SEARCH

SEARCH BY CITATION

Keywords:

  • CRISPR/Cas (clustered regularly interspaced short palindromic repeats);
  • genomics;
  • neurodegeneration;
  • TALEN;
  • zebrafish

Abstract

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References

Zebrafish has become a popular model organism to study human diseases. We will highlight the advantages and limitations of zebrafish as a model organism to study neurodegenerative diseases and introduce zinc finger nucleases, transcription activator-like effector nucleases, and the recently established clustered regularly interspaced short palindromic repeats/clustered regularly interspaced short palindromic repeat-associated system for genome editing. The efficiency of the novel genome editing tools now greatly facilitates knock-out and, importantly, also makes knock-in approaches feasible in zebrafish. Genome editing in zebrafish avoids unspecific phenotypes caused by off-target effects and toxicity as frequently seen in conventional knock-down approaches.


Abbreviations used
ALS

amyotrophic lateral sclerosis

CRISPR/Cas9

clustered regularly interspaced short palindromic repeats/CRISPR-associated 9

DPRs

dipeptide repeats

FTLD

frontotemporal lobar degeneration

KI

knock-in

KO

knock-out

MO

morpholinos

NHEJ

non-homologous end joining

TALEN

transcription activator-like effector nucleases

ZFN

zinc finger nucleases

Zebrafish as an animal model for neurodegeneration

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References

The small size, ease of breeding, and amenability for manipulations have made the zebrafish a very popular research model organism. Zebrafish have several advantages: they have a diploid fully sequenced genome (Howe et al. 2013), which makes them suitable for genetic studies. Large-scale mutagenesis screens have generated hundreds of zebrafish mutants (Driever et al. 1996; Haffter et al. 1996). Currently, mutations in all of the 26 000 genes are systematically isolated and will generate a valuable tool for functional gene analysis in the near future (Kettleborough et al. 2013). Embryos are transparent and therefore ideal for in vivo imaging. Furthermore, the embryos are small enough to perform high-content chemical screens in a 96-well format. Zebrafish are vertebrates and share many anatomical, physiological, and biochemical properties with humans. Therefore, studies of vascular diseases, inflammatory diseases, myelin-specific diseases, or bone disease in zebrafish are feasible, conditions for which invertebrates such as Drosophila melanogaster or Caenorabditis elegans do not offer morphological correlates (Lieschke and Currie 2007). One compound, for example, was identified in an unbiased forward chemical screen for its potential to expand hematopoetic stem cells in zebrafish. This compound is now used in clinical trials to treat patients after irradiation or chemotherapy for an accelerated repopulation of hematopoetic stem cells (North et al. 2007). Such a screen would have not been possible in invertebrates lacking a hematopoetic system. Increased attention has therefore been paid to the zebrafish as a model organism for biomedical research, despite some limitations. Many of the advantages of zebrafish can only be exploited in early stages of development, such as transparency for in vivo imaging or the ability to perform chemical screens. Disease-associated phenotypes are therefore preferably analyzed in embryonic or larval stages. In addition, phenotypes are more easily identified in early developmental stages. Inhibition of the Notch signaling pathway, for example, can be easily identified by disturbed somite boundaries in development (Geling et al. 2002). However, most of the neurodegenerative diseases have a late onset and disease phenotypes might only be recapitulated in aged animals. Amyloid plaques as seen in the brains of Alzheimer's disease patients cannot be observed in larvae and embryos (van Bebber, Schmid, Haass unpublished). At present, it is not even clear if amyloid plaques can be generated at all in aged zebrafish brains. Zebrafish also have a tremendous capacity for regeneration, such that upon spinal cord lesion a fully paralyzed adult fish can again swim normally after only a few weeks (Becker et al. 1997). Adult neurogenesis is also much more abundant than in mammals (Kizil et al. 2012) and consequently precludes the quantification of neuronal cell loss in aged zebrafish brains. Finally, a major drawback of the zebrafish as a model for human diseases is the inefficient technology for gene knock-out (KO) and knock-in (KI). However, major progress has recently been made by the introduction of genome editing techniques into the zebrafish field. In this review, we will discuss these novel techniques based on some recently described zebrafish models for frontotemporal lobar degeneration (FTLD) and amyotrophic lateral sclerosis (ALS).

Genome editing

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References

For many years, the ability to edit a genome was restricted to the mouse community as embryonic stem cell technologies pertaining to homologous recombination was established only in mice. In zebrafish, prior to the availability of genome editing tools, loss of protein function was only temporarily achieved by targeted knockdown (KD) approaches with anti-sense morpholinos (MO) (Eisen and Smith 2008) or anti-sense gripNAs, or alternatively generated by laborious N-ethyl-N-nitrosourea or retroviral mutagenesis screens (Driever et al. 1996; Gaiano et al. 1996; Haffter et al. 1996; Wang et al. 2007; Amsterdam et al. 2011). KD approaches can only temporarily block either translation or splicing and frequently suffer from off-target effects that need to be tightly controlled (Bill et al. 2009). Stable mutations in the genome can circumvent these problems. The way to generate such mutations has dramatically improved with the discovery of genome editing tools, such as zinc finger nucleases (ZFN), transcription activator-like effector nucleases (TALEN), and most recently the clustered regularly interspaced short palindromic repeats/CRISPR-associated 9 (CRISPR/Cas9) system, which now makes genome editing feasible in almost any organism (Gaj et al. 2013).

All genome-editing tools are composed of a sequence-specific DNA targeting subunit and a double-strand DNA cleaving nuclease. The induction of double-strand breaks in the genome initiates two DNA repair pathways, non-homologous end joining (NHEJ) and homology-directed repair (Fig. 1a). NHEJ can ligate the broken DNA strands without a template, and therefore frequently introduces small insertions and deletions (indels). KO animals can be identified by searching for indels generated by incorrect NHEJ repair in the open reading frame of a gene leading to a frame shift or a premature stop codon. Homology-directed repair, on the other hand, repairs double-strand breaks using a homologous template DNA. This can be exploited to introduce foreign DNA at a desired locus through the insertion of flanking homology arms (Fig. 1a) (Zu et al. 2013).

image

Figure 1. Genome editing overview. (a) Schematic representation of genome editing to generate a knock-out and knock-in. First, a nuclease is directed to a target sequence by a specific DNA binding element. The nuclease induces a double strand break that can either be repaired by non-homologous end joining or homology-directed repair. Non-homologous end joining is an error prone repair pathway that often generates insertions or deletions. Indels that lead to a reading frame shift or induce an early stop codon are selected to establish knock-out alleles. The homology directed repair pathway employs a homologous template to repair the double strand break. Integration of foreign DNA at the breakpoint to generate knock-ins can be achieved by providing a donor sequence with the foreign DNA flanked by homologous sequence. (b) Schematic representation of zinc finger nucleases (ZFN). Left and right ZFN bring together the heteromeric Fok1 nuclease to induce a double strand break. Arrays of 3–4 zinc fingers are used, with each zinc finger binding three nucleotides. (c) Schematic representation of transcription activator-like effector nucleases (TALEN). Left and right TALEN generate a heteromeric Fok1 holoprotein to induce a double strand break. Arrays of around 16 TALE elements are used, each TALE recognizing one nucleotide. (d) Schematic representation of the clustered regularly interspaced short palindromic repeats/CRISPR-associated 9 (CRISPR/Cas9) system. The guide RNA (gRNA) is composed of a 20 nucleotide long targeting sequence and a sequence that recruits the Cas9 nuclease. Cas9 (green) generated the double strand break close to the protospacer adjacent motif (PAM) sequence (NGG).

Download figure to PowerPoint

In the following, we will introduce the three techniques most frequently used for genome editing in zebrafish (for more detailed information on engineered endonucleases as searchable database, see http://eendb.zfgenetics.org/een.php).

Zink finger nuclease

ZFN contain zinc finger DNA-binding motifs that target DNA attached to a Fok1 endonuclease subunit (Fig. 1b). To reduce off-target effects, the FOK1 nuclease can be designed as a heterodimer that only cleaves DNA if the left and the right zinc fingers position the two Fok1 nuclease subunits appropriately to generate an active enzyme (Fig. 1b). Each zinc finger moiety binds to three nucleotides and left and right zinc finger DNA-binding motifs are usually composed of arrays of three to four zinc fingers each. The zinc fingers have to be selected and assembled from a library of zinc finger modules for highly efficient target binding. The design of the zinc fingers is quite challenging as it is difficult to predict their binding specificity. In addition, each zinc finger moiety has a different binding affinity to the target sequence depending on the neighboring zinc finger moiety, a feature referred to as context dependence (Sander et al. 2011). An in vitro assay to assess binding specificity of ZFN is therefore mandatory. ZFNs were the first genome-editing tool used in zebrafish (Doyon et al. 2008; Meng et al. 2008) and show variable efficiencies. These differences were mainly because of different strategies of zinc finger selection as well as different in vitro assays (Sood et al. 2013; Xiao et al. 2013b). Laborious design and in vitro tests, as well as high costs if purchased, precluded a breakthrough of this technology in zebrafish.

Transcription activator-like effector nuclease

The bacterial plant pathogen Xanthomonas has the ability to activate transcription in the host cell by transcription activator-like effector (TALE) proteins. The transcription of specific host genes is initiated by TALE to facilitate bacterial infection (Bogdanove et al. 2010). The discovery of the DNA-binding elements from TALE proteins of Xanthomonas was a breakthrough as they have a predictable target binding affinity in contrast to ZFN. The code of the TALE DNA-binding domain has been elucidated and shown to include four distinct repeats, each being able to recognize one specific nucleotide of the DNA. Each repeat is composed of 33–35 conserved amino acids with a variable region at positions 12 and 13. The variable region of the repeat either consists of the amino acids NI, HD, NG, or NN specifically targeting the nucleotides A, C, T, and G (A), respectively (Boch et al. 2009; Moscou and Bogdanove 2009). In contrast to ZFN, each TALE repeat will always recognize its corresponding nucleotide. Therefore, TALE are easy to design. The only sequence preference is the nucleotide T at the 5′ target site (Moscou and Bogdanove 2009). For genome editing purposes, sets of TALEN composed of around 16 TALE repeats are fused to Fok1 nuclease heterodimeric subunits to induce a double-strand break (Fig. 1c). The cloning of the individual TALEN repeats can be time consuming, but can be bypassed by gene synthesis. Furthermore, modular assembly strategies facilitate the assembly of TALENs (Cermak et al. 2011; Kim et al. 2013). A high throughput assembly has even generated TALEN targeting 18 740 human genes, a powerful resource for basic and biomedical research (Kim et al. 2013). Off-target effects of TALEN are rare (Ding et al. 2013; Kim et al. 2013) as the DNA-binding specificity of the TALEN is very high and the target sequence is around 16 nucleotides for each of the left and right TALEN. Because of their very high selectivity, TALENs are currently the genome-editing tool of choice.

Clustered regularly interspaced short palindromic repeats/CRISPR-associated 9

While both ZFN and TALEN use DNA-binding proteins to target DNA and are connected to the Fok1 endonuclease to cleave the DNA, the CRISPR/Cas9 system from bacteria and archea is built on a different concept (Fig. 1d). Here, a RNA strand – not a DNA-binding protein – guides the Cas9 nuclease to the desired genomic locus, and is therefore called guide RNA (gRNA). DNA targeting by gRNA is based on Watson–Crick base pairing and is therefore highly specific and predictable. The gRNA is transcribed from the CRISPR locus that contains short palindromic repeats separated by foreign DNA fragments. This foreign DNA can be from a phage that infected bacteria or plasmid DNA. Upon infection, bacteria can incorporate parts of the invading DNA into the CRISPR locus for subsequent targeting and inactivation of the invading pathogen. This makes the CRISPR/Cas9 system comparable to the innate immune system (Wiedenheft et al. 2012). The nuclease required for inactivation is also encoded by the bacterial genome from a gene that is associated with the CRISPR locus and hence called CRISPR-associated (Cas). Prokaryotes are very diverse and numerous CRISPR/Cas strategies have emerged during evolution, the type II CRISPR/Cas9 system being the most frequently used tool for genome editing (Makarova et al. 2011). The gRNA and Cas9 of the type II CRISPR/Cas9 system have been further adopted for use in eukaryotes. For example, a nuclear localization sequence has been introduced into the Cas9 protein to enhance nuclear transport (Jinek et al. 2012; Cong et al. 2013; Mali et al. 2013). For genome editing in zebrafish, the gRNA can simply be transcribed from a plasmid or transcribed from an oligo and coinjected with Cas9 mRNA to induce double-strand cleavage and mutations (Chang et al. 2013; Hwang et al. 2013; Xiao et al. 2013a). The only sequence requirement for the gRNA target site is a three-nucleotide NGG motif adjacent to the 20-nucleotide target sequence called protospacer adjacent motif (Fig. 1d). Off-target effects might occur more frequently with CRISPR compared with ZFN and TALEN as a much shorter targeting sequence is used. However, outcrossing of edited fish lines can eliminate unwanted side effects caused by off-target genomic modifications. The efficiency to generate somatic mutations in zebrafish is on average 30% with the CRISPR/Cas9 system and comparable to TALEN (Blackburn et al. 2013). However, CRISPR/Cas9 is not as cost-intensive as any of the other editing methods described above and technically much easier, since time-consuming cloning steps or in vitro tests to determine cleavage ability, as required for ZFN, can be avoided. This novel technology will have a tremendous impact on the zebrafish community since the generation of mutants can now be done easily in any laboratory.

Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References

Recent advances in frontotemporal lobar degeneration and amyotrophic lateral sclerosis

FTLD is the second most common form of dementia in patients younger than 65 years. FTLD leads to neuronal cell loss primarily in the frontal and temporal lobes of the brain and is clinically associated with changes in social behavior, changes in personality, and dementia, often associated with language dysfunction (Galimberti and Scarpini 2012). FTLD is a heterogeneous disease with distinct pathological entities [reviewed in (Rademakers et al. 2012)]. Most of the cases are characterized by cytosolic aggregates of the Tar DNA-binding protein of 43 kDa (TARDBP, TDP-43) (Neumann et al. 2006). TDP-43 is a predominantly nuclear RNA/DNA-binding protein that shuttles between the cytoplasm and the nucleus. Nuclear TDP-43 is cleared upon aggregation, suggesting a loss of nuclear function. Interestingly, TDP-43 aggregates are also observed in numerous ALS patients. ALS is characterized by selective loss of the upper or lower motor neurons, leading to muscle atrophy and death most often because of respiratory failure [reviewed in (Robberecht and Philips 2013)]. FTLD and ALS are now believed to be the extreme ends of a disease spectrum based on pathological, genetic, and clinical findings (Neumann et al. 2006). Clinically, many patients with classical ALS have additional signs of FTLD and patients with FTLD often present with motor symptoms. Furthermore, mutations in several genes, such as TDP-43, or the recently discovered hexanucleotide repeat expansion in the C9ORF72 gene can either lead to FTLD or ALS (Morris et al. 2012).

In the following, we will describe the use of genome editing to model FTLD and ALS in zebrafish. The identification of many novel genes associated with FTLD and ALS has given us an exciting opportunity to study the molecular mechanisms of the disease spectrum.

Tar-DNA binding protein of 43 kDa

TDP-43 is a splicing factor and regulates various aspects of RNA metabolism, and its levels are tightly regulated through a regulatory feedback loop (Avendano-Vazquez et al. 2012). Thousands of RNA targets of TDP-43 have been identified by cross-linking and immunoprecipitation (Polymenidou et al. 2011; Tollervey et al. 2011; Xiao et al. 2011); however, the physiological function of TDP-43 is still unclear. FTLD- and ALS-associated mutations in TDP-43 cluster in the C-terminal glycine-rich domain (Fig. 2a). It has been suggested that they increase the propensity to aggregate as nuclear TDP-43 is mainly found in cytosolic aggregates (Johnson et al. 2009). Alternatively, mutations may cause a partial loss of function as the nucleus becomes cleared of TDP-43 upon cytosolic aggregation (Neumann et al. 2006). Interestingly, the zebrafish genome harbors two TDP-43 ortholgues, Tardbp and Tardbp-like (Tardbpl). The second orthologue, Tardbpl, lacks a glycin-rich domain (Fig. 2a). Potential gain and loss of function of TDP-43 has been investigated in zebrafish. Injection of wildtype and mutant human TDP-43 mRNA causes shorter and aberrantly branched primary spinal motor neuron axons in zebrafish (Kabashi et al. 2010; Laird et al. 2010). The motor neuron axon phenotype is more pronounced upon injection of mutant compared with wildtype human TDP-43 in zebrafish (Kabashi et al. 2010). Interestingly, MO-mediated KD of zebrafish Tardbp has been reported to cause a similar motor neuron axon phenotype, leading the authors to speculate that ALS is generated by a combined loss and gain of function mechanism (Kabashi et al. 2010). However, two independent studies have subsequently demonstrated that loss of Tardbp in zebrafish is fully compensated by alternative splicing of Tardbpl (Fig. 2b) and therefore Tardbp mutants do not show any spinal motor neuron axon phenotype. Under wildtype conditions, Tardbpl is mainly expressed as a protein, lacking a C-terminal glycine-rich domain. Upon loss of Tardbp, a novel Tardbpl splice variant (Tardbpl_tv1) with a glycine-rich domain, highly homologous to that of Tardbp, is generated. (Hewamadduma et al. 2013; Schmid et al. 2013) (Fig. 2b). Tardbpl_tv1 can fully compensate for the loss of Tardbp because of the presence of the glycine-rich domain, which is known to be important for interaction with other hnRNPs and splicing (Buratti et al. 2005). Consistent with this finding, the reported Tardbp KD-induced shorter and hyperbranched motor neurons have not been observed in several lines of stable Tardbp mutants generated by genome editing (Schmid et al. 2013). Thus, MO KD phenotypes of Tardbp have to be interpreted with caution, as they may be the result of unspecific off-target effects or the MO injection procedure itself. In contrast to the single mutants, the double homozygous Tardbp and Tardbpl mutants are lethal, characterized by muscle degeneration and shorter spinal motor neuron axon phenotype (Schmid et al. 2013) (Fig. 2c). In addition, they entail mis-patterning of the vasculature and strongly reduced blood circulation (Fig. 2c). A quantitative proteomics analysis identified the muscle-specific actin-binding protein Filamin Ca as a twofold up-regulated protein in double homozygous Tardbp/Tardbpl mutant embryos, whereas all the other identified muscle proteins were down-regulated because of muscle degeneration. This alteration in Filamin C expression could be confirmed in frontal cortex of FTLD patients with TDP-43 pathology. Interestingly, in brain tissues, Filamin C is expressed in vascular smooth muscle cells. This correlation suggests a loss of the function component in FTLD-TDP cases, which may result in vascular symptoms (Schmid et al. 2013) such as disturbed blood flow. Vascular dysfunction might therefore precede neuronal loss in ALS and FTLD. Interestingly, in mice expressing ALS-associated mutant superoxide dismutase 1 (SOD1), vascular malfunction is observed prior to neurodegeneration (Zhong et al. 2008). However, whether vascular dysfunction indeed contributes to the human disease remains to be proved. If this turns out to be the case, the zebrafish would provide an ideal model to study the disturbed signaling pathways responsible for reduced blood flow and identify potential therapeutic targets.

image

Figure 2. The Tar DNA-binding protein (TARDBP) orthologues in zebrafish, Tardbp and Tardbpl, and their double homozygous loss of function phenotype. (a) Schematic representation of human TARDBP with the RNA recognition motif 1 (RRM1), the RNA recognition motif 2 (RRM2) and the glycine-rich domain where frontotemporal lobar degeneration (FTLD) and amyotrophic lateral sclerosis (ALS) associated mutation cluster. (b) Schematic representation of the exon/intron structure of Tardbpl. Upon loss of Tardbp, the tardbpl exon 5 splice donor site is not recognized, and intron 5–6 is transcribed. Translation of this novel transcript variant 1 (Tardbpl_tv1) generates a protein with a glycine rich domain, which is functionally redundant to Tardbpl. (c) Combined loss of Tardbp and Tardbpl leads to vascular mispatterning, shorter spinal motor neuron axons, and muscle degeneration (data from (Schmid et al. 2013)).

Download figure to PowerPoint

Fused in sarcoma

Fused in sarcoma (FUS, TLS) is a nuclear RNA/DNA-binding protein similar to TDP-43 (Dormann and Haass 2013) (Fig. 3A). FUS mutations predominantly cause ALS and are characterized by FUS inclusions. Most mutations affect the nuclear localization signal (NLS) of FUS and result in a mislocalization to the cytoplasm (Bosco et al. 2010; Dormann et al. 2010; Dormann and Haass 2013). Expression of mutant FUS in zebrafish also results in cytosolic mislocalization (Bosco et al. 2010; Dormann et al. 2010) indicating evolutionarily conserved nuclear localization mechanisms. The levels of mislocalized Fus protein inversely correlate with onset of disease. It is currently unclear if a nuclear loss-of-function mechanisms is associated with disease (Dormann et al. 2010) or if mislocalized FUS is toxic because of a cytosolic gain of function. Perhaps both mechanisms act in parallel.

image

Figure 3. Zebrafish orthologues of FUS and GRANULIN and hexanucleotide repeat expansion with associated dipeptide repeat (DPR) translation. (a) Schematic representation of human FUS and zebrafish FUS protein. The zebrafish mutant Fus F500X was generated by zinc finger nucleases (ZFN)-mediated genome editing and deletes the C-terminal NLS (yellow). Motor neurons in these mutants are not altered (shown by antibody staining of motor neuron axons by znp-1) (unpublished data). (b) Schematic representation of human GRANULIN and the four zebrafish orthologues. The individual granulin domains are labeled by letters in human GRANULIN and numbers in the zebrafish orthologues. (c) Schematic representation of the human genomic locus of the C9orf72 gene and the GGGGCC hexanucleotide repeat expansion. (d) the hexaucleotide repeat expansion can be translated in the three reading frames (GA)n, (GP)n, and (GR)n.

Download figure to PowerPoint

In zebrafish, there is only one homologue of FUS. Loss-of-function analysis of zebrafish Fus by MO-mediated KD results in shorter and hyperbranched primary spinal motor neuron axons (Kabashi et al. 2011). In an independent KD study, these phenotypes of Fus were not observed, despite successful depletion of zebrafish Fus protein (Hasenkamp, Schmid, Haass, unpublished). The discrepancy between these two studies will again have to be resolved in stably inherited Fus KO. Expression of mutant human FUS has been reported to result in shorter hyperbrached primary spinal motor neuron axons (Kabashi et al. 2011). As FUS patients do not have higher FUS levels but rather mislocalized and aggregated FUS, these studies only partially resemble pathological FUS. Using ZFN genome editing, a deletion of the C-terminally located nuclear localization sequence was generated at the zebrafish FUS genomic locus (Hasenkamp, Schmid, Haass, unpublished). This mutation resembles the ALS-associated FUS mutations Fus R495X (Bosco et al. 2010) with a premature stop codon and FUS G466VfsX14 (DeJesus-Hernandez et al. 2010), which skips exon 14 also resulting in a C-terminally truncated protein. The Fus R500X fish line thus genetically generates pathological FUS by deletion of the C-terminal nuclear localization sequence. Interestingly, the FusR500X mutants do not show aberrant spinal motor neuron length or branching (Fig. 3a) (Hasenkamp, Schmid, Haass, unpublished), raising the question how well over-expression of a mutant protein can mimic mislocalization of an endogenous protein. The Fus R500X fish is the first animal model that fully recapitulates pathogenic FUS without over-expression. It may serve as a valuable tool for the identification of disease modifying mechanisms and drugs to revert pathological FUS pathology such as mislocalization and aggregation.

Progranulin (GRANULIN, GRN)

The most common genetic cause of FTLD are heterozygous loss-of-function mutations in the GRANULIN (GRN) gene (Baker et al. 2006; Cruts et al. 2006). Interestingly, homozygous GRN mutations lead to Neuronal ceroid lipofuscinoses, a disease characterized by dysfunctional lysosomes (Smith et al. 2012). GRN is a secreted protein with growth factor-like activity involved in the immune response, wound healing, angiogenesis, and many other biological processes (Cenik et al. 2012). It has a modular structure and is composed of 7.5 highly conserved granulin domains in human (Fig. 3b). Proteolytic processing of full-length GRN leads to the liberation of the individual granulin domains, which have opposing activity compared with the full-length protein in inflammation (Cenik et al. 2012). In zebrafish, there are two orthologues of human GRN. Two of them, granulin a (grna) and granulin b (grnb), are paralogues and consist of 12 and 9 granulin domains, respectively (Cadieux et al. 2005) (Fig. 3 B). In addition, the zebrafish genome harbors two shorter granulins, granulin 1 (grn1) and granulin 2 (grn2), which are composed of only 1.5 granulin domains (Cadieux et al. 2005). It has been postulated that grna is the homologue of human GRN based on synteny (Cadieux et al. 2005). MO KD of zebrafish grna and grnb has been reported to cause shorter and hyperbranched spinal motor neuron axons (Chitramuthu et al. 2010; Laird et al. 2010). This phenotype was again not observed in stable single- and double-mutant grna and grnb mutants generated my genome editing (Solchenberger, Haass, Schmid, unpublished data). Furthermore, decreased liver morphogenesis and reduced pax7 positive myogenic progenitor cells have been reported in grna knock-down embryos; however, neither phenotype is correlated with FTLD pathology (Li et al. 2010, 2013). These phenotypes need to be validated in mutant embryos. Loss-of-function analysis of GRN is more straightforward in mouse models that have only one orthologue compared with the four GRN orthologues in zebrafish. However, we still have a very limited understanding of the different functions of full-length GRN versus processed granulins. We hypothesize that the function of the processed granulins in mammals may be represented by Grn1 and Grn2 in zebrafish. The zebrafish model therefore holds the unique potential to facilitate the study of full-length GRN versus processed GRN function, which is difficult to achieve in mammals with one GRN gene.

GGGGCC hexanucleotide repeat expansion in C9orf72

The most common genetic cause of FTLD and ALS is the recently identified GGGGCC hexanucleotid repeat expansion in a non-coding region of the C9orf72 gene (DeJesus-Hernandez et al. 2011; Renton et al. 2011; Gijselinck et al. 2012; Majounie et al. 2012) (Fig. 3c). C9orf72 is a highly conserved gene with unknown function. Possible pathomechanisms for this repeat expansion include haploinsufficency because of reduced C9orf72 transcript levels and sequestration of RNA-binding proteins such as hnRNPA3 (DeJesus-Hernandez et al. 2011; Gijselinck et al. 2012; Majounie et al. 2012; Mori et al. 2013a). In addition, two laboratories have independently identified non-ATG-mediated translation of the repeat into dipeptide repeats (DPRs) (Ash et al. 2013; Mori et al. 2013b). This unusual translation mechanism translates the hexanucleotide repeat expansion in the three reading frames in (GA)n, (GP)n, and (GR)n DPR in the absence of a start methionine (Fig. 3d). DPR proteins are specific to patients with a hexanucleotide repeat expansion and form disease signifying aggregates (Ash et al. 2013; Mori et al. 2013b).

The C9orf72 gene is highly conserved in zebrafish and is called C13H9orf72. KD of C13H9orf72 by MOs has been shown to cause shorter and hyperbranched spinal motor neuron axons. This is currently the first and only animal model addressing the function of C9orf72 (Ciura et al. 2013). It will be important to investigate the effects of C13H9orf72 loss of function in stable KOs for long-term effects and also validate the MO phenotype given the previously described discrepancies between KD and KO phenotypes for Tardbp, Fus, Grna, and Grnb. In addition, it will be required to express the hexanucleotide repeat expansion and DPRs to assess their contribution to neurodegenration. Possibly, a combination of reduced C9orf72 transcript levels, sequestration of RNA-binding proteins by the hexanucleotide repeat expansion and DPRs is responsible for C9orf72-mediated pathology.

Knock-down versus knock-out

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References

As highlighted in the previous examples of TDP-43, FUS, Grna, and Grnb, phenotypes elicited by MO KD are not reproduced in genetic mutants. Similar discrepancies have also been reported for other genes involved in neurodegeneration, such as the Fragile X Mental Retardation gene (Tucker et al. 2006; den Broeder et al. 2009). Thus, the MO phenotypes may have been elicited by off-target effects or the manipulation of the embryos. MOs are known to generate off-target effects that are difficult to control (Eisen and Smith 2008). Only if the MO KD can faithfully recapitulate a known mutant phenotype, it will prove to be a valuable additional tool. With the new genome-editing tools, the generation of mutants is now feasible at any lab, with the possibility of careful comparisons between KD and KO phenotypes avoiding misinterpretations because of unspecific KD effects. In addition, KD phenotypes often suffer from high variability because of the injection procedure and stability of the morpholino or gripNA. Subtle phenotypes and phenotypes that naturally suffer from big variations, e.g., behavioral assays, are therefore much easier to identify and quantify in KO versus KD.

Translational potential

Besides enhancing our understanding of gene function and pathological changes, zebrafish disease models are perfectly suited to identify drug targets that revert the disease phenotype. These can either be other genes that interfere with the phenotype or small chemical compounds. For chemical screens, the zebrafish larvae are extremely well-suited as the embryos are small enough to fit in a 96-well plate and grow in an aqueous environment where drugs can simply be added to the water. The most commonly observed phenotype in FTLD/ALS models are shorter and hyperbranched primary spinal motor neuron axons. For many of the KD models, these cannot be recapitulated in mutants and may be unspecific. However, in the case of the ALS-associated expression of mutant SOD1, this phenotype is well-documented and controlled (Ramesh et al. 2010; Sakowski et al. 2012). Even though it is currently unclear how this developmental phenotype relates to neuronal loss in disease, it has been used as a screening assay for disease modifying activity. The motor neuron phenotype induced by expression of mutant SOD1 and TDP-43 has been successfully screened with a MO library composed of 303 MO for genes that are able to rescue the motor neuron phenotype (Van Hoecke et al. 2012). Loss of the Ephrin receptor EphA4 has been identified as a potent modifier of the motor neuron phenotype (Van Hoecke et al. 2012) almost completely restoring the motor neuron phenotype to wildtype. The protective activity of EphA4 is conserved in rodent models of ALS and represents a promising novel drug target for patients. This was the first example of an unbiased screen in zebrafish that identified a potential therapeutic target for FTLD/ALS treatment, underscoring the zebrafish's potential for drug discovery.

A similar screen was performed with chemical compounds with known neuroprotective effects on mutant TDP-43- and Fus-induced motor neuron phenotypes. Methylene blue, a drug currently in clinical trials to treat Alzheimer's disease, has been identified as a potent protective compound to rescue motor neuron abnormalities and swimming defects upon mutant TDP-43 and Fus expression (Vaccaro et al. 2013). Follow-up studies in rodents will be required to determine if the neuroprotective activity of methylene blue is a valuable therapeutic strategy for FTLD and ALS patients.

Next generation zebrafish disease models

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References

With the highly efficient TALEN and CRISPR/Cas9 system, targeted KI in zebrafish is now within reach. For the TALEN system, somatic KI of a loxP site has been reported (Bedell et al. 2012). As the CRISPR/Cas9 system is equally efficient in the generation of insertions and deletions, we hypothesize that it too can be efficiently used for KI. Many further applications for KI are desirable for zebrafish research: KI of tags to perform biochemistry on zebrafish proteins that often lack suitable antibodies, KI of fluorescent proteins for in vivo imaging, KI of mutations, etc. Furthermore, pairs of genome-editing tools can be designed to induce large deletions or inversions at specific sites (Xiao et al. 2013a), an extremely useful feature to perform promotor analysis or to remove gene clusters. Possible off-target effects and the precision of KI-targeting construct integration with CRISPR/Cas9 have yet to be systematically investigated in zebrafish. However, off-target phenotypes have neither been reported in the analyses of mutants in zebrafish nor are they anticipated to be problematic, as mutants can be outcrossed for several generations to eliminate unwanted off-target mutations. The next generation of zebrafish disease models will introduce the disease-associated mutation into the corresponding zebrafish genomic locus. This approach will most likely eliminate many unspecific phenotypes as a result of over-expression toxicity or MO KD off- target effects. These models will therefore more faithfully reproduce clinical phenotypes.

With the new genome-editing toolbox, many more sophisticated editing approaches will become feasible. TALE and CRISPR can not only be used with nucleases to induce double-strand breaks, but they can also be hooked up to transcriptional activators or repressors (Gaj et al. 2013), opening many new and exciting avenues for manipulations.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References

This work was supported by the European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013)/ ERC Grant Agreement No. 321366-Amyloid (advanced grant to CH).

References

  1. Top of page
  2. Abstract
  3. Zebrafish as an animal model for neurodegeneration
  4. Genome editing
  5. Modeling frontopemporal lobar degeneration and amyotrophic lateralsclerosis using genome editing
  6. Knock-down versus knock-out
  7. Next generation zebrafish disease models
  8. Acknowledgements
  9. Conflict of interest
  10. References