Several Cis-regulatory Elements Control mRNA Stability, Translation Efficiency, and Expression Pattern of Prrxl1 (Paired Related Homeobox Protein-like 1)*

Background: The mechanisms that control the Prrxl1 expression are poorly understood. Results: Several regulatory elements present in Prrxl1 alternative promoters are functionally characterized, including a binding motif for Phox2b required for Prrxl1 expression in visceral sensory neurons. Conclusion: We define diverse regulatory modules, which control the spatiotemporal expression of Prrxl1 in nociceptive neurons. Significance: A new mechanism involved in the ganglion specific action of Prrxl1 is described. The homeodomain transcription factor Prrxl1/DRG11 has emerged as a crucial molecule in the establishment of the pain circuitry, in particular spinal cord targeting of dorsal root ganglia (DRG) axons and differentiation of nociceptive glutamatergic spinal cord neurons. Despite Prrxl1 importance in the establishment of the DRG-spinal nociceptive circuit, the molecular mechanisms that regulate its expression along development remain largely unknown. Here, we show that Prrxl1 transcription is regulated by three alternative promoters (named P1, P2, and P3), which control the expression of three distinct Prrxl1 5′-UTR variants, named 5′-UTR-A, 5′-UTR-B, and 5′-UTR-C. These 5′-UTR sequences confer distinct mRNA stability and translation efficiency to the Prrxl1 transcript. The most conserved promoter (P3) contains a TATA-box and displays in vivo enhancer activity in a pattern that overlaps with the zebrafish Prrxl1 homologue, drgx. Regulatory modules present in this sequence were identified and characterized, including a binding site for Phox2b. Concomitantly, we demonstrate that zebrafish Phox2b is required for the expression of drgx in the facial, glossopharyngeal, and vagal cranial ganglia.

Sensory perception of peripheral stimuli is primarily mediated by different types of afferent neurons, which are located in the trunk dorsal root ganglia (DRG) 2 and send processes to the periphery and the spinal cord. In the spinal cord, specialized neurons integrate and relay the information to somatosensory centers in the brain where appropriate responses are generated (1,2). Spinal sensory neurons differentiate from several classes of proliferating progenitor cells, whose establishment requires the expression of proneural genes encoding for basic helixloop-helix (bHLH) transcription factors (3)(4)(5). The various cell lineages are specified according to the combined expression of a set of homedomain transcription factors that confer neural identity to each class (4,5). One of these factors is Prrxl1 (also known as DRG11). Prrxl1 has emerged as a crucial molecule in the development of the pain-perception circuitry, especially in the establishment of the nociceptive DRG-spinal pathway. Prrxl1 is also expressed in sensory cranial ganglia and their target relay neurons in the hindbrain (6,7). Nevertheless, the role of Prrxl1 in these tissues is poorly studied.
Prrxl1 null mutant mice present a distorted spinal dorsal horn with scarce superficial nociceptive-responsive neurons (8 -10), reduced DRG neuronal population (10), and a marked decrease in nociceptive response capacity in various pain tests (8). Interestingly, although involved in the embryonic differentiation of various subpopulations of superficial dorsal horn excitatory neurons (10), Prrxl1 appears not to be required for the normal development of DRG neurons before birth but rather to be essential for their survival in early postnatal life (9). Although Prrxl1 expression in various cell lineages in DRG and spinal cord is well known, the mechanisms of transcriptional control exerted by different bHLH and homeodomain proteins that modulate Prrxl1 transcription are still poorly understood.
Recently, a Prrxl1 alternative spliced variant was identified, and multiple variants of exon 1 in both Prrxl1 mRNA isoform sequences were discovered, suggesting the existence of various 5Ј-untranslated regions (5Ј-UTRs) controlled by distinct promoters (11). Modulation of gene expression through alternative promoter usage is now widely accepted following evidence gathered in the past years (12,13). According to Baek et al. (14), about 40 -50% of human and mouse genes contain alternative promoters, a condition that seems to be required to initiate transcription in a tissue-specific manner (15)(16)(17). The use of multiple promoters, each one controlling at least one transcription start site (TSS), usually originates different 5Ј-UTRs that might have a role in the control of mRNA stability or translation efficiency (18,19).
Here, we characterize three Prrxl1 5Ј-UTRs variants and the corresponding promoter regions, which may explain the differential involvement of Prrxl1 in the DRG and spinal cord development. We also present in vitro and in vivo evidence that the most evolutionarily conserved Prrxl1 promoter region is sufficient to drive expression to neuronal cells and is regulated by Phox2b specifically in primary afferent neurons.

EXPERIMENTAL PROCEDURES
Animal Care-NMRI mice were bred and housed at the Instituto de Biologia Molecular e Celular, Porto, animal facility under temperature-and light-controlled conditions. The embryonic day 0.5 (E0.5) was considered to be the midday of the vaginal plug. The animals were euthanized (isoflurane anesthesia followed by cervical dislocation), and tissues were collected. Experiments were carried out in compliance with the animal ethics guidelines at Instituto de Biologia Molecular e Celular and approved by the Portuguese Veterinary Ethics Committee.
Wild-type AB/Tuebingen (AB/TU) zebrafish strain were maintained in the breeding colony in CABD, Seville, according to standard procedures. Fertilized eggs were kept at 28°C in E3 medium with 0.003% 1-phenyl-2-thiourea to prevent pigmentation and were staged according to Kimmel et al. (20).
Reverse Transcriptase-PCR-The different 5Ј-UTR-Prrxl1 molecules were amplified by reverse transcriptase-PCR (see supplemental Table S1 for primers) from spinal cord total RNA, extracted from mice at different developmental stages (E11.0, E12.5, E14.5, and E16.5) using the Micro-to-midi total RNA purification System (Invitrogen) following the manufacturer's instructions. The first-strand cDNA synthesis was prepared at 42°C during 1 h from 1 g of total RNA using 200 units of transcriptase enzyme (Bioline) and 500 ng of oligo(dT) [12][13][14][15][16][17][18] (Bioline). To assess for potential contaminants, a control containing all reagents except the reverse transcriptase enzyme was included for each sample. Normalization was performed by amplification of mouse ␤-actin using the primers pair listed in supplemental Table S1. The PCR conditions were the following: denaturation at 94°C for 30 s, annealing at 58°C for 45 s, and elongation at 72°C for 45 s. Thirty-two cycles were performed for the amplification of Prrxl1 5Ј-UTR-B and 5Ј-UTR-C, 29 for 5Ј-UTR-A and ORF, and 20 cycles for ␤-actin. The amplification for each gene was in the linear curve (data not shown). Equal amounts of the PCR products were subjected to a 1% agarose gel electrophoresis and visualized by ethidium bromide staining under UV light source. The signals were acquired by a Kodak digital camera DC290, and the densitometric analyses were conducted using the computational program Kodak 1D Image Analysis Software.
In Vitro Transcription-Translation Assay-The different full-length 5Ј-UTR-Prrxl1 cDNA sequences (see supplemental  Table S1 for primers) were PCR-amplified from mouse E14.5 spinal cord cDNA and cloned by TA overhangs in the pCR2.1 (Invitrogen). The resulting vectors were selected for orientation and used to perform the coupled transcription-translation assay in rabbit reticulocyte lysates using the PROTEINscript II T7 kit (Ambion). A sequence corresponding to nucleotides Ϫ1 to Ϫ50 shared by all isoforms was also cloned into the pCR2.1 plasmid and used as a control. The Prrxl1 expression was measured by Western blotting using our homemade rabbit anti-Prrxl1 antibody as described previously (6) and normalized with a mouse anti-tubulin antibody (Sigma).
For the mRNA stability assays, the different 5Ј-UTR-specific sequences were amplified from E14.5 mouse spinal cord cDNA and cloned by TA overhangs in the pCR2.1 plasmid (Invitrogen). These sequences were then subcloned in the HindIII site of the pGL3-Control vector (Promega) between the SV40 promoter and the luciferase coding region.
To determine the Prrxl1 alternative promoters, the entire and overlapping fragments of the Ϫ1401/Ϫ50-bp region upstream of the start codon were PCR-amplified (for primers see supplemental Table S1) from mouse genomic DNA and cloned in the pBlue-TOPO vector (Invitrogen). The sequences were then subcloned in the HindIII site of the promoter-less vector, pGL3-Basic (Promega).
Site-directed mutagenesis of TATA box and HD element were performed from pGL3-REG1 using the QuikChange Sitedirected Mutagenesis kit from Stratagene and following the manufacturer's instructions. Primers used are depicted in supplemental Table S1.
Cell Culture-ND7/23, HeLa, HEK293, and PC12 cell lines were maintained and grown in Dulbecco's modified Eagle's minimal essential medium (DMEM; Invitrogen) containing 10% fetal bovine serum (Invitrogen) and 50 units/ml penicillin and streptomycin (Invitrogen). All cells were kept at 37°C and 5% CO 2 gas phase. Transfection was performed using Lipo-fectamine2000 agent (Invitrogen), and 24 h later cells were harvested for posterior analysis. Overexpression in differentiated PC12 cells was performed by transfecting PC12 cells, and 6 h later differentiation was induced with 100 ng/ml NGF (Sigma). Luciferase reporter assays were performed 2 days after transfection. To infer the mRNA stability, 24 h after transfection, 100 g/ml actinomycin D (A9415, Sigma) was added to DMEM (Invitrogen), and luciferase activity was measured 0, 3, and 6 h later.
Luciferase Reporter Assays-Transfected cells from a 96-well plate format were resuspended in 50 l of lysis buffer (Promega), and the protein extract was cleared by centrifugation. 5 l of the extract were mixed with the luciferase reagent (Promega), and the signals were measured using a luminometer reader (Tecan). Transfection efficiency was normalized by assessing the ␤-galactosidase activity using 2-nitrophenyl ␤-Dgalactopyranoside (Sigma) as substrate.
Chromatin Immunoprecipitation Assays-For TATA-binding protein (TBP) chromatin immunoprecipitation (ChIP) assays, dorsal spinal cords from E14.5 mouse embryos were dissected and fixed with 2 mM di(N-succinimidyl) glutarate (Sigma) in phosphate buffer saline (PBS) for 45 min followed by 1% formaldehyde in PBS for 10 min and lysed in 50 mM Tris-HCl, pH 8.0, 1% SDS, and 10 mM EDTA. Chromatin shearing was performed using Bioruptor (Diagenode) at high power settings for 60 cycles (30 s on/30 s off). ChIP assays with or without (mock control) mouse monoclonal anti-TBP antibody (Mab-002-100, Diagenode) were performed using 80 g of chromatin/assay in ChIP buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 5 mg/ml BSA), and protease inhibitor mixture (Roche Applied Science). Immunoprecipitates were retrieved with 50 l of Protein G Dynabeads (Invitrogen) per assay and washed once with wash buffer I (20 mM Tris, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS), once with wash buffer II (20 mM Tris, pH 8.0, 250 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS), twice with wash buffer III (10 mM Tris pH 8.0, 250 mM LiCl, 1 mM EDTA, 1% Nonidet P-40, 1% sodium deoxycholate), and once with TE buffer (10 mM Tris, pH 8.0, 1 mM EDTA) and eluted with lysis buffer at 65°C for 10 min. Eluted and input chromatin were subjected to proteinase K (Roche Applied Science) treatment for 2 h at 42°C and reverse-cross-linked at 65°C overnight. Immunoprecipitated and input DNA samples were purified by phenol-chloroform extractions followed by isopropyl alcohol precipitation. DNA sequences were quantified by real-time PCR (primers are listed in supplemental Table  S1) using a StepOnePlus Real Time PCR system (Applied Biosystems) and a SYBR Green chemistry for quantitative PCR (Maxima master mix, Fermentas). Quantities of immunoprecipitated DNA were calculated by comparison with a standard curve generated by serial dilutions of input DNA. Data were plotted as the means of at least two independent ChIP assays and three independent amplifications; error bars represent S.E.
Phox2b ChIP assays were performed essentially as described above, with the following modifications: (i) chromatin samples were extracted from dorsal medulla oblongata of E14.5 mouse embryos; (ii) mouse monoclonal anti-Phox2b antibody (sc-376997, Santa Cruz Biotechnology) was used; (iii) NaCl concentration of ChIP buffer was reduced to 20 mM; (iv) immunoprecipitates were washed 6 times with wash buffer III containing only 0.7% sodium deoxycholate.
Production of Transgenic Zebrafish-A sequence (Ϫ751/Ϫ584 bp) that includes the P3 core promoter was PCR amplified from mouse genomic DNA, cloned in the pCR/GW/TOPO vector (Invitrogen), and then recombined by the Gateway in vitro recombination technology using the Gateway LR Clonase II Enzyme mix (Invitrogen) into a Tol2 vector (22) containing an iroquois enhancer with midbrain activity (Z48; Refs. 23 and 24) and the enhanced GFP reporter gene. This vector was assembled by cloning a SalI/NotI fragment of pCS2eGFP (25) containing the CMV promoter, the enhanced GFP reporter gene, and the poly(A) of SV40 into SalI/NotI restriction sites of a modified pminiTol2/MCS vector (26) that has a fragment of the pUC19 polylinker that goes from EcoRI to HindIII. Z48 iroquois enhancer was isolated from the Z48 TOPO vector (23) by cutting with EcoRI and cloned into NotI restriction site after blunting with Klenow. Finally, a gateway cassette (Invitrogen) was cloned in blunt between the SalI/BamHI restriction sites, replacing the CMV promoter. This vector lacks a promoter and does not drive GFP reporter expression. Therefore, it is useful to test for promoter activity of selected DNA sequences. The following primers were used for amplification: 5Ј-TAAGGCC-CAATAGACCTATC-3Ј and 5Ј-CAGGACCAGAGAAGTGA-CTG-3Ј). About 5 nl of the reaction mix containing 50 ng/l transposase mRNA, 50 ng/l phenol/chloroform purified vec-tor, and 0.05% phenol red were injected in the cell of one-stage zebrafish embryos. The GFP expression was then documented from the next 24 to 72 h.
The fragment containing the Ebox and the HD element was PCR-amplified from human genomic DNA, cloned in the pCR/ GW/TOPO plasmid (Invitrogen), and then recombined to the ZED vector (23) by the Gateway in vitro recombination technology described above. The following primers were used for amplification: 5Ј-GTGGTGGTTGTATCGTTCTC-3Ј and 5Ј-GCATAATTGGCCTTAATCTG-3Ј. The injections were performed as described above. Positive transgenic embryos strongly expressing red fluorescent protein were selected 72 h later. Those F0 embryos were raised, and the F1 generation of embryos expressing GFP was analyzed.
RNA Probe Synthesis and Whole-mount in Situ Hybridization-The drgx and tlx3b sequences were amplified from zebrafish cDNA using the following primers: 5Ј-ATGTTTTACTTTCA-CTGTCCTCCA-3Ј and 5Ј-CATTTCTTATCCGGACCCT-C-3Ј for drgx and 5Ј-TTCGGGTGGTGAGGATGGAC-3Ј and 5Ј-GATTTTGGGATGCAACAGCA-3Ј for tlx3b. PCR products were cloned in the pGEM-T Easy vector (Promega), and a phenol-chloroform purification was performed after linearization with NsiI (for drgx) or NcoI (for tlx3b). Each vector was used as a template for the in vitro synthesis of a DIG-labeled RNA probe for zebrafish drgx and tlx3b.
Wild-type embryos at 48 and 72 h post fertilization were fixed overnight with 4% paraformaldehyde at 4°C. After brief washes with PBS, 0.1% Tween 20 (PBST), they were treated with 10 g/ml proteinase K and fixed for 20 min with 4% paraformaldehyde at room temperature. After a 20-min wash with PBST, embryos were incubated in hybridization buffer for 1 h at 70°C. The respective RNA probe was then added to the hybridization buffer (50% formamide, 2ϫ SSC, and 0.1% Tween 20) to a final concentration of 1 ng/l, and embryos were incubated overnight at 70°C. The next day embryos were sequentially washed at 70°C in solutions containing different concentrations (75, 50, 25, and 0%) of hybridization buffer diluted in 2ϫ SSC (75 mM NaCl and 7.5 mM sodium citrate, pH 7.0). After a further wash with 0.05ϫ SSC for at least 1 h at 70°C, embryos were finally washed with PBST for 10 min at room temperature. Embryos were blocked in 2% normal goat serum/PBST for at least 1 h at room temperature. Anti-digoxigenin coupled with alkaline phosphatase was added in fresh 2% normal goat serum, PBST (1:5000) for 2 h and then allowed to wash overnight in PBST. The next day embryos were washed with AP reaction buffer without MgCl 2 (100 mM Tris, pH 9.5, 100 mM NaCl, and 0.1% Tween 20). Detection was performed with 3.5 l of nitro blue tetrazolium (50 mg/ml) and 1 l of 5-bromo-4-chloro-3indolyl phosphate (50 mg/ml) per 1 ml of complete AP-reaction buffer (100 mM Tris, pH 9.5, 50 mM MgCl 2 , 100 mM NaCl, and 0.1% Tween 20). The signal was allowed to develop for 3-5 h.
Antisense Morpholino Oligonucleotide (MO) Analysis-Antisense MO targeted to the translation initiation site of Phox2b (CATTGAAAAGGCTCAGTGGAGAAGG) was obtained from Gene Tools, LLC, diluted to a working concentration in Milli-Q water (0.4 ng/nl) with 0.05% phenol red, and about 5 nl were injected into 1-to 2-cell-stage embryos.
Statistical Analysis-In the present study, all the data presented (except for ChIP experiments; see above) were derived from at least three independent experiments with three replicates. When necessary, a two-tailed t test was performed. Both mean and S.D. values were calculated and included in the figures.

Prrxl1 5Ј-UTR Variants Present Distinct mRNA Stability and
Translation Efficiency-Prrxl1 is a transcription factor first identified by Saito (27) in a subtractive hybridization screening with rat DRG. The Prrxl1 mRNA sequence described by these authors contained the start codon in exon 2, whereas exon 1 corresponded to a 5Ј-UTR. More recently, by the use of 5Ј-rapid amplification of cDNA ends assays with mice spinal cord RNA extracts, we described two novel variants of Prrxl1 containing alternative exon 1 that gives rise to distinct 5Ј-UTRs (11). BLAST searches of GenBank TM database led us to identify sequences that correspond to all the three 5Ј-UTRs that contained both the first and the last coding exons of the annotated Prrxl1. We named these Prrxl1 5Ј-UTR variants 5Ј-UTR-A, 5Ј-UTR-B, and 5Ј-UTR-C, which encompass, respectively, nucleotides Ϫ622 to Ϫ485, Ϫ484 to Ϫ298, and Ϫ148 to Ϫ85 relative to the start codon (Fig. 1A).
To assess for the presence of the different Prrxl1 5Ј-UTR variants, we performed reverse transcriptase-PCR experiments using a common reverse primer mapping within exon 7 (containing the stop codon), shared by all isoforms, and isoformspecific forward primers mapping within alternative exon 1 (arrows in Fig. 1A). With this primer design, we wanted to make sure that the amplicons contained the entire coding region and to exclude from our analysis other putative non-annotated Prrxl1 splicing variants. The three transcripts are detected in the spinal cord at developmental stages where Prrxl1 expression has been previously reported (6). The levels of amplification are suggestive that 5Ј-UTR-A variant is the most abundant transcript, being detectable three PCR cycles earlier than the other two 5Ј-UTR transcripts (Fig. 1B). The semiquantitative reverse transcriptase-PCR analysis with spinal cord extracts from E11.0 to E16.5, a period that includes the early-born (E10.5-E12.0) and late-born (E12.5-E14.5) neurogenesis waves, showed that 5Ј-UTR-A reaches the maximum level at E14.5, a profile similar to the ORF (Fig. 1B). On the other hand, the highest expression levels of 5Ј-UTR-B and 5Ј-UTR-C are reached at E12.5 (Fig. 1B), which is suggestive of a more preponderant role of these two isoforms during early neurogenesis.
Prrxl1 5Ј-UTR variants result from alternative processing of non-coding exon 1 and, therefore, have no consequences in the Prrxl1 coding region. Because mRNA-untranslated regions have been associated to post-transcriptional regulation mechanisms (18,19), we wondered if 5Ј-UTR sequences confer distinct mRNA stability or translation efficiency to Prrxl1 transcripts. To address this question, we performed coupled transcription-translation in vitro assays using different vectors containing the Prrxl1 ORF associated to each 5Ј-UTR under the control of the bacterial T7 RNA polymerase. The amount of Prrxl1 translated was determined by Western blotting (Fig. 1C). Prrxl1 transcript containing 5Ј-UTR-A was about three times  1B, or 1C). Those untranslated regions were named 5Ј-UTR-A, 5Ј-UTR-B, and 5Ј-UTR-C. B, expression studies of Prrxl1 transcripts containing different 5Ј-UTRs (5Ј-UTR-A, 5Ј-UTR-B, and 5Ј-UTR-C) and Prrxl1 ORF by reverse transcriptase-PCR and gel electrophoresis analysis using mouse spinal cord at different developmental ages (E11 to E16.5). The graph illustrates a typical expression profile, normalized with ␤-actin signal intensity, from three independent experiments. C, Western blot analysis of Prrxl1 in vitro translated from mRNA containing distinct 5Ј-UTR and transcribed by T7 RNA Polymerase. The graph represents the mean of signal intensity (normalized with ␤-tubulin) from three independent experiments. D, luciferase reporter assays in ND7/23 and HeLa cells transfected with vectors containing the different 5Ј-UTR fused to the luciferase encoding gene. E, analysis of the stability of the luciferase mRNA molecule conferred by the distinct 5Ј-UTR. The constructs used in D were used to transfect the ND7/23 and HeLa cell lines. Transcription was halted using actinomycin D and luciferase activity measured at different time points (0, 3, and 6 h of treatment). F, comparison of luciferase expression regulated by different versions of 5Ј-UTR-B. 5Ј-UTR-B contains an upstream ATG (uATG), which was replaced by a GTG sequence (5Ј-UTR-B mut). In C-E, 5Ј-UTR control represents a fragment encompassing nucleotides Ϫ1 to Ϫ85 and shared by all isoforms. DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51 more efficiently translated than the UTR control (a fragment immediately upstream of the start codon encompassing nucleotides Ϫ1 to Ϫ85 and shared by all isoforms). On the contrary, 5Ј-UTR-B led to a loss of translational rate, whereas 5Ј-UTR-C was as expressed as the 5Ј-UTR control. Similar results were obtained in luciferase reporter assays using the ND7/23 cell line transfected with each 5Ј-UTR cloned between the SV40 promoter and the firefly luciferase coding region (Fig. 1D). The ND7/23 cells are an appropriate in vitro model for the study of Prrxl1-associated mechanisms as they endogenously express this transcription factor and display a phenotype characteristic of nociceptive neurons (28). Interestingly, the increased luciferase activity of the 5Ј-UTR-A construct observed in ND7/23 cells was not observed in the non-neuronal HeLa cell line. To test if the neuron-specific activity induced by 5Ј-UTR-A could be due to an increase in the transcript stability, mRNA decay associated to each 5Ј-UTR variant was inferred by measuring the luciferase activity at different time points upon treatment with actinomycin D in both cell lines (Fig. 1E). All the different 5Ј-UTRs promoted a similar mRNA decreasing rate in HeLa cells. In ND7/23 cells, 5Ј-UTR-B-and 5Ј-UTR-C-containing transcripts displayed decay similar to the control, whereas the 5Ј-UTR-A was more stable. This result suggested that the 5Ј-UTR-A sole effect, both in the mRNA translation rate and stability, is conferred by a neuronal context likely mediated by specific RNA-binding proteins.

Regulatory Elements Controlling Prrxl1 Expression
On the contrary, the 5Ј-UTR-B variant reduced the in vitro translation efficiency (Fig. 1C) and protein expression both in ND7/23 and in HeLa cells (Fig. 1D) without interfering with mRNA stability (Fig. 1E). A careful analysis of the 5Ј-UTR-B sequence led us to the identification of an ATG located upstream (uATG) of the Prrxl1 main ATG that could be mistaken as an alternative start codon, modifying the reading frame and thereby explaining the feature of this variant (Fig. 1F). Upstream ORFs are widely recognized as cis-regulatory elements that can affect mRNA translation and thus are the molecular base of severe disorders (29,30). By changing this upstream ATG to GTG, the luciferase activity of the mutated 5Ј-UTR-B increased to values similar to the control sequence ( Fig. 1F), suggesting that 5Ј-UTR-B sequence works as a negative modulator of Prrxl1 expression levels. 5Ј-UTR-C did not confer any particular trait to the mRNA molecule.
Identification of Prrxl1 Alternative Promoter Regions and Evolutionarily Conserved Regulatory Elements-The existence of TSSs specific to each Prrxl 5Ј-UTR suggested that Prrxl1 expression is controlled by a mechanism of alternative promoter usage. To identify these promoter regions and further dissect the mechanisms of Prrxl1 regulation, we selected a region of 1351 bp (Ϫ1401/Ϫ50) upstream of the Prrxl1 translation initiation site (ϩ1) based on the high degree of conservation observed in a genomic alignment of homologous region in various species (human, chick, Xenopus tropicalis, and zebrafish) (Fig. 2). This sequence (named REG-1) was amplified and cloned into the promoter-less pGL3-basic vector. Upon transient transfection into mouse DRG primary cell culture and neuronal derived ND7/23 cells, the luciferase reporter gene expression was activated indicating the presence of promoter activity in the cloned region (Fig. 2). We then evaluated the luciferase activity of shorter overlapping sequences (named REG-2 to REG-16) in ND7/23 cells. When the entire fragment was divided in two (REG-2 and REG-11), we verified that both sequences were able to drive the transcription of the luciferase gene. Because we used a promoter-less vector, this result indicated that each fragment harbored at least one promoter. The reduction of the most distal fragment (REG-2) from its 5Ј end (REG-3 to REG-7) led to the identification of a minimum sequence displaying transcriptional activity (REG-7, Ϫ772/ Ϫ584). This sequence is adjacent to 5Ј-UTR-A TSS and, therefore, was considered to be a promoter region. The same analysis was performed for the fragment most proximal to the Prrxl1 start codon (REG-11), and two minimal fragments eliciting luciferase activity (REG-13, Ϫ622/-481 and REG-14, Ϫ157/ Ϫ50), located in the vicinity of the TSS of 5Ј-UTR-B and 5Ј-UTR-C, were considered as promoter regions. By this analysis, we identified three alternative promoters, named P1 (Ϫ85/ Ϫ157), P2 (Ϫ485/Ϫ604), and P3 (Ϫ622/Ϫ772) that likely control, respectively, the transcription of Prrxl1 5Ј-UTR-C, 5Ј-UTR-B, and 5Ј-UTR-A variants.
Some fragments, even though containing promoter regions, did not display any transcriptional activity (REG-6 and REG-9) (Fig. 2). When these fragments were shortened (REG-7 and REG-10), the luciferase activity increased to values closer to fragments of similar size, which is suggestive of the existence, in the region encompassing nucleotides Ϫ811/Ϫ772, of a regulatory element (termed regulatory region A (RRA)) with the capacity to strongly suppress the transcription of the three alternative promoters. Interestingly, by the 5Ј expansion (REG-4 and REG-8) of the fragments containing the RRA, the luciferase activity increased, again revealing the presence of a new regulatory element in the region Ϫ891/Ϫ922 (termed Regulatory Region B, RRB), with the potential to inhibit the action of the repressive motif RRA and consequently to activate Prrxl1 transcription.
It is also important to note the presence of transcriptional regulatory elements located in the sequence between Ϫ1401 and Ϫ958 bp. The abrogation of this region, which originates REG-8 construct (Ϫ958/Ϫ50), resulted in a strong decrease (about 75%) in the luciferase activity when compared with the full sequence (REG-1). Given that the deletion of the region between Ϫ1401 and Ϫ958 bp did not significantly alter the reporter activity of fragments containing P3 as sole promoter (compare the activity of REG-2 with REG-3 and REG-4 in Fig.  2), we concluded that these elements are required for transcription driven by promoters P1 and P2 rather than for the P3 promoter activity.
Because Prrxl1 expression has only been detected in neuronal tissues, namely all sensory ganglia and second order relay sensory neurons (6), we found it pertinent to investigate whether the promoter regions here identified could also drive transcription in non-neuronal cell models. Thus, we performed similar luciferase reporter assays using HeLa and HEK293 cells and compared the results with those obtained for the ND7/23 cells (Fig. 3A). The longer sequence in analysis (REG-1) exhibited the capacity to drive the luciferase transcription 10 (HeLa) to 15 (HEK293) times lower than when transcription was promoted in the neuronal-derived cells ND7/23 (Fig. 3A). An individual analysis of each promoter and its correspondent luciferase activity demonstrated a neuron-specific activity for promoters P1 (REG-14) and P3 (REG-7), as no relevant activity in the HeLa and HEK293 cells was detected. On the other hand, promoter P2 still led to some expression of the reporter enzyme in non neuronal cells. We assume that the potential displayed by promoter P2 to drive transcription in these cells is probably responsible for the transcriptional activity exhibited by the whole fragment (REG-1) due to the presence of constitutive regulatory elements (see below).
Additionally, both regulatory regions RRA and RRB appeared to act exclusively in the neuronal model, as little differences in the transcriptional activity of the fragments REG-8 (Ϫ958/Ϫ50) and REG-9 (Ϫ811/Ϫ50) in comparison to REG-10 (Ϫ751/Ϫ50) were observed in HeLa and HEK293 cells (Fig. 3A).
Interestingly, a careful search for chromatin modifications in the human REG-1 sequence (chr10:50,603,551-50,604,877; GRCh37/hg19 assembly) using the UCSC Genome Browser and ENCODE annotations (31) of ChIP-seq assays revealed trimethylation in the lysine 4 of the histone H3 (H3K4me3) in all the three promoter regions (Fig. 3B). This is a chromatin signature of promoters actively transcribing protein-coding genes (32,33) and was detected in human embryonic stem cells (H7/ H1-hESC) and in different types of neuronal cells, such as a neuroblastoma cell line (SK-N-SH) and neurons derived from embryonic stem cells (H1-neurons) but not in HeLa cells (Fig.  3B). Moreover, binding events for RNA Polymerase II, TBP, and the enhancer-associated protein P300 (34,35) are also detected in the Prrxl1 promoter regions in H1-neurons, H1-hESC, and SK-N-SH cells but not in HeLa cells. Note that the three-peak pattern observed with RNA Polymerase II and P300 overlaps with the three alternative promoter regions (Fig.  3B). These data reinforce our previous observation that Prrxl1 promoters displayed neuron-specific activity.
The Promoter P3 Displays Neuron-specific Activity-Nucleotide alignment of Prrxl1 alternative promoter sequences from different species revealed that the region comprising promoter P3 presents a high degree of conservation, from zebrafish to human, whereas promoters P1 and P2 appeared to be specific to mammals (Fig. 4). We screened these sequences to identify conserved DNA binding elements for transcription factors using the bioinformatics prediction tool MatInspector from Genomatix. Some putative motifs known to be important for transcriptional regulation during embryonic development were identified.
Promoter P1 has no evident conserved motifs, whereas promoter P2 contains a GC box element (Fig. 4), known to be a binding site for Sp1, Sp3, and Sp4 transcription factors (36). CpG islands are elements often associated with the transcription of genes whose expression is ubiquitous and feature Ͼ50% of vertebrate regulatory sequences (37). The presence of this GC-box may explain the activity displayed by the promoter P2 in HeLa cells (Fig. 3A). Moreover, a conserved binding site for the activator protein 2 family was also predicted on the P2 promoter (Fig. 4). Activator protein 2 transcription factors are general regulators of vertebrate development, controlling the balance between proliferation and differentiation during embryogenesis (38,39).
Analysis of the P3 sequence highlighted a putative TATA motif, a conserved element of some basal promoters (Fig. 4). To test if this sequence could bind a TBP, the binding of a recombinant TBP to an oligonucleotide spanning the P3 TATA motif present in the promoter P3 was tested by EMSA (Fig. 5A). A gel shift, which was impaired by the use of competitor oligonucleotides, was observed, demonstrating the specificity of the binding to this region (Fig. 5A). The incubation with an antibody directed to TBP resulted in a supershift (arrow in Fig. 5A), strongly indicating that P3 is a TATA box-containing promoter. The in vivo association of the TBP to the TATA motif of the P3 sequence was tested in mouse embryonic spinal cord by performing ChIP assays combined with real-time PCR (Fig. 5B). In these experiments a clear enrichment in the binding of the TBP was observed in the P3 region comprising the TATA motif compared with the FIGURE 4. Nucleotide alignment of the regulatory regions A and B and the Prrxl1 alternative promoters. The nucleotide alignment was performed using the UCSC Genome Browser and sequences from several mammals, chick, and zebrafish. The boxes that delimit the regulatory regions A and B are depicted. P3 is the most conserved promoter and contains binding sites for the TFIID complex (TATA box), bHLH (Ebox), and homeodomain transcription factors (HD). P2 and P1 are only present in mammals. P2 contains putative motifs for the binding of SP1-like family transcription factors (GC box) and for the activation protein 2 (AP2). The TSSs of each 5Ј-UTR are also represented. TSS1 corresponds to 5Ј-UTR-C, TSS2 to 5Ј-UTR-B, and TSS3 to 5Ј-UTR-A. Binding sites were predicted by the bioinformatics tool MatInspector (Genomatix). Black peaks represent mammal conservation. The gray box highlights the human sequence used to produce the zebrafish transgenic line (see Fig. 6). Not to scale. DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 36293
downstream site (Fig. 5B, compare position Ϫ737 to ϩ 3247) and the control ChIP (mock).
Moreover, to evaluate the importance of the TATA box in the transcriptional activity displayed by the fragment REG-1 in the ND7/23 cells, we compared the luciferase activity of this fragment with a fragment containing a mutated TATA motif (Fig. 5C). This mutation resulted in a 3-fold decrease of luciferase expression, indicating that this element is required for correct transcription. The remaining luciferase expression observed for the fragment REG-1/TATAmut is probably due to the activity of the alternative promoters P1 and P2, which was unaltered.
Taking into account that the P3 sequence is well conserved among species, we used the zebrafish model to test the potential of this TATA promoter to drive transcription in vivo. One-to two-cell-stage embryos were injected with a vector containing the enhanced-GFP (eGFP) reporter gene under the control of the P3 minimum region combined with a zebrafish iroquois enhancer that could direct the expression of eGFP to the midbrain (24). Twenty-four hours post fertilization (hpf), eGFP was already observed in the midbrain of the embryos (Fig. 5D), an expression that lasted at least until 72 hpf. This result indicated that the P3 region exhibits promoter activity in vivo in this transgenic zebrafish assay.
Adjacent to the TATA motif, the P3 region contains evolutionarily conserved motifs potentially bound by bHLH (Ebox motif) and homeodomain (HD motif) transcription factors (Fig.  4). Transcription factors belonging to these families play important roles in the determination of neuronal fates from undifferentiated progenitor cells (for review, see Refs. 5 and 40). To evaluate their functional relevance, a region of 172 bp containing these sites but excluding the TATA motif was cloned in a Tol2 vector, specifically designed to analyze cis-regulatory elements in zebrafish (23) and carrying the eGFP reporter gene FIGURE 5. Validation of P3 as a TATA-containing promoter. A, electrophoretic mobility shift assay using recombinant TBP and a fluorescent probe (TATA DY680) corresponding to a region of P3 with the TATA box motif. A shift, corresponding to a complex formed between TBP and the probe, was detected, and decreases of intensity when different amounts (2.5ϫ and 5ϫ) of non-labeled competitor were mixed. The supershift (arrow) represents the binding of anti-TBP to the protein-DNA complex. B, chromatin immunoprecipitation-quantitative PCR assays with (anti-TBP) or without (mock) an anti-TBP antibody were performed using dorsal spinal cord chromatins from E14.5 mouse embryos followed by quantitative PCR using primers targeting P3 region (Prrxl1 Ϫ737) and a downstream region (Prrxl1 ϩ 3247). Enrichment is observed in the region where the TATA box is present (Ϫ737). A region corresponding to Prrxl1 intron was used as control (ϩ3247). Mock represents the condition without antibody. C, the evaluation of the functional importance of the TATA box for the transcriptional potential of the fragment REG-1 was assessed by luciferase reporter assays in ND7/23 cells, comparing the REG-1 fragment with a sequence containing a mutated TATA motif (REG-1/TATA mut). D, to evaluate the ability of the P3 core promoter to activate transcription in zebrafish, a region encompassing nucleotides Ϫ584/Ϫ751 bp was cloned upstream to an Irx (Iroquois) enhancer, which drives expression to the zebrafish midbrain (white arrowheads). The vector was injected in zebrafish embryos at one-cell stage, and 48 hpf GFP signal was recorded.
under the control of the gata2a minimal promoter (Fig. 6). This vector was used to generate a zebrafish stable transgenic line. eGFP signal was recorded in F1 embryos at different developmental ages and compared with endogenous expression of drgx, the Prrxl1 zebrafish homologue. The expression of drgx was previously reported (41) as restricted to dorsal spinal cord, DRG, and, in the developing brain, to sensory neuron populations of the midbrain and hindbrain, cranial sensory ganglia, and the habenula. From 48 hpf, a strong eGFP signal was detected in the developing hindbrain and cranial ganglia (Fig. 6, B and F) in a pattern that only partially overlaps the expression of endogenous drgx, as revealed by in situ hybridization (Fig. 6,  A and E). Indeed, eGFP expression was prematurely observed in a posterior region of the hindbrain (marked by an arrowhead in Fig. 6, B and F) that only expresses drgx later at 72 hpf (compare the arrowheads in Fig. 6, C and G, with the arrowheads in Fig. 6, B and F). Due to the very small length of this fragment (172 bp), it is conceivable that repressive elements may be missing. Expression of eGFP in cranial ganglia displayed a spatiotemporal pattern that perfectly matched with drgx staining ( figure  6E-H). At 72hpf, a decrease in the eGFP staining was observed (compare Fig. 6, D and H, with Fig. 6, panels C and G). Albeit the expression of drgx in DRG and spinal cord is well reported in zebrafish (41), the 172-bp sequence does not drive eGFP transcription to these tissues (data not shown). Together, these experiments pointed out that the region of 172 bp containing the conserved Ebox and HD elements was sufficient to drive expression of the reporter gene in drgx-expressing neurons.
Phox2b Controls Prrxl1 Expression by Binding the HD Motif in P3 Promoter-To unravel the trans-acting factors that may be responsible for the modulation of Prrxl1 transcription via Ebox or HD element, we tested the overexpression effect of a set of transcription factors previously implicated in the control of Prrxl1 expression, as described in diverse epistatic studies employing mutant mice. In Tlx3 and Lmx1b null mutant mice, Prrxl1 expression is affected in the spinal cord but not in the DRG (42,43), whereas studies performed with DRG of islet1 inducible conditional knock-out mice suggested that Prrxl1 expression is regulated by islet1 only at an early stage of neurogenesis (44). Moreover, Prrxl1 is activated by Brn3a in the DRG and trigeminal ganglion (45) and repressed by Phox2b in the facial, glossopharyngeal, and vagal cranial ganglia (7). Mash1 and Neurogenin1 (Ngn1) are proneuronal genes implicated in the specification of progenitors cells from which Prrxl1-expressing neurons arise either in the DRG or the neural tube (3, 46 -48). Ngn1 and Mash1 belong to the bHLH transcription factor family and bind the Ebox consensus motif CANNTG (49), whereas the remaining proteins belong to homeodomain family and recognize the HD bipartite element TAATNNNATTA (50,51). Thus, we assessed the expression of luciferase under the control of the REG-1 fragment in ND7/23 cells overexpressing Brn3a, Tlx3, Phox2b, Islet1, Ngn1, Lmx1b, and Mash1 (Fig. 7A). Three transcription factors strongly induced the transcriptional activity of the fragment REG-1. Those were Brn3a, Tlx3, and Phox2b. As homeodomain proteins, they are candidates to bind the HD element present in the P3 promoter. Therefore, we evaluated by luciferase reporter assays the effect of the overexpression of these transcription factors on the REG-1 construct containing the mutated HD motif compared with wild-type REG-1 (Fig. 7B). Tlx3 and Brn3a overexpression displayed the same luciferase activity in both constructs, whereas the induction of Phox2b decreased about 50% when the HD motif is mutated. To ascertain if Phox2b binds directly to the HD element present in Prrxl1 P3 promoter, EMSA and ChIP-PCR were performed (Fig. 7, C and D). By EMSA, a DNA-protein shift was detected after the incubation of an oligonucleotide comprising the HD motif sequence and nuclear protein extracts from ND7/23 cells overexpressing Phox2b. On the . At 72 hpf, drgx expression is more widespread in the hindbrain, and it still maintains its expression in the cranial ganglions and in the tegmentum (see dorsal view (C) and lateral view (G)). B, D, F, and H, analysis of eGFP expression in transgenic zebrafish embryos. eGFP expression is regulated by a module that encloses the gata2 minimal promoter and a region of the Prrxl1 P3 promoter that contains the Ebox and HD elements, excluding the TATA motif (see the scheme). In 48 hpf embryos, eGFP expression is distinguished in the hindbrain and in cranial ganglia (see dorsal view (B), lateral view (F)). The arrowhead indicates a region where eGFP is apparently activated prematurely in the transgenic embryos at 48 hpf (B and F) but whose drgx expression is only detected at 72 hpf (C and G). Although eGFP expression is reduced at 72 hpf, it is still detected in a small region of the anterior hindbrain and in the cranial ganglia (see dorsal view (D) and lateral view (H)). DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51 contrary, overexpression of Tlx3 or Brn3a only led to a weak basal shift in the gel, which is also observed in extracts containing Phox2b overexpressed (marked by an arrowhead in Fig. 7C). We assume that this interaction represents a binding with another homeodomain protein that remains to be identified. The binding of Phox2b on Prrxl1 promoter was further validated by ChIP-PCR assays (Fig. 7D). A chromatin enrichment (relative to the no antibody control, mock) was only observed with primers that amplified the region comprising the HD element (Fig. 7D, position Ϫ737).

Regulatory Elements Controlling Prrxl1 Expression
Co-expression of Phox2b and Prrxl1 or its zebrafish orthologue, drgx, was only reported in some sensory ganglia, namely in the facial, glossopharyngeal, and vagal cranial ganglia, and their target relay neurons in the hindbrain (6, 7). As Phox2b is not detected in the ND7/23 cell line, we performed the same analysis using non-differentiated and differentiated PC12 cells, which endogenously expressed Phox2b and Prrxl1 (52). Again, overexpression of Phox2b increased the luciferase activity of REG-1 fragment, which is impaired by mutating the HD motif (Fig. 7E).
To better investigate the regulation of Prrxl1 expression by Phox2b in vivo, we used a validated specific antisense MO to interfere with the translation of zebrafish Phox2b protein.
According to what was previously reported by Elworthy et al. (53), our zebrafish Phox2b morphants also presented scarce Hu-positive cells in the hindgut at 5 dpf (53%, 8 in 15 embryos). We injected Phox2b MO in the transgenic zebrafish line described above (Fig. 6), which contains eGFP under the control of the 172-bp sequence comprising the Phox2b binding element (HD motif). A decrease in the eGFP expression was observed (Fig. 7F), supporting our hypothesis that Phox2b binds to the HD element present in the P3 promoter region of Prrxl1 and is required for this promoter activity.
Phox2b Is Required for drgx Expression in the Glossopharyngeal, Vagal, and Facial Ganglia-To further address the role of Phox2b in the control of drgx expression, we injected Phox2b MO in wild-type zebrafish embryos at one-to two-cell stage. Expression analysis of drgx by in situ hybridization showed that in 48 hpf morphant animals, drgx is still maintained in the hindbrain, in the tegmentum, and in the trigeminal ganglia (Fig. 8,   FIGURE 7. Phox2b modulates Prrxl1 transcription by binding to the HD element in the P3 promoter. A, the effect on REG-1 transcriptional activity by overexpressing the bHLH transcription factors Mash1 and Ngn1 and the homeodomain proteins Lmx1b, Islet1, Phox2b, Tlx3, and Brn3a in ND7/23 cells was assessed by luciferase reporter assays. The immunoblot confirms the protein expression and corresponds to one representative experiment. B, luciferase activity induced by REG-1 and REG-1/HD mut was measured in ND7/23 cells overexpressing Tlx3, Brn3a, and Phox2b or transfected with an empty pcDNA3.3. C, electrophoretic mobility shift assay using nuclear extracts from ND7/23 cells overexpressing Brn3a, Tlx3, or Phox2b or transfected with an empty pcDNA3.3 and a fluorescent probe (HD DY680) that comprises the HD motif present in the P3 promoter. The shift (arrow) is only observed in ND7/23 cells that overexpressed Phox2b. The arrowhead indicates a basal shift detected in all samples. D, chromatin immunoprecipitation assays with (anti-Phox2b) or without (mock) an anti-Phox2b antibody were performed using dorsal medulla oblongata chromatins from E14.5 mouse embryos followed by quantitative PCR using primers targeting regions, which comprises the HD motif (Prrxl1 Ϫ737) and an upstream control sequence (Prrxl1 Ϫ3568). An enrichment was only observed in the region where the HD element was present (Ϫ737). E, luciferase activity induced by REG-1 and REG-1/HD mut was measured in ND7/23, PC12, and differentiated PC12 cells overexpressing Phox2b. F, analysis of eGFP expression in transgenic zebrafish embryos at 72 hpf, containing the module that comprises the Phox2b-binding site (HD motif) present in the P3 promoter. Expression of eGFP was reduced in Phox2b MO-injected embryos when compared with controls. The fluorescence acquisition settings were exactly the same in both images.
compare drgx expression in control with Phox2b MO). Nevertheless, in 44% of embryos, drgx expression is much weaker or absent in the glossopharyngeal, vagal, and facial ganglia (Fig. 8), suggesting that, as predicted by in vitro assays, Phox2b is required for the transcriptional control of drgx.
To discard the hypothesis that the absence of drgx expression was due to a failure in the development of these ganglia induced by Phox2b knockdown, in situ hybridizations were performed using a tlx3b probe. Tlx3 expression has been shown to extensively overlap with Prrxl1 (10,42) and thus was used here as a reliable marker of neuronal differentiation. As expected, tlx3b expression is maintained in Phox2b morphants (Fig. 8, Phox2b MO tlx3b).

DISCUSSION
Prrxl1 is a transcription factor with an important role in the establishment and maintenance of the nociceptive DRG-spinal cord neuronal circuit. The function of Prrxl1 in this process has been characterized in detail, but the molecular determinants causing its activation or repression are not yet understood. To shed some light on the molecular mechanisms regulating Prrxl1 gene expression, we isolated and characterized the alternative promoters that control the expression of three distinct Prrxl1 5Ј-UTR variants, named 5Ј-UTR-A, 5Ј-UTR-B, and 5Ј-UTR-C. These variants are originated from different exon 1 during Prrxl1 splicing and do not have any consequence in the protein reading frame as the AUG start codon is present in exon 2.
The alternative use of different exon 1 has been recognized as another mechanism of control of gene expression. Within the human and mouse genome, Ͼ3000 genes with multiple first exons have been identified (54). 5Ј-UTR-mediated regulation were shown to modulate gene expression through mechanisms that influence post transcriptional modification of RNA (secondary structure and mRNA stability) and translational efficiency (18). This proved here to be also the case for Prrxl1 5Ј-UTR variants. The 5Ј-UTR-A displayed a neuron-specific effect, increasing both the rate of Prrxl1 protein translation and the stability of the mRNA molecule. This observation implies a contribution of neuronal specific RNA-binding proteins. Taking into account that Prrxl1 has an important role in the development of the neuronal circuit connecting the DRG to the spinal dorsal horn (8,9), possible candidates are Hu RNA-binding proteins (human homologues of Drosophila ELAV), namely the neuronal specific HuC and HuD isoforms, which are long used as markers of neuronal differentiation. Overexpression of HuC or HuD in PC12 cells increased the rate of neuronal differentiation, whereas down-regulation resulted in an impairment of neurite growth (55,56). Likewise, a decrease in neuronal differentiation was observed in HuD null mutant mice (57). Although HuC/D proteins are strong candidates to modulate the neuronspecific activity of Prrxl1 5Ј-UTR-A, their involvement remains to be demonstrated.
On the contrary, the 5Ј-UTR-B reduced mRNA half-life and suppressed mRNA translation, resulting in a decrease in the luciferase expression both in the neuronal-derived ND7/23 and HeLa cells. This was due to the presence of an AUG in the 5Ј-UTR-B, as shown by luciferase reporter gene assays and sitedirected mutagenesis. This AUG is used as an alternative start codon, modifying, therefore, the protein reading frame. AUG codons upstream of the main open reading frame are present in ϳ10% of all mRNAs (58). Although the functional impact of this mechanism has not been investigated in detail, a recent study on male-specific lethal-2 mRNA suggested an important role in negative translational control by increasing initiation of scanning ribosomes at the upstream open reading frame and blocking downstream translation (58). The cis-regulatory upstream ORF present in the Prrxl1 5Ј UTR-B exerts a negative influence on translational efficacy, as seen in Fig. 1C, and is probably responsible for controlling the amount of Prrxl1 protein during early neurogenesis, the stage where 5Ј-UTR-B is most expressed.
By studying deletion derivatives of the Prrxl1 5Ј-flanking region, we demonstrated that transcription of each Prrxl1 5Ј-UTR variants is controlled by specific promoters. The herenamed promoter P1, P2, and P3 controlled, respectively, the expression of 5Ј-UTR-C, 5Ј-UTR-B, and 5Ј-UTR-A. Recent genome wide analyses indicated that alternative promoter usage is a common event that occurs at least with the same order of frequency as alternative splicing, affecting about 52% of human genes (37). On average, there are 3.1 alternative promoters per gene, with the composition of one CpG-island-containing promoter per 2.6 CpG-less promoters. Moreover, it was demonstrated that genes that undergo complex transcriptional regulation often include at least one CpG island-containing promoter, expressed ubiquitously, and are accompanied by other promoters used for tissue-specific or signal-dependent expression (37). Our results on the Prrxl1 promoters are in accordance with these observations. Indeed, a GC-rich region is contained within promoter P2, whereas P3 is a TATA promoter. P2 displayed activity in all cell lines tested, whereas P3 and P1 were regulated depending on the cellular context being only active in neurons (for a summary of these findings see the model in Fig. 9).
Among the three Prrxl1 alternative promoters, P3 is the most conserved and displays in vivo activity in the zebrafish. Reverse transcriptase-PCR experiments also showed that the P3-derived transcript, Prrxl1 5Ј-UTR-A, is more abundant than the other two variants. These two observations led us to conclude that promoter P3 may play a more prominent role in Prrxl1 transcriptional regulation and prompted us to focus our study on transcriptional mechanisms modulating P3 activity. Core promoters were shown to comprise DNA sequence motifs, such as the TATA box, the Initiator (Inr), and the downstream promoter element, located within Ϫ30 to ϩ30 nucleotides relative to the TSS and to mediate the recognition and recruitment of the RNA polymerase II to the transcriptional apparatus (59,60). Prrxl1 P3 core promoter contained a TATA motif located at 35 bp upstream of 5Ј-UTR-A TSS. The functional validation of this motif was attained by in vitro and in vivo evidence, such as (i) TBP interacted with the TATA motif as detected by EMSA and ChIP using mouse spinal cord samples, (ii) mutation of the TATA motif greatly reduced promoter activity, and (iii) P3 core promoter, which contains the TATA element, is sufficient to drive the reporter gene expression in the zebrafish.
Located upstream of the P3 promoter, a highly conserved region (Ϫ1401/Ϫ958) was identified as an important modulator of Prrxl1 transcription. This region contains binding sites for actively transcribing protein such as RNA Polymerase II, TBP, and P300. By sequence deletion experiments, the presence of still uncharacterized regulatory modules was predicted. These elements appeared to be mainly associated with P1 and P2 promoters and thereby to the transcription of 5Ј-UTR-B and 5Ј-UTR-C. Because these Prrxl1 mRNA variants are enriched in early-born neurons of the developing spinal cord, we hypothesize that the modules included in the Ϫ1401/Ϫ958 bp region may be responsible for enhancing temporal-specific transcription controlled by promoters P1 and/or P2 promoters.
We also defined two neuron-specific elements, RRA (Ϫ811/ Ϫ772) and RRB (Ϫ891/Ϫ922), which exhibit the opposite effect on Prrxl1 transcription. RRA has the potential to strongly suppress the transcription of the three alternative promoters, whereas RRB counteracts the action of the RRA repressive motif and consequently induces Prrxl1 transcription (see the model in Fig. 9). Such tight regulation could be understood as a way to modulate Prrxl1 expression in different neuronal types, namely in glutamatergic (Prrxl1-positive) over GABAergic (Prrxl1-negative) neurons during the developing spinal cord. Because these two neuronal populations derived from the same progenitor domain and are mutually exclusive, molecular interrepression between inhibitory and excitatory interneurons has been suggested. For instance, Tlx3 acts as an inhibitor of Lbx1 expression inducing the glutamatergic transmitter phenotype (61), whereas Ptf1a represses Tlx3 specifying a GABAergic cell FIGURE 9. Schematic representation of the regulatory elements that control the expression of Prrxl1 5-UTR variants. Promoter regions and other regulatory elements were identified upstream of Prrxl1 translation start site (ϩ1). Transcription of the three Prrxl1 mRNA variants 5Ј-UTR-C, 5Ј-UTR-B, and 5Ј-UTR-A is controlled by distinct promoters, termed P1, P2, and P3, respectively. 5Ј-UTR-A is the most stable Prrxl1 transcript and is enriched at E14.5 spinal cord, a developmental age associated with late neurogenesis. Located between Ϫ958 and Ϫ772 bp, two preponderant elements were identified: RRA is able to totally suppress the transcriptional activity regulated by the three promoters, and RRB inhibits the repressive effect of RRA resulting in an increase of transcription. In addition, the Ebox and HD motifs located upstream of the TATA box drive the transcription of Prrxl1 to hindbrain and cranial ganglia. Although the bHLH transcription factor that binds the Ebox motif remains to be identified, the homeodomain protein Phox2b binds the HD element and is required for Prrxl1 expression in visceral cranial ganglia. uAUG, AUG codons upstream.
fate (62). A similar mechanism could be envisaged for Prrxl1. In GABAergic neurons, Prrxl1 expression could be prevented by a specific transcription factor acting on the RRA element. This repressive effect could be suppressed in glutamatergic neurons through the RRB element. It would be interesting to assess if one possible RRB binding candidate could be Tlx3 as this transcription factor induces Prrxl1 promoter activity (Fig. 7A) and highly co-localizes with Prrxl1 (10).
Furthermore, the 172-bp 5Ј region adjacent to the P3 core promoter was sufficient to drive specific neuronal activity in zebrafish. GFP expression under the control of putative regulatory elements present in this 172-bp region was only observed in the area in which the transcript of the zebrafish Prrxl1 ortologue, drgx, is detected. Such specificity suggested that important cis-regulatory modules are present in this region, and further analysis of this sequence revealed the presence of a highly conserved putative binding site for homeodomain transcription factors, proteins that are expected to control the expression of Prrxl1 as a function of developmental age and neuronal context. Among the tested candidates, Tlx3, Brn3a, and Phox2b induced Prrxl1 promoter activity, but only Phox2b was able to bind to this HD motif. Co-expression of Phox2b and Prrxl1 or its zebrafish ortologue, drgx, was only detected in the visceral sensory pathway, namely in the facial, glossopharyngeal, and vagal cranial ganglia and their target relay neurons in the hindbrain (6,7,63).We, therefore, presume that the P3 activity observed in the transgenic line (Fig. 6) may be controlled, at least to some extent, by Phox2b. This positive regulation is further supported by silencing experiments with Phox2b morphants (Fig. 7F).
Nonetheless, in the mouse, visceral sensory neurons switch to a somatic fate in the absence of Phox2b acquiring a molecular profile similar to that of somatic sensory neurons at later developmental stages, with higher expression of Prrxl1 and Brn3a (7). This increase in Prrxl1 expression was suggested to be mediated by the increase of Brn3a, which, in turn, is directly repressed by Phox2b (7). On the contrary, Prrxl1 expression in the nuclear TS progenitor domains of the hindbrain is not affected by Phox2b inactivation, suggesting that the repression of Prrxl1 by Phox2b only occurs in the visceral sensory ganglia (7). In accordance, Phox2b is not required for drgx expression in the zebrafish hindbrain. However, contrary to the mouse, Phox2b down-regulation induced a loss of drgx expression in the facial, glossopharyngeal, and vagal cranial ganglia. Binding of Phox2b on the HD motif increased Prrxl1 promoter activity in ND7/23 and PC12 cells. Our data strongly suggest that Phox2b has the potential to work as a direct Prrxl1 activator (see model in Fig. 9).
Although Phox2b is a determinant of visceral fate, its mode of action varies with neuronal types (64,65) likely due to cell typespecific combinations of transcription factor complexes. On the other hand, Prrxl1 is transiently expressed at early stages of the development of the facial-glossopharyngeal ganglion and the distal part of the vagal ganglion, as observed in mice (7) and zebrafish (41). During this particular developmental window, co-expression of Prrxl1, Phox2b, Tlx3, and Islet1 is observed (7). Recently, studies performed in Islet1 inducible conditional knock-out mice (44) suggested that Prrxl1 expression in the DRG is regulated by Islet1 only at an early stage of neurogenesis. Thus, it is conceivable that Phox2b, in combination with Islet1, binds to the HD motif on the P3 Prrxl1 alternative promoter to activate Prrxl1 expression at early stages of sensory ganglia development, whereas later (from E13.5 on), as part of a distinct transcriptional machinery, Phox2b shuts down Prrxl1 expression likely through the repression of Brn3a. Understanding how Brn3a acts on Prrxl1 promoters will add new insight on this mechanism. The present data thus support that the P3 alternative promoter is involved in the ganglion specific action of Prrxl1 (see the model in Fig. 9), which appear to be controlled by Phox2b in the case of visceral sensory neurons.