Deciphering the molecular mechanisms that underlie embryogenesis requires an ability to alter gene function with spatial and temporal precision. Synthetic reagents can be valuable tools in this discovery process, especially in model organisms currently intractable to targeted genomic changes. In particular, morpholino-based antisense oligonucleotides (MOs) have been widely used to inhibit gene expression in metazoans that develop ex utero (Figure 1a),[1–4] and caged versions of these reagents (cMOs) can enable conditional gene silencing through targeted illumination.[5–9] Composed of morpholino-based nucleosides and a phosphorodiamidate backbone, these nuclease-resistant probes are typically injected into zygotes as 25-base oligomers, after which they hybridize to complementary RNAs and disrupt splicing or translation. While cMOs can provide important new insights into embryonic gene function, their utility has been hindered by their empirical design, synthetic complexity, in vivo instability, reliance on complementary inhibitors, and/or use of multiple caging groups. Here we describe next-generation cMOs that overcome these limitations. We demonstrate that cyclic cMOs are effective reverse-genetic tools in zebrafish embryos, and we use these reagents to examine the timing of exocrine fate commitment in the developing pancreas.
We previously developed hairpin cMOs that can be activated with either ultraviolet (UV; 360-nm) or two-photon irradiation (Figure 1b).[5, 10] The RNA-targeting MO is tethered to a shorter, complementary MO inhibitor through a dimethoxynitrobenzyl (DMNB)- or bromohydroxyquinoline (BHQ)-based linker, thereby forming a hairpin structure that is resistant to RNA hybridization. Linker photolysis separates the two oligonucleotides, allowing the 25-base MO to interact with its RNA target. Hairpin cMOs have been used to conditionally inactivate the expression of several zebrafish genes, including no tail-a (ntla), floating head (flh), heart of glass (heg), and ETS1-related protein (etsrp).[5, 10] We have also integrated ntla cMOs, caged fluorophores, fluorescence-activated cell sorting, and microarray technologies to dynamically profile the Ntla-dependent transcriptome. Despite these successes, hairpin cMOs have limitations. Their design involves careful optimization of thermodyamic properties, their purification requires high-performance liquid chromatography (HPLC), and the inhibitory oligonucleotide can exacerbate MO toxicity.
Other oligonucleotide caging strategies have been developed, each with its own advantages and disadvantages. For example, peptide nucleic acid (PNA)-2-OMe RNA chimeras that form hairpins have been used to conditionally inhibit gene function, although the photo-released 2-OMe RNAs may be toxic to embryos. Duplexes composed of the targeting MO and two complementary, tandem oligonucleotides connected by a nitrobenzyl-based linker have also been reported (Figure 1c).[7, 9] While these duplexes are easier to prepare than hairpin reagents, they release two potentially cytotoxic oligonucleotides upon photolysis and can be less stable in vivo. Finally, MOs containing four nitropiperonyloxymethyl-caged thymine bases have been synthesized, bypassing the need for an inhibitory oligomer (Figure 1d). Activating these reagents, however, requires the photolysis of multiple caging groups per oligonucleotide, which can be difficult to achieve without damaging levels of UV irradiation.
To overcome these limitations, we sought to develop cMOs that are easier to synthesize, rely on a single caging group, and do not involve a complementary inhibitor. Since oligonucleotide duplexes have limited tolerance for local curvature, we hypothesized that MO cyclization through a photocleavable linker could enable conditional gene silencing (Figure 1e). A recent report that circular DNA oligonucleotides bind inefficiently to their complementary RNAs in vitro lends further support to this concept.
We tested this new caging strategy by targeting ntla, the zebrafish ortholog of mammalian Brachyury.[14–16] Loss-of-function phenotypes for this T-box transcription factor include ablation of the notochord and posterior mesoderm, ectopic medial floor plate cells, and somite mispatterning. We synthesized a DMNB-functionalized linker with N-hydroxysuccinimide ester and chloroacetamide groups (Scheme 1) and reacted it with a 25-base ntla MO (Table S1) bearing 5′-amine and 3′-alkyl disulfide modifications (Scheme 1). Reduction of the linear MO-linker intermediate yielded the free thiol, which spontaneously reacted with the chloroacetamide to provide the desired macrocycle. Liquid chromatography-mass spectrometry (LCMS) analysis indicated that the macrocyclization reaction went to completion, and the final product was purified by gel filtration. The total yield for the 25-base cyclic ntla cMO synthesis starting with the linker and the targeting MO was 80%, approximately a 10-fold improvement over our hairpin cMO protocol.
We next investigated the efficacy of the cyclic ntla cMO in zebrafish embryos. Since we had previously shown that the conventional, linear ntla MO recapitulates the ntla mutant phenotype at a dose of 1 ng/embryo,[5, 10] we delivered an equivalent amount of the photocleavable cyclic ntla cMO into zygotes and globally irradiated a subset of the embryos with UV light for 10 seconds at 3 hours post fertilization (hpf). The embryos were cultured until 24 hpf, at which point their phenotypes were scored according to four morphological classes as previously described (Figure 2a). The cyclic ntla cMO proved to be comparable to the hairpin reagent in terms of efficacy, if not modestly better, with approximately 75% (n=24) of the cMO-injected embryos exhibiting no developmental defects in the absence of UV light and 90% (n=20) achieving complete ntla mutant phenotypes upon UV irradiation (Figure 2b). Western blot analysis of 10-hpf zebrafish confirmed the UV light-dependent loss of Ntla protein in cyclic ntla cMO-injected embryos (Figure 2c). The cyclic ntla cMO could also be used to knockdown ntla function in a spatially restricted manner; photoactivation of the reagent within the embryonic shield at 6 hpf converted notochord progenitors into medial floor plate cells, as has been reported with hairpin ntla cMOs (Figure 2d).[5, 11]
In principle, the basal activity of cyclic cMOs should decrease with macrocycle size, as local curvature within the smaller MO macrocycles will be more acute. To explore this possibility, we synthesized non-photocleavable cyclic MOs corresponding to 21-, 23-, and 25-base oligonucleotides that target the ntla sequence (Table S1 and Scheme S1). The cyclic oligonucleotides or the corresponding linear MOs were then mixed with equimolar amounts of 25-base complementary RNA (Table S1), and temperature-dependent changes in hypochromicity were assessed (Table 1). Although these in vitro assays cannot recapitulate the complexity of MO/RNA interactions in vivo, the thermodynamic insights they provide can be informative. As expected, the cyclic MOs exhibited reduced affinities for their RNA targets than their linear counterparts, and greater differences in Tm and ΔG values were observed as MO length decreased.
We next tested photoactivatable versions of these cyclic MOs and their linear counterparts in zebrafish embryos. As before, we globally UV-irradiated a subset of the cMO-injected embryos at 3 hpf and evaluated their morphological phenotypes at 24 hpf and Ntla protein levels at 10 hpf. Although the cyclic ntla cMOs exhibited less basal activity as they decreased in size, they also had a commensurate loss of efficacy upon uncaging (Figure 2, b–c). Linear versions of the 21- and 23-base reagents were similarly less potent, indicating that this activity reduction is due to diminished MO/RNA interactions. Thus, within this dosing regimen, the 25-base cyclic cMO appears to provide the optimum dynamic range of caged and uncaged activities.
To explore the generality of cyclic cMOs, we synthesized a 25-base reagent targeting pancreas transcription factor 1 alpha (ptf1a) (Table S1), which is required for exocrine cell differentiation in the developing pancreas.[17–19] Ptf1a is selectively expressed in pancreatic exocrine cells, and their formation can therefore be directly visualized in transgenic zebrafish that express enhanced green fluorescent protein under control of ptf1a promoter-derived cis-regulatory elements [Tg(ptf1a:eGFP) zebrafish]. We injected Tg(ptf1a:eGFP) zygotes with the cyclic ptf1a cMO at a dose of 2 ng/embryo and globally UV-irradiated a subset at 3 hpf for 10 seconds. Another set of Tg(ptf1a:eGFP) zygotes were injected with an equivalent amount of a hairpin ptf1a cMO that was designed according to thermodynamic parameters. The intensity of eGFP fluorescence within the pancreatic field at 3 days post fertilization (dpf) was then scored for each experimental condition (Figure 3a).
The hairpin ptf1a cMO partially impeded pancreatic development even without photoactivation, possibly due to linker degradation or the ability of the reagent to interconvert between stem-loop and linear forms during the three-day period (Figure 3b). In contrast, the majority of Tg(ptf1a:eGFP) zygotes injected with the cyclic ptf1a cMO developed normally in the absence of UV light (76%, n=21) but failed to form exocrine pancreas upon UV irradiation (95%, n=21) (Figure 3b). Endocrine pancreas development was not affected, as gauged by whole-mount in situ hybridization with an antisense probe for insulin (Figure 3c). These findings demonstrate the general efficacy of cyclic cMOs and suggest that these chemical tools might be better probes of later embryological processes than hairpin-based reagents.
To conclude these studies, we used the cyclic cMO technology to examine the temporal requirements of ptf1a function for exocrine pancreas formation. We uncaged the cyclic ptf1a cMO in zebrafish with irradiation conditions optimized for various developmental stages (3 to 52 hpf; Figure S1),. We then assessed exocrine tissue development at 3 dpf by whole-mount in situ hybridization with antisense trypsin (Figure 4) and ptf1a (Figure S2) probes. The resulting expression levels of these exocrine markers followed similar trends, with substantial reductions in transcript levels upon UV irradiation up to 44 hpf. Uncaging of the ptf1a cMO at 48 and 52 hpf, however, had only minor effects on their expression. Assuming that cyclic ptf1a cMO efficacy is not lost during these studies (MOs can perdure in zebrafish for up to five days), our results suggest that pancreatic cells may be committed to exocrine cell fates as early as 2 dpf.
Taken together, our findings demonstrate the efficacy of cyclic cMOs as reverse-genetic tools for embryological studies and highlight their potential advantages over hairpin-based reagents. Cyclic cMOs can be synthesized in two steps starting from commercially available oligonucleotides, and they can be purified without resorting to HPLC. They also bypass the need for inhibitory oligomers and multiple caging groups. Our results further suggest that these cyclic reagents might be more efficacious than current cMO technologies for studying organogenesis and other late-stage developmental processes. These attributes should facilitate the implementation of cMOs by the developmental biology community and enhance future efforts to decipher embryonic gene function.