Title Viable Neuronopathic Gaucher Disease Model in Medaka ( Oryzias latipes ) Displays Axonal Accumulation of Alpha-Synuclein

Homozygous mutations in the glucocerebrosidase (GBA) gene result in Gaucher disease (GD), the most common lysosomal storage disease. Recent genetic studies have revealed thatGBAmutations confer a strong risk for sporadic Parkinson’s disease (PD). To investigate howGBAmutations cause PD, we generated GBA nonsense mutant (GBA-/-) medaka that are completely deficient in glucocerebrosidase (GCase) activity. In contrast to the perinatal death in humans and mice lacking GCase activity, GBA-/medaka survived for months, enabling analysis of the pathological progression.GBA-/medaka displayed the pathological phenotypes resembling human neuronopathic GD including infiltration of Gaucher cell-like cells into the brains, progressive neuronal loss, and microgliosis. Detailed pathological findings represented lysosomal abnormalities in neurons and alpha-synuclein (α-syn) accumulation in axonal swellings containing autophagosomes. Unexpectedly, disruption of α-syn did not improve the life span, formation of axonal swellings, neuronal loss, or neuroinflammation inGBA-/medaka. Taken together, the present study revealed GBA-/medaka as a novel neuronopathic GD model, the pahological mechanisms of α-syn accumulation caused by GCase deficiency, and the minimal contribution of α-syn to the pathogenesis of neuronopathic GD. PLOS Genetics | DOI:10.1371/journal.pgen.1005065 April 2, 2015 1 / 22 OPEN ACCESS Citation: Uemura N, Koike M, Ansai S, Kinoshita M, Ishikawa-Fujiwara T, Matsui H, et al. (2015) Viable Neuronopathic Gaucher Disease Model in Medaka (Oryzias latipes) Displays Axonal Accumulation of Alpha-Synuclein. PLoS Genet 11(4): e1005065. doi:10.1371/journal.pgen.1005065 Editor: Bingwei Lu, Stanford University School of Medicine, UNITED STATES Received: November 10, 2014 Accepted: February 9, 2015 Published: April 2, 2015 Copyright: © 2015 Uemura et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: The cDNA sequences we determined in this study can be found in the European Nucleotide Archive (ENA http://www.ebi.ac. uk/ena, accession numbers LM644999–645003). Funding: This work was supported by Grant-in-Aid for Specially Promoted Research and Scientific Research on Innovative Areas “Brain Environment” from the Ministry of Education, Culture, Sports, Science and Technology of Japan (Grant Number 23111001, http://www.jsps.go.jp/english/index.html). The funder had no role in study design, data


Introduction
Gaucher disease (GD) is the most common lysosomal storage disease and is caused by homozygous mutations in glucocerebrosidase (GBA). Mutations in GBA lead to decreased enzymatic activity of glucocerebrosidase (GCase) and result in the accumulation of its substrates, glucocerebroside and glucosylsphingosine [1,2]. GD is classically divided into three subtypes: a nonneuronopathic form (type 1), an acute neuronopathic form (type 2), and a subacute neuronopathic form (type 3). Visceral manifestations of all forms are characterized by hepatosplenomegaly, cytopenia, and skeletal disease. Pathologically, the accumulation of lipid-laden macrophages, called Gaucher cells, are observed in the affected organs. Neurological manifestations of neuronopathic forms include brainstem dysfunction, intellectual disability, seizures, and myoclonic movement. Pathological features of neuronopathic forms are neuronal loss, astrogliosis, microgliosis, and perivascular accumulation of Gaucher cells [3]. The most severe neuronopathic form, called the perinatal lethal type, has also been reported [4]. Common presentations of patients with the perinatal lethal type are hydrops fetalis and congenital ichthyosis. Almost no residual GCase enzymatic activity is found in these cases. Because currently available therapies are ineffective for neurological manifestations, a strong demand exists for elucidation of the pathological mechanisms and the development of novel therapies.
Parkinson's disease (PD) is the most common neurodegenerative movement disorder. GBA has recently drawn considerable attention because heterozygous mutations in this gene confer a high risk for sporadic PD [5,6]. In addition, patients with type 1 GD also have an increased life-time risk of developing PD [7]. PD patients carrying GBA mutations show intraneuronal accumulation of alpha-synuclein (α-syn) called Lewy bodies and Lewy neurites, which are the pathological hallmarks of sporadic PD [3]. Several cellular, animal, and postmortem studies have indicated an association between GBA mutations and α-syn accumulation. For example, deficiency in GCase enzymatic activity causes lysosomal dysfunction and α-syn accumulation [8,9,10,11,12]. Increased α-syn in turn creates a vicious cycle by inhibiting the trafficking of GCase to lysosomes, thus leading to decreased GCase activity in lysosomes [9]. Consistent with this notion, mouse models overexpressing α-syn and postmortem tissue from patients with PD show reduced GCase activity in the brains [13,14,15]. Although several hypotheses have been proposed, further mechanisms of how GBA mutations contribute to the development of PD remain elusive.
Medaka (Oryzias latipes) are a versatile vertebrate animal model for disease research. These fish are easy to handle, have a relatively short generation time (2-3 months), produce a large number of progeny per generation, and have several inbred strains [16]. Importantly, medaka have an advantage as an animal model of PD due to endogenous α-syn in contrast to invertebrate models that lack α-syn. Moreover, several genetic manipulations can be performed in medaka in addition to established transgenic techniques [17,18,19,20,21]. So far, we have reported genetic PD models of medaka that develop locomotor dysfunction accompanied by the selective loss of dopaminergic and noradrenergic neurons [22,23]. Considering these lines of evidence, medaka have the potential to be a new animal model of PD.
Here, we generated GBA nonsense mutant medaka and found that homozygous GBA nonsense mutant (GBA -/-) medaka are a viable neuronopathic GD model. GBA -/medaka developed remarkable α-syn accumulation in the brains and thus provide novel insights into the association of GBA mutations with α-syn accumulation. Furthermore, we revealed minimal contribution of endogenous α-syn to the pathogenesis of neuronopathic GD in medaka.

Generation of GBA nonsense mutant medaka
We generated GBA nonsense mutant medaka to investigate the mechanisms by which GBA mutation leads to PD. To identify medaka GBA orthologs, we searched the medaka genome database (http://www.ensembl.org/Oryzias_latipes/Info/Index) with the basic local alignment search tool and found only one ortholog of human GBA. We cloned the medaka GBA with reverse transcription-polymerase chain reaction (RT-PCR) and rapid amplification of cDNA ends and found that this gene has 11 exons encoding a protein of 522 amino acids. The amino acid sequence of medaka GBA showed 53% homology to that of human GBA (S1 Fig). Next, we screened a targetinginduced local lesions in genome (TILLING) library for medaka GBA using a high-resolution melting assay [17,24]. We identified a nonsense mutant (W337X) and generated the nonsense mutant medaka by in vitro fertilization (Fig. 1A). We examined GCase activity in the brains of GBA mutants after crossing with heterozygous mutants. GBA W337X/W337X (GBA -/-) medaka showed complete deficiency in GCase activity, and GBA WT/W337X (GBA +/-) medaka showed a decrease in GCase activity of about 50% compared to wild-type (GBA +/+ ) medaka (Fig. 1B). Although humans and mice lacking GCase activity die soon after birth [4,25,26], GBA -/medaka survived for more than 3 months, enabling us to analyze the pathological progression (Fig. 1D). GBA -/medaka showed abnormal rotating swimming movement at 2 months (S1, S2 Movie) and the abnormal appearance of a bent spine at 3 months ( Fig. 1C). High levels of glucocerebroside accumulated in the brains of GBA -/medaka (Fig. 1E), whereas the amount of galactocerebroside, an isomer of glucocerebroside, was not changed. Glucocerebroside with C18 fatty acids was the most dominant type in the brains of GBA -/medaka (S2 Fig), an observation that is consistent with the neuronopathic GD mouse model [27].

GBA -/medaka showed neuronopathic GD-like pathology
We performed pathological analyses of GBA -/medaka. Patients with GD show Periodic acid-Schiff-positive Gaucher cells in affected visceral organs such as the liver, spleen, and bone marrow, whereas GBA -/medaka showed abnormal Periodic acid-Schiff-positive cells in the spleen and kidney, but not in the liver, at 3 months (S3 Fig). Next, we examined the brains of GBA -/medaka and found abnormal cells with large vacuoles mainly in the periventricular gray zone of the optic tectum ( Fig. 2A, B). Transmission electron microscopy revealed that these cells were macrophage-like and had large vacuoles containing filamentous structures (Fig. 2C). Similar filamentous structures are observed in Gaucher cells of patients with GD and mouse models of GD [26,28,29]. The staining intensity with Luxol fast blue was decreased, and singlestranded DNA (ssDNA)-positive cells were observed in GBA -/medaka (Fig. 2D), indicating myelin loss and cell death, respectively. In situ hybridization for apolipoprotein E (APOE), a microglial marker in teleost fish [30], revealed proliferating activated microglia in GBA -/medaka (Fig. 2D). Teleost fish have glial fibrillary acidic protein (GFAP)-expressing radial glial cells (or ependymoglial cells) instead of astrocytes as in mammals [31]. Humans and mice with neuronopathic GD show astrogliosis in their brains [3,32], whereas neither proliferation of GFAP-positive radial glial cells nor elevated levels of GFAP were observed in GBA -/medaka (Fig. 2D, E). To investigate the type of neuronal cells that die in GBA -/medaka, we counted the numbers of tyrosine hydroxylase (TH)-positive dopaminergic neurons in the middle diencephalon, which corresponds to the human substantia nigra [33], TH-positive noradrenergic neurons in the locus coeruleus, and tryptophan hydroxylase (TPH)-positive serotonergic neurons in the superior raphe at 2 and 3 months. GBA -/medaka showed progressive cell loss of all these neurons (Fig. 2F). Consistent with these findings, the amount of TH was decreased in GBA -/medaka (Fig. 2G). Collectively, GBA -/medaka exhibited neuronopathic GD-like pathology including progressive and non-selective neuronal loss.
GBA -/medaka developed α-syn accumulation in axonal swellings containing autophagosomes Medaka express α-syn, which is a protein consisting of 127 amino acids. To investigate α-syn pathology in GBA -/medaka, we created a medaka α-syn antibody against the epitope of amino acids 90 to 104 (S4A Fig). Medaka also express β-synuclein, γ-synuclein-a, and γ-synuclein-b in addition to α-syn, but these other synucleins do not have amino acid sequences homologous to the epitope of the medaka α-syn antibody (S4B Fig). Next, we generated α-syn deletion  Wild-type, heterozygous, and homozygous α-syn deletion mutant (α-syn +/+ , α-syn +/-, and α-syn -/-, respectively) medaka could be distinguished with PCR analysis of α-syn (S4D Fig). RT-PCR analysis of α-syn mRNA revealed that α-syn -/medaka did not express intact α-syn mRNA (S4E Fig). Western blot analysis with the medaka α-syn antibody revealed a 14-kDa band, which was specifically found in α-syn +/+ medaka (S4F Fig). The authenticity of the antibody was confirmed by the lack of immunostaining with the medaka α-syn antibody in the brains of α-syn -/medaka (S4G Fig). Consistent with the findings in mammals, medaka α-syn was mainly found presynaptically with immunoelectron microscopy (S4H Fig). Then, we performed immunohistochemical analysis and found abundant α-syn accumulation in the brains of GBA -/medaka at 3 months (Fig. 3A). α-syn accumulation was also observed at 2 months, but not at 1 month (S5 Fig). Abnormal structures observed with toluidine blue staining were similar to α-syn accumulations in size and distribution (Fig. 3A). Transmission electron microscopy revealed numerous axonal swellings containing vacuoles and other various materials in the brains of GBA -/medaka (Figs. 3B, S6A). A considerable number of these vacuoles had double membranes and sometimes contained mitochondria, suggesting that these vacuoles were autophagosomes. Consistent with this, LC3, an autophagosomal marker, accumulated in the brains of GBA -/medaka ( Fig. 3A). Moreover, immunogold-labeled α-syn was detected in axonal swellings with immunoelectron microscopy (Fig. 3C). Conventional and confocal double immunofluorescence microscopy revealed that accumulated α-syn colocalized with LC3 accumulations (Fig. 3D), and a considerable portion of the α-syn signals in axonal swellings colocalized with LC3 signals (Figs. 3E, S6B). In addition to these findings, ubiquitin also accumulated in α-syn-positive axonal swellings (Figs. 3E, S6B). To examine the α-syn expression level and solubility, we performed sequential biochemical fractionation assays. The total amount of α-syn in the Triton-soluble fraction was not increased in GBA -/medaka. However, when adjusted for the amount of neuron-specific enolase (NSE) and considering the robust neuronal loss in GBA -/medaka, α-syn was significantly increased in GBA -/medaka (Fig. 3F). The decreased amount of neurofilament also reflected the neuronal loss in GBA -/medaka (Fig. 3F). No apparent α-syn band was detected in the Triton-insoluble, SDS-soluble fraction in all genotypes. In agreement with the accumulation of autophagosomes in axonal swellings, the ratio of LC3-II to LC3-I was increased in GBA -/medaka (Fig. 3F).
Meanwhile, we also investigated the phenotypes of GBA +/medaka because heterozygous mutations in GBA are a strong risk for PD [5,6]. However, GBA +/medaka even at 12 months did not show any apparent abnormal phenotypes including α-syn pathology, the numbers of TH-positive neurons, swimming movement, or the amounts of several neurotransmitters (S7A- S7E Fig).

Impairment of the autophagy-lysosome pathway in neurons of GBA -/medaka
Previous studies of GD mouse models reported that p62/SQSTM1 (an autophagic substrate) accumulates in neurons and astrocytes [34], and the number of Cathepsin D-positive puncta is decreased in neurons [35]. These observations prompted us to examine the autophagy-lysosome pathway in GBA -/medaka. Immunohistochemical analysis revealed that ubiquitin-and p62positive aggregates were observed mainly in the perikarya of neurons ( LEL is an excellent teleost microglial marker [36]. Western blot analysis showed that the amounts of ubiquitin and p62 were increased in the brains of GBA -/medaka (Fig. 4C). In contrast, these aggregates did not The phenotypes of GBA -/medaka were rescued by transgenic expression of GBA GBA nonsense mutant medaka have random point mutations in the genome at loci other than GBA. Therefore, we performed a rescue experiment to determine whether the abnormal phenotypes observed in GBA -/medaka were really caused by GBA mutation. We created medaka GBA-expressing vectors driven by a medaka growth-associated protein 43 (GAP-43) promoter (S9A Fig) [37]. GAP-43 mRNA is expressed mainly in nervous system in medaka [37]. We es- Disruption of α-syn in GBA -/medaka did not improve the phenotypes Many studies have reported that accumulated α-syn caused by α-syn overexpression results in neurotoxicity [38]. Neuroinflammation is observed in the brains of patients with PD, and much evidence shows that neuroinflammation promotes disease progression [39]. Moreover, a recent study reported that oligomeric α-syn released from neurons activates inflammatory responses in microglia [40].

Discussion
To date, several genetic animal models of neuronopathic GD including mouse, dog, and sheep have been reported [42]. Gba null mice die within 24 hours of birth due to permeability barrier ***p < 0.001). (D) Conforcal microscope images of GBA -/medaka. Upper panels: ubiquitin (green) and p62 (red). p62-positive aggregates colocalized with ubiquitin-positive aggregates (white arrowheads). Middle panels: LC3 (green) and p62 (red). p62-positive aggregates (yellow arrows) did not colocalize with LC3 accumulations (yellow arrowheads). Lower panels: Cathepsin D (green) and p62 (red). p62-positive aggregates did not colocalize with Cathepsin Dpositive organelles. Nuclei were visualized with DAPI (blue). Scale bars, 10 μm. For all analyses, data are the mean ± SEM. doi:10.1371/journal.pgen.1005065.g004 α-synuclein Accumulation in Gaucher Disease Medaka defects in the skin [26,43]. Thus, conditional knockout K14-lnl/lnl mice with normal GCase activity in their skin were generated [27]. However, use of these mice is limited because they survive only 2 weeks after birth. Mice harboring a Gba missense mutation combined with prosaposin-or saposin C-deficient mice, another neuronopathic GD model, also show neuronopathic abnormalities [29,34] and have the advantage of longer survival than K14-lnl/lnl mice. However, the relevance of these mice to neuronopathic GD is controversial. The present study revealed that GBA -/medaka survive long enough for pathological analysis of disease progression. As an example, GBA -/medaka showed α-syn accumulation at 2 months, not at 1 month after fertilization. Considering these observations, GBA -/medaka are useful as a viable neuronopathic GD model with endogenous α-syn accumulation.
GBA -/medaka showed several phenotypes different from those of mammalian neuronopathic GD model. Firstly, the skin of GBA -/medaka looks intact whereas severe skin lesion is observed in patients with perinatal lethal type GD and Gba null mice [4,43]. Secondly, GBA -/medaka exhibited PAS-positive abnormal cells in spleen and kidney which presumably correspond to human Gaucher cells. In patients with GD, Gaucher cells are not found in kidney, but in liver, spleen and bone marrow. These differences could be explained by the fact that in adult teleost fish kidney has hematopoietic function instead of mammalian bone marrow [44]. Lastly, the proliferation of GFAP-positive radial glial cells was not observed in GBA -/medaka whereas astrogliosis is observed in humans and mice with neuronopathic GD. It was reported that the reaction of GFAP-positive radial glial cells to inflammation is different from that of mammalian astrocytes [31].
Lysosomes play a fundamental role in the autophagic pathway by fusing with autophagosomes and digesting their contents [45]. In general, lysosomal dysfunction results in defective digestion and accumulation of autophagic substrates such as polyubiquitinated proteins, p62, and dysfunctional mitochondria, accompanied by accumulation of autophagosomes. Axonal swellings containing autophagosome-like structures are observed in mouse models of lysosomal storage diseases such as neuronopathic GD mice, Niemann-pick type C1 mice, Cathepsin Ddeficient mice, and Cathepsin B/L double-deficient mice [34,46,47]. A previous study reported that autophagosomes are formed in the distal part of the axon and undergo retrograde transport toward the cell body [48]. Another previous study demonstrated that inhibition of lysosomal function in primary neurons from mouse embryos disrupts the axonal transport of autophagosomes and causes their accumulation in axons [49]. Considering these lines of evidence, we propose that GCase deficiency primarily causes lysosomal dysfunction, leading to disrupted retrograde transport of autophagosomes in axons and formation of axonal swellings with α-syn accumulation (Fig. 6). Contrary to our expectations, p62-positive aggregates did not colocalize with LC3-and α-syn-positive axonal swellings, but were mainly located in the perikarya of neurons. These results suggest the possibility that presynaptic α-syn is transported proximally and degraded in a p62-independent autophagy-lysosome pathway. A previous study supports this hypothesis. Conditional knockout mice lacking ATG7 in TH-positive neurons show α-syn aggregates in swollen axons in the striatum, but not in cell bodies in the midbrain, indicating that autophagic dysfunction initially causes α-syn aggregates in the distal part of axons [50]. In addition, another previous study demonstrated that α-syn aggregation occurs earlier in axons than in neuronal cell bodies in the cardiac sympathetic nervous system in PD patients [51]. Because axonal retrograde transport may be involved in the degradative pathway of α-syn and may be a therapeutic target in PD, further studies are required to elucidate the precise mechanisms.
Because GBA +/medaka as old as 12 months did not show any apparent abnormal phenotypes, we could not directly investigate how heterozygous GBA mutations cause PD. Meanwhile, according to a recent report, induced pluripotent stem cell-derived neurons from PD patients carrying heterozygous GBA mutations show α-syn accumulation, an impaired autophagy-lysosome pathway, and dysregulation of calcium homeostasis [12]. The reason for differences in phenotypes between in vivo and in vitro models is unclear. However, our findings from GBA +/medaka seem to be reasonable because the penetrance of PD in GBA mutation carriers is estimated to be at most 30% by the age of 80 years [52]. Thus, second hits such as environmental factors and other genetic factors are probably required for the development of PD pathology in vivo.
Cellular and animal PD models overexpressing α-syn have provided evidence for the various potential toxic mechanisms of α-syn. However, few studies have demonstrated the pathological role of endogenous α-syn in vivo, which may reflect the authentic role for α-syn in PD. A previous study showed that α-syn null mice are resistant to 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP) toxicity compared with wild-type mice [53]. However, it is unclear whether this improvement is due to attenuation of α-syn toxicity or α-syn-mediated changes in the presynaptic machinery. Another previous study showed that the formation of intraneuronal inclusions and neurodegeneration in 26S proteasome-depleted mice is independent of αsyn [54]. In the present study, disruption of α-syn did not improve the life span, neuronal loss, or neuroinflammation in GBA -/medaka. Moreover, α-syn was not involved in the accumulation of autophagosomes in axons. Our data indicate that α-syn accumulation is a downstream event, and other severe pathological factors may obscure the involvement of α-syn in the pathogenesis of neuronopathic GD in medaka.
In summary, the present study showed that GBA -/medaka are useful as a viable neuronopathic GD model with endogenous α-syn accumulation. Long-term survival of these fish allows us to observe the pathological progression. Our data revealed that GCase deficiency causes lysosomal dysfunction in neurons and α-syn accumulation in axonal swellings containing Neurons of GBA -/medaka showed lysosomal dysfunction, which is reflected in morphologically abnormal structures with decreased Cathepsin D staining intensity, and axonal swellings containing autophagosomes where α-syn and ubiquitin accumulate. p62-positive aggregates, which colocalize with ubiquitin, are not located in axonal swellings, but presumably in soma or dendrites. Considering the previous studies about the axonal transport of autophagosomes, we propose that GCase deficiency primarily causes lysosomal dysfunction, which leads to disrupted retrograde transport of autophagosomes in axons and formation of axonal swellings with α-syn accumulation.
doi:10.1371/journal.pgen.1005065.g006 α-synuclein Accumulation in Gaucher Disease Medaka autophagosomes. Axonal transport of α-syn may play an important role in the mechanisms of GBA mutations leading to PD and may also be a therapeutic target in PD. Furthermore, we demonstrate the minimal contribution of α-syn to the pathogenesis of neuronopathic GD in medaka. GBA -/medaka represent a valuable model for exploring the pathological mechanisms and also provide a new platform for developing treatments in PD with GBA mutations as well as neuronopathic GD.

Ethics statement
Medaka were anesthetized in 0.02% tricaine in fish water and then sacrificed. All experimental procedures used in this study followed national guidelines. The Animal Research Committee of Kyoto University granted a formal waiver of ethical approval and also granted permission.

Maintenance of medaka
Medaka of the Kyoto-cab strain, a substrain of Cab, were maintained at 27°C in a recirculating aquaculture system. Adult fish were kept in a reproduction regimen (14 hr light/10 hr dark). Eggs were kept in a dark box at 28°C.

Cloning of medaka genes
RNA was extracted from wild-type medaka brains with Qiazol (QIAGEN) according to the manufacturer's instructions. cDNA was synthesized using the PrimeScript RT reagent kit (Perfect Real Time) (TaKaRa, #RR037A). To identify medaka GBA, SNCA, SNCGb, p62/SQSTM1, and MAP1LC3B orthologs, we referred to the medaka genome database (http://www.ensembl. org/Oryzias_latipes/Info/Index). Because their cDNA sequences and amino acid sequences were not completely known, we determined their cDNA sequences using a combination of RT-PCR and rapid amplification of cDNA ends, the products of which were generated using Seegene's Capfishing kit (Seegene). The cDNA sequences can be found in the European Nucleotide Archive (ENA, accession numbers LM644999-645003).

Generation of GBA nonsense mutant medaka
GBA mutant medaka were generated as described [17]. To find GBA mutations in the TILLING library, we narrowed down mutated candidates from 5,771 samples using a high-resolution melting assay, followed by determination of the DNA sequences [24]. We screened the TILL-ING library for exons 1-2, exons 5-7, exon 8, and exons 9-11 of GBA. In vitro fertilization was carried out using sperm from a sample with the favorable mutation. To genotype the progeny of GBA WT/W337X mutants, PCR was performed with the following primer set (5 0 -AGGGTTG AAGGGGTTAAGCA-3 0 , 5 0 -TTGTAACCAGTACCGCAGCA-3 0 ), designed HybProbes (5 0 -LC Red 640-CATGTACCAGTGGACG-Phosphate-3 0 , 5 0 -CCTAAGCTTATATCTGCAGG GACTAAACTGT-Fluorescein-3 0 ), and LightCycler 480 Probes Master in LightCycler 480 (Roche) according to the manufacturer's instructions. The genotypes can be distinguished with high-resolution melting curve analysis. GBA WT/W337X mutants were back-crossed to Kyoto-Cab medaka at least seven times and then crossed to obtain GBA W337X/W337X mutants.

Rescue experiment
For the rescue experiment, we established GBA transgenic medaka in which medaka GBA expression was driven by the medaka GAP-43 promoter (Tg(GAP-43:GBA)). We used an insulator located in the upstream region of sea urchin (Hemicentrotus pulcherrimus) arylsulfatase [55,56]. The transgenic construct was flanked by two insulators and included the medaka GAP-43 promoter followed by an internal ribosome entry site (IRES), enhanced GFP (EGFP), and the Simian virus 40 polyadenylation site or the medaka β-actin 3 0 -untranslated region (3 0 UTR) (S8A Fig). The medaka GAP-43 promoter contained a 1.0-kb fragment of the 5 0flanking region of the gene. A DNA fragment of the transgene was inserted into EcoR I/Sal I restriction enzyme sites of the pDs-actb2k-EGFP plasmid. pDs-actb2k-EGFP was constructed by inserting an Xho I/Spe I fragment from pactb2k-EGFP into the Xho I/Spe I site of the pDs-GTDEL4 plasmid, which contains 5 0 -and 3 0 -Ds sequences [21,57]. The resultant vector was injected into the cytoplasm of fertilized Kyoto-Cab eggs before the first cleavage as described [20].

Behavioral analyses
The spontaneous swimming movement of medaka was traced using a video camera positioned above the water tank and analyzed with ethovision XT 5 (Noldus). The water tank was a transparent circular container (20 cm diameter, 2 cm water depth, room temperature). Image acquisition began 5 min after medaka were placed in a new water tank. Data were collected continuously for the subsequent 3 min.

Lipid analyses with supercritical fluid chromatography (SFC)/mass spectrometry (MS)/MS
Medaka brains were stored at-80°C until analyses and homogenized in 1 ml tissue homogenization buffer (250 mM sucrose, 25 mM KCl, 50 mM Tris-HCl, 0.5 mM EDTA, pH 7.4). Aliquots containing 1 mg protein were used for the analyses. Levels of glucocerebroside and galactocerebroside were measured with high-performance liquid chromatography-tandem MS as described with modification using SFC separation [59].

Generation of the medaka α-syn antibody
A synthetic peptide corresponding to amino acids 90-104 of medaka α-syn conjugated to Keyhole Limpet Hemocyanin was used for immunization of rabbits. Serum was obtained 49 days after immunization and purified with affinity chromatography.

Immunohistochemical analyses
Frozen and paraffin sections were used for immunohistochemical analyses. For frozen sections, medaka brains were fixed with 4% (w/v) paraformaldehyde (PFA) in phosphate-buffered saline (PBS) at 4°C for 4 hr and then placed in 30% (w/v) sucrose at 4°C for more than 16 hr. Samples were embedded in Surgipath The sections were incubated at 4°C with primary antibodies or lectin for 1 to 3 days and then processed for visualization. As secondary antibodies, ImmPRESS (VECTOR) was used for diaminobenzidine staining, and Alexa Fluor-conjugated antibodies (Molecular Probes) were used for immunofluorescence. Sections were observed with an Olympus CX41 microscope, a BZ-9000 fluorescence microscope (KEYENCE), and an Olympus FV-1000 confocal laser scanning microscope. For confocal microscope images, Pearson's coefficient correlation (r) was calculated using Olympus software.

In situ hybridization
The vector including a portion of the medaka APOE cDNA sequence was generously provided by Dr. H. Mitani (Tokyo University, Tokyo, Japan) (in submission). In situ hybridization was performed as described [60]. Counterstaining was performed with methyl green (Sigma-Aldrich, #M8884).

High performance liquid chromatography
To measure the amounts of dopamine, noradrenaline and serotonin in medaka brains, high performance liquid chromatography was performed as described [60].

Transmission electron microscopy and toluidine blue staining
Medaka brains were fixed with 4% (w/v) PFA and 2% (v/v) glutaraldehyde (Wako, #072-01961) in 0.1 M phosphate buffer (PB) at 4°C for 16 hr. After rinsing in 0.1 M PB, samples were postfixed with 1% (w/v) OsO 4 in 0.1 M PB for 2 hr. Then, samples were dehydrated, penetrated with ethanol and a propylene oxide series, and embedded in Epon. Sections were obtained with an EM UC6 ultramicrotome (Leica). Sections with a thickness of 1 μm were used for toluidine blue staining. Sections with a thickness of 60 to 80 nm were stained with uranyl acetate and lead citrate and observed with a Hitachi H-7650 transmission electron microscope.

Immunoelectron microscopy
Immunoelectron microscopy using ultrathin cryosections was performed as described [62]. Briefly, brains were quickly removed from medaka and immersed in 4% PFA buffered with 0.1 M PB (pH 7.2) at 4°C for 2 hr, washed thoroughly with 7.5% sucrose in 0.1 M PB (pH 7.2), and embedded in 12% gelatin. The samples were rotated in 2.3 M sucrose in 0.1 M PB overnight at 4°C, placed on a specimen holder (Leica Microsystems), and quickly plunged into liquid nitrogen. Ultrathin cryosections were cut with a Leica UC6/FC6 and UC7/FC7 (Leica Microsystems) at about-120°C. Sections about 60 nm thick were picked up with a 1:1 mixture of 2% methylcellulose and 2.3 M sucrose and transferred to a nickel grid with a carbon-coated Formvar supporting film. The sections were rinsed with PBS containing 0.02 M glycine, treated with 1% bovine serum albumin in PBS, and incubated overnight at 4°C with rabbit anti-medaka α-syn antibody (1:30). They were then incubated with goat anti-rabbit IgG conjugated to 10 nm colloidal gold particles (1:40) (British Biocell International) for 1 hr at room temperature. Immunostained sections were fixed with 1% glutaraldehyde in PBS. After completion of the labeling, the sections were embedded in a thin layer of 2% methylcellulose with 0.4% uranyl acetate (pH 4.0), air-dried, and observed with a Hitachi H-7100 electron microscope. For control experiments, ultrathin sections were reacted only with the gold particle-conjugated secondary antibody.

Statistical analyses
A two-tailed paired Student's t-test or one-way ANOVA with Tukey's post-hoc test was used for analyses. Statistical calculations were performed with Microsoft Excel or GraphPad Prism Software, Version 5.0. of PCR for α-syn. PCR primers were designed to span the deleted region of α-syn, allowing the genotypes to be distinguished with PAGE. (E) RT-PCR for α-syn mRNA. One primer was designed to overlap the deleted region of α-syn. Intact α-syn mRNA was not detected in α-syn -/medaka. (F) Western blot analysis of α-syn and β-actin. A 14-kDa putative α-syn band was observed only in α-syn +/+ medaka, suggesting the authenticity of the medaka α-syn antibody. (G) Immunohistochemistry with medaka α-syn antibody. α-syn immunostaining was observed only in α-syn +/+ medaka. Scale bars, 50 μm. (H) Immunoelectron micrograph of a presynaptic region with immunogold-labeled α-syn. Left panels: presynaptic area with small synaptic vesicles. Right panels: presynaptic area with large synaptic vesicles. The postsynaptic density is visible (arrowheads). Scale bars, 200 nm. Left and middle panels: Neuropil of a GBA +/+ and a GBA -/medaka, respectively. Swellings of both myelinated (arrowheads) and unmyelinated (outlined by dashed lines) axons were found in GBA -/medaka, which contain vacuoles and electron-dense bodies. Scale bars, 2 μm. Right panels: High-magnification images of swellings of myelinated and unmyelinated axons (lower and upper panels, respectively). Scale bars, 500 nm. (B) Co-localization analysis for different markers in axonal swellings of GBA -/medaka. The correlation between LC3, ubiquitin, and α-syn signals are shown as the pixel scatter diagrams and a graph (n = 6). For all analyses, data are the mean ± SEM. (TIF)