https://orcid.org/0000-0001-9596-6040
Abstract
The formation of new genes is a major source of organism evolutionary innovation. Beyond their mutational effects, transposable elements can be co-opted by host genomes to form different types of sequences including novel genes, through a mechanism named molecular domestication. We report the formation of four genes through molecular domestication of Harbinger transposons, three in a common ancestor of jawed vertebrates about 500 million years ago and one in sarcopterygians approx. 430 million years ago. Additionally, one processed pseudogene arose approx. 60 million years ago in simians. In zebrafish, Harbinger-derived genes are expressed during early development but also in adult tissues, and predominantly co-expressed in male brain. In human, expression was detected in multiple organs, with major expression in the brain particularly during fetal development. We used CRISPR/Cas9 with direct gene knock-out in the F0 generation and the morpholino antisense oligonucleotide knock-down technique to study in zebrafish the function of one of these genes called MSANTD2, which has been suggested to be associated to neuro-developmental diseases such as autism spectrum disorders and schizophrenia in human. MSANTD2 inactivation led to developmental delays including tail and nervous system malformation at one day post fertilization. Affected embryos showed dead cell accumulation, major anatomical defects characterized by impaired brain ventricle formation and alterations in expression of some characteristic genes involved in vertebrate nervous system development. Hence, the characterization of MSANTD2 and other Harbinger-derived genes might contribute to a better understanding of the genetic innovations having driven the early evolution of the vertebrate nervous system.
Introduction
The formation of new genes is an important source of evolutionary innovation and adaptation for species (Kaessmann 2010). Indeed, they represent a major substrate for the emergence of new functions and contribute to the birth of novel phenotypic traits that are source of adaptation and speciation. For example, new genes can be generated de novo from scratch from initially nonfunctional sequences, a rare phenomenon, or through the duplication of preexisting genes, which can lead to new functions linked to mutations and relaxed selective constraints (Ohno 1972; Lynch and Conery 2000; Knowles and McLysaght 2009; Toll-Riera et al. 2009). Another source of new genes is the recruitment, also called molecular domestication, of transposable element (TE)-coding sequences (Volff 2006; Moran and Malik 2009; Alzohairy et al. 2013).
TEs are repeated DNA sequences that can insert into novel genomic locations and thus can cause genomic instability through insertion and recombination (Katoh et al. 2002). TEs have been found in every species that have been investigated. However, the quantitative and qualitative composition of TEs in genomes is variable depending on the species (Huelsenbeck and Ronquist 2001). While TEs are mutagenic agents that can have neutral or deleterious effects on genomes (Ohno 1972; Doolittle and Sapienza 1980; Orgel and Crick 1980), they can also serve as material for the formation of new regulatory sequences, new exons or even new genes (Kidwell and Lisch 2000; Warren et al. 2015; Chuong et al. 2017). TEs have been source of major innovations during evolution, as exemplified by vertebrate development (Etchegaray et al. 2021). By the process of molecular domestication, TEs can give rise to new functional genes positively selected in host genomes. Major examples of TE domestication have been documented in vertebrates, such as the RAG genes involved in the adaptive immune system and the SYNCYTIN genes necessary for placenta development in mammals (Mallet et al. 2004; Dupressoir et al. 2011; Kapitonov and Koonin 2015; Etchegaray et al. 2021). Thus, TE molecular domestication can lead to important adaptive innovations. In the human genome, which is composed at least of 45% of TEs (Lander et al. 2001), a hundred cases of protein-coding genes derived from TEs have been identified so far (Volff 2006). However, most of these genes have been poorly characterized, particularly at the functional level. Considering the quantity and diversity of TEs in genomes, their role in the diversification and adaptation of organisms is probably still underestimated (Brandt et al. 2005; Britten 2006; Volff 2006; Alzohairy et al. 2013). Therefore, the identification and functional characterization of new cases of TE-derived genes is important to better understand the formation of novel genes and the factors driving genetic innovation.
In the course of a study aiming to assess the impact of TE molecular domestication on the early evolution of the vertebrate lineage, we have identified several genes domesticated from Harbinger TEs through the comparison of human protein sequences to a vertebrate-wide TE sequence database. Harbinger transposons are DNA transposons present in the genome of protists, plants, insects, worms and vertebrates but absent from mammals (Kapitonov and Jurka 2004). They are generally flanked by terminal inverted repeat sequences (TIRs) and encode two proteins, a transposase with a aspartate/aspartate/glutamate (DDE) endonuclease motif and a SANT-Myb-trihelix motif-containing protein, which we will now refer to as the Myb-like protein. Both genes have been shown to be necessary for Harbinger transposition (Sinzelle et al. 2008; Hancock et al. 2010). The Myb-like protein contains a tri-helix motif with conserved bulky aromatic residues that allows DNA and protein binding. Myb-like proteins are responsible for the nuclear import of the transposase through interaction with its N-terminal end. Thanks to the tri-helix motif, they also bind the TIRs of the transposon, allowing the recruitment of the transposase and thus the excision/insertion of the sequence (Sinzelle et al. 2008).
Two cases of Harbinger-derived genes have been previously identified in vertebrates: HARBI1 and NAIF1 (Kapitonov and Jurka 2004; Sinzelle et al. 2008). HARBI1 is derived from the transposase gene, while NAIF1 has been formed from the second gene encoding the Myb-like protein. The HARBI1 and NAIF1 proteins can directly interact and form a protein complex, NAIF1 allowing the nuclear import of HARBI1. NAIF1 can also bind DNA, but not at the HARBI1 sequence (Sinzelle et al. 2008). NAIF1 has been linked to apoptosis in the context of several cancers and proposed to have antitumoural effects (Lv et al. 2006; Luo et al. 2011; Fu and Cao 2015; Zhao et al. 2015; Kong and Zhang 2018). However, the biological roles of both genes remain largely unknown.
This study describes a family of genes derived from Myb-like genes of Harbinger DNA transposons in jawed vertebrates. We have identified four new genes that have been formed through three to four independent molecular domestication events during vertebrate evolution, three at the base of jawed vertebrates about 500 million years ago and a fourth one possibly in a common ancestor of sarcopterygians ca. 430 million years ago. The Harbinger-derived genes are expressed during zebrafish embryonic development and in zebrafish adult tissues, predominantly in male brain, as well as in human brain during fetal development. Inactivation of one of these genes, MSANTD2, by CRISPR/Cas9 direct knock-out in F0 and morpholino antisense oligonucleotide knock-down techniques in zebrafish led to embryos with severe brain developmental defects and modification of the expression of characteristic genes involved in vertebrate nervous system development. Interestingly, MSANTD2 has been suggested to be associated with neuro-developmental diseases such as autism spectrum disorders and schizophrenia in human (Schizophrenia Working Group of the Psychiatric Genomics Consortium 2014; Lim et al. 2017; O’Brien et al. 2018; Zhang et al. 2020).
Results
Multiple Molecular Domestication Events of Harbinger Transposons Have Formed a New Gene Family in Vertebrates
Identification of New Harbinger Myb-Like-Derived Genes in Jawed Vertebrates
To the best of our knowledge, we performed the first systematic comparison of human protein sequences to an extensive TE sequence database generated during a broad survey in vertebrates, which particularly included many TE families from fish that are absent from mammals and birds (Chalopin et al. 2015). We have identified, in addition to genes already described as TE-derived genes (Volff 2006; Alzohairy et al. 2013), a gene called MSANTD2 (Myb/SANT DNA Binding Domain Containing 2) as a new potential case of molecular domestication from a Harbinger DNA transposon. Indeed, the MSANTD2 predicted protein sequence presented homologies with the Myb-like protein of a Harbinger transposon from the genome of the medaka fish Oryzias latipes, with a conservation score (considering both residue identities and conservation of physico-chemical classes) of 54% in the Myb-like domain region (150 aa). The MSANTD2 gene is located on chromosome 11 in the human genome and is 2,384 base pairs (bp) in length, with four exons encoding a protein of 559 amino-acids (aa). Prediction of conserved domains on the whole sequence of the MSANTD2 protein revealed a single domain, a Myb-like DNA binding domain containing a tri-helix motif. MSANTD2 was different from the two other genes derived from Harbinger transposons previously described in human, HARBI1 that has been formed from a transposase gene and NAIF1 from a Myb-like gene (Kapitonov and Jurka 2004; Sinzelle et al. 2008).
We further identified three additional cases of Harbinger-derived genes in the human genome, called MSANTD1, MSANTD3 and MSANTD4. These three genes encode predicted proteins with similarities to Myb-like proteins of Harbinger transposons. In human, MSANTD1 is on chromosome 4 and 3,164 bp in length, with three exons coding for a 278 aa protein. MSANTD3, on chromosome 9, is a 1,880 bp gene with three exons encoding a 275 aa protein. Finally, MSANTD4, on chromosome 11 and 4103 bp in length, contains three exons coding for a 345 aa protein. Prediction of conserved domains on each MSANTD1, MSANTD3 and MSANTD4 protein also revealed a Myb-like domain containing a tri-helix motif (fig. 1). This suggested functional homology of the MSANTD proteins with the Myb-like proteins of Harbinger transposons, with possible DNA- and/or protein-binding properties.
Multiple alignments of the Myb-like domain of MSANTD proteins and their closest Myb-like proteins from Harbinger transposons. Predicted alpha-helix motifs are represented with dashed squares; bulky aromatic residues, which are essential for alpha helix structure stabilization, are indicated by black stars. The conservation score represented for each residue is measured considering both residue identities and conservation of physico-chemical classes. AG, Anopheles gambiae; DR, Danio rerio; EL, Esox lucius; GA, Gasterosteus aculeatus; GG, Gallus gallus; HS, Homo sapiens; LC, Latimeria chalumnae; OL, Oryzias latipes; ON, Oreochromis niloticus; SS, Salmon salar; TF, Takifugu flavidus; XT, Xenopus tropicalis.
The phylogenetic distribution of Harbinger-derived genes including NAIF1 and HARBI1 was determined using the Ensembl and NCBI databases and verified by blast analysis on metazoan genomes (fig. 2A) (Altschul et al. 1990). All genes were detected only in jawed vertebrates from cartilaginous fishes to mammals, except for MSANTD3, which was absent from both cartilaginous and ray-finned fish genomes but present in sarcopterygians. This suggested the formation of MSANTD1, MSANTD2, MSANTD4, NAIF1 and HARBI1 genes early during vertebrate evolution at the base of jawed vertebrates around 500 Mya, and the occurrence of MSANTD3 in a common ancestor of sarcopterygians around 430 Mya (Lu et al. 2016). Synteny analysis showed that each MSANTD gene was present at the same position in the genome of divergent vertebrate species, indicating that they corresponded to bona fide genes and not to mobile transposon sequences anymore (fig. 2B).
(A) Phylogenetic distribution of Harbinger-derived genes and Harbinger transposons. Presence (+) or absence (–) of these genes in the different lineages is indicated. (B) Synteny analysis of Myb-like-derived genes between human, mouse and spotted gar (nonteleost ray-finned fish) or chicken (for MSANTD3, which is absent from both cartilaginous and ray-finned fish). Species names and genomic locations are shown on the right. For each gene the same color stands for orthologous genes.
Moreover, MSANTD2P1, an intronless processed pseudogene (according to the loss of its protein-coding capacity), probably originating from the retrotransposition of MSANTD2 mRNA, was detected in simians, i.e., from human (on chromosome 21) to marmoset but neither in macaques nor in baboons (supplementary fig. 1, Supplementary Material online). This suggested that MSANTD2P1 appeared at the base of simians about 36–50 Mya (Perelman et al. 2011).
Vertebrate Myb-Like-Derived Genes Originated From Three to Five Independent Molecular Domestication Events of Harbinger Transposons
Predicted Myb-like-derived proteins were compared to Harbinger TE sequences collected from the Repbase database or annotated from sequenced genomes. Multiple sequence alignments were built, comparing the most similar (i.e., with the lowest E-value, all of them <10−5) Harbinger transposon Myb-like proteins from different species with each Myb-like-derived protein (fig. 1). This revealed conservation between the Myb-like domain region of the MSANTD proteins and the Myb-like proteins of Harbinger transposons (covering 115 to 167 aa). The conservation scores (considering both residue identities and conservation of physico-chemical classes) between each MSANTD protein and its closest Harbinger Myb-like proteins were calculated all along the Myb-like domains. For the different MSANTD proteins these scores were estimated between 45 and 57%. Putative alpha helices and aromatic residues, which are essential for Myb-like domain function, were also conserved (fig. 1).
In order to investigate the evolutionary origin(s) of MSANTD genes, phylogenetic trees were built based on protein alignments using the Bayesian method (Huelsenbeck and Ronquist 2001) (fig. 3). While the sequence of the transposase of Harbinger transposons is highly conserved between different families, this is not the case for the Myb-like transposon proteins, which are much more divergent (Kapitonov and Jurka 2004). Such an important sequence divergence was also observed between most MSANTD proteins. Therefore, we were not able to reconstruct reliable general sequence alignment and phylogeny for all MSANTD and transposon Myb-like proteins together. However, MSANTD1 and MSANTD2 were most similar to the same group of Harbinger sequences, and MSANTD3, MSANTD4 and NAIF1 to the same other group of transposon sequences. This allowed generating two different sets of multiple sequence alignments and phylogenies: one for MSANTD1 and MSANTD2 with related Harbinger transposon proteins, and another one for MSANTD3, MSANTD4, NAIF1 and related Harbinger sequences (fig. 3). Phylogenies indicated that each MSANTD sequence from different species formed an independent monophyletic group, and that the closest related Harbinger transposons were different for each MSANTD sequence. Hence, this supported five independent events of molecular domestication, four at the base of jawed vertebrates and a fifth one later in a common ancestor of sarcopterygians for MSANTD3. Phylogenies were also constructed with the Maximum Likelihood method and showed similar results (supplementary fig. 2, Supplementary Material online). However, in this analysis, MSANTD3 and MSANTD4 did not clearly group with a specific Harbinger transposon, and the clustering with transposon was not highly statistically supported for NAIF1. Hence, Maximum Likelihood analysis suggested at least three events of molecular domestication, two for MSANTD1 and MSANTD2 and at least a third one for MSANTD3, MSANTD4 and NAIF1.
Phylogenetic relationships between MSANTD proteins and their closest Myb-like proteins from Harbinger transposons. Trees were constructed using the Bayesian method (Huelsenbeck and Ronquist 2001). Only branch support values higher than 50% are shown. AG, Anopheles gambiae; AM, Alligator mississippiensis; BF, Branchiostoma floridae; CG, Crassostrea gigas; CM, Chelonia mydas; CP, Chrysemys picta; DR, Danio rerio; EL, Esox lucius; GA, Gasterosteus aculeatus; GG, Gallus gallus; HS, Homo sapiens; LC, Latimeria chalumnae; MM, Mus musculus; NV, Nematostella vectensis; OL, Oryzias latipes; ON, Oreochromis niloticus; OS, Oryza sativa; PG, Puccinia graminis; PS, Pelodiscus sinensis; PSt, Puccinia striiformis; SS, Salmon salar; TF, Takifugu flavidus; VV, Vitis vinifera; XT, Xenopus tropicalis). Silhouette images from phylopic.org.
In order to test if some MSANTD genes might have been formed through larger segmental genomic duplications, their flanking genomic regions were compared by synteny analyses. No evidence for paralogous sequences that might have been co-duplicated with MSANTD genes was found, consistent with more local events (fig. 2B).
Taken together, the results indicated that MSANTD1, MSANTD2, MSANTD3 and MSANTD4 are four new cases of vertebrate genes derived from Harbinger Myb-like transposon sequences, in addition to NAIF1. These genes arose from three to four independent molecular domestication events at the base of jawed vertebrates around 500 Mya, with another potential one at the base of sarcopterygians around 430 Mya that generated MSANTD3.
Vertebrate Myb-Like-Derived Genes Evolved under Negative Selection
To further investigate the evolutionary constraints having acted on vertebrate MSANTD genes, we performed a positive/negative selection test using CODEML (Yang et al. 2007). We calculated the dN/dS ratio (ratio between nonsynonymous vs. synonymous substitution rates) as a proxy for selection pressure (supplementary Table 1, Supplementary Material online). All ratios were smaller than 1, reflecting a higher rate of synonymous than nonsynonymous substitutions. These ratios were comparable to those of other genes in genomes (0.066 on average in human-zebrafish comparisons) (Wolf et al. 2009). Hence, MSANTD sequences evolved under negative/purifying selection in vertebrates, i.e., these genes were functionally constrained. The ratio for the MSANTD2P1 pseudogene was closer to 1 compared to other MSANTD genes, in accordance with relaxed constraints and loss of protein-coding capacity.
Harbinger-Derived Genes Are Expressed in Zebrafish during Embryonic Development and Predominantly in Adult Male Brain
The expression of the MSANTD1, MSANTD2, MSANTD4, NAIF1 and HARBI1 Harbinger-derived genes was studied by quantitative PCR (qPCR) in zebrafish embryos (fig. 4A). All these genes were expressed during zebrafish embryonic development, with HARBI1 being the most expressed gene. Except for MSANTD1, which is more expressed at later stages, most genes were more strongly expressed at the first stages of development before the midblastula transition (MBT), suggesting maternal effect. Using in situ RNA hybridization, MSANTD2, which was chosen for further functional analyses (see below), was found to be expressed during zebrafish embryonic development in the whole embryo from 1.25 h post fertilization (hpf) to 17hpf (fig. 4C). From 6hpf, MSANTD2 was more strongly expressed in the anterior side of the embryo, the region leading to the head and the central nervous system. At 24hpf, the expression of MSANTD2 was specifically restricted to the head region, and more particularly to the forebrain, midbrain and hindbrain regions of the brain.
Expression of Harbinger-derived genes in zebrafish embryos and adults. (A) Relative gene expression determined by qPCR compared to the 18S housekeeping gene during embryonic development (from 1hpf to 3dpf). The MBT is shown with an arrow. (B) Relative gene expression determined by qPCR compared to the 18S housekeeping gene in female (F) and male (M) adult tissues. (C) Whole-mount in situ hybridization of MSANTD2 in zebrafish embryos from 1.25hpf to 24hpf using a sense probe as a control. The head or future head region is indicated with ‘h’. fb, forebrain; mb, midbrain; hb, hindbrain. Scale bars: 200 μm.
Expression of Harbinger-derived genes was also studied in zebrafish adult tissues by qPCR (fig. 4B). We observed both a sex- and tissue-biased expression of these genes. Particularly, Harbinger-derived genes were predominantly co-expressed in male but not female brain. As observed in embryos, HARBI1 was also the most expressed gene in adult tissues, with stronger expression in liver and muscle of both males and females.
Harbinger-Derived Genes Are Expressed in Human, Particularly during Brain Development
According to the National Institutes of Health (NIH) genotype-tissue expression (GTEx) project (dbGaP Accession phs000424.v8.p2; GTEx Consortium 2013), all Harbinger-derived genes appeared to be expressed in human brain as well as in some other tissues depending on the gene. Using the BrainSpan Atlas of the Developing Human Brain (www.brainspan.org) (Miller et al. 2014), expression was detected in human brain before and after birth, except for the MSANTD2P1 pseudogene (fig. 5). MSANTD genes were expressed in the whole brain particularly during early fetal development at the first/second trimesters of pregnancy, with decreasing expression in the third trimester (around 10 weeks before birth) (fig. 5A). MSANTD3 and MSANTD4 were the most expressed genes and MSANTD2 and NAIF1 presented the same expression pattern but with lower expression. HARBI1 had a more ubiquitous expression, with higher expression in early fetal development as observed for the MSANTD genes, but also later after 13 years. MSANTD1 presented a more localized expression in a specific brain structure, the striatum, during the second trimester of pregnancy (from 13 to 24 postconceptional weeks [pcw]) (fig. 5B). These results showed that Harbinger-derived genes are particularly expressed during fetal brain development, in whole human brain for most genes or in a more specific brain region (striatum) for MSANTD1.
Expression of Harbinger-derived genes in human brain before and after birth according to donor stages (A) or brain structures (B). For each gene, the expression is shown in log2 reads per kilobase per million (RPKM) for different donor stages (pcw, postconceptional weeks; mos, months; yrs, years) and in different brain structures, represented with multiple colors. Data were obtained from the BrainSpan Atlas (www.brainspan.org) (Miller et al. 2014). The striatum-specific expression of MSANTD1 is indicated with a red arrowhead. Silhouette images are from lifesizesilhouette.com and (Haniffa et al. 2021).
MSANTD2 Inactivation Leads to Zebrafish Embryos with Severe Neuro-Developmental Defects
The biological function of vertebrate Harbinger-derived genes was further investigated by gene inactivation. Gene knock-out was achieved with CRISPR/Cas9 technique and knock-down using morpholino antisense oligonucleotides (Nasevicius and Ekker 2000). The CRISPR/Cas9 protocol was adapted from Wu et al. (2018) in order to produce null phenotypes in the F0 generation of zebrafish, i.e., directly in injected embryos (Wu et al. 2018). This protocol implies the co-injection of four different sgRNAs targeting the same gene at four different loci. In a first screening, single gene inactivation of HARBI1, MSANTD1, MSANTD4 and NAIF1 did not produce any clear and visible phenotype in zebrafish embryos, possibly explained by gene redundancy. Therefore, we focused our analyses on the MSANTD2 gene, which has been suggested to be associated to human neuro-developmental diseases such as autism spectrum disorders and schizophrenia (Schizophrenia Working Group of the Psychiatric Genomics Consortium 2014; Lim et al. 2017; O’Brien et al. 2018; Zhang et al. 2020).
In order to inactivate MSANTD2 by CRISPR/Cas9, zebrafish embryos were injected with one sgRNA or different combinations of two or four sgRNAs. The observed phenotypes were similar with almost all the different combinations of sgRNAs, although the penetrance was variable (fig. 6G; supplementary fig. 3, Supplementary Material online). Therefore, only three treatments (combination of four sgRNAs, combination of sgRNAs 1 and 4, sgRNA 4 only) will be further detailed (fig. 6G). Sequencing of injected embryos revealed mutations at all sgRNA loci, with multiple frameshift nucleotide deletions leading to premature stop codons (supplementary figs. 4–5, Supplementary Material online). When looking at embryos injected with four sgRNAs, read coverage analysis showed that almost all reads (ca. 90%) showed mutations in the first exon at the sgRNA 1 locus, in addition to mutations at the three other sgRNA loci.
MSANTD2 CRISPR/Cas9 embryo phenotypes. Embryos injected with control (A–C, CTRL) or four MSANTD2-directed sgRNAs (D–F, MSANTD2). A/D and B/E present lateral views of whole embryos and of the head region, respectively. C/F show dorsal views of the embryo head region. Embryos injected with four sgRNAs showed developmental delays as well as tail and nervous system malformations compared to control embryos at 24hpf (A–F). Tail curvation and not well-defined somites are shown with black arrowheads (D). Default in neural tube folding and cell aggregates around the nervous system are indicated with white arrowheads and black arrows, respectively (E–F). Not well-formed MHB are shown with white stars (*) (E–F). (G) Proportions of F0 zebrafish phenotypes. Embryos were injected with sgRNA 1 + 2 + 3 + 4, sgRNA 1 + 4 and sgRNA 4 targeting MSANTD2, four scrambled sgRNAs or noninjected (WT), and were scored for phenotypes at 24hpf. Percentages with strong (red), intermediate (orange) or no phenotypes (gray–WT) are shown. Strong phenotypes: developmental delays, tail malformations, nervous system malformations, aggregates; intermediate phenotypes: developmental delays, nervous system malformations, no or few aggregates, no tail malformation. (H–O): Dextran Texas Red injection in brain ventricles of 24hpf (H–K) or 30hpf (L–O) embryos injected with control (H, L, CTRL) or four MSANTD2-directed sgRNAs (I–K, M–O, MSANTD2) (dorsal view). At each stage, three pictures of MSANTD2 CRISPR/Cas9 embryos were compared to a control picture in order to represent phenotype variability. The proportion of embryos presenting the same phenotypes are the following: I—31%, J—47%, K—22% at 24hpf and M—14%, N—43%, O—43% at 30hpf. All cases illustrate neural tubes misfolding (shown with white arrowheads) particularly in the MHB region. We also observed smaller red fluorescent areas in the forebrain and midbrain regions, indicating reduction of these ventricles. F, forebrain, M, midbrain; H, hindbrain. Scale bar 200 µm.
Embryos injected with four sgRNAs showed developmental delays as well as tail and nervous system malformations compared to control embryos at 24hpf (fig. 6A-F). Heads were smaller and tails curved with not well-defined somites (fig. 6D, black arrowheads). Moreover, MSANTD2 CRISPR/Cas9 embryos presented defects in neural tube folding (fig. 6F, white arrowheads), with cell aggregates (i.e., cells that are loosely grouped together) visible around the nervous system (fig. 6E–F, black arrows). Similar phenotypes were observed when embryos were injected with sgRNAs 1 and 4 together, and intermediate phenotypes with sgRNA 4 alone (supplementary fig. 6, Supplementary Material online). In a typical experiment (fig. 6), about 25% of embryos injected with all four sgRNAs presented strong phenotypes (developmental delays, tail malformations, nervous system malformations, cell aggregates) and around 50% intermediate phenotypes (developmental delays, nervous system malformations, no or few cell aggregates, no tail malformation) (fig. 6G). Moreover, MSANTD2 inactivation appeared to severely compromise development, since 40–85% of injected embryos with phenotypes died few days or weeks post injection (compared to 6–16% for control embryos). Similar phenotypes were also observed after injection of morpholino antisense oligonucleotide directed against MSANTD2 (supplementary fig. 7, Supplementary Material online). Developmental delays as well as tail and nervous system malformations were observed in MSANTD2 CRISPR-Cas9 zebrafish embryos at 15hpf and 19hpf too (supplementary fig. 8, Supplementary Material online).
In order to assess the specificity of the phenotypes observed, we performed a rescue experiment by co-injection of MSANTD2 mRNA with the four MSANTD2-directed sgRNAs in zebrafish embryos (supplementary fig. 9, Supplementary Material online). We observed that the mutant phenotypes were strongly rescued, since most embryos presented wild-type (WT) phenotypes or only slight developmental delays. Particularly, we observed a well-developed nervous system in the rescued embryos.
Concerning the nervous system, the midbrain-hindbrain boundary (MHB), which is a well-defined structure of the 24hpf stage of zebrafish development, was not well formed (fig. 6A–F, white stars). In order to characterize nervous system malformations, dextran Texas Red was injected into zebrafish brain ventricles at 24hpf and 30hpf (fig. 6H–O) (Gutzman and Sive 2009). At each stage, three pictures of MSANTD2 CRISPR/Cas9 embryos were compared to a control. All MSANTD2 CRISPR/Cas9 embryos presented brain abnormalities with neural tubes misfolding particularly in the MHB region (fig. 6H–O, white arrowheads), as well as defects in forebrain, midbrain and hindbrain inflation. These phenotypes were observed from 24hpf and were still present at 30hpf. Dead cells were marked in 24hpf embryos with acridine orange staining (fig. 7A–D). We observed numerous dead cells in MSANTD2 CRISPR/Cas9 embryos compared to control embryos. Hence, the cell aggregates we observed might correspond to dead cell areas (fig. 7C–D, regions of cell aggregates are indicated with white stars).
Effects of MSANTD2 inactivation by CRISPR/Cas9 in zebrafish embryos. (A–D) Acridine orange staining of embryos injected with control (CTRL) or four MSANTD2-directed sgRNAs (MSANTD2) at 24hpf. Numerous dead cells were visible in MSANTD2 CRISPR/Cas9 embryos compared to control embryos. Regions of cell aggregates were indicated with white stars. (E–I) Embryos injected with control (CTRL) or four MSANTD2-directed sgRNAs (MSANTD2) in Tg(elavl3:GCaMP3) zebrafish embryos. At each stage, two pictures of MSANTD2 CRISPR/Cas9 embryos were compared to a control in order to represent the variability of impaired neuronal pattern. Differentiated neurons were marked by green fluorescence. The results showed general anatomical defects characterized by different patterns of fluorescence between MSANTD2-mutated and control embryos (shown with white arrowheads). Abnormal pattern of neuronal marking was not only explained by developmental delay, since MSANTD2 mutated embryos are still different from control embryos at later stages. (J–N): Expression of FGF8 (J, K), HER5 (L), PAX2A (M) and DLX2 (N) at 24hpf in embryos injected with control (CTRL, top) or four MSANTD2-directed sgRNAs (MSANTD2, bottom). J present lateral and K/L/M/N dorsal views of the head region of the embryos, respectively. Differences in gene expression patterns between MSANTD2-mutated and control embryos are indicated with black arrowheads. Dd, dorsal diencephalon; hn, hindbrain neurons; oc, otic capsule; os, optic stalks; ov, otic vesicle; pa, pharyngeal arches; pr, proximal part of retina; t, telencephalon; tp, thyroid primordium. For each picture at least n = 10 embryos were observed.
To further study the role of MSANTD2 in nervous system development, this gene was inactivated using the same protocol of CRISPR/Cas9 with four sgRNAs in the Tg(elavl3:GCaMP6s) zebrafish line (fig. 7E–I). This is a transgenic line containing a green fluorescent protein (GFP)-based calcium sensor, with the elav3 promoter fused to the GCaMP6s genetically encoded calcium indicator, marking fluorescently all differentiated neurons (Park et al. 2000; Panier et al. 2013). At each stage, two pictures of MSANTD2 CRISPR/Cas9 embryos were compared to a control to assess the variability of neuronal patterns. From 24hpf to 72hpf, general anatomical defects characterized by aberrant patterning of early neurons were visible in MSANTD2 CRISPR/Cas9 embryos (fig. 7E–I, white arrowheads). Moreover, the abnormal pattern of neuronal marking was not only explained by developmental delay, since MSANTD2 mutated embryos were still different from control embryos at later stages of development.
We studied in 24hpf MSANTD2 CRISPR/Cas9 embryos the expression of characteristic genes involved in vertebrate nervous system development (fig. 7J–N). Because the MHB is a well-defined structure at the 24hpf stage of zebrafish development, we characterized it with the expression of the FGF8, PAX2A and HER5 genes. FGF8 encodes a fibroblast growth factor involved in several processes, including nervous system development particularly for two brain structures, the tectum and cerebellum. FGF8 is also responsible for the maintenance of the MHB together with PAX2A (Nakamura 2001; Chi et al. 2003). In 24hpf control embryos, FGF8 was expressed in telencephalon, dorsal diencephalon, optic stalks, otic vesicle and MHB regions. In addition to MHB, PAX2A was expressed in optic stalks, otic vesicle and hindbrain neurons at 24hpf (fig. 7J,K,M). The MHB is also characterized by the expression of HER5, which is involved in multiple developmental processes and particularly brain development (fig. 7L). Finally, as we observed dead cell aggregates around the nervous system in the MSANTD2 CRISPR/Cas9 embryos (figs. 6 and 7), we questioned whether they might correspond to dead neural crest cells by studying the expression of DLX2, which marks the cranial migratory neural crest cells that form the pharyngeal arches and migrate into the forebrain (Akimenko et al. 1994; Yan et al. 2005; Sperber et al. 2008; Dai et al. 2013).
In MSANTD2 CRISPR/Cas9 embryos, we observed alterations of FGF8, PAX2A and HER5 expression particularly in the MHB region (fig. 7J–M). These genes were still expressed, but the marked areas were different between control and MSANTD2 CRISPR/Cas9 embryos. The expression bands were narrower for FGF8 and PAX2A (fig. 7K,M) and also less deep for FGF8 (fig. 7J). For HER5, the staining into two distinct areas was lost in MSANTD2 CRISPR/Cas9 embryos (fig. 7L). Moreover, the expression of FGF8 in telencephalon and optic stalks was not separated into two different zones but formed a unique and larger area, suggesting defects in the definition and individualization of these structures (fig. 7J,K). MSANTD2 CRISPR/Cas9 embryos lacked the expression of PAX2A in hindbrain neurons (fig. 7M). Finally, the expression of DLX2 was markedly reduced in the telencephalon and pharyngeal arch regions (fig. 7N). In conclusion, these results indicated anatomical defects of the MSANTD2 CRISPR/Cas9 embryos in the MHB and telencephalon regions. Finally, the accumulation of dead cells and the expression patterns of PAX2A and DLX2 also suggested potential implication of MSANTD2 in neural crest cell migration, homing or differentiation into neurons (visible for hindbrain neurons).
SOX10 (SRY-box transcription factor 10) is a marker of neural crest cells, particularly marking the neural crest cell streams at 15hpf and 19hpf (supplementary fig. 8, Supplementary Material online). The most anterior stream (S1), second stream (S2) and third stream (S3) correspond to the cranial neural crest cells that migrate into the mandibular arch, the hyoid arch and posterior pharyngeal arches, respectively (Piotrowski and Nüsslein-Volhard 2000). Cranial neural crest cells can then give rise to multiple derivatives: chondrocytes, osteocytes, cranial sensory ganglia, pigment cells, connective tissue, Schwann and satellite cells (Schilling and Kimmel 1994; Kague et al. 2012). Using a zebrafish Tg(sox10:mRFP) transgenic line, in which SOX10 was fluorescently labeled, we observed defects in SOX10 marking in MSANTD2 CRISPR/Cas9 embryos at both 15hpf and 19hpf, particularly in the S1 stream region. Moreover, we also detected impaired marking in the optic tectum region, a multi-tissue part of the midbrain involved in the visual system. These results suggest potential migratory delays or defects of the cranial neural crest cells in MSANTD2 CRISPR/Cas9 mutant embryos, potentially leading to altered functions of their cellular derivatives.
Discussion
Harbinger Transposons Have Given Rise to a New Gene Family through Recurrent and Concomitant Molecular Domestication Events in Vertebrates
In this work, we report recurrent and concomitant molecular domestication of Harbinger transposons in early vertebrate evolution. We have identified in jawed vertebrates a new family of genes derived from Harbinger elements, and more particularly from their Myb-like gene (figs. 1–3). Indeed, MSANTD1, MSANTD2, MSANTD3 and MSANTD4 presented sequence similarities with the Myb-like domain of proteins from Harbinger transposons (fig. 1). Each MSANTD gene is present as a single copy gene at a conserved position in vertebrate genomes (fig. 2). Hence, MSANTD genes are not transposons anymore but bona fide vertebrate genes. In order to investigate whether MSANTD genes arose from independent molecular domestication or/and sequential duplication events, sequence alignments and phylogenies were constructed (figs. 1 and 3). The sequences of all MSANTD genes and proteins could not be aligned unambiguously together due to high divergence. Indeed, the Myb-like proteins of different families of Harbinger transposons present significant similarities only restricted to a short part of their Myb-like domain. In contrast, Harbinger transposases are much more conserved. After constructing separate phylogenies through Bayesian analyses for MSANTD1/MSANTD2 and MSANTD3/MSANTD4/NAIF1, respectively, we observed that orthologous MSANTD sequences formed monophyletic groups and that for each of them the closest Harbinger transposon was different. However, Maximum Likelihood analysis failed to support some of these preferential phylogenetic relationships between MSANTD genes and Harbinger transposons. Taken together, we propose that the vertebrate family of Myb-like genes derived from Harbinger transposons originated from three to five independent molecular domestication events. Three to four domestications probably occurred at the base of jawed vertebrates about 500 Mya, and a more recent might have led to the formation of MSANTD3 at the base of sarcopterygians approx. 430 Mya.
Interestingly, after comparing 675 human proteins annotated as containing a Myb-related domain with Harbinger Myb-like proteins (data not shown), we identified four additional putative cases of Harbinger-derived genes in the human genome: MYPOP, ZSCAN20, PRDM11 and TSNARE1, TSNARE1 having already been suggested to originate from a Harbinger transposon (Smith et al. 2012). Even if these genes require further analysis, this suggests the presence of additional genes derived from the Myb-like genes of Harbinger transposons in vertebrate genomes.
A processed pseudogene resulted from a duplication by MSANTD2 mRNA retrotransposition (supplementary fig. 1, Supplementary Material online). Duplicated transcribed pseudogenes can directly regulate related functional genes by transcriptional interference through the production of small interfering RNAs, or by recruiting factors silencing the protein-coding gene transcript (Sen and Ghosh 2013). However, MSANTD2P1 is not expressed in human brain, and no more information is available on its expression. Its function, if any, remains to be further investigated.
Harbinger DNA transposons have given rise to multiple novel genes in divergent organisms: at least six in vertebrates, nine in Arabidopsis and seven in Drosophila (Kapitonov and Jurka 2004; Casola et al. 2007; Sinzelle et al. 2008; Liang et al. 2015; Duan et al. 2017; Velanis et al. 2020; Zhou et al. 2021). Hence, it seems that these elements have high propensity to be recruited as new genes. The characteristics of Harbinger transposons, with their two protein-coding open reading frames (ORFs), may be an advantage. These two ORFs encode proteins with domains with widespread functions. Particularly, the Myb-like domain, a DNA- and protein-binding domain, could be repurposed in diverse ways for gene regulation as a transcription factor and/or as a member of a protein interactome (Sinzelle et al. 2008). In addition, separation of the different molecular activities, i.e., DNA breaking/recombination and DNA/protein binding, in two independent ORFs is uncommon for TEs. This might allow a more specific co-option by the host of a single molecular activity without interference of the other.
Harbinger-Derived MSANTD Genes Encode Potential DNA- and Protein-Binding Proteins
We observed that the secondary structure (tri-helix motif) of the Harbinger Myb-like protein has been conserved in the different MSANTD proteins, suggesting conservation of the original molecular properties (fig. 1). Generally, the SANT/myb/trihelix motifs have been shown to have DNA- and protein-binding capacities in multiple transcription factors (Boyer et al. 2004). The Harbinger Myb-like protein is able to bind both the transposon DNA and the transposase protein (Sinzelle et al. 2008). Thus, the MSANTD proteins could act as DNA- and protein-interactors. Accordingly, NAIF1, like the Harbinger Myb-like protein, is able to bind DNA and interact with the Harbinger transposase as well as with the transposase-derived HARBI1 protein (Sinzelle et al. 2008). MSANTD3 has been suggested to work as a transcription factor that binds to DNA, where it can recruit the Polycomb Repressive Complex 2 to regulate neuronal differentiation in P19 mouse cells (Gou 2014). Outside of vertebrates, other genes derived from Harbinger transposons have been identified in Arabidopsis (Liang et al. 2015; Duan et al. 2017; Velanis et al. 2020; Zhou et al. 2021). ALP1 (Antagonist of Like Heterochromatin Protein 1), its paralog HHP1 (HDA6-associated Harbinger transposon-derived Protein 1) and HDP1 (Harbinger transposon-derived protein 1) are derived from transposases, while ALP2 (Antagonist of Like Heterochromatin Protein 2), HDP2 (Harbinger transposon-Derived Protein 2), SANT1, SANT2, SANT3, and SANT4 have been formed from Harbinger Myb-like genes. ALP2 and HDP2 interact with ALP1 and HDP1, respectively, and are involved in chromatin modifying complexes (Liang et al. 2015; Duan et al. 2017; Velanis et al. 2020; Zhou et al. 2021). ALP1 and ALP2 mediate Polycomb Repressive Complex 2 formation (Velanis et al. 2020). HDP1 and HDP2 are part of a histone acetyltransferase complex acting in DNA methylation through the DNA-binding capacity of HDP2 (Duan et al. 2017). Similarly, HHP1, SANT1, SANT2, SANT3, and SANT4 belong to a HDA6 histone deacetylase complex controlling flowering time (Zhou et al. 2021).
Overall, multiple genes derived from Harbinger transposons encode proteins that have kept the DNA- and protein-binding capacities ancestrally present in the transposon Myb-like proteins. Therefore, the MSANTD genes identified in this study may encode transcription factors or other proteins with DNA- and protein-binding activities.
Harbinger-Deriving Genes Are Expressed in Developing and Adult Vertebrate Brain
Expression results indicated that Harbinger-derived genes are expressed in zebrafish during embryonic development, particularly before the MBT for most of them, suggesting potential maternal effect. These genes are also transcribed in adult tissues. We observed that HARBI1 is generally expressed at a higher level than the MSANTD genes. This could favor HARBI1 interaction with multiple MSANTD proteins, as demonstrated with NAIF1 (Sinzelle et al. 2008), particularly in the brain. HARBI1 might also have MSANTD-independent functions, as suggested by the absence of co-expression in some other tissues.
Harbinger-derived genes are expressed in multiple tissues in zebrafish (fig. 4) and human (GTEx Consortium 2013). However, as observed in zebrafish adult male brain, we detected a common expression of Harbinger-derived genes in human brain particularly during early fetal development (figs. 4B and 5), which might favor functional interactions of their proteins in this organ in vertebrates. Harbinger-derived genes are predominantly expressed from 8–9 pcw through the two first trimesters of fetal development. Around 8–9 pcw, a process called neuronal migration starts in fetal brain (Métin et al. 2008; Rahimi-Balaei et al. 2018). Neurons are formed in the neuroepithelium, a neural tube layer, during embryonic development. Neuronal migration corresponds to the processes by which neurons will migrate from their germinal layer to all over the central nervous system, where they will establish connections with other cells. As more and more neurons migrate to their final localization, the different brain structures start to be formed throughout the first and second trimesters of fetal development. Disturbance of neuronal migration has been associated to neurological disorders such as schizophrenia, autism spectrum disorders and epilepsy (Fatemi 2005; Guerrini and Parrini 2010; Muraki and Tanigaki 2015; Pan et al. 2019).
MSANTD1 presents a striatum-specific expression during the second trimester of fetal development in human. The striatum is part of the basal ganglia brain structure, mainly involved in voluntary motor control and related to rewards in social conditions (Báez-Mendoza and Schultz 2013). The general role of the basal ganglia on movement control is conserved in vertebrates (Grillner et al. 2013).
Together, the redundant expression in zebrafish and human brain suggests the potential implication of Harbinger-derived genes in vertebrate nervous system development, potentially in neuronal migration (fig. 5). This is also compatible with works suggesting association of MSANTD2 to schizophrenia and autism spectrum disorders in human (Schizophrenia Working Group of the Psychiatric Genomics Consortium 2014; Lim et al. 2017; O’Brien et al. 2018; Zhang et al. 2020).
MSANTD2, a Gene Involved in Vertebrate Nervous System Development
In order to better understand the biological roles of Harbinger-derived genes in vertebrates, we have further analyzed the effects of the inactivation of the MSANTD2 gene in zebrafish. Expression analyses revealed MSANTD2 expression in brain during development in human but also in zebrafish at 24hpf (figs. 4 and 5). This suggested a possible function of MSANTD2 in vertebrate nervous system development.
Inactivation of MSANTD2 by CRISPR/Cas9-direct gene knock-out in zebrafish produced embryos with severe developmental delays as well as tail and nervous system malformations (fig. 6A–F). We identified defects in neural tube folding, resulting in impaired ventricle formation in forebrain, midbrain and sometimes hindbrain regions (fig. 6H–O). These structural malformations were linked to cellular defects, as we observed accumulation of dead cells and multiple abnormalities in neuronal marking from 24hpf that lasted at least until 72hpf.
In MSANTD2 CRISPR/Cas9 embryos we observed modified expression patterns for the FGF8, DLX2, PAX2A and HER5 genes, which are involved in vertebrate nervous system development. These results revealed brain, and particularly MHB organization defects. Moreover, we found accumulation of dead cells in MSANTD2 CRISPR/Cas9 embryos (fig. 7A–D). The altered expression of DLX2, a gene involved in cranial migratory neural crest cell development, suggested that dead cell accumulation could correspond to neural crest cells. Neural crest cells contribute to multiple cell lineages, including sensory and automatic neurons, glia cells, pigment cells and chondrocytes (Iulianella and Trainor 2003). In zebrafish, cranial neural crest cell migration starts around 13hpf (Rocha et al. 2020). In the Tg(sox10:mRFP) transgenic line, in which SOX10 was fluorescently labeled, we observed defects in SOX10 marking in MSANTD2 CRISPR/Cas9 embryos at both 15hpf and 19hpf, suggesting migratory delays or defects of cranial neural crest cells (supplementary fig. 8, Supplementary Material online). Furthermore, the expression of MSANTD2 in human brain during fetal development in a time lapse where neuronal migration arises, as well as the aberrant pattern of early neurons in MSANTD2 CRISPR/Cas9 zebrafish embryos, might support a role of MSANTD2 in neural crest cell or neuron migration.
The phenotypes observed in our analysis correspond to MSANTD2 F0 generation mutants. Reproducible phenotypes were obtained with different combinations of sgRNAs as well as with morpholino oligonucleotides in a gene knock-down approach. CRISPR-Cas9 mutant phenotypes were rescued by MSANTD2 mRNA. Finally, inactivation of other Harbinger-derived genes in zebrafish did not produce similar phenotypes, indicating specificity of the phenotypes observed for MSANTD2. Hence, in addition to the strong mortality of MSANTD2 mutated embryos, these results strongly support a role for this gene in the development of the vertebrate nervous system.
Conclusion
Vertebrate early evolution has been marked by the emergence of multiple major innovations, which have contributed to the evolutionary success of this lineage. Indeed, vertebrates present new and complex organs, which have allowed the improvement of their movement, sensing and adaptation to their environment. For example, vertebrates have a complex nervous system, which is composed of cranial nerves, spinal cord, ganglia and a brain organized in specialized regions. Bones, cartilages, paired appendages, a complex endocrine system, sensory placodes, the neural crest and an adaptive immune system are also major novelties acquired during early vertebrate evolution.
Ohno proposed that whole genome duplications, generating an extensive expansion of gene repertoires, are major events giving rise to massive innovations and important evolutionary transitions (Ohno 1999). Accordingly, two events of genome duplications have taken place at the base of vertebrates (Dehal and Boore 2005). However, new gene formation by duplication is not the unique mechanism allowing the apparition of major novelties. SYNCYTIN genes, involved in placenta formation in mammals, as well as RAG genes, implicated in the adaptive immune system in vertebrates, testify of the role of TE-derived novel genes in organismal innovation.
In this work, we propose that Harbinger-derived genes could have been contributors of early vertebrate evolution, notably through their role in the evolution of the nervous system development. Further analyses should look at the implication of other TE molecular domestication events in the emergence and evolution of other vertebrate innovations. Hence, the study of TE molecular domestication provides us with important clues on the functional and evolutionary characteristics of new genes, with a broader picture of the genetic basis and dynamics of the emergence and evolution of phenotypic traits.
Materials and Methods
Zebrafish Maintenance
Zebrafish of the strain AB/TU were raised according to standard procedures (PRECI, SFR Biosciences [UAR3444/CNRS, US8/INSERM, ENS de Lyon, UCBL]). Embryos were raised at 28°C. Developmental stages were expressed in hours post-fertilization (hpf) or days post-fertilization (dpf) based on morphological criteria (Kimmel et al. 1995). The Tg(elavl3:GCaMP6s) transgenic line, containing a modified GCaMP (GCaMP is a genetically-encoded calcium indicator) known as GCaMP3 under elavl3 regulatory region (ZFIN ID: ZDB-TGCONSTRCT-180326-1), were also raised according to the same procedures (Park et al. 2000; Panier et al. 2013). The zebrafish Tg(sox10:mRFP) transgenic line, obtained from Florence Ruggiero team (Institute of Functional Genomics, Lyon, France), was also raised according to the same standard procedures. The SOX10 protein, which is expressed in neural crest cells, is fluorescently marked in this transgenic line.
In Situ Hybridization
In situ hybridization (ISH) probes for MSANTD2 were cloned from WT zebrafish cDNA by PCR using the GoTaq polymerase (Promega). PAX2A and HER5 probes were given by Dr. Sebastian Dworkin lab, La Trobe University, Melbourne, Australia.
Zebrafish embryos were collected, removed from their chorion, sorted and fixed in paraformaldehyde (PFA) 4%, dehydrated in methanol and stored at—20°C. ISH was performed following the Thisse Lab protocol (Thisse and Thisse 2008; 2014). Embryos were rehydrated and washed in phosphate buffered saline (PBS)—Tween (PBT) solution. They were permeabilized with proteinase K and fixed in PFA. Each embryo was incubated with probes overnight at 65°C in hybridization mix supplemented with 5% Dextran Sulfate (Millipore). Nonhybridized probes were removed with several washes in formamide and saline-sodium-citrate solutions. Embryos were incubated overnight at 4°C with α-DIG (digoxigenin) antibodies (Roche). Nonfixed antibodies were removed with PBT washes. Probes were revealed with nitro blue tetrazolium chloride - 5-bromo-4-chloro-3-indolyl-phosphate (NBT-BCIP) (Roche). Embryos were fixed in PFA. After removing of the background with ethanol bath, embryos were stored in glycerol 80% at 4°C. Pictures were taken under Leica stereomicroscope and Keyence VHX-7000 microscope.
qPCR
Pools of 3–5 zebrafish adults and 15–20 embryos were used for RNA extraction. RNAs were extracted with Trizol according to the Bio-Rad company protocol and treated with DNaseI. Reverse transcription was performed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific). The following specific primers were designed: for NAIF1 TGAATCACTTTAACGCGGGC, CCGTCTTCAGATCCGACCAT; for HARBI1 CGCTGCGTTTCTAACGTCAC, AGAGTCATCCGCATTGGGAG; for MSANTD1 CAAACCTCTCATCGTCTGGC, AGGCCGTCATCCTCATCATT; for MSANTD2 AGACCCGAGTTCTTCAGATACGAC, GAGAGAAGTCCGTCCACGTTTG; for MSANTD4 TCAAGATGGAGGACGACGAG, GGGAGGATGGAGGGAAAACA. qPCR was performed using SYBR Green following the Bio-Rad protocol. 18S housekeeping ribosomal RNA gene (TCGCTAGTTGGCATCGTTTATG, CGGAGGTTCGAAGACGATCA) was used to normalize gene expression. Results were analyzed with the ΔCt method (Schmittgen and Livak 2008).
Morpholino Knockdown
Two nonoverlapping morpholino antisense oligonucleotides targeting the 5’-UTR of MSANTD2 (GCCATCTTGCTTCTGTTGCTAAGGG, CAGACACGACTGACGGCTTCTTATG) and a control mismatched morpholino (CACACACCAGTGACGCCTTGTTATG) were purchased from Gene Tools and injected from 0.2M to 3M into one-cell embryos (Nasevicius and Ekker 2000). Morphants and mismatched controls were matched per cross. Morphological and phenotypic observations were performed at 1dpf under Zeiss Axio Zoom microscope.
CRISPR/Cas9
For each gene four nonoverlapping single guide RNAs (sgRNAs) were purchased from Synthego (supplementary fig. 2, Supplementary Material online). The sequences of the sgRNAs were selected from Wu et al. 2018 (Wu et al. 2018). For all sgRNAs used, no putative off target site with four or less mismatches and next to a PAM could be detected in the zebrafish genome using the CRISPOR program (Concordet and Haeussler 2018). The Cas9-GFP protein was purchased from TacGene. A mix of four sgRNAs (20 to 30 µM in total) and Cas9-GFP protein (5 to 15 μM) was injected into WT embryos at the one-cell stage (individual sgRNAs as well as combinations of two sgRNAs were also injected). For MSANTD2, sgRNA 1 and sgRNA 2 were located in the first exon and sgRNA 3 and sgRNA 4 in the third and fourth exons, respectively. For each gene knock-out, the experiment was performed at least in duplicate (at least six times for the mix of 4 sgRNAs). Scrambled (random sequence) sgRNAs were used as a negative control and sgRNAs targeting the tyrosinase (TYR) gene as a positive control (its inactivation led to individuals without melanic pigmentation). Crispants and scrambled controls were matched per cross. Embryo survival and phenotypic observations were monitored from 6hpf under Leica stereomicroscope and Zeiss Axio Zoom microscope.
MSANTD2 Mutant Rescue Experiment
MSANTD2 mRNA was transcribed in vitro (mMessage Machine SP6 Transcription Kit, Ambion) from the corresponding cDNA sequences in the pcS2 + vector synthetized by GenScript. For microinjections, the Cas9-sgRNAs mix was combined in a 1:1 volume ratio with MSANTD2 mRNA at a 100 ng/µl concentration. The presence of mutations at the MSANTD2 sgRNA loci in the rescued embryos was verified by Sanger sequencing. For that, injected embryos were collected and DNA extractions were conducted for the three conditions (MSANTD2 sgRNA 1 + 2 + 3 + 4, MSANTD2 sgRNA 1 + 2 + 3 + 4 + MSANTD2 mRNA, scrambled). Multiple pics (mutations) were observed after sequencing at sgRNA loci for the MSANTD2 sgRNA 1 + 2 + 3 + 4 and MSANTD2 sgRNA 1 + 2 + 3 + 4 + MSANTD2 mRNA conditions compared to the scrambled condition, for which only unique pics were observed (no mutations).
Brain Ventricle Imaging
Zebrafish brain ventricle injection was performed according to the protocol developed by Gutzman and Sive (2009). Briefly, embryos were anesthetized with Tricaine (Sigma). Micro-injection was performed in hindbrain ventricle with 1–10nl of dextran Texas Red (5% in 0,2mol/L KCl, Invitrogen). 15 to 30min after injection, images were taken with transmitted and fluorescent lights under a Zeiss AxioZoom Microscope.
Acridine Orange Staining
Embryos were dechorionated and stained with 10μg/mL acridine orange solution for 30 min. Then, embryos were washed three times in E3 medium. Images were taken with transmitted and fluorescent lights under a Zeiss AxioZoom Microscope.
DNA Extraction, PCR Amplification, NGS Sequencing and Sequencing Data Analyses
In order to search for mutations after application of the CRISPR/Cas9 protocol, injected embryos were collected at 24hpf and five DNA extraction replicates were conducted starting from a single embryo for the four conditions (4sgRNAs, sgRNA1-4, sgRNA4 and scrambled). Lysis of embryos was performed in lysis buffer (10mM Tris-HCl pH8–2 mM EDTA pH8–0.2% Triton x-100) with 250 µg/µl proteinase K (Invitrogen) 12 h at 55°C, followed by proteinase K inactivation of 10 min at 95°C. Three fragments of the MSANTD2 gene (exon 1, exon 3 and exon 4) were amplified by PCR. PCR reactions were performed in 25 µl using the GoTaq G2 DNA polymerase kit (Promega), 2 µl of DNA extract and 0.5 μM of each primer set with the following PCR program: 2 min at 94°C, 35 cycles at 94°C 30 s, 60°C 30 s and 72°C 30 s, with a final extension step at 72°C for 5 min. For each condition and for each exon, five PCR tubes (each PCR corresponding to one embryo DNA amplification) were pooled. PCR product purification was carried out according to manufacturer’s recommendations (Nucleospin Gel and PCR Clean-up, Macherey Nagel) and eluted in 30 μl of elution buffer (NE buffer). For each condition, equimolar amounts of the three purified amplicons were used to create a barcoded library with an input of 50ng using the NEBNext Ultra II DNA Library Prep Kit protocol for Illumina. Quantitation and quality assessment of each library was performed on a 4150 Tapestation analyzer using the High Sensitivity D5000 ScreenTape kit (Agilent Technologies). Libraries were mixed at the same equimolar proportions, spiked with approximately 5% PhiX control and sequenced with the Illumina MiSeq sequencer using the Nano Kit v2 reagent (pair-end reads, R1 and R2 read lengths, 260 bp and 259 bp respectively). More than 800K reads were obtained and analyzed using the Galaxy platform (Afgan et al. 2018) using the FastQC, Cutadapt, Bowtie2 and Sort tools to assess the quality of reads, remove adapter sequences, map reads against reference and store aligned sequences, respectively.
TE and Gene Sequence in Silico Analyses
TE-derived genes were identified through sequence similarity with TE sequences from the Repbase database (www.girinst.org) and from annotation of various sequenced vertebrate genomes (Chalopin et al. 2015) using blastp, blastn and tblastn (Altschul et al. 1990). Additional Harbinger Myb-like protein sequences were recovered through blast analysis of the NCBI Genomes (RefSeq Genomes) database (www.ncbi.nlm.nih.gov) using MSANTD and Harbinger transposon sequences as queries. Blast was used with default parameters. A first threshold was set for protein-protein comparisons (E-value <10e-5), which was followed by manual inspection of sequence alignments on the whole predicted sequences, inspection of predicted conserved protein motifs and construction of molecular sequence phylogenies (see below). NCBI, Ensembl, Censor (www.girinst.org) and Genomicus (Muffato et al. 2010) were used to determine the copy number, sequence alignments, phylogeny and synteny of TE-derived genes. Conserved proteins motifs were detected with the NCBI Conserved Domain Search (Marchler-Bauer et al. 2011), InterPro (Apweiler et al. 2000) and PROSITE (Sigrist et al. 2002). Genes and ORFs were predicted with Augustus (Stanke and Morgenstern 2005) and ORFfinder (www.ncbi.nlm.nih.gov/orffinder). NPS-PRABI (Combet et al. 2000) and Jpred4 (Drozdetskiy et al. 2015) were used to predict the secondary structure of proteins. For positive selection tests, the protein-coding sequence of genes from different species were collected from the Ensembl database, aligned as proteins using MUSCLE (Edgar 2004) and then converted back into a nucleic sequence alignment. A phylogenetic tree was then built with the PhyML package (see below). Positive/negative selection tests were performed using CODEML (Yang 2007). The tests were run based on an alignment of coding sequences from spotted gar, zebrafish, tetraodon, stickleback, platyfish, coelacanth, chinese softshell turtle, mouse, macaca, marmoset, human, chimpanzee and chicken.
For phylogenetic analysis, nucleotide and amino-acid sequences were aligned with MAFFT (Katoh et al. 2002). Phylogenetic trees were built using maximum likelihood with PhyML (Guindon and Gascuel 2003) and with MrBayes (Huelsenbeck and Ronquist 2001) using a mixed model (estimated by the Protest-3 software (Darriba et al. 2011) and 500,000 generations of Bayesian inferences.
Gene accessions numbers are HGNC:33741, HGNC:26266, HGNC:23370, HGNC:29383, HGNC:25446 and HGNC:26522 for MSANTD1, MSANTD2, MSANTD3, MSANTD4, NAIF1 and HARBI1, respectively. Transposon sequences and alignments are available upon request.
Supplementary Material
Supplementary data are available at Molecular Biology And Evolution online.
Acknowledgments
This work was supported by grants from the Agence Nationale de la Recherche (ANR) and from the Ecole Normale Supérieure de Lyon. We acknowledge the contribution of the PRECI fish facility of the SFR Biosciences (UAR3444/CNRS, US8/INSERM, ENS de Lyon, UCBL) and of the IGFL’s PSI Sequencing platform. We are grateful to Marilyne Malbouyres, Sandrine Bretaud and Florence Ruggiero (Matrix biology and pathology team, Institut de Génomique Fonctionnelle de Lyon) for their help in Crispr-Cas9 methodology, and to Marie Sémon (ENS Lyon) and Christoph Winkler (University of Singapore) for discussions. EE thanks the ‘Fondation pour la Recherche Médicale’ (FRM) for financial support through an end of PhD program fellowship.
Author Contributions
Experiments were designed by E.E. and J.N.V. and performed by E.E., manuscript was drafted by E.E. and amended by J.N.V., the project was supervised by J.N.V. and co-supervised by D.B., M.N. and Z.H.T.
Data Availability
Data are available on request.
