A versatile toolkit for CRISPR-Cas13-based RNA manipulation in Drosophila

Advances in CRISPR technology have immensely improved our ability to manipulate nucleic acids, and the recent discovery of the RNA-targeting endonuclease Cas13 adds even further functionality. Here, we show that Cas13 works efficiently in Drosophila, both ex vivo and in vivo. We test 44 different Cas13 variants to identify enzymes with the best overall performance and show that Cas13 could target endogenous Drosophila transcripts in vivo with high efficiency and specificity. We also develop Cas13 applications to edit mRNAs and target mitochondrial transcripts. Our vector collection represents a versatile tool collection to manipulate gene expression at the post-transcriptional level. Supplementary information The online version contains supplementary material available at 10.1186/s13059-020-02193-y.

The recent introduction of the class II type VI CRISPR/Cas13 system further expands the existing technology in significant ways. Like Cas9, Cas13 uses a guide RNA (CRIS PR-RNA, aka crRNA) to identify its substrate, which is RNA rather than DNA (Fig. 1b). Cas13 enzymes have two distinct catalytic activities: (i) an RNAse activity that is mediated by two higher eukaryotic and prokaryotic nucleotide (HEPN)-binding domains and (ii) a gRNA maturation activity, possibly a combination of activities located in the HEPN2 and Helical-1 domains [21,22]. There are currently four subtypes identified in the Cas13 family, including Cas13a (aka C2c2), Cas13b, Cas13c, and Cas13d. All Cas13 family members are smaller than Cas9, with Cas13d being the smallest protein. The small size of Cas13 proteins makes them suitable for molecular genetics (Fig. 1c). All Fig. 1 Functional overview of CRISPR/Cas9 and CRISPR/Cas13 systems. a Schematic of Cas9 mechanism in genome editing. This system requires the recruitment of CRISPR-associated protein Cas9 (blue) to the target site recognized by the guide RNA (gRNA: orange). Target site cleavage by Cas9 is ensured by the presence of the protospacer adjacent motif (PAM) (green), a sequence that immediately follows the target site. The PAM will determine the Cas9 cleavage site, which lies about three nucleotides upstream of the PAM. b Schematic of the Cas13 RNA cleavage mechanism. This system requires the pre-assembly of Cas13 (green) with the CRISPR-RNA (crRNA: red) to recognize target RNAs. Upon RNA-binding, Cas13 will undergo a conformational change and induce the catalytic activity of its nuclease domains, resulting in the cleavage of target transcripts. c Comparisons of Cas9 size with different Cas13 subtypes (a-d). Polypeptide sizes are indicated as the number of amino acids. d Relative structural representation of different Cas13 subtypecompatible crRNAs. All four subtype crRNAs carry a direct repeat to facilitate the binding with their corresponding Cas13 enzyme, as well as a spacer sequence specific for the target transcript. Cas13bcompatible crRNAs carry a direct repeat at the 3′-end while compatible crRNAs for Cas13a, c, and d carry the direct repeat at the 5′-end Cas13 enzymes require a 60-66-nucleotide-long crRNA to ensure target specificity [2,3,23]. Similar to the gRNA in the CRISPR/Cas9 system, the crRNA used by Cas13 forms a short hairpin structure next to a short spacer sequence (28-30 nucleotides) that is specific to the target transcript (Fig. 1d). Since CRISPR/Cas13 mediates RNA degradation, it holds the promise to replace or complement RNA interference (RNAi) approaches or other systems that interfere with transcript levels, such as CRISPRi. Despite being a powerful tool, RNAi often suffers from low efficiencies or off-target effects, whereas Cas9-based CRISPRi requires a protospacer adjacent motif (PAM), thus limiting the flexibility by which target sequences can be selected [24][25][26][27][28]. It is desirable to examine whether CRISPR/Cas13 can offer better specificity and efficiency than these other interference techniques.
Drosophila melanogaster is a versatile genetic model organism that is used to study a wide variety of biological processes. Traditional techniques to analyze gene function in Drosophila include the generation of mutations via chemical mutagens and transposable P-elements, or the use of transgenes to trigger RNAi and to express cDNAs for gain-of-function studies via the Gal4/UAS system [29][30][31][32][33]. Like other model organisms, the CRISPR/Cas9 endonucleases have been quickly adopted by Drosophila researchers [10,24,25,[34][35][36][37][38][39]. CRISPR-based techniques are remarkably precise and, therefore, ideal for replacing, validating, and complementing traditional approaches, in particular procedures relying on the expression of RNAi or cDNA transgenes [40,41]. Also, the large worldwide collection of gRNAs stocks has ensured the quick adaptation of CRIS PR/Cas9 into mainstream Drosophila research [6,42,43]. Given the potential of CRIS PR/Cas13-based methods to replace current techniques, we explored its feasibility and reliability in Drosophila.
Our lab studies signaling pathways that control ecdysone and heme biosynthesis in the larval prothoracic gland (PG), which is part of a larger structure called the ring gland. The PG is a popular model for investigating fundamental aspects of insect endocrinology and allows for the study of external cues that control the timing of ecdysone pulses [44]. Recently, we carried out a genome-wide PG-specific RNAi screen that identified 1906 genes with critical roles in larval development [45]. In follow-up experiments, however, we often were unable to validate the RNAi-induced phenotypes by independent RNAi lines, either because no such lines existed or because other RNAi lines did not replicate the phenotype. This prompted us to develop CRISPR-based methods that could validate the RNAi results by an unrelated methodology. We previously generated two CRISPR/Cas9 toolkit collections and could use them to validate some RNAi phenotypes. However, specific issues still exist, including inconsistent gRNA efficiency and early lethality. We sought to investigate the possibility of adapting the CRISPR/Cas13 system for interference and other potential applications of this system in Drosophila melanogaster.
We generated and evaluated the catalytic activity of Drosophila codon-optimized Cas13 (a-d) variants in a cell line derived from Sg4 embryonic cells. We refer to these Cas13 variants as CasFA[n], CasFB[n], CasFC[n], and CasFX[n], respectively (F = fruit fly, A-C indicates the Cas13 subfamily, CasFX is the fly version of CasRX, and [n] indicates variant number) (Fig. 2a-d). "CasRX" was coined by Konermann et al. for the Cas13d ortholog isolated from Ruminococcus flavefaciens XPD3002 to distinguish it from other Cas13d variants [46]. Since we generated fly-optimized versions of CasRX, we refer to these versions as CasFX. Once we had identified a fly-optimized Cas13 variant, we used this variant to adapt existing Cas13 mammalian cell culture applications for Drosophila cells, such as transcript tracking and RNA modification [23,[47][48][49][50][51][52]. These ex vivo procedures formed the basis for generating a collection of transgenic CRISPR/Cas13 tools designed for in vivo RNA targeting. In particular, we generated four Cas13 transgenic lines, namely two that either ubiquitously express CasFB or CasFX, and two that express either CasFB or CasFX under UAS control. The UAS lines allow tissue-specific expression of CasFX and CasFB by crossing them to Gal4expressing flies. As proof-of-principle that these Cas13 transgenes work effectively Fig. 2 Efficiency evaluation of Drosophila codon-optimized Cas13 variants. a-d qPCR analysis showing eCFP transcript levels in Sg4 cells as a function of the different Cas13 variants that were expressed in these cells (a-d, respectively). Shown are relative fold changes of eCFP transcript being targeted by two independent crRNAs, crRNA 1 (red) and crRNA 2 (cyan). Data were normalized to eCFP expression levels when using a blank crRNA as a control (blue dotted line = 1). * = p value < 0.05, ** = p value < 0.01, *** = p value < 0.001, ns = not significant, p values based on Student's t test, error bars represent 95% confidence intervals. e Fluorescence changes of eCFP across samples targeted by the Cas13/crRNA 2 complex. Fluorescence levels were measured using ImageJ and normalized to signals obtained with a blank crRNA (control). Nuclei were stained with Nuclear Green LCS1 (ab138904). Color was adjusted for color-blind-friendly purpose. eCFP and DsRed fluorescence were measured using their native fluorescence properties (no antibody staining). Scale bar = 50 μm in vivo, we generated seven crRNA transgenes to target three genes we are studying in our lab.

Results
Generation and characterization of Drosophila-optimized Cas13 enzymes We generated ten Cas13 variants for each of the four Cas13 family members (a-d) by optimizing different codon subsets for codon usage in Drosophila. Specifically, we made ten constructs based on the Leptotrichia wadei Cas13a gene (LwaCas13a), ten variants based on the Prevotella sp. P5-125 Cas13b gene (PspCas13b), ten versions based on the Fusobacterium perfoetens Cas13c gene (FpeCas13c), and ten forms of the Ruminococcus flavefaciens XPD3002 Cas13d gene (aka CasRX) (Additional file 2: Table S1). We chose these Cas13 orthologs for the following reasons: (i) based on studies in mammalian and plant cells, LwaCas13a, PspCas13b, and CasRX showed improved and robust catalytic efficiency when compared to other Cas13 orthologs [23,46,50,53], (ii) unlike some Cas13 orthologs, the Cas13 genes we chose for our studies do not require a specific protospacer flanking sequence (PFS) for efficient target RNA identification [23,46,50,53]. In the case of PspCas13b, the original study, which was performed in Escherichia coli, showed that the PFS is necessary for RNA cleavage activity. However, when the same enzyme was tested in mammalian cells and plants, the PFS was no longer required [48,49,52]. Finally, (iii) we also selected Cas13c, since only a few studies have examined this Cas13 subtype [23].
To evaluate the RNA degradation efficiency of these fruit fly-optimized Cas13 enzymes, we needed to establish a stable reporter gene cell line. For this, we used the PhiC31 integrase system to generate a dual-reporter transgene in the Drosophila embryo cell line Sg4-PP-27F [54] that simultaneously expressed eCFP (enhanced Cyan Fluorescent Protein) and DsRed (Discosoma Red fluorescent protein) (Additional file 1: Fig. S1A, C). Sg4 is one of four embryonic cell lines isolated from the original Schneider's line 2 (S2) and differs from the popular S2 cells in a range of transcriptional properties [55]. Importantly, Sg4-PP-27F cells were modified from the original Sg4 cells by adding a PhiC31 docking site to the second chromosome [54]. The inserted eCFP and DsRed transgenes are each controlled by the ubiquitous actin 5C promoter (act5C). To ensure this transgene's stability, we added a Neo R gene cassette, which encodes aminoglycoside kinase and ensures cell survival in the presence of G418 antibiotics [56]. We refer to this new transgenic cell line as Sg4_CD (C = eCFP, D = DsRed), and our subsequent cell culture experiments were based on this line. To transform the Sg4_CD cell line with appropriate vectors, we generated plasmids that harbored a single copy of a given Cas13 variant and a single crRNA (the vector allows for adding multiple crRNAs). These constructs, here referred to as pC13cr01, allowed us to simultaneously express Cas13 as well as its crRNA in transfected cells (Additional file 1: Fig. S1D, Additional file 3: Table S2). To ensure stable transfection, we also included the PURO gene in the pC13cr01 vector. The PURO gene encodes the puromycin Nacetyltransferase, which allows cells to survive in media supplemented with puromycin [57,58] (Additional file 1: Fig. S1D, Additional file 3: Table S2). Thus, the presence of two resistance markers allowed for dual selection during the transfection experiments. Besides testing the Drosophila-optimized Cas13 variants, we also examined the efficiency of the original Cas13 orthologs in the Sg4_CD cell line (Additional file 2:  Tables S1, Additional file 3: Table S2).
We measured the efficiency of the Cas13 variants by targeting one of the two reporter gene mRNAs and quantifying mRNA levels via qPCR. To accomplish this, for each Cas13 variant, we used two independent single crRNAs targeting eCFP mRNA (crRNA1 and crRNA2, Fig. 2), while the DsRed mRNA was not targeted and served as a control (Additional file 1: Fig. S2, Additional file 4: Table S3). To ensure that any observed differences derived only from the catalytic activity of the Cas13/crRNA complex, and not from either Cas13 or the crRNA itself, we also tested the eCFP expression level in the presence of a non-targeting (NT) Cas13/crRNA complex. In our hands, the different Cas13 variants showed a wide range of RNA-targeting efficiency, with some of the variants failing to trigger RNA degradation. The original Cas13a (aka LwaCas13a) showed roughly 35-40% eCFP knock-down efficiency, while the best-performing Drosophila variant, CasFA5, was only slightly better and exhibited 47% efficiency (Fig. 2a). For the Cas13b (aka PspCas13b) variants, we measured 45-51% efficiency for the original Cas13b enzyme, while the best-performing Drosophila variants were CasFB5 and CasFB8, both of which were 65-70% efficient (Fig. 2b). The Cas13c group was the least efficient in knocking down eCFP, with the best line, CasFC4, only accomplishing a 25% knock-down (Fig. 2c). In contrast, the Cas13d group performed best, displaying 82% efficiency for the original Cas13d (CasRX) enzyme, whereas the CasFX4 variant was even better and reached a 90% knock-down (Fig. 2d).
To validate these qPCR data, we quantified the protein levels of eCFP and DsRed based on their fluorescence and Western blotting. We selected the best-performing enzyme variants from all four groups, namely three CasFA variants, four CasFB versions, one CasFC enzyme, and six CasFX forms. We then assessed the efficiency of the eCFP knock-down via immunofluorescence (Fig. 2e) and Western blotting (Additional file 1: Fig. S3A-D). Both approaches showed comparable results and confirmed that CasFX4 was the overall most efficient Cas13 enzyme of the entire cohort, showing~90% and9 5% efficiency on the mRNA and protein levels, respectively. Next, we sought to investigate whether the subcellular localization of Cas13 would affect the enzyme's catalytic activity. Since mRNAs mature in the nucleus but are translated in the cytoplasm, we wondered if Cas13 performance could be improved by identifying which cellular compartment is optimal for Cas13 activity. To test this, we selected the original Cas13 variants and their corresponding best-performing Drosophila counterparts (CasFA5, CasFB8, CasFC4, and CasFX4) and fused them either with a nuclear localization signal (NLS) or a nuclear export signal (NES) (Additional file 1: Fig.  S3E). These constructs were based on similar designs from other studies and our approaches (Additional file 1: Fig. S3F) [8,10,12,35,39,[59][60][61][62]. Then, as described above, we again examined how efficiently eCFP was knocked down. Overall, we observed similar efficiencies when the same Cas13 variant was tested in the nucleus or cytoplasm, indicating that the catalytic activity of these Cas13 variants was independent of the subcellular localization (Additional file 1: Fig. S3G). For LwaCas13a, PspCas13b, and CasRX, this result is consistent with a previous study in plants [52]. Since we found no significant differences, we decided to use Cas13 variants without any localization signal for experiments that followed.
Together, these data suggested that the Cas13 variants retain their RNA-cleaving activity in Drosophila Sg4_CD cells, but efficiencies varied considerably. Among the Drosophila codon-optimized Cas13 enzymes we generated, we noticed consistent and robust efficiency of two CasFB versions (namely CasFB5 and CasFB8) and the overall best Cas13 variant, CasFX4.
Evaluating the collateral activity of Drosophila-optimized Cas13 variants Studies in Escherichia coli showed that once the Cas13/crRNA complex is bound to its target RNA, the HEPN-nuclease domains become active and are capable of cleaving not just the intended target, but also RNA molecules that are in the vicinity of the Cas13/RNA complex, resulting in the non-specific RNA degradation referred to as "collateral activity" (Additional file 1: Fig. S4A) [21][22][23]63]. Subsequent studies reported that the collateral activity of Cas13 varied from system to system. While non-specific RNA degradation was detected in human U87 glioblastoma cells [63], no collateral activity was detected in human embryonic kidney 293FT cells or in the plant Nicotiana benthamiana [23,49,50]. To test for collateral activity in our hands, we examined the best-performing Cas13 variants using the same transgenic cell line Sg4_CD. Specifically, we co-expressed eCFP, DsRed, and Neo R independently, each with an act5C promoter. Since eCFP, DsRed, and aminoglycoside kinase (encoded by Neo R gene) are foreign genetic components, we reasoned that manipulating their expression via Cas13 would not have a significant impact on the physiology of SG4_CD cells. The idea was to target eCFP with specific crRNAs in the presence of Cas13 and monitor the expression of DsRed as a readout for collateral activity. Both eCFP and DsRed were presumed to be highly expressed in a coordinate fashion, since the act5C promoter controlled each transgene. As such, if the interference activity of Cas13 was not specific to eCFP, we expected to detect differences in DsRed expression via qPCR. Using this approach, our data showed that the selected Cas13/crRNA complexes only affected target eCFP expression, while DsRed expression appeared unperturbed (Additional file 1: Fig. S4B).
These data suggest that the tested Cas13 enzymes did not have any detectable collateral activity, at least not in the Drosophila Sg4_CD cell line.

Testing the fidelity of Drosophila Cas13 variants
Our efforts identified several Cas13 versions that efficiently degraded target RNAs in Drosophila cells while exhibiting no detectable collateral activity. Next, we wanted to assess how mismatches between crRNAs and their cognate target RNA would affect RNA degradation as a means to define Cas13 fidelity. In particular, we were curious as to whether Cas13 would display higher fidelity-and as such, lower off-target ratesthan RNA interference (RNAi), which is widely used in a variety of research models, ranging from cell culture to whole organisms [64][65][66]. While RNAi is an attractive and powerful tool, its usability is often hampered by its off-target activity, which can make it challenging to interpret phenotypes, and validation strategies involving codonmodified genes/cDNAs are cumbersome and harbor pitfalls [26,67,68]. Other validation strategies include non-overlapping RNAi constructs targeting distinct regions on the mRNA, classic mutants, or conditional CRISPR/Cas9 approaches. To test the propensity of our Cas13 enzymes to degrade off-target RNAs due to small sequence differences, we selected the six top-performing variants for which we had not detected any collateral activity (CasFA5, CasFB5, CasFB8, CasFC4, CasFX4, and CasFX8). Specifically, we generated mismatches in the crRNA-2 spacer sequence and measured the ability to degrade its target RNA, eCFP. To indicate the mismatch location, we referenced the position of the altered nucleotide relative to the stem loop-forming direct repeat of the crRNA. The nucleotide at position 1 represents the one closest to the DR, and the highest number corresponds to the nucleotide farthest away from the DR.
Among all variants that we tested, all had a central region that appeared to be intolerant to single mismatches. The CasFA5, CasFB5, and CasFB8 variants showed some tolerance to single mismatches outside the core region, namely nucleotides #1-3 at the 5′-end and nucleotides #28 and higher at the 3′-end. In contrast, the core region showed no tolerance to mismatches ( Fig. 3a-d). Remarkably, CasFC4, CasFX4, and CasFX8 variants showed no tolerance for mismatches throughout the entire range, including the extreme 5′ and 3′ ends. To examine this further, we tested the outermost nucleotides for both CasFX variants (position #1 and #30). Even single mismatches at either end of the spacer region abrogated interference activity, indicating that these two variants are highly specific and have the lowest off-target potential (Fig. 3e, f). Since four of the variants had some tolerance towards a single mismatch, we further examined mismatch tolerance by introducing more than one mutation per crRNA. Specifically, we generated constructs encoding two, three, or four mismatches in the eCFP-crRNA. In all tested conditions, we included at least one mismatch from the extreme 5′ or 3′ end of the spacer. In our hands, none of the Drosophila Cas13 variants exhibited tolerance to crRNAs with mismatches of more than one nucleotide (Additional file 1: Fig. S4 C-H). These data are in agreement with other studies using similar approaches [51,69,70]. Taken together, this suggests that the Drosophila Cas13 variants tested here are highly specific and display no tolerance to a single mismatch in the core region of the spacer, and none of the enzymes were functional with two mismatches in the crRNA. The CasFA, CasFB5, and CasFB8 variants did tolerate a single mismatch located at either end outside the core region. In contrast, the CasFC4, CasFX4, and CasFX8 variants appeared to require a perfect match of the entire spacer region to mediate interference. We conclude that the CasFX4 and CasFX8 variants will likely have the lowest off-target rate while retaining optimal RNA-targeting efficiency among the Cas13 enzymes tested here.

Nuclease-dead CasFX for applications involving transcript detection
The CRISPR/Cas9 system has been modified to allow for non-nuclease activities, such as for transcription interference (CRISPRi) as well as transcriptional activation (CRIS PRa) [8,10,12,39]. Similarly, the Cas13 system can also be adapted for other purposes and may be more suitable for certain applications than CRISPR/Cas9-based methods. For instance, the ability to target RNA instead of DNA has the advantage that it is reversible. Also, Cas13 may allow for the development of techniques that cannot be accomplished by the corresponding CRISPR/Cas9 approaches: By abolishing the nuclease activity of Cas13 while retaining its RNA-binding capability, one could use the enzyme to specifically target RNAs to track these transcripts in the cell. Another option would be to fuse Cas13 with different protein domains to affect post-transcriptional processing of target mRNAs, e.g., altering transcript splicing or stability. Specific efforts have been made to investigate these applications with promising results [48-50, 52, 71].
We were particularly interested in a nuclease-deficient Cas13 variant as a tool to validate specific RNA-protein interactions. For our proof-of-principle approach, we selected the Cas13 variant with the most consistent, robust, and specific interference activity, CasFX4 (hereafter referred to as simply CasFX), and introduced quadruple mutations in the catalytic HEPN domains (R239A/H244A/R858A/H863A). These mutations abolish the nuclease activity but not RNA-binding activity in the CasRX variant [50,52,71] (Fig. 4a). We first tested whether the mutant CasFX still retained nuclease Relative expression levels of eCFP when using different Cas13 variants and crRNAs that carry a range of single mismatches along the eCFP-crRNA-2. Data were normalized to samples treated with a blank crRNA (control = C). eCFP expression levels in Cas13/ wild-type (WT) crRNA samples were also included as a reference. * = p value < 0.05, ** = p value < 0.01, *** = p value < 0.001, ns = not significant, p values based on Dunnett's post hoc test, error bars represent 95% confidence intervals activity by testing our validated crRNAs against eCFP in the Sg4_CD cell line. As expected, the mutant CasFX failed to interfere with the expression level of eCFP, whereas the wild-type variant worked efficiently (Fig. 4b, c). We conclude that this mutant CasFX variant, similar to the corresponding variants in other species, lost its nuclease activity. We hereafter refer this variant as dCasFX (d = dead).
To assess whether crRNA-guided dCasFX would specifically interact in a nondestructive manner with its intended target mRNA, we tested its ability to co-IP a protein known to bind to the same mRNA. As such, immunoprecipitation of dCasFX should pull down the mRNA as well as its bound protein, which can be detected via Western blotting. This approach is useful to validate the RNA-binding activity of dCasFX, as well as the interaction between mRNA and the interrogated protein. To test this, we used an isoform of the ferritin heavy chain 1 mRNA (Fer1HCH-RA), which carries a canonical iron-responsive element (IRE) at its 5′-end. This IRE allows iron regulatory protein 1A (IRP1A), the Drosophila ortholog of human iron regulatory protein 1 (IRP1), to bind to the Fer1HCH-RA mRNA [72][73][74][75][76]. Specifically, we used the IRP1A C450S form [74], which is constitutively RNA-binding. We then designed a series of crRNAs that direct dCasFX to its target, Fer1HCH-RA, and tested whether immunoprecipitation of dCasFX would also pull down IRP1A. We transfected and lysed cells containing the dCasFX and crRNA components and mixed this lysate with a second sample obtained by lysing cells containing transfected Fer1HCH-RA mRNA and IRP1A C450S . By combining the two lysates, the dCasFX/crRNA enzyme should bind to the Fer1HCH-RA mRNA/IRP1A C450S complex. If the interaction occurs, immunoprecipitation of dCasFX (via its added HA tag) is expected also to pull down IRP1A C450S (Fig. 4d).
(See figure on previous page.) Fig. 4 Properties of modified Cas13 variants. a Schematic of nuclease-dead CasFX (dCasFX) activity. dCasFX carries quadrupl e point mutations that abolish its nuclease activity. As a result, the dCasFX/crRNA complex can be recruited and bind to target transcripts, but it cannot cleave the RNA. b Evaluation of Cas13 cleavage efficiency of dCasFX compared to wild-type CasFX. qPCR data represent expression levels of eCFP. Data were normalized to samples treated with blank crRNA (control). * = p value < 0.05, ** = p value < 0.01, *** = p value < 0.001, p values based on Student's t test, error bars represent 95% confidence intervals. c eCFP fluorescence when targeted by either CasFX or dCasFX. Nuclei were stained with nuclear green DCS1 (Abcam ab138904). Color was adjusted for color-blind-friendly purpose. eCFP and DsRed fluorescence were measured using their native fluorescence property without using antibody staining. Scale bar = 50 μm. d Schematic of dCasFX for the validation of RNA-protein interactions. dCasFX and crRNA targeting Fer1HCH-RA mRNA were transfected together in one sample. Fer1HCH-RA and IRP1A C450S , a constitutively RNAbinding form of IRP1A that interacts with the iron-responsive element (IRE) in the Fer1HCH-RA mRNA, were transformed together in a different sample. The two samples were each lysed and combined, followed by immunoprecipitation (IP) of dCasFX (utilizing the attached HA tag) to test for the presence of IRP1A in the pull-down assay. e Western blot showing the IP of dCasFX combined with different crRNAs along Fer1HCH-RA mRNA and the detection of IRP1A in corresponding samples. f Functional schematic of CasFX that carries a mitochondrial localization signal (CasFX mt ). At the N terminus, CasFX mt is fused with the tim23 mitochondrial signal sequence. Upon binding with crRNA, the complex will localize into mitochondria and target mitochondrial-encoded transcripts. g Mitochondrial localization of CasFX mt . Nuclei were stained with DAPI (blue) while mitochondria were stained with mitotracker green (Cell signaling 9074S) and CasFX polypeptide was stained with anti-HA antibody (magenta). Scale bar = 25 μm. Color was adjusted for colorblind-friendly purpose. h The relative expression level of mitochondrial-encoded transcripts, COXI and COXII, targeted by RNAi, CasFX O , and CasFX mt . Data were normalized to samples treated with no transfected plasmid (control). * = p value < 0.05, ** = p value < 0.01, *** = p value < 0.001, ns = not significant, p values based on Dunnett's post hoc test, error bars represent 95% confidence intervals. i Western blotting of COXI and COXII when being targeted by RNAi, CasFX O , and CasFX mt A key question for this strategy was how far the recognition site for dCasFX/crRNA needed to be away from the IRE to allow binding of both proteins, dCas13 and IRP1A, to the Fer1HCH-RA mRNA. To this end, we generated nine different crRNAs, representing binding sites spaced~150 bases apart to roughly cover the entire 1.7 kb Fer1HCH-RA mRNA. One of the sites (crRNA #3) partially overlapped with the IRE site, which served as a control to disrupt IRP1A binding. Using this strategy, we found that immunoprecipitation of dCasFX successfully pulled down IRP1A, as long as the cRNA binding site was sufficiently removed from the IRE. As expected, this interaction appeared to be dependent on the distance between the crRNA target site and IRE sequence, since an insufficient distance should cause steric hindrance between the two proteins (Fig. 4e). As a control, we used a non-targeting (NT) crRNA to ensure the interactions we observed were specific. The control showed that immunoprecipitation of dCasFX with a non-Fer1HCH-RA mRNA-targeting cRNA was not able to pull down IRP1A.
We also tested whether we can simply detect immunoprecipitated Fer1HCH-RA mRNA via real-time PCR (qPCR). In the absence of IRP1A, dCasFX appears to bind to the Fer1HCH-RA mRNA efficiently, and we found no significant differences between the nine different crRNAs (Additional file 1: Fig. S5A). Interestingly, when we repeated the experiment in the presence of IRP1A, we noticed a~4-fold reduction of immunoprecipitated Fer1HCH-RA mRNA when we used cRNAs #1-4 (Additional file 1: Fig.  S5B). This is consistent with the results for co-immunoprecipitated IRP1A (Fig. 4e), suggesting that competition between IRP1A and dCasFX (bound to crRNAs #1-4) affected the RNA-binding ability of both proteins. We conclude that dCasFX is a reliable tool to validate interactions between a protein and its candidate target RNA. In addition to RNA immunoprecipitation, dCasFX could potentially also used for other in vivo studies, such as locating a transcript of interest to elucidate its subcellular localization or for co-localization studies, or to determine whether a given protein is bound to its target RNA or unbound.

Targeting mitochondrial RNAs via Cas13
Like CRISPR/Cas9, Cas13 needs to form a complex with a crRNA before it can identify and cleave its target transcript [22,23]. Since the Cas13/crRNA complex harbors a single protein, it can be easily tagged with a mitochondrial targeting sequence to cleave RNA in mitochondria, which is not feasible with RNAi. Drosophila mitochondria contain multiple copies of circular DNA (mtDNA), which encode tRNAs, rRNAs, and polypeptides important for oxidative phosphorylation. The study of mitochondrial genes is important, because mutations in mtDNA can cause devastating human disorders, such as Leber's hereditary optic neuropathy, which causes blindness [77][78][79]. To modify CRISPR/Cas13 applications for mitochondrial-encoded transcripts, we added a sequence encoding an N-terminal mitochondrial targeting peptide derived from the nuclear-encoded translocase of the inner mitochondrial membrane 23 (tim23) gene. For this approach, we generated a modified version of our highly efficient CasFX variant, which we termed CasFX mt . The CasFX mt /crRNA complex is predicted to be imported into the mitochondrial matrix, where it should bind to and cleave the target transcripts (Fig. 4f, g).
To test the functionality and efficiency of the CasFX mt variant, we co-transfected CasFX mt with constructs encoding a crRNAs against either mitochondrial cytochrome c oxidase subunit I (mt:CoI, aka COXI) or mitochondrial cytochrome c oxidase subunit II (mt:CoII, aka COXII). Both COXI and COXII are highly expressed mitochondrialencoded genes critical for oxidative phosphorylation [80][81][82]. We analyzed the expression levels of COXI and COXII via qPCR as well as Western blots. To put these results into context, we generated RNAi samples against each of these targets and used the original CasFX (CasFX O , O = original) variant, which lacks the mitochondrial sequence, as a control. In our hands, RNAi targeting either COXI or COXII had no significant effect on the expression of these two transcripts. Similarly, CasFX O /crRNA produced no significant effects (Fig. 4h, i). In stark contrast, CasFX mt caused a 4-5-fold reduction of the COX transcripts and resulted in a substantial drop in protein levels as well (Fig. 4h,  i). To ensure that this result was reproducible, we tested additional RNAi as well as crRNA sequences, all of which target COXI or COXII transcripts (Additional file 1: Fig.  S2). In all cases, the observed results were comparable (Additional file 1: Fig. S5C-E), suggesting that CasFX mt is a useful tool to target mitochondrial-encoded transcripts.

Cas13-ADAR2 for RNA modification
One intriguing aspect of CRISPR/Cas13 has focused on the modification of RNA, which led to two approaches, namely "RNA editing for programmable A to I replacement" (REPAIR) and "RNA editing for specific C to U exchange" (RESCUE) [47,50]. These methods allow for programmable adenosine-to-inosine editing as well as cytosine-to-uridine editing, respectively. The ability to modify genetic information at the RNA level may be advantageous because, unlike Cas9, which causes a permanent change in the genome, RNA modifications via Cas13 are reversible due to RNA turnover [8,12,39,74]. As such, Cas13-based approaches may be suitable for future therapies, where Cas13 could be used to repair missense mutations in transcripts without affecting a patient's genome.
In the REPAIR systems used in mammalian cells, the nuclease-dead PspCas13b was fused to the RNA-modifying domain of Adenosine Deaminase Acting on RNA 2 (ADAR2). In their original approach, Cox et al. found that the first REPAIR version (REPAIRv1) had substantial off-target activity. Subsequently, they generated REPAIRv2, which harbored two point mutations in the ADAR2 domain (T375G and E488Q). This version showed high specificity and robustness in mammalian cells [50].
Given its success in mammalian cell systems, we wondered whether a Cas13-ADAR fusion would be functional in Drosophila. The insect ADAR protein appears to function similarly to its human counterpart [83], suggesting that constructs based on mammalian ADAR2 would work in Drosophila. We first fused the above-described dCasFX to the mutant human ADAR2 domain that carries equivalent mutations as the REPAIRv2 we mentioned earlier. We refer to this construct as FREPAIRv2 (F = fruit fly) and tested for its editing efficiency (Fig. 5a). To test for Cas13-ADAR2 activity, we generated a system that uses a dual-reporter transgene in the Drosophila embryo cell line Sg4-PP-27F. Similar to the earlier described Sg4_CD line; this cell line carries the independently expressed eCFP and DsRed transcription units in the genome, each with their own actin5 promoters. However, unlike the Sg4_CD line, we introduced a point mutation into the eCFP coding region that converts a tryptophan residue 57 (W57*) TGG into an early stop codon (TGA), which we refer to as eCFP*. Also, we termed this new cell line "Sg4*" line to distinguish it from the original Sg4_CD (Additional file 1: Fig. S1B). Next, we co-expressed FREPAIRv2 and an eCFP-crRNA, which carries a single mismatch A to C at the position that corresponds to the introduced stop codon (Fig. 5a,  b). If the FREPAIRv2 is capable of editing its target RNA encoded by eCFP*, the stop codon should be reverted to the wild-type tryptophan residue (W57), and the resulting full-length eCFP should be detectable via Western blotting and, if efficiency is sufficiently high, via fluorescence from the restored CFP. Using this strategy, we found that we were able to detect fluorescence at a wavelength of 405 nm as early as 36 h after Fig. 5 Adaptation of the REPAIRv2 system to modify RNA in Drosophila Sg4 cell culture. a Schematic for the Drosophila-modified REPAIRv2 system (FREPAIRv2), to modify a mutant eCFP transcript. Mutant eCFP carries an early stop codon that normally encodes Tryptophan at residue 57 (W57*). By generating an A to C mismatch in the crRNA spacer that corresponds to the stop codon, the ADAR2 DD domain will change the equivalent adenosine (A) to inosine (I). Inosine will be treated as guanosine by the translation machinery. b Schematic of FREPAIRv2 outcome. Originally, the mutant eCFP transcript harbors a stop codon at position 57, which will generate a short polypeptide with 56 amino acids. However, once modified by FREPAIRv2, codon 57 will be reverted to wild-type tryptophan and restore the production of a full-length polypeptide. c Western blotting monitoring eCFP productions relative to transfection time. d Fluorescence emitted by eCFP relative to transfection time. Nuclei were stained with nuclear green DCS1 (Abcam ab138905). Color was adjusted for color-blind-friendly purpose. eCFP and DsRed fluorescence were measured based on their natively emitted fluorescence. Scale bar = 50 μm. e Schematic of crRNAs that we used for FREPAIRv2. We considered two criteria for the crRNA design: (i) mismatch distance from the first nucleotide and (ii) spacer length. f Editing rate and off-target rate of FREPAIRv2 concerning mismatch distance when spacer length was kept at a constant 50 nucleotides. Error bars represent standard deviation. g Editing rate and off-target rates of FREPAIRv2 in relation to spacer lengths when the mismatch distance was kept at the constant position 26. Error bars represent standard deviation transfection, indicating that detectable levels of eCFP had been produced. eCFP fluorescence continued to increase, with substantially higher levels at the 60-h time point (Fig. 5d). When we conducted Western blots to validate these data, we saw corresponding results, with detectable eCFP protein at 36 h and progressively higher levels from 42 to 60 h after transfection (Fig. 5c). We conclude that Cas13-ADAR2 works effectively in Drosophila and can be used to modify target mRNAs, such as reverting transcripts carrying missense mutations without altering the genome.
For the above approach, we followed a similar path that was used in the original study [50] where the mismatch (C → A) was placed in the center of the crRNA spacer, measured at the 26th nucleotide of 50 nucleotides (nt) spacer, relative to the stem loop-forming direct repeat of the crRNA. To evaluate the editing efficiency in correlation to mismatch position and spacer length, we tested a series of crRNA constructs with the same spacer length of 50 nt; however, we changed the relative mismatch distance to the hairpin by increments of 8 nt (Fig. 5e). We then performed reverse transcription and sequenced a minimum of ten randomly selected eCFP cDNAs per construct. This was followed by sequencing to assess the fraction of clones that harbored the repaired codon for tryptophan #57, expressed as editing rate (Fig. 5f). Based on our findings, the crRNA that carried the mismatch at position 26 relative to the hairpin ("mismatch distance", Fig. 5e) resulted in the highest efficiency (Fig. 5f), consistent with other studies [50]. We then tested the effect of varying spacer length while keeping the mismatch distance at 26 nt. We tested spacer lengths from 30 to 80 nt, and in all cases, we observed similar efficiencies, all of which were comparable to a 50-nt spacer (Fig. 5g). Based on these findings, we conclude that FREPAIRv2 works best when using a mismatch distance of 26 nt, whereas the spacer length did not appear to affect the editing efficiency [50].
To evaluate the off-target tendencies of FREPAIRv2 in Drosophila cells, we examined the cDNA sequences for additional A → I modifications, which is straightforward since inosine is recognized as guanosine by the reverse transcriptase. However, we scored any unpredicted sequence deviations as potential off-target events and plotted them relative to the mismatch distances and spacer lengths (Fig. 5f, g). This strategy revealed that some off-target effects persisted, albeit at a low level across all crRNAs that we tested. Given that these effects are random, and distributed across multiple RNA molecules, it appears likely that this off-target activity has no or inconsequential impact on phenotypes. However, future studies are needed to improve the specificity of this editing system further.

Generation and characterization of transgenic Cas13 flies
Our data demonstrated that Cas13 works well in Drosophila Sg4 cells and can be used for purposes beyond RNA cleavage. We next sought to generate transgenic fly lines carrying Cas13 variants and characterize their efficacy in vivo. We were also interested in creating a system that allows for temporal and spatial control over Cas13 expression. To this end, we have previously created a Drosophila toolkit for CRISPR/Cas9 based on Gateway-compatible cassettes that allow researchers to insert specific enhancers that drive the expression of the Cas transgene in a tissue of interest [8,59]. While this generates more upfront work compared to Gal4/UAS-based systems driving the expression of Cas9 [12,37], it does simplify the downstream workflow. Also, it reduces unspecific effects since one requires fewer transgenes to build the necessary fly genotype. We, therefore, decided to create a similar Cas13 toolkit. In total, we manufactured two general Cas13 vectors, one based on CasFB and one that uses CasFX, both of which displayed the highest catalytic efficiency in Sg4_CD cells. For our in vivo strategy, we limited our efforts to constructs that would interfere with RNA expression (Additional file 1: Fig. S6A). Based on these all-purpose vectors, we then generated four transgenic lines for further characterization, named here act_CasFB, UAS-CasFB, act_CasFX, and UAS-CasFX (Additional file 1: Fig. S6A). For the generation of crRNAs, we used the previously described multiplexed pCFD5 vector and implemented changes suitable for Cas13 crRNA processing [12]. We refer to the new plasmids as i) pC13B, which expresses CasFB-compatible crRNAs under control of the U6:3 promoter and ii) pC13X, which expresses CasFX-compatible crRNAs under control of the U6:3 promoter (Additional file 1: Fig. S6B, C). Both plasmids will ubiquitously express the tRNA:crRNA units. As the tRNA is processed, its cleavage will result in the release of mature crRNAs that form complexes with Cas13 enzymes. The cloning procedures for these new crRNA plasmids are overall similar to those for the pCFD5 vector, but since some differences exist, we include a detailed protocol in the supplementary material (see supplemental method S1).
When we crossed the Cas13-compatible crRNAs (referred to as 13B for CasFBcompatible cRNAs and 13X for CasFX-compatible crRNAs) targeting either phm, dib, or IRP1A with either ubiquitously expressed or PG-specific Cas13 variants, we observed the same developmental defects we found with our previous strategies (Figs. 6a-d, S2, and S7, Additional file 5: Table S4), indicating that Cas13 worked effectively in Drosophila. The fact that phm 13B , phm 13X , dib 13B , and dib 13X individuals were rescued to adulthood when reared on 20E-supplemented media [8,59] and that IRP1A 13B , as well as IRP1A 13X animals, reached adulthood when dietary iron was provided [74], strongly suggested that the activity Cas13 was highly specific (Figs. 6a-c and S7).
In addition to the above phenotypic analysis, we evaluated dib expression levels via qPCR. We compared the results to other tissue-specific loss-of-function techniques, including samples from two independent RNAi lines and samples from one line where we used transcriptional interference via dead Cas9 (dCas9) to target dib. We found that the two RNAi lines reduced dib expression by 30-40%, whereas the CRISPRi approach via dCas9 lowered dib expression by 50-60%. Concerning the new Cas13 lines, CasFB reduced dib expression by 55-65%, equivalent to the dCas9 data. Remarkably, CasFX showed the strongest knock-down and robustly reduced dib expression by 80-90% (Fig. 6e). These data indicated that Cas13 transgenes work in vivo and may exceed the efficacy of other techniques.
We also tested the ability to target multiple transcripts with a single transgene. For this, we used the pC13X vector and generated a dual-crRNA transgenic line (termed dI 13X ) that ubiquitously expressed a crRNA-targeting dib mRNA as well as a crRNA targeting the IRP1A transcript (Fig. 6f). Target sites for either of these transcripts were the same as before (Figs. 6a, c and S2). As expected, the animals arrested development at the L3 stage, similar to targeting the dib and IRP1A transcripts individually. Consistent with this, neither 20E-nor iron-supplementation alone could rescue these double knock-downs; however, a diet supplemented with both 20E and iron caused a significant rescue to adulthood (Fig. 6g). This makes sense since the two cRNAs interfered with ecdysone production and the regulation of cellular iron homeostasis. To assess whether the simultaneous knock-down of two genes was as efficient as targeting these genes individually, we evaluated dib and IRP1A expression levels via qPCR. We found no significant difference in any of these approaches suggesting that there is no penalty when targeting two genes at the same time (Fig. 6h).

RNA degradation efficiency of Cas13 in Drosophila
We evaluated eleven variants of each reported Cas13 ortholog in Drosophila Sg4 cells, including the well-characterized variant from the original studies and ten Drosophilaoptimized variants. Among all Cas13 enzymes tested, we observed a wide range of efficiencies, even between the versions from the same ortholog. Among them, CasRX and its Drosophila-optimized variants CasFX appeared to have the highest efficiency. For the Cas13a and Cas13b variants, we also identified the optimized variants with reliable efficiency. Even though they were less efficient than CasFX, these variants may still prove useful in circumstances where only a moderate knock-down is desired. On the other hand, Cas13c variants did not significantly alter the expression of target transcripts. We hypothesize that this was caused by several factors: (i) Cas13c is the least characterized Cas13 enzyme, and it might use a mechanism that differs from the other Cas13 enzymes. (ii) Even though the low efficiency of Cas13c was in agreement with previous studies conducted in other species, we cannot rule out the possibility that the Cas13c variants we used were not ideally suited for Drosophila, and (iii) Cas13c might still require a PFS for optimal activity in the fruit fly. Future studies will need to address this.
We noticed that the expression of the PspCas13b and CasRX variants resulted in considerable toxicity when animals were homozygous for these transgenes, causing lethality during the first (L1) or second (L2) instar larvae (Additional file 1: Fig. S8). Interestingly, animals heterozygous for PspCas13b and CasRX transgenes showed no significant lethality. In contrast, animals homozygous for our Drosophila-optimized Cas13 transgenes, namely CasFB and CasFX, showed only moderate lethality, with 51 to 58% reaching adulthood, respectively (80-85% is expected in wild-type populations). As expected, animals heterozygous for these transgenes appeared normal (Additional file 1: Fig. S8). The lethality of Cas13 transgenic animals was also reported in a recent study [88], similar to the results of early versions of Cas9 in Drosophila [8,10]. Since we observed a wide range of efficiencies between the variants, it is possible that each variant also exhibits different levels of toxicity. While the reasons for the relatively high lethality of the original PspCas13b and CasRX constructs (in a homozygous setting) remain unclear, our data suggest that each variant is unique and that perhaps using codon-optimized versions help to reduce the toxicity associated with Cas13.

Beyond RNA cleavage
A few studies have shown that Cas13 may be useful in a broad range of applications and not just RNA cleavage. In this study, we have demonstrated that dCasFX can validate RNA-protein interactions by using an appropriately designed crRNA. We also showed that by adding a mitochondrial localization sequence, one could recruit the CasFX mt /crRNA complex into mitochondria and target mitochondrial-encoded transcripts. We also adopted the REPAIRv2 system from mammalian cell culture into Drosophila Sg4 cells and showed that this system, FREPAIRv2, can efficiently modify target transcripts with an overall low off-target rate. We have not tested other potential applications; however, in theory, Cas13 can be modified for many approaches to study RNA, including splicing, transcript stabilization, or RNA localization.

From in vitro to in vivo
A significant part of the work presented here was based on cell culture experiments. These approaches were ideal to economically evaluate the efficiencies of multiple Cas13 versions in Drosophila. However, our ultimate goal is to establish CRISPR/Cas13 approaches for in vivo studies in model organisms, which has not been accomplished yet. Based on our results of transgenic CRISPR/Cas13 flies, CasFX and CasFB can efficiently target and cleave transcripts of interest in vivo, and as such, represent a compelling alternative to existing methods. This study may also help scientists working with other model organisms to optimize their approach for implementing Cas13 in vivo.

The CRISPR/Cas13-based toolkit
This study has generated two collections of Cas13/crRNA toolkits to study in either cell culture or organisms. For the cell culture toolkit, we have produced the pC13cr01 vectors, which allow the co-transfection of Cas13 variants and the crRNA corresponding to the target transcript. With this vector, one only needs to digest the crRNA backbone with the BbsI enzyme and clone the target site for the crRNA, similar to the generation of the Cas9-compatible gRNA system in pCFD5 or pCFD6 plasmids. For in vivo work, we also established a similar system with Cas13 transgenes already available from our study. Researchers will need to generate their crRNAs against the target transcript. For this, we provide the pC13B and pC13D vectors with the same cloning procedure as pCFD5. We also provided a supplemental method section with a detailed description of the cloning procedures. On the other hand, the UAS-based versions of Cas13 transgenes will also allow scientists to spatially and temporally manipulate Cas13 activity and study transcript of interest at desired tissues.

Conclusions
Just like CRISPR/Cas9 allows for the manipulation of DNA, Cas13 enables us to target any transcript of interest. This is beneficial for approaches where researchers do not want to alter the DNA of the gene of interest, since Cas13 controls gene expression on the RNA level, similar to RNAi. Furthermore, current evidence suggests that Cas13, especially variants from the Cas13d family, display minimal off-target tendencies, and this might help quell concerns regarding RNA targeting. Even though it might be too early to make conclusions about the off-target activity of Cas13, we believe that its high specificity holds excellent promise for future applications. Also, the ability to modify Cas13, such as targeting Cas13 to mitochondria, further expands the range of future applications for this methodology.

Generation of Drosophila-optimized Cas13 orthologs (DmCas13)
To generate fruit fly codon-optimized Cas13 variants, the original Cas13 nucleotide sequences were evaluated by using two independent web tools: (i) ATGme (https:// atgme.org) and (ii) OPTIMIZER (http://genomes.urv.es/OPTIMIZER) [92,93] with the customized codon usage frequency specific for Drosophila [94][95][96]. The two indices, namely the codon adaption index (CAI) and the effective number of codons (ENC), were used to obtain the optimized sequences. CAI has value ranges from 0 to 1 and is used to evaluate the similarity between codon usage of a gene and codon usage of the reference group [97]. Thus, at least in theory, the higher the CAI value, the higher is the gene expression [98,99]. On the other hand, ENC is a measure of codon usage bias with values between 20 and 61. Since the expression of a gene is usually dependent on the availability of tRNA species, one would expect that genes with higher expression will use a smaller subset of codons recognized by the most abundant tRNAs, resulting in lower ENC values [100]. Taking these two factors into consideration, we picked the top 10 variants per Cas13 subtype for further investigation (Additional file 2: Table S1). We reasoned that it would not suffice just to choose the top-scoring variant, and therefore, we also selected other high-scoring sequences. We generated the selected variants via a combination of mutagenesis of the original Cas13 sequences and fusing gBlocks gene fragments from Integrated DNA Technologies (IDT) (Additional file 4: Table S3).

Generation of transgenic cell lines
The original Sg4-PP-27F (#238) cell culture line was obtained from Drosophila Genetics Resource Center (DGRC) and grown in the HyClone SFM4 Insect cell culture (SFM4) medium (GE Lifesciences SH30913.02) with 1% (v/v) streptomycin-penicillin (Sigma P4333) following standard procedures (Invitrogen). To generate the transgenic Sg4_CD cell line, Sg4-PP-27F cells were co-transfected with two different plasmids, where one plasmid contained the PhiC31 integrase gene, and the other was the dualreporter plasmid (Additional file 1: Fig. S1A). The dual-reporter transgenic construct also contained a Neo R gene, which allows for resistance to Geneticin G418 (Sigma 4727878001). Then, 48-72 h after co-transfection, cells were washed and grown in SFM4 medium supplemented with G418 at the final concentration of 200 μg/ml. Transfected cells were maintained on this type of medium (SFM4 with 1% streptomycinpenicillin and 200 μg/ml G418) for at least four passage rounds before being tested for the integration of transgenic constructs via Sanger sequencing.

DNA extraction from cells
Cells were grown, and DNA was extracted as previously described [59]. In brief, cells were collected as pellets and filled with 20 μl of DNA extraction buffer (10 mM Tris-HCl pH 8.2, 25 mM NaCl, 1 mM EDTA pH 8.0, 0.2% v/v Triton X-100, 1× proteinase K (AM2546)). The mixture was vortex for 3 × 30 s and incubated at 37°C for 30 min before heat-inactivated at 95°C for 5 min. Cell lysates were centrifuged at 12,000×g at 4°C for 5 min, and the supernatant was transferred to a new collection tube. One microliter of supernatant was used for PCR amplification at the genomic region spanning target sites. PCR products were purified using the HighPrep™ PCR reagent from MagBio (AC-60005) following the manufacturer's protocol.

Cell culture transfection
Cells were grown in SFM4 medium with 1% streptomycin-penicillin, 200 μg/ml G418, and transfected by the calcium phosphate-based method (Invitrogen). To study the effects of different Cas13 variants, puromycin was added to the medium on the second day after transfection at the final concentration of 5 μg/ml, similar to what was previously used [112,113]. Cells were collected 7 days after transfection to ensure the turnover of already translated eCFP polypeptides [114]. Transfected cells were washed in ice-cold 1× PBS for 3 × 5 min and collected for later experiments.

Cell immunostaining
On the first day of the transfection experiment, coverslips were pre-cleaned in 70% ethanol and placed into a transfection plate (Sigma CLS3516). Cells were then seeded and transfected following the standard procedures (Invitrogen). This allows cells to adhere to the coverslips for subsequent immunostaining. Subsequent transfection procedures were carried out as described in the cell culture transfection section. Seven days after transfection, coverslips were transferred to a clean transfection plate for immunostaining, while cells in the supernatants were collected for cell lysis and protein extraction.
Samples were fixed in 1× PBS 4% formaldehyde (Thermo Fisher #28906) for 15 min at room temperature (RT) with gentle shaking followed by washing in 1× PBS 0.3% Triton (Sigma #T9284) (PBS3T) for 3 × 10 min. Samples were blocked at RT for 30 min in blocking solution (1× PBS3T 5% normal goat serum (Abcam ab138478)) and incubated in primary antibody dilution buffer (antibody diluted in 1× PBS3T and 1% BSA) for 1 h at RT. Samples were then washed in 1× PBS3T for three times with 10 min each, incubated in secondary antibody dilution buffer for 1 h at RT, and then washed in 1× PBS3T with either 1:50,000 DAPI (Cell Signaling #4083) or 1:2000 Nuclear Green DCS1 (Abcam ab138905) for 3 × 10 min. Samples were mounted in Vectashield mounting medium (#VECTH1000). Pictures were taken on Nikon Eclipse 80i Confocal C2+ microscope/camera. We used the following reagents: a monoclonal mouse anti-HA-tag antibody (Abcam ab18181) at the ratio of 1:1000 for 3xHA tagged Cas13 orthologs, mitotracker green (Cell signaling 9074S) at the concentration of 400 nM for staining mitochondria, monoclonal mouse anti-MTCO1 (Abcam ab14705) at the ratio of 1: 2000, and monoclonal rabbit anti-MTCO2 (Abcam ab79793). Secondary antibodies were obtained from Abcam and used at the ratio of 1:2000, including goat anti-mouse IgG H&L Alexa Fluor 488 (ab150113), goat anti-mouse IgG H&L Alexa Fluor 555 (ab150114), and goat anti-rabbit IgG H&L Alexa Fluor 555 (ab150078). eCFP and DsRed signals were captured based on their fluorescence properties without antibody staining. For quantification of the eCFP signal, the mean pixel values of the images were analyzed using ImageJ as the corrected total cell fluorescence (CTCF) following the formula: CTCF = selected cell intensity − (area of the chosen cell × background intensity). The CTCF values were averaged from all biological replicates and normalized to the normalized average CTCF values of no-targeting (NT) crRNA samples.

Western blotting of cell extracts
For cell lysis and Western blotting, 7-day post-transfection cells were collected by centrifugation at 1000×g for 10 min at 4°C and supernatant was removed as much as possible. Cells were washed in ice-cold 1× PBS for 3 × 10 min and lysed in 90 μl lysis buffer (1× PBS, 1% Triton, 1× proteinase K inhibitor) by vortexing for 15 s every 10 min for up to 1 h. Cell lysate was mixed with fresh 4× Laemmli buffer (0.25 M Tris pH 6.8, 8% SDS, 40% glycerol, 25% β-mercaptoethanol, 0.2% bromophenol blue) at the ratio of 3:1 (v/v). Forty microliters of the mixture (1/3 total volume) was loaded on 12.5% SDS gel. Later steps, including gel electrophoresis and Western blotting, were carried out following the manufacturer's (Abcam) instructions. To detect eCFP, monoclonal rabbit anti-GFP-tag antibodies (Invitrogen G10362) were used at a ratio of 1:1000, followed by incubation with a goat anti-rabbit IgG H&L HRP secondary antibody (Abcam ab97051) at a ratio of 1:20,000. To detect DsRed, monoclonal mouse anti-DsRed antibody (Santa Cruz sc-390909) was detected at the ratio of 1:1000, followed by incubation with a goat anti-mouse IgG H&L HRP secondary antibody (Abcam ab97023) at the ratio of 1:20, 000. To detect COXI and COXII, monoclonal mouse anti-MTCO1 antibody (Abcam ab14705) and monoclonal rabbit anti-MTCO2 antibody (Abcam ab79393), respectively, were used at a ratio of 1:500. We used monoclonal mouse anti-β-tubulin antibodies (Sigma 05-661) at a ratio of 1:10,000 to detect tubulin, which served as a loading control. Blots were scanned for image acquisition with a ChemiDoc imaging system (Bio-Rad), and band intensity was measured using ImageJ.

Survival studies
Experiments were carried out as previously described [8,59,74]. In brief, 50 embryos per replicate were collected in 1-h intervals and transferred to vials containing appropriate media. Larval survival was scored for every stage. At least three independent crosses (= three biological replicates) were carried out per experimental condition. Modified media were prepared by adding compounds (e.g., iron or 20E) during the preparation process. For iron-enriched media, a 1 M stock solution of ferric ammonium citrate (FAC) (Sigma #F5879) was used to make a medium with a final concentration of 1 mM FAC. For 20-hydroxyecdysone (20E)-supplemented media, the final concentration was 0.33 mg/ml. For dib 2 , phm E7 mutants or transgenic lines that ubiquitously knock-down dib or phm, fresh embryos were immersed for 5 min in 1× PBS containing 20E at the final concentration of 0.11 mg/ml [59]. Survival rates were normalized to the number of embryos used per replicate. Error bars represent standard deviation (data is normally distributed).

Embryo injection
PhiC31 constructs were injected at 500-600 ng/μl concentrations. Injections were performed either at the University of Alberta or Da Lat University using standard procedures [115]. In total, 300-500 embryos were injected per construct. Surviving adults were backcrossed to w 1118 (for Cas13 transgenes) or y 2 cho 2 v 1 (for crRNA transgenes) and used to generate independent lines.

Quantitative real-time PCR (qPCR)
Studies were performed as previously described [8,74]. The extracted RNA (Qiagen RNeasy extraction kit) was reverse-transcribed via the ABI High-Capacity cDNA synthesis kit (Thermo Fisher #4368814). Synthesized cDNA was used for qPCR (QuantStudio 6 Flex) using KAPA SYBR Fast qPCR master mix #Sigma KK4601). For each condition, three biological samples were each tested in triplicate. Samples were normalized to rp49 based on the ΔΔCT method. Error bars represent 95% confidence intervals.

Statistical analysis
For the survival studies, survival rates were normalized to the starting number of embryos (50 embryos per replicate). Error bars represent standard deviation (data is normally distributed). In the FREPAIRv2 editing experiment, the editing rate was calculated as the percentage of samples with correct modification out of the total number of sequenced samples. The off-target rate represents the percentage of samples with incorrect modifications out of the total number of sequenced samples. Error bars represent standard deviation (data is normally distributed). In qPCR reactions, samples were normalized to rp49, a housekeeping gene, and based on the ΔΔCT method [116], error bars represent 95% confidence intervals and contain the error for the calibrator (which is shown without error bars). For multiple comparisons to the same control, we used one-way ANOVA, followed by Dunnett's test. For multiple pair-wise comparisons (in the RNA immunoprecipitation experiments), we applied one-way ANOVA coupled with Tukey's honestly significant difference (HSD) test. At least three biological samples and three technical replicates were analyzed per condition. For quantification of the eCFP signal in immunostains or Western blots, the mean pixel values of the images were analyzed using ImageJ as the corrected total cell fluorescence (CTCF) using the formula: CTCF = selected cell intensity − (area of the chosen cell × background intensity). The CTCF values were averaged from all biological replicates and normalized to the average CTCF values of no-targeting (NT) crRNA samples. Graphs, standard error calculations, t-tests, Dunnett's tests, and Tukey HSD were conducted in Microsoft Excel, SPSS (IBM), and Prism 8 (GraphPad). All data were normally distributed.

Peer review information
Yixin Yao and Kevin Pang were the primary editors on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.

Review history
The review history is available as Additional file 7.

Availability of data and materials
The datasets supporting the conclusions of this article are available in the Source Data file at Figshare repository [117]. The pC13cr01 vector collection for cell culture work, p13X and p13B vectors for generating in vivo crRNA transgenes have been deposited at the Drosophila Genetic Resource Center (DGRC) and tentatively scheduled to be available in December 2020. Fly strains carrying CasFB and CasFX will be available at the Bloomington Drosophila Stock Center starting January 2021 as a part of the National Institute of Health project (NIH P40ODO18537). Import permit for cell lines Sg4_CD and Sg4* has been obtained and the live stocks will be sent to DGRC in October 2020. Fly strains, plasmids, and cell lines can also be obtained from our lab upon request.
Ethics approval and consent to participate Experiments in this study were conducted in Drosophila cells and live organisms following standard protocols. No human samples were used. No ethics approval is required in this study.