The mammalian genome contains nine voltage-gated sodium channel α-subunit genes (SCNA genes), along with an evolutionarily-related gene encoding a non-voltage-gated sodium channel (Yu and Catterall,2003). During axonal myelination, SCN8A replaces SCN2A as the major sodium channel isoform at the nodes of Ranvier. SCN8A sodium channels are responsible for postnatal increases in sodium current in motoneurons (Burgess et al.,1995; Garcia et al.,1998). The reduced expression of SCN8A sodium channels is associated with a broad spectrum of locomotor impairments. Missense and nonsense mutations of SCN8A gene result in motor end-plate disease, characterized by ataxia, tremor, muscle weakness, and dystonia. The severity of neurological phenotypes in mice with different SCN8A allelic mutations is related to the amount of remaining normal SCN8A mRNA (Raman et al.,1997; Meisler et al.,2001; Meisler et al.,2004). High-frequency firing as well as resurgent and persistent currents is reduced in Purkinje cells of SCN8A mutant mice and mice with Purkinje cell–specific SCN8A knockout. Thus, SCN8A sodium channels play an essential role in the motor circuit (Raman et al.,1997; Levin et al.,2006).
The zebrafish scna gene family consists of four sets of duplicated sodium channel genes, designed as scn1aa/ab, scn4aa/ab, scn5aa/ab, and scn8aa/ab (Novak et al.,2006). Scn8aa/ab shares sequence homology with mammalian SCN8A gene family. Zebrafish scn8aa gene is expressed in both Rohon-Beard neurons (RB) and trigeminal ganglia (trg) at around 16–17 hr post fertilization (hpf) when these neurons start to extend axons (Kuwada et al.,1990; Tsai et al.,2001). Zebrafish scn8aa sodium channels play critical roles in the development of normal motilities during early embryonic stages. An increase in the number of scn8aa sodium channels is responsible for the developmental upregulation of RB sodium current (Pineda et al.,2005). Knockdown of scn8aa gene increases the survival of RB and delays the dorsal projection of secondary motoneurons (Pineda et al.,2006). Furthermore, knockdown of scn8aa gene attenuates spontaneous contraction, tactile sensitivity, and swimming of zebrafish embryos (Chen et al.,2008). It is suggested that scn8aa sodium channels are essential for early embryonic locomotor activities.
Transcriptional regulation of sodium channel genes during development is highly dynamic (Spitzer and Ribera,1998; Spitzer et al.,2002). The expression of SCN1A, SCN2A, and SCN8A sodium channels increases, while the expression of SCN3A sodium channels decreases. The promoters of mammalian neuronal-specific SCN2A, SCN3A, and SCN8A genes as well as muscle-specific SCN5A gene have been identified (Kraner et al.,1992,1999; Drews et al.,2005,2007). Repressor element 1 (RE1) prevents extra-neuronal expression of rat SCN2A gene (Chong et al.,1995; Lunyak et al.,2004; Ma,2006). The conserved elements located in exon 1 of mouse SCN8A, human SCN8A, chicken SCN8A, and zebrafish scn8aa are sufficient for brain-specific expression in transgenic mice (Drews et al.,2005). Whether this regulatory fragment is sufficient to drive the expression of mouse SCN8A in the spinal cord and during embryonic development remains unknown. Here, we investigate the transcriptional regulatory mechanism of zebrafish neuronal-specific scn8aa gene. We provide evidence that four enhancing elements, including three newly identified elements and one conserved element shared with mouse SCN8A gene, cooperatively regulate the transcription activity of zebrafish scn8aa in the brain and spinal cord.
A 15-kb Putative Regulatory Fragment of Zebrafish Scn8aa Restricted GFP Expression in a Neuronal-Specific Manner
To study the transcriptional regulation of zebrafish scn8aa, a 15-kb putative regulatory fragment was isolated from zebrafish genomic library. The transcriptional start site, as determined by 5′ RACE, was designed as +1 bp. The 15-kb genomic fragment encompassed 7.4 kb 5′ upstream region, 283 bp exon 1, and 7.2 kb intron 1. One putative TATA box (TATAAATA) was located at +74 bp. There was a 1.5-kb gap between the 3′ end of the 15-kb genomic fragment and the translation start codon of scn8aa (Fig. 1A).
To determine whether the 15-kb fragment is sufficient to modulate the expression of scn8aa in a broad spectrum of neuronal types, we cloned the fragment into a modified pEGFP-ITR reporter vector. The resulting plasmid, designated scn8aa:GFP, was injected into zebrafish embryos at the one-cell stage for transient transgene expression analysis. GFP was expressed in the anterior lateral line ganglia, posterior lateral line ganglia, trg, and the ventral region of the hindbrain at 24 hpf (Fig. 1B, a). GFP protein was detected in RB and interneurons in the spinal cord (Fig. 1B, c). GFP expression pattern of scn8aa:GFP recapitulated the expression pattern of endogenous scn8aa mRNA at 24 hpf (Fig. 1B, b and d). At 48 hpf, GFP-positive cells were observed in the hindbrain and spinal cord (Fig. 1B, e and g). It was consistent with the profile of scn8aa mRNA expression at this stage (Fig. 1B, f and h). GFP- and scn8aa mRNA- positive cells became more abundant in the brain, spinal cord, and retina at 72 hpf (Fig. 1B, i–l). The olfactory sensory neurons, retinal neurons, optic tectum, and ventral and dorsal projecting motoneurons were GFP-positive at 72 hpf (Fig. 1B, m–p). The accumulation of GFP protein in the cytoplasm of spinal neurons allowed us to identify subtypes of interneurons (Fig. 1B, q–w). Seven of the nine classes of spinal interneurons, including three classes of ascending interneurons (DoLA, CoPA, and CoSA), three classes of descending interneurons (VCoD, VeMe, and CiD), and bifurcating interneurons (CoBL) (Hale et al.,2001; Higashijima et al.,2004), were GFP-positive. The 15-kb fragment was sufficient to drive GFP expression in all of scn8aa mRNA-expressing neurons, with the single exception of dorsal root ganglia (DRG) in the transient expression assay shown in Table 1.
Table 1. The Expression Profile of Endogenous Scn8aa and Transient Transgene Scn8aa:GFP During Zebrafish Developmenta
Central nervous system
Peripheral nervous system
Expression is classified as present (+), absent (−), or not determined (ND).
The absence of GFP in DRG may be due to lack of cis elements responsible for the GFP expression in DRG in the 15-Kb fragment. Alternatively, little scn8aa:GFP DNA remains in the transient transgene expression analysis when neuronal progenitor cells begin to differentiate into DRG after 36 hpf. To answer this question, we established stable transgenic lines. One female and one male transgenic scn8aa:GFP founder fish (F0) were isolated. GFP transgene was expressed by approximately 1 and 13.3% of F1 progeny from the female F0 and male F0, respectively. The highly mosaic germ-line transmission from transgenic founder fish has been described (Stuart et al.,1988). GFP expression patterns in embryos derived from the two founder fish were identical. Therefore, seven lines derived from the female founder, referred to as Tg(scn8aa:GFP) lines, were analyzed further. F1 fish was crossed with wild type fish to generate heterozygotic F2 offspring. Among three independent matings, 49.6% (114/230, N = 3) of F2 offspring carried scn8aa:GFP transgene. A single chromosomal insertion in F1 zebrafish was predicted based on Mendelian genetics. However, the possibility of multiple insertions close to each other in one chromosome cannot be excluded.
Next, we investigated whether GFP expression temporally and spatially recapitulates scn8aa mRNA expression during development in Tg(scn8aa:GFP) line. GFP was first detected in RB and trg at 16 hpf in the transgenic offsprings (Fig. 2A, B), when scn8aa mRNA was first detected. By 24 hpf, GFP was expressed in the lateral line ganglia, hindbrain, and interneurons (Fig. 2C, D). GFP mRNA was first detected at 16 hpf and increased at 24 hpf, as determined by in situ hybridization (see Supplemental Figure 1, which can be viewed online). Increased GFP expression was observed in the retina, optic tectum, and ventral hindbrain of heterozygotes at 48 hpf (Fig. 2E). GFP protein was accumulated in the dorsal longitudinal fascicle containing axons of RB and spinal interneurons (Fig. 2F). However, only a limited number of GFP-positive neurons were present in the cerebellum and hindbrain (Suppl. Fig. 2). This phenomenon could be due to the dosage effect, as GFP expression in the hindbrain was markedly greater in homozygotes than heterozygotes at 48 and 72 hpf (Fig. 2G–L).
At 72 hpf, GFP expression was first detected in DRG located in the anterior trunk (Fig. 2M, N). The axonal projections of primary and secondary motoneurons also contained GFP at this stage (Fig. 2O). These axons could be projected by scn8aa-expressing CaP primary motoneurons and dorsal projecting secondary motoneurons (Pineda et al.,2006). Confocal microscopy revealed that GFP was present in trg, facial (VII), glossopharyngeal (IX), and vagus (X) sensory ganglia (Fig. 2P). The optic tectum, retina, and hindbrain expressed GFP abundantly at 72 hpf (Fig. 2Q). GFP protein was detected in the optic tectum and retina, but not diencephalon, which was similar to the expression pattern of endogenous scn8aa mRNA (Fig. 3A, B). In the homozygous Tg(scn8aa:GFP) line, GFP protein was detected in the retinal ganglia cell layer, inner plexiform layer, and part of the inner nuclear layer (Fig. 3C, D). The dendrites of interneurons located at the proximal side of the inner nuclear layer projected into the inner plexiform layer. Some GFP-positive cells in the inner nuclear layer were ON-type amacrine cells based on their morphological features (Connaughton et al.,2004). However, GFP expression by other types of interneurons cannot be ruled out, because some GFP-positive cells in the inner nuclear layer possessed no axonal projection at this stage. In contrast to GFP, scn8aa mRNA was broadly expressed in the inner portion of the inner nuclear layer (Fig. 3E). In the trunk, GFP-positive cells were present at the lateral region of the spinal cord (Fig. 3F, G). The location of GFP-positive cells matched the position of scn8aa-expressing neurons, which migrate from the ventricular zone to the lateral region of the spinal cord during differentiation (Fig. 3H, I). The ventricular zone of the hindbrain and spinal cord contained actively proliferating cells, labeled by BrdU incorporation (Fig. 3J, K). Early differentiated neurons labeled by anti-Hu antibody in the hindbrain and spinal cord were partially overlapped with GFP- and scn8aa mRNA-positive neurons (Fig. 3L, M).
Deletion Analysis of Zebrafish Scn8aa Neuronal-specific Regulatory Region
Deletion constructs of scn8aa:GFP (re-termed construct #1 for the deletion analysis) were generated to identify regulatory elements directing the neuronal-specific expression in the 15-kb genomic region. The percentage of GFP-positive embryos, average number of GFP-positive neurons per embryo, and GFP intensity were analyzed following transient expression of each deletion construct (Fig. 4A). The deletion of genomic fragments spanning nucleotides +1,340 to +3,826 (construct #4) or −7,327 to −1,621 (construct #5) did not alter GFP expression, indicating that no element essential for GFP expression was located in these regions. In contrast, the deletion of nucleotides spanning +6,373 to +7,631 (construct #6) reduced GFP expression in all neuronal types examined. Neuronal GFP expression was also reduced by deletion of nucleotides spanning +851 to +2895 (construct #8). The removal of nucleotides spanning +525 to +851 (construct #9, deletion of +525 to +2895) resulted in a greater reduction in the number of GFP-expressing cells and GFP intensity. The further deletion of nucleotides spanning +52 to +525 from construct #9 (construct #10, deletion of +52 to +2895) completely ablated GFP expression. These results demonstrated that four regulatory elements spanning nucleotides +52 to +525, +525 to +851, +851 to +1,340, and +6,373 to +7,631 functioned cooperatively to enhance GFP expression in the nervous system.
The conserved brain-specific regulatory elements of mouse SCN8A are located between nucleotides +141 and +252 at the 3′ end of exon 1 of zebrafish scn8aa (Drews et al.,2007). Since mutations in the brain-specific regulatory elements of mouse SCN8A reduces reporter expression in cultured cells (Drews et al.,2007), we determined whether these conserved elements functioned in an equally potent manner in zebrafish by deleting nucleotides spanning +52 to +525 of #5 to generate construct #11. Surprisingly, the deletion of this region did not reduce the percentage of GFP-positive embryos or the number of GFP-positive neurons (Fig. 4A). However, the GFP intensity of #11-injected embryos in these neurons was lower than that of the construct #5-injected embryos. The prolonged exposure was necessary to observe GFP expression of construct #11 in the spinal cord (Fig. 4B) and brain (Fig. 4C). Analysis of the overall expression profile for each deletion construct revealed that construct #1 and #5 had almost identical percentages of GFP expression among the central and peripheral neuronal types (Fig. 5). There was no mosaic expression in #1- and #5- injected embryos in the transient expression assay. GFP-positive neurons were observed in all of the neuron types examined. The percent of embryos containing GFP-positive neurons derived from the neural crest, including trg, cranial neurons, post lateral line neurons, and RB neurons, was reduced in #6-, #8-, and #9-injected embryos. For these three deletion constructs, the reduction in GFP expression in the hindbrain and spinal cord was less obvious due to the large cell number in these regions.
Putative Transcription Factor Binding Sites Identified in the Regulatory Regions of Zebrafish Scn8aa
The restriction enzyme sites used to generate the deletion constructs are shown in Figure 6A. MATCH (Kel et al.,2003) and MatInspector (Cartharius et al.,2005) programs were used to identify putative transcription factors binding sites in nucleotides +52 to +1,340 and +6,373 to +7,631 of scn8aa (Fig. 6B, C). An RE1-like element was present between nucleotides +7,520 to +7,540. However, deletion of sequences between +6,373 and +7,631 (construct #3 and #6) did not induce ectopic expression of GFP in non-neuronal cells, indicating that the RE1-like element does not suppress scn8aa expression in non-neuronal cells. Putative binding sites for Sox, AML1, En-1, TCF11, Meis1, cdxA, myb, FREAC1, GATA1, GATA2, CAP, and OCT1 were identified, based on a threshold for the core matrix of 100% and the whole matrix of 99.4%. Previous studies indicated that Sox, En1, and AML1 regulated the expression of neuronal genes (Matise and Joyner,1997; Bergsland et al.,2006; Marmigere et al.,2006). Four perfectly matched Sox binding sites [(A/T)(A/T)CAA(A/T)G] and 21 one-mismatched Sox sites were located within the genomic fragment spanning nucleotides +52 to +1340 (Fig. 6B). In addition, one perfectly matched Sox binding site and 15 one-mismatched Sox sites were located within the genomic fragment spanning nucleotides +6,373 to +7,631 (Fig. 6C).
The zebrafish scn8aa gene is expressed in a broad spectrum of neurons in embryos. The expression level is increased in the larval stage and sustained into adulthood, which is responsible for the locomotor activities. We showed that a 15-kb regulatory fragment of zebrafish scn8aa drove the expression of GFP in a temporal and spatial pattern that recapitulated the expression of endogenous scn8aa mRNA. The presence of all four elements in construct #1 and #5 drove the highest GFP expression. The neuronal-specific GFP expression of construct #6, #8, and #9 was retained, despite the reduced percentage of GFP-positive embryos in neural crest–derived neurons. Distant regulatory elements are responsible for the expression of SNAP25 and islet1 genes in motoneurons and sensory neurons (Hwang and Lee,2003; Uemura et al.,2005). However, our data suggested that the same set of elements is utilized to control scn8aa expression in motoneurons and sensory neurons.
The expression pattern of the Tg(scn8aa:GFP) line recapitulated that of scn8aa mRNA, with only one discrepancy. Namely, GFP-positive neurons were scattered in the inner nuclear layer, while scn8aa mRNA was homogenously distributed throughout this region. In zebrafish, the inner nuclear layer of the retina contains four classes of interneurons: bipolar, horizontal, amacrine, and inner plexiform cells (Connaughton et al.,2004). The limited number of GFP-positive cells in the inner nuclear layer may be due to the absence of specific regulatory elements that are responsible for controlling GFP expression in certain types of interneurons.
Spinal interneurons coordinate body movements for fish to turn away from a stimulus. Spontaneous contractions of zebrafish embryos are produced by a neural network that includes interneurons, possibly VeLD, CoPA, and CiD interneurons (Saint-Amant and Drapeau,2001). We found that GFP was expressed in CoPA and CiD interneurons, as well as DoLA, CoSA, VeMe, and CoBL interneurons. The function of scn8a-expressing interneurons will be confirmed when the reduced scn8aa expression in these interneurons specifically disrupts embryonic locomotor activities (Meisler et al.,2004; Chen et al.,2008).
The regulatory regions of the zebrafish scn8aa gene contain putative binding sites for En-1, Runx1/AML1, and the Sox gene family, which play essential roles in neuronal differentiation during neurogenesis. En-1 is expressed in the hindbrain, mid/hindbrain junction, and spinal interneurons (Matise and Joyner,1997). Runx1/AML1 directs differentiation of a set of sensory neurons in DRG during development (Marmigere et al.,2006). The Sox gene family participates in neural crest development and CNS neurogenesis (Wegner and Stolt,2005). Sox5 is expressed in premigratory and migratory neural crest cells (Morales et al.,2007). Sox4 and Sox11 are expressed in the post-mitotic differentiating neurons and serve as transcription activators for establishing pan-neuronal protein expression (Bergsland et al.,2006). Due to the overlapping expression profiles of zebrafish scn8aa and these transcription factors, further study on mutating these binding sites will be performed.
The correlation between the motor dysfunction and reduced expression of human, mice SCN8A, and zebrafish scn8aa strongly warrants the study of transcriptional regulation of this specific subtype of sodium channels (Meisler et al.,2004; Trudeau et al.,2006; Chen et al.,2008). This study has uncovered the genomic regions that direct neuronal expression of zebrafish scn8aa during development, and the results provide a strong starting point for the complete characterization of these regulatory elements.
Adult zebrafish (Danio rerio, Oregon AB line) from the Institute of Zoology, Academia Sinica in Taiwan, served as the breeding stock. Embryos were staged by hours and days post fertilization (dpf). Embryos either transiently or stably expressing the scn8aa:GFP transgene were raised in embryonic medium containing phenylthiourea at 28.5°C.
Zebrafish Genomic Library Screening and 5′ RACE
The lambda FIXII zebrafish genomic DNA library was screened using a 32P-labeled 388-nucleotide cDNA probe containing sequences spanning exon 1, intron 1, and part of exon 2 of scn8aa. Eighteen positive clones obtained in the first screen were subjected to the second screen using a 174-nucleotide cDNA probe that contained sequences spanning exon 1 and intron 1. Six clones remained positive in the second screen, with the longest containing the 15-kb genomic fragment. The 15-kb genomic insert was released by NotI digestion and cloned into pBluescript SK+ vector (Clontech) for further sequencing (accession number EU600239). To identify the transcription start site, mRNA of the 7-dpf larva was reverse transcribed using the scn8aa-specific primer 1 (5′CTTGTTCTCGTCGTCATCATCGCGA3′) and 5′RACE kit (Boehringer Mannheim). The nested PCR conditions were 5 cycles of 15 sec at 94°C, 30 sec at 73°C, and 2 min at 68°C; 5 cycles of 15 sec at 95°C, 30 sec at 70°C, and 2 min 68°C; and 25 cycles of 15 sec at 95°C and 2 min at 68°C. Scn8aa-specific primer 2 (5′GCGGCTTGGCTTTCTCCTCCT3′) and primer 3 (5′TTATGTACCCTCCACGGCAGCC3′) were successively used for the first and second round of PCR. The PCR products were cloned into the pGEMT-easy vector, and colony hybridization was performed using a 174-nucleotide cDNA probe. Positive colonies were confirmed by PCR using scn8aa-specific primer 4 (5′TCCCAACACGACAACTAAAATTAACGCC3′) and primer 5 (5′AGTTTATGATGGAAAGGAAAAA3′).
The 15-kb genomic fragment of scn8aa was inserted into NotI site of the modified multiple cloning sites (MCS) in pEGFP-ITR vector (Hsiao et al.,2001). The resulting construct was used to establish the transgenic lines (designated as scn8aa:GFP) or was used as the starting material (designated as construct #1) for the deletion analysis. Deletion constructs were generated through the restriction enzyme digestions. Construct #2, which contained a large 3′-end deletion, was generated by self-ligating ApaI-digested construct #1. Construct #3 with a shorter 3′-end deletion was generated through self-ligation of end-filled HindIII- and NcoI-digested construct #1. Construct #4 was generated by self-ligating of HindIII internal deletion of construct #1. Construct #5 was generated through inserting the 10-kb ApaI fragment of construct #1 into the ApaI site of pEGFP-1 (Clontech).
To narrow down the regulatory elements containing enhancing activities, we generated six additional constructs. Construct #6 was created by end filling and self-ligating PacI (in 10 kb)/XmaI (in MCS)-digested construct #5. Construct #7 was created by digesting construct #5 with EcoRV at +2,532 nt and PacI at +6,373, and self-ligating. Construct #8 was created by digesting construct #5 with EcoRV at +851 and +2,895 nt, and self-ligating. Construct #9 was created by performing PCR amplification of the fragment spanning nucleotides −1,680 to +525 in construct #5 using primer 6 (5′GTCGACGGTACCGCGGGCCC3′) and primer 7 (5′GCGATATCGACATGTTATTTGAAAGGGGG3′) and inserted into KpnI (in MCS)/EcoRV (at +2,895 nt)-digested #5. The elements conserved in human and mouse were deleted by performing PCR amplification of the fragment spanning nucleotides −1,680 to +52 nucleotide of construct #5 using primer 6 and primer 8 (5′GCGATATCTAGATCGCTGTGCGAGAA3′). The PCR product was inserted into KpnI (in MCS)/EcoRV (at +2,895 nt)-digested #5 to create construct #10. The PCR amplified fragment spanning nucleotides +525 to +2,895 was inserted into EcoRV-digested construct #10 to create construct #11, which specifically deletes the fragment spanning nucleotides +52 to +525.
Microinjection of Zebrafish Embryos and Generation of Stable Scn8aa-GFP Transgenic Lines
Deletion constructs were purified using the Qiagen plasmid kit. Deletion constructs #1 and #4 were linearized with FseI, constructs #2 and #3 with ClaI, and constructs #5–11 with EcoRI. The linearized DNA was injected into the cytoplasm at the one-cell stage using an IM 300 microinjector (Narishige, Japan). Injected embryos that had normal morphology and contained one GFP-positive neuron were scored as GFP-expressing embryos. FseI-linearized construct #1 (250 ng/ml in 0.1 M KCl) was combined with 0.2% phenol red and injected into the embryos. The injected embryos with the highest GFP expression levels were raised to sexual maturity. F0 fish were crossed with one another, and at least 100 embryos were screened for GFP-positive offspring. The founder fish producing GFP-positive offspring were mated with wild type fish to identify the F0 fish with germ line transmission.
Whole-Mount In Situ Hybridization and Immunohistochemistry
Whole-mount in situ hybridization was performed as described (Tsai et al.,2001). Hybridization was performed with a digoxigenin-labeled antisense riboprobe containing the carboxyl-terminal and 3′-noncoding regions of scn8aa. After hybridization, embryos were washed and incubated with anti-digoxigenin antibody, and the color development was performed.
Whole-mount immunohistochemistry was performed as described (Chen et al.,2008). Briefly, embryos were incubated with rabbit anti-GFP antibody (1:5,000) and washed with PBS. The bound antibody was detected using a biotinylated secondary antibody (1:400, Jackson Laboratory). The color was developed using ABC kit (Vector) and DAB with nickel enhancement. Embryos were embedded in agar-sucrose for preparing cross-sections.
Two publicly available TFBS prediction tools, MATCH (Kel et al.,2003) and MatInspector (Cartharius et al.,2005), were used to annotate the putative binding sites of transcription factors in the regulatory regions of scn8aa. The TFBS database Transfac 6.3 (Matys et al.,2003) was used for both tools on all vertebrate matrices. To understand whether the neuronal-specific expression of scn8aa is also controlled by RE-1, we constructed a position weight matrix from known RE1 sites (Schoenherr et al.,1996). Next, PROMSCAN program was used to search for the consensus RE-1 (5′TTCAGCACCACGGACAGCGCC3′) in the 15-kb genomic fragment of zebrafish scn8aa.
We thank Dr. H.J. Tsai for providing the pEGFP-ITR vector. We thank Dr. S.P. Huang and C.J. Huang for assistance with the microinjection technique. We thank S.J. Lin and Dr. K.M. Ho for assistance in screening the zebrafish genomic library.