Recent progress in genome engineering techniques in the silkworm, Bombyx mori



Rapid advances in genome engineering tools, such as zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), and the clustered regularly interspaced palindromic repeats/CRISPR-associated (CRISPR/Cas) system, have enabled efficient gene knockout experiments in a wide variety of organisms. Here, we review the recent progress in targeted gene disruption techniques in the silkworm, Bombyx mori. Although efficiency of targeted mutagenesis was very low in an early experiment using ZFNs, recent studies have shown that TALENs can induce highly efficient mutagenesis of desired target genes in Bombyx. Notably, mutation frequencies induced by TALENs can reach more than 50% of G0 gametes. Thus, TALENs can now be used as a standard tool for gene targeting studies, even when mutant phenotypes are unknown. We also propose guidelines for experimental design and strategy for knockout experiments in Bombyx. Genome editing technologies will greatly increase the usefulness of Bombyx as a model for lepidopteran insects, the major agricultural pests, and lead to sophisticated breeding of Bombyx for use in sericulture and biotechnology.


The domesticated silkworm, Bombyx mori, has been reared by humans for thousands of years and used as a model insect in studies of insect physiology and genetics (Tazima 1978; Goldsmith et al. 2005). Recent successes in germline transformation and whole-genome sequencing of Bombyx represent especially significant milestones in Bombyx genetics (Tamura et al. 2000; The International Silkworm Genome Consortium 2008). Tamura et al. (2000) are the first to establish a highly efficient system for germline transformation of Bombyx using the piggyBac transposon. Since then, various genetic tools have been developed for functional characterization of genes and for heterologous expression of recombinant proteins in Bombyx. These include a GAL4/UAS system (Imamura et al. 2003; Kobayashi et al. 2011), an enhancer trap system (Uchino et al. 2008), a Cre-mediated recombination system (Duan et al. 2013), and an in vivo lipofection/electroporation system (Ando & Fujiwara 2012; Kamimura et al. 2012). These tools are useful for gain-of-function studies such as overexpression or ectopic expression of genes of interest; however, as Bombyx appears to lack efficient systemic RNA interference (RNAi) machinery (Terenius et al. 2011; Kobayashi et al. 2012), loss-of-function studies using RNAi are generally difficult to achieve, with the exception of embryonic RNAi (Quan et al. 2002; Yamaguchi et al. 2011; Nakao 2012). Several successful larval RNAi experiments adopted special techniques to express double-stranded RNAs, including the use of viral vectors (Uhlirova et al. 2003) and introduction of transgenes (Subbaiah et al. 2013) to express RNA hairpins. Recently, Yamaguchi et al. (2011) showed that injection of short-interfering RNAs induces strong gene silencing effects in Bombyx embryos. Nevertheless, gene knockdown by RNAi has a serious limitation in that gene silencing is transient and incomplete. Thus, establishment of a practical gene targeting system has long been awaited in Bombyx genetics. Although there is only one example of targeted gene disruption in Bombyx, in which an endogenous gene was disrupted by homologous recombination (HR) using a baculovirus vector (Yamao et al. 1999), the efficiency of gene targeting was very low and this method cannot be used as a standard technique.

Recent advances in genome engineering tools, such as zinc finger nucleases (ZFNs) (Urnov et al. 2010), transcription activator-like (TAL) effector nucleases (TALENs) (Joung & Sander 2013), and the clustered regularly interspaced palindromic repeats/CRISPR-associated (CRISPR/Cas) system (Cho et al. 2013; Hwang et al. 2013a,b; Jiang et al. 2013), have enabled efficient gene knockout experiments in a wide variety of model and non-model organisms. Takasu et al. (2010) reported the first example of gene targeting in Bombyx using ZFNs. After that, Sajwan et al. (2013) and Ma et al. (2012) showed that TALENs are able to disrupt target genes more efficiently than ZFNs and thus are more suitable for gene knockout studies in Bombyx. Most recently, Takasu et al. (2013) developed a new TALEN backbone vector optimized for the use in Bombyx. Intriguingly, TALENs assembled in this vector yielded very high (>50%) germline mutation rates.

Here, we review the recent progress in gene targeting techniques in Bombyx and suggest a guideline for successful knockout experiments. Targeted gene disruption using engineered nucleases has now become a standard and easy technique in Bombyx.

Zinc finger nucleases

Zinc finger nucleases are the first-generation tool for genome engineering (Urnov et al. 2010). They are artificial enzymes consisting of a zinc finger DNA binding domain and the endonuclease domain from the FokI restriction enzyme (Kim et al. 1996). The zinc finger domain can be custom-designed to bind to a specific site of the genome, and the FokI domain causes a double-stranded break (DSB) at the target site. This frequently induces a short insertion/deletion mutation resulting from non-homologous end joining (NHEJ), a repair process for DSBs. The first successful gene targeting using ZFNs in insects was reported in the fruit fly Drosophila melanogaster (Bibikova et al. 2002). Although genes encoding ZFN proteins were expressed as transgenes in the original protocol (Bibikova et al. 2002), subsequent studies have shown that direct injection of mRNAs for ZFNs to embryos works very well in Drosophila (Beumer et al. 2006, 2008) and in other animals (Doyon et al. 2008; Geurts et al. 2009; Xiao et al. 2013).

Takasu et al. (2010) reported the first example of gene targeting in a non-Drosophila insect. The target genes they chose were the two Bombyx genes, BmBLOS2 (Fig. 1) and Bmwh3, both of which are essential for the formation of urate granules in the larval epidermis (Quan et al. 2002; Komoto et al. 2009; Fujii et al. 2010) that confer an opaque white color to larval skin. Loss of the urate granules in the epidermis results in the visible “oily” phenotype with translucent skin. Importantly, as BmBLOS2 and Bmwh3 mutants display the oily phenotype in a cell-autonomous manner, disruption of the genes is visible as oily mosaics of the skin in injected G0 animals (Fig. 1). Takasu et al. (2010) tested two pairs of ZFNs for BmBLOS2 and one pair for Bmwh3. They found that one ZFN pair for BmBLOS2 (BL-1, Table 1) yielded a high proportion of oily mosaic animals in G0: 72% of fifth-instar larvae showed a mosaic pattern of translucent and normal skin. As BmBLOS2 gene locates on the Z chromosome (female W/Z and male Z/Z in Bombyx; Fujii et al. 2010), it was expected that the proportion of mosaic G0 animals would be higher in female than in male G0 larvae; however, the proportion was comparable between the sexes (70% vs. 56%), indicating that the BL-1 ZFN pair induced both mono- and bi-allelic mutations with similar efficiencies. Germline transmission of mutated alleles was also examined. Among 16 350 G1 larvae from the cross of G0 males with normal (uninjected) females, 46 larvae (0.28%) displayed the oily phenotype and all of the oily animals were females, as expected. Sequencing analysis showed that the most frequent changes at the target site were short deletions and insertions. On the other hand, BL-2 did not yield oily mosaics or germline mutations, and BW-1 (targeting Bmwh3 gene) yielded a moderate number of G0 oily mosaics (22%) but germline mutations were not found in approximately 4000 G1 larvae screened.

Figure 1.

Targeted disruption of BmBLOS2 gene. (A) Genomic structure of BmBLOS2 gene. The red boxes indicate exons of BmBLOS2. In distinct oily (od) mutants, exons 1–3 are deleted from the genome, which results in the translucent skin called the “oily” phenotype (Fujii et al. 2010). (B) A representative oily mosaic larva (G0) injected with a transcription activator-like effector nuclease (TALEN) pair targeting BmBLOS2 gene. In a severe case, about half the larva's skin exhibited the oily phenotype. (C) A magnified ventral view of the abdomen of an oily mosaic G0 larva. (D and E) G1 oily larvae at the third (C) and fifth (D) larval instar. Red arrows indicate oily mutants induced by TALENs.

Table 1. Efficiency of zinc finger nucleases (ZFNs) in BmBLOS2 mutagenesis
ZFNStrainEggs injected% Hatched% Somatic mosaics% Yielders% Germline mutantsReference
BL-1 pnd w-1 48051725–90.28Takasu et al. (2010)
BL-2 pnd w-1 4804000Takasu et al. (2010)

These results suggest that although targeted gene disruption can be achieved in Bombyx using ZFNs, mutation rates in Bombyx are very low compared to those reported in Drosophila. Thus, ZFN-mediated knockout experiments in Bombyx still require great effort, especially when targeting genes with unknown phenotypes.


Transcription activator-like effector nucleases are the second-generation tools for genome editing (Joung & Sander 2013). Similar to ZFNs, TALENs consist of a DNA binding domain, which originates from bacterial TAL effectors, and a FokI nuclease domain, which induces DSBs at the target site (Christian et al. 2010; Miller et al. 2011). The highly modular structure of the DNA binding domain of TAL effectors and its simple DNA binding code (Boch et al. 2009; Moscou & Bogdanove 2009) enables researchers to design and assemble custom TALENs with ease. TALENs have been also used for targeted gene disruption in many organisms including animals, plants, and yeast (Joung & Sander 2013; Xiao et al. 2013), and successful gene knockout experiments have been reported in Drosophila (Liu et al. 2012), Bombyx (Takasu et al. 2013; Ma et al. 2012; Sajwan et al. 2013), and crickets (Watanabe et al. 2012).

In Bombyx, two research groups reported the targeted gene disruption using TALENs. Both groups disrupted the same target gene, BmBLOS2 (Fig. 1), but used different architectures (Table 2): the TAL effector domain of TALENs used by Sajwan et al. (2013) had 287 and 232 amino acid (aa) residues at the N- and C-terminal sides of the DNA binding domain, respectively, while those used by Ma et al. (2012) had truncations at both sides. As also reported for other animals (Miller et al. 2011), truncated forms exhibited higher activities in Bombyx. Whereas germline mutation rates of Sajwan et al. (2013) were comparable to those using ZFNs (0.05–0.7% vs. 0.28%), those of Ma et al. (2012) ranged from 0.4% to 61%. Importantly, Ma et al. (2012) also demonstrated that co-injection of two pairs of TALENs induced large (~800 bp) heritable deletions in the genomic region flanked by target sites of each TALEN pair. This suggests that more sophisticated forms of genome editing can be conducted in Bombyx, such as introduction of large deletions in gene clusters and regulatory regions, and induction of large genomic inversions.

Table 2. Efficiency of transcription activator-like effector nucleases (TALENs) in BmBLOS2 mutagenesis
TALEN pairTALEN architectureRecognition site (bp)Spacer (bp)Target site in BmBLOS2StrainEggs injected% hatched% somatic mosaics (G0)% yielders% germline mutantsDose of mRNA (ng/μL)Reference
N-terminus (aa)C-terminus (aa)LeftRight
  1. Designed and constructed by Cellectis bioresearch. Germline mutants ranging from 1 to 61% per brood.

BLT-1287232282421exon 3 pnd w-1 480371510.70.7400Sajwan et al. (2013)
BLT-2287232222419exon 3 pnd w-1 47821106.30.5400Sajwan et al. (2013)
BLT-3287232212020exon 3 pnd w-1 4794960.020.05400Sajwan et al. (2013)
B224063181816exon 2Nistari968234631NA700Ma et al. (2012)
B324063181816exon 3Nistari521192714NA700Ma et al. (2012)
B324063181816exon 3Nistari331212311NA400Ma et al. (2012)
B324063181816exon 3Nistari299121022NA200Ma et al. (2012)
BLTS-213663222419exon 3 pnd w-1 19252857931000Takasu et al. (2013)
BLTS-413663211918exon 3 pnd w-1 191328210010.41000Takasu et al. (2013)
BLTS-513663211915exon 3 pnd w-1 191389810010.71000Takasu et al. (2013)
BLTS-613663151915exon 3 pnd w-1 19151100100501000Takasu et al. (2013)
BLTS-713663151912exon 3 pnd w-1 19242050.21000Takasu et al. (2013)
BLTC13645161615exon 2 pnd w-1 4325588394.4400Takasu et al. (2013)

To further increase the efficiency of TALENs in Bombyx, Takasu et al. (2013) recently developed a new TALEN backbone vector optimized for use in Bombyx. The new vector, named pBlueTAL (Fig. 2), is based on the truncated TALEN architecture of Miller et al. (2011), and codon usages for the N- and C-terminus of the TAL effector and FokI domains are optimized for insects. In addition, the 5′-UTR of the TAL effector domain is replaced by that of Bombyx cytoplasmic actin 3 gene and a Kozak sequence is introduced to the site surrounding the start codon. Esp3I sites are also included to make the vector compatible with the Golden Gate Assembly Kit by Cermak et al. (2011). Takasu et al. (2013) examined the efficiency of the new vector by targeting the BmBLOS2 gene by changing spacer lengths flanked by left and right TALEN binding sites. As shown in Table 2, efficiency of these TALENs was greatly increased compared to previous studies. The proportion of oily mosaics reached as high as 82–100% in G0 larvae hatched from eggs injected with TALENs having 15-, 18-, and 19-bp spacers, and most or all (% yielders = 79–100%) of these G0 larvae yielded at least one oily mutant in the next generation. The TALEN architecture of Miller et al. (2011) (63 aa at the C-terminus of TAL effector domains) was shown to have an optimal spacer length of 14–16 bp. In good agreement with this, TALEN pairs with 15-bp spacers (BLTS-5 and BLTS-6 in Table 2) yielded the highest germline mutation rates (10.7–50%). On the other hand, TALEN pairs with suboptimal spacers (12-bp and 19-bp) exhibited much lower mutagenic activities (0.2% and 3% germline mutants, respectively), although these efficiencies were still higher than those observed in ZFNs (Takasu et al. 2010).

Figure 2.

Schematic representation of pBlueTAL vector. The pBlueTAL is a transcription activator-like effector nuclease (TALEN) backbone vector optimized for use in Bombyx (see main text). Repeat-variable di-residue (RVD) modules can be introduced to the vector by Golden Gate Assembly as described in Cermak et al. (2011). This vector is also compatible with vectors developed by Sakuma et al. (2013) for 6-module assembly. After assembly, the vector can be linearized at the XbaI site for in vitro transcription from the T7 promoter.

Takasu et al. (2013) also tested the new TALEN backbone vector by targeting another autosomal gene, red egg (Bm-re), which is responsible for the color of the egg serosa (Osanai-Futahashi et al. 2012). When a single TALEN pair with a 16-bp spacer was tested, most (72%) of the injected eggs displayed mosaic colors of red (mutant) and dark brown (normal). Notably, all the injected G0 moths carried germline mutations, and the mutation rate in G0 gametes was estimated to be as high as 77.3%.

These studies clearly demonstrate that TALENs can disrupt genes with surprisingly high efficiencies in Bombyx. Given the high efficiency of the newly developed TALEN backbone vector, it would not be difficult to disrupt genes with unknown phenotypes, where screening of mutations should rely solely on molecular diagnostics such as CEL-I assay (Oleykowski et al. 1998; Kulinski et al. 2000).

CRISPR/Cas system

Clustered regularly interspaced palindromic repeats/Cas is a new system, most recently adapted to induce targeted mutagenesis. In this system, custom guide RNAs are created in cultured cells or embryos to direct site-specific DNA cleavage by the Cas9 endonuclease (Cho et al. 2013; Hwang et al. 2013a,b; Jiang et al. 2013). This system is based on a bacterial immune system in which the Cas9 is guided to its specific target DNA by CRISPR RNAs (crRNAs) and the trans-activating CRISPR RNA (tracrRNA) (Brouns et al. 2008; Jinek et al. 2012). Several procedures have been developed for using the CRISPR/Cas system in vivo. A two-component system has proven to be effective in Drosophila, in which the crRNA and tracrRNA are fused into a single RNA called synthetic guide RNA (sgRNA), and the Cas9 is expressed from an expression vector (Gratz et al. 2013) or is provided as mRNAs synthesized in vitro (Bassett et al. 2013; Yu et al. 2013). Strong advantages of this system are that plasmids for expressing custom sgRNAs can be easily and rapidly assembled by ligating pairs of short, annealed oligonucleotides into sgRNA vectors, and Cas9 can be expressed from a pre-made vector (Hwang et al. 2013b). These features enable high-throughput production of sgRNAs.

Our pilot experiments suggest that the CRISPR/Cas system can also be applicable to Bombyx (Fig. 3). We used a two-component system of Hwang et al. (2013b), in which custom sgRNA and Cas9 mRNA are transcribed in vitro from T7 vectors. We examined mutagenic efficiencies by targeting the two sites of BmBLOS2 gene. As in the experiments with ZFNs and TALENs, we injected each of two sgRNAs (od1 and od2; 50 ng/μL) into preblastoderm embryos of Bombyx strain pnd w-1 together with Cas9 mRNA (300 ng/μL). As shown in Table 3, od1 sgRNA yielded a small number of mosaic G0 larvae (8.8%, 3/34), but germline mutants were not found in G1 individuals (0/10,526). On the other hand, od2 sgRNA yielded a relatively high proportion of mosaic G1 larvae (58%, 7/12). One of the four fertile G0 male moths yielded oily G1 larvae (25% yielders) and the overall proportion of germline mutants from four G1 broods was 0.46% (6/1292). This germline mutation rate was comparable to that of ZFNs (0.28%, Takasu et al. 2010). We performed the same experiment using another strain, N4, which is also a non-diapausing strain. The od1 sgRNA did not yield mosaics or germline mutants. Although od2 sgRNA yielded a relatively high number of mosaics (57%, 8/14), germline mutants were not found in 2375 G1 larvae. Notably, all of the mosaic G0 larvae were females in these experiments. Such sex-biased appearance of oily G0 mosaics was not observed in previous experiments using ZFNs or TALENs. This indicates that bi-allelic mutations rarely occurred in od1 or od2 sgRNA-injected embryos, probably due to the weak activity of sgRNA and/or Cas9 used in this study.

Figure 3.

Gene targeting of BmBLOS2 using Clustered regularly interspaced palindromic repeats (CRISPR)/Cas9. (A) Target sites of sgRNAs. The od1 and od2 sgRNAs target exons 3 and 4 of BmBLOS2 gene (red letters). Note that od2 sgRNA targets the reverse strand. Proto-spacer adjacent motif (PAM) is indicated by black arrows. Cleavage sites are shown by arrowheads. Intron (small letters)-exon (capital letters) junctions are shown by dashed lines. The pDR274 vector (Hwang et al. 2013b) was used for od1 sgRNA synthesis. For od2, a single nucleotide substitution was introduced to pDR274 (G>A, 1 bp downstream of T7 transcription start site) to enable targeting the sequence of GA-N18 -NGG. (B) Photo of an oily mosaic G0 larva induced by od2 sgRNA. Arrows indicate areas of translucent skin. (C) Oily mutant G1 larvae (arrows) induced by od2 sgRNA. (D) Sequences of induced mutations in the BmBLOS2 gene. The target site of od2 sgRNA is shown by red letters. The two mutant alleles were identified from six oily larvae, both of which had short insertions. In one mutant allele (od2_#1, middle line), a 12-bp region within the target site was duplicated (grey underscore). Numbers in parentheses on the right indicate the number of individuals carrying the mutant allele.

Table 3. Efficiency of clustered regularly interspaced palindromic repeats (CRISPR)/Cas9 in BmBLOS2 mutagenesis
sgRNAExonStrainEggs injected% hatchedNo. fifth instar G0 larvaeNo. somatic G0 mosaics (%)% yielder males (No. G0 males yielding oily mutants/total G0 males)% germline mutantsa (No. oily G1 mutants/total G1 larvae hatched)
  1. a

    Only progeny from the cross of pnd w-1 females with G0 males were screened.

od13 pnd w-1 1926641340 (0)3 (8.8)0 (0/32)0 (0/10 526)
3N4487117100 (0)0 (0)0 (0/16)0 (0/3465)
od24 pnd w-1 19218570 (0)7 (100)25 (1/4)0.46 (6/1292)
4N4487115140 (0)8 (57)0 (0/12)0 (0/2375)

Our results suggest that the CRISPR/Cas system can be used for gene targeting in Bombyx, although it requires further improvement for practical use. For example, optimization of the Cas9 expression vector for Bombyx would be promising, as that used in this study (pMLM3613) is codon-optimized for zebrafish (Hwang et al. 2013b). In addition, optimal concentrations of sgRNAs and Cas9 mRNAs should be determined.

A guideline for gene knockout experiments in Bombyx

Although ZFNs, TALENs, and CRISPR/Cas work in Bombyx, TALENs would be a promising first choice given their high activities. Indeed, we have rarely failed in generating knockouts using TALENs. Construction of TALEN vectors is more complex and requires more time than that of sgRNA vectors (5 vs. 2 days by standard procedures) (Cermak et al. 2011; Hwang et al. 2013b); however, this disadvantage can be overcome by the high efficiency of TALENs as long as one does not attempt to disrupt many genes. From our experience in targeting more than 20 genes, including phenotypic marker genes and genes with unknown phenotypes, we here suggest a guideline for gene knockout experiments in Bombyx using TALENs (Fig. 4). This guideline could also be generally adapted to other insect systems in which microinjection into early embryos is possible.

Figure 4.

A mating scheme and screening strategy for gene targeting in Bombyx. A proposed guideline for successful gene targeting experiments in Bombyx. See main text for details.

Target design

The number of repeat-variable di-residue (RVD) modules, each of which recognizes one nucleotide of DNA sequence, and the length of spacer should be determined based on each TALEN architecture. For example, when assembled in the pBlueTAL vector (having 63 aa residues at the C-terminus of the TAL effector domain), TALENs will work best with 15–20 RVD repeats and 14- to 16-bp spacers (Takasu et al. 2013; Miller et al. 2011). Public tools for finding target sites and designing RVD repeats are available (e.g., TALEN targeter, (Doyle et al. 2012). Several guidelines have been proposed for the design of efficient TALENs, such as requirement of a specific nucleotide at a specific site, and percent composition of TAL binding sites (Reyon et al. 2012; Streubel et al. 2012). We usually follow these guidelines as long as the software identifies a binding site at the desired location. Designed binding sites are then subjected to BLAST search against the whole genome sequence of Bombyx to minimize the possibility of targeting multiple loci. As target sites of TALENs can be flexibly designed by changing the number of RVD modules and the spacer size within the aforementioned ranges, most loci can be targeted.

Construction of TALENs

Among several methods developed for construction of TALENs, golden gate assembly (GGA) described by Cermak et al. (2011) would be best on the single-lab scale. The original protocol of Cermak et al. (2011) calls for simultaneous assembly of 10 RVD modules; a newly developed method for 6-module assembly by Sakuma et al. (2013) will greatly improve the efficiency of construction. Vector kits for both the original GGA and the 6-module assembly are available from Addgene; they are compatible with the pBlueTAL vector, as Esp3I sites are present in the vector so that assembled RVD modules can be incorporated in frame. After assembly, the TALEN vectors can be linearized at the XbaI site downstream of the SV40 terminator for in vitro mRNA synthesis. It is very important to follow the exact procedures described in Cermak et al. (2011) and Sakuma et al. (2013), including use of the same kits and enzymes as described there. Construction of TALEN vectors is usually complete in 5 days. We do not perform validation of the assembled TALENs before injection (e.g., single strand-annealing assay in yeast [Doyon et al. 2008] or in cultured cells [Sakuma et al. 2013]), as almost all TALENs we have used have exhibited sufficient activity to obtain germline mutations. Without such validation, activity can be evaluated after injection by recording the number of G0 somatic mosaics (if observable) or by molecular diagnostics such as CEL-I assay (Oleykowski et al. 1998; Kulinski et al. 2000) using genomic DNAs of dead larvae or unhatched eggs.

Mating scheme and screening strategy to establish knockout lines

As one generation of Bombyx takes approximately 50 days and three generations are required to obtain G2 moths carrying a defined mutation allele (i.e., ~150 days after injection at the earliest), it is important to make a well-organized plan to optimize the likelihood of achieving successful targeting in the first trial. Here, we propose a mating scheme and screening strategy for targeting an autosomal gene with unknown phenotype, which should be the most difficult case since induced mutations should be “tracked” solely by molecular techniques.

The first goal is to obtain an adequate number of G0 founder moths. In our experience, 10 G0 founder individuals is generally sufficient for genes that do not cause lethality in injected animals, since the value of % yielders sometimes reaches 100% and many types of mutant alleles can be recovered from a single G0 founder. On the other hand, when targeting genes whose mosaic animals are expected to have a high probability of mortality, the number of eggs injected should be increased to maximize the chance of obtaining fertile adults.

We recommend crossing G0 moths with parental or wild type (wt) strains and avoiding sibling G0 crosses. Sibling crosses can reduce the number of G1 broods to be screened and the labor associated with screening; however, as germline mutation rates often exceed 50% in Bombyx, this “jumbles” the genotype of G1 animals and makes it difficult to establish a line carrying a single mutant allele in the following generations.

The next goal is to identify G1 broods to which mutant alleles are transmitted from G0. CEL-I assay is a powerful tool for this purpose and its detection limit is approximately 1% in our experience, that is, the presence of 1% mutant haplotype in a DNA pool can be called positive. We usually sample approximately 40 neonate larvae or eggs from each brood and investigate the presence of mutant alleles by CEL-I assay. Samples from the same brood can be mixed in a single tube, but we recommend dividing them into several tubes (5–10 larvae/tube) for DNA extraction and subsequent CEL-I assay so that frequency of germline mutations can be roughly estimated in each brood and compared among broods. Finally, 4–6 CEL-I-positive G1 broods are chosen based on the estimated frequency of germline mutations and the presence of large insertions/deletions inferred from fragment size in the CEL-I assay.

The final goal is to establish G2 broods that carry a defined, favorable mutant allele. For this, 50–100 larvae per G1 brood are reared and subjected to sibling mating. After crossing, each G1 moth is given an ID number and subjected to genotyping, that is, CEL-I assay and subsequent DNA sequencing. Alternatively, genotyping can be done using the legs of living moths before crossing or using the hemolymph of living larvae.

After identification of CEL-I-positive G1 moths, nucleotide sequences of induced mutations are determined. Since CEL-I-positive G1 moths are considered to be heterozygous for the target gene, PCR products cannot be sequenced directly and should be subcloned into a cloning vector for sequencing. Three types of products are amplified by PCR: wt/wt and mutant/mutant homoduplexes and a wt/mutant heteroduplex; we usually sequence 10 clones per individual. Finally, G2 broods carrying favorable mutant alleles are chosen to establish knockout lines. We preferentially choose null alleles caused by large insertions/deletions, as they are easily genotyped by PCR in following studies. Careful consideration should be given before discarding hypomorphic alleles.

Anticipated problems

Several problems are anticipated in gene targeting experiments. As strange as it may sound, we have found that TALENs with very high cleavage activity do not always give a good result. This is manifested when a target gene causes high lethality in mosaic G0 animals. For example, we faced a problem in generating knockout lines for Kr-h1, a repressor gene of larval-pupal metamorphosis (Minakuchi et al. 2009; Kayukawa et al. 2013), as almost all the G0 larvae became mosaic animals with larval and pupal cuticles in the skin and these mosaics eventually died before reaching pupal stage (Fig. 5). In such cases, TALENs with very high cleavage activities are not appropriate for use in generating germline mutants, whereas they are useful for mosaic analysis in injected G0. A reliable method for circumventing this problem has not been established; however, given that germline mutants can be recovered from non-mosaic G0 animals as well as mosaic ones in our experience, we suggest lowering the mutation rates induced by TALENs to reduce the frequency of G0 mosaics. For example, one may change injection parameters to lower mutation rates by reducing the concentration of TALEN mRNAs and/or the volume of mRNAs injected. It would also be helpful to design and use new TALEN pairs with lower cleavage activity. This can be achieved by diverging from some of the design guidelines; for example, allowing runs of six or more A + T in TALE binding sites, which is not favorable in a standard experiment (Streubel et al. 2012), will reduce the cleavage activity.

Figure 5.

Larval and pupal mosaic phenotypes in G0 larvae targeted for Kr-h1. (A) Photo of larva with larval (white) and pupal (brown) mosaic skin. (B) Magnified lateral view of a mosaic animal. Pupal cuticles are distributed in the skin in a patchy manner.

Before designing custom TALENs, we recommend sequencing of target genes in recipient strains to examine the presence of inter-strain nucleotide polymorphism, as this may give false-positive fragments in CEL-I assay. Special care for intra-strain polymorphisms should be also taken when non-isogenic strains are used as recipients (e.g., pnd w-1 and other hybrid strains).

Off-target effects of ZFNs, TALENs and CRIPSR/Cas have been extensively studied in mammalian systems (Hockemeyer et al. 2011; Pattanayak et al. 2011) and it has been suggested that the off-target mutations induced by CRISPR/Cas occur much more frequently than those induced by ZFNs and TALENs (Fu et al. 2013). Although it has not been clarified whether the same is true in insects, off-target effects should also be considered. Potential off-target sites should be checked for the induction of unintended mutations, by sequencing or other methods such as CEL-I assay or high-resolution melt analysis. Alternatively, established mutant lines can be outcrossed with another strain to observe a genetic linkage between a mutant phenotype and a genotype of a target locus in F2 or a backcross population. Several tools for computational prediction of TALEN off-targets, such as Paired Target Finder (Doyle et al. 2012) and TALENoffer (Grau et al. 2013), which are applicable to custom input genome data, have been developed.

Future perspectives

Targeted gene disruption has now become a standard and easy technique in Bombyx. However, efficiency of gene knock-in is still very low in Bombyx. For example, in a green fluorescent protein (GFP) knock-in experiment using the BL-1 ZFN pair (see Table 1), 8 out of 85 injected G0 larvae (9.4%) became GFP mosaics, but only one GFP knock-in larva was found in 11 770 G1 larvae (0.0085%) (Y. Takasu, unpubl. data, 2012; Fig. 6). Efficient gene knock-in systems should be developed for more sophisticated genome editing in the future. One promising approach is to increase the frequency of homology-directed repair (HDR) after DSB. Interestingly, the frequency of HDR after DSBs is greatly increased in Drosophila mutants lacking a functional DNA ligase IV (lig4), an essential component of NHEJ. It is thus of interest to create a Bombyx lig4 mutant as well as mutants for other NHEJ components to see if the frequency of HDR is increased. It is also noteworthy that many variants of TALEN and CRISPR/Cas systems have been developed for other purposes, such as activation or suppression of genes of interest, including TALE activators (Crocker & Stern 2013; Konermann et al. 2013), TALE repressors (Cong et al. 2012), and a CRISPR interference system (Qi et al. 2013). These tools can be also used for genetic manipulation of Bombyx.

Figure 6.

Knock-in experiments using zinc finger nucleases (ZFNs). A green fluorescent protein (GFP) knock-in vector (200 ng/μL) consisting of a GFP cassette (A3-GFP-SV40T) and homology arms (left 2217 bp and right 2079 bp) was co-injected with BL-1 ZFN mRNAs (400 ng/μL) into eggs of pnd w-1 strain. Eight out of 85 G0 larvae (9.4%) became GFP mosaic animals (A and B, indicated by arrows), while only one of 11 770 G1 larvae (0.008%) carried a germline knock-in allele (C, indicated by arrow).

In conclusion, rapid progress in genome editing technologies in recent years has initiated a new era for Bombyx molecular genetics. This will greatly increase the usefulness of Bombyx as a model for lepidopteran insects, the major agricultural pests, and lead to sophisticated breeding of Bombyx for use in sericulture and biotechnology.


We thank Dr Hideki Sezutsu, Dr Keiro Uchino, Dr Isao Kobayashi, Dr Toshiki Tamura, Dr Tetsuro Shinoda, and the members of Transgenic Silkworm Research Unit of National Institute of Agrobiological Sciences (NIAS), Japan, for microinjection experiments and discussion. We also thank Ms Mayumi Tomiyama for helping with silkworm rearing and stock maintenance. We are grateful to Dr Takahiro Kusakabe (Kyushu University) for advice on purification of CEL-I from celery. We also thank Dr Tetsushi Sakuma and Dr Takashi Yamamoto (Hiroshima University) for advice on construction of TALENs and for providing vectors for 6-module assembly. This work was supported by MEXT/JSPS KAKENHI Grant Numbers 22128004, 23688008, and 25252059, and by NIAS Strategic Research Fund to TD.