Establishment of CRISPR/Cas9-based knock-in in a hemimetabolous insect: targeted gene tagging in the cricket Gryllus bimaculatus

In what is often called the ‘post-genomic era’ (Wainberg et al., 2021), massive advances in nucleic acid sequencing chemistry over the past two decades have given scientists access to greater volumes of gene sequence data than ever before (Kulski, 2016; Papageorgiou et al., 2018). However, this wealth of genomic information has highlighted two major gaps in our understanding of gene function and evolution. First, comparative genomic data and increased taxon sampling in functional genetics and developmental and cellular biology have revealed that the biology of many traditional laboratory model organisms is not representative of the broader clades to which they belong (Goldstein and King, 2016). Second, our ability to accurately deduce gene function from sequence data is limited to those genes that display high sequence and structural conservation (Ashburner et al., 2000; Gene Ontology Consortium et al., 2020), and tools for manipulating gene function have been developed for only a small fraction of organisms (Russell et al., 2017). Addressing these problems calls for both increased taxon sampling and development of techniques to enable targeted alteration of gene function in understudied organisms. Here, we address both of these issues by developing a method for targeted genome editing, including both knockout and knock-in editing, in a basally branching insect model organism, the cricket Gryllus bimaculatus.

G. bimaculatus is an emerging model organism in a variety of fields of biology (Horch et al., 2017a; Kulkarni and Extavour, 2019). Ease of husbandry (Horch et al., 2017b), detailed developmental staging tables (Donoughe and Extavour, 2016), established gene expression analysis methods (Horch et al., 2017b), and an assembled and annotated genome (Ylla et al., 2021) make this cricket a highly amenable hemimetabolous laboratory model system (Kulkarni and Extavour, 2019). In contrast to the ontogenetically derived model Drosophila melanogaster, many aspects of cricket embryogenesis are thought to resemble putative ancestral developmental modes of insects (Davis and Patel, 2002). For example, the function of several axial patterning genes has been analyzed and compared with that of their D. melanogaster homologues, revealing that the gene regulatory networks governing axial patterning have undergone considerable evolutionary change across insects (Mito et al., 2007, 2008; Matsuoka et al., 2015). G. bimaculatus is also used for the analysis of gene function in tissue regeneration (Mito et al., 2002; Nakamura et al., 2007; Bando et al., 2013), as this cricket can regenerate amputated organs, including legs and antennae, after several rounds of juvenile molts. In addition, G. bimaculatus is used for the analysis of gene function and neuronal circuits in neuronal activity, including learning, memory and circadian clocks (Hedwig and Sarmiento-Ponce, 2017; Matsumoto et al., 2018; Mizunami and Matsumoto, 2017; Tomiyama et al., 2020). Moreover, several species of cricket are being farmed as a new food source for humans because of their high protein and nutrient content (Huis et al., 2013).

Genome editing techniques using artificial nucleases were previously established in G. bimaculatus (Watanabe et al., 2012). However, construction of artificial nucleases is laborious. More recently, use of the clustered regulatory interspaced short palindromic repeat (CRISPR)/associated Cas9 nuclease (CRISPR/Cas9) has emerged and been verified as an efficient tool for genome editing in several arthropod species, including the fruit fly Drosophila melanogaster (Gratz et al., 2013), the beetle Tribolium castaneum (Gilles et al., 2015), the mosquito Aedes aegypti (Kistler et al., 2015), multiple butterfly species (Li et al., 2015; Matsuoka and Monteiro, 2018; Zhang et al., 2017), the amphipod Parhyale hawaiensis, (Martin et al., 2016), the clonal raider ant Ooceraea biroi (Trible et al., 2017), the European honeybee Apis mellifera (Kohno and Kubo, 2018) and the milkweed bug Oncopeltus fasciatus (Reding and Pick, 2020). Among crickets, we and others have reported CRISPR/Cas9-mediated knockouts in Acheta domesticus (Dossey et al., 2023) and in G. bimaculatus (Bai et al., 2023; Nakamura et al., 2022). Knock-in has also been reported for A. domesticus (Dossey et al., 2023), but this approach has not been used to generate the types of enhancer trap or protein trap lines that have proven so beneficial to advancing developmental biology research in other insects (Morin et al., 2001). Here, we briefly review this technique, and investigate and demonstrate its utility for targeted gene trap knock-in genome modification in G. bimaculatus, the most widely used orthopteran model for developmental biology.

In the CRISPR/Cas9 system, short guide RNAs (sgRNAs) recruit Cas9 nuclease to the target sequence, and Cas9 then introduces a double strand break (DSB) at the target sequence. The presence of DSBs triggers the activity of the DNA repair machinery of the cell, either non-homologous end joining (NHEJ) or homology directed repair (HDR). NHEJ is an error-prone machinery, such that insertions or deletions can be generated at the break point (Branzei and Foiani, 2008). By using artificial nucleases to trigger NHEJ, we have previously succeeded in generating G. bimaculatus mutant lines (Watanabe et al., 2012). HDR, however, would offer more precise repair machinery, as the break is repaired through use of a homologous template. By supplying a donor template containing sequence homologous to the target, in principle a desired donor sequence can be integrated into the genome though HDR. Such gene knock-ins, although highly desirable for detailed analysis of the function of genomic regions, are more difficult to achieve than gene knockouts because of the low efficiency of HDR in eukaryotes (Hagmann et al., 1998). Although success with HDR has been reported in some insects, including the silk moth Bombyx mori (Ma et al., 2014; Zhu et al., 2015), multiple mosquito species (Hammond et al., 2016; Kistler et al., 2015; Purusothaman et al., 2021) and mosquito cell lines (Rozen-Gagnon et al., 2021), the beetle Tribolium castaneum (Gilles et al., 2015), the medfly Ceratitis capitata (Aumann et al., 2018), and the squinting bush brown butterfly Bicyclus anynana (Connahs et al., 2022), our attempts at HDR-based knock-in techniques have never succeeded in G. bimaculatus (T.N., unpublished observations). Recently, an efficient gene knock-in method through NHEJ was developed in the zebrafish Danio rerio (Auer et al., 2014). In this method, both the genome and the donor vector are cleaved in vivo, then the terminal genomic and donor sequences are combined through NHEJ. The method is efficient and can integrate longer constructs into the genome than knock-ins achieved through HDR (Auer et al., 2014; Yoshimi et al., 2016). Bosch and colleagues (Bosch et al., 2019) subsequently reported that this knock-in strategy also works in D. melanogaster.

Here, we present evidence that the CRISPR/Cas9 system functions efficiently in G. bimaculatus. We demonstrate the utility of this technique for developmental biology by performing functional analysis of the G. bimaculatus orthologues of the Hox genes Ultrabithorax (Gb-Ubx) and abdominal-A (Gb-abd-A). Furthermore, using a donor vector containing an autonomous expression cassette, we demonstrate that gene knock-in by a homology-independent method works efficiently in G. bimaculatus. We show that this homology-independent gene knock-in method can be applied to identify mutant individuals simply by detecting marker gene expression in this cricket. Efficient targeted genome editing, now including both knockout and gene-tagging knock-in techniques, will pave the way for making this cricket a much more sophisticated model animal for functional genetic laboratory studies.

To determine whether the CRISPR/Cas9 system was functional in the cricket, we first tried to perform a targeted gene knock-out of the laccase 2 (Gb-lac2) gene (Table 1), which regulates tanning of the arthropod cuticle after molting (Arakane et al., 2005). We chose this gene because of its easily detectable loss-of-function phenotype, and because we had previously successfully generated stable mutant lines for this gene by using artificial nucleases (Watanabe et al., 2012). sgRNA target sites were first chosen using the CasOT tool (Xiao et al., 2014), then we chose the final target sequence from among these candidate sequences based on the number of mismatches relative to the other sequences in the genome (>3 mismatches in the whole sgRNA sequence) and GC content (70±10% for the whole sgRNA sequence). Based on these criteria, we designed sgRNAs against the fifth exon of Gb-lac2, which is close to the target regions of the previously reported artificial nuclease experiment (Fig. 1D; Table S1) (Watanabe et al., 2012).

We co-injected 0.5 μg/μl sgRNA and 0.5 μg/μl of Cas9 mRNA into 232 fertilized cricket eggs within 1-3 h after egg laying (AEL) (Table 2). Five days after injection, we evaluated the frequency of mutant alleles in individual eggs using the Surveyor nuclease assay (Qiu et al., 2004) (see Materials and Methods for detailed methods and procedures). We detected cleaved fragments in 13 out of 29 eggs examined (Table 2, Fig. 1C).

We observed mosaic pigmentation of the cuticle in 92% of G₀ hatchlings that emerged from the individual eggs injected with the Gb-lac2 sgRNA and Cas9 mRNA, consistent with Cas9-mediated interruption of the Gb-lac2 gene in some, but not all, somatic cells of the G₀ hatchlings. We raised these hatchlings to adulthood (Fig. 1A′), and crossed these G₀ adults with wild-type crickets of the opposite sex, to determine the efficiency of germline transmission of the Cas9-induced Gb-lac2 mutations to the G₁ generation. We found that 47.4% of those injected with the Gb-lac2 gRNA and Cas9 mRNA transmitted the mutation to their offspring (Fig. 1B). To determine the nature of the Gb-lac2 Cas9-induced alleles, we isolated genomic DNA from each line and analyzed the sequence of the Gb-lac2 locus. We found that several different types of indel mutations were introduced at the target locus (Fig. 1D). These results indicate that this CRISPR/Cas9-mediated genome editing system is functional in the cricket.

To compare phenotypes obtained with targeted gene disruption with those obtained with the RNA interference (RNAi) method that has hitherto been the most common method of performing functional genetics in this cricket (Mito and Noji, 2008), we used the CRISPR/Cas9 system to perform functional analyses of the G. bimaculatus ortholog of the Hox gene Ultrabithorax (Gb-Ubx) (Table 1). We have previously examined RNAi-induced phenotypes for Gb-Ubx in developing abdominal segments (Barnett et al., 2019), providing a basis for comparison with CRISPR-induced mutants. We designed sgRNA for a sequence within an exon upstream of the homeodomain (Fig. 2A), and co-injected 0.5 μg/μl of this sgRNA and 1 μg/μl Cas9 mRNA into 167 fertilized cricket eggs within 1-3 h AEL. Seven days after injection, we extracted genomic DNA from a small subset of injected eggs and performed the Surveyor assay to determine the efficiency of gene targeting. We found that indel mutations had been induced at the Gb-Ubx locus in all examined eggs (n=16) (Fig. 2B). The remaining 151 injected G₀ embryos gave rise to ten adults, which we backcrossed individually to wild-type adults of the opposite sex. We randomly chose ∼30 G₁ eggs from each of the ten G₀ crosses, extracted genomic DNA from the pooled embryos, and performed the Surveyor assay. We found that six out of ten G₀ crickets transmitted Gb-Ubx mutations to the next generation. We selected one of the six G₁Gb-Ubx^CRISPR lines, which had a frame-shift mutation in the Gb-Ubx locus, for further phenotypic analysis. These Gb-Ubx^CRISPR mutants displayed two different classes of phenotype. The first was contraction of the T3 leg. Wild type G. bimaculatus adults have large, conspicuous T3 jumping legs. However, heterozygous Gb-Ubx^CRISPR mutants had smaller T3 legs than wild-type animals (Fig. 2C). Homozygous Gb-Ubx^CRISPR mutants obtained in the G₂ generation had T3 legs that were even smaller than those of heterozygotes, almost the same size as T1/T2 legs (Fig. 2C). This specific phenotype is not directly comparable with previously reported Gb-Ubx RNAi experiments (Barnett et al., 2019) because embryos in those experiments were not reared to hatching. However, the Gb-Ubx^CRISPR phenotypes were in good correspondence with those previously observed for Ubx RNAi in the cricket Acheta domesticus (Mahfooz et al., 2007). The second phenotype was transformation of the A1 appendage. Wild-type G. bimaculatus germ band stage embryos possess two appendage-like organs on the A1 segment called pleuropodia (Fig. 2D; Rathke, 1844; Wheeler, 1892). Instead of the pleuropodia present in wild-type adults, the appendage outgrowths on the A1 segment of homozygous Gb-Ubx^CRISPR mutants were transformed towards leg-like structures, which is consistent with detection of expression of Gb-Dll and other leg patterning genes in the leg-like structures. (Fig. 2D and Fig. S1). This phenotype matches that previously observed in Gb-Ubx RNAi embryos (Fig. 2E; Barnett et al., 2019). Gb-Ubx^CRISPR heterozygous mutants were fertile but homozygous mutants were lethal. Therefore, to maintain this line, Gb-Ubx^CRISPR heterozygous mutant animals of separate sexes were crossed to each other, and we performed the Surveyor assay to isolate heterozygous mutants among their offspring.

To confirm whether the production of Gb-Ubx protein was indeed disrupted by these CRISPR-induced mutations, we performed immunostaining with the ‘UbdA’ monoclonal antibody FP6.87 (Kelsh et al., 1994), which recognizes both Ubx and Abd-A proteins, and was previously reported to crossreact in multiple Gryllus species, including G. bimaculatus (Barnett et al., 2019; Mahfooz et al., 2004). In wild-type embryos, the UbdA antibody revealed the expected combination of the Gb-Ubx and Gb-Abd-A expression patterns (Fig. 2E and F; Barnett et al., 2019; Mahfooz et al., 2004). In Gb-Ubx^CRISPR embryos, however, although the UbdA expression domain was retained in the abdomen where Gb-abd-A is expressed, expression was undetectable in the T3 and A1 segment where Gb-Ubx is expressed (Fig. 2F), suggesting that the CRISPR-induced Gb-Ubx mutations interfered with Gb-Ubx protein production.

We further characterized the phenotype of Gb-Ubx^CRISPR embryos by observing the Gb-Dll expression pattern (Fig. 2D). In wild-type embryos, Gb-Dll is expressed strongly in the presumptive tarsus of all three developing T1, T2 and T3 legs (Fig. 2D). In the tibia, however, Gb-Dll is expressed differently in T1/T2 and T3 legs. Specifically, in T3 there is a sharply defined border between a distal domain of high expression and a proximal domain of lower expression, but T1 and T2 lack such a border and display moderate proximal expression, as well as strong distal expression as reported by Niwa et al. (1997) (Fig. 2D). In Gb-Ubx^CRISPR embryos, Gb-Dll expression in the presumptive tibia of the T3 leg lacked a strong boundary between high and low tibial expression levels, and instead resembled the expression pattern of wild-type T1 or T2 legs (Fig. 2D). In addition, the T3 leg-specific patterns of multiple leg-patterning genes were absent in Gb-Ubx^CRISPR embryos, which is consistent with disruption of the Gb-Ubx locus (Fig. S1). Taken together, these results suggest that the CRISPR/Cas9 system induced mutations specifically at the Gb-Ubx locus, which disrupted Gb-Ubx function.

To optimize the genome-editing procedure, we wished to evaluate whether and how the timing of injection affected NHEJ mutagenesis. For detailed assessment of this mutagenesis, we therefore performed in-depth analysis of the CRISPR mutants using next-generation sequencing.

Our previous study had revealed early cellular dynamics during cricket embryogenesis (Nakamura et al., 2010), allowing us to assess whether specific mutagenesis events were correlated with cellular behaviors during early development. As in D. melanogaster (Foe and Alberts, 1983), early mitotic divisions in G. bimaculatus embryos are syncytial, meaning that mitosis takes place without cytokinesis, resulting in multiple energids (nuclei surrounded by aqueous cytoplasm but lacking a unique lipid bilayer) within a single cell membrane (Donoughe et al., 2022; Donoughe and Extavour, 2016; Nakamura et al., 2010; Sarashina et al., 2003). To evaluate whether and how the timing of injection affected mutagenesis outcomes, we chose four early embryonic time points after the 1 h embryo collection period, as follows (Fig. S2A): (1) at the 1 h injection time point, when energids start to migrate from the center of the egg to the cortex; (2) at the 3 h injection time point, when energids continue to become distributed throughout the yolk, accompanied by mitotic cycles; (3) at the 5 h injection time point, when energids have become nearly uniformly distributed throughout the egg cortex and begin tangentially oriented nuclear division; and (4) at the 9 h injection time point, which is 1-8 h before cellularization (Donoughe and Extavour, 2016). We co-injected 0.5 μg/μl of the Gb-Ubx sgRNA, the Gb-lac2 sgRNA or an sgRNA targeting abdominal-A (Gb-abd-A); see the section ‘Knock-in of donor vector sequence at the Gb-abd-A locus’) described above, and 1 μg/μl of Cas9 mRNA into the eggs at each of these time points. At 5 days AEL, we isolated genomic DNA from three individual embryos for each injection time point and used it for amplicon sequencing. We examined the sequences of the on-target site and of the single highest predicted potential off-target site for each of the Gb-Ubx and Gb-abd-A genes. For each sample, we performed amplicon sequencing with three replicates.

For Gb-Ubx on-target gene disruptions, we found that the rate of NHEJ-induced mutations decreased with the age of the embryo at injection (Fig. S2B). The rate of NHEJ-induced mutations at the studied off-target site was less than 1.3% for all injection time points (Fig. S2C), suggesting that off-target effects may be minimal in this system. The same trend was also observed for Gb-abd-A on-target mutations (Fig. S2D-E; see section ‘Targeted mutagenesis of the Gb-abd-A locus via knock-in’). This result is well correlated with the phenotypic severity observed in the G₀ hatchlings emerging from the Gb-lac2^CRISPR embryos. Gb-lac2^CRISPR embryos injected at the two earlier time points (1 h and 3 h) gave rise to hatchlings with broad patches of white cuticle (Fig. S2H). In contrast, the embryos injected at 5 h showed milder phenotypes (Fig. S2H) and the embryos injected at 9 h showed little detectable phenotype (Fig. S2H).

In addition to targeted sequence deletions, targeted sequence knock-in is a highly desirable technique that would expand our ability to understand the functions of genomic regions of interest. We had previously attempted to achieve targeted gene knock-ins through HDR, but this method has not worked in G. bimaculatus in our hands to date (T.N., unpublished observations). In a homology-independent knock-in method reported for D. rerio and D. melanogaster (Auer et al., 2014; Bosch et al., 2019), both genome and donor vector are cleaved in vivo, then the cut ends of genome and donor vector are combined through NHEJ. This method is more efficient than the homology-dependent method, potentially because NHEJ is highly active throughout the cell cycle in eukaryotes (Hagmann et al., 1998). However, owing to the nature of NHEJ, the orientation of integration of the donor vector sequence cannot be controlled. In addition, indel mutations are generated at the junction point. To try to circumvent these issues, which might otherwise prevent functional knock-in, we generated a donor vector containing an autonomous expression cassette comprising the G. bimaculatus actin (Gb-act) promoter followed by the eGFP coding sequence (Nakamura et al., 2010). As a sgRNA recognition site, we included a partial DsRed gene sequence (Auer et al., 2014), which is native to the coral Discosoma sp. (Baird et al., 2000) and not present in the cricket genome. We predicted that successful knock-in of this donor sequence into the genome would result in eGFP expression being driven by the Gb-act promoter regardless of the orientation of the insert or any potential induced indel mutations. To try to further increase the utility of this tool to facilitate identification of targeted gene disruptions, we targeted knock-in of the donor sequence to an exon of the target gene, which we anticipated would result in disruption of target gene function. Our goal was to be able to identify such successfully knocked-in individuals by detectable eGFP expression in the known expression domains of the target gene.

We chose the Gb-Ubx locus for this targeted knock-in strategy and used the same sgRNA as that used for the knockout experiment described above (see section ‘Targeted mutagenesis of the Gb-Ubx locus via knockout’, Fig. 2A). We co-injected 50 ng/μl of sgRNA for the Ubx locus, 50 ng/μl of sgRNA for the donor vector, 100 ng/μl of Cas9 mRNA and 100 ng/μl of donor vector into fertile cricket eggs. By 7 days after injection, four out of 85 injected embryos (4.7%) showed mosaic eGFP expression in the T3 trunk and leg (Fig. 3B′). Of the 85 injected embryos, 18 individuals (21.2%) grew to adulthood. We crossed them individually with wild-type counterparts of the opposite sex and evaluated eGFP expression in their offspring. One out of the 18 G₀ crickets (5.6%) produced G₁ embryos with eGFP expression in a pattern identical to that of Gb-Ubx (Fig. 3C′; Barnett et al., 2019; Matsuoka et al., 2015; Zhang et al., 2005). The eGFP expression was detectable through the eggshell even at late embryonic stages (Fig. 3C′). At adult stages, G₁ knock-in crickets showed detectable eGFP expression in the hind wing and T3 legs (Fig. S3A′,B′).

To confirm the integration of donor sequence into the genome, we performed PCR and sequence analysis. We designed specific primers for each 5′ and 3′ junction point (a and b in Fig. 3A). All three examined embryos showed the expected amplicon size for both junctions (Fig. 3G), suggesting that at least two copies of the donor vector fragment were integrated into the genome. Sequence analysis further confirmed the integration of the donor plasmid into the genome, and that indel mutations were generated at each junction (Fig. 3H). To determine how many copies of donor sequence were integrated into the genome, we performed copy number estimation by quantitative RT-PCR. The copy number of the donor plasmid was estimated as the copy ratio of the eGFP gene to that of the endogenous orthodenticle gene (Gb-otd), which is known to have only one copy in the genome (Nakamura et al., 2010; Ylla et al., 2021). The results of this analysis indicated that three copies of the donor plasmid were likely integrated into the genome (Fig. S4A).

To determine whether the function of the target gene was indeed disrupted by this knock-in/knockout (KI/KO) strategy, we examined eGFP expression and morphology in G₂Gb-Ubx^CRISPR-KI embryos. Among the G₂Gb-Ubx^CRISPR-KI embryos, we found they displayed one of two different intensities of eGFP expression (Fig. 3E′,F′). We determined their genotype by quantitative RT-PCR, and found that the individuals with strong eGFP expression were homozygous mutant, and the individuals with weak eGFP expression were heterozygotes (Fig. S4B). Consistent with the genotyping result, crickets with weak eGFP expression displayed no detectable morphological abnormalities (Fig. 3E,E′). Crickets with strong eGFP expression, however, had smaller T3 legs and formed leg-like structures rather than pleuropodia on the A1 segment (Fig. 3F,F′). These phenotypes, which were the same as those observed in the Gb-Ubx^CRISPR homozygous mutant (Fig. 2C,D), further suggested that the strong eGFP-expressing crickets were homozygous mutants.

To confirm the efficiency and utility of this method, we next chose the Hox gene Gb-abdominal-A (abd-A) as a target (Table 1). We designed sgRNAs for the sequence within the exon immediately upstream of the homeodomain (Fig. 4A). We co-injected 50 ng/μl of sgRNA for the Gb-abd-A locus, 50 ng/μl of sgRNA for the donor vector, 100 ng/μl of Cas9 mRNA and 100 ng/μl of donor vector into fertilized cricket eggs.

Of 38 injected G₀ embryos, five showed mosaic eGFP expression in the abdomen (Fig. 4B). Four G₀ adults were individually backcrossed with wild-type counterparts of the opposite sex to obtain multiple G₁ crickets. We obtained one stable transgenic line, in which eGFP expression in G₂ embryos was similar to the previously documented expression pattern of Gb-abd-A transcript (Barnett et al., 2019; Matsuoka et al., 2015; Zhang et al., 2005) (compare Fig. 4B with Fig. S1K). In a replicate injection experiment, we obtained a second such transgenic line (Table 2). PCR and sequence analysis confirmed that one of the two lines contained the plasmid fragment in the sense orientation, and the second line contained the plasmid fragment in the antisense orientation (Fig. 4C). Copy number estimation analysis results suggested that a single plasmid fragment was integrated into the genome in each line (Fig. S4B).

In the Gb-abd-A^KI-exon lines, eGFP expression was detectable in nymphs even through the cuticle (Fig. S3C′,D′). We further detected eGFP expression in adult male and female internal organs (Fig. S5). In wild-type females, a pair of ovaries, each comprising hundreds of ovarioles, is located in the anterior abdomen (Nandchahal, 1972). Mature eggs are located at the posterior of each ovariole, and eggs are subsequently moved further posteriorly through the oviduct. The posterior end of the oviduct is connected to the uterus, where fertilization takes place, located at the base of the ovipositor (Fig. S5C″). In Gb-abd-A^KI-exon mutant females, eGFP expression was detected in the posterior portion of the oviduct (compare Fig. S5B′,B″ with Fig. S5A′). We found, however, that the oviduct and uterus were not connected in Gb-abd-A^KI-exon mutant females, and these females were not able to lay eggs. Gb-abd-A^KI-exon mutant females also generated ectopic ovipositors (Fig. S6U). In Gb-abd-A^KI-exon males, ubiquitous eGFP expression was detected throughout the testis (compare Fig. S5E′,E″ with Fig. S5D′). The observed eGFP expression in females is reminiscent of the expression pattern of D. melanogaster abd-A in the developing female genital disc [which gives rise to the somatic reproductive structures, including the oviduct in this fruit fly (Epper, 1983; Sánchez and Guerrero, 2001)] and in the adult oviducts (Foronda et al., 2006).

We also detected eGFP expression at the anterior tip of the testis in Gb-Ubx^KI-exon adult males (Fig. S5G). abd-A expression has not, to our knowledge, been previously detected in the D. melanogaster male genital disc (Freeland and Kuhn, 1996), which gives rise to male somatic reproductive structures. However, high-throughput sequencing data from the modENCODE project do report abd-A expression in the adult D. melanogaster testis (Brown et al., 2014).

To confirm whether the eGFP expression observed in Gb-abd-A^KI-exon and Gb-Ubx^KI-exon animals reflects the endogenous expression of target genes in a given tissue, we performed quantitative RT-PCR on eGFP-positive and eGFP-negative tissues. In a previous study, we performed an RNA-seq analysis using G. bimaculatus adult brain and gonad tissue from both sexes (Whittle et al., 2021). In this dataset, we did not detect expression of Gb-Ubx or of Gb-adb-A in the brain (Fig. S7A,B). We therefore used brain as a negative control tissue for the expression of Gb-Ubx and Gb-abd-A. We found a clear correlation between eGFP and Gb-abd-A expression levels in the oviduct (eGFP-positive) and the uterus (eGFP-negative) in Gb-abd-A^KI-exon females (Fig. S7C,D). In the adult testis of Gb-abd-A^KI-exon males, where we observed strong eGFP expression, we confirmed that the expression of eGFP and Gb-abd-A was also well correlated (Fig. S7E). These results indicate that eGFP expression correlates well with Gb-abd-A expression, and are consistent with the interpretation that eGFP expression in these knock-in crickets reflects target gene expression.

In contrast, we observed no correlation between Gb-Ubx and eGFP expression levels in either eGFP-positive (anterior tip of ovary) or eGFP-negative (rest of the ovary) tissues (Fig. S7F,G). It might be possible that Gb-Ubx is expressed at low levels throughout the ovary and that we were therefore unable to detect eGFP expression in this organ, except for at the tips of ovaries. We found that the Gb-Ubx expression level in the wild-type ovary and ovary tip was comparable with the level detected in the brain; in other words, it was effectively undetectable (Fig. S7H). These results suggest that Gb-Ubx is not expressed, or is expressed only at levels undetectable in our transcriptome (Whittle et al., 2021), at the tip of wild-type ovaries, even though eGFP expression was detected there in Gb-Ubx^KI-exon adults (Fig. S5G). Taking into account the fact that three copies of the donor sequence were integrated into the genome in the Gb-Ubx^KI-exon line (Fig. S4), it might be possible that the observed eGFP expression in the tip of the ovary from Gb-Ubx^KI-exon was an artifact due to the multiple copies of the expression cassette integrated at the Gb-Ubx locus, rather than being reflective of a true wild-type Gb-Ubx expression domain. It is also possible that the Gb-act promoter used in the expression cassette might be more or differently active compared with the endogenous Gb-Ubx promoter. The fact that the Gb-Ubx^KI-exon line did not show any abnormality in ovary development is consistent with this notion. Taken together, we conclude that it is possible to visualize target gene expression by knocking an expression cassette into a target gene locus, but careful verification should be carried out before drawing conclusions about domains and levels of target gene expression.

Kimura and colleagues (Kimura et al., 2014) demonstrated that, in D. rerio, the homology-independent method could be applied for trapping endogenous enhancer activity by inserting a donor sequence containing an expression cassette into the 5′UTR of genes of interest. We aimed to apply this technique to G. bimaculatus by attempting to knock-in a donor vector into the intronic region of Gb-abd-A (Fig. 5A).

We co-injected an sgRNA against an intron of Gb-abd-A, together with all other relevant reagents, as described above (sgRNA against the donor vector, donor vector and Cas9 mRNA) (Fig. 5A). Of 100 injected eggs, two eggs showed mosaic expression of eGFP in the abdomen (Fig. 5B). When the donor sequence was inserted into an exon in the previous experiment (see the section ‘Knock-in of donor vector sequence at the Gb-abd-A locus’), eGFP expression was accompanied by a phenotype of ectopic leg-like structure development on abdominal segments (Fig. 4B; Fig. S6), as previously observed in Gb-abd-A RNAi experiments (Barnett et al., 2019). However, when the plasmid fragment was inserted into an intron, the region expressing eGFP did not generate ectopic leg-like structures in G₀ embryos (Fig. 5B). This apparent absent or minimal loss-of-function phenotype in the intron knock-in embryos might explain the relatively high survival rate of the intron-targeted G₀ embryos (28% of injected G₀ embryos survived to adulthood) compared with that of the exon-targeted G₀ embryos (4.9% to 8.5% of 41 and 47 injected embryos survived to adulthood; Table 2).

We obtained one Gb-abd-A^KI-intron line, in which we confirmed that one donor vector sequence was likely integrated into the target region in a forward orientation (Fig. 5C). We carefully inspected the morphology of the Gb-abd-A^KI-intron adult crickets to assess whether potential post-embryonic functions of the target gene were affected by insertion of the donor sequence into an intron. G₁ heterozygous Gb-abd-A^KI-intron females did not show the supernumerary ovipositors observed in G₁ heterozygous Gb-abd-A^KI-exon female adults (Fig. S6W compared with S6U). For further confirmation, we examined the morphology of G₂ homozygous Gb-abd-A^KI-intron mutants. Approximately 25% of examined G₂ eggs showed strong eGFP expression, which we interpret as likely indicative of a homozygous mutant. All G₂ embryos with strong eGFP expression generated leg-like structures on the abdomen (Fig. 5B), suggesting that the function of the target gene was somewhat affected in the homozygous condition, unlike in the mosaic condition exhibited by G₀ embryos (Fig. 5B). Thus, even though the knock-in into an intron somewhat affected target gene expression, G₀ embryos with intronic insertions did not show abnormalities, unlike the embryos with exonic insertions. This increases the developmental success of intronic knock-in embryos and, accordingly, the possibility of establishing a transgenic line.

In the present study, we demonstrated that targeted knockout and knock-in by using the CRISPR/Cas9 system works efficiently in the cricket G. bimaculatus. We performed functional analysis of CRISPR/Cas9-induced mutations in the Hox genes Gb-Ubx and Gb-abd-A during embryogenesis and at post-embryonic stages. We found that the cleavage efficiency of the CRISPR/Cas9 system was much higher than that previously reported for artificial nucleases in this cricket (Watanabe et al., 2012, 2014). We demonstrated that gene knock-in via a homology-independent method is effective in this cricket, and successfully applied it to functional analysis of Hox genes by knocking a donor sequence into an exon of the target gene to disrupt the function of the target gene (KI/KO). In addition, we succeeded in trapping endogenous gene activity using this method and revealed a number of new expression domains that had not been previously observed with traditional methods (Barnett et al., 2019; Matsuoka et al., 2015; Zhang et al., 2005). This homology-independent method is technically simpler than the homology-dependent methods, as the donor plasmid does not need to be newly made for each target region.

Delivering the proper amount of genome editing constructs at the proper time is important for efficient outcomes with CRISPR/Cas9. For example, gene editing in embryos injected around 3 h AEL, at which stage energids are distributing throughout the yolk, was highly efficient, but both detectable mosaic phenotypes (of Gb-lac2 crispants) and the rate of NHEJ decreased with later injections at blastoderm stages (Fig. S2B,D,H). Thus, we suggest that optimization of delivery conditions can be achieved by using pigmentation genes as an index.

In our study, reproducibility and severity of phenotypes generated with CRISPR/Cas9 were greater than those obtained with RNAi (Fig. 2D,F). We speculate that the efficiency of RNAi-mediated knockdown may be influenced by when and at what levels the target gene is expressed. In the case of Gb-abdA and Gb-Ubx, these genes are expressed much later in development (embryonic stage 5, about 1.5 days after egg laying) than the stage at which we performed injections (within 3 h after egg laying) (Matsuoka et al., 2015; Donoughe and Extavour, 2016). Correspondingly, much higher concentrations of dsRNA proved necessary to produce even mild phenotypes (5-6 µg/µl; Fig. 2E), compared with those typically used for RNAi against most genes in this cricket (1-2 µg/µl; e.g. Donoughe et al., 2014). In contrast, indel mutations generated at earlier stages by genome editing techniques resulted in clearly detectable, severe phenotypes (Fig. S2). Several studies have demonstrated that genome editing techniques can sometimes be adequate for functional analysis of target genes in mosaic G₀ individuals (e.g. Daimon et al., 2015; Martin et al., 2016; Matsuoka and Monteiro, 2018). However, it is often difficult or impossible to unambiguously identify mutant cells in such mosaics. In this regard, establishment and maintenance of stable mutant lines as performed herein, allows less ambiguous phenotypic analysis in this cricket.

Although the CRISPR/Cas9 system is efficient, it offers little to no conditionality, which can complicate study of the many genes that act pleiotropically during development (Minelli, 2016). For example, in the case of Gb-abd-A, the gene acts to repress leg formation in the abdomen at embryonic stages (Fig. S6), while at adult stages, it regulates proper development of female genitalia (Fig. S7). Likely because of this latter phenotype, we were unable to obtain homozygous Gb-abdA^CRISPR animals. To overcome this problem, sophisticated genetic methods, such as balancer chromosomes (Miller et al., 2019), will need to be developed in the future. In this regard, RNAi offers more options for conditional control of gene function. By controlling the timing of injection of dsRNA, target gene activity can be knocked down at any desired developmental stage in G. bimaculatus (Dabour et al., 2011; Nakamura et al., 2008; Takahashi et al., 2009). Thus, although the CRISPR/Cas9 system is a powerful new tool for gene function analyses, RNAi remains a useful technique for this system.

Homology-independent knock-in methods will expand our ability to analyze the function of target genes in this hemimetabolous insect model. Here, we demonstrated one such application, the KI/KO method, which allows the isolation of mutants without PCR-based genotyping. When analyzing mutant phenotypes, affected individuals must typically be distinguished either by their morphology or by molecular methods to detect changes in target gene product levels or functions. In the case of the Gb-Ubx mutant, we would have needed to distinguish subtle differences in the T3 and A1 embryonic segments (Fig. 2C), requiring destructive sampling. Moreover, antibodies against target genes may not be routinely available in many cases. The KI/KO method allows us to distinguish mutant individuals based on marker gene expression. Even heterozygous and homozygous mutants can sometimes be distinguished based on the intensity of marker gene expression. A similar strategy was employed in mosquitos via HDR (McMeniman et al., 2014). In the present study, we could easily identify the eGFP expression resulting from the KI/KO event because it matched the previously characterized expression pattern for Gb-Ubx (Barnett et al., 2019; Matsuoka et al., 2015; Zhang et al., 2005). However, for target genes with previously uncharacterized expression domains, analysis may be more complex.

The promoter used in all expression cassettes herein is the same one used in a previous study to drive ubiquitous constitutive expression (Nakamura et al., 2010). Nevertheless, all knocked-in lines showed an eGFP expression pattern that was spatially and temporally restricted like that of the target gene. We speculate that the promoter in the expression cassette acts as a minimal promoter, and that the observed eGFP expression results from trapping endogenous enhancer activity. The eGFP expression was not caused by fusion to the endogenous gene product, as both the line containing an inverted orientation of the donor sequence, and the knock-in line targeting an intronic region, showed similar eGFP expression patterns. To enhance the usefulness of this method, identification and use of a ubiquitous and strong promoter could in principle drive exogenous marker gene expression in the whole embryo without being subject to positional effects.

A remarkable feature of the homology-independent knock-in method is the length of sequence that can be integrated. In case of knock-in through HDR, a few kb of sequence can be integrated into the genome in arthropods (Gilles et al., 2015; McMeniman et al., 2014). In this study, through NHEJ, at least 6 kb of plasmid sequence was integrated into the genome. Furthermore, in some cases, three copies of plasmid sequence were integrated in tandem into the genome (Fig. S4). In this case, we speculate that first the donor plasmids were digested and combined via NHEJ, and then the combined fragment was knocked into the genome via NHEJ, suggesting that homology-independent knock-in might be able to integrate several tens of kb of sequence into the genome. For example, one study showed that a 200 kb BAC vector could be integrated into a rodent genome through a similar strategy (Yoshimi et al., 2016). This method might therefore be used for direct functional comparison of genomic regions by exchanging homologous regions between related species of interest.

The efficiency of knock-in through NHEJ is high, but to improve its feasibility as a technique for functional genetic analysis in this cricket, future studies may be able to further enhance efficiency by optimizing at least one of two parameters. The first parameter is enhancing the expression level of a fluorescent cassette. Empirically, G₀ crickets showing mosaic eGFP expression tend to transmit their knocked-in transgene to their offspring. To increase the efficiency of obtaining knock-in lines, future efforts should therefore focus on increasing the number of mosaic marker gene expression cassettes in G₀ embryos. We also observed that some expression cassettes seemed to show higher expression levels than others and were therefore easier to detect in mosaic G₀s. Inclusion of inducible expression elements, such as a heat shock promoter or a modified Gal4/UAS system, might help to enhance the activity of the expression cassette. However, enhancement of fluorescent cassette expression may also cause artificial or ectopic expression, as we observed in the ovaries of Gb-Ubx^KI-exon (Fig. S5G″). The second parameter is introducing insulator sequences into the donor cassette. Positional effects might in principle prevent the full potential activity of the expression cassette. In this study, our vector plasmid contained insulators of the sea urchin Hemicentrotus pulcherrimus arylsulfatase gene (Takagi et al., 2011) on either side of the expression cassette (see Materials and Methods), but we nonetheless detected eGFP expression in a pattern matching that of the target gene. To achieve more effective insulation, future studies might evaluate several different combinations of insulator orientations, which can affect insulator activity (Tchurikov et al., 2009). Alternatively, other insulators, such as that of the gypsy retrotransposon (Modolell et al., 1983), might be additional options for future optimization (Carballar-Lejarazú et al., 2013).

In conclusion, we provide evidence that the CRISPR-Cas9 system works well for both knock-in and knockout in the cricket G. bimaculatus. In depth analysis of CRISPR-Cas9-induced mutations revealed optimized injection timing. In addition, we succeeded in the targeted functional knock-in of exogenous sequences into the genome through NHEJ, resulting in expression-tag reporter lines for target genes.

All adult and juvenile Gryllus bimaculatus were reared in plastic cages at 26-30°C and 50% humidity under a 10 h light, 14 h dark photoperiod. They were fed on artificial fish food (Tetra) or Purina cat food (178046). For microinjections, fertilized eggs were collected on a wet kitchen towel in a plastic dish and incubated at 28°C, as previously described (Barry et al., 2019; Watanabe et al., 2017).

sgRNA target sequences were designed and their off-target sites were predicted with the CasOT program (Xiao et al., 2014). From the suggested candidates, we selected target sequences having high GC content (around 70%), beginning with guanidine for efficient in vitro transcription by the T7 promoter, and with off-target sites containing at least three mismatches as per Ren and colleagues (Ren et al., 2014). Two synthetic oligonucleotides (5′-ATAG-N₁₉-3′ and 5′-AAA-N₂₀) were annealed and inserted into the BsaI site of the modified pDR274 vector (Addgene 42250) to expand its utility (the GGN₁₈NGG sequence was present in the original pDR274, whereas in the modified vector, a GN₁₉NGG sequence was used). We confirmed insertion by Sanger sequence analysis.

For sgRNA synthesis, the template for in vitro transcription was digested from the vectors generated as described above with DraI. For Cas9 mRNA synthesis, the template for in vitro transcription was digested from pMLM3613 (Addgene 42251) with PmeI. Both sgRNA and Cas9 mRNA were in vitro transcribed using mMESSAGE mMACHINE T7 Kit (Life Technologies, AM1344), and purified by ethanol precipitation. For the Cas9 mRNA, we attached a poly-A tail by using a poly-A tailing Kit (Life Technologies, AM1350). The concentration of synthesized RNAs was estimated by NanoDrop and gel electrophoresis.

The eGFPbait-2A-RFP donor plasmid was generated in a pUC57 vector by commercial artificial composition (GeneScript). The DsRedbait-G′act-eGFP donor plasmid was generated based on the eGFPbait-2A-RFP donor plasmid. First, 2A-RFP was digested using BglII and NotI. Gb-act-eGFP was also digested from a pXL-BacII- G′act-eGFP vector and ligated to generate the eGFPbait-G′act-eGFP vector. Then, we digested this eGFPbait vector using BglII and SacII. We amplified DsRedbait with primers (5′ to 3′) DsRed_fwd: GCTCAGATCTCTTGGAGCCGTACTGGAAC and DsRed_rev: GTACGAGCTCCATCACCGAGTTCATGCG. The amplicon was ligated to generate the DsRedbait-G′act-eGFP donor plasmid. The DsRedbait-2×Ars_rev-G′act-eGFP-2×Ars_fwd donor plasmid was generated based on the DsRedbait-G′act-eGFP donor plasmid. The Ars insulator sequence ArsInsC from H. pulcherrimus (Takagi et al., 2011) was amplified from an ArsInsC-containing plasmid (a kind gift from Associate Professor Naoaki Sakamoto, Hiroshima University, Japan) and integrated on either side of the expression cassette in the donor plasmid.

Cas9 mRNA, sgRNA and donor vectors were injected into 2-5 h AEL cricket eggs. Cricket eggs were aligned in a groove 0.7 mm deep and 0.7 mm wide made with 2% agarose in 1×phosphate-buffered saline (PBS) using a custom mold as previously described (Barry et al., 2019; Watanabe et al., 2017) and filled with 1×PBS. Needles for injection were made by pulling glass capillaries with filament (Narishige catalogue, GD-1) with a pipette puller (Sutter Instrument catalogue, P-1000IVF), using the following pulling program: (1) ×3 Heat; 858, Pull; 0, Velocity; 15, Time; 250, Pressure; 500; and (2) ×1 Heat; 858, Pull; 80, Velocity; 15, Time; 200. To minimize the invasiveness of the injection, the tips of the pulled needles were sharpened and ground to a 20° angle by using a Micro Grinder (Narishige, EG-400). Approximately 5 nl of solution was injected into eggs with a Micro Injector (Narishige, IM300). After injection, eggs were moved to a fresh Petri dish and submerged in fresh 1×PBS containing 50 U/ml penicillin and 50 μg/ml streptomycin (15070-063, Thermo Fisher), and incubated at 28°C. During the incubation period, the 1×PBS with penicillin and streptomycin was replaced every day. We observed fluorescent protein expression at the stages when the target gene was known to be expressed. Genomic DNA was extracted from 7 days AEL eggs and adult T3 legs, and used for insertion mapping and sequence analyses. After 2 days of incubation, injected cricket eggs were moved to wet filter paper in a fresh Petri dish for hatching.

After Cas9 nuclease digests a target sequence, the disrupted sequence is repaired by either the NHEJ or the HDR cell machinery. To confirm a KO mutation, we searched for errors repaired by the NHEJ pathway, which sometimes introduces or deletes nucleotides at the digested site during the repair process. As disruption or repair are unlikely to take place identically in all cells of an injected G₀ embryo, G₀ animals are expected to contain heterogeneous sequences at the CRISPR targeted site, and thus to be heterozygous for a putative Cas9-induced indel. To confirm the activity of the sgRNAs, the Surveyor Mutation Detection Kit (Transgenomic) was used. This assay relies on a ‘surveyor’ nuclease that can recognize and digest a heteroduplex DNA structure. First, genomic DNA was extracted from whole eggs or part of the T3 leg by a phenol chloroform method, as previously described (Barry et al., 2019; Watanabe et al., 2017). Subsequently, ∼200 bp of the targeted region was amplified by PCR from genomic DNA (Table S2). PCR conditions were optimized to reduce non-specific amplification or smearing. To create the putative heterogeneous DNA structure for the nuclease assay, the PCR product was heated to 98°C for 5 min, and then re-annealed by gradually cooling down to 30°C. Half of the PCR product was digested with the Surveyor nuclease, and the other half was used as a negative control and incubated without the nuclease. Digestion was confirmed by agarose gel electrophoresis. For sgRNAs that yielded indels in the target sequence, digest of the PCR product by the Surveyor nuclease is expected to produce split fragments around the CRISPR targeted site relative to the negative control; the latter should not be digested by the nuclease, and thus should remain intact and run at the same size as the original amplicon. Positive PCR products were extracted from the gel, purified with the QIAquick Gel Extraction Kit (Qiagen, 28506) and sub-cloned into the pGEM-Teasy vector (Promega, A1360) using TA-cloning. The vectors were used for Sanger sequence analysis.

After a 1 h egg collection, eggs were incubated for the desired length of time at 28°C. We co-injected 0.5 μg/μl sgRNA and 1 μg/μl Cas9 mRNA into fertilized cricket eggs after each of these incubation periods. Five days after injection, genomic DNA was extracted individually from three eggs from each of the four tested injection times; the latter analysis was performed in biological triplicate. Amplicon sequence analysis was performed by using MiSeq (Illumina), and the preparation of DNA libraries and sequencing reactions were performed according to the manufacturer's instructions. We read ∼10,000 reads for on-target regions and ∼50,000 reads for off-target regions. We chose off-target sites that were most similar to the target site relative to other sequences in the genome. As we mentioned above about the criteria for selecting target site, the off-target site has three mismatch sequences compared with the target site. The assembly of output paired end reads was performed by using CLC Genomic Workbench (CLC Bio, QIAGEN Digital Insights). The relative proportions of reads containing indels and substitutions in the individual eggs were calculated with the online-tool CRISPResso (Pinello et al., 2016). We used the Integrative Genomic Viewer (Broad Institute) for investigation of the distribution of indels and substitutions (Thorvaldsdóttir et al., 2013).

Genomic DNA was extracted from eGFP-positive eggs of each line. Owing to the specifics of this knock-in method, two types of insertion of vector fragment (sense and antisense orientations) would be expected to occur; we therefore performed PCR using primers designed against either side of the putative junction. PCR was performed using target region-specific (upstream or downstream of sgRNA recognition site) and donor vector-specific primers (sequence within eGFP for forward integration and M13Fw for reverse integration). Primer sequences are listed in Table S2. Positive PCR products were extracted from the gel, purified by using the QIAquick Gel Extraction Kit (Qiagen catalogue #28506), and sub-cloned into the pGEM-Teasy vector (Promega catalogue #A1360) using TA-cloning. The vectors were used for Sanger sequence analysis.

Embryos were dissected in 1×PBS and fixed with 4% paraformaldehyde in 1×PBS+0.1% Tween (PBT) for 1 h at 4°C The fixed embryos were dehydrated stepwise in 25%, 50%, 75% and 100% methanol in 1×PBT with 5 min per wash. The dehydrated embryos were stored in 100% methanol at −30°C. Whole-mount in situ hybridization with digoxigenin (DIG)-labeled antisense RNA probes was performed as previously described (Niwa et al., 2000; Zhang et al., 2005). Immunohistochemistry was performed as follows. Fixed embryos were rehydrated stepwise in 75%, 50% and 25% solutions of methanol/PBT and finally in 100% PBT for 5 min in each solution. After blocking with 1% bovine serum albumin (BSA) (Thermo Fisher) in PBT for 1 h at room temperature, embryos were incubated with an anti-UbdA antibody FP6.87 (Kelsh et al., 1994) (Developmental Studies Hybridoma Bank) diluted 1:200 in 1% BSA/PBT overnight at 4°C. After washing with PBT three times, embryos were incubated in 1% BSA/PBT for 1 h at room temperature, and then incubated with Alexa Fluor 488-conjugated goat anti-mouse IgG(H+L) (Invitrogen, A32723) diluted 1:400 in 1% BSA/PBT for 1 h at 4°C. After washing the embryos with PBT once for 10-60 min, embryos were counterstained with DAPI (Sigma, 10236276001) stock solution 1 mg/ml diluted 1:1000 in PBT for 10 min, and then washed with PBT twice for 10-60 min per wash. PBT was then substituted with 25% and 50% glycerol/PBT to clear embryos for microscopy.

To estimate the number of plasmid fragments integrated into the genome via NHEJ events, we performed quantitative RT-PCR using genomic DNA from individual 5-day-old embryos of wild type, Gb-Ubx^KI-exon and Gb-abd-A^KI-exon lines, and compared relative quantity values of the inserted eGFP gene with an endogenous gene, Gb-otd, which is present in a single copy in the G. bimaculatus genome (Ylla et al., 2021). Genomic DNA was extracted from embryos using Cica geneus Total DNA Prep kit (for Tissue) (Kanto Chemical), according to the manufacturer's protocol. Real-time quantitative PCR was performed using the power SYBR Green PCR Master Kit (Applied Biosystems, 4368577) and an ABI 7900 Real Time PCR System (Applied Biosystems), as described previously (Nakamura et al., 2008). Primer sequences are listed in Table S2.

To examine whether the eGFP expression observed in gonads of Gb-Ubx^KI-exon and Gb-abd-A^KI-exon reflects the endogenous expression of Gb-Ubx or Gb-abd-A in those tissues, we performed quantitative RT-PCR using total cDNA from tissues with or without eGFP expression. We dissected out gonads and brain in 1×PBS, and then transferred the dissected tissues to RNA later (Thermo Fisher). Total mRNA was extracted with RNeasy MinElute Cleanup Kit (Qiagen), according to the manufacturer's protocol. Equal amounts of total RNA were used for reverse transcription reaction with SuperScript III Reverse Transcriptase (ThermoFisher). Quantitative RT-PCR was performed with THUNDERBIRD SYBR qPCR Mix (TOYOBO) and LightCycler 96 (Roche). The expression levels of eGFP, Gb-Ubx and Gb-abd-A were normalized to the level of Gb-beta-tubulin. Primer sequences used were as previously described (Barnett et al., 2019).

We thank N. Sakamoto for the Ars plasmid, the members of the Extavour and Mito labs for discussion, and the Harvard Faculty of Arts and Sciences Bauer Core Facility for Illumina sequencing.