Hexb enzyme deficiency leads to lysosomal abnormalities in radial glia and microglia in zebrafish brain development

Abstract Sphingolipidoses are severe, mostly infantile lysosomal storage disorders (LSDs) caused by defective glycosphingolipid degradation. Two of these sphingolipidoses, Tay Sachs and Sandhoff diseases, are caused by β‐Hexosaminidase (HEXB) enzyme deficiency, resulting in ganglioside (GM2) accumulation and neuronal loss. The precise sequence of cellular events preceding, and leading to, neuropathology remains unclear, but likely involves inflammation and lysosomal accumulation of GM2 in multiple cell types. We aimed to determine the consequences of Hexb activity loss for different brain cell types using zebrafish. Hexb deficient zebrafish (hexb−/−) showed lysosomal abnormalities already early in development both in radial glia, which are the neuronal and glial progenitors, and in microglia. Additionally, at 5 days postfertilization, hexb −/− zebrafish showed reduced locomotor activity. Although specific oligosaccharides accumulate in the adult brain, hexb−/−) zebrafish are viable and apparently resistant to Hexb deficiency. In all, we identified cellular consequences of loss of Hexb enzyme activity during embryonic brain development, showing early effects on glia, which possibly underlie the behavioral aberrations. Hereby, we identified clues into the contribution of non‐neuronal lysosomal abnormalities in LSDs affecting the brain and provide a tool to further study what underlies the relative resistance to Hexb deficiency in vivo.

1991). β-Hexosaminidase plays a role in ganglioside metabolism and hydrolyzes GM2 gangliosides into GM3 gangliosides in lysosomes, which in turn is further degraded by other lysosomal enzymes. Loss of either HEXA or HEXB results in accumulation of the ganglioside GM2, and therefore these disorders are clustered as GM2 gangliosidoses. GM2 accumulation is mainly found in brain cells, including neurons and glia (Kawashima, Tsuji, Okuda, Itoh, & Nakayama, 2009;Kyrkanides et al., 2012;Tsuji et al., 2005). It is unknown however, what molecular and cellular events precede pathology and cause SD or Tay-Sachs disease.
In the most common and severe form of SD, symptoms appear during infancy. At about 3 to 6 months after birth, after an initially normal postnatal development, development slows down and skeletal muscles weaken, causing loss of acquired motor abilities such as turning over, sitting upright, and crawling (Hendriksz et al., 2004). Most symptoms are neurological and include an exaggerated startle response to loud noises, seizures, vision and hearing loss, intellectual disability, and paralysis. There is no treatment available and infantile Sandhoff patients typically do not live longer than 5 years (Hendriksz et al., 2004). Due to the very early onset and rapid disease progression, the initial cellular events leading to neuronal loss remain obscure.
SD neuropathology involves neuronal loss, the presence of increased numbers of activated microglia-the brain's macrophages-and astrocytes (Jeyakumar et al., 2003;Myerowitz et al., 2002;Sargeant et al., 2012). Expansion of both microglia and astrocyte numbers are in fact found in many brain disorders including epilepsy, other LSDs, multiple sclerosis, and neurodegenerative disorders, and likely play a role in pathogenesis (Fakhoury, 2018;Geloso et al., 2017;Joe et al., 2018;Lee et al., 2016;Oosterhof et al., 2018;Ponath, Park, & Pitt, 2018;Zhao et al., 2018). In SD mice, GM2 storage is found in lysosomes of neurons, but also in lysosomes of astrocytes and microglia (Kawashima et al., 2009;Kyrkanides et al., 2012;Tsuji et al., 2005). Based on pathology and gene expression data, it is still unclear which cell types are responsible for the initiation and progression of the pathology. In SD mice, microglial activation was found to precede neurodegeneration (Wada, Tifft, & Proia, 2000). In microglia, HEXB is remarkably highly expressed compared to expression in other brain cell types and even considered to be a microglial signature gene (Artegiani et al., 2017;Bennett et al., 2018;Butovsky et al., 2014;Gosselin et al., 2017;Hickman et al., 2013;Oosterhof et al., 2017). Given that microglia are highly phagocytic, lysosomal function is critical for processing ingested material and HEXB could also be important for microglial function. Several studies showed that suppression of microglial or astrocytic inflammation could reduce SD pathology, suggesting both glial cells might be involved in SD pathogenesis (Tsuji et al., 2005;Wada et al., 2000;Wu & Proia, 2004). It is important to determine if, and to what extent, HEXB deficiency also affects glial function or if gliosis is a consequence of neuronal problems directly caused by the loss of HEXB. Analysis of hexb −/− mouse fetal neuronal stem cells (NSCs), also called radial glia, revealed a more frequent differentiation towards the astrocyte lineage rather than neurons, indicating abnormal differentiation of brain cells . SD patient IPS cell-derived organoids also showed aberrant neuronal differentiation (Allende et al., 2018). These studies suggest that HEXB deficiency could intrinsically affect neuronal differentiation and induce increased astrocyte numbers, which could be an initiating event in the pathogenesis of SD.
Potential therapeutic options for SD include gene therapy, immune suppression, and bone marrow transplantation. Gene therapy in SD mouse models, applied by intravenous viral transduction from postnatal day one and two, prevented pathology (Niemir et al., 2018). However, the subclinical disease process preceding the onset of symptoms makes early intervention using gene therapy difficult to implement. Immune suppression or bone marrow transplantations improve neurological function in SD mice, likely by reducing gliosis (Abo-Ouf et al., 2013;Ogawa et al., 2018;Wada et al., 2000;Wu, Mizugishi, Bektas, Sandhoff, & Proia, 2008). Although these therapies seem very promising, it is important to understand to what extent such approaches are targeting the cause, rather than inhibiting the symptoms. Therefore, understanding the early cellular pathogenic processes will facilitate the development of an effective treatment for SD.
The disease course in these animals is similar to that in humans, showing the conserved importance of proper processing of gangliosides. To understand SD pathogenesis, we aimed to visualize the earliest lysosomal abnormalities to discern the affected cell types during brain development in vivo. We used the zebrafish as a model since their offspring develop ex utero allowing the visualization of pathology in vivo using live imaging, they have a relative small size and are transparent, and have previously been used to study lysosomal (dys)function (Berg et al., 2016;Festa et al., 2018;Keatinge et al., 2015;Li et al., 2017;Louwette et al., 2013;Shen, Sidik, & Talbot, 2016). Moreover, zebrafish harbor a single HEXB homolog, hexb, and transgenic reporter lines are available to label different cell types. We generated an SD zebrafish model in which we investigated the earliest emergence of cell-type specific lysosomal abnormalities in the developing brain. Already in the first days of brain development, we identified increased lysosomal volume in radial glia and altered lysosomal morphology in microglia. With the present study, we identified some of the cellular consequences of loss of Hexb enzyme activity during embryonic brain development, showing early effects on glia, which possibly underlie the behavioral aberrations.
2 | RESULTS 2.1 | Direct mutagenesis causes Hexb enzyme deficiency and glial lysosomal abnormalities Zebrafish Hexb shares~64% sequence identity with the human protein, and is encoded by a single hexb gene. We targeted hexb by injecting fertilized zebrafish oocytes at the one-cell stage with Cas9 protein/gRNA complex targeting exon 1 of hexb. To determine the efficiency of targeting the zebrafish hexb locus, the generation of insertions and deletions (InDels) was determined by Sanger sequencing genomic DNA from individual 4 dpf larvae (n = 24). This was followed by sequence decomposition using TIDE showing that, on average, 94% of the alleles harbor InDels (i.e., mutagenesis efficiency) ( Figure 1a) (Brinkman, Chen, Amendola, & van Steensel, 2014). We refer to the F0 generation of larvae that are directly injected with Cas9 protein/gRNA complex as crispants.
HEXA and HEXB enzyme activity are both affected in SD patients. Therefore, we analyzed β-Hexosaminidase (Hexa and Hexb) enzyme activity in dissected heads of control and hexb crispants. Tail genomic DNA of these larvae showed highly efficient (97-100%) InDel generation, and, consistent with loss of Hexb activity, hexb crispants showed >95% detectable reduction in substrate conversion by β-Hexosaminidase A + B ( Figure 1b).
To visualize lysosomes in vivo we used lysotracker (LT) staining, which labels all acidic organelles including lysosomes, and to visualize microglia we used transgenic zebrafish in which GFP expression is driven by the mpeg1-promotor (Berg et al., 2016;Ellett, Pase, Hayman, Andrianopoulos, & Lieschke, 2011;Shen et al., 2016). At 4 dpf, although there were no major differences in microglial numbers, F I G U R E 1 Efficient mutagenesis in hexb crispants shows abnormal lysosomes in microglia and radial glia. (a) graph shows mutagenesis efficiency for 24 injected fish at 4 dpf. Each dot represents 1 larva. (b) hexb crispants show almost no enzymatic conversion of substrate by β-Hexosaminidase A + B at 4dpf. (c) Representative images of microglia in 8 dpf larvae imaged in the depicted area. Scale bar represents 20 μm. Quantification of lysotracker (LT+) area within microglia relative to the total microglial area in 8 dpf larvae. Each dot represents 1 larva. (d) Expression pattern of the tgBAC(slc1a2b:Citrine) transgenic line at 4 dpf, with magnification/sub-stack showing the radial fibers in the optic tectum, and a schematic showing radial glia orientation and direction of neurogenic migration along the radial fibers. Scale bars represent 100 and 20 μm, respectively. (e) Representative images of radial glia at 4 dpf in controls and in hexb crispants with quantification of relative LT+ signal per measured total area. Each dot represents 1 larva. Scale bars represent 10 μm. Error bars represent SD [Color figure can be viewed at wileyonlinelibrary.com] lysosomal compartments of hexb deficient microglia were found to be enlarged compared to those in controls, suggesting that Hexb deficiency results in lysosomal abnormalities in microglia (Figure 1c,d).
Unexpectedly, however, we also observed numerous LT+ speckles outside microglia that were more prominently present in hexb crispants ( Figure 1c; Movie S1). These speckles are most likely lysosomes of another cell type that is not fluorescently labeled in the mpeg1 reporter line.
Astrocytes have been shown to accumulate gangliosides GM2 and GA2 in postnatal Hexb −/− mice, and showed increased proliferation (Kawashima et al., 2009). In zebrafish, radial glia has both astrocytic and radial glia/neural stem cell properties (reviewed in: [Lyons & Talbot, 2014]). For example, homologs of two mammalian astrocytes genes, Slc1a2, encoding the glutamate transporter GLT-1, and Gfap, are expressed in radial glia in zebrafish (slc1a2b and gfap) (McKeown, Moreno, Hall, Ribera, & Downes, 2012). Similarly, astrocyte endfeet in mammals express tight junction proteins, which in zebrafish are also expressed particularly in radial glia (Corbo, Othman, Gutkin, Alonso Adel, & Fulop, 2012;Marcus & Easter, 1995). We hypothesized that the lysosomal puncta observed outside microglia in hexb crispants may reside in radial glia (RG). We used bacterial artificial chromosome 2.2 | Early abnormal lysosomal phenotypes both in microglia and RG of hexb −/− larvae To validate and further investigate the Hexb deficient phenotype, which we observed in hexb crispants, we generated stable mutants containing a 14 bp frame-shifting deletion in the first exon of hexb, leading to 19 out of frame amino acids followed by a premature stop codon ( Figure 2a). To test whether this causes β-Hexosaminidase (Hexa + Hexb) deficiency, we measured β-Hexosaminidase A + B activity in lysates from control and hexb −/− larval heads, and also in adult brain and internal organs-containing: intestine, liver, spleen, pancreas, and gonads-, which showed that activity was reduced by >99% ( Figure 2b). Consistently, analysis of specific oligosaccharides, which accumulate, as a consequence of HEXB deficiency, in adult brain and internal organs-containing: intestine, liver, spleen, pancreas, and gonads-, showed accumulation in both brain and internal organs from hexb −/− fish, but was hardly or not detectable in brain or organs from controls ( Figure 2b). Surprisingly, adult hexb −/− zebrafish were viable and do not show obvious motility defects.
We wanted to further explore the consequences of Hexbdeficiency on brain function and development, and determined by LT staining when lysosomal aberrations in microglia first appeared. We first quantified LT-stained surface in control and hexb −/− microglia in 5 dpf larvae, which in contrast to the small but significant increase in lysosomal volume that we observed in 8 dpf hexb crispants, showed no changes ( Figure 2c). However, detailed inspection of lysosomes in hexb −/− and control microglia showed changes in lysosomal morphology. Hexb −/− microglia showed fewer lysosomes consisting of a single round compartment, but more lysosomes that seemed to consist of clustered small lysosomes. To quantify this, we defined three lyso- Since the hexb −/− larvae presented with lysosomal abnormalities in microglia preceding those in RG, it is possible that abnormalities in RG lysosomes occur secondary to a defect in microglia. Therefore, we analyzed LT-staining in RG in csf1r mutant fish, containing almost no microglia during early development (csf1r DM ) (Oosterhof et al., , 2019. These csf1r DM larvae did not retain LT+ speckles in radial fibers of radial glia at 5 dpf, as we did observe in hexb −/− larvae (Figure 3b-  (Burgess & Granato, 2007;Kokel et al., 2010;MacPhail et al., 2009). Locomotor activity over repeated light-dark cycles was recorded by infrared imaging and measured by automated tracking.
Although control and hexb −/− larvae both responded to light/dark switches, hexb −/− larvae showed reduced locomotor activity, which was consistently detected at 4, 5, and 6 dpf (n = 24 per group, per experiment) (Figure 4a). This test mainly focuses on the behavioral response to shifts in light intensity, whereas we were mainly interested in general locomotor activity. Therefore, we developed a locomotor assay consisting of gradually increasing and decreasing light intensities, mimicking more physiological relevant changes regulating circadian behavior. Hexb −/− larvae showed >50% less locomotor activity during the light periods in this dusk/dawn cycle assay (Figure 4b).
These findings indicate that reduced locomotor activity of hexb −/− larvae is not due to a defect in the behavioral response to light/dark changes, but due to overall reduced locomotor activity compared to controls. Thus, hexb −/− larvae show abnormally low locomotor activity, which coincides with the timing of appearance of glial lysosomal abnormalities in the brain. SD pathology is characterized by extensive loss of neurons. During normal vertebrate brain development extensive apoptosis occurs, and in zebrafish developmental apoptosis in brain development is particularly high at 3 and 4 dpf (Cole & Ross, 2001;Sierra et al., 2010;Southwell et al., 2012;van Ham, Kokel, & Peterson, 2012;Xu, Wang, Wu, Jin, & Wen, 2016). To determine whether there is more apoptosis in hexb −/− brains, we used transgenic zebrafish expressing secreted fluorescently labeled Annexin V to analyze apoptotic cells (Morsch et al., 2015). We did not detect differences in the number of Annexin V+ apoptotic cells or debris at 3, 4, 5, and 7 dpf (Figure 4d). We also performed TUNEL labeling, which stains late-apoptotic cells, in 5 dayold brains and observed despite the low numbers of late-apoptotic cells an~twofold increase in the hexb −/− brains (Figure 4c). Taken together, Hexb deficiency in zebrafish leads to reduced locomotor activity, more late-apoptotic cell numbers in the brain at Day 5, but no increase in overall apoptotic debris between Days 3 and 7.

| DISCUSSION
Here we generated hexb −/− zebrafish, which are β-Hexosaminidase enzyme-deficient, and their pathology mimics molecular, cellular and behavioral phenotypes that are also observed in HEXB-deficient Sandhoff disease patients. Already within 5 days of development, hexb −/− larvae presented with abnormal lysosomes in microglia and in radial glia. The appearance of abnormal lysosomes in glia coincided with reduced locomotor activity. These phenotypes were observed both in Cas9/hexb gRNA injected and in hexb −/− larvae, which indicates that crispants are a very suitable method to rapidly screen for pathologic phenotypes, also for other lysosomal or metabolic diseases. Based on our zebrafish model, the early presentation of abnormal lysosomes in glial cells implies that pathogenesis might be characterized by dysfunctional glial cells in the first stages of SD.
Both in controls and hexb −/− RG, we observed abundant lysosomal speckles in the radial fibers, at 3 and 4 dpf. In controls, the lysosomal speckles were largely undetectable by 5 dpf, whereas in hexb −/− fish these abundant RG lysosomes persisted or remained enlarged.
Enlarged lysosomal compartments are a characteristic of LSDs. This suggests that, to resolve lysosomes in RG, β-Hexosaminidase enzyme activity is required, and could imply that extensive processing of sphingolipids such as GM2 occurs at this stage of brain development.
To determine whether microglial function might be involved in the resolution of LT+ speckles in radial glia we analyzed csf1r DM larvae, which are almost devoid of microglia . The absence of LT+ speckles in radial glia of microglia-less mutants shows that the resolution of speckles occurs independently of microglia. Therefore, Hexb activity seems required cell autonomously to resolve LT+ speckles in radial glia. We did however find increased volume of LT staining in RG of csf1r DM larvae, but these appear different from the LT+ speckles observed in hexb mutants, as they are larger in size and located in phagolysosome-like structures. Since radial glia or astrocytes may take over phagocytic functions usually performed by microglia, these large LT+ lysosomes could be the result of phagocytic activity of radial glia in the absence of microglia in csf1r DM larvae (Puñal et al., 2019). Importantly, RG in vertebrates are neuronal stem cells, and are thought to produce most neurons, but also astrocytes and oligodendrocytes (Doetsch, 2003). This process requires extensive membrane remodeling involving synthesis and degradation of membrane lipids, including glycolipids, in which lysosomes play a key role (Jaishy & Abel, 2016;Kolter & Sandhoff, 2010;Sild, Van Horn, Schohl, Jia, & Ruthazer, 2016). Consistent with this idea, differentiation of RG into astrocytes in mice was accompanied by an increase in lysosomes and autophagosomes (Gressens, Richelme, Kadhim, Gadisseux, & Evrard, 1992). Furthermore, differentiation of SDpatient iPSC-derived RG to neurons and astrocytes was skewed towards astrocyte development (Ogawa, Kaizu, et al., 2017). Based on F I G U R E 4 Locomotor activity assay shows abnormal locomotor activity in hexb −/− larvae. (a) Representative graph showing the locomotor responses of larvae to alternating light and dark periods indicated by white and black squares in one single experiment. The dot plot shows the quantification of the sum distance moved during all the light and all the dark periods. (b) Representative graph showing the total distance travelled by larvae during the dusk-dawn routine (total time: 3 hr 12 min). Quantification of the total distance moved throughout the experiment excluding the dark period. (a, b) n = 24 larvae per genotype. Grey shading shows the standard error of the mean. (c) Representative images of control and hexb −/− larvae at 4 dpf transiently expressing secreted annexin V tagged with mVenus. Quantification of annexin V positive objects at 3, 4, 5, and 7 dpf. Scale bar represents 100 μm. (d) Representative images of TUNEL staining at 5 dpf with quantifications. Each dot represents one larva. Scale bar represents 20 μm. Error bars represent SD unless stated otherwise [Color figure can be viewed at wileyonlinelibrary.com] these findings, the persistence or enlargement of lysosomal speckles that we observed in hexb −/− RG in vivo, could be the result of impaired glycolipid metabolism, essential for membrane remodeling, and this could influence differentiation towards neurons and glia.
The hexb mutant locomotor phenotype is likely the consequence of alterations in the brain, as we did not observe obvious alterations in peripheral motorneurons and neuromuscular junctions (data not shown), whereas aberrant lysosomes in glia are already detected at these early stages. The reduced locomotor activity could be the result of lysosomal dysfunction in either microglia or RG. Microglia have essential functions for brain hemostasis, including clearance of apoptotic cells, myelin remodeling and synaptic pruning (Hagemeyer et al., 2017;Paolicelli et al., 2011;Reemst, Noctor, Lucassen, & Hol, 2016;Safaiyan et al., 2016;Sierra et al., 2010;Wlodarczyk et al., 2017). Defects in such processes could perhaps contribute to defects in locomotor control in the hindbrain leading to locomotor aberrations (Chong & Drapeau, 2007;Kimura et al., 2013). RG on the other hand produce neurons in the brain and alterations in neurogenesis during early development likely affects brain circuitry formation, which could also explain the reduced locomotor activity.
Gangliosidoses due to HEXA or HEXB mutations are associated with severe neurological symptoms that often precede early lethality. These diseases have been described in several mammalian species, including dogs, cats, and Muntjak deer, and even in birds as Hexa deficiency has been observed in Flamingo's (Cork et al., 1977;Kanae et al., 2007;Kolicheski et al., 2017;Martin et al., 2004;Rahman et al., 2012;Torres et al., 2010;Zeng et al., 2008). hexb deficient zebrafish seem to recapitulate early disease stages, but lack extensive neurodegeneration. In addition, adult homozygous hexb −/− zebrafish, even though they lack Hexb activity, are viable and do not show obvious behavioral symptoms. To the best of our knowledge, zebrafish are the first vertebrate species tested that do not succumb to pathology caused by loss of Hexb activity. Mice deficient in Hexa do not have a severe phenotype as they can bypass the need for HEXBA heterodimer activity by an alternative pathway requiring a sialidase, NEU3, to degrade gangliosides (Seyrantepe et al., 2018). Mass spectrometry analysis of oligosaccharides in zebrafish brains, however, did show the presence of specific oligosaccharides, which accumulate as a consequence of HEXB deficiency, suggesting there is no alternative conversion route active in zebrafish. This suggests that zebrafish are, perhaps to a certain extent, resistant to Hexb deficiency. In addition, zebrafish, in contrast to mammalian species, can regenerate brain tissue and produce new neurons throughout life (Becker & Becker, 2008;Becker & Becker, 2014;Caldwell et al., 2019;Kizil, Kaslin, Kroehne, & Brand, 2012).
In conclusion, hexb −/− zebrafish show loss of enzyme activity, accumulation of oligosaccharides, lysosomal abnormalities in glia and reduced locomotor activity, reminiscent of the molecular, cellular, neuropathological, and motility aspects of Sandhoff disease. The very early accumulation of lysosomes in radial processes of RG and raspberry-shaped lysosomes in microglia, occurring apparently independently of cell death in the brain, may represent early events in pathogenesis. These Hexb deficient zebrafish could be used to further dissect cellular mechanisms that render them resistant to the effects of Hexb deficiency, which leads to disease in various other species, and to identify small molecules that suppress the earliest emergence of cellular pathogenic features.

| Generation of the tgBAC(slc1a2b:Citrine) reporter line
To generate the tgBAC(slc1a2b:Citrine) re01tg reporter line we used the following clone DKEY-49F8 (HUKGB735F0849Q) in the pIndigoBAC-536 vector. To perform recombination to insert tol2 sites and the cytoplasmic Citrine, the BAC recombineering protocol developed by Bussmann and Schulte-Merker (2011) was used. Primers are described in Table 1.

| TUNEL staining
For staining larvae were euthanized on ice, followed by fixation in 4% paraformaldehyde in PBS at 4 C overnight, permeabilized by proteinase K (10 μg/mL) in PBST treatment for 40 min and refixed in PFA for 20 min at room temperature. TUNEL staining was performed as previously described using the Click-iT TUNEL Alexa Fluor 647 Kit (Invitrogen, Carlsbad) (C10247) (van Ham, Mapes, Kokel, & Peterson, 2010). TUNEL+ cells in the whole brain were quantified.

| Mass spectrometric analysis of oligosaccharides
Tissues were obtained from adult zebrafish, euthanized on ice water, by microdissection as described . Tissue samples were homogenized in 80 μL water, containing 18 μmol/L triacetylchitotriose and 18 μmol/L hexa-acetylchitohexaose as internal standards, using a Potter-Elvehjem tube, followed by sonication at 130 W for 10 s on ice. Insoluble material was removed by centrifugation at 13,000g for 10 min. The supernatant was dried under nitrogen flow at 40 C and the resulting residue was dissolved in 80 μL water. The following oligosaccharides were determined semiquantitatively using highresolution mass spectrometry: Hex2-HexNAc2, Hex3-HexNAc3, and Hex3-HexNAc4 (Bonesso et al., 2014). Chromatography was performed using a Thermo Scientific Ultimate 3,000 UHPLC system, equipped with a BEH Amide column (Waters 2.1 × 50 mm, 1.7 μm particle size). Analytes were separated by a gradient LC method using mobile phase consisting eluent A (10 mM ammonium acetate in 10:90 water/acetonitrile + 0.1% w/v ammonium hydroxide) and B (10 mM ammonium acetate in 70:30 water/acetonitrile adjusted to pH 9 with ammonium hydroxide) at a flow T A B L E 1 Primers used for BAC recombineering  Images were processed with ImageJ software.

| β-Hexosaminidase activity assay
Heads were dissected from 4 dpf control larvae and hexb crispants or mutants, after euthanizing them with ice water and a tricaine overdose, and analyzed for enzymatic activity. Adult pools of 2 or 3 dissected heads were grouped according to their genotype and homogenized using sonication in 100 μL H 2 O. For adult organs, animals were euthanized using ice water, Protein concentration was measured using BCA reagent by spectrophotometry after 1-hr incubation at 37 C (Pierce).
The substrate, 5 mM MU-β-GlcNAc in 0.2 M Na-phosphate/0.1 M citrate buffer, pH 4.4 + 0.02% (w/v) NaN 3 , was added to each lysate pool, including one replicate per sample, and incubated for 1-hr at 37 C in which β-Hexosaminidase will form 4-MU. After addition of 0.5 M Na 2 CO 3 buffer, pH 10.7, 4-MU fluorescence was measured and normalized to the blank (substrate alone). The final read-out consisted of nmoles of 4-MU released per hour per mg of protein.

| Locomotor activity assays
To assess the locomotor activity of zebrafish larvae from 3 to 5 dpf, locomotor activity assays were performed using an infrared camera system (DanioVision™ Observation chamber, Noldus) and the using EthoVision ® XT software (Noldus). Control (n = 24) and hexb −/− (n = 24) zebrafish larvae, in 48 well plates, were subjected to two different light/dark routines. The White Light routine consisted of a 30 min habituation period, followed by 4 cycles of 15 min of light (100%)/ 15 min darkness 15 (2.5 hr total). The Dawn routine 30 min habituation in the dark (0% light intensity), followed by the routine described in Table 3 and comprised 3 hr 12 min. Each experiment was performed twice for the three ages (3, 4, and 5 dpf) with n = 24 controls, and n = 24 mutants per experiment. Distance traveled (mm) per second was measured.

| Statistical analysis
To measure mpeg1:GFP and LysoTracker volumes in confocal microscopic imaging data, images were analyzed using automatic thresholding in Fiji/ImageJ. The area of the LysoTracker inside mpeg1:GFP+ cells was measured, by generating a mask for microglia in FIJI, and compared to the total mpeg1:GFP+ area. For the Lysotracker areas of the radial glia, images were automatically thresholded and the total LT+ area per image was counted using the Analyze Particles plug-in (FIJI). For the radial glia measurements, quantification the total LT+ area relative to total imaged area was used. To quantify the annexin V signal we used the 3D object counter plug-in (FIJI). For image processing and quantitative analysis, Excel (Microsoft), Prism (Graphpad) and ImageJ were used. Statistical significance was calculated using Student's t test. Error bars indicate standard deviation (SD) and p < .05 was considered significant. Symbols in the graphs (*; **; ***) display levels of significance (p < .05; p < .01; p < .001 respectively).

CONFLICT OF INTEREST
No authors declare competing interests.