|Home | About | Journals | Submit | Contact Us | Français|
Synthetic and nonnatural oligonucleotides have been used extensively to interrogate gene function in zebrafish. In this review, we survey the capabilities and limitations of various oligonucleotide-based technologies for perturbing RNA function and tracking RNA expression. We also examine recent strategies for achieving spatiotemporal control of oligonucleotide function, particularly light-gated technologies that exploit the optical transparency of zebrafish embryos.
Chemical tools are increasingly being used to interrogate the mechanisms of embryonic development, augmenting more traditional genetic and surgical approaches. Molecular probes can facilitate the perturbation and detection of embryonic gene function,1 and these structurally diverse reagents can be divided into two major classes: small molecules and oligonucleotide-based reagents. As described elsewhere in this special issue of Zebrafish, the former has been invaluable for studying cell signaling dynamics during zebrafish development. Small molecules can alter protein function in a direct, rapid, and reversible manner, exemplified by the use of cyclopamine and dorsomorphin to block zebrafish Hedgehog and bone morphogenetic protein signaling, respectively.2,3 Yet, extrapolating the efficacy and pathway selectivity of small molecules to other embryonic processes is not always straightforward. Small molecule–protein interactions require complementary molecular shape, polarity, and electrostatics, and identifying new chemical modulators of signaling proteins frequently relies upon high-throughput screens and serendipity rather than molecular design.
In light of these challenges, oligonucleotide-based reagents provide an alternative strategy for controlling or observing gene function, exploiting the use of Watson–Crick base pairing by DNA and RNA molecules. The simple hydrogen bonding architecture that mediates adenine-thymine/uracil and guanine-cytosine interactions allows for the design of synthetic or nonnatural oligomers that hijack the Nature's genetic machinery in a sequence-specific manner. Moreover, a variety of artificial oligomeric structures can be accommodated since the sugar-phosphate backbone in DNA and RNA is not directly involved in the molecular recognition of their nucleic acid bases. Guided by these basic principles, several oligonucleotide-based strategies for studying the genetic mechanisms of embryonic development have been created, many of which are efficacious in zebrafish. These approaches are primarily used to target RNA molecules, which are generally more accessible than their double-stranded DNA counterparts, and depending on the targeted sequence, the oligonucleotide reagents can inhibit RNA splicing, processing, or translation with genome-wide specificity. Oligonucleotide-based methods can also be used to detect specific RNA transcripts in whole embryos, and collectively these tools have yielded critical insights into how genetic programs dynamically regulate embryogenesis.
In this review we survey the use of synthetic and nonnatural oligonucleotides to investigate gene function in zebrafish. We examine the capabilities of each method, as well as recent strategies for implementing oligonucleotide-based reagents in a tissue-specific manner. We also discuss the limitations of current oligonucleotide technologies and future challenges for the chemical biology and zebrafish communities.
RNA molecules both encode and regulate protein expression, and several steps in this process can be modulated by synthetic and nonnatural oligonucleotides. The precise activity of a given oligomer depends in part on the targeted region within the endogenous RNA molecule (Fig. 1). Maternal and zygotic mRNAs can be targeted by antisense approaches that either promote enzymatic transcript degradation or sterically block the transcriptional start site and flanking 5′ untranslated regions (UTRs). Alternatively, antisense oligomers can selectively target zygotic mRNAs by base-pairing to splice junctions and disrupting mRNA processing. This latter approach can be used to insert introns or skip exons, frequently resulting in a reading-frame shift and introduction of a premature stop codon.4 In principle, splice-blocking morpholinos (MOs) could also be used to selectively remove protein domains by skipping exons in-frame, leading to gain-of-function or dominant-negative gene products. Noncoding RNAs such as microRNAs (miRNAs) can similarly be perturbed by synthetic oligomers, which can hybridize to pri-miRNA or pre-miRNA excision sites and inhibit their processing, or to the mature miRNA itself. Alternatively, miRNA function can be modulated by directing oligonucleotides against noncoding RNA-binding sites in the 3′ UTRs of targeted transcripts.
The chemical structure of the exogenous oligonucleotide also influences the mechanism by which it abrogates RNA function and the perdurance of the oligomer in vivo. Oligonucleotides that replicate or closely mimic natural nucleic acids can trigger RNA-degrading enzymes, while structurally divergent oligomers typically act as a steric block of endogenous RNA-interacting molecules. During the past two decades, nucleic acid chemists have created a diverse array of oligonucleotide-based technologies, each with unique biophysical and biological properties (Fig. 2). Certain oligomers utilize natural nucleic acids, such as synthetic small interfering RNAs (siRNAs), and their genetically encodable precursors, short hairpin RNAs (shRNAs). Others incorporate chemical modifications of varying complexity. Synthetic oligonucleotides with altered sugar structures include 2′ O-methyl RNA, 2′ fluoro RNA (FNA), 2′ methoxyethyl RNA, and locked nucleic acids (LNAs).5 The backbone phosphate esters can also be changed to phosphorothioates. In fact, the entire sugar–phosphate backbone can be replaced with a nonnatural scaffold, as exemplified by MOs and peptide nucleic acids (PNAs).6,7 In this section we summarize how exogenous nucleic acids have been used to modulate RNA function in zebrafish.
MO oligonucleotides are the most commonly used antisense reagents in zebrafish models, and hundreds of MOs have been reported as tools for studying teleost development. These synthetic oligomers utilize a morpholine–phosphorodiamidate backbone rather than one composed of repeating sugars and phosphodiesters, increasing their stability in vivo while maintaining their ability to target endogenous RNAs through Watson–Crick base pairs. Their widespread application by multiple laboratories provides an extensive account of their efficacy and limitations as functional genomic probes, and the lessons learned from MO-based approaches can be applied to other antisense technologies.
Unlike siRNAs, phosphorothioate DNA, and other oligomers that are closely related to endogenous nucleic acids, MOs do not cause RNA degradation through the recruitment of an RNA-induced silencing complex (RISC) or RNAse H. Instead, the MO/RNA heteroduplex is rendered sterically unable to interact with RNA-binding proteins.8 MOs are generally designed to block RNA translation by hybridizing to the start site and flanking 5′ UTR9 or to generate nonfunctional splice variants by targeting intron–exon junctions.4 More recent applications of MOs have exploited the role of noncoding RNAs in gene regulation, and these reagents have been used to block miRNA binding sites in mRNAs, to prevent pri- and pre-miRNA processing, and to sequester mature miRNAs from RISC.10
The use of MOs in zebrafish has been detailed in two recent reviews,11,12 and we summarize the basic principles here. MOs are typically used as 25-base oligomers and at doses of 50–1000fmol per zebrafish embryo. Although they can be injected directly into the animal cell at the one-cell stage as is typically done with plasmid DNA or mRNA, MOs can also be delivered to the yolk underlying the blastomeres before the eight-cell stage, since the nonionic molecules are readily taken up by the animal cells via cytoplasmic bridges.9 MO perdurance in vivo is believed to be primarily dilution-limited,12 and the MO concentration in each cell during embryogenesis will depend on the number of preceding cell divisions and cell size. In practice, MOs can maintain their efficacy for 3 days and in some cases persist for up to 5 days, allowing gene function at larval stages to be explored.
As with any specific binding interaction, MO activity is dose dependent, and the reagents can be used to study phenotypes associated with a range of target gene expression levels. Previous studies have quantified MO activity in vivo through the use of green fluorescent protein (GFP)9 or firefly luciferase13 reporters. For example, an MO directed against the 5′ UTR of GFP mRNA was able to block protein expression with an IC50 of 3.5μM, which corresponds to a binding free energy value of approximately −7.5kcal/mol. A more potent MO targeting no tail-a (ntla) inhibited endogenous Ntla protein expression with an IC50 of 0.48μM (binding free energy of −8.7kcal/mol).14 These apparent binding energies do not reflect the actual dissociation constants of MO/mRNA duplexes in zebrafish embryos but instead are aggregate descriptors of MO/RNA affinity, RNA accessibility, and MO interactions with other cellular components. For comparison, the ntla MO/RNA duplex has an in vitro binding free energy value of −28.1kcal/mol—an apparent affinity that is 1014 times stronger. This significant difference between in vitro and in vivo MO/RNA interaction strengths and the broad range of MO IC50s in zebrafish embryos underscore the complexity of MO function in live organisms. Since computational algorithms that reliably delineate accessible regions in targeted RNAs are lacking, obtaining effective MOs against specific genes remains largely an exercise in trial and error. However, once an effective MO has been identified, the in vivo activities of these synthetic oligonucleotides can be modeled as simple two-state thermodynamic equilibrium, allowing one to correlate MO doses with specific protein expression levels.14
Another challenge that has emerged with the use of MOs is the importance of confirming that the resulting phenotypes are due to specific gene silencing rather than off-target effects. The most common side effect of MO injection is p53 activation, leading to apoptosis predominantly in anterior neural tissues and the notochord.15 The mechanism by which MOs activate p53 remains unclear, and approximately 20% of all MOs induce this off-target phenotype. Unfortunately, predicting which MOs will cause p53-mediated apoptosis is not yet possible since this activity does not correlate with a known consensus sequence or secondary structure. These off-target effects can be mitigated, however, by coinjecting a p53 MO with the synthetic oligonucleotide targeting the gene of interest.15 Since p53 is not required for the normal development of fish and mammals, the coinjected p53 MO is generally well tolerated by zebrafish embryos.
In addition to eliminating nonspecific MO toxicity, the on-target effect of MOs must be carefully evaluated and validated. Although 25-base MOs are designed to perfectly complement the targeted RNA sequence, these reagents can still functionally interact with sequences containing up to four base mismatches or with at least 15 contiguous complementary bases.13 Theoretical calculations predict that one out of every two MOs will bind, although weakly, to an off-target translational start site or intron–exon junction.11 However, the chance of two nonoverlapping MOs producing the same off-target phenotype is very low. These considerations have led to recommended guidelines for correlating MO-induced phenotypes with on-target MO activities.11 At least two nonoverlapping MOs designed against the gene of interest—typically a translational-start-site MO and a splice-junction MO—should induce similar phenotypes, and coinjection of these reagents should have a synergistic effect.13 Indeed, combining two MOs at low to intermediate doses is preferable to using a single MO at high concentrations, since the latter condition increases the likelihood of off-target interactions. Target gene knockdown in MO-injected embryos should also be independently confirmed. This can be achieved through specific antibodies when they are available, or in the case of splice-junction-targeting MOs, the mis-spliced RNA can be detected by RT-PCR and sequence-verified. Mis-splicing can also be detected by in situ hybridization, as inappropriately spliced transcripts are typically sequestered in the nucleus.16 Finally, MO-induced phenotypes can be validated by rescuing them with MO-resistant mRNA that encodes the gene of interest, although in practice this can be complicated by developmental defects associated with global gene overexpression.
Although the development of PNAs was contemporaneous to that of MOs, these polyamide oligomers are less commonly used to alter RNA function in zebrafish embryos. This is due in part to the relative insolubility of PNAs composed of the standard pseudopeptide backbone, which makes it difficult to achieve functional PNA concentrations in vivo. As a result, hydrophilic, negatively charged PNA analogs (ncPNAs; also known commercially as gripNAs™) that contain an alternating trans-4-hydroxyl-L-proline/(2-aminoethylamino)methyl-phosphonate scaffold have been developed as an alternative technology for inhibiting zebrafish gene expression.17
Like MOs, ncPNAs can be delivered by microinjection into the yolk of zebrafish embryos and exhibit similar tissue distributions. Since ncPNAs hybridize to RNA more tightly than MOs, 18-base pair ncPNA/RNA heteroduplexes are energetically equivalent to the standard 25-base pair MO/RNA complex, and it has been observed that ncPNAs and MOs targeting the same 5′ UTR mRNA sequence inhibit translation at similar doses.17 Perhaps reflecting these differing optimum oligomer lengths, the two synthetic oligonucleotides exhibit distinct tolerances for base pair mismatches. While an ntla ncPNA produces a mutant phenotype in 96% of injected embryos, a single mismatch in the ncPNA sequence reduces phenotypic penetrance to 30%.17 The same study also demonstrated that a chordin (chd) ncPNA could be rendered phenotypically inactive with just two base pair mismatches, contrasting the ability of an equivalently mismatched chd MO to induce mild to medium loss-of-function chd phenotypes in nearly all injected embryos.
In addition to this greater differential between on-target and off-target efficacy, studies suggest that ncPNAs and MOs targeting the same RNA region may not necessarily produce the same p53-dependent toxicity. For example, an ncPNA targeting the dharma translational start site can effectively phenocopy the genetic mutant, whereas an MO targeting the same region produces severe necrosis in the head and a shortened, degenerated body axis.17 The ncPNA reagent is therefore preferred for inhibiting dharma expression.18 This is not to say that ncPNAs are free of deleterious off-target effects, especially at higher doses,17 as p53-mediated apoptosis has been observed with a splice-blocking ncPNA targeting wnt5b.15 The two synthetic oligonucleotide classes should therefore be viewed as complementary reagents that can be used to achieve specific RNA perturbations in whole organisms. Yet in spite of its potential as a research tool, relatively few ncPNAs targeting zebrafish genes have been published. Since all of these reported ncPNAs have targeted genes expressed early in development, it is possible that these amide bond-containing synthetic oligonucleotides have lower in vivo stability than MOs, which utilize a phosphorodiamidate backbone. Consistent with this idea, ncPNA and MO reagents targeting the same region in microphthalmia-associated transcription factor-a (mitfa; also known as nacre) exhibit divergent activities, with only the MO oligonucleotide successfully blocking the expression of this later-acting gene (S.C. Ekker, personal communication). Nevertheless, it would be constructive for the zebrafish community to further explore the utility of ncPNAs and related reagents as reverse-genetic tools.
While the zebrafish community has predominantly utilized MOs and ncPNAs to modulate RNA function in vivo, the pharmaceutical industry has largely focused on synthetic oligonucleotides that incorporate sugar or phosphodiester analogs rather than a complete replacement of the sugar-phosphate backbone. Structural changes at the 2′ hydroxyl position of the ribose ring, including the ring-locking modification in LNAs, can improve oligomer base-pairing affinity by promoting a C3′-endo ribose conformation,19 and alkylation of the 2′ hydroxyl mimics the posttranscriptional mRNA capping mechanism, affording protection from endonucleases.19 Oligomer stability in vivo can also be increased by substitution of the phosphate linkages with phosphorothioates.5 For example, the U.S. Food and Drug Administration–approved drug fomivirsen (also known by the brand name Vitravene™) is a 21-base phosphorothioate oligonucleotide that mitigates cytomegalovirus (CMV) retinitis by inhibiting the translation of viral mRNA.20 Cholesterol-conjugated oligomers that contain both 2′ O-methyl and phosphorothioate modifications have also been delivered into mice by intravenous injection and successfully blocked endogenous miRNA function.21
Unfortunately, these commonly used synthetic oligonucleotides appear to be toxic to zebrafish embryos, as both phosphorothioate DNA and 2′ O-methyl RNA induce nonspecific cell death at a dose of approximately 500fmol/embryo.22 These effects are similar to the toxicity associated with short single-stranded DNAs, suggesting that these oligonucleotide analogs might irreversibly trigger DNA damage checkpoint responses and cell cycle arrest. Whether lower doses of modified RNA oligomers can be functionally efficacious in zebrafish remains to be established. A 2′ O-methyl RNA dose of 50fmol/embryo has been used to inhibit coinjected let7 miRNA,23 but higher oligomer doses sufficient to alter endogenous miRNA function were not tolerated.10 The efficacy of LNAs in zebrafish has not been thoroughly examined, although our preliminary attempts to knockdown gene expression with these reagents have been unsuccessful (A.J. Firestone and J.K. Chen, unpublished results).
Despite these challenges, the use of sugar or phosphodiester analogs in oligonucleotide-based tools should not be discounted by the zebrafish community. Phosphorothioates and 2′ O-methyl ribose systems are not the only useful oligonucleotide modifications. A recent study reported that a 21-base pair FNA duplex (siFNA) was tolerated by zebrafish embryos at doses that could inhibit expression of a coinjected GFP plasmid (approximately 30fmol/embryo).24 As nucleic acids research continues to yield oligonucleotides with novel structures, the zebrafish community's chemical toolbox might finally include alternatives to MOs and ncPNAs that are more potent, less toxic, easier to synthesize, and less expensive.
The discovery in the 1990s that double-stranded RNA can activate an endogenous gene-silencing mechanism called RNA interference has transformed functional genomic studies in many biological systems.25 RNA interference is now the preferred gene-silencing method for cultured cells, and it is widely applied to interrogate gene function in fruit flies, nematodes, and planaria. The endonuclease Dicer processes the exogenous double-stranded RNAs into siRNAs, which are loaded onto the multicomponent RISC.26 Alternatively, siRNAs can be delivered directly into cells or model organisms for RISC incorporation. siRNAs can also be generated in situ by expression of shRNAs, as these hairpin oligonucleotides are Dicer substrates as well. RISC then uses one siRNA strand to recognize and degrade complementary RNAs.
Since zebrafish embryos utilize Dicer to generate endogenous miRNAs,27 one might expect siRNAs to be effective zebrafish gene-silencing reagents. However, the use of RNA interference to study zebrafish development has been problematic. Severe developmental abnormalities and growth arrest after the mid-blastula transition were observed in embryos injected with just 7.5pg of double-stranded RNA against a GFP transgene.28 In another study, exogenous double-stranded RNAs against tbx16 spadetail, dharma, and ntla actually induced global RNA degradation29 in manner reminiscent of the interferon response induced in mammalian cells exposed to double-stranded RNAs greater than 23 base pairs in length.30 A few research groups have reported the use of double-stranded RNA to induce mutant phenotypes in zebrafish, but the molecular specificity of these perturbations has not been extensively explored.31–33 For reasons that are not clear, even siRNAs can produce nonspecific patterning defects in zebrafish, as was observed with 50fmol/embryo doses of siRNAs against lamin A, lamin B2, or kinesin family member 11 (kif11).34 These general defects were reduced at lower siRNA doses, but no specific phenotypes were achieved. Yet, the same siRNAs were able to yield the expected loss-of-function phenotypes in ZFL, SJD, and ZF4 zebrafish cell lines,34 indicating that siRNA off-target effects may be cell-type specific.
Despite these discouraging and confounding results, RNA interference might be achievable in zebrafish through the use of shRNA technologies. shRNAs appear to circumvent the toxicity caused by microinjection of double-stranded RNAs or siRNAs, at least when expressed through microinjected plasmids or stably integrated transgenes. Constitutive shRNA expression through the use of U635 or CMV36 promoters induced a modest knockdown of GFP, Ntla, and VEGF-A protein levels at plasmid doses that did not cause toxic off-target effects.37 Driving shRNA expression through a hybrid U6/CMV promoter can enhance RNA interference efficiency, reducing Ntla protein to approximately 20% of wild-type levels and producing a ntla mutant phenotype in about half of the injected embryos.38 Transgenic zebrafish that stably express shRNAs through a pri-miR-30-derived stem-loop precursor have also been generated.39 In this system, the miR-shRNA is embedded into an intron from β-actin and driven by an RNA polymerase II-responsive promoter, either ubiquitously or in a tissue-specific manner. For example, miR-shRNAs targeting gata1a have been stably expressed in developing hemangioblasts, resulting in a 50% reduction of gata1a mRNA levels and hematopoietic reprogramming.39 Realizing the full potential of shRNAs in zebrafish will require further optimization of these approaches, as they do not yet approach the gene-silencing efficacies of MOs or ncPNAs.
As discussed above, MOs and ncPNAs are generally microinjected into the yolk of zebrafish embryos before the eight-cell stage, after which they are distributed uniformly among the animal cells. These reagents therefore constitutively perturb RNA function throughout the embryo, frequently producing phenotypes that are comparable to a zygotic loss-of-function mutant. Conditional activation of MOs and ncPNAs would permit spatiotemporal control of RNA-dependent processes, allowing functional genomic analyses at specific developmental stages and/or individual tissues. Toward that goal, a number of oligonucleotide caging strategies have been developed and applied to zebrafish (Fig. 3). These approaches take advantage of the ex utero development and optical transparency of zebrafish embryos, attributes that make them particularly amenable to photochemical manipulations. In this section we summarize current technologies for the photoactivation of exogenous oligonucleotides in vivo. Each method relies upon a distinct set of structural modifications to block oligonucleotide activity, and a variety of light-sensitive caging groups have been employed.
One of the first reported applications of caged oligonucleotides in zebrafish involved derivatization of the phosphodiester backbone of cDNAs and mRNAs (Fig. 3A).40 In principle, the addition of photocleavable chromophores to these phosphate groups would abrogate interactions with transcriptional or translational machinery, preventing expression of the encoded genes after microinjection of the caged oligonucleotides into zebrafish embryos. Irradiation of the injected embryos would then restore oligonucleotide function. This method is synthetically appealing because the oligomers can be generated in vitro and then caged. For instance, to make light-activatable GFP, β-galactosidase, and eng2a mRNA, the corresponding in vitro transcription products were reacted with 6-bromo-4-diazomethyl-7-hydroxycoumarin (Bhc-diazo), resulting in random alkylation of the backbone phosphodiesters.40
In practice, however, this caging strategy has significant limitations. First, the stochastic nature of the caging reaction results in a heterogenous population of oligonucleotides with varying degrees of modification. To minimize the fraction of incompletely caged plasmid DNA or mRNA and achieve acceptable levels of basal activity, an average of 1 in every 30 phosphodiester groups must be caged.40 As a result, each oligonucleotide molecule contains tens if not hundreds of Bhc groups. Re-activating these caged oligonucleotides through irradiation consequently requires the removal of multiple chromophores, a low-probability event, since individual photochemical reactions rarely achieve 100% yields. Second, modification of the phosphodiester groups with Bhc-diazo produces a phosphotriester bond, which is susceptible to hydrolysis or, in the case of RNAs, nucleophilic attack by the 2′ hydroxyl of the ribose sugar. Either reaction will degrade the caged oligonucleotide. Consistent with these limitations, irradiating the caged GFP, β-galactosidase, and eng2a mRNAs in vivo with long-wavelength ultraviolet light (365nm) restores only a fraction of oligonucleotide activity (in the case of caged β-galactosidase mRNA, this was quantified to be less than 20% of the activity observed with unmodified transcripts).40
Subsequent studies have yielded similar findings, illustrating the general limitations of this approach.41 Perhaps more promising results have been reported for an FNA-based siRNA equivalent (siFNA) that was chemically modified with 1-(4,5-dimethoxy-2-nitrophenyl)-diazoethane.24 The 2′ fluoro modifications in the caged siFNAs preclude 2′ hydroxyl-mediated hydrolysis of the caged phosphotriester, enhancing the stability of these oligonucleotides. As has been observed with regular siRNAs, siFNAs caused nonspecific embryonic toxicity at doses of 50fmol/embryo or higher. Interestingly, caged siFNA reagents were not toxic at the same concentration.24
Since the molecular recognition of DNA and RNA is primarily mediated through Watson–Crick base pairing, derivatizing individual nucleic acid bases with photocleavable groups is another strategy for caging oligonucleotide function (Fig. 3B). Unlike the phosphodiester modifications described above, conjugating chromophores to pyrimidines and purines is best conducted on nucleoside monomers, before solid-phase synthesis of the oligonucleotide chain. (6-Nitropiperonyloxylmethyl)-caged thymidine42 is now commercially available for this purpose, and these photoactivatable nucleosides have been incorporated into phosphorothioate DNA to regulate gene expression in cultured cells.43 In this application, three caged thymidines were required to prevent an 18-base phosphorothioate oligomer from inducing the RNAse H-dependent degradation of Renilla luciferase mRNA in transfected cells. Irradiation of the cells with 365-nm light restored the activity of the transfected oligonucleotide to levels approaching that of the uncaged oligomer.43
The application of this approach in zebrafish has not yet been published, but in principle caged thymine bases could be incorporated into MO or ncPNA oligonucleotides. In both cases it will be critical to achieve an appropriate balance between modifying enough thymine-containing monomers to sufficiently inhibit oligomer activity and limiting the number of caging groups to promote photoactivation efficiency. It may also be advantageous to explore caged versions of adenine, guanine, and cytosine. Since ncPNAs appear to be more sensitive to base-pair mismatches than MOs, these anionic oligonucleotides may be better suited for photoactivation strategies that utilize caged bases.
The requirement for multiple backbone or base modifications to completely abrogate oligonucleotide activity may limit the dynamic range of these reagents, since the overall efficiency of photoactivation decreases with each additional caging group. Gating oligonucleotide activity with a single light-sensitive chromophore would circumvent this issue; however, regulating the extensive hydrogen-bonding network of an 18- to 25-base oligomer with an individual caging group is an equally challenging proposition.
One strategy for controlling oligonucleotide hybridization with a single photochemical reaction is to conjugate the RNA-targeting oligomer to a complementary sequence through a photocleavable linker (Fig. 3C). The resulting molecule forms a hairpin through intramolecular base pairing, abrogating its ability to bind to RNA targets. Linker photolysis detaches the complementary inhibitor from the targeting oligonucleotide, resulting in a significantly weaker intermolecular base-pairing interaction and causing oligonucleotide/RNA hybridization to be energetically favored.
This approach has been implemented in zebrafish through the synthesis of caged MO hairpins directed against the ntla gene. Using this reagent, ntla silencing can be achieved at different developmental stages and in specific tissues, revealing multiple roles of this transcription factor in mesoderm development.44 Subsequent studies have refined the technology by simplifying procedures for caged MO synthesis.14 Caged MO hairpins can now be assembled from two commercially available MO oligomers and a readily synthesized bifunctional linker. Depending on the chromophore incorporated into the crosslinker, caged MO activation can be achieved with conventional 365-nm light sources or with 820-nm light using two-photon irradiation. Basic guidelines for caged MO design have also been established by synthesizing caged ntla MOs with divergent structures and quantitatively comparing their in vitro binding affinities and in vivo gene-silencing activities.14 Through these design criteria, caged MOs that undergo at least a 30°C decrease in oligomer/inhibitor duplex melting temperature upon linker photolysis can be achieved, ensuring an efficacious dynamic range of activity. The generality of this approach has been further validated by the application of these algorithms to caged MOs against floating head, heart of glass, and ets varient gene 2.14
Caged ncPNA hairpins have been developed as well, in this case by conjugating 18-base ncPNA oligonucleotides to complementary 2′ O-methyl RNA sequences of various lengths through a nitrobenzyl-based linker.45 These reagents have been used to induce light-dependent chd and dharma mutant phenotypes in zebrafish embryos, suggesting that they act in a manner similar to that of caged MOs. It is perhaps surprising, however, that the photoreleased 2′ O-methyl RNA is not associated with nonspecific developmental defects, as this synthetic oligonucleotide has been found to be toxic to zebrafish embryos.22 General criteria for designing caged ncPNAs against other genes have not yet been established by quantitative studies, but the lessons learned from caged MO hairpins should be extrapolatable to these reagents.
Caged oligonucleotide hairpins do have certain limitations. The inclusion of an inhibitory MO or 2′ O-methyl RNA oligomer may increase the chance of off-target effects, especially after linker photolysis. In addition, designing an appropriate inhibitory oligomer is a compromise between maximizing intramolecular duplex formation and minimizing intermolecular hybridization upon photoactivation of the reagent. This balance will be increasingly difficult to achieve as MO or ncPNA potency decreases, since the higher reagent concentrations required to obtain the desired mutant phenotype will promote binding of the targeting oligonucleotide to the photoreleased inhibitor.
Although the efficacy of caged MO and ncPNA hairpins in zebrafish embryos has been established, they are not yet commercially available and therefore must be manually synthesized. An alternative methodology that utilizes a single photocleavable moiety has been recently developed and marketed, exploiting the relative ease with which RNA can be chemically synthesized. In this strategy, the 25-base targeting MO oligonucleotide is hybridized to an RNA oligomer composed of two 12-base fragments separated by a short nitrobenzyl-based linker, generating a heteroduplex called a PhotoMorph™ (Fig. 3D).22 Strand exchange between the PhotoMorph and endogenous RNAs is energetically disfavored, presumably because of the relative inaccessibility of the latter binding partner. Photocleavage of the exogenous RNA oligomer, however, generates two 12-base fragments that are unable to effectively block MO binding to its RNA target, and PhotoMorphs have been used to conditionally modulate GFP, Ntla, E-cadherin, and Rheb expression in zebrafish embryos.22
As with the previous caging strategies, the PhotoMorph approach has both strengths and weaknesses. The synthetic simplicity and sole utilization of commercially available reagents is appealing, but the reliance of this methodology on an RNA-based inhibitor compromises reagent stability. Unfortunately, caging MO activity by intermolecularly hybridizing the oligomer to endonuclease-resistant molecules such as 2′ O-methyl RNA and phosphorothioate DNA was found to be ineffective because these latter oligonucleotides exhibited significant embryonic toxicity.22 Although the PhotoMorph RNA does not produce these off-target effects, its gradual degradation in vivo leads to a time-dependent increase in MO activity, and an empirically determined excess of PhotoMorph RNA—typically five to ten molar equivalents—must be coinjected with the MO to completely cage its activity. Technically this creates a situation similar to having multiple caging groups per oligonucleotide, since more than one PhotoMorph RNA molecule must be photolyzed to uncage each MO oligomer. Consistent with this mechanistic description, MO activity is not fully restored upon 365-nm irradiation when sufficient quantities of PhotoMorph RNA are used to completely block MO function.22
Although the application of synthetic oligonucleotides in zebrafish research has primarily focused on altering RNA function, nucleic acid derivatives have also been used to observe the expression patterns of RNA molecules. For example, mRNA in situ hybridization in paraformaldeyde-fixed tissues is a technique familiar to the zebrafish community and developmental biologists in general,46 and this method typically uses in vitro transcribed antisense RNA probes that contain digoxigenin-modified uridine nucleosides. Once hybridized to its mRNA target, the chemically modified RNA probes can be detected with antidigoxigenin antibodies conjugated to a reporter enzyme, such as alkaline phosphatase, and chromogenic substrates. Distinct transcripts can also be observed simultaneously by costaining with orthogonally labeled RNA probes.47 In situ hybridization reagents containing fluorescein-modified uridine are also commonly used for this purpose, whereas biotin-labeled RNA probes are less effective in zebrafish since they are associated with high background signals.47 New methods for simultaneously detecting two or more RNAs in zebrafish embryos will likely be developed, fostered by the versatility of azide-alkyne coupling reactions (also known as “click chemistry”)48 and the ongoing development of alkyne-derivatized nucleotides.49
The emerging roles of noncoding RNAs in vertebrate development and physiology have also spurred efforts to detect miRNA expression in zebrafish embryos and larvae. The short length of these RNAs (typically 16–18 bases) renders them undetectable by standard RNA probes since these duplexes readily dissociate. The remarkable affinity of LNA-containing oligomers for RNA (each LNA substitution of an RNA nucleoside within double-stranded RNA increases the duplex melting temperature by approximately 5°C)50 provides a convenient solution to this problem, and digoxigenin-labeled LNA probes have now been used to detect over 100 different zebrafish miRNAs.51,52 As new nucleic acid derivatives with greater RNA-binding affinity and in vivo stability are developed, it may even be possible to achieve the detection of mRNAs and miRNAs in live zebrafish rather than in fixed tissues.
Advances in nucleic acid chemistry have played a critical role in establishing the zebrafish as a model organism for biomedical research. Synthetic oligonucleotides have provided key insights into the functions of individual genes during zebrafish development, especially in the absence of homologous recombination technologies for targeted gene knockdowns. Nonnatural oligomers have also helped reveal dynamic expression of these genes in embryonic and adult tissues. The recent development of several caging strategies promises to further extend these research capabilities by allowing scientists to ask not only what specific genes do, but also when and where.
As zebrafish are increasingly used to model vertebrate development, physiology, and disease, the demand for new oligonucleotide-based tools will continue to grow. Reagents that overcome bottlenecks associated with existing technologies are particularly needed. For example, current antisense technologies utilize membrane-impermeant oligomers that must be microinjected, electroporated, or transfected into zebrafish cells. Methods that can deliver these reagents to multiple embryos uniformly and simultaneously, ideally by just adding the oligonucleotides to the culture medium, would enable the zebrafish community to use these chemical tools more extensively in their research. Large numbers of embryos could be rapidly subjected to specific genetic perturbations, and loss of gene expression at different embryonic and adult stages could be readily achieved. Toward this goal, several technologies for the intracellular delivery of oligonucleotides have been described,53 including cell-targeting ligands (cholesterol and oligonucleotide aptamers), cell-penetrating peptides (polyarginine and peptides derived from the Drosophila Antennapedia and HIV-1 Tat proteins), and nanoparticles. In particular, functional MO delivery into mouse tissues has been achieved via intravenous injection of MOs covalently conjugated to polyarginine peptides or guanidinium dendrimers.54 The latter reagents, commonly known as Vivo-Morpholinos™, are commercially available and have been recently used to deliver a splice-junction-targeting MO into adult zebrafish thrombocytes, although with relatively low efficacy.55 The utility of these approaches in zebrafish models remains to be established, and significant technical hurdles remain. For example, current cell-penetrating technologies may not be able to achieve uniform oligonucleotide distributions in organismal tissues due to various biological barriers.56
Developing new methods for the conditional control of oligonucleotide function is another fertile area for future research. While caged MOs and ncPNAs illustrate the potential of conditional synthetic reagents, there are certain drawbacks to current technologies. Most caging strategies rely on chemical groups that are sensitive to ultraviolet light, which can induce DNA damage and has limited tissue penetrance. Next-generation chromophores that can undergo photochemical reactions at lower doses of ultraviolet light, are sensitive to longer wavelengths, or have greater cross sections for two-photon irradiation could help mitigate these issues. It may even be possible to develop chromophores that enable the orthogonal activation of two or more oligonucleotides using different wavelengths of light.
Other approaches that should be explored include the development of caged oligonucleotides that can be triggered through nonphotochemical mechanisms. For example, enzymatically gated oligomers could be used in combination with transgenic lines that express the activating enzyme in a tissue-specific and, perhaps, inducible manner. This method would enable conditional oligomer activation with a spatial precision and complexity that would be difficult to achieve with light. Technologies that permit reversible control of oligonucleotide function would also be transformative, as they would allow researchers to turn gene expression on and off at will.
Finally, the chemistry and developmental biology communities would benefit from a renewed effort to develop oligonucleotide-based tools for the observation of RNA. Nonradioactive methods for the detection of RNA transcripts by in situ hybridization were developed 20 years ago, and advances since that time have been largely incremental. New synthetic tools have the potential to change how biologists observe the molecular mechanisms that underlie tissue patterning and homeostasis, as exemplified by the use of LNAs to detect miRNAs in zebrafish. Hybridization probes with greater affinity and/or increased detection sensitivity might even obviate enzymatic amplification steps in RNA detection protocols, permitting the real-time observation of transcripts in live embryos.
Realizing these goals will require the concerted efforts of chemists, developmental biologists, and even zebrafish chemical biologists. If history is any guide, the zebrafish community will play a leading role in pioneering many of these interdisciplinary approaches. Its openness to new practitioners has facilitated the integration of emerging scientific concepts, as well as the development of novel technologies. Continuing this collaborative, interdisciplinary spirit will lead to new exciting possibilities with this versatile model organism.
We thank X. Ouyang, A.J. Firestone, and S.C. Ekker for helpful comments on this article. We also gratefully acknowledge financial support from the NIH Director's Pioneer Award (DP1 OD003792), the NIH/NIGMS (R01 GM072600), the March of Dimes Foundation (1-FY-08-433), and the California Institute for Regenerative Medicine (T1-0001).
No competing financial interests exist.