell proliferation and differentiation(Marshall, 1995). Fgf8 signaling also activates the expression of the transmembrane protein Sef(similar expression to Fgf genes), which acts as a negative feedback regulator to limit the duration of Ras-MAPK signaling by inhibiting the tyrosine phosphorylation of Fgf receptors(Fürthauer et al., 2002; Tsang et al., 2002; Kovalenko et al., 2003). Here,we have demonstrated that the expression of Sef fails to be induced in the mid-hindbrain region of Pax2^–/– mouse embryos, similar to Pax2.1 (noi) mutant embryos of the zebrafish (Tsang et al.,2002). As Sef expression is also not activated at the mid-hindbrain boundary of Fgf8 (ace) mutant embryos(Fürthauer et al., 2002),it is possible that the absence of Sef transcripts in Pax2^–/– embryos is an indirect consequence of the failed induction of Fgf8 expression(Ye et al., 2001). However, it is equally likely that Pax2 directly activates the Sef gene in cooperation with Fgf signaling.

Pax2 also regulates the midbrain-specific expression of the tandem PH-domain-containing protein 1 (Tapp1) gene. Tapp1 was identified as an adaptor molecule that specifically binds to the lipid phosphatidylinositol 3,4-bisphosphate [PI(3,4)P₂] via its C-terminal pleckstrin homology (PH) domain (Dowler et al.,2000). PI(3,4)P₂ is generated by the inositol 5′-phosphatase SHIP from PI(3,4,5)P₃, which in turn is produced through phosphorylation of PI(4,5)P₂ by the phosphatidylinositol 3′-kinase (PI3K)(Rohrschneider et al., 2000). Both PI3K and SHIP are activated by stimulatory and inhibitory tyrosine kinase receptors, respectively, in agreement with the role of their second messengers PI(3,4,5)P₃ and PI(3,4)P₂ in recruiting different PH domain-containing effector proteins to the plasma membrane (Rohrschneider et al.,2000). The PI(3,4,5)P₃-dependent recruitment and activation of the Akt/PKB, PDK1 and Btk kinases promotes cell survival and proliferation (Rohrschneider et al.,2000). SHIP antagonizes these pathways by metabolizing the lipid ligand PI(3,4,5)P₃ of these kinases to PI(3,4)P₂, which functions as a membrane docking site for adaptors such as Tapp1 (Dowler et al.,2000; Kimber et al.,2002; Marshall et al.,2002). Tapp1 is constitutively associated though its PDZ domain-binding motif with the protein tyrosine phosphatase-like protein 1(PTPL1/FAP1), which dephosphorylates receptors and adaptor proteins at the plasma membrane, thus further inactivating PI3K signaling(Kimber et al., 2003). Hence,the Pax2-dependent expression of Tapp1 may contribute to feedback inhibition of PI3K signaling during midbrain development.

The Pax2-regulated gene Ncrms is transcribed into a non-coding RNA that was initially identified because of its higher abundance in alveolar rhabdomyosarcoma compared with the embryonic subtype of this pediatric muscle tumor (Chan et al., 2002). Recently, it has been shown that non-coding RNA genes are almost as prevalent as protein-coding genes in the mammalian genome(Cawley et al., 2004). Among these RNAs, the Ncrms transcript belongs with its size of 1.25 kb to the family of long non-coding RNAs, which include the H19, Air and Xist transcripts (Reik and Walter, 2001; Sleutels et al.,2002; Wutz et al.,2002). Analogous to the regulatory functions of these known non-coding RNAs, it is conceivable that the Ncrms transcript is involved in the control of mid-hindbrain-specific gene expression.

Pax2 also controls the expression of the transcriptional regulators En2 and Brn1 in addition to the transcription factors Pax5 and Pax8 in the developing mid-hindbrain region. These data indicate a key role for Pax2 in the activation of distinct transcriptional programs at the onset of mid-hindbrain development. The homeodomain protein En2 is required for normal development of the cerebellum (Joyner et al.,1991; Millen et al.,1994). Interestingly, a 1.0 kb enhancer of the En2 gene contains two Pax2/5/8-binding sites that are essential for directing lacZ transgene expression at the mid-hindbrain boundary(Song et al., 1996). However,mutation of these two sites in the En2 locus only minimally affects the initiation of endogenous En2 transcription(Song and Joyner, 2000). Our observation, that the mid-hindbrain-specific expression of En2completely depends on Pax2 function, points to the presence of yet unidentified functional Pax2/5/8-binding sites that must lie outside of the 1.0 kb enhancer in the En2 locus. In contrast to En2, En1expression is reduced but not absent at the mid-hindbrain boundary of Pax2^–/– embryos(Ye et al., 2001). Hence, Pax2 acts upstream of En genes in the genetic cascade of mid-hindbrain development,consistent with the fact that both En1 and En2 are not required for the initiation, but for the maintenance, of mid-hindbrain-specific gene expression(Liu and Joyner, 2001b) in marked contrast to Pax2 (Pfeffer et al.,2000; Ye et al.,2001) (this study). Moreover, the combined inactivation of En1 and En2 results in a similar mid-hindbrain phenotype(Liu and Joyner, 2001b) as mutation of Pax2 (Favor et al.,1996; Bouchard et al.,2000) in agreement with the regulation of both En genes by Pax2.

The gene coding for the POU domain transcription factor Brn1 was identified as a direct target of Pax2 by two criteria. First, Brn1 expression fails to initiate in the mid-hindbrain region of Pax2^–/– embryos, in contrast to wild-type or Pax5^–/– embryos (see Fig. S3 in the supplementary material). Second, the 5′ region of Brn1 contains two high-affinity Pax2-binding sites, which are essential for initiating transgene expression in the mid-hindbrain region. Additional regulatory sequence located outside of the 6.2 kb 5′ region analyzed are, however,required for the subsequent maintenance of Brn1 expression throughout the entire midbrain and hindbrain, which may involve auto- and cross-regulatory interactions of POU proteins with a consensus octamer sequence in the upstream element B (see Fig. S1 in the supplementary material). Of the four members of the class III Pou gene family, only the closely related Brn1 and Brn2 genes(Bürglin and Ruvkun, 2001)are abundantly expressed in the mid-hindbrain region of E8.5 embryos. Interestingly, Pax2 also regulates the Brn2 gene in this brain region, although less stringently than Brn1 (see Fig. S2 in the supplementary material). Owing to the overlapping expression patterns,midbrain and cerebellum development is largely normal in mice that lack either Brn1 or Brn2 (Schonemann et al., 1995; Nakai et al.,1995; Nakai et al.,2003), whereas cerebellum patterning is severely affected in double-mutant mice (Sugitani et al.,2002). Here, we have complemented these loss-of-function analyses by ectopic expression experiments in chick embryos to further study the role of Brn1 and Brn2 in mid-hindbrain development. Ectopic expression of the transcriptional repressor EnR-Brn1 in combination with Otx2 strongly induced Fgf8 expression in the hindbrain of chick embryos in marked contrast to full-length Brn1 or the transcriptional activator VP16-Brn1. These data and the fact, that endogenous Brn1 is expressed throughout the mid-hindbrain region, suggest that Brn1 functions as transcriptional activator in mid-hindbrain development. In agreement with this conclusion, Brn1 and Brn2 were shown to be important activators of the protease nexin-1 gene at the mid-hindbrain boundary in transgenic analysis as well as in transient transfection experiments (Mihailescu et al., 1999). According to this hypothesis, the EnR-Brn1 fusion protein functions as a dominant-negative regulator to block the transcriptional activity of full-length Brn1(Fig. 7H). The EnR-Brn1 protein may thus lead to ectopic Fgf8 expression by preventing the Brn1/2-dependent expression of a so far unknown repressor of the Fgf8gene (Fig. 7H). This hypothesis implicates Brn1 and Brn2 in restricting Fgf8 expression to the posterior domain of the mid-hindbrain boundary, but fails to explain why Pax2 is able to activate Fgf8 in this location despite the presence of Brn1/2 proteins. Other regulators that are induced at the Otx2/Gbx2 boundary may antagonize the function of Brn1/2 or the postulated repressor to facilitate Fgf8 expression in Gbx2⁺ cells of the mid-hindbrain boundary. Although further experiments including Fgf8promoter analyses are required to elucidate the precise molecular nature of the Pax2-Brn1/2-Fgf8 regulatory interactions, our experiments have identified the Brn1/2 proteins as novel regulators controlling the expression of the IsO signal Fgf8.

We thank R. McEvilly for providing the rat Brn1 cDNA clone, A. Joyner for the En2 probe, K. Paiha for FACS sorting and C. Hartmann for critical reading of the manuscript. This research was supported by Boehringer Ingelheim, the Austrian Industrial Research Promotion Fund and a grant from the Canadian Institutes for Health Research (MOP67219 to M.B.).