The N17 domain mitigates nuclear toxicity in a novel zebrafish Huntington’s disease model

Since our earlier experiments had demonstrated significant toxicity leading to premature death of transgenic zebrafish upon over-expression of polyQ expanded HTT, we decided to create Cre-loxP conditionally inducible transgenic fish. This system allows us to maintain the transgenic lines by bypassing any toxicity by keeping the transgene switched off until expression of Cre recombinase is introduced, which also allows us spatial and temporal control of expression. We chose to use the strong, ubiquitous bactin2 promoter provided with the tol2kit to drive transgene expression [42]. Each transgene was assembled with the bactin2 promoter, floxed mCherry, followed by one of four HTT-exon1 cassettes (25Q, 97Q(m), ΔN17-25Q, or ΔN17-97Q(m)) fused to GFP flanked by tol2 transposon sites (Fig. 1a). The polyQ coding DNA sequences were mutated at wobble sites as reported previously [29] to stabilize the transgene and prevent the genomic instability associated with CAG repeats. Transgenic germlines were established for each transgene, one line was maintained for HTT-exon1 and HTT-ΔN17-exon1 control fish while two separate lines were maintained for mHTT-exon1 and mHTT-ΔN17-exon1. Transgenic lines have been maintained to at least F₆ generation with Mendelian inheritance patterns suggesting a single insertion site is present in each line. Additionally, the use of the Tol2 system should result in single copy transgene insertions. To confirm that these transgenes were inducible by Cre, we injected embryos with Cre mRNA and observed loss of mCherry expression and induction of GFP expression in all lines (Fig. 1b and data not shown). HTT-exon1 L1 and HTT-ΔN17-exon1 L1 lines exhibited ubiquitous, diffuse GFP expression at all time points examined while mHTT-exon1 and mHTT-ΔN17-exon1 lines initially had ubiquitous expression but exhibited GFP⁺ protein aggregates by 5 days post fertilization (dpf), most prominently in trunk muscles (Fig. 1c). To measure the expression level of each transgene in the different lines, western blots and qRT-PCR were performed. Each line was crossed to heat shock inducible Cre (HS:cre) transgenic fish and then heat shocked at shield stage resulting in robust, ubiquitous recombination. To evaluate protein levels, Western blots were performed on pools of embryos from three separate clutches harvested at 5 dpf for the HTT-GFP fusion protein and α-tubulin as a loading control (Fig. 1d, upper panel). Densitometry was used to measure relative protein abundance (Fig. 1d, lower panel). Embryos lacking HS:cre never show GFP expression (data not shown). HTT-exon1 L1 and HTT-ΔN17-exon1 L1 both exhibit strong GFP expression upon recombination at a size slightly larger than the predicted ~37 kDa and ~35 kDa, suggesting post-translational modification was present. Similarly, both mHTT-exon1 and mHTT-ΔN17-exon1 appear at sizes larger than their predicted ~46 kDa and ~44 kDa molecular masses, which may result from aberrant conformation due to the expanded polyQ and/or post-translational modification (Fig. 1d, upper panel). mHTT-exon1 L1 and L2 exhibit similar levels of expression that are significantly lower than line HTT-exon1 L1, HTT-ΔN17-exon1 L1, and both mHTT-ΔN17-exon1 lines (p < 0.01, ANOVA, Bonferroni posthoc test) (Fig. 1d). mHTT-ΔN17-exon1 L1 and L2 also exhibit similar levels of expression that are lower than HTT-ΔN17-exon1 L1 but with only L2 reaching statistical significance (p < 0.01, ANOVA, Bonferroni posthoc test) (Fig. 1d). We were surprised to observe the lower protein levels for the mHTT-exon1 lines as compared to the mHTT-ΔN17-exon1 lines since the transgene design and promoter were identical. To test whether transgene mRNA expression was responsible for the difference, we performed qRT-PCR analysis on mRNA extracted from each line. For this analysis, pools of ten GFP⁺ embryos from six independent clutches were assayed for each line. mRNA levels were less varied between lines than protein, with mHTT-exon1 L1 the only one to reach statistical significance in comparison to HTT-exon1 L1 (p < 0.05, ANOVA, Bonferroni posthoc test) (Fig. 1e). It is important to note that mRNA levels between mHTT-exon1 lines and mHTT-ΔN17-exon1 lines are similar. This is in contrast to the protein levels, suggesting that post-translational mechanisms may be affecting protein abundance. In fact, the N17 region of HTT has been described as a modifier of protein stability, harboring both phosphorylation [26, 27] and ubiquitination [17] sites that regulate protein degradation. Therefore, it appears that the intact N17 domain leads to greater protein instability in this zebrafish model and its deletion accelerates protein accumulation.

Each HTT transgenic line was crossed with HS:cre and the embryos heat shocked at shield stage to induce recombination or left as non-heat shocked controls. HTT-exon1 L1 (n = 21) and HTT-ΔN17-exon1 L1 (n = 26) both exhibited robust GFP expression and developed normally with no behavioral changes up to 26 weeks of age (Fig. 2a and c and Additional file 1: Movie S1 and Additional file 2: Movie S2). mHTT-exon1 L1 (n = 16) and L2 (n = 16) both survived early development even though mHTT-exon1 protein aggregates were present by 5 dpf. These fish behaved normally until 16 weeks of age, at which point mild behavioral abnormalities, such as abnormal swimming and immobility alternating with jerky movement, were noted in a small number of fish in mHTT-exon1 L1 with more fish displaying these behaviors over time (Fig. 2c). mHTT L2 has not displayed any behavioral abnormalities up to 32 weeks of age. mHTT-ΔN17-exon1 L1 (n = 10) and L2 (n = 24), in contrast, developed a robust, progressive movement deficit (Fig. 2b and c, and Additional file 3: Movie S3 and Additional file 4: Movie S4). mHTT-exon1, mHTT-ΔN17-exon1 L1, and mHTT-ΔN17-exon1 L2 all exhibit statistically significant differences in survival curve analysis as compared to each other as well as HTT-exon1 L1, HTT-ΔN17-exon1 L1, and mHTT-exon1 L2 (Fig. 2c, Kaplan Meier with Log Rank, p < 0.001). We grouped the observed behaviors into three stages according to their severity (Fig. 2d): Stage 1 – abnormal swimming, immobility alternating with jerky movement; Stage 2 – loss of lateral stability, corkscrew swimming (see Fig. 2b); Stage 3 – loss of vertical stability, inability to coordinate swimming, death. mHTT-exon1 L1 fish begin exhibiting symptoms at 16 weeks and slowly progress so that by 44 weeks 50 % of the fish have reached Stage 3 (Fig. 2e). However, 40 % of the fish are still unaffected. mHTT-ΔN17-exon1 L1 begins exhibiting Stage 1 symptoms by 5 weeks of age and progresses to Stage 3 by 10–12 weeks (Fig. 2f). mHTT-ΔN17-exon1 97Q L2 does not exhibit Stage 1 symptoms until 8 weeks and progresses more slowly with Stage 3 not reached for all fish until after 20 weeks (Additional file 5). These experiments have been repeated in multiple generations with similar results. Non-heat shocked control transgenic fish for all lines, including both mHTT-ΔN17-exon1 lines, develop normally and do not show abnormal movement before 1 year of age (the last observation made), although some leaking recombination was noted. Overall, we have found that mHTT-ΔN17-exon1 transgenic fish consistently exhibit a progressive movement deficit, while mHTT-exon1 fish show a later-onset and more slowly progressing behavioral impairment. Importantly, the control HTT-exon1 and HTT-ΔN17-exon1 fish are indistinguishable from wild-type fish. Since the phenotypes of the transgenic fish were analyzed up to 11 months of age, it is possible that the mHTT-exon1 zebrafish could develop more severe disease-like phenotypes as they age.

To examine if brain atrophy, a hallmark of neuropathology in HD, was present in the mHTT-ΔN17-exon1 fish, brain weight from Stage 3 fish was compared to non-heat shocked fish and control HTT-ΔN17-exon1 with and without heat shock. Fish from each group were age matched and raised under equal population density and tank size. This is an important consideration for zebrafish since they exhibit density dependent growth [43]. The average body length and weight of each group of fish was not significantly different from each other (n = 8 for each group) (Fig. 3, a and b). The zebrafish brains from each line (excluding the olfactory bulbs and eyes) were dissected and weighed. Interestingly, heat shocked, Stage 3 mHTT-ΔN17-exon1 fish exhibited a significantly smaller brain weight than the brains from non-heat shocked mHTT-ΔN17-exon1 fish or HTT-ΔN17-exon1 fish with or without heat shock (Fig. 3c) (ANOVA with Bonferroni post hoc, p < 0.05). This result suggests that mHTT-ΔN17-exon1 fish exhibit brain atrophy similar to that seen in HD patients.

The 17N-terminal amino acids of HTT have been reported to regulate protein stability, toxicity, and sub-cellular localization. In a mouse BACHD model with N17 deleted in the context of the full length HTT gene, dramatically accelerated and exclusive nuclear accumulation of mHTT fragments were demonstrated [28]. In the context of mHTT-exon1, an HD-relevant pathogenic fragment [32], N17 is necessary for nuclear export/cytoplasmic targeting in vitro [44, 45]: its relevance in mediating nuclear versus cytoplasmic pathology in vivo remains unclear. Therefore, we assessed if the localization of aggregates was different in transgenic HD fish expressing mHTT-exon1 with or without intact N17. First, we immunostained brain sections of Stage 3, HS:cre/mHTT-ΔN17-exon1 fish, HS:cre/HTT-ΔN17-exon1 fish, HS:cre/mHTT-exon1 fish, and HS:cre/HTT-exon1 fish with the anti-human HTT antibody S830 [46] (Fig. 4a–d). HTT-ΔN17-exon1 and HTT fish exhibit diffuse staining throughout the brain with no visible aggregates (Fig. 4a and c). mHTT-ΔN17-exon1 and mHTT-exon1 fish both exhibit robust aggregate staining throughout the brain (Fig. 4b and d). The aggregates in mHTT-exon1 appear as small puncta with occasional larger accumulations (Fig. 4b) while the aggregates in mHTT-ΔN17-exon1 appear mainly as large oval shaped staining reminiscent of nuclei (Fig. 4d).

To examine the cellular and subcellular localization of these aggregates in more detail, we combined immunofluorescent staining for neurons (anti-HuC) with GFP staining of the HTT-exon1 fusion protein and DAPI for nuclei. We focused our examination on the ventrolateral nucleus of torus semicircularis, a brain region with strong HuC staining and a mixture of cell bodies and neuropil. In HTT-exon1 and HTT-ΔN17-exon1 fish, diffuse GFP staining was present throughout the brain including neurons (Additional file 6). Close examination by confocal microscopy demonstrated that HTT-exon1 is largely excluded from the nucleus while HTT-ΔN17-exon1 is detectible in the nucleus suggesting the nuclear export function of N17 is intact in vivo (Additional file 6). mHTT-exon1 fish exhibited robust aggregate formation, with the majority localized away from HuC positive neuronal cell bodies (Fig. 4e–h). These aggregates are likely present in the neuropil (e.g. axons and dendrites), but it is also possible that a subset of them are present in non-neuronal cells. Occasional GFP⁺ aggregates are present in the peri-nuclear regions stained by HuC or weakly and diffusely in DAPI positive nuclei (Fig. 4e–h, arrowheads highlight GFP⁺ nuclei). mHTT-ΔN17-exon1 fish, on the other hand, exhibited strong and predominant nuclear accumulation of mHTT-ΔN17-exon1 that was found almost exclusively in the HuC positive neurons (Fig. 4i–l). Neuropil aggregates similar to those seen in mHTT-exon1 fish were also present, but were much less abundant. 3D confocal projections clearly demonstrated the neuropil localization of HTT aggregates in mHTT-exon1 fish (Fig. 4m) and nuclear accumulation in the mHTT-ΔN17-exon1 fish (Fig. 4n). Quantification of HTT-GFP location demonstrates significant enrichment of nuclear mHTT-∆N17-exon1 in HuC positive neurons of mHTT-ΔN17-exon1 fish (Fig. 4o, p < 0.05, ANOVA with Bonferonni post hoc). No difference was detected in cell density, percent HuC positive cells, or aggregate density (nuclear accumulation was scored as a single aggregate) in mHTT-exon1 versus mHTT-ΔN17-exon1 fish (Additional file 7). Taken together, our study is consistent with the in vitro studies in cells and mouse study with BACHD-∆N17 mice that N17 is a crucial element in preventing nuclear pathogenesis for mHTT-exon1, and our study provides the first in vivo evidence that N17 can prevent nuclear accumulation and aggregation of a known pathogenic mHTT fragment, mHTT-exon1, in a vertebrate model of HD.

HD is generally thought to be caused by neuron specific dysfunction and atrophy. However, since HTT expression is ubiquitous it is possible that disease specific pathologies are occurring in other cell types. HTT-polyQ expression in glia [47], immune cells [48], skeletal muscle [49] and vasculature [50, 51] have all been suggested to contribute to HD. A recent study using conditional BAC transgenic mouse models of HD (BACHD) revealed distinct and synergistic roles of full-length mHTT expressed in cortical pyramidal neurons and striatal MSNs in eliciting multiple behavioral deficits and selective neurodegeneration [52]. However, these prior models lack the overt and progressive movement disorder seen in our mHTT-ΔN17-exon1 fish model, and it is unclear if mHTT-exon1 lacking N17 is ubiquitously toxic or may retain pathogenic specificity similar to the intact mHTT-exon1. To begin answering such a question, we crossed mHTT-ΔN17-exon1 fish with Cre driver lines with selective Cre expression in lineages for neurons (elavl3-Cre) [53], glia (gfap-Cre) [54], skeletal muscle (mylpfa-Cre) [55], and endothelial (etv2-Cre) [56] cells. Each cross gave expected tissue and cell-type specific mHTT-exon1 transgene expression (Fig. 5a–d). GFP⁺ fish from each cross were selected and raised for behavioral observation (n = >20 for each genotype). Only the neuron specific, elavl3:cre/mHTT-ΔN17-exon1 fish developed movement abnormalities reminiscent of the ubiquitously activated mHTT-ΔN17-exon1 (Fig. 5e, p < 0.001, Kaplan Meier analysis). Using our previously described behavioral categories, we found that elavl3:cre/mHTT-ΔN17-exon1 fish exhibit all of the progressively deteriorating movement behaviors described in the ubiquitously expressing fish (Fig. 5f). It should be noted that the movement abnormalities observed in elavl3:cre/mHTT-ΔN17-exon1 fish appeared and progressed more slowly than ubiquitously expressing fish, suggesting the possibility that non-neuronal cells modify the effect of mHTT on neurons in this model.

To examine the behavior of these fish in a more quantitative manner, we performed field potential recordings [57] of freely behaving elavl3:cre/mHTT-ΔN17-exon1 (n = 6) and elavl3:cre/HTTΔN17 fish (n = 6) fish (Fig. 6). elavl3:cre/mHTT-ΔN17-exon1 fish were categorized as Stage 1 at the time of this experiment while elavl3:cre/HTT-ΔN17-exon1 fish were categorized as healthy. Field potential recordings are performed by placing individual fish into a small chamber with recording electrodes at each end. When a fish moves the electrical activity of the muscle creates a potential across the electrodes that can be recorded over time. This gives an indirect measurement of each fish’s motor activity. elavl3:cre/HTT-ΔN17-exon1 fish exhibit short bursts of activity corresponding to swimming and turning movements within the chamber (Fig. 6a, top panel). elavl3:cre/mHTT-ΔN17-exon1 fish exhibited bouts of prolonged activity (Fig. 6a bottom panel and c). However, elavl3:cre/mHTT-ΔN17-exon1 fish initiated movements less frequently than elavl3:cre/HTT-ΔN17-exon1 fish (Fig. 6b). Overall, elavl3:cre/HTT-ΔN17-exon1 and elavl3:cre/mHTT-ΔN17-exon1 fish did not differ in the total amount of time they were active during these trials (Fig. 6d). These results support the behavioral observations that Stage 1 fish exhibit bouts of jerky movement followed by inactivity.

In summary, our cell-type-specific expression of mHTT-ΔN17-exon1 revealed a remarkable specificity of toxic mHTT fragments to neurons while its toxicity to several non-neuronal cell types including glia, muscle and blood vessels are unremarkable. This result suggests the neurons are particularly sensitive to the nuclear polyQ toxicity in the context of mHTT-exon1, and the presence of N17 domain can delay such toxicity in vivo.

All zebrafish use was approved by the University of California, Los Angeles Animal Care and Use Committee. This work was performed under IACUC protocol ARC # 2001-074-41 issued by OFFICE OF ANIMAL RESEARCH OVERSIGHT, University of California Los Angeles. Adults, juveniles, and zebrafish embryos were maintained according to standard zebrafish methods [64]. Heat shock was performed on shield stage embryos for 30 min at 38.5 °C after which embryos were transferred to a standard zebrafish incubator at 28.5 °C to be raised.

All transgenes were generated using the Multisite Gateway System (Life Technologies) and the Tol2kit developed for zebrafish [42]. p3E-HTT Exon 1 25Q-EGFP, p3E-HTT Exon 1 97Q-EGFP, p3E-ΔN17 HTT Exon 1 25Q-EGFP, and p3E − ΔN17 HTT Exon 1 97Q-EGFP were generated by BP reaction of each respective PCR product flanked with B2F and B3R sequences to pDONR™ P2R-P3. Transgene plasmids were created by LR multisite reaction of pDESTTol2pA, p5E-bactin2, pME-loxPmCherryloxP, and each respective HTT p3E vector. Resulting plasmids were injected into single cell zebrafish embryos along with Tol2 mRNA using standard methods to efficiently generate germline integrations. Resulting F₀ fish were raised and their offspring screened for ubiquitous mCherry expression. Positive F₁ embryos were raised to establish each line. Each line was derived from separate F₀ fish to control for integration effects.

Cre driver lines were generated by LR multisite reaction with pDESTTol2pA, promoter specific p5E vector, pME-Cre, and p3E-pA. p5E-elav3l was generated by cloning the 3.1 kilobase zebrafish elav3l (also known as HuC) promoter, into pDONR™ P4R-P1R using the BP reaction. Similarly p5E-gfap was generated by subcloning the gfap promoter from the Tg(gfap:GFP) plasmid [54] into pDONR™ P4R-P1R. p5E-mylpfa [65] and p5E-etv2 [56] were previously reported. Each transgenic line was generated as described above. However since these lines do not have a fluorescent marker gene, founders were identified by outcross to ubi:Switch fish [66] which exhibit a GFP to mCherry fluorescent switch in tissues expressing Cre recombinase. Independent lines with strong tissue specific recombination and minimal leakiness were maintained for each promoter.

Quantitative RT-PCR was performed on cDNA generated using Superscript III Reverse Transcriptase (Life Technologies, 18080044) from total RNA extracted from 5 day post fertilization larvae using Trizol reagent (Life Technologies, 15596026) on a Stratagene Mx3005P qPCR system. Three pools of ten embryos from separate clutches of each transgenic line were analyzed. Oligo(dT)12-18 (Life Technologies, 18418–012) was used to prime the reverse transcriptase reaction. Primers were: βactin1F, TGTTTTCCCCTCCATTGTTG; βactin1R, ACATACATGGCAGGGGTGTT; EGFPqF, ACGTAAACGGCCACAAGTTC; EGFPqR, AAGTCGTGCTGCTTCATGTG.

Western blots were performed on protein extracted from 5 day post fertilization embryos using standard methods. Five embryos were pooled for each sample into 25 μl of 2X SDS Loading Buffer. Protein samples were run on a 12 % Tris-glycine acrylamide gel (Thermo Scientific, 25247). Protein was transferred onto a PDVF membrane (Thermo Scientific, 88520), and then probed with rabbit anti-GFP primary antibody 1:2000 dilution (Life Technologies, A11122), stripped and then reprobed with mouse anti-tubulin antibody 1:10,000 dilution (Sigma, T5168). Secondary antibodies were anti-rabbit-peroxidase (Sigma, A0545) and anti-mouse-peroxidase (Sigma, A9044) both at 1:5000 dilutions. Chemiluminescent detection was performed using Lumi-Light western blotting reagent (Roche, 12015200001) and HyBlot CL autoradiography film (Denville Scientific, E3018). Film was digitized on an HP Scanjet G3110 and the image adjusted in Adobe Photoshop CS4. ImageJ was used to perform densitometric measurements.

Fish were raised at similar densities and monitored weekly for abnormal movement. Fish were grouped into four categories: 1) Healthy – normal swimming; 2) Stage 1 – abnormal swimming, immobility alternating with jerky movements; 3) Stage 2 – loss of lateral stability, corkscrew swimming; 4) Stage 3 – loss of vertical stability, inability to coordinate movement, death. When fish were no longer able to swim for feeding they were killed for further analysis.

For more detailed behavioral analysis we performed electric field potential recordings on freely behaving adult elav3l:cre/HTT-ΔN17 or, Stage 1 categorized, elav3l:cre/mHTT-ΔN17 fish (n = 6 each) according to the methods of Issa et.al. [57]. Fish were placed into a recording chamber and allowed to acclimate to the chamber for 5 min. Then field potentials generated from muscles during movement were recorded for 5 min for each fish. The recording chamber was 11 cm long × 4 cm wide × 3 cm deep containing double distilled water with a resistance of ~ 15 MΩ. Water depth was 2.75 cm. Room temperature was set to 25 °C. Electric field potentials were recorded using one pair of electrodes (0.5 cm exposure, 0.1 cm thick of bare copper medal) placed on either end of each testing chamber. Electric signals were amplified 1000-fold using an AC differential amplifier (AM-Systems model 1700, Carlsborg, WA USA), low-pass filtered at 300 Hz and high-pass filtered at 500 Hz. All signals were digitized and stored with a Digidata-1322A and acquired using Axoscope software (Molecular Devices, Inc., Sunnyvale, CA, USA). Data was annotated manually and movement frequency, duration, and total time were calculated.

To harvest tissue for cryo-sectioning, adult fish were deeply anesthetized in Tricaine solution (Sigma, A5040) and then killed by exsanguination. The head was removed as was the top of the skull. Fish heads were fixed overnight in 4 % paraformaldahyde/PBS solution at 4 °C with constant movement on a nutator mixer. The following day the brain was dissected out intact from the tip of the forebrain to the hindbrain spinal cord junction. The brain was then fixed overnight in 4 % paraformaldahyde/PBS solution plus 5 % sucrose at 4 °C with constant movement on a nutator mixer. The following day fixed tissue was put through an increasing sucrose gradient: 5, 10, 12.5, 15, and 20 % in PB buffer, 1 h each and then placed at 4 °C with constant movement on a nutator mixer overnight. The following day tissue was infiltrated with a mix of 2:1 20 % sucrose solution to OCT Compound (Tissue-Tek, 4583) for 1 h and then snap frozen in 100 % OCT in aluminum foil molds placed into a 2-methylbutane (Sigma, M32631), dry ice bath. Blocks were stored at −80 °C until they were cut at 16 μm for immunostaining on a Leica model CM3050S cryostat.

Immunohistochemistry was performed on cryosections using the polyclonal sheep anti-human HTT primary antibody S830 [46] at 1:1000 dilution. Biotinylated anti-sheep secondary (Vector Laboratories, BA-6000) was used at 1:200 dilution and the Vectastain ABC Elite kit (Vector Laboratories, PK-6100) plus Vector SG substrate (Vector Laboratories, SK-4700) used to perform the color reaction. Sections were then counter stained with Nuclear Red (Vector Laboratories, H-3403) and mounted with Permount (Fisher, SP15-100).

Immunofluorescence was performed on sections using manufacturer’s recommended protocol and dilution for each antibody. Primary antibodies were rabbit anti-GFP (Life Technologies, A11122), 1/200 dilution, and anti-HuC/HuD (Life Technologies, A21271), 5 μg/mL. Secondary antibodies were goat anti-rabbit Alexa488 IgG (Life Technologies, A11008), 1/200 dilution, and donkey anti-mouse Rhodamine (Life Technologies, A16016), 1/200 dilution. Stained sections were mounted in Prolong Gold Antifade with DAPI (Life Technologies, P36935).

Images were captured on an Axioskop 2 plus microscope (Zeiss) using 5×, 10×, or 40× objectives with a Hamamatsu ORCA-ER camera and Volocity 5 software (Improvision). Confocal images were collected on a Zeiss LSM 510 with a 63× oil immersion objective (NA 1.4). Adobe Photoshop CS4 was used to adjust brightness and contrast and assemble composite images.

To quantify GFP⁺ protein aggregates in HuC stained sections, ImageJ was used to count cell number (DAPI positive nuclei), HuC⁺ cells, and aggregate number. Maximum intensity projections of 133.6 μm × 133.6 μm × 4.7 μm z-stacks were analyzed from three non-consecutive sections of each fish. Images were acquired such that bright GFP+ aggregates did not over saturate the image. Therefore, the background level of GFP expression appears lower than it actually was. Three fish were examined for each genotype.

Student’s t-test was used for single comparisons, ANOVA with Bonferonni posthoc test was used for multiple comparisons, and Kaplan Meier survival analysis with Log Rank posthoc test was used for survival curve comparisons with p < 0.05 set as significant for each. All statistical analysis was performed using SPSS 14.0 software.