The Zebrafish Homologue of the Human DYT1 Dystonia Gene Is Widely Expressed in CNS Neurons but Non-Essential for Early Motor System Development

DYT1 dystonia is caused by mutation of the TOR1A gene, resulting in the loss of a single glutamic acid residue near the carboxyl terminal of TorsinA. The neuronal functions perturbed by TorsinA[ΔE] are a major unresolved issue in understanding the pathophysiology of dystonia, presenting a critical roadblock to developing effective treatments. We identified and characterized the zebrafish homologue of TOR1A, as a first step towards elucidating the functions of TorsinA in neurons, in vivo, using the genetically-manipulable zebrafish model. The zebrafish genome was found to contain a single alternatively-spliced tor1 gene, derived from a common ancestral locus shared with the dual TOR1A and TOR1B paralogues found in tertrapods. tor1 was expressed ubiquitously during early embryonic development and in multiple adult tissues, including the CNS. The 2.1 kb tor1 mRNA encodes Torsin1, which is 59% identical and 78% homologous to human TorsinA. Torsin1 was expressed as major 45 kDa and minor 47 kDa glycoproteins, within the cytoplasm of neurons and neuropil throughout the CNS. Similar to previous findings relating to human TorsinA, mutations of the ATP hydrolysis domain of Torsin1 resulted in relocalization of the protein in cultured cells from the endoplasmic reticulum to the nuclear envelope. Zebrafish embryos lacking tor1 during early development did not show impaired viability, overt morphological abnormalities, alterations in motor behavior, or developmental defects in the dopaminergic system. Torsin1 is thus non-essential for early development of the motor system, suggesting that important CNS functions may occur later in development, consistent with the critical time window in late childhood when dystonia symptoms usually emerge in DYT1 patients. The similarities between Torsin1 and human TorsinA in domain organization, expression pattern, and cellular localization suggest that the zebrafish will provide a useful model to understand the neuronal functions of Torsins in vivo.


Introduction
Dystonia is characterized by sustained, patterned involuntary muscle contractions resulting in abnormal postures and twisting movements [1]. The most common form of hereditary dystonia is linked to the DYT1 locus; initial clinical manifestations of DYT1 dystonia usually occur during late childhood and adolescence [2]. DYT1 dystonia is caused by mutations in the TOR1A gene encoding TorsinA [3]. Nearly all patients harbor an in-frame trinucleotide GAG deletion in exon 5 of TOR1A, resulting in the loss of a single glutamic acid (DE) near the carboxyl terminal of TorsinA. The mutant allele gives rise to dystonia that is inherited as an autosomal dominant trait. Only approximately 40% of mutation carriers manifest dystonia [2]. The reasons for this reduced penetrance are not known; several hypotheses have been suggested, including modifying genes and environmental stressors [4].
TorsinA is a member of the AAA + superfamily of proteins (ATPase associated with various cellular activities). Torsins are found only in metazoan animals, and a unique subset has arisen in the vertebrate lineage [4]. Members of the AAA + family have roles in a diverse array of cellular functions [5]. In vitro studies have implicated human TorsinA in numerous cellular processes including cytoskeletal dynamics [6], synaptic vesicle cycling [7] and the secretory pathway [8,9]. TorsinA is expressed in a wide variety of cell types [10] and colocalizes predominately with endoplasmic reticulum (ER) markers [11]. Mutant TorsinA [DE] shows aberrant cellular localization, being redistributed from the ER to the nuclear envelope (NE) in some cell lines [12], and forming cytoplasmic membranous whorls in others [11]. Similar to mutant TorsinA [DE], disruption of the Walker B ATP hydrolysis domain of TorsinA by mutagenesis also resulted in relocalization to the NE [9,12]. Because comparable Walker B domain mutations in other AAA + family members exhibit stabilization of substrate interactions [13,14], the similar redistribution of TorsinA by ATP hydrolysis domain and DE mutations led to the hypothesis that both mutations prevent disengagement of TorsinA from a NE resident protein [15]. However, accumulating data suggest that the ATP hydrolysis domain and DE mutants may not be mechanistically equivalent; the two mutants differ in the formation of membranous whorls [15] and in the strength of co-immunoprecipitation with two putative NE substrates [16].
Although these studies have started to elucidate the cellular functions of Torsins, the mechanisms by which mutant Torsi-nA [DE] causes dystonia are not understood. Despite the dramatic clinical abnormalities, brain tissue from DYT1 dystonia patients is histopathologically unremarkable at autopsy, suggesting that aberrant activity or connectivity in neural circuits might underlie the pathophysiology of dystonia [17]. Consequently, there has been significant interest in generating model systems to gain insights into the functions of TorsinA in neurons and motor circuits in vivo. In C. elegans, mutations in the Torsin-related ooc-5 gene disrupted spindle orientation and PAR protein polarity at the 2-cell stage of development, thereby preventing asymmetric divisions and cell fate determination [18]. The D. melanogastor genome contains a single Torsin family member, dtorsin. Knockdown of dtorsin in the retina by RNA interference altered the cellular organization of pigment granules, suggesting a role in intracellular transport [19]. Analysis of dtorsin 2/2 mutants suggested that dtorsin may act as a positive regulator of GTP cyclohydrolase, an enzyme important in the production of BH4, a limiting cofactor in dopamine synthesis [20]. In mice, multiple strategies have been employed to generate a transgenic model of dystonia. Although these models have yielded insights into the neuronal mechanisms perturbed by expression of TorsinA [DE], none of these models exhibits clinical dystonia [21][22][23][24][25][26][27]. Inactivation of endogenous murine TOR1A by homologous recombination caused perinatal lethality, despite the absence of overt developmental morphological abnormalities. Transgenic mice overexpressing human TorsinA [DE] under the immediate-early promoter from cytomegalovirus (CMV) did not develop dystonia, although a deficit in motor learning was identified [25], as well as a reduced response to amphetamine [28] and changes in the bidirectional plasticity of glutamatergic synapses between cortical neurons and striatal medium spiny neurons [29]. Although these results confirm an important role for TorsinA in motor control, the current strategies have not resulted in a model that would allow physiological, pharmacological and molecular studies to understand the pathophysiology of dystonia and develop treatments for this condition.
There has been increasing interest in the use of zebrafish models to provide novel insights into mechanisms of human disease [30], and a zebrafish model of DYT1 dystonia could be very useful. Zebrafish can be manipulated to be transparent during larval development, allowing the deployment of fluorescent reporter proteins and chemical probes to monitor activity in neural circuits [31], dynamic changes in gene expression [32] and cellular movement and morphology in vivo [33]. Zebrafish can be genetically modified using simple and widely available methods, allowing hypotheses concerning molecular mechanisms to be tested rapidly. Since zebrafish produce large clutches of embryos, screens for genetic modifiers are feasible. In addition, the ability to house zebrafish larvae in 96-well plates facilitates exposure to chemical libraries, permitting discovery of small molecule modifiers and candidate novel therapeutic agents [34]. With specific relevance to dystonia models, the availability of automated neurobehavioral assays would allow high-throughput chemical or genetic screens to be driven by phenotypic abnormalities relevant to the function of motor circuits [35]. Finally, the zebrafish presents an accessible vertebrate model for neurophysiological studies to determine the effects of genetic manipulations on the function of motor systems [36,37]. Zebrafish models have recently been used to study several other movement disorders, including Parkinson's disease [38,39], Huntington's disease [40], and ataxias [41]. The high degree of phylogenetic conservation of genes and molecular pathways relevant to disease pathogenesis, and the many neurochemical and neuroanatomical similarities shared between vertebrates of different classes [42,43], suggests that insights yielded by zebrafish models will be applicable to human disease.
As a first step towards generating a zebrafish model of DYT1 dystonia, we identified the zebrafish homolog of TorsinA. Here, we show that the zebrafish genome contains a single tor1 gene, expressed as two splice variants, in contrast to the dual TOR1A and TOR1B paralogues found in tetrapod genomes. Zebrafish Torsin1 shares a similar protein domain structure with human TorsinA; like its human homologue mutations in the ATP hydrolysis domain of Torsin1 resulted in redistribution of the protein to the nuclear envelope. Similar to murine TOR1A, loss of tor1 during early development did not cause morphological defects in the nervous system, loss of dopamine neurons or deficits in spontaneous movement. Taken together, our findings provide an essential basis for further studies to elucidate the functions of Torsin proteins in CNS function and for future development of a zebrafish DYT1 dystonia model.

Ethics Statement
All studies were carried out in accordance with NIH guidelines for animal care and use, and with approvals from the University of Pittsburgh Institutional Animal Care and Use Committee.

Torsin1 Antibody and Western Blots
Torsin1 antibody was generated by inoculating rabbits with the peptide CPDKEVVEKMAHD derived from the C-terminus of Torsin1, linked at its N terminus to keyhole limpet hemacyanin (New England Peptide). Crude antiserum was affinity purified against the peptide using a column (SulfoLink, Pierce, Rockford, IL) as previously described [47]. Transfected cells or adult zebrafish brains were sonicated in RIPA buffer (10 mM Tris, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% Triton X-100, and 1% sodium deoxycholate, pH7.4) and protein concentrations were measured using BCA assay. Equal amounts of protein were resolved on a 10% polyacrylamide gel, transferred to nitrocellulose, and blocked with 10% milk in TBS (50 mM Tris-HCl, 150 mM NaCl, 0.2% Tween-20). Torsin1 affinity-purified antibody was used at a concentration of 1:1000, and was detected using HRP-conjugated proteinA (1:4000; GE Healthcare, Piscataway, NJ). For peptide competition experiments, Torsin1 antibody was pre-incubated with 760 mM peptide in PBS for 2.5 hours at 37uC.

Cell Culture
HEK cells were maintained in DMEM supplemented with 10% FBS and 50 mg/mL of each penicillin and streptomycin and maintained at 37uC in a humidified, 5% CO 2 incubator. MN9D cells [48] (a gift from Dr. Alfred Heller, University of Chicago) were maintained in DMEM high glucose, supplemented with 10% FBS and 50 mg/mL of each penicillin and streptomycin and maintained at 37uC in a humidified, 5% CO 2 incubator. Cells were grown to 80% confluence and transfected with 2 mg plasmid using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). The following day, cells were transferred to Poly-D coated glass coverslips and allowed to adhere overnight. Cells were fixed in 4% PFA, washed in 1X PBS, counterstained with DAPI and mounted to slides for imaging.

Morpholino Knockdown of tor1
To examine the function of tor1 during development we employed morpholino oligonucelotides (MOs) to target tor1 expression. MOs bind complementary nucleic acid sequences with high affinity, allowing sequence-specific inhibition of splicing or translation by targeting splice consensus sequences in pre-mRNA or the translational initiation sequences of mRNAs [49].

Spontaneous Movement Assays
Spontaneous movement was measured at 5 dpf as previously described [50,51]. Embryos (at least 16 per group) were transferred to individual wells of a 96-well plate. Spontaneous movement was recorded using a video camera, and the resulting video files were analyzed using custom MatlabH algorithms [50] to determine: mean velocity (total displacement/total time of recording); active velocity (total displacement/time spent moving); % time moving (time spent moving/total time of recording); mean rest duration and mean duration of active episodes. All data sets were analyzed by both parametric (ANOVA) and non-parametric (Kruskal-Wallis) tests. Most of the data sets at this age point were normally distributed as evaluated by D'Agostino & Pearson's test. For clarity, raw data are shown along with mean and standard error.

Identification of a Zebrafish TorsinA Homologue
To identify zebrafish proteins with homology to human TorsinA, the zebrafish RefSeq and non RefSeq mRNA databases (www.ncbi.nlm.nih.gov; final accession date: August, 11, 2011) were interrogated by TBLASTN using the human TorsinA protein sequence as a probe. Eight sequences (six transcripts and two predicted transcripts) with homology to human TorsinA were further analyzed. The predicted translations of the zebrafish sequences were aligned with Torsin protein sequences derived from human, chicken and xenopus, as well as the nearest C. elegans and drosophila homologs, using the ClustalX 2.1 algorithm, [52], and a dendrogram was constructed to illustrate the inferred evolutionary relationships between the proteins, using the BEAST program [53] (figure 1). Two putative zebrafish proteins appeared most closely related to Torsin1 family proteins from other vertebrate species. Putative zebrafish orthologues of mammalian Torsin2 and Torsin3 family proteins were also detected. Comparison of the mRNA and genomic sequences for the two putative zebrafish Torsin1 proteins showed that the two transcripts are derived from a single gene by inclusion of alternative 59 exons (figure 2). We refer to the gene (si:ch73-178d14.1) as tor1, and its two transcript variants as tor1_tv1 (NM_001200015) and tor1_tv2 (BC_050957). The resulting proteins are referred to as Torsin1a and Torsin1b respectively. The tor1 gene maps to zebrafish chromosome 21, and spans approximately 11.78 kb of sequence. An expressed sequence tag (EST; AA542632) encoding amino acids 14-140 of Torsin1a was previously reported as TOR1C [54].
The zebrafish genome also contains an additional cluster of four adjacent genes on chromosome 23 predicted to encode Torsin-like proteins. Sequence comparison suggests that these putative proteins are most closely related to Torsin1 and Torsin3 family proteins, although they show less homology than either Torsin1a or Torsin1b to human TorsinA. We found evidence that two of these genes (NP_001124104 and NP_957150) are transcribed. Each of these genes shows divergent exon structure with respect to human TOR1A, suggesting that they are more distantly related than tor1 to TOR1A (see below). The Torsin-like gene cluster was not found in other teleosts and was located in a region of chromosome 23 with little conservation of synteny with respect to other teleosts (not shown).
In order to better understand the relationships between zebrafish and human Torsins, we next examined the genomic context of Torsin1 family genes in the human, and how this is conserved in other vertebrate species (figure 3). In human, the dual torsin1 family members TOR1A and TOR1B are adjacent and inverted with respect to one another. The order of the flanking genes PRRX2-PTGES -TOR1A -TOR1B -C9orf78-USP20-FNBP1 is conserved from mammals to amphibians (figure 3A). In coelacanth, a later diverging fish species, this gene order is also conserved. However, similar to other fish species, coelacanth has a single tor1 gene instead of the dual TOR1A and TOR1B genes found in xenopus and human. This suggests that tor1 was duplicated at the root of the tetrapod lineage.
In earlier diverging fish species (figure 3B), substantial genomic rearrangements with respect to tetrapods and coelacanth disrupt the conservation of synteny surrounding the Torsin1 family genes. fnbp1 is found in close proximity to tor1 in fugu, stickleback and tetraodon, but is located <2 MB upstream of tor1 in the zebrafish genome. Interestingly, the genes immediately flanking TOR1A and TOR1B in human are found in syntenic blocks separated from tor1 in earlier-diverging fish species. Thus ptges and usp20 are adjacent to one another, but located .350 kb from tor1, in fugu, stickleback and teraodon; in zebrafish, these genes and c9orf78 are located on a different chromosome to tor1. However, tor1 otherwise shows conserved syntenic relationships in all four fish species. Given the proximity of tor1 and fnbp1 in three of the species, and the single tor1 gene in coelacanth despite conservation of synteny with human a TOR1A and TOR1B, these data are consistent with the other findings reported here suggesting that zebrafish tor1 and human TOR1A and TOR1B may share a common genomic origin.
tor1 Transcriptional Start Sites, Full cDNA Sequences, Transcript and Genomic Organization We cloned zebrafish tor1 cDNA and mapped its transcriptional start sites by rapid amplification of cDNA ends (RACE), using RNA isolated from adult zebrafish brain tissue. 59RACE was carried out using the tobacco acid pyrophosphatase (TAP) method to allow specific ligation of the RACE adapter to the 59 end of mRNA transcripts (figure 4A). tor1 cDNA was amplified using a 39 primer complementary to exon 2 of the tor1 mRNA (which is shared by both transcript variants) and a 59 RACE adaptor primer, yielding a single PCR product. TAP 2 controls confirmed that amplification of this product was dependent on hydrolysis of the 7-methylguanylate mRNA cap prior to adapter ligation; RT-PCR controls confirmed that reverse transcription was equivalent in both TAP 2 and TAP + samples. These controls show that the boundary between the RACE adapter and the cDNA sequence corresponds to the 59 end of the transcript and hence the transcriptional start site in the genomic sequence. 59RACE products were cloned and sequenced: tor1_tv1 and tor1_tv2 were found to initiate transcription from different promoters located at the 59 end of the gene; the unique first exons in each transcript splice into a common mRNA at exon 2 (figure 2). A single major transcriptional start site was found in tor1_tv1, whereas tor1_tv2 showed two different 59 termini (figure 4B, C). As is commonly found in ubiquitously-expressed genes, neither promoter region contained a consensus TATA box or initiator motif [55].
The 39 end of the tor1 transcript was cloned by 39RACE ( figure 4D). A single polyadenylation signal (GATAAA) was located at position 1862 with respect to the first nucleotide of tor1_tv1. This poly(A) signal is located 24 bp upstream of the poly(A) tail, and is shared by both transcript variants. After allowing for addition of the poly(A) tail, the sizes of the complete tor1_tv1 (1881 bp) and tor1_tv2 (1886 bp) cDNA sequences are compatible with the single <2.1 kb tor1-hybridizing band found on northern blot analysis of brain RNA (figure 4E).
The tor1 cDNA sequences were next mapped to the Zv9 assembly of the zebrafish genomic sequence and the exon boundaries and splice sites determined (figure 5). Both zebrafish tor1 and human TOR1A transcripts contain 5 exons. The phases of all splice boundaries are conserved between TOR1A and tor1. In addition, exons 2, 3, and 4, which encode the majority of the open reading frame, share identical lengths between the human and zebrafish sequences. This striking conservation of genomic organization between the human TOR1A and zebrafish tor1 genes provides further evidence that they descend from a common ancestral gene.

Expression of tor1 in Zebrafish
We next determined the temporal and spatial expression patterns of tor1. mRNA for both tv1 and tv2 was detected by RT-PCR during early embryonic development (not shown). However, isoform-specific probes complementary to the short exon sequences unique to each transcript yielded insufficient signal in whole mount RNA in situ hybridization assays to localize tor1 transcripts. Consequently, we employed a larger cRNA probe complementary to the entire tor1 open reading frame (ORF) shared by both transcripts. This tor1 ORF cRNA probe specifically detects a single tor1 transcript on Northern blot (figure 4E) and shares less than 60% sequence identity with the closest related gene; consequently, the probe is expected to hybridize specifically with tor1 transcripts under the high stringency conditions used for RNA in situ hybridization. At 12 hours post fertilization (hpf), tor1 was expressed throughout the embryo (figure 6A). A sense control probe did not show hybridization at this time point, suggesting that the observed staining pattern was not attributable to nonspecific hybridization. At later developmental points, the pattern appeared unaltered, but the expression level progressively reduced and was barely detectable by 96 hpf. In the adult, both tor1_tv1 and tor1_tv2 were detected by RT-PCR in several tissues including brain, foregut, hindgut, muscle, and gonad (figure 6B). Relative to bactin1, expression of both isoforms was lower in brain and muscle compared to gut and gonad, suggesting reduced expression of tor1 in post-mitotic tissues. In addition, the relative expression of the two transcripts differed between tissues. For example, tor1_tv1 was more abundant in foregut, whereas tor1_tv2 was expressed more strongly in the hindgut. These data suggest that expression of the two transcripts is regulated independently by distinct cis-regulatory elements, as might be predicted from their separate promoters.

Zebrafish Torsin1 Proteins and their Subcellular Localization
The amino termini of Torsin1a and Torsin1b, encoded by exons 1A and 1B, are divergent. However, both are predicted to contain a cleavable ER retention signal, followed by an alphahelical domain and both Torsin1a and Torsin1b are 336 amino acids in length. Alignment of zebrafish Torsin1a and Torsin1b with human TorsinA and TorsinB revealed conservation of key functional domains, including consensus sequences for the ATPbinding Walker A domain, the ATP-hydrolysis Walker B domain, and the Sensor 1, and Sensor 2 domains (figure 7A). Both zebrafish isoforms are 59% identical and 78% similar to TorsinA and 57% identical and 77% similar to TorsinB. Human TorsinA and TorsinB share 66% identity and 77% similarity. The highest degree of divergence between zebrafish Torsin1 and human TorsinA was observed within the amino terminus ( figure 7B). Similarly, maximal divergence between Human TorsinA and TorsinB was also within the amino terminus.
The DYT1 dystonia mutation causes deletion of one glutamate residue from an EE motif in the carboxyl terminal of TorsinA. Although the pathophysiological properties of this mutation are yet to be fully elucidated, one striking property of the mutant is relocalization of the protein from the endoplasmic reticulum to the nuclear envelope [12]. The homologous region in zebrafish Torsin1 contains the amino acids HD (figure 7A), precluding mutagenesis studies to determine whether the zebrafish protein undergoes similar relocalization following introduction of the same mutation. However, redistribution of human TorsinA to the nuclear envelope is also observed when null mutations are introduced into the Walker B ATP hydrolysis domain of TorsinA [9,12]. To investigate whether the subcellular localization of Torsin1a and Torsin1b can be altered by similar mutations, we constructed zebrafish Torsin1-eGFP fusion proteins and examined their localization in transfected cells. Both Torsin1a and Torsin1b isoforms primarily localized to the cytosol ( figure 8A, B) in the mouse dopaminergic MN9D cell line. In a small percentage of cells, Torsin1-eGFP fusion protein could also be detected at the nuclear envelope ( figure 8C). This distribution pattern is consistent with the diffuse intracellular localization observed with human TorsinA [11]. Next, we examined whether mutations in the conserved ATP binding (Walker A) and ATP hydrolysis domains (Walker B) of Torsin1 altered the cellular redistribution of Torsin1. Similar to human TorsinA, mutation (K114T) of the ATP binding domain of Torsin1 [13,56] did not alter the cytoplasmic localization of the Torsin1-eGFP fusion proteins (figure 8A, C). However, Torsin1-eGFP[K114T] was not detected in the nuclear envelope of transfected cells, suggesting that ATP binding is Figure 1. Identification of zebrafish Torsins. Human TorsinA was used as a probe to interrogate the NCBI zebrafish RefSeq and non RefSeq RNA databases by TBLASTN. Eight different zebrafish mRNA and predicted mRNA sequences with homology to TorsinA were translated and aligned with Torsins from C. elegans, D. melanogaster, Xenopus, Chicken, and Human. A dendrogram was constructed to illustrate the inferred evolutionary relationships between Torsin family proteins from different species. doi:10.1371/journal.pone.0045175.g001 required for Torsin1 to enter the nuclear envelope. Conversely, mutation (E177Q) of the ATP hydrolysis domain [13,56] caused accumulation of Torsin1-GFP in the nuclear envelope (figure 8B, C). These results show that (i) similar to human TorsinA, zebrafish Torsin1 can enter the nuclear envelope; and (ii) mutation of the ATP hydrolysis domain of Torsin1 results in its sequestration in this location. Zebrafish Torsin1 proteins thus share key aspects of dynamic cellular localization with human TorsinA, and this depends on at least some of the same functional domains. This suggests that zebrafish Torsin1 and human TorsinA may share a common set of interacting proteins.

Zebrafish Torsin1 Proteins -glycosylation and Expression in Adult Brain
In order to better characterize Torsin1, we next developed a polyclonal antibody against a peptide sequence near the carboxyl terminal, which is divergent between Torsin1 and other zebrafish Torsins (figure 9A). Polyclonal antisera to this antigen were raised in rabbits, and affinity-purified against the peptide. The purified antibody was initially tested by probing western blots of protein lysates from HEK293 cells, which were transfected with a construct expressing tor1_tv2 under the CMV immediate-early promoter, or empty vector as a negative control (figure 9B). A single immunoreactive species of 35 kDa was detected in cells expressing tor1_tv2, whereas no immunoreactive proteins were present in cells transfected with empty vector. This demonstrates that the antibody can recognize Torsin1 expressed in cultured cells. Next, we analyzed adult zebrafish brain lysate by western blot (figure 9C). Two bands were detected with apparent molecular weights of 47 kDa and 45kDa. Detection of both bands was prevented by pre-incubation of the antibody with the cognate peptide, and neither band was present when blots were probed with pre-immune serum. These data confirm that the 47 kDa and 45 kDa species contain the cognate peptide sequence, and therefore most likely represent Torsin1. tor1 is expressed at low levels in embryos, and we did not detect convincing Torsin1immunoreactive signal during early development. Consequently we were not able to further confirm specificity of the antibody for Torsin1 in vivo by showing loss of immunoreactive signal after morpholino knockdown of tor1 expression (see discussion). The apparent molecular weight of Torsin1 differed between cultured cells and the zebrafish brain in vivo. It is known that Torsins are     subject to post-translational modifications, including signal peptide cleavage and N-glycosylation [11,57]. Pre-incubation of brain lysate with peptide:N-glycosidase F (PNGase F), an amidase that cleaves between asparagine and the innermost N-acetyl-glucosamine (GlcNAc) residue of the oligosaccharide chain of N-linked glycoproteins, altered the electrophoretic mobility of the Torsin1immunoreactive species, which migrated at 36-38 kDa following PNGaseF treatment ( figure 9D). This mobility shift demonstrates the presence of N-linked carbohydrate chains totaling approximately 9 kDa. The consensus 'sequon' or target sequence for Nglycosylation is N-X-S or N-X-T, where X is any residue except for proline. There are two predicted N-glycosylation sites in Torsin1, at residues 63 and 164 (figure 9E). The lower apparent molecular weight observed in cultured cells suggests that posttranslational processing of Torsin1 differs between transformed human kidney cells and the zebrafish brain in vivo.
Sections from adult zebrafish CNS were labeled with Torsin1 antibody. Immunoreactive signal was seen in multiple different brain regions and was abolished by pre-incubation of the antibody with the cognate peptide (figure 9F), indicating that the tissue epitope recognized by the antibody is shared by the peptide, and is therefore likely to represent Torsin1. Torsin1-related immunoreactivity was abundant in grey matter regions, such as the optic tectum, cerebellum, brainstem nuclei and central grey matter of the spinal cord ( figure 9F-K). In contrast, white matter tracts, including the optic nerves and tracts, the medial longitudinal fasciculus and long descending tracts of the spinal cord, did not show Torsin1-related immunoreactivity ( figure 9G, K). Cytoplasmic labeling was apparent in many neurons of the brainstem, optic tectum and diencephalon ( figure 9H). Torsin1-related immunoreactivity was also abundant in the neuropil and some neurons showed punctate staining around the perimeter of their cytoplasm, raising the possibility that Torsin1 may also be expressed in nerve terminals (figure 9J). These data suggest that Torsin1 is expressed in the cell body, dendrites and possibly also terminals of neurons in multiple CNS regions, but is not expressed in myelinated fibers.

Physiological Expression of tor1 is not Required for Early Development of Motor Circuits
To determine the function of tor1 during larval development, we targeted its expression by using morpholino antisense oligonucleotides (MOs). We opted to use splice site MOs to inhibit pre-mRNA processing, because knockdown could be verified readily by RT-PCR. Initial studies showed that a combination of two MOs targeting the exon2/intron2 and exon3/intron3 splice donor consensus sequences caused both exons 2 and 3 to be excluded from the transcript ( figure 10A). This resulted in a frame shift and formation of a premature stop codon, which abrogated expression of the Walker B domain and both Sensor domains ( figure 10B). This combination of MOs efficiently disrupted expression of tor1 mRNA up to up to 4 days post fertilization (dpf); densitometry estimated that less than 10% of the tor1 RT-PCR product at 24 hpf was derived from the full-length transcript ( Figure 10C). To exclude phenotypes attributable to adverse effects from microinjection, and non-specific toxicity of MOs, we compared tor1 knockdown zebrafish with wild-type animals, and with animals injected with control non-targeting MO.
There were no differences between uninjected, control MO injected, and tor1 MO injected embryos in any of the parameters we measured. For clarity, data are shown for WT and tor1 knockdown animals only. Zebrafish from all three experimental groups showed normal morphological development through day 7 (not shown). We next evaluated development of dopaminergic neurons, because the function of the dopamine system is disrupted in some patients with dystonia [17]. Whole mount RNA in situ hybridization was employed to label developing dopaminergic neurons using a probe specific to the slc6a3 transcript encoding the dopamine transporter, which is expressed in dopamine neurons [58,59]. We did not observe any differences between wild type larvae and tor1 morphant larvae in the position of dopaminergic neurons or in the intensity of slc6a3 expression at 48 hpf or 72 hpf ( figure 10D and data not shown). The total number of diencephalic slc6a3 + neurons was determined in each population of zebrafish; knockdown of tor1 did not affect the number of dopamine neurons at 48 or 72 hpf (figure 10E and data not shown). We conclude that, similar to murine TOR1A 2/2 lines [22], knockdown of tor1 in zebrafish did not give rise to abnormalities in the number or position of dopamine neurons during development.
Finally, we asked whether knockdown of tor1 provoked abnormalities of motor function by employing a previously described assay [50,51] to quantify spontaneous propulsive movements in uninjected larvae, control MO injected larvae, and Torsin1 morphant larvae. Compared with controls, larvae lacking physiological expression levels of Torsin1 did not show statistically significant differences in mean velocity, percent time moving, active velocity, mean duration of movement bursts, or mean rest duration between movements at 4-6 dpf (figures 10F-J show data at 5 dpf). At earlier developmental time points before 4 dpf, when tor1 knockdown was most prominent, larval propulsive movements are infrequent and variable, making it difficult to detect differences between experimental groups [51]. To evaluate zebrafish between 1-4 dpf, we employed a second method that determines changes in pixel grayscale values from frame to frame of a video stream, rather than detecting displacement of the larval centroid. The pixel quantification method is sensitive to both propulsive and non-propulsive movements, such as coiling and reorientation movements, and consequently has previously allowed us to detect differences in motor activity between experimental groups at early developmental points [60]. However, even using this method, we did not detect differences between the motor activity of controls and tor1 knockdown animals between 1-4 dpf (data not shown). Taken together, these findings suggest that wildtype expression levels of Torsin1 are not required for early morphological development and function of the motor system.

Discussion
The data presented here suggest that a novel zebrafish gene, tor1, is a single zebrafish homolog of the mammalian TOR1A and TOR1B genes. Conservation of sequence and genomic organization provide strong evidence that the zebrafish and human genes share a common ancestor, and we found notable similarities between the expression patterns of the human and zebrafish mRNAs, and the domain organization and subcellular localization of the proteins. There is limited conservation of synteny between teleost and tetrapod genomes in the immediate vicinity of the tor1 genes, suggesting that substantial genomic reorganizations occurred in this region during evolution. One of the more striking differences is the presence of adjacent TOR1A and TOR1B genes Figure 10. Torsin1 is non-essential for early development of the motor system. A: Single-cell zebrafish embryos were microinjected with morpholino oligonucleotides designed to block splicing of tor1 exons 2 and 3 (see figure 2). At the times indicated between 24 and 120 hours postfertilization, RNA was harvested from developing zebrafish and subjected to RT-PCR using primers specific for tor1_tv1 (upper panel) or tor1_tv2 (lower panel). The pictures show ethidium bromide stained agarose gels after the PCR products were resolved by electrophoresis. The positions of PCR products corresponding to wild-type tor1 mRNA and the transcript lacking exons 2 and 3 are shown. B: A schematic representation is shown of the predicted protein products produced by transcripts lacking exons 2, 3 or both exons. It is not known if any of these truncated products is stable. C: The % tor1 RT-PCR product arising from the wild-type, full-length mRNA at each time point following MO injection was estimated by densitometry. in tetrapods, whereas the genomes of zebrafish, fugu, stickleback, tetraodon and coelacanth all contain a single tor1 gene. These data strongly suggest that an ancestral tor1 locus was duplicated during the fin-to-limb transition, resulting in the occurrence of distinct TOR1A and TOR1B genes before the emergence of amphibious tetrapods. Whether the separate TOR1A and TOR1B genes have divergent functions remains to be determined; TorsinA and TorsinB show complementary expression patterns [61] and it is possible that they mediate similar cellular functions in different groups of cells, similar to the numerous other paralogous proteins with distinct expression patterns that show redundant functions [62]. In support of the possibility that TorsinA and TorsinB have redundant functions, there is evidence that TorsinB may partially compensate for the functional effects of mutations in TorsinA: morphological abnormalities of the nuclear envelope in fibroblasts derived from DE/DE knock-in mice were exacerbated by loss of TorsinB [61]. Consequently, it is possible that the single tor1 gene in zebrafish will prove advantageous for determining the role of Torsin1 in neurons, since loss-of-function phenotypes might not be mitigated by compensatory functions provided by closely related proteins. Similar to the relationship between human TorsinA and TorsinB, zebrafish Torsin 1a and Torsin 1b are most divergent within the N-terminal region. It is possible that discrete functions of TorsinA and TorsinB are distributed between the two isoforms encoded by the tor1 gene in zebrafish. Compatible with this idea, the expression pattern of tor1 showed similarities to both TOR1A and TOR1B. The relatively weak expression of both tor1_tv1 and tor1_tv2 in the brain, compared to gut and gonads resembles the expression of mammalian TOR1B, which is more abundantly expressed in somatic cells than neurons. Conversely, Torsin1related immunoreactivity in the zebrafish nervous system appeared to closely resemble the expression pattern of TorsinA in the mammalian nervous system, with prominent localization in the cell body and processes of neurons and absence from white matter tracts. The development of isoform-specific antibodies to zebrafish Torsin1a and Torsin1b will allow clarification of whether the expression patterns of these two proteins are complementary and similar to mammalian TorsinA and TorsinB.
In order to characterize the zebrafish Torsin1 protein, we developed a novel antibody directed towards its C-terminus. We showed that: (i) the antibody recognizes a protein produced by expression of cDNA encoding tor1 in cultured cells; and (ii) preincubation of the antibody with its cognate peptide prevented it from detecting epitopes on either western blot or tissue sections. These controls support the conclusion that the antibody specifically detected Torsin1 in zebrafish brain. However, loss of immunoreactive signal in genetic knockout tissue is generally considered a 'gold standard' for proving antibody specificity [63]; in this case, the absence of clearly demonstrable Torsin1 immunoreactivity during early development precluded confirmation of specificity by MO-mediated knockdown of the tor1 transcript. Formal proof that the antibody specifically detects Torsin1 in the zebrafish brain therefore awaits the development of stable tor1 2/2 lines. Nonetheless, the availability of this antibody allowed some initial conclusions to be drawn about the nature of Torsin1: it is expressed as two species in CNS, a major 45 kDa form and a minor 47 kDa form; both species are heavily Nglycosylated, showing loss of 9 kDa carbohydrate after deglycosylation; and the protein localizes to neuronal somatic cytoplasm and possibly also nerve terminals within the neuropil, but was not detected within axons. These properties are similar to those previously reported for human TorsinA, and are compatible with an endoplasmic reticulum subcellular localization in neurons. The origin of the two protein species is unclear; the predicted molecular weights of Torsin1a and Torsin1b are similar (38.2 kDa and 38.4 kDa, respectively) and approximately the size of the major protein isoform found after deglycosylation. It is possible that different post-translational modifications are responsible for the two forms, or that signal peptide cleavage is directed to a different site in the nascent protein by each of the unique Ntermini.
Relocalization of TorsinA from the endoplasmic reticulum to the nuclear envelope is a hallmark of the dystonia-associated DE mutation [12]. This relocalization can also be induced by mutations in the ATP-hydrolysis domain of TorsinA [9,12]. Although we were not able to introduce the DE mutation into Torsin1, because of limited sequence conservation with respect to the human protein in this region of the C-terminus, we have demonstrated that zebrafish Torsin1 can be made to accumulate in the nuclear envelope by the introduction of a putative ATPhydrolysis null mutation in the Walker B domain. Interestingly, this relocalization was observed in a heterologous murine cell line. Trafficking of Torsins between nuclear envelope and endoplasmic reticulum is thought to depend on their interactions with specific substrates and with other Torsin molecules in hexameric complexes. This raises the possibilities that the zebrafish protein may interact with: (i) one or more mammalian nuclear envelope proteins, possibly some of the same proteins that interact with endogenous mammalian Torsins; or (ii) endogenous mammalian TorsinA to form complexes. These possibilities are not mutually exclusive and both are equally compatible with the data presented. For example, even if zebrafish Torsin1 did not recognize murine nuclear envelope substrates, disruption of its ATPase domain may have prevented a Torsin1/TorsinA heterohexamer from disengaging from a substrate. These data show an intriguing degree of functional interaction between the heterologous proteins and imply that at least some of the cellular roles may be phylogenetically conserved. It will be of interest to determine whether human TorsinA behaves similarly in zebrafish cells, since this will have significant implications for generating a zebrafish model of DYT1 dystonia based on transgenic expression of human TorsinA [DE]. Zebrafish homologues of many of the human proteins known to interact with TorsinA have been described, and we have identified candidates for others by database searching (Table S1). It will be of considerable interest to determine whether these proteins interact with Torsin1, suggesting phylogenetic conservation of biochemical properties and functions.
We did not identify any morphological or neurobehavioral abnormalities resulting from reduced Torsin1 expression early in development, although the use of RNA in situ hybridization to label dopamine neurons does not allow examination of neuronal morphology and so the data do not exclude, for example, abnormalities of dendritic branching. The absence of prominent phenotypes in morphant larvae could be attributable to incomplete knockdown of Torsin1 (figure 10A, C) or compensatory functions provided by other Torsin proteins. However, these data are also consistent with findings in murine TOR1A 2/2 lines, in which the first observed phenotypic abnormality was perinatal lethality due to lack of feeding [22]. In zebrafish, the analogous developmental point occurs after 5 dpf, when patterning of the nervous system is nearly complete and larvae begin actively pursuing food. Unfortunately, by this time point, the expression of both tor1 transcripts had largely recovered from transient MOmediated knockdown (figure 10A, C); consequently it remains unclear whether loss of tor1 in zebrafish at later developmental points would provoke similar abnormalities to those observed in mice lacking TorsinA. DYT1 dystonia patients usually first show symptoms during late childhood or adolescence. Consequently, it is expected that critical CNS functions of TorsinA, of relevance to the pathogenesis of dystonia, could occur much later during development. In order to address these possibilities and to elucidate the role of Torsin1 during later development, it will be necessary to develop stable tor1 2/2 lines. This would not only allow unambiguous evaluation of Torsin1 functions at later time points, but will also facilitate validation of Torsin1 antibodies for further biochemical studies.
In conclusion, our data suggest that the zebrafish may provide a powerful model system to study the neuronal functions of Torsin1 in vivo, and an opportunity to elucidate the molecular evolution of Torsins. Our work provides essential background for future studies on tor1 gene function and raises the intriguing possibility of generating a transgenic zebrafish model of DYT1 dystonia, with potential applications in understanding pathogenesis and developing novel treatments.

Supporting Information
Table S1 Zebrafish homologs of TorsinA-interacting proteins. Human proteins known to interact with TorsinA are shown in the first column and details of their genes shown in the second column. The third column shows details of the proposed zebrafish homologues of these genes (references are shown where available; otherwise, Sager et al this publication). The final column shows the degree of homology between the human and zebrafish proteins as % amino acid residues that are identical or homologous between human and zebrafish. (PDF)