Comparing robotic and manual injection methods in zebrafish embryos for high-throughput RNA silencing using CRISPR-RfxCas13d

The zebrafish is a vertebrate model organism with many experimental advantages, such as large progeny, embryo transparency and amenability to genetic manipulation by microinjection. However, microinjection of the small zebrafish embryos is laborious and time-consuming. Therefore, better high-throughput methods for the rapid microinjection of zebrafish embryos would significantly facilitate studies of gene function, pharmacological screens and disease models.

The authors' laboratory has shown that a robotic system allows for high-throughput injections of CRISPR-Cas9 and DNA constructs into zebrafish zygotes with an efficiency comparable to that of manual injections [1]. Moreover, similar types of embryo injection robots were efficiently used for zebrafish tuberculosis infection studies and compound screens [2–5].

In addition to DNA modifications via CRISPR-Cas9 to generate F0 knockouts [6,7], robotic injection also offers the possibility of generating and screening phenotypes resulting from various RNA depletion methods. For instance, RNA interference technologies and morpholinos offer an approach to RNA knockdown in several organisms. However, RNA interference technologies have failed to be established in a zebrafish model [8,9] and morpholinos have shown toxicity, off-target effects, innate immune responses and discordant phenotypes with mutant animals [10–13].

Furthermore, CRISPR-Cas systems were first described as bacterial adaptive immune systems against phages [14–19]. They are organized in two main classes: class 1 with Cas effectors composed of multiple subunits and class 2 with monomeric large proteins; on top of that, specific subtypes within classes depend on the Cas endonuclease and its mechanism of action (Table 1). From class 2, Cas9 and Cas12 are used for many genomic engineering applications [20–24] and have been applied in the zebrafish model [6,25–27]. Moreover, the recently discovered class 2 type VI Cas13 systems are RNA-guided ribonucleases applied for transcriptomic engineering (Cas13a–Cas13d) [28–33]; in particular, the CRISPR-RfxCas13d (CasRx) system induces efficient knockdown of both zygotically and maternally provided RNAs in animal embryos [34].

The authors used automation to quantitatively evaluate the relative efficiencies of CRISPR-CasRx and CRISPR-Cas9 to inactivate the function of the TBXTA gene at the RNA or DNA level, respectively, and generate the resulting no-tail phenotype [35]. They show that high-speed robotic injection of the CRISPR-CasRx achieves survival rates and knockdown efficiencies comparable with manual injections, thus opening the possibility of high-throughput interrogation of the transcriptome in a tunable manner.

Materials & methods

Zebrafish maintenance & embryo production

Leiden University researchers follow the European rules for the use of animals. Zebrafish wild-type strains ABTL are maintained and bred according to standard conditions [36]. One-cell stage ABTL embryos were used in this study.

CasRx TBXTA gRNA production

The gRNAs for CasRx targeting TBXTA [34,37] gene were produced by fill-in polymerase chain reaction (PCR) with Phire HotStart II DNA polymerase (Thermo Fisher Scientific, F122S; primers in Supplementary Table 1, Integrated DNA Technologies (IDT) in a thermal cycler (BioRad, T100) with 10 uM final concentration of each primer and the following program: 3 min at 95°C, 35 cycles of 30 s at 95°C, 30 s at 60°C and 30 s at 72°C, and a final step at 72°C for 5 min. The PCR product was in vitro transcribed (Epicentre, ASF3507) under the control of promoter T7 for 12 h at 37°C and DNAse treated with TURBO-DNAse for 20 min at 37°C. Finally, the gRNA was precipitated with 3 M ammonium acetate and resuspended in nuclease-free water. gRNA integrity was verified with electrophoresis (Supplementary Figure 3), quantified by Nanodrop 2000 (Thermo Fisher Scientific) and titrated until high phenotypical penetrance was obtained. The authors observe that quantifying gRNAs after in vitro transcription can yield inaccuracies, which may lead to an overestimation of the amount of gRNAs being utilized. However, they mitigated this issue by titration of the gRNA employed.

Cas9 & CasRx mRNA production

Plasmids containing Cas9 and CasRx genes, pT3Ts-nCas9n (Addgene 46757) and pT3Ts-RfxCas13d-HA (Addgene 141320), respectively, were linearized with XbaI (NEB R0145L) for 3 h at 37°C and then in vitro transcribed under control of promoter T3 (Ambion AM1348) for 3 h at 37°C and DNAse treated with TURBO-DNAse for 20 min at 37°C. Capped mRNA from these reactions was purified with RNAeasy MiniKit (Qiagen 74104). RNA integrity was verified by electrophoresis (Supplementary Figures 1 & 2) and quantified with Nanodrop 2000 (Thermo Fisher Scientific).

CasRx protein expression & purification

Competent Rosetta DE3 pLysS (Novagen 70956) was transformed with the plasmid pET28b-RfxCas13d-His (Addgene 141322), which contains the recombinant CasRx gene. Cells were grown for 3 h at 37°C and then induced with 0.1 mM isopropyl β-D-1-thiogalactopyranoside for another 3 h. Cells were washed once with 20 mM Tris-HCl pH 7.5 and the pellet was frozen at -80°C before the next steps.

Pellets were resuspended with lysis buffer (50 mM HEPES*KOH pH 7.5, 500 mM KCl, 10% v/v glycerol, 1 mM dithiothreitol and 10 mM imidazole) and sonicated (QSonica, Q125) using the following program: 3 s on, 10 s off, 120 cycles, 30% amplitude. Lysate was centrifuged and then filtered using a 0.2 μm cellulose acetate filter. CasRx protein was then purified using a HisPur™ Ni-NTA spin column (Thermo Fisher Scientific, 88226) and washed with 20 column volumes of lysis buffer by gravity. Elution was performed in five steps with elution buffer (50 mM HEPES*KOH pH 7.5, 500 mM KCl, 10% v/v glycerol, 1 mM dithiothreitol and 500 mM imidazole) diluted with lysis buffer to get increasing concentrations of imidazole (10 mM, 30 mM, 50 mM, 100 mM and 200 mM). Fractions containing purified CasRx were pooled and then concentrated and dialyzed with dialysis buffer (50 mM HEPES-KOH pH 7.5, 250 mM KCl, 1 mM dithiothreitol and 10% glycerol) using a 50K Amicon Ultra-15 (Millipore UFC905024) until the concentration of 3 µg/ml was reached. Protein concentration was estimated with Nanodrop 2000 (Thermo Fisher Scientific) and extraction quality was verified by SDS-PAGE (Supplementary Figure 4). Single-use aliquots of 4 μl were stored at -80°C.

Cas9 protein & gRNAs

Alt-R™ – Cas9 protein, Alt-R™ crRNA and Alt-R™ tracrRNA were purchased from IDT. The same gRNAs (Supplementary Table 1) were used in both mRNA and protein injections of Cas9. Alt-R CRISPR systems from IDT are optimized genome editing tools with predesigned gRNAs, to produce on-target double-strand breaks using high-purity solutions.

Production of mRNAs, gRNA & protein

CasRx and Cas9 (see methods) showed a high rate of capped mRNA production from linearized DNA template (∼100-fold ng of RNA α ng of DNA, data not shown), with little degradation after purifying, as shown in the agarose gels (Supplementary Figures 1 & 2). Similarly, the gRNA generation showed higher efficiency (∼800-fold ng of RNA α ng of DNA, data not shown), having a reaction time four-times longer than mRNA reaction and with low degradation (Supplementary Figure 3). The CasRx protein production was induced and purified with a Ni-NTA column followed by concentration with an Amicon ultracentrifugation column. The authors obtained a concentration of 3 μg/ml with a purity of approximately 85% (Supplementary Figure 4).

Microinjection solutions & injection quantities

Cas9 gRNAs were prepared by annealing crRNA with tracrRNA using duplex buffer IDT and incubating at 95°C for 5 min. Three different gRNAs were pooled and then mixed with Cas9 mRNA. The final quantity injected with 1 nl per embryo was 200 pg of Cas9 mRNA and 1070 pg of gRNA (357 pg each). Control groups were only injected with 200 pg Cas9 mRNA. To produce the Cas9 ribonucleoprotein, the same gRNAs were mixed with Cas9 protein IDT and incubated at 37°C for 5 min. The final quantity injected with 1 nl per embryo was 5029 pg of Alt-R Cas9 protein and 1070 pg of gRNA (357 pg each). Control groups were only injected with 5029 pg Cas9 protein. In the case of CasRx, the mRNA was mixed with the gRNA targeting TBXTA transcript. The final quantity injected with 1 nl per embryo was 200 pg of mRNA and 300 pg of gRNA. Control groups were only injected with 200 pg CasRx mRNA. CasRx protein was mixed with gRNA targeting TBXTA transcript. The final quantity injected with 1 nl per embryo was 3 ng of protein with 300 pg of gRNA. Control groups were only injected with 3 ng CasRx protein.

Manual microinjection

Manual microinjection was performed using a Pneumatic Pico-Pump (PV 820, WPI) and all embryos were injected into the yolk during one-cell stage with ~5–8 psi depending on the needle pore size. Commercial needles with the tip size within 6–7 μm (Clunbury Scientific LLC) were used. All experiments were performed with at least three replicates.

Automated robot microinjection & software updates

The detailed procedure of using the upgraded automated microinjection system (Life Science Methods BV, The Netherlands) is described in Figure 1. The same type of commercial needles from manual injection was used for robotic injection. Some software updates were carried out to improve the automated injection throughput and efficiency. A deep learning network was updated to recognize the yolk center and the first cell interface. Another deep learning network was added to recognize the shape of the droplet and compute to volume. The speed increased from 1.8 s per injection in the previous version to 1.0 s per injection on average in the current system.

Results & discussion

Establishment of robotic injection procedures

In this study, the authors developed and upgraded a robot based on previous versions [1,3,5]. The new version of the robot is a one-box fully integrated device with touch screen and provides optimal injection speed into early-stage zebrafish embryos (Figure 1A). An agarose grid plate is prepared using a 9 × 100 well mold (Figure 1B & Supplementary Video 1). Zebrafish embryos can be placed in each grid well (Figure 1C & Supplementary Video 2). It takes approximately 2 min to position the eggs within the 9 × 100 grid wells. Subsequently, a needle is mounted into the needle holder, and needle tip (xy direction) and height (z direction) are calibrated (Figure 1D & Supplementary Video 3). For droplet calibration, a machine learning algorithm is added to recognize the shape of the droplet in a well filled with mineral oil and the volume of the droplet is given automatically (Figure 1E & Supplementary Video 4). Injections can be chosen to be either into the middle of the yolk (Figure 1F& Supplementary Video 5) or close to the first cell (Figure 1F & Supplementary Video 6). Statistics on the top left of the screen give an overview of the number of embryos in each category during injection (Figure 1F & Supplementary Videos 5 & 6). Eggs where a first cell cannot be recognized (such as “no-cell” type) can be chosen to be injected or skipped. The skipped embryos are registered, and these eggs can be killed using a blunt needle (Figure 1G & Supplementary Video 7). After the killing is finished, the embryos can be easily transferred to petri dishes for culture (Figure 1H & Supplementary Video 8). The count of the destroyed eggs can be verified, and this count should match the total number of eggs that were not injected, displayed at the top left corner of the screen.

In this study, the authors presented a detailed procedure for utilizing an advanced zebrafish embryo injection robot and conducted a comparative analysis of injection outcomes between robotic and manual methods. The robotic system offered several advantages. The implementation of this robot allows individuals without any prior zebrafish egg injection experience to perform large-scale injections effortlessly. Additionally, the inclusion of a user-friendly touch screen interface enhances the operational convenience and provides a visually appealing display of the injection processes, ensuring a comfortable experience for the operator. Moreover, the robot offers the advantage of precise injection volumes by automatically calculating droplet sizes using two-dimensional image analysis. This automated computation significantly reduces errors associated with manual estimation, minimizing the potential translation of such errors into inaccuracies in injected volume. Finally, the robotic system incorporates a feature whereby it records the positions of noninjected eggs and subsequently performs selective euthanasia, ensuring that only the injected embryos remain alive for further analysis.

The results indicated that both the robotic and manual injection approaches yielded similar percentages of phenotypes and mortality rates, specifically regarding the knockdown of the TBXTA gene. However, it is noteworthy that the robotic method exhibited superior speed, being three-times faster than manual injections. This improvement in speed is attributed to the expanded region of interest, which allows for simultaneous evaluation of the next well during the ongoing injection process. On average, the robot achieved an injection speed of 1.0 s per injection, as demonstrated in Supplementary Videos 5 & 6.

Comparison of manual & automated robotic injection

To compare injection results between manual and robotic injections, both CRISPR-Cas9 and CRISPR-CasRx systems targeting the TBXTA gene were used. The Cas9 gRNAs achieved more than 90% of F0 biallelic knockouts, which were validated by DNA sequencing [6]. Additionally, the CasRx gRNA used to reproduce this phenotype was previously validated by reverse transcription–quantitative PCR, resulting in an average transcript level decrease of 60% [34]. Knockdown of the TBXTA gene using both systems resulted in different levels of tail and notochord defect phenotypes at 30 hpi (Figure 2A). Moreover, the knockdown efficiency of both Cas9 and CasRx when administered through a robot was very similar to that achieved using a manual injector (Figure 2B & C). Embryos graded with phenotype 1 were more frequent with CasRx treatment. A higher percentage of embryos with severe phenotype (type III) and a lower percentage with mild phenotypes (type I and type II) were observed in Cas9 protein than in Cas9 mRNA groups (Figure 2B). However, there were no clear differences in the percentages of embryos displaying different levels of phenotypes between the groups treated with CasRx protein and those treated with CasRx mRNA (Figure 2C). Since in previous papers it has been shown that the position of the injection in the zebrafish embryos results in different penetrance in gene silencing technologies, we compared the effect of the position of injection in the embryo. Injection into the cell interface and yolk center of zebrafish embryos showed no difference in the phenotype (Figure 2B & C). The mortality rate of injected embryos was similar between manual and robotic injection (Table 2). However, the speed of the robot (59 ± 3 embryos/min, 2169 injected embryos) was three-times faster compared with the manual injections (20 ± 3 embryos/min, 990 injected embryos; Table 2).

It has been observed that Cas9 protein injections lead to more pronounced phenotypic effects compared with Cas9 mRNA injections in terms of TBXTA gene knockout. This disparity may be attributed to the ability of Cas9 protein to perform DNA double-strand breaks from one-cell stage at the moment of injection; on the other hand, the mRNA would have to be translated for the protein to cause a DNA double-strand break. Moreover, this study also highlighted the potential of the CasRx system at the RNA level. The CasRx system for knockdown has undergone significant improvements through upgraded protocols and platforms in recent years [37]. In the present study, the authors utilized previously validated CasRx gRNAs targeting TBXTA mRNA [34,37], taking advantage of the enhanced CasRx system. Given that CasRx protein is not commercially available, the authors expressed and purified the protein in their own laboratory. Both the purified CasRx protein and CasRx mRNA injections resulted in high phenotypic penetrance. The percentage of phenotypes observed in the CasRx protein group was similar to that in the mRNA group. This low difference between mRNA and protein injection of CasRx may be because the target gene, TBXTA, is involved in embryo notochord formation, which occurs relatively late in zebrafish development. Nevertheless, if an early development gene is studied, the CasRx protein might have a higher phenotypic penetrance. Moreover, after the CasRx mRNA injection, protein translation in the embryo has not yet been studied in different developmental stages and the phenotypic penetrance would also depend on the target mRNA levels in these stages. Another interesting observation is that grade I phenotypes can be seen with the CasRx experiments. This could be explained by the previous validation for the used gRNA, which showed a reduction of mRNA levels of about 60% of the TBXTA transcript by reverse transcription–quantitative PCR [34,37], as the authors expect the same efficiency in their experiments. This reduction would still allow for a low expression of this gene in the embryo. Additionally, it was demonstrated that disrupting a single nucleotide in the tail of the gRNA (CasRx binding domain) and using a 23-nucleotide binding sequence significantly improves the phenotypic penetrance [38]. Specifically, the percentage of grade III no tail cases doubles from 35% to 70% with this chemically synthetized, highly efficient gRNA structure [37].

In the context of gene editing, traditional practices have involved injecting DNA/mRNAs/gRNAs into the first cell of zebrafish embryos. However, recent studies, such as the work by Kroll et al. [6], have demonstrated that injection directly into the yolk is adequate for achieving F0 knockouts using the highly efficient CRISPR-Cas9 technique. In the present study, the authors utilized a combination of multiple gRNAs targeting the TBXTA gene and observed that yolk injections resulted in a substantial occurrence of tail and notochord defects. Moreover, they found that the proportion of phenotypes caused by yolk injection closely resembled that resulting from cell interface injection using the robotic system. It is important to note that these injections were administered at the one-cell stage, shortly after fertilization, minimizing the time elapsed before the cell started to inflate. Consequently, the authors propose that the timing of injection, specifically immediately following fertilization and prior to cellular expansion, holds greater significance than injecting into the zygotic cell itself [6].

Conclusion

In conclusion, this study establishes the efficacy and advantages of robotic automated injection methods in zebrafish embryos for gene silencing experiments. The developed robotic system, with its user-friendly touch screen interface and automated droplet size calculation, streamlines large-scale injections, making it accessible even to individuals without prior zebrafish egg injection experience. The precision in injection volumes, error reduction and the unique feature of recording and selectively euthanizing noninjected eggs contribute to the system's efficiency. Our results demonstrate comparable phenotypes and mortality rates between manual and robotic injections, particularly in the context of the knockdown of the TBXTA gene. On the other hand, the robotic method surpasses manual injections in speed, being three-times faster. Additionally, we observed that Cas9 protein injections result in more pronounced phenotypic effects compared with Cas9 mRNA injections for TBXTA gene knockout. Furthermore, our study highlights the potential of the CasRx system at the RNA level, emphasizing its efficacy in achieving high phenotypic penetrance, particularly in early developmental genes. Overall, our study contributes to the understanding of the benefits and feasibility of robotic automated injection methods in zebrafish embryos for gene silencing experiments. The combination of precise injection control and increased efficiency provided by the robotic system holds promise for accelerating research in early embryonic development and other areas that require large-scale injections. Furthermore, the exploration of the CasRx system showcases the potential of RNA-targeting CRISPR effectors for precise and potent gene knockdown at high throughput.

Future perspective

Zebrafish automation is expected to gain increasing importance in screening research requiring high throughput, such as drug screening, gene and transcript function analysis, toxicity assessments and pathogen studies. We anticipate that our approach can serve as a foundation for further technical advancements. These could include improvements in throughput levels; enhanced injection accuracy, particularly in specific embryo or larval body regions; and increased ease of needle loading in the case of blockages. Such technologies may also find application in other models, such as organoid cultures. It is evident that the CRISPR-Cas toolbox will undergo further refinement, unlocking new possibilities for modifying DNA, RNA and even protein functions. The swift integration of this technology into our developed robotic pipeline is feasible. In particular, the exploration of the CasRx system highlights the potential of RNA-targeting CRISPR effectors for precise and potent gene knockdown at high throughput and with higher accuracies, as high-fidelity versions are being developed.