Photoactivatable molecules are versatile probes of cellular and organismal physiology, as they allow biochemical control with spatiotemporal precision.^[1] We and others have developed caged morpholino oligonucleotides (cMOs) that can perturb targeted RNAs in vivo,^[2–8] and these optochemical tools have been used to interrogate the functions of individual genes, such as the zebrafish transcription factors no tail-a (ntla) and ets variant gene 2 (etv2).^{[9, 10]} Morpholino (MO) technologies that can similarly decipher the combinatorial actions of two or more genes would extend this approach to more complex biological systems; however, current cMOs have overlapping spectral properties that preclude differential control. We report here the development of spectrally differentiated cMOs that enable sequential gene silencing via wavelength-selective illumination. We demonstrate the efficacy of these probes in zebrafish embryos and use them to examine the mechanisms of mesoderm patterning.

Our approach builds upon our previous studies of cyclic cMOs in zebrafish models.^[7] Antisense MOs are typically 25 bases in length and designed to complement splicing or translational start sites in targeted RNAs. The resulting MO/RNA duplexes have limited tolerance for backbone curvature, and MO activity therefore can be caged by cyclizing the oligonucleotide with a photocleavable linker. Illumination then re-linearizes the antisense reagent and restores its function. MO cyclization has certain advantages over earlier caging methodologies, which utilize hairpin structures,^{[2, 3]} MO/RNA or MO/MO duplexes,^{[4, 6]} or modified bases:^[5] cyclic cMOs are easy to synthesize, rely on a single optically gated trigger, and obviate the need for auxiliary oligonucleotides.

Due to their modular design, cyclic cMOs can be prepared with a variety of linkers, and our first-generation reagents utilized 4,5-dimethoxy-2-nitrobenzyl (DMNB)-based tethers, which are readily cleaved by 365-nm light. The experimental scope of these reagents would be significantly expanded if cMOs targeting distinct RNA sequences could be differentially photoactivated. Wavelength-selective photo-deprotection of thiols has been achieved using 2-nitrobenzyl (NB) and ([7-bis(carboxymethyl)amino]coumarin-4-yl)methyl (BCMACM) chromophores,^[11] and cyclic guanosine monophosphate (cGMP) and cyclic adenosine monophosphate (cAMP) signalling have been separately regulated using NB-caged protein kinase G and BCMACM-caged cAMP, respectively.^[12] In addition, 4-carboxymethoxy-5,7-dinitroindolinyl-caged glutamate and BCMACM-caged γ-aminobutyric acid have been used to achieve bimodal control of each neurotransmitter,^[13] and phosphoamino acids have been selectively caged with 1-(2-nitrophenyl)ethyl and [7-(diethylamino)coumarin-4-yl]methyl (DEACM) groups.^[14] We therefore envisioned that NB- and DEACM-based linkers could be used to develop wavelength-selective cMOs, allowing the sequential inactivation of two genes in whole organisms (Figure 1).

We first prepared a NB-containing linker with N-hydroxysuccinimide ester and chloroacetamide groups, which can react with 5′ amine- and 3′ disulfide-functionalized MOs to generate macrocyclic oligonucleotides. In comparison to our previous DMNB-based tether, this photocleavable linker has a blue-shifted absorption maximum, minimizing photolysis at wavelengths greater than 400 nm. The bifunctional reagent was synthesized from commercially available 1-(2-nitrophenyl)ethane-1,2-diol (1) in eight steps (Scheme 1). The primary alcohol of 1 was tosylated and reacted with methylamine, and the resulting 1,2-amino alcohol 2 was condensed with methyladipoyl chloride to yield the ester 3. The secondary alcohol in 3 was then conjugated with ethylenediamine through 1,1-carbonyldiimidazole (CDI)-mediated activation, and the primary amine was capped with 2-chloroacetyl chloride to yield compound 4. Methyl ester hydrolysis and N-hydroxysuccinimide coupling subsequently provided the fully functionalized NB linker 5.

We next prepared two DEACM-based linkers with differing absorption properties. The addition of electron-donating groups to the C7 position of the coumarin ring induces a bathochromic shift, and DEACM protecting groups are efficiently photolyzed at wavelengths over 400 nm.^[15] We therefore synthesized a DEACM-containing linker in ten steps (Scheme 2), starting with the oxidation of commercially available 7-diethylamino-4-methylcoumarin (6). Allylation of the aldehyde 7 provided the secondary alcohol 8, which was then t-butyldimethylsilyl (TBDMS)-protected and subjected to hydroboration and oxidation to afford the primary alcohol 9. Activation with CDI followed by treatment with ethylene-diamine and 2-chloroacetyl chloride yielded the intermediate 10. The final DEACM linker 11 was then obtained after alcohol deprotection and N,N′-disuccinimidyl carbonate (DSC) coupling.

Since recent studies have shown that malononitrile functionalization further red-shifts the DEACM absorption profile,^[16] we also synthesized the corresponding diethylaminocoumarylidenemalononitrilemethyl (DEACM-MN)-containing linker in eight steps (Scheme 3). Aldehyde 7 was reacted with nitromethane in the presence of N,N,N′,N′-tetramethylethylenediamine (TMEDA) to afford the nitroalcohol 12. The TBDMS-protected alcohol was converted into the corresponding thiocoumarin 13 with Lawesson’s reagent, and the dicyanocoumarin 14 was obtained by Ag(I)-promoted malononitrile condensation. The nitro group was subsequently reduced with zinc and acetic acid, and the resulting amine was coupled to 6-(2-chloroacetamido)-hexanoic acid to yield the intermediate 15. Alcohol deprotection and DSC coupling then provided the final DEACM-MN linker 16.

We used the NB, DEACM, and DEACM-MN bifunctional linkers to prepare cMOs, using procedures analogous to those we previously developed for DMNB-caged reagents (Scheme 4).^[7] Each N-hydroxysuccinimide ester- and chloroacetamide-functionalized tether was coupled to a ntla-targeting MO (5′-GACTTGAGGCAGACATATTTCCGAT-3′; anti-start codon underlined) modified with a 5′ amine and 3′ disulfide. The corresponding MO-linker amides were treated with immobilized triscarboxyethylphosphine (TCEP) to reduce the terminal disulfides, and the resulting thiols spontaneously reacted with the linker chloroacetamides to form the desired macrocycles. Any remaining linear MOs were removed from the reaction mixtures using iodoacetyl-functionalized resin, and the cyclic cMOs were purified by size-exclusion chromatography.

With these spectrally distinct cMOs in hand, we evaluated their activities in vivo. Ntla is a T-box transcription factor required for the differentiation of axial mesoderm into notochord cells (Figure 2a), and ntla mutants/morphants exhibit re-specification of notochord progenitors into medial floor plate cells, loss of posterior mesoderm, and mispatterned somites.^{[17, 18]} We individually microinjected the DMNB, NB, DEACM, and DEACM-MN cyclic ntla cMOs into zebrafish zygotes (115 fmol/embryo for all reagents except for the DEACM-MN cMO; see below) and either briefly illuminated the embryos at 3.5 hours post fertilization (hpf) with 365-, 405-, or 470-nm light or maintained them in the dark. The zebrafish were then scored at 24 hpf for ntla loss-of-function phenotypes as previously described (Figure 2b).^{[3, 7]} As expected, the DMNB and NB cyclic ntla cMOs were efficiently uncaged by 365-nm light, inducing strong ntla morphant phenotypes in 89% and 86% of the embryos injected with these respective reagents (Figure 2c). The DMNB cMO was more sensitive to 405-nm light; 36% of the embryos in this cohort had partial mesodermal defects, in comparison to only 14% of the NB cMO-injected, 405-nm-irradiated zebrafish. DMNB and NB cMO activation was even less efficient with 470-nm light, with 86% and 91% of the embryos developing normally under these respective conditions.

In contrast to these observations, the DEACM and DEACM-MN cyclic ntla cMOs were highly responsive to 405- and 470-nm irradiation (Figure 2c). For example, 95% of the embryos injected with either reagent had morphological phenotypes consistent with a complete loss of ntla function upon 470-nm illumination. The DEACM and DEACM-MN cMOs, however, were only partially activated by 365-nm light, resulting in a range of mesodermal deficits. The DEACM-MN cMO exhibited more dark activity relative to that of its DEACM counterpart, perhaps due to reduced in vivo stability. A lower embryonic dose of the malononitrile-containing reagent (87 fmol) was therefore required to minimize basal gene silencing.

Taken together, our results suggest that NB and DEACM cyclic cMOs could be used in combination to sequentially silence genes. To investigate this possibility, we focused on two additional regulators of zebrafish mesoderm development, the T-box transcription factor Spadetail (Spt/Tbx16) and the homeobox-containing repressor Floating head (Flh). Spt controls the differentiation of non-axial mesoderm, and spt null zebrafish lack trunk somites.^{[19, 20]} Muscle precursors in these mutants are mislocalized to the tailbud, leading to the hallmark “spade tail” morphology. Spt function is largely restricted to the adaxial and paraxial mesoderm during gastrulation and somitogenesis, and its transcription within axial tissues is inhibited by Flh (Figure 3a).^{[21, 22]} Accordingly, flh mutants inappropriately express spt within the midline, leading to the re-specification of notochord progenitors into muscle.^[23] The hierarchical relationship between Spt and Flh provides a convenient system for interrogating genetic interactions; spt mutants have reduced myogenesis, flh exhibit ectopic axial muscle, and spt/flh double mutants phenocopy spt loss-of-function defects.^[21]

We first recapitulated these phenotypes using conventional MOs targeting each transcription factor (spt MO: 5′-CTCTGATAGCCTGCATTATTTAGCC-3′; flh MO: 5′-GGGAATCTGCATGGCGTCTGTTTAG-3′). In comparison to wildtype zebrafish, spt morphants (50 fmol/embryo) were deficient in adaxial and paraxial muscle by 12 hpf, as determined by diminished expression of myogenic differentiation 1 (myod1) (Figure 3b–c).^[24] Zebrafish injected with the flh MO (100 fmol/embryo) exhibited axial myod1-positive cells, and spt/flh double morphants had myod1 expression patterns similar to that of embryos injected with the spt MO alone. We next pursued the combinatorial control of spt and flh function using spectrally differentiated cMOs. We synthesized a NB cyclic spt cMO and a DEACM cyclic flh cMO as described above and injected the reagents into zebrafish zygotes (50 and 100 fmol/embryo, respectively), either alone or in combination. The resulting embryos were then either briefly irradiated with 365-, 405-, or 470-nm light at 3.5 hpf or cultured in the dark throughout their development, and we stained the embryos for myod1 transcripts at 12 hpf. Zebrafish injected with the NB spt cMO alone had dramatically reduced adaxial and paraxial myogenesis upon 365-nm illumination (100% penetrance) but minimal defects upon exposure to 405- or 470-nm light (19% and 11%, respectively) (Figure 3d). Conversely, about 90% of embryos injected with the DEACM flh cMO had ectopic axial muscle after 405- or 470-nm irradiation, with 365-nm light resulting in a less penetrant phenotype. To confirm that these spectral differences can be exploited to achieve sequential gene knockdowns, we co-injected the NB spt and DEACM flh cMOs into zebrafish zygotes and first irradiated the embryos at 3.5 hpf with 405-nm or 470-nm light. A subset of the animals was then exposed to 365-nm light, and the resulting myod1 expression was scored at 12 hpf (Figure 3d). Both 405- and 470-nm irradiation selectively activated the DEACM flh cMO, as shown by axial myod1 expression (75% and 83%, respectively). As expected, subsequent 365-nm irradiation restored the spt loss-of-function phenotype (100% and 93%, respectively). Thus, the NB and DEACM cMOs enabled wavelength-selective control of spt and flh function in zebrafish embryos.

To conclude our studies, we utilized these spectrally differentiated cMOs to examine the timing by which ectopic spt function induces axial muscle development in flh-deficient embryos. We co-injected the NB spt and DEACM flh cMOs into zebrafish zygotes and irradiated the embryos at 3.5 hpf with 405-nm or 470-nm light to inhibit flh function as before. We then activated the spt cMO with 365-nm light at time points ranging from 5 to 7 hpf. Light-induced spt silencing at 5 hpf resulted in nearly complete loss of adaxial and paraxial myod1 transcription (100% and 86% penetrance for the 405- and 470-nm flh cMO photoactivation protocols, respectively) (Figure 4). In contrast, spt cMO photolysis at 7 hpf predominantly yielded embryos with flh morphant phenotypes (67% for both flh cMO uncaging protocols). Our time-course studies therefore indicate that Spt acts during gastrulation (6–9 hpf) in flh mutants to redirect notochord precursors toward muscle cell fates. This action precedes axial myod1 expression in flh mutants, which is evident by the 3- to 5-somite stage (10–11 hpf).^[23]

In summary, we have achieved the first sequential inactivation of organismal gene function using reverse-genetic chemical probes. Our studies demonstrate that NB- and DEACM-based cyclic cMOs can be differentially activated by 365- and 405/470-nm light, expanding the scope of caged antisense technologies. Such synthetic tools complement TALEN (transcription activator-like effector nucleases) and CRISPR (clustered regularly interspaced short palindromic repeats)/Cas genome-editing technologies, which have been successfully employed in zebrafish.^{[25, 26]} Cyclic cMOs can be synthesized and applied within days, allowing rapid loss-of-function studies with nearly complete phenotypic penetrance and a high degree of spatial, temporal, and dosage control. In comparison, genetic approaches can require multiple animal generations for full implementation, have more limited conditionality, and are subject to Mendelian phenotypic frequencies.

We anticipate that these optochemical tools will be valuable probes of the genetic interactions that dynamically regulate tissue formation and function. The commercial availability of 5′ amine- and 3′ disulfide-functionalized MOs and the ease with which they can be cyclized should make this chemical approach widely available to the scientific community. Our MO caging strategy is also compatible with a variety of linkers, facilitating the implementation of other chromophores. As new caging groups with divergent spectral properties continue to be developed,^[27] it is likely that future cMO technologies will eventually enable the orthogonal photoregulation of three or more genes.