Loss of Bace2 in zebrafish affects melanocyte migration and is distinct from Bace1 knock out phenotypes


  • Frauke van Bebber,

    1. German Center for Neurodegenerative Diseases (DZNE), Munich, Germany
    2. Adolf-Butenandt Institute, Biochemistry, Ludwig-Maximilians University Munich, Munich, Germany
    Search for more papers by this author
  • Alexander Hruscha,

    1. German Center for Neurodegenerative Diseases (DZNE), Munich, Germany
    Search for more papers by this author
  • Michael Willem,

    1. Adolf-Butenandt Institute, Biochemistry, Ludwig-Maximilians University Munich, Munich, Germany
    Search for more papers by this author
  • Bettina Schmid,

    Corresponding author
    1. German Center for Neurodegenerative Diseases (DZNE), Munich, Germany
    2. Adolf-Butenandt Institute, Biochemistry, Ludwig-Maximilians University Munich, Munich, Germany
    3. Munich Centre for Systems Neurology (SyNergy), Munich, Germany
    • Address correspondence and reprint request to Christian Haass or Bettina Schmid, German Center for Neurodegenerative Diseases (DZNE), Munich and Adolf-Butenandt Institute, Biochemistry, Ludwig-Maximilians-University, Schillerstrasse 44, 80336 Munich, Germany. E-mails: christian.haass@dzne.lmu.de; bettina.schmid@dzne.lmu.de

    Search for more papers by this author
  • Christian Haass

    Corresponding author
    1. German Center for Neurodegenerative Diseases (DZNE), Munich, Germany
    2. Adolf-Butenandt Institute, Biochemistry, Ludwig-Maximilians University Munich, Munich, Germany
    3. Munich Centre for Systems Neurology (SyNergy), Munich, Germany
    • Address correspondence and reprint request to Christian Haass or Bettina Schmid, German Center for Neurodegenerative Diseases (DZNE), Munich and Adolf-Butenandt Institute, Biochemistry, Ludwig-Maximilians-University, Schillerstrasse 44, 80336 Munich, Germany. E-mails: christian.haass@dzne.lmu.de; bettina.schmid@dzne.lmu.de

    Search for more papers by this author


Alzheimer's disease is the most frequent dementia. Pathologically, Alzheimer's disease is characterized by the accumulation of senile plaques composed of amyloid β-peptide (Aβ). Two proteases, β- and γ-secretase proteolytically generate Aβ from its precursor, the ß-amyloid precursor protein (APP). Inhibition of β-secretase, also referred to as beta-site APP cleaving enzyme (BACE1) or γ-secretase is therefore of prime interest for the development of amyloid-lowering drugs. To assess the in vivo function of zebrafish Bace1 (zBace1), we generated zBace1 knock out fish by zinc finger nuclease-mediated genome editing. bace1 mutants (bace1−/−) are hypomyelinated in the PNS while the CNS is not affected. Moreover, the number of mechanosensory neuromasts is elevated in bace1−/−. Mutations in zebrafish Bace2 (zBace2) revealed a distinct melanocyte migration phenotype, which is not observed in bace1−/−. Double homozygous bace1−/−; bace2−/− fish do not enhance the single mutant phenotypes indicating non-redundant distinct physiological functions. Single homozygous bace1 mutants as well as double homozygous bace1 and bace2 mutants are viable and fertile suggesting that Bace1 is a promising drug target without major side effects. The identification of a specific bace2 −/− associated phenotype further allows improving selective Bace1 inhibitors and to distinguish between Bace 1 and Bace 2 inhibition in vivo.


Inhibition of BACE1 protease activity has therapeutic importance for Alzheimer's disease. Analysis of BACE1 and BACE2 knock-out zebrafish revealed that they exhibit distinct phenotypes. bace1 mutants display hypomyelination in the PNS and supernumerary neuromasts while in bace2 mutants the shape and migration of melanocytes is affected. These phenotypes are not further enhanced in the viable double mutants. Our data suggest that blocking BACE1 activity is a safe therapeutic approach.

Abbreviations used

Alzheimer's disease


anterior lateral line neurons


amyloid precursor protein


beta-site APP cleaving enzyme


neuregulin 1


zinc finger nucleases

Age-related dementias such as Alzheimer's disease (AD) are dramatically increasing in our aging societies presenting a huge socio-economical burden. To date, no treatment targeting the cause of AD is available. AD is pathologically characterized by the accumulation of extracellular plaques consisting mainly of amyloid β peptide (Aβ) and intracellular neurofibrillary tangles composed of hyperphosphorylated Tau protein. The main component of plaques in Alzheimer's disease brains, the 37–43 amino acid (aa) Aβ, is generated through progressive proteolytic processing from the β-amyloid precursor protein (APP) (Haass and Selkoe 2007). Cleavage by β-secretase, also known as Beta-site APP Cleaving Enzyme (BACE)1 initiates the release of Aβ (Vassar and Citron 2000; Selkoe 2001). This generates a membrane bound C-terminal fragment of APP (C99) and results in the release of truncated full length APP (βAPPs). C99 is the immediate substrate for subsequent intra-membrane cleavage by γ-secretase (Haass 2004; Haass and Selkoe 2007), which finally releases Aβ into the extracellular space. In contrast to BACE1, α-secretase cleaves within the Aβ domain and has therefore anti-amyloidogenic properties (Haass 2004).

Targeted disruption of BACE1 in mice showed that it constitutes the sole β-site APP cleaving activity (Luo et al. 2001; Roberds et al. 2001; Harrison et al. 2003), whereas the BACE1 homolog, BACE2, does not exhibit an amyloidogenic function. BACE1 knock-out (KO) mice do not display any overt phenotype. They are viable and have initially been characterized as phenotypically rather normal (Luo et al. 2001; Roberds et al. 2001). However, subtle phenotypes such as reduced weight (Harrison et al. 2003; Dominguez et al. 2005), increased anxious behavior (Harrison et al. 2003), mild cognitive deficits (Ohno et al. 2004; Laird et al. 2005), hyperactivity, and enhanced lethality of pubs (Dominguez et al. 2005), hypomyelination (Hu et al. 2006; Willem et al. 2006), impaired axon guidance (Rajapaksha et al. 2011; Hitt et al. 2012), seizures (Hu et al. 2010), reduced voltage-gated sodium channel (Na(v)1) beta2-subunits (beta2) (Kim et al. 2007, 2011), and retinal pathology (May et al. 2011; Cai et al. 2012) have been described. So far in only a few cases phenotypical alterations could be directly related to a distinct BACE1 substrate, namely Na(v)1 beta2 (Wong et al. 2005; Kim et al. 2007), neuregulin 1 (Nrg1) type III (Hu et al. 2006; Willem et al. 2006), the neural adhesion molecules L1, and the close homolog of cell adhesion molecule L1 (Hitt et al. 2012; Kuhn et al. 2012; Zhou et al. 2012). Neuregulin is particularly interesting because the BACE1 KO phenocopies the heterozygous NRG1 type III KO by presenting hypomyelination and increased numbers of axons within remark bundles (Hu et al. 2006; Willem et al. 2006). The effects of the BACE1 KO on central nervous system myelination is still controversial (Hu et al. 2006; Willem et al. 2006). The BACE2 KO did not result in hypomyelination and failed to enhance the BACE1 myelination phenotype (Willem et al. 2006), suggesting distinct functions of these closely related proteases. However, very little is known about BACE2 function in vivo and no phenotypes have been reported for mice lacking BACE2. So far it has only been described that a homozygous double KO of BACE1 and BACE2 enhances the lethality observed in BACE1−/− animals for so far unknown reasons (Dominguez et al. 2005).

BACE1 is an atypical membrane anchored aspartyl protease (Hussain et al. 1999; Sinha et al. 1999; Vassar et al. 1999; Yan et al. 1999; Lin et al. 2000). The only other membrane bound aspartyl protease is its close paralog BACE2 (Saunders et al. 1999; Bennett et al. 2000; Farzan et al. 2000). No true BACE1 and BACE2 orthologs have been identified in invertebrates, because the closest homologous protein lacks the transmembrane domain characteristic for BACE1 and BACE2 (Kotani et al. 2005). BACE proteases evolved rather late during evolution and have exclusively been found in vertebrates thus far. A zebrafish BACE1 ortholog, zBace1, which is highly homologous to human BACE1 and shares both protease active sites as well as the transmembrane domain, has been identified (Moussavi Nik et al. 2012). We now report that zebrafish also have a BACE2 ortholog. Therefore, the zebrafish comprises a very attractive new vertebrate model organism to study Bace function in a whole organism approach, specifically as the small size of zebrafish larvae allows chemical screens and their transparency make them ideal for in vivo imaging.

Material and methods

Zebrafish maintenance and transgenic lines used

The wildtype strain AB was used in all experiments unless stated otherwise. All embryos were kept at 28.5°C in E3 media (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, and 0.33 mM MgSO4) supplemented with 10−5% methylene blue and were staged according to Kimmel et al. (1995). All experiments were performed in accordance with animal protection standards of the Ludwig-Maximilians University Munich and were approved by the government of Upper Bavaria (Regierung von Oberbayern, Munich, Germany).

The transgenic line Tg(foxd3:GFP), which drives expression of GFP under the control of the forkhead box D3 promoter (foxd3) (Gilmour et al. 2002) was used for analysis of Schwann cells. The tg(foxd3:GFP) transgenic line is referred to as foxd3:GFP throughout the article. The double transgenic line Tg(claudin k:Gal4); tg(14xUAS:GFP) that drives membrane-bound GFP expression under control of the claudin k promoter (Munzel et al. 2012) was used for analysis of myelination and is referred to as claudin k:GFP throughout the manuscript. For the zBace2 functional analysis the bace2hu3332 allele (provided by the Hubrecht laboratory and the Sanger Institute Zebrafish Mutation Resource) was backcrossed for 3 generations and used in all experiments.

Cloning of genes and sequence analysis

Full-length zbace1 cDNA sequence was obtained by amplification from EST clone IMAGE 6972051. Alignment was obtained by comparison of full zBace1 protein sequence to human BACE1 (hBACE1) (>gi|6912266|ref|NP_036236.1) using CLC Main Workbench 6. Full-length zbace2 cDNA sequence was amplified from EST clone IMAGE 5777048. The subsequent alignment was created using CLC Main Workbench 6 to compare zBace2 protein to hBACE2 (>gi|19923395|ref|NP_036237.2).

Zinc finger nucleases and identification of induced genomic lesions in zbace1

CompoZr custom zinc finger nucleases (ZFN) were designed for the zbace1 locus (Sigma, St Louis, MO, USA). ZF nucleotide recognition sequences are highlighted in upper case:

zbace1 ZFN set 2 forward: 5′-gaGTCGCGGTGtGTGTCGcgcggctctg-3′

zbace1 ZFN set 2 reverse: 5′-gcACAGCAGAAGCAGACgcgccgcgcgc -3′

Genomic lesions induced by ZFN were identified by PCR amplification of the ZFN cut site and subsequenct restriction fragment length polymorphism analysis.

PCR amplification of genomic DNA was performed with the following primers:


zbace1 reverse: 5′-ACTGTGGGGGCTCCTGTG-3′

Primer zbace1 reverse was used for sequencing.

Eight microliters of the 18 μL PCR reactions were subsequently digested at 37°C for 3 h with the enzyme HinfI (NEB) and used for restriction fragment length polymorphism analysis.

mRNA injections

mRNA was synthesized in vitro using the mMessage mMACHINE kit (Life Technologies, Carlsbad, CA, USA) according to the manufacture's instructions. ZFN mRNA was injected at a concentration of 400 ng/μL. zbace1 mRNA was injected at a concentration up to 600 ng/μL to assure high protein expression. Successful translation of injected mRNAs was controlled by western blot analysis.


Knock-down (KD) of Nrg1 type III in zebrafish was done by microininjection of 2-4 pL of 1 mM antisense gripNA (Active motif, Carlsbad, CA, USA) targeting the splice donor site of the nrg1 type III-specific exon.

The following grip was used (exonic sequence in upper case): Nrg1 type III grip:


RT-PCR, qRT-PCR, and western blot analysis

Total RNA was extracted from wildtype zebrafish of the indicated age using RNeasy Mini Kit (Qiagen, Hilden, Germany). First strand cDNA synthesis was performed with the M-MLV reverse transcriptase (Invitrogen, Carlsbad, CA, USA) and random hexamer primers (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions.

Primers used for detection of zbace1 by semi-quantitative RT-PCR (28 cycles):

zbace1 for: 5′-atgagactctggagccgttctt-3′

zbace1 rev: 5′-gatgacctcatagtaccactcacg-3′

Primers used for detection of actin by semi-quantitative RT-PCR (26 cycles for zbace1, 35 cycles for zbace2):



Primers used for detection of zbace2 by semi-quantitative RT-PCR (35 cycles):



For qRT-PCR, elongation factor 1 alpha (elf1α), ribosomal protein L13A (rpl13a), and zbace1 specific primer sets and the SsoFastTM EvaGreen® Supermix (Bio-Rad Laboratories, Hercules, CA, USA) were used to amplify cDNA. Analysis of this reaction was performed employing the CFX384 Real-Time System (Bio-Rad).

Primer used for qRT-PCR:







Western blots were performed according to standard methods. In summary, zebrafish of the indicated age were anesthetized and frozen in liquid nitrogen. Samples were homogenized in 4× Laemmli buffer using a motorized pellet pestle (Fisher Scientific, Hampton, NH, USA) and incubated at 95°C for 10 min. Lysates were subjected to centrifugation at 13 000 g for 1 min, and 20 μl of the supernatant was loaded on an 8% sodium dodecyl sulfate–polyacrylamide gel and subsequently transferred to a polyvinylidene difluoride membrane. Western blot analysis was carried out using the anti-hBACE1 antibody 3D5 ((Zhao et al. 2007), gift from Robert Vassar, WB 1:1000) and anti-α-Tubulin (Sigma T6199, WB 1:8000). Membranes were developed with ECLplus (GE Healthcare, Little Chalfont, UK).

Staining of neuromasts

Whole larvae were incubated for 10 min at 5 mM N-(3-triethylammoniumpropyl)-4-(6-(4-diethylamino)phenyl)hexatrienyl)pyridinium dibromide (FM4-64; Molecular Probes, Eugene, OR, USA) in E3 and washed twice for 5 min in E3.


Neuromast numbers were quantified and compared using a Student's t-test. Data are presented as mean values ± standard deviation (s.d., error bars). p-Values (p) below 0.05 were considered significant.

Image acquisition

Zebrafish embryos and larvae were dechorionated, anesthetized with Tricaine (0.016% w/v), and mounted in 1.5% low melting agarose on coverslips. DIC images were taken with an Axioplan2 compound microscope (Zeiss, Oberkochen, Germany). Fluorescence was imaged using an LSM510 META inverted confocal microscope (Zeiss) and assembled in Photoshop 8.0 (Adobe Systems, San Jose, CA, USA). Brightness and contrast were adjusted using ImageJ.


We cloned the zebrafish BACE1 ortholog (zbace1) to assay Bace1 function in zebrafish. We identified a single zbace1 gene in the zebrafish genome by database screening, which translates to a 505 aa protein as shown in a previous report (Figure S1a) (Moussavi Nik et al. 2012). Comparison of protein sequences revealed that zBace1 is 74% identical to hBACE1. The protease active sites containing the critical aspartate residues are 100% conserved (Figure S1a). Moreover the very C-terminus of BACE1 known to be important for trafficking (Walter et al. 2001) is highly conserved.

Semi-quantitative and quantitative RT-PCR revealed that zbace1 mRNA is expressed at all embryonic and larval stages analyzed with the lowest expression at 1 day post-fertilization (dpf) (Fig. 1a and b). We next investigated the temporal expression of zBace1 protein by Western blot using the monoclonal antibody 3D5 directed against the ectodomain of hBACE1 (Zhao et al. 2007). To confirm that antibody 3D5 cross-reacts with zBace1 protein we over-expressed zBace1 protein by zbace1 mRNA injection. In lysates of zebrafish embryos over-expressing zBace1 a protein of 64 kDa was detected at 5 h post-fertilization (hpf), which was not observed in uninjected age-matched siblings (Fig. 1c). Endogenous zBace1 protein was detected in lysates of whole zebrafish larvae from 3 dpf onwards and in adult zebrafish brain (Fig. 1c).

Figure 1.

zBace1 is expressed early in development. (a) zbace1 mRNA expression determined by semi-quantitative RT-PCR (upper panel). Lower panel shows the expression of actin mRNA as a control. (b) Quantitative RT-PCR for bace1, normalized to the housekeeping genes elongation factor 1 alpha (elf1α) and ribosomal protein L13A (rpl13a). (c) Developmental profile of zBace1 protein. As a positive control, total lysate of zebrafish embryos injected with zbace1 mRNA are loaded, while total lysates of the respective age were loaded to detect endogenous zBace1 with the monoclonal antibody 3D5 directed against the ectodomain of hBACE1. The size of the endogenous zBace1 corresponds to the size of the positive control. α-Tubulin is used as a loading control.

To reveal the physiological function of zBace1 in zebrafish we generated several bace1 loss of function alleles in zebrafish by genome editing using ZFN (Fig. 2a) (Urnov et al. 2010). mRNA of a ZFN set that targets bace1 in close proximity to the start codon was injected into wildtype eggs and embryos were raised to adulthood (F0 founder fish). Genomic DNA of progeny of injected founder fish (F1) was analyzed around the ZFN cleavage site by restriction fragment length polymorphism and subsequent sequencing for frameshift mutations. We analyzed 221 F1 fish and identified 50 mutation carriers with 15 different mutations. 9 out of the 15 identified mutations are deletions leading to a frameshift that causes a premature stop codon before the proteolytically active sites. We established and further analyzed 5 independent mutant lines for zbace1 that lead to premature stop codons (Fig. 2b). Western blot analysis with the zBace1 specific antibody 3D5 confirmed that no zBace1 protein is detected in homozygous bace1 mutants (bace1−/−) (Fig. 2c) and that they thus represent null alleles. Even if short N-terminal fragments would be translated, which may not be detected by 3D5 such peptides would all lack both active sites of the protease and would therefore be proteolytically inactive. bace1−/− are viable and fertile and do not show gross morphological defects. As hypomyelination has been described as the primary morphological phenotype in BACE1 deficient mice (Hu et al. 2006; Willem et al. 2006), we asked if myelination is affected in bace1−/− larvae. We took advantage of the transparent larvae and crossed a transgenic line expressing membrane-bound green fluorescent protein (GFP) under the control of the myelin specific claudin k promoter (Munzel et al. 2012) into a bace1−/− background (bace 1−/−; claudin k:GFP). The claudin k:GFP transgene highlights oligodendrocytes and Schwann cells, the differentiation of which can be followed in living claudin k:GFP transgenic animals from 2 dpf onwards. Consistent with the mouse phenotype, bace1−/−; claudin k:GFP larvae display severely reduced myelination of Schwann cells of the PNS that ensheath the posterior lateral line nerve (PLLN), first observed at 3 dpf (Fig. 3a and b). No hypomyelination of these Schwann cells is detected in bace1+/; claudin k:GFP zebrafish larvae (Fig. 3b). Interestingly, oligodendrocytes, the glial cells of the CNS, myelinating and ensheathing the Mauthner axons, do not display a defect in myelination in Bace1 KO zebrafish larvae (Fig. 3a and b). Hypomyelination of PNS glial cells can still be observed at 5 dpf, because the myelin sheaths around the anterior lateral line neurons (ALLNs) are severely thinner in bace1−/−; claudin k:GFP when compared with claudin k:GFP larvae (Fig. 3c and d). Again no hypomyelination of these Schwann cells is detected in heterozygous bace1+/; claudin k:GFP zebrafish larvae (Fig. 3b). CNS oligodendrocytes myelinating the Mauthner axons of bace1−/−; claudin k:GFP are still indistinguishable form claudin k:GFP larvae at 5 dpf (data not shown). In summary, hypomyelination in bace1−/− is restricted to the PNS and is not observed in the CNS.

Figure 2.

Generation and analysis of bace1 loss of function alleles. (a) Schematic representation of wildtype zebrafish Bace1. White stars: protease-active sites (DTGS and DSGT); black box: transmembrane domain (TMD). Lower panel: schematic representation of the predicted genomic organization of bace1 around the start codon. Red arrows indicate genomic localization of the target sequence of the ZFN set used. (b) Scheme of translated mutant bace1 alleles with respective allele designation (allele number specifies the last nucleotide of the newly introduced stop codon). (c) Western blot analysis with BACE1-specific monoclonal antibody 3D5 shows no protein expression of zBace1 in homozygous mutation carriers. Upper panel: antibody 3D5 detects the 64 kDa endogenous zBace1 protein in total lysates from wildtype larvae at 3 and 5 dpf but not from age-matched bace1mde81/mde96 mutants. The asterisk marks an unspecific band. Lower panel: anti-α-Tubulin serves as a loading control.

Figure 3.

Myelination in the PNS is delayed in bace1−/− mutants. (a) Schematic representation of myelination of Mauthner axons (CNS, dotted arrows) and lateral line nerves (PNS, arrows) in zebrafish larvae as visualized by claudin k:GFP. (b) Dorsal views of claudin k:GFP, bace1+/−; claudin k:GFP, bace1−/−; claudin k:GFP, and Nrg1 type III KD; claudin k:GFP larvae at 3 dpf expressing the claudin K:GFP transgene visualizing myelinating oligodendrodytes (CNS, arrowheads) and Schwann cells (PNS, arrows). At 3 dpf myelination of the posterior lateral line nerves is severely reduced in bace1−/−; claudin k:GFP and absent in Nrg1 type III KD; claudin k:GFP larvae. (c) Myelination of anterior lateral line nerves (PNS, arrows) in zebrafish larvae, lateral view. (d) Lateral views of the head region of claudin k:GFP, bace1+/−; claudin k:GFP, bace1−/−; claudin k:GFP, and Nrg1 type III KD; claudin k:GFP larvae at 5 dpf. Compared with claudin k:GFP larvae, myelination of anterior lateral line nerves (arrows) is severely reduced in bace1−/−; claudin k:GFP, and absent in Nrg1 type III KD; claudin k:GFP larvae at 5 dpf. Scale bars in panels (b) and (d) represent 100 μm.

NRG1 type III has been identified as a physiological substrate of mouse BACE1 and processing of NRG1 type III by mBACE1 is essential for normal myelination (Hu et al. 2006, 2008; Willem et al. 2006). We therefore asked if functional requirement of BACE1 for NRG1 type III mediated myelination is conserved in zebrafish. A gripNA-mediated KD of Nrg1 type III (Nrg1 type III KD) in zebrafish resulted in hypomyelination of the PNS (Fig. 3b and d). The myelin sheath around the ALLNs in bace1−/−; claudin k:GFP is only reduced compared with claudin k:GFP larvae, whereas there is hardly any myelin around the ALLNs seen in age-matched Nrg1 type III KD; claudin k:GFP larvae (Fig. 3d). Thus hypomyelination occurs in both, the bace1 mutant as well as the Nrg1 type III KD, but the hypomyelination phenotype of Nrg1 type III KD was more severe than in bace1−/− larvae, especially at 5 dpf (Fig. 3d). This is in agreement with the finding that BACE1 KO mice display hypomyelination of similar strength as Nrg1 type III heterozygous mice (Michailov et al. 2004; Taveggia et al. 2005). Recently and in addition to our finding, a nrg1 type III loss of function mutant was identified in a screen for zebrafish mutants with an increased number of mechanosensory neuromasts (Perlin et al. 2011). To further substantiate the phenotypic similarities between bace1 and nrg1 type III mutants, we analyzed the number of neuromasts in bace1−/− larvae. bace1−/− mutants display supernumerary neuromasts (Fig. 4a and b), showing striking similarities to the reported nrg1 type III−/− phenotype at 5 dpf (Perlin et al. 2011).

Figure 4.

Lateral line glia are not affected in bace1−/− zebrafish. (a) Schematic representation of a foxd3: GFP (green) larvae labeled with N-(3-triethylammoniumpropyl)-4-(6-(4-diethylamino)phenyl)hexatrienyl)pyridinium dibromide (FM4-64; red) to reveal neuromasts at 5 dpf, anterior to the left, lateral view. (b) foxd3:GFP-positive lateral line glia are present in bace1−/− larvae (lower panel) and no obvious difference glial cells ensheathing the lateral line nerve can be detected (dotted arrow). Loss of zBace1 leads to a supernumerary neuromast phenotype at 5 dpf in bace1−/− larvae (lower panel). Scale bar in panel (b) represents 100 μm.

In contrast to the hypomyelination of peripheral nerves, CNS derived oligodendrocytes ensheathing the Mauthner axons are normally myelinated in Nrg1 type III KD; claudin k:GFP larvae similar to our findings in bace1−/−; claudin k:GFP (Fig. 3b). To elucidate the importance of zBace1 for myelination further, we asked if the observed hypomyelination of bace1−/− could be a secondary effect caused by abnormal Schwann cell development. We therefore analyzed Schwann cells using a transgenic line expressing GFP under the control of the foxd3 promoter (foxd3:GFP), which is described to label lateral line associated glial cells from 2 dpf onwards (Gilmour et al. 2002). No difference in GFP-positive glial cells can be detected comparing wildtype; foxd3:GFP and bace1−/−; foxd3:GFP zebrafish at 2 and 3 dpf (data not shown). By 5 dpf, bace1−/−; foxd3:GFP larvae with expression of GFP in the lateral line glia accompanying the PLLN are still indistinguishable from foxd3:GFP larvae (Fig. 4b). Moreover, GFP-positive Schwann cells are detected in foxd3:GFP as well as in bace1−/−; foxd3:GFP larvae at all locations where neuromasts are observed (Fig. 4b). Thus, Schwann cells are present but fail to myelinate. In line with these findings, a tap assay did not reveal any obvious differences in the escape response of bace1−/− larvae compared with wildtype siblings indicating that the lateral line sensory organ is fully functional and initiates the appropriate escape response (data not shown). In conclusion, the hypomyelination defect observed in bace1−/− larvae is not caused by the absence or mismigration of PNS glial cells but rather by reduced BACE1-dependent Nrg1 type III signaling.

To extend our Bace functional studies, we next investigated the physiological function of Bace2 (zBace2) in zebrafish. We identified one hBACE2 ortholog in the zebrafish genome, coding for a 503 aa protein. zBace2 is highly conserved to hBACE2 (Figure S1b). zbace2 mRNA is expressed throughout early development (Fig. 5a). To determine the physiological function of zBace2 in zebrafish, we analyzed a bace2 allele generated by ENU (N-ethyl-N-nitrosourea)-mutagenesis and identified by TILLING performed by the zebrafish mutation project (http://www.sanger.ac.uk/Projects/D_rerio/zmp/). Sequencing of ENU (N-ethyl-N-nitrosourea)-induced mutation carriers confirmed a C to A conversion in the bace2 resulting in a premature in-frame stop codon that leads to a severely truncated zBace2 protein lacking both protease active sites (Fig. 5b and c). Homozygous bace2 mutant larvae (bace2−/−) become distinguishable from their wildtype siblings at 3 dpf. Melanophores of bace2−/− larvae are more dilated compared with wildtype melanophores (Fig. 6). Moreover, the stereotypic migration paths of melanophores are disrupted in bace2−/−, which is best seen around the yolk sac extension and the tail fin (Fig. 6). zbace2 mRNA expression has been described in migrating neural crest cells by whole mount in situ hybridization (Thisse and Thisse 2004). The observed melanophore phenotpye in bace2−/− larvae is therefore consistent with the reported expression of zbace2 because melanophores are neural crest derivatives.

Figure 5.

zbace2 expression during development and bace2 mutant. (a) zbace2 mRNA expression determined by semi-quantitative RT-PCR (upper panel). Lower panel shows the expression of actin mRNA as a control. (b) bace2 mutants carry a C to A mutation that leads to an early stop and results in a truncated zBace2 protein. Sequencing reads of wildtype and bace2+/ zebrafish. (c) Schematic representation of wildtype and mutant zBace2 protein.

Figure 6.

Pigmentation phenotype of homozygous bace2 mutants. Pigment phenotype of bace2−/− at 3 dpf. All panels: lateral view, anterior to the left. Left panel: whole fish; middle and right panel: details of boxed regions around end of yolk sac extension and fin, respectively, where the difference in migration of melanophores and their shape is most obvious.

The loss of zBace1 and zBace2 leads to distinct phenotypes, namely hypomyelination and supernumery neuromasts versus abnormal pigmentation. In order to directly address if zBace1 and zBace2 confer redundant functions in zebrafish, we analyzed double homozygous bace1; bace2 zebrafish using a phenotype that can reliably be quantified. We determined the number of neuromasts in bace1 and bace2 single mutants as well as double homozygous bace1; bace2 larvae in comparison to wildtype larvae (Fig. 7a and b). While the number of neuromasts of bace2−/− larvae is similar to wildtype controls, bace1−/− larvae display a significantly higher number of neuromasts than age-matched wildtype siblings at late 5 dpf (Figs 4b, 7b and 7c). Next, we analyzed Nrg1 type III KD larvae. Compared with bace1−/− larvae, the number of neuromasts is even further increased in Nrg1 type III KD larvae indicating that a bace1−/− neuromast phenotype can be enhanced and hence a synergistic function of zBace2 or other enzymes could potentially be uncovered. We used this quantifiable neuromast phenotype to determine, whether loss of zBace2 enhances the zBace1 phenotype, which would be indicating a redundant function. Double homozygous bace1; bace2 mutants are indistinguishable from bace1−/− larvae regarding neuromast numbers (Fig. 7b and c). We also addressed the question, whether the loss of zBace1 can reciprocally further increase the pigment phenotype caused by loss of zBace2. Double homozygous bace1; bace2 mutants have an indistinguishable pigment phenotype compared with bace2−/− larvae (Fig. 7d). In summary, none of the phenotypes observed in the single homozygous mutants is more pronounced in bace1−/−; bace2−/− zebrafish, supporting that the two genes convey non-redundant functions in zebrafish.

Figure 7.

bace1 and bace2 double homozygous mutants reveal no redundancy regarding neuromast numbers and pigmentation. (a) Schematic representation of the morphology of the lateral line system of zebrafish larvae modified after (Van Trump and McHenry 2008), schematic view from dorsal, anterior to the left. Distribution of neuromasts along the body axis. (b) Dorsal views of wt, bace1−/−, bace2−/−, bace1−/−; bace2−/−, and Nrg1 type III KD larvae at 5 dpf stained with N-(3-triethylammoniumpropyl)-4-(6-(4-diethylamino)phenyl)hexatrienyl)pyridinium dibromide (FM4-64) to visualize neuromasts. While wt and bace2−/− larvae have the same amount of trunk neuromasts, bace1−/− larvae have an elevated number of trunk neuromasts, the amount of cranial neuromasts is not altered. Additional loss of zBace2 protein does not further increase the amount of neuromasts in bace1−/−, whereas Nrg1 type III KD leads to a further increase in neuromasts compared with bace1−/−. Scale bar represents 100 μm. (c) Quantification of trunk neuromasts (n = 100; except for bace1−/−; bace2−/−:= 10). (d) Tail fins of wt, bace1−/−, bace2−/− and bace1−/−; bace2−/− at 3 dpf, lateral view.


The type I transmembrane aspartyl proteases BACE1 and BACE2 are highly conserved within vertebrates and have not been identified in invertebrates thus far, suggesting a vertebrate specific function. This hypothesis is in accordance with the proposed functions of substrates cleaved by BACE1 in the immune system (Kitazume et al. 2001; Lichtenthaler et al. 2003) and in myelination (Hu et al. 2006; Willem et al. 2006). In zebrafish, we identified and functionally characterized the highly conserved zBace1 and zBace2 proteins. Given the advantages of the zebrafish system such as ease of manipulation and power of genetics the zebrafish is a very powerful tool to study the physiological function of Bace proteases. We generated bace1−/− mutant zebrafish that lack zBace1 protein and show that zBace1 is necessary in vivo to stimulate myelination, a physiological requirement that is conserved from fish to mice (Hu et al. 2006; Willem et al. 2006). bace1−/− mutants as well as Nrg1 type III KD larvae share the PNS-specific hypomyelination phenotype. Myelin sheaths around PLLNs and ALLNs of bace1−/−; claudin k:GFP are thinner but not absent at 5 dpf when compared with claudin k:GFP controls showing that loss of zBace1 delays the initiation of myelination rather than leading to a complete block in the myelin formation consistent with previous findings in murine models (Hu et al. 2006; Willem et al. 2006). As there are controversial reports regarding the role of mBACE1 in myelination of the CNS (Hu et al. 2006; Willem et al. 2006; Brinkmann et al. 2008) we also analyzed myelination of glial cells of the CNS. In contrast to the observed hypomyelination phenotype of peripheral nerves in bace1−/−; claudin k:GFP zebrafish, the initiation and degree of myelination of glial cells ensheathing CNS Mauthner axons are indistinguishable from claudin k:GFP at all stages analyzed thus demonstrating that CNS myelination is not affected. Regarding the major discrepancy in the field concerning the necessity of BACE1 for CNS myelination our data clarifies that the process of CNS myelination is independent of zBace1.

The degree of hypomyelination and the increase in neuromast number is more pronounced in Nrg1 type III KD; claudin k:GFP when compared with bace1−/−; claudin k:GFP larvae. We conclude that zBace1 is not the only protease that processes zNrg1 type III to gain a signaling competent molecule, and therefore only generates a partial nrg1 type III loss of function phenotype. Indeed, apart from BACE1, the proteases ADAM10 and ADAM17 have been reported to also cleave NRG1 type III (Falls 2003; Taveggia et al. 2005; Hu et al. 2006; Willem et al. 2006; La Marca et al. 2011; Velanac et al. 2012). To our knowledge our bace1−/− zebrafish is the first mutant described with an elevated number of neuromasts despite normal developing Schwann cells indicating that indeed impaired zNrg1 type III cleavage and signaling is responsible for the myelination phenotype. This hypothesis is further corroborated by the fact, that a soluble NRG1 type III β epidermal growth factor (EGF)-like domain is able to partially rescue the hypomyelination of bace1−/− zebrafish larvae (D. Fleck, under revision). In summary, the hypomyelination of the PNS observed in all BACE1 deficient animal models generated so far is the most prominent loss of function phenotype. This finding underlines the impact of Nrg1 type III as being a zBace1 substrate, which mediates physiological relevant processes comparable to the γ-secretase substrate Notch (Geling et al. 2002). If and to what extent this finding affects the development of BACE1 inhibitors remains to be shown. However, myelination occurs in higher vertebrates immediately after birth, thus treatment of adults should not affect myelination.

Until now, very little information is available regarding the function of BACE2. However, as therapeutic inhibition of BACE1 might as well affect BACE2 (Vassar 2001, 2002; Citron 2002a,b; Stachel et al. 2004), knowledge of the physiological requirement of both proteases is crucial. Zebrafish carrying mutations in zbace2 are characterized by the mismigration of melanophores. At first, the abnormal migration pattern of melanophores resembles a described zebrafish pigment mutant, parade (Kelsh et al. 1996). Unlike in parade, the migration of another pigment cell type of zebrafish, the iridiophores, is not abnormal in bace2−/− mutants (data not shown) suggesting a different molecular target to be affected in bace2−/− mutants. The underlying molecular mechanism of the pigment phenotype is still elusive and requires further investigation. The abnormal distribution of melanophores observed in bace2−/− zebrafish may be consistent with an abnormal coat color of BACE2 KO mice that is greyish instead of black (D. Dominguez and B. De Strooper, personal communication) again supporting the conserved function of BACE proteases among vertebrates.

At last, the zebrafish bace1−/−; bace2−/− double mutant allowed us to address zBace1 and zBace2 redundancy in vivo. Using two different read-outs, the neuromast and the pigment phenotype, we determined that neither the additional loss of zBace2 did further enhance the loss of function phenotype of bace1−/− mutant larvae, nor did the loss of zBace1 in bace2−/− mutants increase the loss of function phenotype observed in single bace2−/− zebrafish larvae. Both results demonstrate that zBace1 and zBace2 are proteases with fundamentally different physiological functions.

To evaluate the relevance of our zBace1 and zBace2 studies in zebrafish, it is important to consider the role of BACE1 in AD patients. BACE1 cleaves APP, which is consecutively cleaved by γ-secretase to generate Aβ. Aβ is, according to the Aβ hypothesis, the main component of amyloid plaques and the disease initiating neurotoxic agent in AD (Hardy and Selkoe 2002; Hardy 2009). Generation of Aβ depends on cleavage of APP by BACE1 followed by γ-secretase. For both proteases, additional substrates aside APP are reported but as none of the BACE1 substrates seem to be essential the development of BACE1 inhibitors is still a promising therapeutic avenue. Our results confirm that in vertebrates BACE1 inhibition is not detrimental because zebrafish Bace1 KO are also viable and fertile. The hypomyelination phenotype of the bace1−/− mutants can be used to screen for specific BACE1 inhibitors in vivo, as it constitutes an easy and affordable functional read-out. Moreover the pigment phenotype of the bace2−/− mutants allows to determine in vivo if inhibitors designed against BACE1 also inhibit BACE2. Thus, we have developed the tools to analyze BACE inhibitors in vivo important for evaluating therapeutic intervention in the course of AD.


We thank Katrin Strecker, Mathias Teucke, Stefanie Topp, and Rielana Wiechert for excellent technical support and Daniela Würzinger and Roberto Rojas Rojas for fish care. We thank the Hubrecht laboratory for providing the zebrafish KO allele bace2hu3332 and the Sanger Institute Zebrafish Mutation Resource (ZF-MODELS Integrated Project; contract number LSHG-CT-2003–503496; funded by the European Commission) also sponsored by the Welcome Trust [grant number WT 077047/Z/05/Z]. We thank Robert Vassar for the antibody 3D5. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 596 to CH & BS and to CH & MW) and the NGFNplus (to CH and MW). The research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement n°321366-Amyloid. We thank the Hans & Ilse Breuer Foundation for the confocal microscope, and Sophie Rothhämel for critical discussion. The authors declare no conflict of interest.