A method for CRISPR/Cas9 mutation of genes in fathead minnow (Pimephales promelas)

Abstract

Clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 genome editing allows for the disruption or modification of genes in a multitude of model organisms. In the present study, we describe and employ the method for use in the fathead minnow (Pimephales promelas), in part, to assist in the development and validation of adverse outcome pathways (AOPs). The gene coding for an enzyme responsible for melanin production, tyrosinase, was the initial target chosen for development and assessment of the method since its disruption results in abnormal pigmentation, a phenotype obvious within 3 to 4 d after injection of fathead minnow embryos. Three tyrosinase-targeting guide strands were generated using the fathead minnow genome browser (setac.org/fhmgenome) in tandem with CRISPOR v. 4.0 (http://crispor.tefor.net, Haeussler et al. 2016). The strands targeted two areas: one stretch of sequence in a conserved region that demonstrated homology to EGF-like or laminin-like domains as determined by Protein Basic Local Alignment Search Tool (BLASTp) in concert with the Conserved Domain Database, and a second area in the N-terminal region of the tyrosinase domain. To generate one cell embryos, in vitro fertilization was performed, allowing for microinjection of hundreds of developmentally-synchronized embryos with Cas9 proteins complexed to each of the three guide strands. Altered retinal pigmentation was observed in a portion of the tyr guide strand injected population within 3 d post fertilization (dpf). By 14 dpf fish without skin and swim bladder pigmentation were observed. Among the three guide strands injected, the guide targeting the EGF/laminin-like domain was most effective in generating mutants. CRISPR greatly advances our ability to directly investigate gene function in fathead minnow, allowing for advanced approaches to AOP validation and development. The technique could also produce novel molecular level insights into the basic biology of fathead minnow and other fishes.

1. Introduction

The precision genome editing technology, clustered regularly interspaced short palindromic repeats (CRISPR/Cas9) is revolutionizing the life sciences. CRISPR RNA strands can be designed to guide Cas9 protein to specific sequences of DNA. Upon binding, the Cas9 enzyme makes a double-stranded cut in the region targeted. When the organism’s primary natural repair system, the non-homologous end joining (NHEJ) pathway, repairs the double-stranded DNA break, various base insertion and deletions (indels) may be generated and sustained by the genome, producing premature stop codons that result from frameshift mutations. There are wide ranging applications of this technology. For example, if applied early in development, ideally in the single cell stage of an embryo immediately after fertilization, the mutation can be transmitted to all cells in the developing organism. Assuming the mutation results in a loss of gene and/or protein function, the consequences of the gene modification can be studied to gain significant insights into the function of specific gene/protein targets of interest.

While the use of CRISPR/Cas9 has exploded in many areas of biological research, utilization in the aquatic toxicology research community has been more limited. In part, this is due to the field’s primary focus on examining the adverse, apical effects of chemical and non-chemical stressors on species of concern. Nonetheless, investigators have started to employ CRISPR to improve understanding of mechanisms of toxicity and the role of specific proteins in mediating adverse effects of chemicals. For example, CRISPR/Cas9 approaches were used to generate aryl hydrocarbon receptor (AhR2) mutations in both zebrafish (Danio rerio) and Fundulus heteroclitus to better understand the mechanisms underlying toxicity mediated by AhR2 (Aluru et al. 2015; Garcia et al. 2018). The technology has also been used to explore the role of multi-resistance protein 1 (Mrp1) in cadmium and benzo[a]pyrene-exposed zebrafish embryos (Tian et al. 2017).

The objective of the present study was to provide a foundation for further exploring the utility of CRISPR/Cas9 for mechanistic ecotoxicology research. While CRISPR has now been successfully applied in a number of fish species including zebrafish, medaka (Fang et al. 2018), and rainbow trout (Cleveland et al. 2018), we aimed to develop methods to support the use of CRISPR/Cas9 in the fathead minnow (Pimephales promelas), a model small fish that has been widely used in regulatory aquatic toxicity testing for decades (Ankley and Villeneuve, 2006). As a proof of concept, we used CRISPR/Cas9 to create a targeted mutation in the gene coding for tyrosinase (tyr), an enzyme essential to melanin production. Disruption of tyr has resulted in abnormal pigmentation, an easily observable phenotype, in previous CRISPR/Cas9 knockout studies with zebrafish and medaka (Jao et al., 2013; Ota and Kawahara, 2014; Fang et al. 2018). Based on these results and sequence similarities between zebrafish and fathead minnow, we expected to observe abnormal pigmentation in fathead minnow embryos as early as 3 to 4 d after injection. Because tyrosinase mutation leads to a readily observed phenotype, screening an injected population for mutations can easily be achieved, thereby simplifying the process of evaluation and optimization. As part of this work, methods for in vitro fertilization that allowed reliable production of single-cell stage embryos for injection were developed. After injecting with appropriately designed guide strands and gaining proficiency in fathead minnow embryo microinjection techniques, tyr mutants were generated. Mutations identified based on the altered pigmentation phenotype were verified at the genetic level by sequencing analysis.

2. Materials and Methods

2.1. CRISPR/Cas9 crRNA strand design for tyrosinase

For effective gene knockout, it is often necessary to screen multiple guide strands for the gene target. Ideally, guide strands should base pair in the coding region, sufficiently downstream of alternative start codons and far from the C-terminal end of the protein. Targeting the first 5–65% of exonic sequence (Doench et al., 2016, and Sanson et al., 2018) generally allows for many options given the prevalence of protospacer adjacent motif (PAM) sites for Cas9 binding. In the present study, to increase the likelihood of an effective knockout, three guide strands were designed to target three unique regions of the tyr sequence and deployed simultaneously. The fathead minnow tyr sequence was obtained from the fathead minnow genome browser (Saari et al., 2017; GenBank accession #MH392277; Supplementary Information), and three tyr-targeting CRISPR RNA (crRNA) strands were generated using CRISPOR v. 4.0 (http://crispor.tefor.net/, Haeussler et al. 2016). The selected strands targeted exonic sequence within roughly the first half of the gene, with one strand targeting the laminin-type epidermal growth factor-like (EGF-like) domain and the other two strands targeting the tyrosinase domain (Fig. 1). Mutation by insertion/deletion (indel) using CRISPR/Cas9 can disrupt the gene in a manner that either destroys the corresponding protein’s function when mutation occurs in a crucial area, or results in frameshift mutations and premature stop codons that prevent the protein from being translated in its entirety (Doudna, J.A. and Jiang, F. 2017).

Figure 1.

Zebrafish tyrosinase protein with features labeled. SP = signal peptide, EGF-like = laminin-type epidermal growth factor-like domain, tyrosinase domain, and TM = transmembrane domain. A, B, and C label the locations where guide strands A, B, and C base pair with the corresponding tyrosinase DNA.

Each Alt-R® CRISPR-Cas9 crRNA strand included a twenty nucleotide, target-specific sequence on the 5’ side of an NGG protospacer adjacent motif (PAM) site (Table 1). Strands were purchased through Integrated DNA Technologies (https://www.idtdna.com). The three guide strands used (A, B and C) had high specificity scores (96–99%), and 17, 10 and 8 predicted off-target hits, respectively, in fathead minnow (http://crispor.tefor.net/, Haeussler et al. 2016). Among strand A’s off-target hits, one was a three base mismatch and sixteen of them possessed four mismatches. All of strand B and C’s off-target hits contained four mismatches. In addition to considering the off-targets in fathead minnow, strands were cross-referenced with those generated at http://crispr.mit.edu for zebrafish. The selected strands scored best in both analyses, with high specificity to the desired target and minimal off-target effects.

Table 1.

Fathead minnow crRNA strands selected for RNP preparations to knockout tyrosinase using the CRISPR/Cas9 system (http://crispor.tefor.net/, Haeussler et al. 2016). The guides’ specificity scores range from 0 (worst) to 100 (best) and indicate guide quality, including possible off-target complementarity. The number of off target sites from within the fathead minnow genome are noted (off-target counts). Each of these off-target areas possess three or four mismatches to the twenty base crRNA sequence.

Strand	crRNA sequence	Specificity score (%)	Off-target counts
A	AACUACAUGGGCAUCGACUG	97	17
B	GCGUGUACGAUUUGUUCGUG	96	10
C	CGACGACUACAACCUUCGAG	99	8

Three ribonucleoproteins (RNPs) targeting tyr were prepared and co-injected using a protocol adapted from Integrated DNA Technologies (https://www.idtdna.com/pages/docs/default-source/catalog-product-documentation/user-method_crispr-cas9_zebrafish_v7_web.pdf?sfvrsn=8). In short, a 3 μM guide RNA (gRNA) solution was prepared by combining 0.5 μL of 100 μM Alt-R^(R) CRISPR-Cas9 crRNA, 0.5 μL of trans-activating crRNA (tracrRNA), and 16 μL nuclease-free duplex buffer. To prepare an empty vector negative control injection solution, the crRNA was replaced with nuclease-free duplex buffer. The solutions were complexed by incubating at 95°C for 5 min and subsequently cooled to 25°C. Ribonucleoproteins for the three target regions were assembled into one batch by combining 3 μL of each of the three gRNA solutions (9 μL total) with 9 μL of 0.5 μg/ μL Alt-R® S.p. Cas9 nuclease (Integrated DNA Technologies), incubating at 37°C for 10 min, and then cooling to 25°C. An injection solution was prepared by adding 1 μL of 1% phenol red to 9 μL of the triple RNP solution, resulting in a 0.45 μM final concentration of each tyr RNP. Empty vector negative control injection solution was prepared similarly by combining 3 μL of the crRNA-free gRNA solution and 3 μL of Cas9 solution, incubating, and adding 0.7 μL of 1% phenol red.

2.2. In vitro fertilization and microinjections

Control adult fathead minnows were obtained from an on-site culture facility at the US Environmental Protection Agency (USEPA) in Duluth, MN. All procedures were conducted in accordance with approved Animal Care and Use Guidelines. Reproductively mature adults and embryos were housed in a flow-through system with control UV-treated, filtered Lake Superior water (LSW). Water temperature was maintained at 25±1^ᵒC, and photoperiod was set to 16:8 light:dark. Animals were fed brine shrimp ad libitum. An in vitro fertilization (IVF) procedure based on the zebrafish protocol (https://wiki.zfin.org/display/prot/Embryo+Production+By+In+Vitro+Fertilization) was performed in order to generate hundreds of developmentally-synchronized one-cell embryos for microinjection of the CRISPR/Cas9 complexes. In brief, sexually mature male and female fathead minnows were paired with a spawning substrate and monitored for spawning behavior. Upon observation of initial spawning behaviors, such lateral contact against the spawning substrate and body vibration between the male and female, the adult pairs were selected for in vitro fertilization. Fish were anesthetized in buffered tricaine methanesulfonate (MS-222; Finquel; Argent) prior to collecting sperm and eggs. Sperm was collected into capillary tubes from at least three males and transferred to Hank’s buffered salt solution (HBSS; 0.137 M NaCl, 5.4 mM KCl, 0.25 mM Na₂HPO₄, 0.44 mM KH₂PO₄, 1.3 mM CaCl₂, 1.0 mM MgSO₄, 4.2 mM NaHCO₃). Sperm solutions were pooled and placed on ice prior to IVF. Unfertilized eggs were collected by gently squeezing the abdomen of four anesthetized females. If eggs were not easily expressed, a new female was selected. To the degree possible, eggs were expelled in straight lines directly onto glass microscope slides covered with 1 mL of control LSW. Due to the adhesive properties of the fathead minnow chorions, the eggs affixed themselves to the glass slides. The eggs were fertilized by gently pipetting the pooled sperm onto the slide containing the eggs. The slides were briefly swirled and allowed to rest for 2 min at room temperature after which 5 mL of control LSW was added to the slides, contained within a petri dish. Fertilization was assessed under a dissecting microscope. A total of three slides, each containing at least 30 fertilized eggs, were prepared per female, one for microinjection of the targeted CRISPR/Cas9 complex (“Tyr injected”), one for microinjection of control solution (“Control injected”), and one used as an uninjected control (“Control”).

Microinjections were conducted using a Narishige IM 300 microinjector and Narishige micromanipulator. Each pre-pulled needle (Tritech Research, Los Angeles, CA) was calibrated to deliver approximately 34 nL of injection solution. Immediately following IVF, one to two-cell eggs, adhered to the glass slides, were injected directly into the yolk sac with either the tyrosinase RNP or the control RNP. One slide of eggs per female were not injected and saved as control organisms.

2.3. Grow-out and phenotype observations

Slides containing eggs were placed in mesh-bottomed (0.5mm x 0.5mm mesh) baskets and submerged in LSW under gentle aeration until hatched. Dead embryos were removed from the slides daily, and retinal and body pigmentation were assessed at 3, 4, 14, and 34 d post-fertilization (dpf). Images were captured using a Nikon SMZ1270 microscope and NIS-Elements D (Version 4.6) imaging software. Upon hatching at approximately 5 dpf, larvae from each treatment were combined and released from the egg baskets into 10 L aquaria filled with ~ 5 L of LSW in a flow-through system. The fish were fed live brine shrimp (sp. Artemia) twice daily. Whole body samples were collected at 14 and 34 dpf, where the larvae were anesthetized with buffered MS-222, photographed, flash frozen in liquid N₂, and stored at −80°C until analyzed.

2.4. Mutation confirmation

Three randomly selected larvae from each of the control and control injected treatments, and six larvae from the tyr injected treatment, selected based on a strong phenotypic response at 14 dpf, were prepared for mutation analysis. DNA was extracted from the whole bodies using a crude digestion method in which the individual larvae were incubated in 50 mM NaOH at 95⁰C for 10–20 min. Samples were cooled to 4⁰C, and a 10% volume of 1M Tris pH 8.0 was added. The extracted samples were used to evaluate the potentially mutated regions by targeted amplification using polymerase chain reaction (PCR). All PCRs were conducted using JumpStart Taq Polymerase kits according to the manufacturer’s protocol (Sigma-Aldrich, St. Louis, MO), with each 30 μL reaction containing 3 μL of the DNA extract, and 0.5 μM forward and reverse primers (Table 2). Amplicons were verified by electrophoresis on a 2% agarose gel and purified using a QIAquick® Gel Extraction kit (Qiagen) according to the manufacturer’s protocol. The DNA concentrations were determined using a Nanodrop ND-1000 Spectrophotometer (Nanodrop Technologies; Thermo Fisher Scientific). The resulting amplicons were sequenced by classic Sanger sequencing (University of Minnesota Genomics Center, St. Paul, MN) for each of the forward and reverse sequences targeted with the PCR primers (Table 2). The frequency and locations of the sequence mutations were identified for each of the PCR products using Tracking of Indels by Decomposition (TIDE, https://tide.nki.nl, Brinkman et al., 2014) analysis. The TIDE algorithm delineates mutations by position and efficiency based on the chromatograms obtained via Sanger sequencing (Supplementary Information). This tool is vital to the analysis since the action of the non-homologous end joining repair system in each organism results in varied DNA sequence, leading to overlapping chromatograms that are not simply interpreted. Using TIDE, amplicon sequences from the control injected fish were compared with three sequences from the uninjected control fish for both the forward and reverse based sequences. The six results (three each, forward and reverse) for efficiency of indel formation were averaged (±SD) to determine the overall mutation efficiency (Con 1–3, Fig. 5). Amplicon sequences from the six tyr injected fish were compared with the three forward and three reverse sequences from the control injected fish (Tyr1–6, Fig. 5).

Table 2.

Fathead minnow primer sequences targeting three tyrosinase (tyr) CRISPR RNA (crRNA) strands for polymerase chain reaction and mutation verification analysis. FW = forward primer, RV = reverse primer.

Target strand	Amplicon size (bp)	Primer	Sequence (5′ → 3′)
tyr-crRNA-A	301	FW	GGTGTGCGGATCCCTTTC
		RV	GGTGTTCTTCGCCAGATTCA
tyr-crRNA-B	203	FW	CCTGAATCTGGCGAAGAACA
		RV	ATTCGTGCGCAAAGTCAATATC
tyr-crRNA-C	261	FW	CTGTAGTTTATGTGTACGGGTGT
		RV	CGTGTTCCTGAAGCTCATGT

Figure 5.

Mutation efficiency represented by percentage efficiency of indel formation in guide strands A, B, and C. Control injected (Con1–3) fish mutation efficiency was calculated against uninjected control fish. Tyrosinase CRISPR/Cas9 (Tyr1–6) fish simultaneously injected with all three tyr-specific guide strands A, B, and C. Sequences for Tyr1–6 fish were compared to control injected sequences to calculate mutation efficiency. Each bar represents the mean (± SD) mutation efficiency when measured against six sequences.

2.5. Statistical analyses

Percent survival data were arcsin transformed prior to analysis, and normality was assessed using a Kolmogorov-Smirnov test. Data were analyzed using a one-way analysis of variance (ANOVA) followed by Duncan’s post-hoc test. Data are presented as mean ± standard error (SE), and differences among treatments were considered significant at p < 0.05. Statistical analyses were conducted using Statistica 10 (StatSoft).

3. Results

3.1. Survival and phenotypic observations

Survival ranged from 50% (±11) to 76% (±31) (mean ± SD) among the injected cohorts, but there were no statistically significant differences in survival between the CRISPR/Cas9-injected, control injected, or uninjected controls at any of the time points (Fig. 2). All fish appeared healthy and had a similar appetite based on behavioral observations. All larvae had hatched by 5–6 dpf.

Figure 2.

Percentage of survival of control (uninjected), control injected, and tyrosinse (tyr) CRISPR/Cas9 injected fathead minnow embryos/larvae at 1, 4, and 14 days post-fertilization (dpf). For 1 and 4 dpf, bars represent mean ± SE derived from egg survival from n=4 spawning females. All larvae were pooled at 14 dpf and total percent data were calculated.

There were no apparent differences in phenotypes between the uninjected and control injected fish. Upon the occurrence of retinal pigmentation in the controls at 3 dpf (Fig. 3A and 3C), it became apparent that several of the tyr injected fish had mosaic pigmented eyes (Fig. 3E). This mosaic retinal pattern is consistent with previous studies examining tyr CRISPR/Cas9 knockdown in zebrafish (Ota and Kawahara, 2014). By 14 dpf, the appearance of melanocytes on the skin and around the swim bladder was obvious in the control fish (Fig. 3B and 3D), but complete lack of pigmentation was observed in several of the tyr injected fish (Fig. 3F).

Figure 3.

Live images of pigmentation modifications in 4 day post-fertilization (dpf) fathead minnow embryos and 14 dpf larvae following injection of a CRISPR/Cas9 complex targeted for tyrosinase knockdown. (A-B) Uninjected controls; (C-D) Injected controls; (E-F) Injected with tyrosinase targeted CRISPR/Cas9 complex.

3.2. Mutation efficiency

In total, 241 embryos were injected with the tyr RNP, 247 embryos were injected with the control RNP, and 275 embryos were used as uninjected controls. Of the fish sampled at 14 dpf, 50% of the tyr injected fish displayed altered pigmentation (n=12 Tyr injected). At 34 dpf, 44% of the the tyr injected fish had altered pigmentation (n=18 Tyr injected). In considering these values, it is important to note that the sub-sampling at 14 and 34 dpf was not random; fish were selected based on the most dramatic phenotypic alterations in order to obtain organisms which could be used to confidently confirm the mutation with subsequent genetic sequencing.

A TIDE analysis was used to evaluate the mutation efficiencies for each of the three guide strands that were co-injected. Sample data from two individual TIDE analyses of the strand A amplicons are shown in Figure 4. Variation between control injected and uninjected samples were minor, and mutations were not statistically significant (Fig. 4A). Control injected sequences were used as the comparison sequence for all potential tyr mutant fish. Sequence variation between tyr injected fish displaying phenotype and control injected fish was as high as 83.5% (Fig. 4B).

Figure 4.

Tracking of Indels by Decomposition (TIDE) analysis was employed to determine mutation efficiency as a percent of insertion/deletion (indel) formation in individual fish. Two analyses are shown: A. Control injected fish compared to uninjected control fish and B. Tyrosinase mutant fish compared to control injected fish.

Control injected and uninjected samples showed minor, statistically insignificant mutation efficiencies for all samples using guide strands A, B, and C. In the tyr mutant fish, guide strand A was the most effective, with the phenotypically identified mutants yielding a 12–85% mutation rate (Fig. 5). Overall, the three fish with the most visually apparent phenotypes (i.e., lack of pigmentation) were those with 53, 81, and 85% indel efficiency with guide strand A; therefore, phenotypic appearance and mutation rates agreed. In addition to the TIDE results, one of the fish with the tyrosinase phenotype also provided Sanger sequencing data with a non-frameshift mutation that allowed for a clear interpretation of the chromatogram. As shown in Table 3, the affected area is rich in cysteine residues, and this fish had a cysteine, as well as glycine and glutamatic acid residues deleted. Guide strand B was found to be ineffective, while strand C resulted in minor formation of indels compared to control injected fish (Fig. 5).

Table 3.

Amino acid sequence alignment (Sievers et al., 2011) of uninjected control versus tyrosinase knockout (KO) fathead minnow larvae determined using classic Sanger sequencing.

Sample Type	Amino acid sequence
Control	NYMGIDCGECKFGFFG
Tyrosinase KO	NYMGIE - - - CKFGFFG

4. Discussion

Logistical challenges of CRISPR/Cas9 with fathead minnows

In developing and optimizing a CRISPR/Cas9 system injection protocol for fathead minnows, there were technical challenges to overcome. First was a need to control the synchronization and timing of fertilization. In the case of zebrafish in culture conditions, initiation of spawning behaviors and fertilization are tightly synchronized with light (Blanco-Vives and Sánchez-Vázquez 2009). In contrast, fathead minnow spawning and fertilization typically occur asynchronously during the early morning hours (U.S. EPA, 2002). Additionally, zebrafish broadcast their eggs, meaning eggs can be readily siphoned from the bottom of a tank and easily transferred via pipets. Fathead minnows lay their eggs on the undersurface of a nesting object and the eggs both stick to one another and the substrate. A consequence of this is that the eggs are easily damaged during removal from the substrate, particularly within the first few hours of fertilization before they harden (Gast and Brungs, 1973). To circumvent these issues and obtain large batches of single cell-stage eggs for microinjection, we adopted an in vitro fertilization approach. Rather than try to remove eggs from a substrate, their adhesive property was used as an advantage to prepare groups of eggs stuck to a glass side that could be fertilized simultaneously and then rapidly injected as a group. Viability and fertilization success depended largely on the maternal origin, with some females providing groups of eggs that were largely non-viable, while others produced mostly viable eggs. Selection of females that were exhibiting reproductive behavioral cues markedly improved the success rate. Overall, the procedure developed could readily support the preparation of several hundred eggs for injection in a single day, with the ability to inject approximately 50–75 eggs within the 10 minutes that the eggs were out of the water. The in vitro fertilization technique and mounting of embryos to glass slides not only proved to be critical for CRISPR injections, but holds promise as a useful application in other microinjection scenarios such as chemical treatment of embryos or other gene knockout approaches.

Design of the tyrosinase guide strands and mutation efficiencies

The guide strands in our study were designed to target three regions of tyr in the first half of the gene (Fig. 1) in areas of importance as determined using the Conserved Domain Database (Wang et al., 2016). Strand A base-paired with sequence in the laminin-type EGF-like domain, while B and C targeted the tyr domain. Strand A led to a strong tyr mutant phenotype which matched the higher mutation efficiencies determined with TIDE analysis. The fish with 85% mutation efficiency demonstrated a predominant deletion of nucleotides, resulting in the mutation of aspartic acid to glutamic acid followed by the omission of three amino acids: cysteine, glycine and glutamic acid (CGE) from the EGF-like domain of the protein (Table 3). Cysteine is the most important of these residues; cysteine clusters are present in tyrosinase enzymes where, as in most proteins, they are critical for folding and therefore function (Garcia-Borron and Solano, 2002).

In the laminin-type EGF-like domain, one cysteine cluster contains a highly conserved redox protein CXXC motif (Fomenko and Gladyshev, 2003), present as CGNC in mice and CGEC in fathead minnow (Garcia-Borron and Solano, 2002, and Table 3). Since the dominant mutation deleted the CGE of CGEC in fathead minnow, this mutant tyr most likely failed to form the proper disulfide link essential for proper folding. Notably, the mutation of CGNC in mice to CGNS results in albino mice (Garcia-Borron and Solano, 2002; Beermann et al., 2004).

As seen in the low mutation efficiencies using TIDE analysis, strands B and C were comparatively ineffective. Retrospectively, the region within the tyr domain (Fig. 1) where strand C targeted was not critically conserved and thus not an ideal candidate for CRISPR/Cas9 targeting. Strand B targeted a conserved area, but mutation efficiency was very low, and data on function of the amino acids targeted within this region was not found. Although it was recently determined that indel formation results in a frameshift mutation, on average, 80% of the time (Chakrabarti et al., 2019), and a fraction of these mutations result in stop codons, the most effective strategy for targeting is to identify regions where in-frame mutations are just as problematic as the generation of new stop codons, as demonstrated by the successful outcome observed with strand A.

Injection of three different guide strand containing RNPs was a tactic used here to expedite the strand screening process, yielding mutant fish in the shortest time possible. As recently demonstrated, injection of multiple effective guide strands can also greatly improve mutation efficiency in the G0 zebrafish, with four strands giving optimal mutation without unacceptable biological effects such as lethality or overt malformations (Wu et al., 2018). While the triple injection approach decreased the number of microinjections and fish growth periods required, PCR and analysis by TIDE was still required for each of the three targeted regions. In future studies, quality guide strand development should be the first goal, targeting a critical region when possible. Potential guide strands will be screened with a Cas9 in vitro nuclease assay to determine targeting and cutting ability prior to injection (Turner et al., 2018). One or several active RNPs will be injected at increased concentrations, following the methods of Wu et al. (2018).

Potential applications for adverse outcome pathway development and evaluation

One of our primary applications of interest for a precision gene editing technology such as CRISPR was the potential to use a highly specific, non-chemical intervention to perturb a biological system. The adverse outcome pathway (AOP) framework was developed as a systematic approach for assembling and evaluating the biological understanding and empirical evidence that underlies the ability to infer a toxicological hazard (i.e. adverse apical outcome) from the perturbation of a specific molecular target or pathway (i.e., a molecular initiating event) (Ankley et al. 2010; Vinken et al. 2017). Molecular or pathway-based perturbation of biological systems can plausibly be evaluated in assays (primarily in vitro) that are low cost and high throughput compared to animal testing focused on the direct observation of apical outcomes (Krewski et al. 2010). To support the premise that batteries of mechanistically-focused screening assays could be employed to help identify the toxicological hazards a chemical may pose, the AOP framework aims to describe the linkage between a biological perturbation and the adverse outcome(s) it may cause, in a manner that is independent of chemical (Villeneuve et al. 2014). That is, an underlying assumption of the approach is that any chemical (or other type of stressor), that perturbs a given biological target or pathway could potentially elicit the same toxicological outcome, given adequate dose, duration of exposure, etc.

Adverse outcome pathway development is often supported by studies that employ chemicals with well-established specificity to act on specific molecular targets of interest (e.g., known to bind a given receptor, inhibit a specific enzyme, etc.). However, there are many toxicologically relevant targets and/or pathways for which chemicals with highly specific, target selectivity are not available. Likewise, establishing that a chemical is specific for a particular molecular target is difficult to achieve experimentally, often leaving open questions as to whether the toxicity observed is due to the chemical-target interaction of interest, or whether off-target/non-specific effects may be responsible for the toxicological effect(s) observed. This is illustrated by a pair of studies with 2-mercaptobenzothiazole (MBT), a chemical known to inhibit thyroid peroxidase (TPO). Experiments conducted in both fathead minnow and zebrafish provided support for an AOP linking TPO inhibition to impaired swim bladder inflation in fish (Nelson et al. 2016; Stinckens et al. 2016; https://aopwiki.org/aops/159). However, experimentally, neither study could rule out the possibility that MBT may have elicited an effect on the swim bladder through an alternative mechanism independent of its effect on TPO. In concept, a precision gene editing approach like CRISPR/Cas9 would allow for specific perturbation of TPO, with minimal likelihood for off-target effects, allowing one to establish the connection between the molecular initiating event and apical outcome with greater certainty. Thus, CRISPR/Cas9 could prove to be a highly effective tool for the evaluation of AOPs and ultimately supporting the proposed vision for toxicity testing in the 21st century (Krewski et al. 2010; Villeneuve and Garcia-Reyero, 2011).

Although the present study demonstrated the ability to successfully produce precision targeted mutations in the fathead minnow tyr gene, there remain several technical challenges for the approach to be practical for routine use in evaluating hypothesized AOPs. The first is the mutation efficiency. In the present study, of the selected fish on each sampling day, only 44–50% of the individuals injected with tyr RNPs displayed a clear mutant phenotype. Because only a small portion of the “treated” cohort realize the mutation, large sample sizes need to be injected in order to obtain an adequate sample size to evaluate the hypothesis. This is further complicated by the variability in the insertion-deletion mutations introduced during non-homologous end joining repair of the double strand breaks. Even where a phenotypic mutation was evident, there was clearly a gradient of effectiveness among different fish. That gradient correlated closely with the mutation efficiencies detected by TIDE. Consequently, even in the subpopulation of the injected cohort that received a mutation, the nature and effect of the mutation varied from fish to fish. This introduces additional uncertainty into the ability to associate a targeted mutation with the hypothesized outcome. If complete or near-complete mutation can be achieved by using four RNPs at higher concentrations (e.g., Wu et al., 2018), some of this uncertainty will be ameliorated. Finally, in the case of tyr there was an obvious phenotype that could easily be used to identify mutants. However, for many targets of potential interest the phenotypic result of the mutation may not be readily observed. Therefore, there is a need for low cost-efficient methods to rapidly screen a large number of injected organisms for mutations in order to separate those that received a successful mutation from those for which mutation did not occur.

One potential solution is to use CRISPR/Cas9 to generate a mutant line that can be cultured and used for subsequent experimentation. While this could work for some applications, it would not be feasible in cases where a mutation in the target of interest is lethal during development and/or impairs successful reproduction, although both outcomes could still provide information of toxicological value. Optimization of the guide strands and injection protocols have potential to improve injection efficiencies, as does injection of increased amounts of soluble RNPs (Burger et al. 2016). Additionally, newer CRISPR/Cas9 approaches that facilitate insertion of donor DNA and homology-directed repair to induce consistent frameshift mutations (Cox et al. 2015) would improve results compared to relying on indel formation from non-homologous end-joining. There are also emerging approaches that enable direct irreversible conversion of one DNA base into another without the need for double strand breaks (Komor et al. 2016), a method that has been used to silence genes by introducing stop codons (Kuscu et al. 2017). Such “base-editing” approaches could provide more control over the mutations generated. Consequently, although there are currently obstacles to the ready application of CRISPR/Cas9 for AOP evaluation, the technology is developing rapidly and many of the current technical obstacles can be overcome.

5. Conclusion

The present study demonstrated successful application of the CRISPR/Cas9 method to induce targeted mutations in a gene of interest in the fathead minnow. Three guide strands to knock out tyr function were designed for use in the fathead minnow. The guide strand targeting the highly conserved cysteine-rich region was the most effective. Following refinement and optimization of methods for in vitro fertilization and microinjection, fathead minnow embryos were successfully treated and phenotyped, and mutations were verified through sequence analysis. While further optimization is required to make the method effective for routine use in the evaluation of hypothesized AOPs, the present study suggests that application of these approaches to the fathead minnow, as a prominent model organism in aquatic toxicology, is feasible.

Abbreviations

AOP

Adverse outcome pathway

CRISPR

Clustered Regularly Interspaced Short Palindromic Repeats

FHM

Fathead minnow

IVF

In vitro fertilization

RNP

ribonucleoprotein

Tyr

Tyrosinase

ZF

Zebrafish