Characterization of genetic loss-of-function of Fus in zebrafish.

The RNA-binding protein FUS is implicated in transcription, alternative splicing of neuronal genes and DNA repair. Mutations in FUS have been linked to human neurodegenerative diseases such as ALS (amyotrophic lateral sclerosis). We genetically disrupted fus in zebrafish (Danio rerio) using the CRISPR-Cas9 system. The fus knockout animals are fertile and did not show any distinctive phenotype. Mutation of fus induces mild changes in gene expression on the transcriptome and proteome level in the adult brain. We observed a significant influence of genetic background on gene expression and 3'UTR usage, which could mask the effects of loss of Fus. Unlike published fus morphants, maternal zygotic fus mutants do not show motoneuronal degeneration and exhibit normal locomotor activity.

amyotrophic lateral sclerosis

nuclear localization sequence

Fus

Ewsr1 family of RNA-binding proteins

low complexity domain

C-terminal domain (of RNA polymerase II)

gene ontology

hours post fertilization

days post fertilization

maternal zygotic

morpholino

Introduction

RNA-binding proteins that carry aggregation-prone prion-like domains have recently gained attention due to their involvement in many cellular processes and disease. One of the most studied of these proteins is FUS (“Fusion,” or TLS, “translocated in liposarcoma”). FUS belongs to a conserved family of RNA-binding proteins (FET family, from the names of the members FUS, EWSR1, and TAF15) which shuttle between the nucleus and cytoplasm, are predominantly nuclear in steady-state and have long N-terminal domains of low complexity (Q,S,G,Y-rich stretches, further called LC domains). In addition, they contain RGG motifs, one RRM and one RanBP2-type Zinc finger domain. FUS plays roles in regulation of splicing, especially co-transcriptional splicing of long neuronal pre-mRNAs, 3′end definition of mRNAs, DNA repair and the formation of nuclear foci: paraspeckles and Gems. The LC domain provides FUS with the ability of phase transition, enabling it to form RNP foci. Mutations in the FUS nuclear localization sequence (NLS) which are associated with the disease ALS (amyotrophic lateral sclerosis) lead to FUS depletion from the nucleus and its aggregation in the cytoplasm of motoneurons, and eventually to motoneuronal death. While both the loss of function of FUS in the nucleus, and toxicity of the cytoplasmic FUS aggregates can contribute to neurodegeneration, most recent data suggest that a gain of function of FUS is detrimental. Several independently produced rodent models show ALS-like neurodegeneration phenotypes upon overexpression of wild-type FUS, while the loss-of-function of FUS alone produces phenotypes different from ALS in mice. Zebrafish injected with antisense morpholino oligonucleotides against fus, as well as overexpressing mutant fus mRNA display motoneuron degeneration and swimming defects. Given recent controversies in the field regarding the non-specific effects of morpholinos and the readily available CRISPR-Cas9 technique we created a genetic knockout zebrafish model to study fus loss-of-function. We show that contrary to morpholino, and similar to mouse Fus knockout, genetic loss-of-function of Fus does not cause an ALS-like phenotype in zebrafish. We also report high-throughput transcriptomic and proteomic profiling of the adult zebrafish fus knockout brain.

Results

Fus knockout alleles in zebrafish

We used CRISPR-Cas9 technology to disrupt the zebrafish fus gene with a single guide RNA targeting exon 3 (Fig. 1A). All subsequent analyses were performed on the allele that deletes 8 base pairs, resulting in a frameshift and a premature stop codon (“Δ8” allele) (Fig. 1A, B). The Δ8 fus mRNA was still present in knockout animals, although it was approximately 2 times less abundant than the wt mRNA (Fig. 1C). The published antibodies used to detect zebrafish Fus protein on Western blot failed to detect Fus in wild type embryos in our hands (not shown). Therefore we used mass spectrometry (MS) in order to determine if the Δ8 fus mRNA was still able to produce protein. To minimize a possible contribution of maternally deposited protein, we analyzed 5 day-old embryos from a heterozygous incross. No peptides unique to Fus protein were detected in any of the 4 replicates of fus embryos, whereas 2 Fus-unique peptides were present in ¾ replicates of the wild type siblings (Table 1). This analysis led us to conclude that Fus protein was not produced and the fus Δ8 allele is a loss-of-function mutation.

Figure 1.

Generation and validation of the fus knockout zebrafish. (A) fus alleles generated by CRISPR-Cas9. A screenshot from the UCSC genome browser shows aligned sequences from heterozygous F1 animals. CRISPR target site in fus exon 3 is underlined in red. (B) An example of a PAGE gel for genotyping the Δ8 allele. M = Low molecular weight marker. Wild type (75bp) and fus−/− (67bp) products can be discriminated. (C) Quantification of fus mRNA relative to WT mRNA in 5dpf embryos. (D) Normal morphology of the WT and fus−/− embryo at 6 d post fertilization (dpf). Scale bar is 1mm.

Table 1.

Identification of Fus by mass-spectrometry (MS). Peptides unique to Fus protein were only identified in the MS analysis of the WT 5dpf embryos and adult brain, but not in the Fus knockout. Cross (x) signifies that the peptide was identified in the respective sample. Below: Fus full protein sequence, detected peptides unique for Fus are bold and underlined.

		wild type				fus−/−
Sample	Sequence	repl. 1	repl. 2	repl. 3	repl. 4	repl. 1	repl. 2	repl. 3	repl. 4
Embryo	AAIDWFDGKDFNGNPIK	X
	TGLPMINLYTDR		X		X
Brain	AAIDWFDGKDFNGNPIK	X	X	X	X
	CSNPSCGNLNFSWR	X	X	X
	TGLPMINLYTDR	X

MASNDYGQTSSHGYGGYGGQSGQSYSQPSAQNYSQQSYGGYNQSSESSSAPYNQGGYSSNYGQSQSGGYGSQAPSQGYSQSSQSYSSGGYSNTSQPPPAQSGGYSQQSSYSGYNQSSPASAPGGYSSSSQSSGYGQQQQQSGGGYGGSGGQSGGYGSSGGQSSGFGGSGGQHQSSQSGGGSYSPSPNYSSPPPQSYGQQSQYGQGGYNQDSPPMSGGGGGGGYGGQDGGYSQDGRGGRGRGGGFGGRGAGGFDRGGRGGPRGRGGMGMGDRGGFNKFGGPRDHGAGGPNMQEQDNSDNNTIFVQGLGDDYTVDSVADYFKQIGIIKVNKKTGLPMINLYTDRETGKLKGEATVSFDDPPSAKAAIDWFDGKDFNGNPIKVSFATRRAEFGRGGSSGGMRGGRGRGGPMGRGGFGGGRGGGGGGGGFQGNNGGGSGNGGGQQRAGDWKCSNPSCGNLNFSWRNECNQCKEPKPEGSGGGMSPMGGGFGGERGRSGFDRGGFRGRGGDRGGFRGGRGGDRGGFGPGKMDSRGDHRHDRRDRPY

Transcriptome and proteome analysis of fus knockout brain

Fus homozygous knockout fish, including maternal zygotic mutants, are viable, fertile and appear morphologically normal (Fig. 1D). In search for molecular consequences of loss of Fus, we performed gene expression analysis on adult brains of fus animals and wild type siblings. Fus is ubiquitously expressed with the highest expression found in brain, as shown by in situ hybridization and qPCR (Fig. S1A, B), immunostaining and as inferred from publicly available RNA-Seq data. Also, numerous splicing changes have been reported for FUS loss-of-function in the mouse brain and spinal cord. We therefore considered brain to be the most relevant tissue to search for gene deregulation upon loss of Fus. We performed paired-end sequencing of the polyA+ transcriptome from brains of 5 fus and 4 wild type animals (Fig. 2A). All animals were 8-month old female siblings. The most significantly changed gene was fus itself, and only 5 other transcripts changed over 2-fold (Fig. 2B and Table S1). We selected 16 genes and validated their fold changes with qPCR using independently dissected brains (Fig. S1C). Changes in alternative splicing are a hallmark of the FUS-dependent regulation in mouse brain. However, in the zebrafish fus−/− brain we did not reliably identify strong changes in alternative splicing using several available tools (DEXSeq, MISO, cuffdiff, data not shown). Likewise, identified FUS targets with long introns reported to be downregulated in FUS knockdown mouse brain failed to show comparable changes in zebrafish (Fig. S2). In parallel to the transcriptome analysis, we quantified proteins using 4 replicates of the RNA-profiled brain lysates using label-free quantitative mass spectrometry. In accordance with the transcriptome, protein steady-state levels remained unchanged upon loss of Fus (Fig. 2C and Table S2).

Figure 2.

Transcriptome and proteome of the fus−/− brain. (A) Adult. zebrafish brains from either fus−/− or WT siblings were used to simultaneously quantify mRNAs by RNA sequencing and proteins by mass-spectrometry. In total, transcriptomes of 5 fus−/− and 4 WT, and proteomes of 4 fus−/− and 4 WT animals were measured. (B) Volcano plot of mRNA fold changes displays the fold change (on the x axis) and p-value (on the y axis). fus is the most significantly changed gene on mRNA level (labeled in red) (C) Quantification of the proteome by label-free mass-spectrometry. Label-free quantification was performed as previously described. Median expression of the proteins in WT brain (y axis) is plotted against the expression in fus−/− brain (x axis) on the log scale. Proteins not detected in one of the samples were assigned an arbitrary value of 22. Fus is among the highest expressed proteins that were detected in WT and missing from the knockout (marked in red).

Genetic background affects alternative 3′UTR usage

During analysis of the RNA-sequencing data, we noticed that some genes display large variation in gene expression among siblings (biological replicates), independent of the fus genotype (Fig. S3). Zebrafish cannot be maintained as inbred lines because of inbreeding depression, and the background of our line is mixed (AB × TU). We reasoned that genetic background effects may prevail over subtle effects of fus knockout. To address this issue more systematically, we chose to focus on alternative polyadenylation site usage, which is relatively straightforward to quantify and visualize. We used isoSCM to de novo detect 3′UTR isoforms, and quantified their differential expression. Next, we called single nucleotide polymorphisms (SNPs) from the RNA-Seq data taking advantage of the sufficient read coverage over the 3′UTRs. Indeed, the top differential 3′UTRs isoforms often followed the distribution of SNPs, and not the fus genotype (Fig. 3). Thus, we are able to detect differential isoform expression in our RNA-Seq data, but the isoform usage seems to be more strongly influenced by local genetic background than by the expression of Fus.

Figure 3.

Examples of alternative 3′UTR usage in different genetic backgrounds. Shown are screenshots from IGV genome browser. The relevant genes are slc37a4b, pomp and imp3. For convenience, RNA-Seq coverage tracks are shown in blue only for the minus strand. Called SNPs that have been filtered (see methods) are represented by colored bars above the gene track. Dark blue are homozygous variants, dark red bars are heterozygous; homozygous reference or absent calls are not shown. The isoSCM track shows putative alternative 3′ends called by isoSCM. The scale for RNA-Seq coverage is normalized to 1,000,000/(total read count) and is the same for all tracks (shown in gray in square brackets on the right).

Fus knockout embryos lack motoneuronal defects

It has been previously reported that zebrafish fus morphants display defects in motoneurons. These defects comprise severe shortening of motoneuronal axons and their excessive branching, abnormal formation of neuromuscular junctions and impaired locomotion. We asked if some of these defects could be observed in maternal zygotic fus knockout (MZfus−/−) embryos. To address the structure of primary motoneuronal axons we crossed fus mutants to the Nbt:dsRed reporter line expressing DsRed fluorescent protein under the control of neuronal specific promoter (Xenopus neural-specific β tubulin promoter). We imaged transgenic embryos at the age of 36 hours post fertilization (hpf) (Fig. 4A). We did not observe shortening of motoneuronal axons in both zygotic and maternal zygotic fus−/− embryos compared with wild type siblings (n = 5 wild type, 6 zygotic and 7 maternal zygotic fus embryos). More detailed, quantitative analysis also did not reveal strong differences in branching between wt and mutant axons (Fig. S6). We further assessed motor activity of fus−/− embryos by performing the touch-evoked escape response assay. In this assay, 2 day old embryos are placed in the middle of a dish with embryonic medium and the tail is slightly touched with forceps. Healthy embryos respond to the touch by swimming away (escaping). Embryos with motoneuron impairment are expected to swim a shorter distance. We quantified the length of the swim tracks and did not find a significant difference between wild type and fus embryos (Fig. 4B). Since we could not see the locomotor phenotype upon loss of Fus, we tried to reproduce the published fus morpholino experiments using the same line (AB × TU) in which the fus mutant was generated. Surprisingly, we could not detect a significant reduction in swim distance for the fus morpholino injected embryos (Fig. S4A, left panel), even though we confirmed that the fus morpholino was functional by co-injecting GFP mRNA carrying the fus morpholino target site (Fig. S4B). We reasoned that our outbred strain might have a different response to morpholino injection. We therefore performed the touch response assay on fus MO-injected fish with an AB background and observed a trend for reduced swim distance in MO-injected, but not in control MO-injected embryos. Nevertheless, the difference was still not significant (p = 0.103) (Fig. S4A, right panel).

Figure 4.

Motoneuron morphology and touch evoked escape response of fus knockout larvae. (A) Confocal images of trunks of 36hpf larvae expressing neuron-specific DsRed. Maternal zygotic Fus knockout larvae show normal overall primary motoneuron axon morphology indistinguishable from WT. Maximal intensity projection of a confocal stack is shown; scale bar is 100µm. (B) Touch evoked escape response of 2dpf (48–52hpf) larvae. Swim distances in mm for individual larvae tracks are plotted. P-value is from Kruskal-Wallis test. n indicates the number of larvae tested for each group.

Discussion

We report the generation of a fus knockout zebrafish model and its phenotypic characterization. Maternal zygotic fus knockouts are healthy and fertile. In contrast to one of the mouse knockouts, fus knockout zebrafish are not radiation sensitive (data not shown). The brain transcriptome and proteome was only slightly affected by the loss of Fus protein, with most changes being less than 2-fold. This is in line with the most recent RNA-Seq data from the mouse FUS knockout brain where only ∼10% of the changes were more than 1.5-fold. Overall, our results with Fus knockout zebrafish add to the idea that, in vertebrate model organisms, Fus loss-of-function alone does not elicit the neurodegenerative phenotype seen in FUS-related diseases, supporting recent mouse data. It is tempting to speculate that other members of the FET family, Taf15 and Ewsr1, may compensate for the loss of Fus in vertebrates. Drosophila, which has only one homolog of FET proteins, Cabeza, does show reduced survival and impaired locomotion upon its inactivation, and simultaneous depletion of FUS and TAF15 in human motoneurons results in gene expression profile similar to that in ALS patient-derived neurons bearing the ALS mutation FUSR521G. It is interesting to note as well that in zebrafish the genetic knockout of another ALS-related RNA-binding protein Tardbp apparently causes neurodegeneration followed by muscular atrophy, but only if its paralog Tardbpl is also inactivated. FET proteins are similar in structure, localization, have partially overlapping targets and interact with each other. Notably, none of the FET proteins was upregulated in our Fus knockout (Fig. S1C), suggesting that genetic compensation does not take place. However, since zebrafish has 4 FET proteins as compare with 3 in all mammals, it is still plausible that upregulation is not necessary to compensate for the loss of Fus. Simultaneous knockout of several FET proteins in zebrafish may elicit stronger phenotype.

We used outbred zebrafish (AB × TU) to generate fus knockout animals. This strategy was beneficial, since embryos in TU background injected with highly concentrated Cas9 and guide RNA showed poor survival. However, we observed that the genetic background can strongly influence gene expression and alternative isoform usage. These effects could have obscured mild expression changes resulting from the loss of Fus. In addition, our data suggest that effects of morpholino injections can manifest differently depending on the genetic background of the embryos. In line with this, several FUS knockout mouse models show a wide spectrum of phenotypes which range from postnatal lethality in more inbred lines to surviving adults with mild morphological and behavioral defects in an outbred line. Since the high density of SNPs, and outbred nature of zebrafish is more similar to that in humans than that in inbred mouse lines, zebrafish can represent a very useful model to study the effects of Fus, or that of any other protein, within the context of naturally occurring genetic variation.

Data

Raw RNAseq data are deposited in GEO (GSE85554). Proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD004876e.

Methods

Extended methods can be found in the Supplementary material.

Fish husbandry and strains

Fish were maintained, raised, and staged as described before. Unless stated otherwise, in all experiments a mix of AB and TU strain was used. All experiments were performed according to German animal welfare law (licenses 23 177–07/G 13–5–087 (CRISPR-Cas9 knockouts), 23177–07/A 15–5–001 OES (tail fin clips)). CRISPR-Cas9 knockouts were generated as described.

mRNA sequencing

Strand-specific polyadenylated RNA libraries were prepared using the TruSeq Stranded RNA Library Preparation Kit and sequenced on an Illumina HiSeq2000 using the 2 × 100bp read protocol. RNA-sequencing reads were mapped to the zebrafish genome (Zv10) using the aligner STAR (version 2.4.1d). Read count Table was produced with featureCounts (v. 1.4.6) using Ensembl database gene annotation (release 80). Differentially expressed genes were called using R package DESeq2.

Mass spectrometry

Mass spectrometry sample preparation and measurement was done as described except a 50 cm C18 column and a 240 minute LC gradient was used. Samples were analyzed with MaxQuant (version 1.5.2.8) against the supplied Uniprot zebrafish database (39,559 entries) with standard settings, except LFQ values were based on unique peptides. For data analysis, information about identified peptides was extracted from the evidence table. On the protein level, the median LFQ values were calculated for each condition (WT, fus−/−) from the protein groups file and plotted as a scatterplot in R using the ggplot2 package.

Confocal imaging

Zebrafish trunks were imaged on Leica SP5 confocal microscope. Maximal projection images were generated with Fiji software.

Morpholino injections

Fus morpholino (GGCCATAATCATTTGACGCCATGTT) and 5-mismatch control morpholino (GCCCATAATGATTTCACGGCATCTT) were from Genetools. Approximately 1nl drop of 1mM Fus morpholino was injected into 1-cell stage embryos.

Touch evoked response assay

Embryos at 2dpf were placed in a 15cm dish with embryonic medium, lightly touched with forceps at the tail, and movements were video recorded using a Basler acA2000–165um Monochrome USB camera with the speed of 40 frames per second. A ruler was placed next to the dish to calibrate the track lengths in mm. The videos were processed in Fiji using a customized macro script and MTrack2 plugin (http://valelab.ucsf.edu/∼nstuurman/ijplugins/MTrack2.html). Only complete tracks were considered for statistical analysis.