Introduction

Fragile X syndrome (FXS) is one of the most common known causes of inherited mental retardation with a frequency of 1∶4000 males and 1∶6000 females [1]. In almost all cases, FXS is due to the expansion of the unstable CGG trinucleotide repeat sequence in the 5′ untranslated region of the FMR1 gene [2], [3]. Once the repeats exceed 200 units (full mutation), the gene is silenced due to the consequent hypermethylation of the CpG island and CGG repeat. Thus, no mRNA is produced, and the lack of the gene product, FMRP, is responsible for the mental retardation in fragile X patients [4]. Other clinical features include macroorchidism, autistic behaviour, epileptic seizures, hyperactivity, attention deficits and mild craniofacial abnormalities [1].

FMRP is a ubiquitously expressed RNA-binding protein, including two KH domains and an RGG box, with high expression levels in brain and testis [5], [6]. The protein can bind to RNAs containing a G-quartet structure and forms together with many other mRNAs and proteins a messenger ribonucleoprotein (mRNP) particle [7], [8]. The dynamics and transport of mRNP particles over long distances within the dendrites of neurons is established by movement along microtubules [9].

The development of mouse models of FXS has facilitated cellular studies on the underlying molecular basis of this loss-of-function disorder [10], [11]. Fmr1 knock-out mice recapitulate the typical characteristics of FXS, including behavioural abnormalities, learning deficits and audiogenic seizures. Microscopic analysis of brain material from both FXS patients and Fmr1 knockout mice has shown dendritic spine abnormalities [12]–[17]. The discovery of a spine morphological phenotype indicates a possible defect in synaptic plasticity in FXS. The precise physiological function of FMRP is still not defined; therefore, the role of FMRP at the synapse has become a central research interest. Compelling evidence predicts a model in which FMRP is involved in the regulation (repression) of local protein synthesis at the synapse, which is triggered group 1 mGluR (mGluR1 and mGluR5) activation. Thus, a lack of FMRP may lead to uncontrolled (exaggerated) protein synthesis at the synapse upon group 1 mGluR stimulation and may underlie the enhanced hippocampal and cerebellar LTD found in Fmr1 knock-out mice [16], [18], [19]. Interestingly, some behavioural abnormalities could be rescued in Fmr1 knock-out mice using mGluR5 antagonists [20], [21]. Recently, a rescue of the spine morphological phenotype could be established in cultured Fmr1 knock-out hippocampal neurons using two different mGluR5 antagonists [21].

In 2006, Tucker et al. reported the use of zebrafish embryos to model FXS [22]. Instead of a knock-out approach, a knock-down strategy was applied using microinjection of morpholinos (MOs) into 1–2 cell stage embryos. MOs are antisense oligonucleotides, in which the deoxyribose is substituted with an N-morpholino ring. They can bind to a target mRNA and prevent either translation or normal splicing for up to 4 days. Hence, inhibition of translation is transient and may not result in a complete loss-of-function. Injection of fmr1 specific MOs resulted in abnormal axonal branching, changes in trigeminal ganglion number and craniofacial abnormalities. Most of these abnormalities in zebrafish embryos could be rescued using MPEP, an mGluR5 antagonist, or by fmr1 overexpression [22].

In the present study, we generated two independent fmr1 knock-out alleles using TILLING (targeted induced local lesions in genomes). TILLING combines random induced mutations by ENU treatment and subsequent screening for null mutations [23]. We provide a characterization of both homozygous and transheterozygous mutants with special emphasis on the phenotypic features reported earlier in the fmr1 knock-down study [22].

Results

Isolation of Two Fmr1 Mutant Alleles

In order to develop a genetic model in which the effects of FMRP on brain development can be easily studied during development we screened for knock-out alleles in the zebrafish system. From a randomly mutagenized library we isolated two independent mutant alleles: hu2787 defines a C to T change in the coding region of fmr1 (ENSDARG00000037433), leading to the introduction of a premature stop at codon position 113 (Figure 1A). The mRNA derived from this allele is less stable than that derived from the wild-type fmr1 locus. This is illustrated in Figure 1B using whole mount in situ hybridisation with an fmr1 specific probe, on a batch of embryos obtained from a cross between heterozygous parents. Presumably this is the result from a well-known phenomenon named nonsense-mediated-decay (NMD). Furthermore, using a C-terminal antibody we are unable to detect expression of Fmr in neurons using immunocytochemistry on paraffin sections, whereas a high expression could be detected in neurons from wild type zebrafish. Figure 1C illustrates high Fmr expression in Purkinje cells in the cerebellum and neurons in the telencephalon. In addition, we determined Fmr expression in total brain homogenates by Western blot analysis. Consistent with the immuno-stainings, no Fmr was detectable (Figure 1D). The second allele we isolated, hu2898, has a mutated splice acceptor site at the end of the 7th intron. Sequencing of splicing products from this allele shows that hu2898 leads to the use of an alternative splice acceptor site 2 bases downstream of the original site. This induces a frameshift with regard to the original reading frame and an opal stop codon 27 nucleotides downstream (Figure 1A). Animals carrying any combination of the two mutant alleles show loss of Fmr in both immunocytochemistry (Figure 1C) and Western blot analysis (Figure 1D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Fmr1 mutant alleles.

A) Illustration of the fmr1 gene product. The different domains are indicated, along with the sites where isolated mutant alleles will affect the protein. B) Whole mount in situ hybridisation with an fmr1 specific probe. C) Immuno staining of wild type and mutant brain sections using Fmr specific antibodies. Some Purkinje cells in the mutant have been outlined. D) Brain lysates were analyzed by western blot, using an Fmr specific antibody. Lanes 1 and 3 contain wild type samples. Lane 2 contains hu2787/hu2787 lysate. Lane 4 contains hu2787/hu2898 lysate. The upper arrow points at Fmr. The lower arrow points at an a-specific band that serves as a loading control.

https://doi.org/10.1371/journal.pone.0007910.g001

Lack of Fmr1 Does Not Induce Craniofacial Defects

The results described above contrasts with morpholino induced fmr1 knock-down studies that have been published before [22]. More specifically, it was demonstrated that these morphants display aberrant expression of three markers: axial, dlx-2a and islet-1. We therefore analyzed the expression of these genes by in situ hybridisation in fmr1 MZ null embryos. The results of these experiments are depicted in Figure 2A. We observed no significant differences between wild-type and fmr1 null embryos in any of the analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Phenotypic assays on wild-type and fmr1 mutant embryos.

A) Wild type and mutant embryos were analyzed using whole mount in situ hybridisation using probes against dlx-2a, axial and islet-1. B) The width of Meckel's cartilage was measured in wild type (n = 9) and MZ fmr1 mutant (n = 11) embryos. The angle of this structure with regard to the anterior-posterior axis was also measured in wild-type (n = 6) and fmr1 mutant (n = 9) embryos. Indicated errors represent SD. C) Neurite branching was measured on Rohon-Beard neurites using the monoclonal antibody zn-12. Plotted is the branching frequency per 1000 µm in both wild-type and MZ stop mutant embryos. In total n = 25 neurites (wild-type) and n = 28 neurites (MZfmr1) were traced in a total of 8 embryos of each genotype. Error bars represent SD.

https://doi.org/10.1371/journal.pone.0007910.g002

In addition, we measured the width of Meckel's cartilage and the angle it makes to the anterior-posterior axis, to address whether these mutant animals develop abnormalities that may be related to the craniofacial defects seen in fragile X patients, as described in the morpholino knockdown morphants [22]. In both the MZ homozygous stop mutants and embryos derived from homozygous stop mutant mothers and homozygous splice mutant fathers (not shown) the width as well as the angle of this structure is indistinguishable from that in wild-type animals (Figure 2B).

Discussion

We describe the generation of two fmr1 knockout alleles in zebrafish, and as such provide a new genetic model system to study FXS, a highly prevalent form of inherited mental retardation. FXS is caused by the loss of the gene product of fmr1, Fmr. FXS models have been described in multiple systems and from these models it has become clear that FMRP is acting at the synapse to regulate the translation of target mRNAs upon group 1 mGluR stimulation and whose protein products mediate synaptic strength [16], [18]. Fmr1 knock-out mice exhibit exaggerated translation of target mRNAs at the synapse. Potentially, such a process can be well affected by mGluR antagonists that would ameliorate the phenotypic outcome of FXS [24].

Establishing a zebrafish model for FXS is very useful in this context, as the zebrafish embryo is amenable to large scale, small molecule drug screens. Supporting this idea was the finding that morpholino induced knock-down of fmr1 (fmr1 morphant) in the zebrafish led to embryonic phenotypes that could in principle be used as a read-out in drug screens [22]. Tucker et al. [22] reported neurite branching defects and changes in trigeminal ganglion neuron number following fmr1 knock down [22]. Interestingly, treatment with MPEP, an mGluR5 antagonist, rescued most of these abnormalities, indicating a connection between mGluR5 signalling and fmr1 function in neurite branching and number of trigeminal ganglion neurons. Furthermore, it suggested that small molecule drug screens in the zebrafish may indeed be an affective manner of finding bio-active lead-compounds that are good starting points for developing drugs beneficial for FXS patients.

Tucker et al [22] also reported craniofacial dysmorphology as a result of fmr1 knock-down, and this again could be rescued by treatment with MPEP. This is a curious finding, since the role of synaptic connections between neurons in cranial cartilage development is to date totally unexplored, and hence it is not clear whether indeed MPEP would be expected to affect craniofacial defects caused by loss of Fmr.

We here characterize two fmr1 mutant alleles in the zebrafish, both of which lead to loss of detectable Fmr. In contrast to the above-mentioned MO study, however, we cannot find any gross phenotypical effects caused by these alleles. We checked specifically the above-mentioned phenotypes, craniofacial abnormalities and neurite branching phenotypes, but find no significant differences between wild type and mutant siblings. Fmr1 mutant fish are also completely fertile, and incrosses between homozygous mutant males and females result in normally developing embryos, indicating that maternally provided mRNA and/or protein is not rescuing first generation fmr1 mutants.

What could be causing the observed phenotypes in the morphants [22], when genetic fmr1 null animals do not display these defects? First we explore why genetic mutation of fmr1 may miss FXS-relevant phenotypes. Redundancy could potentially be an issue. The morpholinos used could affect the fmr1 homologues fxr1 and fxr2, which are both present in zebrafish. This seems unlikely, however, given the fact the sequence comparison between the morpholino and fxr-1/2 genes shows very little complementarity. Fmr1 itself could be duplicated in zebrafish. However, the most recent genome annotation shows no indication of a duplicated fmr1 gene, and on western blot we detect no protein in fmr1 homozygous mutant tissue. This makes the presence of a closely related, functional fmr1 copy unlikely. Finally, potential phenotypes may be rescued by modifier loci; loci that genetically interact with fmr1 and of which particular alleles may suppress phenotypes triggered by loss of Fmr. Despite the fact that our zebrafish strains show no sign of selection for or against homozygous fmr1 mutants, this is an option that is difficult to eliminate. Extensive outcrossing into the zebrafish strains used in the studies by Tucker et al [22] would be required to test this hypothesis.

There is, however, a more likely potential explanation: the morpholino-induced phenotypes may not be related to loss of Fmr. Morpholino oligonucleotides are well known to cause phenotypes unrelated to knock-down of the intended gene. In fact, 15–20% of MOs used in zebrafish show off-targeting effects that are mediated by p53-induced apoptosis [25]. In the study from Tucker et al. [22] the number of analyzed morphants is very limited. For instance, altered dlx-2a, fgfr1 and axial expression could only be observed in 17/30, 11/30 and 3/30 fmr1 morphants, respectively; for neurite branching phenotypes no numbers are given related to the penetrance of the defect; injection of antibodies against alpha-acetylated tubulin resulted in a dramatic axon defect only in 3/30 and axon defasciculation in 13/30 fmr1 morphants. Finally, the craniofacial dysmorphology could only be observed in 9/15 fmr1 morphants.

In summary, we find the loss of fmr1 in zebrafish at most induces very subtle phenotypes that are not readily detectable using light-microscopy and techniques like immunocytochemistry and in situ hybridisation, at least in the strains used in our laboratory. It remains well possible that subtle defects are induced by lesions in fmr1, and that these may be used to develop sensitive and robust essays to probe fmr1 function, which may in turn be used for screening of small molecules libraries in order to find drugs suitable for treatment of FXS. At present, however, we have to conclude that the phenotypes as described by Tucker et al [22] may be based on morpholino induced artefacts, and as such not useful to study fmr1 function in the zebrafish.

Materials and Methods

Acknowledgments

We would like to thank Lies-anne Severijnen, Edwin Romme, Laleh Davanian and Marit de Haan for technical assistance. We thank Edwin Cuppen and Ewart de Bruijn for assistance in screening ENU mutant zebrafish libraries. We thank E. Meijering for analysis of neurite branching using NeuronJ plugin through Image J Software (NIH) software.