Heritable and Lineage-Specific Gene Knockdown in Zebrafish Embryo

Background Reduced expression of developmentally important genes and tumor suppressors due to haploinsufficiency or epigenetic suppression has been shown to contribute to the pathogenesis of various malignancies. However, methodology that allows spatio-temporally knockdown of gene expression in various model organisms such as zebrafish has not been well established, which largely limits the potential of zebrafish as a vertebrate model of human malignant disorders. Principal Finding Here, we report that multiple copies of small hairpin RNA (shRNA) are expressed from a single transcript that mimics the natural microRNA-30e precursor (mir-shRNA). The mir-shRNA, when microinjected into zebrafish embryos, induced an efficient knockdown of two developmentally essential genes chordin and α-catenin in a dose-controllable fashion. Furthermore, we designed a novel cassette vector to simultaneously express an intronic mir-shRNA and a chimeric red fluorescent protein driven by lineage-specific promoter, which efficiently reduced the expression of a chromosomally integrated reporter gene and an endogenously expressed gata-1 gene in the developing erythroid progenitors and hemangioblasts, respectively. Significance This methodology provides an invaluable tool to knockdown developmental important genes in a tissue-specific manner or to establish animal models, in which the gene dosage is critically important in the pathogenesis of human disorders. The strategy should be also applicable to other model organisms.


Introduction
Understanding the development and disease-associated molecular and cellular processes in model organisms has largely relied on gene loss-of-function approaches. Homologous recombinationmediated gene knockout has not yet been achieved in zebrafish, due to the difficulty of generating embryonic stem cell line. The generation of zebrafish knockout has instead taken use of TILLING (targeting induced local lesions in genomes) strategy, in which a library of ENU-mutagenized F1 animals are generated and kept either as a cryopreserved sperm or as a living stock, and the DNA of these animals is screened for genetic lesion in specific exons [1]. Recently, heritable targeted gene disruption with designed zinc-finger nucleases has been reported to inactivate zebrafish golden, ntl and vascular endothelial growth factor-2 receptor genes [2,3]. While these technologies will undoubtedly speed up the dissection of signal transduction pathways/networks during development and evolution, a conditional knockdown system with stably titering down gene dosage in a tissue-specific fashion is unavailable. This latter strategy is critically important toward establishing zebrafish as a vertebrate model of human pathological conditions and diseases [4], because increasing evidence has demonstrated that haploinsufficiency and epigenetic suppression of tumor suppressor genes, other than complete mutational inactivation or permanent removal of genetic material from the host genome, might be a preferred mechanism in promoting cell transformation [5]. For instance, mice carrying hypomorphic Sfpi1 enhancer allele that reduces Pu.1 expression to 20% of normal levels develop acute myeloid leukemia (AML), while a 50% or even a 100% loss of Pu.1 expression only induc accumulation of abnormal myeloid precursors [6]. Recent studies also indicate that haploinsufficiency of RPS14 and even lower levels of a-catenin expression ranging from 10 to 30% of normal contribute to the pathogenesis of hematological malignant disorders [7,8]. These data suggest that a narrow window of reduced expression of a tumor suppressor is crucial for acute myeloid leukemia and solid tumor development.
RNA interference (RNAi) using either chemically synthesized small interfering RNAs (siRNA) or DNA-based vector systems expressing small hairpin RNAs (shRNA) driven by RNA polymerase (pol) III promoter has been proved to be an efficient method to mediate sequence-specific, post-transcriptional silencing of virtually any gene in various model organisms [9]. The shRNAmediated knockdown using either pol III or pol II promoter has been utilized to knockdown gene expression in mammalian cells and animals in a regulated fashion [10,11]. In combination with a natural backbone of the primary miR-30 microRNA (miRNA), higher amounts of synthetic shRNAs can be produced from the pol III promoter than from the simple hairpin design [12]. This miRNA-based shRNA (mir-shRNA) can also be produced by pol II promoter in cultured cells [13,14], which offers several advantages over the pol III promoter, including simultaneous expression of several miR-shRNAs from a single polycistronic transcript, and regulated or tissue-specific expression [15]. Unfortunately, a heritable and tissue-specific knockdown of gene expression has not yet been developed in animal model organism such as zebrafish, which largely restricts its genetic potential as a vertebrate model of human disorders associated with reduced expression of etiologic genes.
Here, we show that a miRNA-based shRNA (mir-shRNA), when embedded in an intron of b-actin genomic fragment that is in-frame linked to a fluorescent protein-coding reporter gene and placed under the ubiquitous or tissue-specific pol II promoter, is able to efficiently knockdown the expression of chromosomally integrated and endogenous genes in a heritable and tissue-or cellspecific fashion. Cells with reduced expression of targeted genes can be visualized and dynamically traced owing to the expression of the nontoxic actin-tagged fluorescent protein.

Efficient Knockdown of Reporter Gene In Vivo by Mir-shRNA
It has been previously shown that the 59 and 39 flanking sequences of miRNA precursor are crucial for miRNA processing and maturation [16], and the hairpin shRNA can be expressed from a synthetic stem-loop precursor flanked by the 59 and 39 flanking sequences of either human miR-30 [14] or mouse miR-155 gene [13]. We first identified zebrafish homologues of mammalian miR-30 and miR-155 genes based on their sequence identity (data not shown), and cloned both zebrafish pri-miR-30e (409 bp) and pri-miR-155 (447 bp) genomic precursor sequences into the pCS2 + vector ( Figure 1A. and data not shown). Coinjection of in vitro synthesized capped pri-miR-30e mRNAs and sensor EGFP mRNAs containing two tandem perfectly complementary target sites (26PT for miR-30e binding) in its 39UTR (EGFP-26PT mir30e ; Figure 1B) into one-cell stage embryos, resulted in a striking decrease of both EGFP fluorescence and proteins ( Figure 1C, D, right panels). As a control, injection of miR-155 did not have any effects on the expression of EGFP-26PT mir30e ( Figure 1B-D, left panels). Compared with the mir-30e, injection of the capped miR-155 mRNAs showed less efficiency to knockdown the EGFP sensor containing 26PT for miR-155 binding (EGFP-26PT mir155 ) (data not shown). The result suggests that the 409 bp of pri-miR-30e precursor contains flanking sequence capable of directing the production of functionally mature miR-30e in vivo. We therefore used miR-30e precursor as a backbone in the following experiments.
We next replaced the miR-30e stem-loop sequence with a 24 nt-long linker containing two Bbs I restriction sites that allowed insertion of a synthesized shRNA EGFP-ORF stem-loop and preserved all flanking sequences intact ( Figure 2A). The resultant construct mir-shRNA EGFP-ORF contained the same sequence (including a di-nucleotide bugle [17]) as the native miR-30e precursor, except that the strand of the mir-30 hairpin stem has been replaced with the 22 nt-long sequences complementary to EGFP open reading frame (ORF) at the position of 121-142 ( Figure 1A and Figure 2A). Northern blot analysis showed that the mir-shRNA EGFP-ORF mRNAs, when injected into one-cell stage embryo, gave rise to abundant mature shRNA EGFP-ORF fragment in 12 and 24 hours post fertilization (hpf) embryos ( Figure 2B). To test whether the mir-shRNA EGFP-ORF was able to knockdown the EGFP expression, we first microinjected the EGFP-ORF sensor (containing the targeted site within its ORF; Figure 2C, top) with either miR-30e precursor control or mir-shRNA EGFP-ORF into one-cell stage embryos ( Figure 2D). Surprisingly, no obvious knockdown effect was observed in 24 hpf embryos as evidenced by fluorescence and Western blot analysis ( Figure 2E, I, left panels), although the same shRNA EGFP-ORF under H1 promoter was able to efficiently knockdown the EGFP expression in transfected 293T cells ( Figure S1).
We next tested additional sensors EGFP-39UTR-16INS and EGFP-39UTR-26INS, in which one and two copies of the same targeted site were introduced into the SV40-39UTR of EGFP sensors ( Figure 2C, middle). An obvious reduction on both EGFP fluorescence and proteins was observed ( Figure 2F, G; Figure 2I, right panels). The same result was also obtained when the targeted site was inserted into the 39UTR of the DsRed sensor ( Figure 2C, bottom; Figure 2H). In contrast, injection of miR-30e precursor control had no detectable effects on EGFP expression ( Figure 2E-I, left panels). These results suggest that the targeted sites within the 39UTR preferably conferred the knockdown effects by mir-shRNAs, consistent with previous observations [18].

Efficient Knockdown of Endogenous Target Genes by Mir-shRNA
To test whether the mir-shRNA could knockdown endogenous genes in zebrafish, we selected chordin and alpha-catenin that were expressed during early embryogenesis. It has been shown that lossof-function of chordin results in embryonic ventralization with the expansion of mesodermal hematopoietic tissue at the expanse of neuroectodermal development [19]. Significantly reduced expression of alpha-catenin has been observed in the leukemia-initiating cells of del(5q)-associated acute myeloid leukemia/myelodysplastic syndrome and in the invasive solid tumors [8,20].
Because the local secondary structure and the free energy (DG) of 39UTR might affect the accessibility by mir-shRNA [16], we selected two sequences within the 39UTR of chordin gene, which could be potentially targeted by mir-shRNA chordin-39UTR-1 and mir-shRNA chordin-39UTR-2 , respectively ( Figure 3A). These two sequences were selected with mFold software [21] based on the DG of these sites and their flanking sequence (60 bp 59 and 39), which the mir-shRNA chordin-39UTR-1 appeared to have lower DG than mir-shRNA chordin-39UTR-2 ( Figure 3B). The capped mir-shRNA chordin-39UTR-1 and mir-shRNA chordin-39UTR-2 was individually microinjected into one-cell stage embryos and whole-mount in situ hybridization (WISH) analysis with a dig-labeled antisense probe was performed to evaluate the level of chordin transcripts. While the chordin transcripts were appropriately detected in the dorsal shield of wild-type or mir-shRNA EGFP-ORF control-injected embryos at 6 hpf as previously reported [22] ( Figure 3C, left panel, white arrowhead), a dramatic reduction of chordin transcripts was observed in the embryos injected with 200 pg of mir-shRNA chordin-39UTR-1 , but not with the same amount of mir-shRNA chordin-39UTR-2 likely due to its higher DG (3.7 v.s. 0.2 kcal/ mol) ( Figure 3B and 3C, white arrowheads). As a result, an enlarged blood ICM ( Figure 3C, black arrowhead, n = 61/99) with increased gata-1 expression (black arrow) and partial loss of neural tissues (white arrow) were observed only in mir-shRNA chordin-39UTR-1 -injected embryos at 24 hpf, which were comparable to the embryos injected with 0.8 ng of chordin-specific morpholino oligonucleotides [19] ( Figure 3D).
The mir-shRNA a-catenin-39UTR-1 and mir-shRNA a-catenin-39UTR-2 were also designed to target two regions within the 39UTR of alpha-catenin gene ( Figure 4A). WISH analysis showed that the alpha-catenin was maternally expressed (data not shown) and ubiquitously detected in wild type or control mir-shRNA EGFP-ORFinjected embryos at 8 hpf ( Figure 4B, left panel). In contrast, a significant reduction in the alpha-catenin transcripts was consistently observed in the 8 hpf embryos injected with 160 pg of either mir-shRNA a-catenin-39UTR-1 or mir-shRNA a-catenin-39UTR-2 ( Figure 4B, right panels). Consistently, quantitative Western blot analysis showed that the alpha-catenin proteins were dramatically decreased to 26% of normal level at 22 hpf ( Figure 4C). To determine whether the mir-shRNA a-catenin-39UTR-1 can confer gene knockdown in a dosage-dependent fashion, embryos were injected with the same amount of duplex 06, duplex 16 and duplex 26, which harbored zero, one and two copies of shRNA a-catenin-39UTR-1 , respectively ( Figure 4D, top). Northern and Western blot analyses showed that as expected, injection of duplex 26 generated about one-fold more shRNAs ( Figure 4D, bottom) and one-fold less a-catenin proteins than injection of duplexes 16 and control at 22 hpf embryos ( Figure 4D, bottom). Consistently, injection of fourplex 46 also generated one-fold more shRNAs than injection of fourplex 26 ( Figure 4E), suggesting that the a-catenin protein could be further reduced (data not shown). Thus, the experimental design presented here provided not only an efficient means to screen and identify mir- shRNA capable of reducing target gene level, but also a feasible tool to titer down the gene dosage in a controllable manner.

Intronic Mir-shRNA Expression and Genetic Tractability under Pol II Promoter
Natural miRNAs lying within the intron of protein-coding genes have been shown to be co-transcribed with message mRNAs under ubiquitous or tissue-specific pol II promoters [16]. We designed a cytomegalovirus (CMV) prompter-driven expression cassette in which the zebrafish b-actin genomic fragment containing an intact exon 2 (123 base pairs), an intact intron 2 (364 base pairs) and the first 21-base pairs of exon 3, was in-frame fused to the DsRed-Express (DsRed-EX) reporter followed by a bovine growth hormone (BGH) poly (A) site as 39UTR ( Figure 5A). After injection of the plasmid cassette into one-cell stage embryos, the injected embryos showed red fluorescence due to the expression of the chimeric b-actin-DsRed protein, and no any morphological and developmental abnormalities were observed during embryogenesis ( Figure 5B). The precise splicing of intron 2 from the b-actin-DsRed fusion gene in vivo was confirmed by amplification of a predicted size of RT-PCR product and subsequent sequencing ( Figure 5C, D). Furthermore, co-injection of the plasmid expression cassette carrying an introduced mir-shRNA EGFP-ORF or mir-shRNA EGFP-SV40-1 at the Bgl II restriction site within the intron 2, with the EGFP-SV40 reporter plasmid ( Figure 5E), resulted in a dramatic decrease of EGFP fluorescence with correct splicing of the mir-shRNA-containing intron 2 in 22 hpf embryos ( Figure 5F, G). The results provide a proof-ofconcept that knockdown of chromosomally integrated or endogenous genes under a tissue-or lineage-specific pol II promoter might be feasible.

Heritable and Lineage-specific Knockdown of Chromosomally Integrated and Endogenous Genes in the Developing Erythroid Progenitors and Hemangioblasts during Embryogenesis
We previously established a transgenic reporter line Tg(zgata-1:EGFP-SV40) with stable expression of EGFP under the erythroid-specific gata-1 promoter [23]. The transgenic line is unique in that the EGFP expression can be detected in multiple tissues including the midbrain, forebrain, dorsal neurons other than in the erythropoietic ICM, which has also been observed in previous transgenic line with the same gata-1 promoter [24]. Thus, this line offers a unique advantage as a reporter line to detect mir-shRNA mediated knockdown effects in multiple lineages and tissues within an individual animal.
We screened and established 6 transgenic lines stably expressing the intronic mir-shRNA EGFP-SV40-1 under the same gata-1 promoter. One of the lines designated as Tg(zgata-1:mir-shRNA EGFP-SV40-1 -actin-DsRed-BGH) line 1 was selected to determine its knockdown potency because the DsRed fluorescent proteins were also observed to be expressed in the same tissues as the reporter line Tg(zgata-1:EGFP-SV40). When the homozygous Tg(zgata-1:EGFP-SV40) reporter line was crossed to heterozygous Tg(zgata-1:mir-shRNA EGFP-SV40-1 -actin-DsRed-BGH) line 1 ( Figure 6A), 421 (52.9%) and 375 (47.1%) of 796 F2 embryos collected from multiple crosses were DsRed + and DsRed 2 , respectively, suggesting a dominant Mendelian ratio. The EGFP expression in both mRNA and protein levels was dynamically evaluated in the DsRed + and DsRed 2 sibling embryos at 24, 48 and 72 hpf. The results demonstrated a significant reduction of EGFP fluorescence and transcripts in the midbrain (MB), hindbrain (HB), dorsal neurons (DN) and caudal hematopoietic tissue (CHT) only in the DsRed + embryos at 48 and 72 hpf, compared with the appropriate expression of EGFP in the DsRed 2 siblings at the same developmental stages ( Figure 6B, C, arrows; Figure S3). Western blot analysis further confirmed the results that a 45%, 58% and 62% of total EGFP protein was lost in the DsRed + embryos at 24, 48 and 72 hpf, respectively ( Figure 6D). The results indicate that the mir-shRNA EGFP-SV40-1 is able to mediate the cell subtype-specific knockdown of a chromosomally integrated gene in a genetically heritable manner.
To further test the knockdown effects on endogenous genes, we selected an erythroid-specific gata-1 gene to evaluate the knockdown the erythroid-specific gata-1 gene in the developing hemangioblasts. A transgenic line Tg(zlmo2:mir-shRNA gata-1 -actin-DsRed-BGH) line 3 stably expressing the mir-shRNA gata-1 under the hemangioblastic lmo2 promoter [25,26] was established ( Figure 7A). Fluorescence and WISH analyses showed that the DsRed fluorescence and transcripts were specifically expressed in the lmo2-positive vascular endothelial cells and hematopoietic progenitors at the ICM and posterior blood island (PBI) as observed previously [25] ( Figure 7A). As expected, a 50% reduction of gata-1 transcripts was only observed in the PBI of all DsRed + siblings at 22 hpf ( Figure 7B, arrows). More importantly, the pu.1 transcripts (a myeloid progenitor-specific gene) were concomitantly increased in the same region of all DsRed + sibling observed at 22 hpf ( Figure 7C, arrows). The result is consistent with previous observations that reciprocal negative regulation between pu.1 and gata1 determines myeloid versus erythroid fate [27].

Discussion
In this study, we have developed a novel methodology that uses a microRNA-based shRNA (mir-shRNA) to reduce the dosage of a given gene in a controllable and tissue-specific manner. Although the miRNA-based shRNA knockdown strategy has been successfully used to mediate efficient and specific knockdown of genes in vitro, its use in combination with cell-or tissue-specific pol II promoter in animals is still absent. The backbone of miR-30 is one of the most frequently used microRNA sequence to direct the processing and maturation of shRNA, because its stem sequence could be substituted with exogenous sequences that match different target genes and to produce 12 times more mature shRNAs than simple hairpin designs [12,14], and its ability to prevent interferon-stimulated gene expression and associated offtarget effects and toxicity in cultured cells and mouse brain [17,28].
Although the sequences targeted by mir-shRNA in this study are derived from 39UTR, the mir-shRNA should be able to target sequences within other part of a given transcript such as the open reading frame as described previously [12,17]. The observations that the targeted site in the 39UTR, rather than in the ORF of EGFP, confer robust knockdown effects by mir-shRNA EGFP-ORF (Figure 2), suggest that the mir-shRNA might preferably target sequence in the 39UTR in vivo. Interestingly, the similar phenomenon has also been observed in cultured Schneider S2 cells, although the underlying mechanism remains elusive [18]. On the other hand, to optimize the site that mediate maximal knockdown effects, two to three potential target sequences for a given gene should be designed with mFold software and selected based on the predicted secondary structure and DG. Furthermore, as shown in Figure 4D and 4E, taking use of mir-shRNA duplex or fourplex also provides a potential means to maximally knockdown the target genes whose dosage can be regulated in a controllable fashion.
The use of pol II promoter-driven mir-shRNA expression cassette provides a unique advantage in that the cells or tissues with reduced expression of target gene can be genetically traced and visualized in the transparent zebrafish embryos, because of the simultaneous expression of the chimeric red fluorescent protein, bactin-DsRed. Transgenic embryos and adults stably expressing this chimeric fluorescent protein appear morphologically and developmentally normal and have been fertile for three generations, suggesting a lack of detectable toxicity. Given the facts that reduced expression of many disease-associated and developmentally important genes due to either epigenetic inactivation or haploinsufficiency, contribute to the pathogenesis of myeloid malignancies and tumorigenesis [6,8], the methodology described in this study highly complements the recently reported zinc fingermediated gene knockout strategy in zebrafish, and provide an  invaluable tool to knockdown disease-associated gene in specific tissues or cells in the model organisms. In combination with Cre-loxP recombination and drug-inducible strategy [25,29], the pol II promoter-driven mir-shRNA knockdown system could be further optimized to prevent embryonic lethality from reduced expression of the developmentally crucial genes and tumor suppressors.

Fish care
The maintenance, breeding and staging of zebrafish lines (Tubingen and Shanghai) were performed as described previously [30].

Cloning and plasmid construction
The precursor sequences of zebrafish mir30e (409 bp) and mir155 (447 bp) were cloned from the genomic DNA of Tubingen adult fish into the pCS2 + vector. The 68-bp mir30e stem-loop region was replaced with a linker sequence containing two Bbs I sites using two-step PCR. The shRNA sequences were synthesized as DNA oligonucleotides (Invitrogen) and inserted at the Bbs I sites. The sensor sequences (26PT) were synthesized as DNA oligonucleotides (Invitrogen) and placed into the 39UTR of the pCS2 + -EGFP plasmid. The genomic sequence of zebrafish b-actin containing an intact exon 2 and intron 2, and the first 21-bp of exon 3 was in-frame infused to the open reading frame of DsRed-39UTR BGH, and cloned into the pCS2 + plasmid through BamH I and EcoR I sites. The mir-shRNA was inserted into an endogenous Bgl II site within the intron 2. The resultant intronic mir-shRNA was then inserted downstream of the gata-1 promoter or lmo2 promoter and cloned into the I-SceI-containing plasmid. All primer sequences were available in Table S1.

Microinjection and establishment of transgenic zebrafish line
All capped mRNAs were synthesized with SP6 mMessage mMachine (Ambion) and microinjected into one-cell stage embryos. The transgenic plasmids flanked by the I-SceI sites were prepared with endotoxin-free miniprep kit (Axygen). Microinjec- tion was performed at one-cell stage embryos with 2 nl of injection solution containing 40 pg/nl of DNA, 0.56I-SceI buffers and 0.5 units/ml I-SceI meganuclease (New England Biolabs). Injected embryos were raised to sexual maturity (F0 founders) and crossed to wild-type zebrafish to generate F1 progeny, which were screened for red fluorescent DsRed expression in the ICM at 24 hpf. The DsRed + F1 embryos were raised to adults to establish the stable transgenic lines. Embryos were imaged using a Zeiss SteREO Discovery V12 fluorescent stereomicroscope.
shRNA Northern blot analysis and RT-PCR Total RNAs of embryos injected with capped mRNAs were extracted with Trizol (Invitrogen), and separated on 12% of UREA-PAGE gel. Northern blot was probed with dig-labeled antisense probe, and visualized using DIG luminescent detection kit for nucleic acids (Roche). RT-PCR was performed with Onestep RT-PCR kit (Qiagen) as previously described [30].

Whole-mount mRNA in situ hybridization
Whole-mount mRNA in situ hybridization was performed as described previously [30]. Dig-labeled antisense probes of a-catenin and chordin were generated from a 975-nt cDNA fragment encoding N-terminal 325 aa of a-catenin, and entire 570 bp 39UTR, respectively.

Cell culture and transfection
HEK293T cells were cultured in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% calf serum in an atmosphere containing 5% CO2. H 1 pol III promoter-driven shRNA and EGFP reporter (pEGFP-C1, Clontech) and DsRed plasmid were cotransfected with a ratio (20:1:2) into the HEK293T cells using the calcium phosphate method.