Defective Escape Behavior in DEAH-Box RNA Helicase Mutants Improved by Restoring Glycine Receptor Expression

RNA helicases regulate RNA metabolism, but their substrate specificity and in vivo function remain largely unknown. We isolated spontaneous mutant zebrafish that exhibit an abnormal dorsal bend at the beginning of tactile-evoked escape swimming. Similar behavioral defects were observed in zebrafish embryos treated with strychnine, which blocks glycine receptors (GlyRs), suggesting that the abnormal motor response in mutants may be attributable to a deficit in glycinergic synaptic transmission. We identified a missense mutation in the gene encoding RNA helicase Dhx37. In Dhx37 mutants, ribosomal RNA levels were unchanged, whereas GlyR α1, α3, and α4a subunit mRNA levels were decreased due to a splicing defect. We found that Dhx37 can interact with GlyR α1, α3, and α4a transcripts but not with the GlyR α2 subunit mRNA. Overexpression of GlyR α1, α3, or α4a subunits in Dhx37-deficient embryos restored normal behavior. Conversely, antisense-mediated knockdown of multiple GlyR α subunits in wild-type embryos was required to recapitulate the Dhx37 mutant phenotype. These results indicate that Dhx37 is specifically required for the biogenesis of a subset of GlyR α subunit mRNAs, thereby regulating glycinergic synaptic transmission and associated motor behaviors. To our knowledge, this is the first identification of pathologically relevant substrates for an RNA helicase.


Introduction
RNA helicases belong to a large DNA/RNA helicase superfamily and are classified into Upf1-like, Ski2-like, RIG-I-like, NS3/ NPH-II, DEAD-box, and DEAH-box subfamilies (Bleichert and Baserga, 2007). These RNA helicases take part in various aspects of RNA metabolism. In yeast, the DEAH-box protein Dhr1p regulates the synthesis of ribosomal RNA (rRNA) and is involved in the preribosomal complex (Colley et al., 2000;Grandi et al., 2002). However, the biological function of the vertebrate homolog Dhx37 has not been investigated. In vivo physiological functions of RNA helicases remain largely unknown, with the exception of a few involved in disease (Moreira et al., 2004;Boon et al., 2007;Pena et al., 2007). Preference for certain substrate RNAs was reported only for RNA helicase A (Hartman et al., 2006). Whether other RNA helicases have specific targets remains unknown.
Zebrafish represent a useful model for motor study. At 2 d postfertilization (dpf), tactile stimulation evokes an escape behavior, which consists of an initial turn and subsequent swimming (Saint-Amant and Drapeau, 1998). Forward genetics has identified zebrafish mutants as being defective in this escape behavior (Granato et al., 1996). For example, bandoneon mutants carry mutations in glrbb, encoding the glycine receptor (GlyR) ␤b subunit. Since GlyR ␣ and ␤ subunits form heteromeric pentamers, with the ␤ subunit being essential for the synaptic clustering of GlyRs (Lynch, 2004), the GlyR ␤b mutants exhibited a loss of inhibitory glycinergic synaptic transmission. Due to the consequent loss of the reciprocal inhibition between the left and right sides of the spinal cord, motor neurons activate simultaneously on both sides (Grillner, 2003;Fetcho et al., 2008), which results in bilateral muscle activation and thus dorsal flexure of the body in GlyR ␤b mutants . The same motor deficit is observed in zebrafish embryos treated with a high concentration of strychnine, a GlyR antagonist (Granato et al., 1996;McDearmid et al., 2006). On the other hand, the application of strychnine at a lower concentration causes a transient dorsal bend followed by swimming.
In this study, we characterized a zebrafish mutant, dhx37 nig1 , that exhibits a tactile-evoked dorsal bend, followed by swimming. Positional cloning revealed a missense mutation in the DEAHbox RNA helicase Dhx37. In dhx37 nig1 mutants, mRNA levels of selected GlyR ␣ subunit genes were decreased, whereas rRNA levels were unchanged. Indeed, our RNA analyses showed that Dhx37 binds to a subset of GlyR transcripts and regulates their splicing. Overexpression of selected recombinant GlyR ␣ subunits restored the normal touch response in Dhx37-depleted embryos. Thus, Dhx37 plays an essential role in the biogenesis of GlyR ␣ subunit mRNAs and is indispensable for normal escape behavior in vertebrates.

Materials and Methods
Animals. Zebrafish were bred and raised according to the guidelines set forth by the National Institute of Genetics. The dhx37 nig1 mutation was found in zebrafish breeding stock obtained from a pet shop. Animals from either sex were used in this study.
Video recording. Embryonic behaviors were recorded using a highspeed camera at 200 frames per second (HAS-220, Ditect), as previously described .
Mapping, cloning, mRNA rescue, and antisense knockdown. A mutant carrier fish was crossed with a WIK fish for meiotic mapping. Cloning, mRNA rescue, and antisense knockdown were performed as described previously . The following primers and morpholino oligonucleotides (MOs) were used. We performed injection of MOs and/or RNAs in three independent experimental trials and obtained consistent results: To pharmacologically increase glycinergic transmission for behavioral rescue, NFPS (N[3-(4Ј-fluorophenyl)-3-(4Ј-phenylphenoxy)propyl] sarcosine, Santa Cruz Biotechnology), a blocker of Gly transporter 1 (GlyT1), was injected into mutants as described previously .
RNA analysis. Northern blotting was performed using a full coding sequence as a probe. Quantitative PCR (qPCR) was performed using a KAPA SYBR Fast qPCR Kit (Kapa Biosystems). The following PCR primers were used: GlyR ␣1: Immunostaining. Immunostaining of spinal cord sections was performed as described previously . The following anti- Figure 1. Zebrafish nig1 mutants display a dorsal bend in a tactile-evoked escape behavior. A, At 2 dpf, wild-type embryos showed a lateral turn (level 1). B, In the presence of 30 M strychnine, some wild-type embryos exhibited a dorsal bend followed by swimming (level 2). C, In the presence of 70 M strychnine, most wild-type embryos displayed a dorsal bend without subsequent swimming (level 3). D, Mutant embryos responded to touch with dorsal flexure of the body followed by swimming. E, F, In the presence of 30 and 70 M strychnine, mutants responded to tactile stimulation with a dorsal bend; this was sometimes but not usually followed by swimming. G, The histogram represents the ratio of embryos exhibiting level 1, 2, or 3.

Movie 1.
Touch-induced escape behavior at 2 dpf. Clip 1, Tactile stimulation of wild-type zebrafish embryo via the tail induced an escape response. Clip 2, Most nig1 mutants responded to touch with a dorsal bend of the body followed by escape swimming. Clip 3, Some of the nig1 mutants exhibited a strong dorsal bend without subsequent swimming. Clip 4, Level 1 (normal), exhibiting a lateral turn and subsequent burst swimming. Clip 5, Level 2 (mild), Exhibiting a bend to the dorsal side instead of a lateral turn followed by swimming of Ͼ2 cm. Clip 6, Level 3 (severe), Exhibiting a dorsal bend without escape swimming. Clip 7, Some NFPS-injected mutants responded to touch without dorsal bend at the onset of the escape.

Results
dhx37 nig1 mutants display abnormal escape behavior due to a defect in glycinergic synaptic transmission A recessive mutation, nig1 (later referred to as dhx37 nig1 ), was identified in our breeding stock of zebrafish. At 2 dpf, tactile stimulation of wild-type zebrafish induced an escape response that consists of an initial turn to the lateral side and subsequent swimming (Movie 1, Clip 1). The initial turn is executed by activation of the trunk muscles on one side, and the subsequent swimming is mediated by alternating muscle contractions on either side of the trunk. By contrast, most nig1 mutants (92%, 44/48 mutants) responded to touch with a dorsal bend of the body followed by swimming (Movie 1, Clip 2). The rest of the mutants (8%, 4/48 mutants) exhibited a strong dorsal bend without subsequent swimming (Movie 1, Clip 3). Since a dorsal flexure of the trunk is typically seen in embryos treated with strychnine, a GlyR antagonist, we assayed the touch response following exposure to a low or high dose of strychnine in the bath and classified the response into the following three groups: level 1 (normal), exhibiting a lateral turn and subsequent swimming (Movie 1, Clip 4); level 2 (mild), exhibiting a dorsal bend followed by swimming of Ͼ2 cm (Movie 1, Clip 5); and level 3 (severe), exhibiting a dorsal bend without escape swimming (Ͻ2 cm; Movie 1, Clip 6). In the absence of strychnine, all of the wild-type embryos exhibited a normal response (Fig. 1 A, G). Application of 30 M strychnine induced level 2 (21%, 17/81) and level 3 (38%, 31/81) responses in wild-type embryos (Fig. 1B). This strychnine-induced motor deficit was dose dependent, because treatment with 70 M strychnine worsened the response (level 3: 88%, 57/65; Fig. 1C). Without strychnine, most mutant embryos exhibited the level 2 phenotype (Fig. 1D). Treatment with 30 M strychnine induced severe motor deficits (level 3: 78%, 40/51; Fig. 1E), as did treatment with 70 M strychnine (level 3: 87%, 48/55; Fig. 1F ). These results suggest that glycinergic transmission is compromised in nig1 mutants. Although nig1 mutants did not show apparent developmental defects, they became thinner and died at 7-10 dpf, possibly from an ineffective motor response and feeding difficulties.
To see whether glycinergic transmission is affected in nig1 mutants, we measured miniature glycinergic currents in motor neurons at 2 dpf. In wild-type zebrafish, spontaneous glycinergic synaptic currents in the presence of TTX, CNQX, APV, bicuculline, and D-tubocurarine, which block Na ϩ channel, AMPA, NMDA, GABA A , and nACh receptors, respectively, were observed at 0.39 Ϯ 0.17 Hz (n ϭ 7; Fig. 2A), with a frequency comparable to a previous report (0.33 Ϯ 0.07 Hz; Hirata et al., 2005). However, the spontaneous events were less frequently seen in mutants (0.08 Ϯ 0.05 Hz, n ϭ 7, p Ͻ 0.01; Fig. 2B), suggesting that glycinergic synaptic transmission is indeed decreased in mutants. GlyT1 is known to remove glycine from the synaptic cleft, and the blockade of GlyT1 enhances glycinergic transmission Eulenburg et al., 2005;Mongeon et al., 2008). We therefore injected NFPS, an inhibitor of GlyT1, into the ventricles of mutant embryos. Although 28% of NFPS-injected mutants (5/18 mutants) became immobile, 17% of the mu- . mRNA levels for GlyR ␣1, ␣3, and ␣4a subunits are decreased in dhx37 nig1 mutants. A, The graph represents the ratio of mutant transcripts compared with wild-type fish estimated by qPCR. B, Northern blots of total RNA. Note that mRNA levels of GlyR ␣1, ␣3, and ␣4a subunit and gephyrin b decreased in mutants. C, RT-PCRs to detect unspliced products in mutant transcripts. Unspliced transcripts for GlyR ␣4a and gephyrin b were retained in mutants. Genome lanes refer to the validation of unspliced products by genomic PCR. ϩRT, With reverse transcription; ϪRT, without reverse transcription. D, HEK293 cells were transfected with GlyR cDNAs and Dhx37-FLAG expression vector. Cell extracts were subjected to immuneprecipitation with anti-FLAG, followed by RT-PCR. Transcripts for GlyR ␣1, ␣3, and ␣4a subunits but not of ␣2 subunits were coprecipitated with Dhx37-FLAG. The presence of Dhx37-FLAG in the immunoprecipitates was verified by Western blotting with an anti-FLAG. tants (3/18 mutants) responded to touch without dorsal bend at the onset of the escape Movie 1, Clip 7). These results indicate that glycinergic synaptic transmission is compromised in nig1 mutants.
To determine whether nig1 mutants have morphological defects at glycinergic synapses, the distribution of gephyrin, GlyRs, and acetyl tubulin was assessed by immunolabeling of spinal cord sections. Gephyrin and acetyl tubulin were found within the lateral region of the mutant spinal cord as in the wild type (Fig. 2C,D). Synaptic localization of GlyR clusters was also found at the lateral spinal cord in wild type but with lower intensity in mutants (Fig. 2 E, F ). Quantitative analysis confirmed that GlyR signals in mutant spinal cord was lower (0.56 Ϯ 0.11, n ϭ 6, p Ͻ 0.05; Fig. 2G) compared with wild-type zebrafish (n ϭ 6). Western blotting of whole embryo protein extracts with anti-GlyR␣ also showed that GlyR ␣ subunit levels were decreased in mutants (0.37 Ϯ 0.15, n ϭ 6, p Ͻ 0.01; Fig. 2 H, I ) compared with wild-type zebrafish (n ϭ 6), whereas the amount of gephyrin protein (1.06 Ϯ 0.21, n ϭ 6) and acetyl tubulin protein (1.10 Ϯ 0.14, n ϭ 6) was unchanged. These results, along with those of our behavioral assays and electrophysiology testing, indicate that glycinergic transmission is compromised due to a reduction of GlyR subunit levels in nig1 mutants.

nig1 encodes the DEAH-box RNA helicase Dhx37
We next performed genetic mapping of nig1 locus and identified a missense mutation (L489P) in an RNA helicase, DEAH-box protein 37 (Dhx37; GenBank accession number AB739007). To address whether the L489P mutation is responsible for dhx37 nig1 mutant phenotype, we injected control (green fluorescent protein) RNA, wild-type Dhx37 RNA, or Dhx37 L489P RNA into recently fertilized embryos from dhx37 nig1 heterozygous carriers and assayed touch responses. A quarter of control RNA-injected progeny (29%, 22/75) and Dhx37 (L489P) RNA-injected progeny (27%, 20/75) displayed the mutant response, whereas most of the progeny injected with wild-type Dhx37 RNA (94%, 61/65) exhibited normal escape behavior, and only 6% (4/65) displayed an abnormal response. The molecular identification of dhx37 as the causative gene was also supported by knockdown of Dhx37 using antisense MOs. Antisense Dhx37 MO1, which blocks translation of Dhx37, was injected into wild-type embryos. Most MO1-injected embryos (82%, 71/87) displayed an abnormal touch response, whereas all control MO-injected embryos exhibited a normal response. Similarly, injection of Dhx37 MO2, which interferes with splicing of dhx37 mRNAs also induced a mutant phenotype (91%, 84/92). These RNA rescue and antisense knockdown experiments confirm that the dhx37 nig1 mutation is responsible for the abnormal dorsal bend in the mutant escape behavior.
Some RNA helicases are involved in pre-mRNA splicing. Since qPCR can detect transcripts that are not fully spliced (i.e., could have a retained intron at another junction distant to the qPCR site), this method does not accurately measure the levels of fully spliced mRNAs. To quantify the amount of fully spliced mRNA, we performed Northern blotting and confirmed that GlyR ␣1, ␣3, and ␣4a subunits and gephyrin b mRNAs were decreased in dhx37 nig1 mutants (␣1: 0.34 Ϯ 0.17, n ϭ 3, p Ͻ 0.05; ␣3: 0.26 Ϯ 0.12, n ϭ 3, p Ͻ 0.05; ␣4a: 0.24 Ϯ 0.11, n ϭ 3, p Ͻ 0.01; gephyrin b: 0.37 Ϯ 0.15, n ϭ 3, p Ͻ 0.05; Fig. 3B). The level of actin ␤1 mRNA was unchanged in mutants (1.03 Ϯ 0.14, n ϭ 3). We also tried to detect GlyR ␤a mRNA, but specific bands were not detectable even in wild-type fish, presumably because the expression level is very low at this stage (Hirata et al., 2005). The amounts of 28S and 18S rRNA were comparable between wild- Figure 4. Expression of GlyR ␣1, ␣3, or ␣4a subunits restored the normal touch response in Dhx37-deficient embryos. A, Injection of MO2, which blocks the splicing of Dhx37, into wild-type embryos led to a dorsal bend response. Coinjection of either GlyR ␣1, ␣3, or ␣4a subunit RNAs with MO2 decreased the abnormal behavior and restored the normal response. Mixed application of GlyR ␣1, ␣3, and ␣4a subunit RNAs with MO2 also restored the normal escape. Neither GlyR ␣2 nor ␣4b subunit RNAs had an effect on the motor recovery. The histogram represents the ratio of embryos exhibiting level 1, 2, or 3. B, All of the wild-type embryos injected with control MO exhibited a lateral turn in the escape response. Injection of GlyR ␣1, ␣3, or ␣4a MOs alone induced the level 2 response (␣1 MO: 8%; ␣3 MO: 21%; ␣4a MO: 65%). Mixed application of ␣1 and ␣3 MOs, ␣1 and ␣4a MOs, or ␣3 and ␣4a MOs increased the level 2 response rate compared with single MO injections. Most of the embryos injected with all three MOs (␣1, ␣3, and ␣4a) exhibited the severe level 3 phenotype. Injection of GlyR ␤a MO did not affect the touch response. GlyR ␤b MO-injected embryos displayed level 3 motor deficits. type and mutant fish (28S rRNA: 0.98 Ϯ 0.07, n ϭ 3; 18S rRNA: 1.04 Ϯ 0.13, n ϭ 3). Our qPCR and Northern blot analyses indicate that mRNA levels for GlyR and gephyrin are decreased in dhx37 nig1 mutants, whereas rRNA biogenesis remains unchanged.
We next assessed the splicing of GlyR and gephyrin genes by RT-PCR, revealing that unspliced transcripts for the GlyR ␣4a subunit are increased in dhx37 nig1 mutants (Fig. 3C). Similarly, unspliced gephyrin b products were upregulated. By contrast, splicing defects in actin ␤1 transcripts were not observed in dhx37 nig1 mutants. We also attempted to detect unspliced transcripts for GlyR ␣1 and ␣3 subunit genes, but they were not accessible, because all of the introns are too long to amplify alongside normal short products. These results suggest that Dhx37 is involved in pre-mRNA splicing and that Dhx37 has substrate specificity for certain transcripts important for glycinergic synapse function.
The substrate specificity of Dhx37 was further investigated using an RNA immunoprecipitation assay. HEK293 cells were cotransfected with GlyR cDNAs that contained 5Ј-and 3Ј-UTR sequences together with a FLAG-tagged Dhx37 expression vector. Cell extracts were immuneprecipitated with anti-FLAG or with control IgG and subjected to RT-PCR. FLAG-Dhx37 immunoprecipitates contained GlyR ␣1, ␣3, and ␣4a subunit transcripts, but not ␣2 subunit transcripts (Fig. 3D). Taken together, this assay clearly indicates that Dhx37 has substrate specificity for selected GlyR ␣ subunit transcripts.

Discussion
In summary, we have identified a zebrafish dhx37 nig1 mutant that displays an abnormal dorsal bend at the beginning of the escape response. GlyR ␣1, ␣3, and ␣4a subunit mRNAs were downregulated and mis-spliced in dhx37 nig1 mutants. The depletion of GlyR ␣1, ␣3, and ␣4a subunits is responsible for the motor deficits observed, since these could be improved by restoring expression of these GlyR ␣ subunits. Mutations in human DHX37 may cause startle disease/hyperekplexia associated with defective glycinergic transmission.

Substrate specificity of Dhx37
The physiological significance of RNA helicases has been characterized in only a few cases. For example, mutations in the RNA helicase senataxin are responsible for neurodegenerative diseases (Moreira et al., 2004). Retinitis pigmentosa is caused by a dysfunction of pre-mRNA splicing factor 8 (Boon et al., 2007;Pena et al., 2007). However, the pathologically relevant substrate of these RNA helicases remains unknown. We demonstrated that Dhx37 interacts with selected GlyR ␣ subunit mRNAs and specifically regulates biogenesis of these transcripts. Does Dhx37 have the other targets? Most embryos injected with MO2 and Dhx37 RNA displayed a normal touch response (90%, 28/31), while embryos injected with MO2 and mixture of GlyR ␣1, ␣3, and ␣4a subunit RNAs showed motor recovery at a lower efficiency (56%, 18/32). This suggests that Dhx37 has targets other than GlyR subunit mRNAs.

Subunit composition of GlyRs
The initial characterization of purified GlyRs suggested that GlyRs were composed of 3␣ and 2␤ subunits (Langosch et al., 1988), but recent functional and structural studies revised the stoichiometry to 2␣3␤ (Grudzinska et al., 2005;Dutertre et al., 2012;Yang et al., 2012). Although our knockdown experiments indicate that ␣4a appears to be the major GlyR ␣ subunit involved in the escape behavior, the ␣1 and ␣3 subunits also contribute. We also showed that the ␤b but not ␤a subunit is essential for touch responses. This is consistent with our previous finding that the ␤b subunit is expressed by 1 dpf, while the expression of ␤a is very low until 3 dpf . Taken together, it is likely that GlyR ␣1, ␣3, and ␣4a subunits coassemble with the GlyR ␤b subunit, either in different receptor complexes or in heteromeric combinations, and that knockdown of all three ␣ subunits is required to mimic the severe GlyR ␤b-deficient phenotype.