Manipulation of the Tyrosinase gene permits improved CRISPR/Cas editing and neural imaging in cichlid fish

Direct tests of gene function have historically been performed in a limited number of model organisms. The CRISPR/Cas system is species-agnostic, offering the ability to manipulate genes in a range of models, enabling insights into evolution, development, and physiology. Astatotilapia burtoni, a cichlid fish from the rivers and shoreline around Lake Tanganyika, has been extensively studied in the laboratory to understand evolution and the neural control of behavior. Here we develop protocols for the creation of CRISPR-edited cichlids and create a broadly useful mutant line. By manipulating the Tyrosinase gene, which is necessary for eumelanin pigment production, we describe a fast and reliable approach to quantify and optimize gene editing efficiency. Tyrosinase mutants also remove a major obstruction to imaging, enabling visualization of subdermal structures and fluorophores in situ. These protocols will facilitate broad application of CRISPR/Cas9 to studies of cichlids as well as other non-traditional model aquatic species.

repair the chromosome. NHEJ is an imprecise process that often leads to insertions or deletions (indels). If an indel results in a frameshift mutation of the coding sequence of a gene, a loss-of-function mutation is likely to result. HDR utilizes a DNA template to guide repair, offering the ability to insert exogenous DNA sequences. In practice, however, template insertion via HDR is difficult to achieve in fish species, though some progress has been made [18][19][20][21][22] . We focus on the use of CRISPR/Cas to generate loss-of-function (LoF) mutations, as they can be rapidly and inexpensively created, and these mutations provide insights into development, physiology, behavior, and evolution.
In this paper we generate LoF mutations in the Tyrosinase (Tyr1) gene, which encodes a key enzyme for eumelanin synthesis. This strategy has been useful in zebrafish, medaka, and lizard mutagenesis [23][24][25] . The mutant phenotype of nonpigmented melanocytes is easily visible early in development and stable in adults, making Tyr1 gRNA microinjections an efficient tool to assess mutagenesis rates and troubleshoot CRISPR/Cas9 protocols. Further, these eumelanin-deficient mutants enable analysis of visual signaling and facilitate in vivo activity-based imaging of both embryonic and adult brains. Here we outline the processes that we have developed to generate mutant and transgenic cichlids, including genetic target selection, reagent preparation, embryo manipulation, genotyping, and husbandry. These approaches should be applicable to species beyond A. burtoni, including cichlids more generally and other fish species that can be raised in aquatic laboratories.
The general approach to gene editing in cichlids is described here, using a targeting strategy for a pigmentation gene, Tyrosinase (Fig. 1). First, we collect and inject single-cell embryos with CRISPR gRNA and Cas9 to induce DNA breaks and subsequent mutations in the gene of interest. Then we screen injected fish by PCR to identify animals carrying indels at the desired site and select genetic mosaic animals to breed. Offspring of injected animals are screened for mutations to identify heterozygous founders of mutant lines. We cross heterozygous animals to generate homozygous mutant cichlids (plus wildtype and heterozygous controls) that we assay for phenotypes of interest.
The success of gene editing depends on the efficiency of several steps. First, a large number of embryos should be obtained for injection because there is embryo death due to natural causes as well as the injection itself. Further, the gene target may not be mutated in the germline, which is necessary to establish a genetic line. Finally, not all mutations are likely to be of large effect, and thus it is desirable to recover frameshift mutations that are large and therefore facilitate easy genotyping. Here, we describe approaches to optimize each step and maximize the number of edited animals recovered. The genome editing workflow, summarized in Fig. 1, consists of the design and synthesis of gRNA and genotyping primers, spawning and egg collection, micro-injection of gRNA/ Cas9, determining and quantifying cut efficiency of gRNA, and recovery and maintenance of mutant lines. While we focus on the generation of loss-of-function alleles through frameshift mutations, the approaches here can be generalized to the creation of other mutant lines. The detailed protocols provided in supplementary materials will improve gene editing efficiency for a variety of cichlid species.

Results
Optimizing cichlid embryo recovery and survival rates. To improve gene editing, we first sought to maximize recovery of fertilized single-cell embryos. We asked whether spawning could be facilitated by intraperitoneal injections of Ovaprim (Syndel), a commercially available mixture of gonadotropin-releasing hormone analogue and dopamine receptor antagonist 26 . We found that Ovaprim injection ( Fig. 2A) shortens the time to egg laying by ~ 5 days (Fig. 2B). It also increases the yield of eggs laid by ~ 2-fold (Fig. 2C,D), without affecting egg viability (Fig. 2E). Interestingly, the increased fecundity effect of Ovaprim is observed in the following ovarian cycle, though period of the cycle returns to normal.
Generation of knockout or transgenic lines requires a reliable source of embryos. We designate several 80-120 L tanks that house a single sexually mature male and a cohort of sexually mature wildtype females, separated by a barrier to control the onset of spawning (Fig. 3A). Because A. burtoni have an approximately 1-month long ovarian cycle 27 , we include sexually mature females in sufficient numbers (10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20) to increase the likelihood of a spawning. We have historically observed a high, but variable rate of embryo death after injection of CRISPR components. We hypothesized that standardizing the method by which embryos are manipulated prior to injection and/or irregularities in microinjection needles may contribute to this variability. First, we designed a 3D-printed mold for an agarose embryo holder to secure the embryos during injections (Fig. 3B,C). This reduced and standardized manual handling compared to prior approaches 28 . This had the benefit of improving speed of embryo injection by ~ 2-fold. We find that embryo survival in the agarose wells is comparable to unhandled embryos (survival rate for uninjected embryos in beaker, 67 ± 4%; in wells 63 ± 4%; p = 0.363, Student's t test).
Microinjection needles are generated from glass capillary tubes (Fig. 3D,E) and breakage yields needles of variable size. Variation in microinjection needles could impact both the volume of reagents injected as well as the extent of damage to the embryo. To test whether this variation contributes to embryonic death rate, we screened needle widths, and correlated these to survival rate. We found that bores larger than 12.5 µm are associated with lower post-injection survival rates (Fig. 3G). Thus, needle widths from 7.5 to 12.5 µm permit maximal survival. Other factors associated with needle shape could also contribute to embryo death rate. For example, the angle or unevenness of the needle tip breakage (Fig. 3E) could affect death rate, but these were not quantified. Punctured embryos may release some contents of the yolk after injection (Fig. 3F). We observed that leakage of yolk during injection often results in low survival (Fig. 3I), which may be due to internal pressure caused by large embryos fit into small holes in the injection mold. Hence, we designed our injection mold with varied sizes of pegs of increasing diameter from 1.81 to 2.13 mm, allowing us to fit each embryo into an appropriately sized well. www.nature.com/scientificreports/ gene editing parameters. The Tyrosinase (Tyr1) gene has been shown to be necessary in medaka, zebrafish, and anole lizards for producing eumelanin from tyrosine 24,25,29 . We synthesized gRNAs that target the intramelanosomal, common central domain of Tyr1, a sequence identical at the nucleotide level across cichlids (Fig. 5A). We injected these along with Cas9 protein into single-cell embryos (Fig. 3H). We monitor embryo survival for approximately 10 days post-fertilization (dpf) to identify issues that contribute to low survival rates (Fig. 3J). By 3 days post-fertilization, uninjected embryos developed black, ramified melanocytes. In contrast, we found that ~ 50% of injected embryos exhibited a reduction in the number of melanocytes (Fig. 4A), suggesting that Tyr1 gene function is necessary for eumelanin synthesis in cichlids. We were surprised to find ramified, fluorescent cells scattered across the surface of Tyr1 CRISPR-injected cichlid embryos, but not in control embryos (Fig. 4D). These cells also appeared at 2-3 dpf, and their number was inversely proportional to the number of black melanocytes. This suggests that in the absence of functional Tyr1, melanocytes accumulated a fluorescent metabolite derived from tyrosine. Further, we speculate this is not tyrosine itself due to its green, rather than blue, fluorescence emission 30 . As CRISPR-injected embryos are mosaics of cells bearing different mutations or unmodified alleles, the conversion of melanocytes from black (Tyr-positive) to fluorescent (Tyr-mutant) permits a quantification of gene editing efficiency. To test whether this metric reflected genomic mutation rate, we quantified the ratio of melanocyte conversion and obtained genomic DNA from embryos to assess indel rates. We adapted a PCR amplification approach 31 to evaluate mutation induction in injected embryos using PCR size analysis. This approach quantitatively and rapidly detects size polymorphisms that result from indel mutations (Fig. 4C), the sequence changes most likely to affect gene function. We found a positive correlation between indel rate and black-to-fluorescent melanocyte conversion (Fig. 4D). Notably, melanocytes can be rapidly counted, and can be performed at an early stage. It thus will enable a rapid exploration of CRISPR parameters, and to troubleshoot problematic reagents.

Manipulation of the cichlid
Previous work using CRISPR/Cas to generate mutant cichlids utilized single guide RNAs (sgRNAs) that incorporate both crRNA and tracrRNA [32][33][34] . Recent work shows that using separate crRNA and tracrRNA as is utilized naturally in S. pyogenes (dual-guide RNA system; dgRNA) drives more efficient gene editing in zebrafish than does sgRNA 35 . We synthesized sgRNA and dgRNA ( Supplementary Fig. 1A) that target the same Tyr1 site to directly test whether this finding holds true in cichlids. We find high gene-editing rates at the Tyr1 locus with either sgRNA or dgRNA systems ( Supplementary Fig. 1B).
Homozygous Tyr1 mutant cichlids enable in situ imaging. Tyr1 CRISPR-injected fish remain mosaic animals through adulthood (Fig. 4B). We crossed injected fish to wild-types and screened offspring using Tyr1 locus-specific fragment analysis. We identified a fish heterozygous for a 20 bp deletion at Tyr1 (Tyr1 d20/+ ) (Fig. 5C). An intercross of Tyr1 d20/+ animals yielded 1/4 of offspring devoid of eumelanin from 2 dpf through We injected Ovaprim (0.5 μL per g of fish) into females 10 days after spawning and monitored for spawning behavior in the following two reproductive cycles. Females were monitored every day until they were observed carrying a brood. (B) Ovaprim advanced the reproductive cycle around 5 days, but timing of the following cycle was unchanged. (C) Ovaprim increased number of eggs laid compared with saline-injected females, an increase observed in the following reproductive cycle as well. (D) Ovaprim increased number of eggs laid, even when controlling for mass of female. (E) Ovaprim did not affect the overall survival rates of the embryos. Two-tailed Mann-Whitney U test was used to compare groups. www.nature.com/scientificreports/ adulthood (Fig. 5B). In place of black melanocytes, fluorescent cells can be readily observed from 2 to 10 dpf (Fig. 5D). The eumelanin deficiency remains a stable into adulthood (Fig. 5B), implying that developmental patterning signals from melanocytes remain intact 36 .
The absence of eumelanin in Tyr1 d20/d20 cichlids also permits unobstructed visualization of subdermal features (Fig. 6A,B). As proof-of-principle, we unilaterally injected one eye of larval cichlids with the lipophilic fluorescent tracer DiI to label retinal ganglion cell (RGC) projections to the brain. After 3 days of diffusion, we imaged the (J) Embryo development from 2 dpf (days post-fertilization) to 5 dpf. Note that eumelanin pigmentation on yolk appears at 2 dpf; pigmented optic cups with lens placodes become discernible at 4 dpf; tail region separates from yolk (hatching) by 5 dpf. www.nature.com/scientificreports/ optic tectum, to visualize RGC targets in situ. Imaging revealed dense RGC axonal arbors that wrapped around the dorsal and lateral edges of the tectum contralateral to the injected eye (Fig. 6C,D).

Discussion
Many fundamental insights about evolution, development, physiology, and medicine have been derived from the study of organisms that are not regularly genetically modified. The advent of the CRISPR/Cas system enables reverse genetic approaches in species beyond traditional species such as Drosophila and mouse. Our protocol utilizing this system efficiently edits the genome in the non-traditional model species, A. burtoni. This African cichlid is a model organism for studies from behavioral neuroscience to ecology and evolution; CRISPR/Cas provides genetic insights into mechanisms of development and physiology. We targeted Tyr1 gene as an example, and further demonstrated that CRISPR/Cas9 editing rates at the Tyr1 locus as determined by tissue biopsy are correlated with the deficiency in melanin. The mechanism by which the mutation leads to the loss of pigmentation remains unclear; while we suggest a frameshift 5' to a critical copper binding domain is causative, alterations in splicing could contribute. Nonetheless, the externally visible phenotype can be screened in the embryos as early as 2-3 dpf, which makes the Tyr1 gRNA-injection a fast and reliable tool to quickly validate CRISPR/Cas activity in vivo, and to troubleshoot problematic reagents and protocols.
Tyrosinase mutant animals will be broadly useful as models for in vivo neural imaging. Zebrafish (Danio rerio) are commonly used as models for live neural imaging in a variety of contexts due to the transparency of www.nature.com/scientificreports/ larvae [37][38][39][40] . Further, recent advancements in gene editing have facilitated the development of novel genetic lines that circumvent the interference of skin pigmentation in adults 39,41,42 . Our Tyr1 d20/d20 line also lacks melanocyte pigmentation, opening the prospect of in vivo neuroimaging in A. burtoni. Our DiI tracer injections demonstrate the ability to map the neural arborizations of retinotectal connections in an intact cichlid. These approaches have the potential to reveal important insights on the neural pathways of information processing and evolutionary differences across species. Beyond simply facilitating development of CRISPR/Cas9 protocols, the Tyr1 mutation provides an opportunity for exploration of behavior and neurophysiology. A. burtoni males participate in a complex social hierarchy in which dominant and subordinate males differ in pigmentation, physiology, and reproductive opportunities 14 . One of the starkest distinguishing features of dominant males is the "eyebar": a vertical line of melanin-rich cells across the cheek (Fig. 5B, top). Melanosomes in eyebar cells rapidly disperse and contract, enabling dynamic responses to social stimuli, and is thought to be used as a signal between rivals 43,44 . Because Tyr1 d20/d20 individuals lack all eumelanin (Fig. 5B, bottom), use of mutant males in territory defense assays enables manipulation of this signaling between males. Similarly, Tyr1 d20/d20 males will make it possible to determine whether females rely on these pigmentation patterns as a mate quality cue. This protocol for using CRISPR is well-suited to creating loss-of-function mutations, but there is also a need for efficient, targeted introduction of specific gene editing by knock-in mutations 45 . This approach can test how a genomic region has evolved to differentially regulate a phenotype across species by transferring a sequence from one species to the orthologous locus in another. Knock-in mutations also enable the expression of a transgene using the complete gene regulatory environment of a locus, permitting faithful recapitulation of transgene expression. Such a system would enable conditional gene manipulation as well. For example, the Cre/loxP system enables site-specific or temporally delimited mutations 46 , but loxP sites must be inserted into the genetic locus of interest, and Cre lines benefit from the faithful gene expression described above. Progress has been made in targeted sequence insertions 19,20,22 , aided by chemical treatments and specific sequence features that increase HDR efficiency [47][48][49][50] . Homology-independent approaches for sequence knock-in may be more attractive given natural biases towards NHEJ over HDR 51 , and have demonstrated success in zebrafish and medaka 20,52,53 . Testing for repair of the mutated A. burtoni Tyr1 locus will provide a platform on which cichlid knock-in experiment conditions may be rapidly optimized. Correct insertion of a repair template will restore pigmentation in the Tyr1 d20/d20 fish, providing a visible, quantifiable readout of gene modification. As an alternative to genome editing by knock-in to obtain conditional mutants, transposon-based systems have the potential to permit cell-type specific manipulations. Prior work in cichlids has shown that the Tol2 transposon system efficiently catalyzes transgene insertion and permits cell-type-specific transgene expression 28,[54][55][56][57][58] . This system could be adapted to express CRISPR components using cell type-specific promoter sequences, thereby facilitating spatially restricted mutagenesis 59 .
A major advantage of cichlids as research models is the diversity of form and function across species. However, benefits accrue to research communities from a focus on a limited number of species. This allows the sharing of transgenic or CRISPR-edited lines, leveraging the size of the community to increase the number of experiments Adult Tyr1 d20/d20 animals (bottom) exhibit no eumelanin compared to wild type animals (top). (C) PCR fragment analysis shows that Tyr heterozygous mutants (Tyr1 d20/+ among F 1 offspring) exhibit one additional peak resulting from indel mutation (middle) compared to wild type (top); while Tyr homozygous mutants only have one peak corresponding to a 20 bp deletion (bottom). Wild type peak is denoted by asterisk. (D) A. burtoni homozygous for Tyr1 d20/d20 mutation exhibit a loss of pigment in eye as well as neural crest-derived melanocytes that cover the yolk and body. This phenotype is observable starting from 2 dpf (Top panel). The converted melanocytes which exhibit green fluorescence are also observed starting from 2 to 3 dpf (bottom panel). www.nature.com/scientificreports/ Figure 6. Tyrosinase mutant A. burtoni permit brain imaging in situ. (A) Dorsal view of the skin covering the midbrain in wildtype (left) and Tyr1 d20/d20 (right) fish at 9 dpf, imaged using transmitted light. Pigmented melanocytes are absent above the brain in Tyr1 d20/d20 fish, enabling underlying imaging of brain structures. (B) Dorsal view (left) and lateral view (right) of the head structure with schematic brain illustrations in Tyr1 d20/d20 larvae at 9 dpf. Note that in the DiI injection experiment, those images were focused on dorsal and lateral edges of optic tectum (regions in red squares). E eyes, Tel telencephalon, TeO coptic tectum, Cb cerebellum, HB hindbrain. (C,D) Contralateral retinotectal projections were visualized by confocal microscopy in 9 dpf Tyr1 d20/d20 cichlids after unilateral eye injection with DiI. Z-projections from dorsal (338 µm depth) (C,D, left) and lateral (304 µm depth) aspects (C,D, right) reveal retinal ganglion cell axons terminating throughout the peripheral optic tectum. Different colors depict the depth of brain tissues from the center surface (red) to the edge of the optic tectum (magenta). Scale bars 200 µm. www.nature.com/scientificreports/ possible using existing lines. Furthermore, a focus on a limited set of species increases the available experimental tools, bioinformatic resources, and more 17 . A. burtoni is a well-suited model system as there is already a wealth of data and experimental protocols developed. In addition, as a basal Haplochromine cichlid, it is an ideal species with which to test hypotheses generated in the speciose radiations of Haplochromines in Lakes Malawi and Victoria. Thus, genetic engineering in A. burtoni promises to reveal novel genetic mechanisms for phenotypic diversity in behaviors, physiology, and anatomy.

Methods
Animals. Fish were bred and used at the University of Maryland from a colony derived from Lake Tanganyika 60 , according to the guidelines of the University of Maryland animal care and use committee. Tanks received recirculated water with constant pH (8.0-8.2) and salinity (320-480 ppm). Fish were housed in 30 L tanks prior to embryo collection, then in 6-well plates and 1.2 L tanks during growth. All animal maintenance and experimental procedures were approved by the Institutional Animal Care and Use Committee of the University of Maryland (protocol 1046257-1). The study was carried out in compliance with the ARRIVE guidelines.
Fecundity testing. For each trial, we selected two size-match adult females (mass differences < 1.0 g), and each was injected intraperitoneally with either Ovaprim (Syndel) or saline at 0.5 μL per gram of body mass at 10 days after previous spawning. We then tagged injected females with visible elastomer tags (Northwest Marine Technologies) to distinguish treatments. After injection, fish were monitored in an isolated 2 L tank 30 min before returned to community housing. These females were co-housed with a dominant male which was separated by a transparent divider. We removed the divider for 30 min and monitored females for spawning behavior daily across the following two reproductive cycles, until they were observed carrying a brood. Embryos were collected 30 min after initial spawning and fertilization, and their survival was tracked until 12 dpf.

CRISPR injections.
Microinjection needles are generated from glass capillary tubes (GC100F-10, 1.0 mm O.D; 0.58 mm I.D; Harvard Apparatus) by first using a micropipette puller (Sutter, P-97) to separate the tube into two needles and then opening the tip by gently tapping the needle on a taut Kimwipe to break the tip to ~ 10 µm diameter (see Supplementary Methods Microinjection of Cichlid Embryos for detail). Needle bore width was defined as the long axis of the needle's angled opening (not the distance orthogonal to the axis of the tube). Resulting needles are screened by microscopy to confirm sizes are ≤ 12.5 µm. All oligonucleotides were ordered from IDT; crRNA and tracrRNA were synthesized with AltR modifications. sgRNA was synthesized as in 31 by annealing an oligo with gene-specific sequence Tyr1 sgRNA2 with a universal lower oligo (

Oligonucleotide name Sequence (5'-3') Notes
TyrFlankF -M13  TGT AAA ACG ACG GCC AGT cag gtt ttg caa gtc cac aga c  M13 binding site in uppercase   TyrFlankR-pigtail  GTG TCT Ttc ttt ctc act gca tta cac cc  Pigtail sequence in uppercase   Tyr1 sgRNA2  TTA ATA CGA CTC ACT ATA ggt tcc atg tca tca gta caG TTT TAG AGC TAG AAA TAG C  Tyr1 binding sequence in lowercase   Universal lower oligo  AAA AGC ACC GAC TCG GTG CCA CTT TTT CAA GTT GAT AAC GGA CTA GCC TTA TTT TAA CTT  GCT ATT TCT AGC TCT AAA  www.nature.com/scientificreports/ mutation prevalence and identifying mutant cichlids for details). Peak prevalence relative to wild type was analyzed using Peak Scanner and fragmentanalysis.com. Briefly, we quantified allele frequency within the mosaic tissue by using Peak Scanner to measure the area under peaks of each amplicon length. The ratio of the area under the wildtype peak to sum of areas under all peaks represents the fraction of unmodified alleles.

Melanocyte quantification.
To quantify black and fluorescent melanocytes, Tyr1 gRNA injected embryos at 5 dpf were chilled in ice-cold tank water for 30 s to immobilize them, and then were transferred to a 6 well plate to detect green fluorescence under a stereomicroscope (Nikon, SMZ18). We used a soft paintbrush (Robert Simmon, size 1) to flip the embryos to ensure that we counted all melanocytes. After melanocyte quantification, we obtained genomic DNA by removing ~ 1 mm 3 of posterior tissue from each embryo for fragment analysis for assessing indel rates. The level of deficiency in melanin was calculated as the number of fluorescent melanocytes divided by the number of total melanocytes (Fig. 4). We further categorized these Tyr1-injected embryos by pigmentation phenotypes using a fluorescence stereomicroscope (Leica M165 FC) using band-pass filter cubes for the detection of green fluorescence (ET GFP-M205FA/M165FC Excitation: 470/40, Emission: ET525/50). If the percentage of fluorescent (converted) melanocytes cells is smaller than 10%, the embryo is categorized as 'normal pigmentation'; if the percentage of fluorescent melanocytes cells is between 10 and 50%, then it is categorized as 'sparse pigmentation'; if the percentage of fluorescent melanocytes cells is larger than 50%, then it is categorized as 'significantly diminished pigmentation' (Supplementary Fig. 1B).
Retinotectal projection labeling. We fixed larval cichlids at 9 dpf overnight in 4% paraformaldehyde and washed them in PBS. DiI (Life Technologies V22889) was pressure injected to fill the eye cavity unilaterally and allowed to diffuse for 3 days at 28 °C. We imaged tectal RGC arborizations in situ using a Zeiss LSM800 confocal microscope and a 10 × 0.5NA objective.