Properties of gene knockdown system by vector-based siRNA in zebrafish



RNA interference (RNAi) has emerged as a powerful tool to silence specific genes. Vector-based RNAi systems have been developed to downregulate targeted genes in a spatially and temporally regulated fashion both in vitro and in vivo. The zebrafish (Danio rerio) is a model animal that has been examined based on a wide variety of biological techniques, including embryonic manipulations, forward and reverse genetics, and molecular biology. However, a heritable and tissue-specific knockdown of gene expression has not yet been developed in zebrafish. We examined two types of vector, which produce small interfering RNA (siRNA), the direct effector in RNAi system; microRNA (miRNA) process mimicking vectors with a promoter for RNA polymerase II and short hairpin RNA (shRNA) expressing vector through a promoter for RNA polymerase III. Though gene-silencing phenotypes were not observed in the miRNA process mimicking vectors, the transgenic embryos of the second vector (Tg(zU6-shGFP)), shRNA expressing vector for enhanced green fluorescence protein, revealed knockdown of the targeted gene. Interestingly, only the embryos from Tg(zU6-shGFP) female but not from the male fish showed the downregulation. Comparison of the quantity of siRNA produced by each vector indicates that the vectors tested here induced siRNA, but at low levels barely sufficient to silence the targeted gene.


Since the zebrafish (Danio rerio) was introduced as an embryological and genetic system (Streisinger et al. 1981), it has become one of the most important model organisms to study biological processes in vivo. Because zebrafish has many of the strengths found in invertebrate models, such as a large number of offspring, a short generation time, and transparent embryos, it allows the use of various approaches and techniques, especially large-scale forward genetic methods (Driever et al. 1996; Haffter et al. 1996) and reverse genetic methods. Reverse genetics is a powerful tool to elucidate the functions of genes, but gene knockout through homologous recombination has not yet been achieved in zebrafish. Instead, antisense morpholino oligonucleotides (MO) have often been used as a loss-of-function approach in zebrafish (Nasevicius & Ekker 2000; Shinya et al. 2001). However, one problem about MO is that injection into a 1-cell stage embryo silences the target gene in the whole body of the embryo, and also transiently works only for a few days in vivo. TILLING (targeting induced local lesions in genomes) strategy is another approach to knockout a gene in zebrafish, in which a library of ENU-mutagenized F1 fish are generated and the DNA of each is screened for mutation in the target gene (Wienholds et al. 2002, 2003). Furthermore, new genome editing technologies have come out recently: custom zinc finger nucleases (ZFNs) (Doyon et al. 2008; Meng et al. 2008) and transcription activator-like effector nucleases (TALENs) (Huang et al. 2011; Sander et al. 2011). These nucleases are able to induce heritable disruption of targeted genes through targeted double-stranded breaks due to the enzymes. While these technologies are undoubtedly powerful tools to dissect molecular mechanisms of biological phenomena, a conditional knockdown system enabling stable tittering down gene dosage in a tissue-specific and/or stage-specific fashion is unavailable in zebrafish.

RNA interference (RNAi) is the process of double-stranded (ds) RNA-dependent, post-transcriptional gene silencing, first identified in Caenorhabditis elegans (Fire et al. 1998). dsRNA in cells is digested by Dicer to yield small interfering RNA (siRNA) 21–23 nt in length, and involves the degradation of the homologous mRNAs (Agrawal et al. 2003). This mechanism has been observed in a wide range of eukaryotes, including fungi, plants and animals (Agrawal et al. 2003). In fact, introduction of siRNA has been reported to silence a targeted gene efficiently in zebrafish cells (Dodd et al. 2004; Gruber et al. 2005). DNA-based vector systems to generate siRNA have proved to be an efficient method to mediate heritable, sequence-specific silencing of genes in various model organisms (Katahira & Nakamura 2003; Wiznerowicz et al. 2006). Conditional knockdown has been achieved in mammalian cells and animals by shRNA transcribed by RNA polymerase III (Ventura et al. 2004). In combination with a natural backbone of the precursor or primary microRNA (miRNA) transcripts, shRNAs can be produced in higher amounts (Chang et al. 2006). This miRNA-mimicked shRNA can be generated by RNA polymerase II promoter (Du et al. 2006), which offers several advantages over the RNA polymerase III promoter, including easier regulation in temporal and/or spatial expression. Recently, long dsRNA decapped by ribozyme in the cassette, which is transcribed by RNA polymerase II has also successfully induced knockdown of endogenous genes (Hou et al. 2011).

Here, with the aim of constructing conditional gene knockdown systems in zebrafish, we tested two types of DNA-based vector to generate siRNAs in zebrafish cells. One was a vector mimicking miRNA process and the transcripts were produced by RNA polymerase II promoter. The other was an shRNA expression vector through RNA polymerase III. As a targeted gene for model cases, three genes were selected; one exogenous gene: enhanced green fluorescence protein (egfp), and two endogenous genes: sonic hedgehog (shh) and zebrafish bone morphogenetic protein (zbmp2b). Furthermore, to elucidate the relationship between the siRNA quantity and the successful knockdown of the targeted genes, the relative amounts of siRNA were examined in the siRNA injected embryos and the transgenic embryos, using the case of egfp target.

Materials and methods

Fish maintenance and strains

The use of zebrafish for experimental purposes was conducted in accordance with the guidelines of the National Institute of Genetics (NIG). Zebrafish were maintained at 27°C and embryos were collected from natural crosses of adult fish. Collected embryos were maintained in 0.3 × Danieau's solution (17.4 mmol/L NaCl, 0.21 mmol/L KCl, 0.12 mmol/L MgSO4, 0.18 mmol/L Ca(NO3)2, 1.5 mmol/L Hepes at pH 7.6 (Kane & Kishimoto 2002)) at 28.5°C. Embryos were staged according to hours postfertilization (hpf) at 28.5°C and morphological criteria (Kimmel et al. 1995). India strain was used in most experiments, and the only exception was SAGFF73A, a Gal4 transgenic line (Asakawa et al. 2008), to induce maternal and zygotic expression ubiquitously (provided by Professor K. Kawakami, NIG). This Gal4 line is lethal in the homozygous, so that it was maintained in the heterozygous. To generate transgenic strains, microinjection was performed in 1- to 2-cell stage embryos with 500 pl of injection solution containing 25 ng/μL of DNA and 25 ng/μL of tol2 transposase mRNA. Injected embryos were raised to sexual maturity and crossed with wild-type zebrafish to generate F1 progeny, which were screened by using green fluorescent d2EGFP at 24 hpf for pTG-d2egfp, red fluorescent DsRed-express at 24 hpf for pTG-dsR-shGFP, pTG-dsR-NC, pT2AUAS-dsR-shGFP, pT2AUAS-dsR-shshh and pT2AUAS-dsR-shbmp2b, and polymerase chain reaction (PCR) at 3 dpf (days postfertilization) for pzU6-shGFP and pzU6-NC with the primers: LG2_Sd and A155 (see Supporting Information). F1 heterozygous fish were then intercrossed to establish stable transgenic strains.

siRNA design, mRNA synthesis and injection

The sequences of siRNAs used in this study are shown in Table 1. sishh-1 to sishh-7 were siRNAs designed for shh gene (NM_131063), and sibmp2b-1 and sibmp2b-2 were for zbmp2b gene (NM_131360). Potential siRNA target sites in the zebrafish were determined using a web-based software, IDT SciTools RNAi Design ( for sishh-1 to sishh-3, siSearch ( for sishh-4, and sishh-5 to sishh-7 were designed by siRNA Target Sequence Design Service (B-bridge International, Sunnyvale, CA, USA). sibmp2b-1 was from siDirect ( and sibmp2b-2 were designed by siRNA Target Finder supplied by Ambion ( Synthesized siRNA duplexes were obtained from B-bridge International, Inc. for sishh-5 to -7, and from Nippon EGT (Toyama, Japan) for the rest. siRNAs were injected into 1- to 2-cell stage embryos. Capped sense RNAs were synthesized using the mMESSAGE mMACHINETM large scale in vitro transcription kit (Ambion) from the plasmids containing zebrafish h1m-egfp and tol2 transposase (Kawakami et al. 2004). 15 pg mRNA was injected into each of the one-cell stage embryos.

Table 1. Sequences of siRNAs against the egfp, shh, and zbmp2b
siRNAsSense/antisense strandsTarget genesPositions
  1. Positions correspond to coding sequences of the target mRNAs. Underlined sequences represent mismatched bases.




egfp 443–461



shh 384–402



shh 998–1016



shh −61 to −43



shh 1136–1154



shh 281–299



shh 426–444



shh 770–788



zbmp2b 157–175



zbmp2b 711–729

Cloning and construction

For construction of h1 m-egfp, the full-length of histone h1m (h1m) cDNA (NM_183071) was amplified by PCR and inserted into pCS2+ containing EGFP such that the EGFP is placed in the C terminus of H1M linked by six amino acids (GGSSGG, the vector was kindly provided by Dr Y. Kishimoto, NIG). pTG-d2egfp was constructed by exchanging a NotI-BamHI fragment of pd2EGFP-N1 (Clontech) containing d2egfp with egfp in pT2AL200R150G provided by Professor K. Kawakami (Urasaki et al. 2006). To generate pTG-dsR-shGFP, dsRed-miR cassette (Fig. 2A) was first cloned into pCS2+ (Rupp et al. 1994) (pCS2-dsR-shGFP). PCR amplified fragment of dsR-shGFP with SV40 poly(A) from pCS2-dsR-shGFP was then replaced with egfp with SV40 poly(A) of pT2AL200R150G. pCS2-dsR-shGFP was generated by two steps as follows. miR-30a precursor (Wienholds et al. 2005) mimicking shGFP fragment was first cloned between EcoRI and XbaI sites of pCS2+ , and then, dsRed-express fragment from pDIREN-RetroQ-DsRed-Express was inserted at BamHI site of the pCS2+ containing shGFP. miR-30a precursor mimicking shGFP fragment was obtained by amplification of miR-shGFP (see Supporting Information) by primers (miR-long-Eco and miR-long-Xba, see Supporting Information). dsRed-miR cassette of pTG-dsR-NC was also replaced from pCS2-dsR-NC by the same method used for pTG-dsR-shGFP. pCS2-dsR-NC was generated by replacing miR-30a precursor mimicking shGFP fragment of pCS2-dsR-shGFP with PCR amplified shGFP-NC fragment. This fragment was obtained by amplification of miR-shGFP-NC (see Supporting Information) by primers (miR-short-Eco and miR-short-Xba, see Supporting Information). For pT2AUAS-shGFP, dsRed-miR cassette was amplified from pTG-dsR-shGFP, and cloned in pT2AUASMCS (provided by Professor K. Kawakami) between BglII and XhoI sites (Asakawa et al. 2008). pT2AUAS-shshh was generated by replacing miR-30a precursor mimicking shGFP fragment of pT2AUAS-shGFP with a fragment obtained by amplification of miR-shshh (see Supporting Information) by primers (miR-short-Eco and miR-short-Xho, see Supporting Information). pT2AUAS-shbmp2b was also generated by the same method, amplifying miR-shbmp2b (see Supporting Information) with primers (miR-short-Eco and miR-short-Xho). Zebrafish U6 small nuclear RNA (snRNA) was identified using a BLAST search against the zebrafish genome. To obtain its promoter, about 300 bp of 5′ flanking sequence were amplified by PCR using a primer set, LG2_Sd and LG2_Ab (see Supporting Information) with genomic DNA of India strain. PCR product was cloned into pSIREN (Clontech) by replacing with human U6 snRNA promoter region (pzU6). The sequence of the cloned promoter is shown in the Supporting Information. Sense and antisense long oligos for shRNA to generate siGFP and siGFP-NC, respectively, were annealed, and cloned into pzU6 by following the protocol supplied by Clontech (pzU6-shGFP and pzU6-shGFP-NC). The sequences of the oligos are described in the Supporting Information.

RT-qPCR, RNA extraction and northern blot

Relative expression of the target mRNA was examined by reverse transcription quantitative PCR (RT-qPCR). RT-qPCR was performed in an LightCycle 480 Real-Time PCR system (Roche Applied Science), following the protocol supplied by the manufacturer. β-actin gene was applied as an internal control. All the used primer sequences are shown in the Supporting Information. t-test was used to determine significant reduction of mRNA against the control fish. cDNA from each embryo for RT-qPCR was prepared according to a protocol given by the manufacturer with some modifications (Cells-to-cDNA II kit, Ambion). Total RNA from embryos at 2.5, 10 and 30 hpf was obtained using Trizol reagent (Invitrogen) by following the manufacturer's instructions. Northern blot analysis was performed as described previously with some modifications (Valoczi et al. 2004; Giraldez et al. 2005). Five micrograms of total RNA per lane were loaded and probed by DIG labeled LNA (locked-nucleic acid)–modified DNA oligonucleotide for siGFP. LNA probe sequence is shown in the Supporting Information.


Confirming inhibition of target genes by siRNA in zebrafish embryos

To confirm gene-silencing through siRNA in zebrafish embryos, we first tested the effects of synthetic siRNA for exogenous egfp, and endogenous shh, and zbmp2b genes. For egfp gene, the effective sequence of siRNA (siGFP) has already been shown in several organisms, such as Drosophila, chick and mammalian cells (Ui-Tei et al. 2004). Therefore, it would be suitable for a target gene in order to confirm the presence of gene-silencing system in vivo. Two endogenous genes, shh and zbmp2b, were selected as target genes for model cases in this study, because they are early functioning genes so that transient gene silencing by siRNA injection can easily be detected. In addition, the silenced phenotypes can be compared to the mutant phenotypes (van Eeden et al. 1996; Mullins et al. 1996; Kishimoto et al. 1997; Schauerte et al. 1998).

First, the inhibitory effect of siGFP was examined: h1m-egfp mRNA was injected into the embryos at 1-cell stage, followed by injection of 9 fmol siGFP or siGFP-NC as a negative control. The level of EGFP fluorescence in siGFP injected embryos was slightly lower than that observed in siGFP-NC embryos at 9 hpf (data not shown), and the difference was clear at 24 hpf (Fig. 1B,D). In accordance with the above observation, RT-qPCR for egfp revealed that the amount of h1m-egfp mRNA was significantly reduced in the siGFP injected embryos (= 1.1 × 10−10): about 35% of that in the siGFP-NC embryos at 10 hpf. These data indicate that siRNA with an appropriate sequence can silence the target gene in zebrafish embryos.

Figure 1.

Down regulation of the targeted genes by siRNAs. (A–D) Embryos at 24 hpf injected with h1 m-egfp mRNA and siGFP-NC (A, B) and siGFP (C, D). EGFP fluorescence in siGFP injected embryos (D) was clearly weaker than those injected siGFP-NC (B). B, D are fluorescent fields of A, C. (E) Seven sequences of sishh were injected into zebrafish embryos, and relative expressions of shh mRNA for each were quantified by RT-qPCR at 10, 12, 14, 16 and 18 hpf. Thick bar indicates the amount of shh mRNA equal to the control (siGFP injected embryos). Asterisks indicate the significant reduction of mRNA compared to the control. Error bars, S. E. (F, G) Trunk region of 28 hpf embryos in the lateral view. In the siGFP injected control embryo (F), V-shaped somites (arrowhead) were observed. On the contrary, somites (arrowhead) were malformed as U-shaped in the sishh-2 injected embryo (G). (H) Dorsalized phenotype in the sibmp2b-2 injected embryos at 26 hpf. Various levels of dorsalization were observed dose-dependently in sibmp2b-2 injected embryos (normal to dorsalized C1–C4). C1 to C4 corresponds to the dorsalized phenotypes classified in the previous study (Kishimoto et al. 1997). The numbers on the right are the number of embryos that exhibited each phenotype in the two doses of sibmp2b-2 (18 and 36 fmol) injection. The phenotypic distribution was significantly different between the two doses (= 0.0021 in Fisher's exact test). Scale bar, 400 μm in A–D, 80 μm in F, G, 250 μm in H.

For the shh gene, seven sequences of the siRNA (sishh-1 to −7) were examined because there is a sequence preference in siRNA. The same amount of the siRNA (18 fmol) was injected into embryos, and the relative expression of shh mRNA against the siGFP injected embryos as a common control was examined at 10, 12, 14, 16 and 18 hpf (Fig. 1E). sishh-2 was the most effective, especially about 40% expression of shh mRNA compared to the siGFP injected at 10 hpf, followed by sishh-5, −6 and −3. The malformed development, which has not been reported in shh mutants, was also observed most severely in the sishh-2 injected embryos, followed by sishh-7, −5 and −6 (data not shown). However, by reducing the amount of injected siRNA to 0.72 fmol, sishh-2 injection could produce the U-shaped somites, which was shh mutant phenotype (12 out of 27 embryos, 44.4%), with less other malformed development (Fig. 1F,G). The phenotype frequency was similar to that reported in shh-MO injection (Nasevicius & Ekker 2000), but half of the embryos with the phenotype also revealed bending tail, which might be the off-targeting effect. The other siRNAs for shh did not show the mutant phenotype by injection at this amount (data not shown). Therefore, sishh-2 seemed to be the best sequence for inhibition in the tested seven sequences, although the amount of the siRNA might need to be controlled accurately.

Zbmp2b regulates early dorsoventral patterning in zebrafish embryos through its ventralizing activity, so that the mutant showed the severe dorsalized phenotype (Kishimoto et al. 1997). We examined two sequences of the siRNA for zbmp2b gene (sibmp2b-1 and −2). Both sibmp2b-1 and sibmp2b-2 exhibited the dorsalized phenotype when they were injected into early embryos, but the malformed development, which has not been reported in zbmp2b mutants was observed in sibmp2b-1 injected embryos (data not shown). Injection of sibmp2b-2 revealed various levels of dorsalization in a dose dependent manner with almost no other malformed development (Fig. 1H). This observation indicated that sibmp2b-2 could inhibit the zbmp2b function. Indeed, zbmp2b mRNA was reduced at least by about 60% in sibmp2b-2 injected embryos (Table 2), indicating that sibmp2-2 was an excellent siRNA sequence for silencing zbmp2b.

Table 2. Effect of sibmp2b-2 injection
HpfRelative expressionP-valuea
Avg. ± SD
  1. a

    P-value of t-test.

40.699 ± 0.1360.024
60.923 ± 0.0680.076
80.574 ± 0.2990.052
100.674 ± 0.2020.042
120.986 ± 0.1460.432

miRNA process mimicking vectors with a promoter for polymerase II

For conditional knockdown, siRNAs need to be generated where and when we aim. Here, we tested a type of vector that expressed mRNA of dsRed-express followed by miR-30a precursor mimicking sequences including its 3′ and 5′ flanking regions (dsRed-miR cassette, Fig. 2A). This transcript generates DsRed-express protein as a marker and an shRNA, which is expected to be processed to siRNA in zebrafish cells. As the mRNA is transcribed by RNA polymerase II, it is relatively easy to regulate the expression temporally and/or spatially by using known promoters and enhancers of various genes. Three target genes were tried using this type of vector, as described below.

Figure 2.

Transgenic embryos with miRNA processing mimicking vectors. (A) Schematic representation of the dsRed-miR cassette. (B) Northern blot analysis revealed generation of siGFP processed from miR-shGFP. An arrowhead indicates siGFP. TG: Tg(EF1α:dsR-shGFP) heterozygous embryos at 30 hpf, wt: wild-type embryos at 30 hpf and PC: siGFP (0.1 pmol). (C–E) d2EGFP fluorescence (E) of Tg(EF1α:d2egfp) with Tg(EF1α:dsR-shGFP) (lower) was normally observed just like with Tg(EF1α:dsR-shGFP-NC) (upper), although DsRed was expressed in both transgenic embryos (D). (F–H) Normal V-shaped somites were observed in both embryos from SAGFF73A crossed to Tg(EF1α:dsR-shshh) (H) and Tg(EF1α:dsR-shGFP) (G), respectively, although DsRed was expressed in both transgenic embryos (F, upper: Tg(EF1α:dsR-shGFP), lower: Tg(EF1α:dsR-shshh)). (I) Relative amount of shh against the control was not reduced in the embryos at 10 hpf, but slightly reduced at 24 hpf. (J, K) Normal embryos were obtained from the females heterozygous with SAGFF73A and EF1α:dsR-shbmp2b, crossed to Tg(EF1α:dsR-shbmp2b) males (K, lower) and Tg(EF1α:dsR-shGFP) males (K, upper), respectively, although DsRed was expressed in both transgenic embryos (J, upper: Tg(EF1α:dsR-shGFP), lower: Tg(EF1α:dsR-shbmp2b)). (C–H, J, K) Lateral view. Scale bar, 400 μm in C–E, J, K; 500 μm in F; 80 μm in G, H.

One vector, pTG-dsR-shGFP was designed to express shRNA generating siGFP (shGFP) under control of Xenopus EF1α enhancer/promoter which induces ubiquitous expression in zebrafish embryos (Johnson & Krieg 1994; Kawakami et al. 2004). Using LNA-modified DNA oligonucleotide probes, Northern blot analysis showed generation of siGFP in the embryos from pTG-dsR-shGFP transgenic zebrafish (Tg(EF1α:dsR-shGFP), Fig. 2B, arrowhead). Then, Tg(EF1α:dsR-shGFP) was crossed with transgenic zebrafish of pTG-d2egfp (Tg(EF1α:d2egfp)), which express d2EGFP under control of Xenopus EF1α enhancer/promoter. d2EGFP is a destabilized variant of EGFP by fused amino acids 422–461 of the degradation domain of mouse ornithine decarboxylase (MODC) to its C-terminal end (Li et al. 1998). We used d2EGFP because a rapid protein turnover of d2EGFP was expected to make the fluorescent level sensitive to siGFP. As a negative control, a transgenic strain, Tg(EF1α:dsR-shGFP-NC) was prepared, which carried a transgene pTG-dsR-NC expressing DsRed-express and siGFP-NC. Though almost the same strength of DsRed-express expression was observed in the embryos from both Tg(EF1α:dsR-shGFP) and Tg(EF1α:dsR-shGFP-NC) crossed with Tg(EF1a:d2egfp) (Fig. 2D), there was no clear difference in the level of green fluorescence between them (Fig. 2E). In the mRNA level of d2egfp, there was no significant difference between them at 24 hpf (= 0.980).

shh, the essential gene in early development (Schauerte et al. 1998), was another target to knockdown. The Gal4-UAS system was used for this gene in order to control the timing of siRNA expression. Transgenic zebrafish of pT2AUAS-shshh (Tg(UAS:dsR-shshh)) was prepared, in which dsRed-express and sequences to generate shRNA for shh (shshh; generating sishh-2) were coded under control of UAS sequences. Tg(UAS:dsR-shshh) was crossed to SAGFF73A, which is a Gal4 line expressing Gal4 maternally and ubiquitously in zebrafish embryos. Their embryos showed a considerable level of red fluorescence of DsRed-express, but revealed normal development with V-shaped somites just like those from the cross of SAGFF73A and Tg(UAS:dsR-shGFP), transgenic zebrafish of pT2AUAS-shGFP as a negative control (Fig. 2F–H, compare to Fig. 1G). The phenotype of bending tail, which was observed in sishh-2 injected embryos, was also not observed. RT-qPCR analyses also showed no difference in shh expression level at 10 hpf (= 0.803), but revealed slight reduction of shh mRNA (about 91.5% of the control) at 24 hpf (= 0.043, Fig. 2I).

The last target gene was zbmp2b; however, the zbmp2b homozygous mutants do not develop beyond the embryonic stage (Kishimoto et al. 1997), dsRed-miR cassette, which expresses dsRed-express and sibmp2b-2 through the processing of shbmp2b was also placed under control of UAS sequences. Transgenic zebrafish of the vector, Tg(UAS:dsR-shbmp2b), was crossed to SAGFF73A, but the embryos with considerable fluorescence of DsRed-express developed and grew up to adults normally (data not shown). Consistent with the above observation, there were no differences in zbmp2b mRNA expression levels between the shbmp2b expressing embryos and the shGFP expressing embryos at 11 hpf (= 0.983). Zbmp2b is expressed and functions quite early in development (Nikaido et al. 1997). Therefore, we examined embryos with maternal and zygotic expression of shbmp2b. All the embryos were normal in the cross of SAGFF73A males and females heterozygous in both loci of SAGFF73A and UAS:dsR-shbmp2b (Fig. 2J,K). zbmp2b mRNA was also normally expressed in those embryos at 7 hpf (= 1.000).

shRNA expressing vector through a promoter for polymerase III

All of the three genes failed to show clear knockdown phenotypes by using polymerase II promoters as described above. Thus, we further tested the other type of vector in which shRNA was directly transcribed by RNA polymerase III. The major advantages of polymerase III promoters are their demonstrated high activity in most cell types and the small size of the expression cassettes. Taking these, various vectors based on the polymerase III promoters have been used and achieved the robust levels of knockdown (Ventura et al. 2004; Chang et al. 2006; Wiznerowicz et al. 2006). To examine this type of vector in zebrafish, we cloned about 300 bp of 5′ flanking sequence for zebrafish U6 snRNA on LG2 (Supporting Information). Here, we focused on the analysis for this type of vector on the shGFP efficiency, because the LNA-modified DNA oligonucleotide probe allowed us quantitative analysis for siGFP. Transgenic zebrafish, Tg(zU6:shGFP), was established by injecting a plasmid with the promoter followed by shGFP sequence (pzU6-shGFP, Fig. 3A). Tg(zU6:shGFP-NC) was also prepared as a negative control. Northern blot analysis revealed the generation of siGFP in the Tg(zU6:shGFP) embryos (Fig. 3B, lane 1 and 2).

Figure 3.

Transgenic embryos with shRNA expressing vector through a promoter for RNA polymerase III. (A) Schematic representation of the vector in which zebrafish U6 snRNA promoter (zU6 pro) drives shRNA. (B) Northern blot analysis of 30 hpf (lane 1, 2) and 2.5 hpf (lane 3, 4) embryos obtained from Tg(zU6-shGFP) male (lane 1, 3) and female (lane 2, 4). 0.1 pmol siGFP was loaded as a control in lane 5. More siGFP (arrowhead) was detected in the embryos from Tg(zU6-shGFP) female than those from the male at 30 hpf. (C–F) Down regulation of d2EGFP was observed only in the cross of Tg(zU6-shGFP) female and Tg(EF1α:d2egfp) male (E, F, easy to see the difference in the head region) but not in the cross of Tg(zU6-shGFP) male and Tg(EF1α:d2egfp) female (C, D). Upper is the control embryos crossed with Tg(zU6-shGFP-NC) and lower is Tg(zU6-shGFP) embryos. (G) Relative amount of d2egfp against the control was reduced in the embryos from Tg(zU6-shGFP) female, but not in the embryos from Tg(zU6-shGFP) male. Scale bar, 400 μm.

Tg(zU6-shGFP) and Tg(zU6:shGFP-NC) were crossed to Tg(EF1α:d2egfp), respectively, and examined d2EGFP expression in the 28 hpf embryos. The green fluorescence of d2EGFP was observed in the same level between the embryos from Tg(zU6-shGFP) male and those from Tg(zU6:shGFP-NC) male (Table 3, Fig. 3C,D). Unexpectedly and interestingly, downregulation of d2EGFP was observed in the embryos from Tg(zU6-shGFP) females compared to those from Tg(zU6:shGFP-NC) females (Table 3, Fig. 3E,F). The difference in the mRNA of d2egfp was also detected only in the embryos obtained from the cross of Tg(zU6-shGFP) females and Tg(EF1α:d2egfp) males. RT-qPCR analyses at 30 hpf (Fig. 3G), revealed that d2egfp mRNA was expressed equally in Tg(zU6-shGFP) male cross compared to Tg(zU6:shGFP-NC) male cross, but was significantly reduced to about 70% of the control (Tg(zU6:shGFP-NC) female cross) in the Tg(zU6-shGFP) female cross (Table 3, P = 1.5 × 10−9).

Table 3. Parental sex-dependent downregulation of d2EGFP
FemaleMaleFluorescence levelaP-valueb
  1. a

    The green fluorescence level compared to the embryos from Tg(EFlα:d2egfp) and Tg(zU6-shGFP-NC).

  2. b

    P-value of t-test for d2egfp mRNA.

Tg(EF1α:d2egfp)Tg(zU6-shGFP)Not changed0.754
Tg(zU6-shGFP)Tg(EF1α:d2egfp)Reduced1.5 × 10−9

Although maternal transcript of d2egfp was likely to be present at quite a low level by RT-qPCR of embryos at 2.5 hpf, d2EGFP fluorescence was not observed maternally in Tg(EF1α:d2egfp) eggs (data not shown). We could not completely exclude the possibility that the maternal transcript led to the parental sex difference on the downregulation of the target gene, but considering the rapid turnover of d2EGFP, the faint maternal supply of d2egfp mRNA was insufficient to affect the level of the green fluorescence at 28 hpf or d2egfp mRNA level at 30 hpf. On the other hand, there seems little or no maternal supply of siGFP, as the band was not detected in northern blot analysis at 2.5 hpf (Fig. 3B, lane 3 and 4). However, a larger amount of siGFP was likely to be generated in embryos from Tg(zU6-shGFP) female than those from the male (Fig. 3B, compare lane 1 and 2). This different expression level of siGFP might be caused by the positional effect of the transgene, or by the cloned zU6 snRNA promoter itself. As we obtained only one transgenic strain for Tg(zU6-shGFP), the reason for this female-biased expression is unclear at present. Anyway, a larger amount of siGFP in embryos from Tg(zU6-shGFP) female might cause the above phenomenon: parental sex dependent downregulation.

Relative amount of siRNA generated from the vectors

We selectively used effective siRNA sequences for the shRNA expressing vectors; however, only one vector, pzU6-shGFP, exhibited the knockdown phenotype, and only in the embryos from Tg(zU6-shGFP) female at that. Considering that the embryos from Tg(zU6-shGFP) female produced larger amounts of siGFP than those from Tg(zU6-shGFP) male (Fig. 3B, lane 1 and 2), the siRNA amount produced in the embryos might affect the success/failure of knockdown by the vectors. To examine this possibility, we checked the relative amount of siRNA produced by three types of vectors tested in this study: the vector with zU6 promoter, and two miRNA process mimicking vectors with EF1α promoter and UAS, respectively. As there were transgenic zebrafish carrying each of the three vectors, the levels of siGFP were compared among embryos as follows: siGFP injected and originated from Tg(zU6-shGFP) female, Tg(zU6-shGFP) male, Tg(EF1α:dsR-shGFP) female, Tg(EF1α:dsR-shGFP) male, Tg(UAS:dsR-shGFP) female, and Tg(UAS:dsR-shGFP) male. Embryos from Tg(UAS:dsR-shGFP) female and male were obtained by crossing with SAGFF73A male and female, respectively.

At 10 hpf, a strong signal of siGFP was detected in siGFP injected embryos by northern blot, and about 12% of the signal was observed in the embryos from Tg(zU6-shGFP) female (Fig. 4). The embryos from Tg(zU6-shGFP) male expressed siGFP in less than half of the embryos from Tg(zU6-shGFP) female. In the cases of transgenics with miRNA process mimicking vectors, siGFP was detected at only about 1% or less of siGFP injected embryos. The sexual bias in the siGFP level was only found in Tg(zU6-shGFP), but not in the other two transgenic strains. These data suggest that the vectors tested here did generate siRNA but not in sufficient quantities to inhibit the target gene clearly.

Figure 4.

Relative quantification of siGFP in embryos at 10 hpf. Northern blot analysis of 9 fmol siGFP injected embryos (lane 1), and embryos from Tg(zU6-shGFP) female (lane 2), Tg(zU6-shGFP) male (lane 3), Tg(EF1α:dsR-shGFP) female (lane 4), Tg(EF1α:dsR-shGFP) male (lane 5), Tg(UAS:dsR-shGFP) female (lane 6), and Tg(UAS:dsR-shGFP) male (lane 7). Embryos from Tg(UAS:dsR-shGFP) female and male were obtained by crossing with SAGFF73A male and female, respectively. An arrowhead indicates siGFP. Figures at the bottom indicate the amount of siGFP for each lane, relative to the siGFP injected (lane 1).


In this study, we tried to establish a conditional gene knockdown system through siRNA by two types of vectors, miRNA process mimicking vectors with a promoter for RNA polymerase II and shRNA expressing vector through a promoter for RNA polymerase III. The miRNA process mimicking vector was examined by using functional siRNA sequences for three genes: exogenous d2egfp and endogenous shh and zbmp2b. However, gene-silencing phenotypes were not observed in any cases. On the other hand, the shRNA expressing vector achieved downregulation of d2egfp at 70%, and in the level of phenotype, d2egfp inhibition in the embryos obtained from the transgenic male of pTG-d2egfp crossed with the transgenic female of the shGFP expressing vector. Consistently, siGFP was expressed most strongly in those embryos from Tg(zU6-shGFP) female compared to the others: embryos from Tg(zU6-shGFP) male and from transgenics of the miRNA process mimicking vectors. These results indicate that the vectors tested here could induce siRNA, but at low amounts barely sufficient to silence the targeted gene.

Both of the two endogenous genes tested here failed to show the knockdown phenotypes by the miRNA process mimicking vectors, though synthetic siRNAs have both been confirmed effective. In the vectors constructed for these genes and egfp, dsRed-express and shRNA precursor were transcribed as a single mRNA (Fig. 2A), and thus, DsRed-express functioned as an indicator for siRNA expression. Tg(UAS:dsR-shGFP), Tg(UAS:dsR-shshh) and Tg(UAS:dsR-shbmp2b) revealed similar levels of DsRed-express fluorescence, when they crossed with the Gal4 line (Fig. 2F, for example). Since we observed the correlation between the strength of DsRed-express fluorescence and the expressed siGFP level (data not shown), the amounts of sishh and sibmp2b in the transgenic embryos could be estimated as about the same to siGFP produced in Tg(UAS:dsR-shGFP). Thus, siRNAs in transgenics of pT2AUAS-dsR-shRNA vectors seemed <1% of siRNA-injected embryos at 9 fmol (Fig. 4). Considering that 18 fmol injection of sibmp2b-2 led to the mildest phenotype (C1 in Fig. 1H) of zbmp2b mutant in most embryos, siRNA produced in Tg(UAS:dsR-shbmp2b) might be too little to silence the target gene. In the case of shh, the mutant phenotype was observed in the siRNA injection at a quite small amount, 0.72 fmol (Fig. 1F,G). Probably, because of this strong inhibitory effect of sishh-2, such a small amount of siRNA could only reduce shh mRNA slightly. Still, it was unlikely to be enough to show the phenotypes including the malformed development of sishh-2. In summary, the failure of clear knockdown in two endogenous genes, shh and zbmp2b, seems to have resulted from the shortage of siRNA.

It is reported that a similar type of vector, in which shRNA was expressed from a single transcript mimicking the natural miRNA-30e precursor, achieved efficient knockdown by targeting 3′ UTR of genes, just like the natural miRNA function (Dong et al. 2009). Furthermore, another group has recently evaluated the same vector system, and reported that shRNAs were able to target the genes tested mapped to both the 3′ UTR and to the open reading frame (De Rienzo et al. 2012). Two major points are different in our vector design from the study of De Rienzo et al. (2012). The first difference was the promoters; we used Xenopus EF1α enhancer/promoter with rabbit β-globin intron to enhance expression (Kawakami et al. 2004) and UAS sequence repeated tandem five times, and in their vectors, β-actin promoter, no tail promoter and miR-124 promoter were used. It is likely that the difference in the promoter activity affects the quantity of siRNA. The second difference was the sequence of mimicked miRNA precursor and its flanking region. We referred to miR-30a precursor with only about 20 bp 5′ flanking sequence (see Supporting Information), and De Rienzo et al. used 409 bp of genomic sequence, which included miR-30e precursor. As it has been previously shown that the 5′ and 3′ flanking sequence of miRNA precursor have an influence on miRNA processing (Giraldez et al. 2005; Dong et al. 2009), this backbone difference was likely to lead to an insufficient amount of siRNAs resulting in the failure of the most vectors in our study.

d2EGFP was successfully downregulated in the transgenic fish of shGFP expressing vector through zU6 snRNA promoter, although the downregulation was limited only in the case of female cross. Furthermore, we confirmed 5 to 12% of siRNA was produced in the Tg(zU6-shGFP) embryos, compared to the 9 fmol siRNA injected (Fig. 4). Together with the observation that sishh-2 raised the knockdown phenotype at 0.72 fmol injection (8% of 9 fmol), there is a possibility that shshh under the control of zU6 promoter may work as a knockdown vector. In this case, to keep as a transgenic line, it is necessary to generate a conditional system to control the activity of zU6 promoter by using Tet-on/off or Cre/loxP systems. In combination with highly efficient siRNA sequences and expression inducible systems, pzU6-shRNA may be a promising vector for conditional gene-silencing in zebrafish.

Finally, our study suggests two key findings for successful vector-based siRNA systems in zebrafish. One is the inhibitory efficiency of the selected siRNA sequence, supported by the observation as follows: Tg(UAS:dsR-shshh) and Tg(UAS:dsR-shbmp2b) carried the vectors with the same design except for siRNA sequences, but only Tg(UAS:dsR-shshh) showed the reduced mRNA of the target gene. Further analyses are required to clarify sequence preference for the zebrafish siRNA system. The other key finding is how to achieve a considerable amount of siRNA in zebrafish cells. To produce siRNA more efficiently, optimization of promoters and processing steps for siRNA are necessary to establish a more useful and widespread knockdown system in zebrafish.


This work was supported by the Program for the Promotion of Basic Research Activities for Innovative Biosciences (BRAIN) from the Bio-Oriented Technology Research Advancement Institution of Japan. We thank Dr K. Kawakami for providing SAGFF73A, tol2 transposase cDNA, pT2AL200R150G and pT2AUASMCS, Dr Y. Kishimoto for providing pCS2 + containing egfp with a linker, and Mr T. Mizutani at B-Bridge International for providing the service designing siRNA sequence and siRNA. We thank Ms M. Kojima, K. Saka, T. Hoshikawa, A. Takagi, Y. Yoshida, Y. Kida, N. Suzuki, E. Ashikawa, and R. Maeda for their helpful maintenance of the zebrafish facility.