Targeted cell ablation in zebrafish using optogenetic transcriptional control

Tissue development, homeostasis and regeneration rely on functional networks composed of diverse cell types. To probe these mechanisms, several genetically encoded technologies have been developed to achieve cell ablation with spatiotemporal control, including cytotoxic proteins (Chelur and Chalfie, 2007; Kurita et al., 2003; Slanchev et al., 2005; Smith et al., 2002, 2007) and enzymes that can covert prodrugs into cytotoxic molecules (Bridgewater et al., 1995; Chu et al., 2007; Curado et al., 2007; Jung et al., 2007; Pisharath et al., 2007; Springer and Niculescu-Duvaz, 2000). Both of these approaches can be combined with tissue-specific promoters to spatially restrict ablation; however, cis-regulatory sequences that are exclusive to the targeted cells are not always available, resulting in collateral damage to other tissues. For example, nitroreductase and nitrofuran-based toxins have been used to ablate motor neurons in zebrafish larvae (Ohnmacht et al., 2016), providing an alternative to mechanically induced spinal cord lesions (Becker and Becker, 2001; Briona et al., 2015; Mokalled et al., 2016; Reimer et al., 2009; Yu et al., 2011). Although this chemical-genetic technique was able to elicit regenerative responses in the central nervous system (CNS), the promoter used for these studies also led to nitroreductase expression in the heart and pancreas. As result, metronidazole treatment also induced heart damage and edema (Ohnmacht et al., 2016).

Light-inducible technologies have the potential to achieve targeted cell ablation in a more precise manner. Toward that goal, chromophores that generate reactive oxygen species (ROS) upon exposure to light have been applied in various systems (Buckley et al., 2017; Bulina et al., 2006; Makhijani et al., 2017; Qi et al., 2012; Sarkisyan et al., 2015; Xu and Chisholm, 2016). However, these methods can require sustained irradiation for efficient cell ablation, limiting their ability to target specific cell populations in dynamic systems, and ROS sensitivity also varies between cell types (Williams et al., 2013). We reasoned that light-inducible expression of cytotoxic proteins or toxin-activating enzymes might represent a more general and versatile approach. In particular, we envisioned this could be achieved using a photoactivatable transcription factor that can drive the expression of exogenous transgenes. One candidate construct for this strategy is a synthetic transcription factor, GAVPO, that was originally developed for light-inducible gene expression in mammalian systems (Wang et al., 2012). The GAVPO transactivator contains a Gal4 DNA-binding domain, a variant of the VIVID light-oxygen-voltage (LOV) domain and the p65 transcriptional activation domain. Upon exposure to blue light, the LOV domain undergoes a photo-reductive reaction that forms a covalent adduct between its flavin adenine dinucleotide chromophore and a neighboring cysteine residue. The adduct stabilizes a conformational state that promotes LOV domain self-association, generating a transcriptionally active GAVPO dimer. Light-activated GAVPO then drives the expression of upstream activating sequence (UAS)-controlled genes until it reverts to the inactive monomer with a half-life of ∼2 h (Wang et al., 2012).

In this report, we investigate the efficacy of light-inducible cell ablation in zebrafish using two genetically encoded cell-ablation technologies and the GAVPO transactivator. We demonstrate that the cytotoxic ion channel variant M2^H37A (Lam et al., 2010; Smith et al., 2002, 2007) acts through non-apoptotic pathways to kill neurons and other cell types with greater efficacy and faster kinetics than the nitroreductase/nitrofuran system. We further establish GAVPO as an effective tool for achieving photoactivatable gene transcription in zebrafish, using both embryos injected with GAVPO mRNA and stable transgenic lines with neuronal expression of the transactivator. Finally, we show that integrating the GAVPO and M2^H37A technologies enables targeted cell death in zebrafish embryos and larvae. We anticipate that optogenetic cell ablation will be a versatile approach for studying developmental and regenerative biology.

To develop conditional models of CNS damage, we pursued neuronal ablation through two genetically encoded technologies: bacterial nitroreductase (NTR) and viral ion channel M2. The Escherichia coli gene, nsfB, encodes a flavoprotein that can reduce a variety of nitroaromatic compounds (Zenno et al., 1996). Among the substrates of this nitroreductase (NTR) is the prodrug metronidazole, which is reductively converted into a DNA interstrand crosslinking agent that triggers cell death. M2 normally functions as a proton-selective channel during influenza infection (Shimbo et al., 1996; Wharton et al., 1994), and a single point mutation in the transmembrane domain (H37A) converts M2 into a constitutively active, non-selective cation channel (Gandhi et al., 1999; Wang et al., 1995). This loss of specificity renders M2^H37A cytotoxic (Lam et al., 2010; Le Tissier et al., 2005; Smith et al., 2002, 2007). The availability of M2-targeting antiviral drugs, such as rimantadine and amantadine (Intharathep et al., 2008; Schnell and Chou, 2008), also enables pharmacological control of M2^H37A function.

Exogenous expression of NTR in combination with metronidazole has been widely used to selectively ablate cells in zebrafish (Curado et al., 2007, 2008; Pisharath and Parsons, 2009; Pisharath et al., 2007). However, the M2^H37A channel has not been applied previously in this model organism. We therefore injected zebrafish embryos with varying amounts of M2^H37A mRNA and monitored their embryonic development. Toxicity was initially observed at the blastula stages, and the embryos were unable to proceed through gastrulation (Fig. 1A,B). Injecting increasing amounts of M2^H37A mRNA rapidly killed 100% of the embryos. Culturing the M2^H37A-expressing embryos in the presence of 100 µg/ml rimantadine enabled some to develop normally until 24 h post fertilization (hpf). However, the channel blocker was unable to fully protect against channel-mediated toxicity, demonstrating the need to control M2^H37A expression.

We next examined whether M2^H37A can be used to ablate zebrafish neurons by establishing a system for Gal4/UAS-dependent expression in these cells. We generated heterozygous Tg(UAS:M2^H37A;myl7:mCherry) zebrafish, using five tandem UAS sequences to minimize basal M2^H37A expression (Ma et al., 2013), and crossed them with a homozygous Tg(tuba1a:Gal4VP16;myl7:GFP) line, which restricts Gal4VP16 expression to the developing nervous system (Goldman et al., 2001). We then selected progeny that contained one copy of each transgene, which could be confirmed by the presence of both GFP and mCherry fluorescence in the heart (driven by the myl7 promoter of each construct) (Fig. 1C). The double transgenic embryos expressed M2^H37A in neural tissues (Fig. S1), and they exhibited CNS deficits that appeared as early as 24 hpf and progressed over time (Fig. 1C). Rimantadine rescued these defects, confirming the role of M2^H37A channel activity in neuronal loss (Fig. 1C).

To compare the ablative efficacies of the M2^H37A and NTR/metronidazole systems, we also crossed homozygous Tg(UAS:NTR-mCherry) (Goldman et al., 2001) and homozygous Tg(tuba1a:Gal4VP16;myl7:GFP) zebrafish, and cultured the resulting embryos in medium supplemented with DMSO or 5 mM metronidazole starting at bud stage (10 hpf). In contrast to zebrafish with neuronal M2^H37A expression, metronidazole treatment of the NTR-expressing progeny did not induce CNS defects until 48 hpf, suggesting the two ablative technologies kill neural cells with different kinetics and possibly through distinct mechanisms (Fig. 1D). We next used TUNEL staining to detect apoptosis in M2^H37A- and NTR-expressing embryos (Fig. S2A). Although we observed overt developmental defects as early as 24 hpf in M2^H37A-expressing embryos, significant TUNEL staining was not detected until 32 hpf, and TUNEL-positive cells were distributed in regions within and near the CNS. Because TUNEL staining detects DNA fragmentation that occurs later in the apoptotic process, we also stained the embryos for activated caspase 3, the initiator caspase for apoptosis. Similar to our TUNEL results, we did not observe activated caspase 3 until 32 hpf (Fig. S2B). In comparison, NTR-expressing embryos treated with 5 mM metronidazole first exhibited phenotypic defects at 48 hpf, which coincided with robust TUNEL staining localized to the CNS (Fig. S2A). Thus, M2^H37A appears to kill zebrafish neurons through necrosis, which, when continuously expressed in the developing CNS, subsequently causes the apoptosis of surrounding cells. This progressive mechanism of tissue loss is similar to that associated with CNS damage (Oyinbo, 2011), potentially making M2^H37A a useful tool for studying neural injury and recovery.

Using cytotoxic proteins to ablate specific cell populations, including regions within a tissue of interest, requires spatiotemporal control of their expression or function. To accomplish this goal, we pursued light-dependent NTR and M2^H37A expression in zebrafish embryos and larvae. A number of strategies for photoactivatable gene expression have been described, including GAVPO (Fig. 2A), EL222, an EL222-derived construct (TAEL) and two-hybrid-like systems based on CYR2-CIB1 or PhyB-PIF3 heterodimerization (Liu et al., 2012; Motta-Mena et al., 2014; Reade et al., 2017; Shimizu-Sato et al., 2002). Of these technologies, EL222, TAEL and CYR2-CIB1have been applied previously in zebrafish, but they are limited by the minutes-scale lifetimes of their light-activated states. The PhyB-PIF3 system requires co-administration of phycocyanobilin, an exogenous chromophore that cannot penetrate zebrafish embryos (Buckley et al., 2016). As GAVPO uses endogenous flavin and has a light-state half-life of 2 h, we envisioned that this optogenetic tool could afford transcriptional control with transient illumination, enabling genetically encoded ablation of dynamic cell populations.

We assessed GAVPO activity in Tg(UAS:NTR-mCherry) zebrafish by injecting transgenic zygotes with various amounts of GAVPO mRNA (Fig. 2B). The embryos were then raised in the dark until 6 hpf, after which half were exposed to blue light (470 nm) from a light-emitting diode (LED) lamp for 2 h. NTR-mCherry levels were subsequently assessed by fluorescence microscopy at 24 hpf. GAVPO expression induced ubiquitous NTR-mCherry expression in a light- and dose-dependent manner (Fig. 2B,C), and the majority of animals developed normally when the GAVPO mRNA was injected at a dose of 100 pg/embryo (Fig. 2D). At 200 pg doses, GAVPO mRNA led to embryonic deformity or lethality. In comparison, Gal4VP16 mRNA caused developmental defects even when only 25 pg was injected per embryo, corroborating previous reports of Gal4VP16 toxicity (Köster and Fraser, 2001; Scott et al., 2007) (Fig. 2D). Taken together, these findings indicate that the GAVPO system can be used to convey light-dependent gene expression in multiple zebrafish cell types and may provide a less-toxic alternative to the currently available Gal4 driver lines.

We next examined whether GAVPO can be coupled with targeted illumination to achieve both spatial and temporal control of gene expression. We injected heterozygous Tg(UAS:NTR-mCherry) zygotes with GAVPO mRNA and irradiated the embryonic shield at 6 hpf for 20 s, using an epifluorescence microscope equipped with a 470/40 nm filter, a 20× objective and an iris diaphragm. Zebrafish fate-mapping studies have shown that cells within this dorsal organizing center give rise to the notochord, head mesoderm and hatching gland (Kimmel et al., 1990; Shih and Fraser, 1996), and we observed red fluorescence in these tissues at later developmental stages (Fig. 3A). To determine whether the expressed NTR-mCherry can impart nitrofuran sensitivity, we treated a cohort of the shield-irradiated embryos with either 5 mM metronidazole or DMSO vehicle alone from 10 to 32 hpf. We then fixed and immunostained the embryos for activated caspase 3. Although basal levels of caspase 3 activation increased with metronidazole treatment, we did not observe any evidence of GAVPO/NTR-mediated apoptosis in the notochord (Fig. S2C). Thus, even levels of NTR-mCherry expression that can be readily detected by fluorescence microscopy are not sufficient to convey nitrofuran-induced apoptosis within this time frame.

As our studies with Tg(tuba1a:Gal4VP16;myl7:GFP) zebrafish indicated that M2^H37A can ablate neurons with faster kinetics than the NTR/metronidazole system, we investigated whether the GAVPO transactivator could be used to express cytotoxic levels of this constitutively active channel. We injected varying amounts of GAVPO mRNA into heterozygous Tg(UAS:M2^H37A;myl7:mCherry) zygotes, establishing 30 pg/embryo as the maximum dose that permits normal embryonic development in this transgenic line under dark-state conditions (Fig. 3B). Higher concentrations of GAVPO mRNA caused patterning defects, and RT-PCR analyses confirmed that this coincides with increased light-independent M2^H37A expression (Fig. 3C). We then injected Tg(UAS:M2^H37A;myl7:mCherry) zygotes with 30 pg/embryo of GAVPO mRNA, and focally irradiated the shield with blue light at 6 hpf for 20 s using the epifluorescence microscope. The resulting embryos exhibited deformities of the notochord and head when imaged at 30 hpf (Fig. 3D), matching the transgene expression we observed in Tg(UAS:NTR-mCherry) zebrafish that had been injected with GAVPO mRNA and shield irradiated (Fig. 3A). In addition, culturing the focally irradiated embryos in the presence of the M2 blocker rimantadine, prevented these developmental defects.

Having established the ability of transiently expressed GAVPO to convey light-inducible cell ablation, we explored the efficacy of transgenic lines that stably express GAVPO in a neuron-specific manner. To target GAVPO expression to post-mitotic neurons, we generated zebrafish carrying GAVPO under control of the neuronal promoter elavl3 (also known as HuC) (Kim et al., 1996; Park et al., 2000a,b). Multiple founders were screened for single insertions of the GAVPO expression construct and raised to generate two stable lines. When crossed to homozygous Tg(UAS:NTR-mCherry) zebrafish, both lines yielded embryos that exhibited photoinducible, neuron-specific mCherry fluorescence in the expected Mendelian ratios (Fig. S3A). One of the heterozygous lines exhibited more robust mCherry fluorescence in the developing spinal cord, and we incrossed this line to generate fertile, homozygous Tg(elavl3:GAVPO) zebrafish. GAVPO expression in this line recapitulated endogenous elavl3 expression (Fig. S3B).

We next tested the ability of the elavl3:GAVPO transgene to convey light-inducible gene expression in neurons. We crossed homozygous Tg(elavl3:GAVPO) and homozygous Tg(UAS:NTR-mCherry) zebrafish to obtain embryos that contained one copy of each transgene. The double transgenic embryos were irradiated at different developmental stages for 2 h using the blue LED lamp and imaged 8 h post irradiation (hpi). We found that the mCherry fluorescence coincided with GAVPO expression at 36 and 48 hpf (Fig. S3B,C), and correlated with the duration and intensity of blue-light illumination (Fig. 4A-C, Fig. S4). Focal irradiation using the epifluorescence microscope induced mCherry expression in the targeted neurons within the zebrafish brain, whereas global irradiation concurrently induced mCherry expression in the spinal cord (Fig. 4D-F). We also examined the sensitivity of the GAVPO transactivator to ambient light by exposing the double transgenic embryos to a white LED lamp (intensity at 470 nm=1.8 mW/cm²). These embryos exhibited substantially lower levels of mCherry expression in comparison with those irradiated with blue light for the same duration (Fig. S4C).

We then crossed homozygous Tg(UAS:M2^H37A;myl7:mCherry) zebrafish with the homozygous Tg(elavl3:GAVPO) line and exposed their Tg(elavl3:GAVPO;UAS:M2^H37A;myl7:mCherry) progeny to blue light at 48 hpf for varying lengths of time. As determined by whole-mount immunostaining, M2^H37A levels correlated with the duration of blue-light irradiation (Fig. S5). Rimantadine treatment led to higher M2^H37A expression levels, likely owing to its ability to inhibit the cytotoxic channel and promote the survival of M2^H37A-expressing neurons.

Finally, we sought to confirm that Tg(elavl3:GAVPO;UAS:M2^H37A;myl7:mCherry) zebrafish exhibit light-dependent neuronal loss. As M2^H37A-induced cell death is not associated with caspase 3 activation or DNA fragmentation that can be detected by TUNEL staining, we monitored neuronal ablation by optical microscopy. We injected Tg(elavl3:GAVPO;UAS:M2^H37A;myl7:mCherry) zygotes with an elavl3:mCherry-CAAX construct to label post-mitotic neurons in a mosaic manner (Fig. 5A), facilitating their visualization and quantification. We raised the embryos to 48 hpf, globally irradiated a subset with the blue LED lamp for 8 h, and imaged the zebrafish at 12 hpi. The irradiated animals had significantly fewer intact neurons in the hindbrain than those maintained in the dark or cultured in rimantadine-containing medium after exposure to blue light (Fig. 5A). The spinal cords of irradiated Tg(elavl3:GAVPO;UAS:M2^H37A;myl7:mCherry) embryos also exhibited reduced axon bundle widths (Fig. 5B). In contrast to the Gal4VP16-driver line with high levels of constitutive neural M2^H37A expression (Fig. 1C), we did not observe gross abnormalities in the irradiated transgenic animals and caspase 3 staining of these embryos had limited apoptosis (Fig. S5B). Thus, the effects of transient M2^H37A expression can be restricted to the targeted tissues.

Our studies demonstrate that GAVPO can be used to achieve optogenetic control of gene expression in zebrafish. Although it has previously been reported that the GAVPO mRNA disrupts zebrafish development and reduces embryo viability (Reade et al., 2017), we readily generated viable, fertile transgenic lines, indicating that the photoactivatable transactivator is not inherently toxic. The efficacy of GAVPO in zebrafish opens the door to new experimental approaches for this model organism, as this VIVID LOV domain-based system has certain advantages over other optogenetic approaches. For example, the homodimeric GAVPO transactivator can be introduced into organisms as a single construct, whereas CYR2/CIB1- and PhyB/PIF-based technologies are heterodimeric systems (Buckley et al., 2016; Liu et al., 2012). The photoactivated state of GAVPO is also long-lived, with a mean lifetime that is more than two orders of magnitude greater than that of EL222 and its derivatives (τ∼2 h versus 30 s) (Motta-Mena et al., 2014; Wang et al., 2012). As a result, sustained GAVPO activity can be achieved with a single pulse of blue light, allowing transcriptional control of specific cell populations in developing embryos and other dynamic living systems. The reliance of GAVPO on endogenous flavin chromophores and its compatibility with the myriad of existing Gal4/UAS systems are other major strengths.

These attributes highlight the versatility of GAVPO as an optogenetic tool and, in principle, its functionality could be enhanced through modifications of its LOV domain. For example, the light-state lifetime of the VIVID LOV domain can be increased up to tenfold by mutating the flavin-binding site (Zoltowski et al., 2009). Introducing these residue changes into the GAVPO transactivator could prolong transcriptional responses to a single illumination pulse. We also observed that GAVPO mRNA was more toxic to Tg(UAS:M2^H37A;myl7:mCherry) embryos than their Tg(UAS:NTR-mCherry) counterparts, even when the mRNA-injected zebrafish were cultured in the dark. As M2^H37A is constitutively ablative and NTR is not, we interpret these results as evidence of dark-state GAVPO activity. Consistent with this model, rimantadine attenuates GAVPO mRNA-dependent toxicity in Tg(UAS:M2^H37A;myl7:mCherry) embryos. We therefore anticipate that basal activation of UAS-driven transgenes will increase with GAVPO concentration, and strong promoters may constrain the dynamic range of this photoactivatable transactivator. This limitation could be overcome by using alternative cis-regulatory sequences or by screening multiple transgenic lines, or, alternatively, GAVPO variants with lower dark-state dimer affinities could be developed.

Our studies also establish M2^H37A expression as an effective strategy for ablating specific cell populations in whole organisms. In addition to providing the first demonstration that M2^H37A is cytotoxic in zebrafish, we reveal important differences between this non-selective cation channel and the NTR/metronidazole system. M2^H37A expression kills cells with greater efficacy than NTR-mediated prodrug activation, as evidenced by their respective activities in the zebrafish CNS and axial mesoderm. M2^H37A and NTR/metronidazole also ablate cells through distinct mechanisms. M2^H37A appears to initially trigger neuronal necrosis, and surrounding cells can subsequently undergo apoptosis, at least when the cytotoxic channel is continuously expressed in the targeted tissue (e.g. using a CNS Gal4VP16 driver). In comparison, NTR-mediated activation of metronidazole induces programmed cell death. These differences make M2^H37A a promising tool for studying injury, a process that involves both necrosis and apoptosis. For example, previous studies have shown that spinal cord injuries lead to sustained apoptosis that can last for weeks (Crowe et al., 1997; Oyinbo, 2011). M2^H37A could be particularly advantageous for investigating molecular and cellular responses to CNS damage, as commonly used models of spinal cord injury rely on mechanical perturbations that can concurrently damage muscle and skin (Becker et al., 1997; Bhatt et al., 2004). The sensitivity of M2^H37A to rimantadine also enables pharmacological tuning of this cytotoxic channel, allowing one to minimize basal activity and maximize inducibility.

Finally, by combining the GAVPO and M2^H37A, we have achieved optical control of cell ablation in a whole organism. In comparison with other genetically encoded methods for cell ablation, the GAVPO/M2^H37A system can eliminate cells of interest without requiring tissue-specific promoters. When GAVPO is expressed ubiquitously, cell targeting can be guided by morphological cues alone, as illustrated by our ablation of progenitors within the zebrafish shield. Moreover, optically inducing M2^H37A expression within the shield did not appear to cause neighboring cells to apoptose, contrasting the non-cell-autonomous effects we observed with Gal4VP16-driven M2^H37A expression in the developing CNS. Thus, it may be possible to use the GAVPO/M2^H37A system to elucidate the developmental fates of other embryonic cell populations. Alternatively, cis-regulatory sequences can be used to restrict GAVPO expression to certain tissues, enabling the ablation of specific cell populations within these promoter-defined regions. The latter approach could be used to generate tissue injuries in a spatially defined manner. Based on these capabilities, we anticipate that the GAVPO/M2^H37A system will be a versatile tool for deconstructing developmental and regenerative mechanisms in whole organisms.

Adult zebrafish [wild-type AB strain and Tg(UAS-E1b:NTR-mCherry) (Davison et al., 2007); 3-18 months] were obtained from the Zebrafish International Resource Center. The transgenic line Tg(tuba1a:Gal4VP16;myl7:GFP) (Goldman et al., 2001) was a generous gift from Philippe Mourrain (Stanford University, CA, USA). All zebrafish lines were raised according to standard protocols. Embryos were obtained through natural matings and cultured at either 28.5°C or 32°C in E3 medium. The culture medium was also supplemented with 0.003% (w/v) PTU to prevent pigmentation. All animal procedures were approved by the Administrative Panel on Laboratory Animal Care at Stanford University (protocol 10511) and the University of Wyoming (protocol 2018017KM00326-02).

To generate GAVPO mRNA, the GAVPO-coding region with an upstream Kozak site was PCR amplified from pGAVPO plasmid (a gift from Yi Yang, East China University of Science and Technology, Shanghai, China) using the following primers containing BamHI and XbaI sites (underlined): 5-GAATGGATCCGCCACCATGAAGCTACTGTC-3′ and 5′-GAACTCTAGAGTGTACATTACTTGTCATCATCGTC-3′. The resulting PCR product was digested and ligated into pCS2+ to generate pCS2-GAVPO. The insertion was sequenced to confirm ligation fidelity. pCS2-M2^H37A (a gift from Tim Mohun, Francis Crick Institute, London, UK) and pCS2-GAVPO constructs were linearized with SacII, and their corresponding mRNAs were synthesized using in vitro SP6-dependent runoff transcription (Invitrogen).

To generate the Tol2-elavl3:GAVPO plasmid, the GAVPO-coding region with an upstream Kozak sequence was PCR amplified using the following primers: 5′-CCACCTGCAGATAATTGTTTAAACCACTCCGCCACCATGAAG-3′ and 5′-AGTAAAACGACGGCCAGGATCCACCGGTCTGCTATTACTTGTCATCATCGTC-3′. Tol2-elavl3:GCaMP6s plasmid (Misha Ahrens; Addgene 59531) was digested with AgeI, and Tol2-elavl3:GAVPO was assembled from the resulting vector and GAVPO PCR product using Gibson Assembly Master Mix (NEB). The entire plasmid was sequenced to confirm ligation fidelity.

To generate the Tol2-5xUAS-TATA:M2^H37A;myl7:mCherry construct, the M2-coding region was PCR amplified using the following primers, which contain HindIII and ApaI sites (underlined): 5′-GAACAAGCTTGCCACCATGAGTCTTCTAACCG-3′ and 5′-GAATGGGCCCTTACTCCAGCTCTATGTTGAC-3′. This PCR product was then digested and ligated into pU5 (a gift from Yi Yang) to generate pU5-M2^H37A. The 5xUAS-TATA:M2^H37A region was then PCR amplified using the following primers: 5′-ACACAGGCCAGATGTGGGCCCGTACTTGGAGCGGCCGCA-3′ and 5′-GTCTGGATCATCATCGATGCGGCCGCAAGCCATAGAGCCCACCG-3′. pBH (a gift from Michael Nonet; Washington University, St Louis, MO, USA) was digested with ApaI and NotI, and the Tol2-5xUAS-TATA:M2^H37A;myl7:mCherry plasmid was assembled from the digested pBH and 5xUAS-TATA:M2^H37A PCR product using Gibson Assembly Master Mix (NEB). The entire plasmid was sequenced to confirm ligation fidelity.

To generate the Tol2-elavl3:mCherry-CAAX plasmid, the mCherry- and CAAX-coding regions with an upstream Kozak sequence were PCR amplified using the following primers: mCherry, 5′-TATATTTTCCACCTGCAGATAATTACCGGTGCCACCATGGTGAGCAAG-3′ and 5′-TAAAACGACGGCCAGGATCCACGCGTCTATTACATAATTACACACTTTGTCTTTGACTTC-3′; CAAX, 5′-TATATTTTCCACCTGCAGATAATTACCG GTTCCGCCACCATGGTGAGC-3′ and 5′-AGTAAAACGACGGCCAGGATCCACACGCGTCTATTACATAATTACACACTTTGTCTTTGA CTTCTTTTTCTTCTTTTTAC-3′. Tol2-elavl3:GCaMP6s was digested with AgeI, and Tol2-elavl3:mCherry-CAAX was assembled from the digested plasmid and PCR products using Gibson Assembly Master Mix (NEB). The insert was sequenced to confirm ligation fidelity.

Transgenic zebrafish were generated using Tol2-mediated transgenesis as previously described (Suster et al., 2009). For each line generated, plasmid DNA and Tol2 mRNA were premixed and co-injected into one-cell-stage embryos (12-50 pg of plasmid; 50 pg of mRNA). Fish were raised to adulthood and mated with wild-type AB fish to identify founders with germline transmission, and those yielding F2 generations with monoallelic expression were used to establish transgenic lines. Both heterozygous and homozygous Tg(elavl3:GAVPO) and homozygous Tg(UAS:M2^H37A;myl7:mCherry) lines were used in subsequent studies.

All experimental procedures with GAVPO-expressing zebrafish used a Wratten #29 filter (Kodak) to minimize GAVPO activation by the microscopy light sources required for embryological procedures, monitoring and imaging. For global irradiations, a plate containing E3 medium and embryos was mounted onto a mirrored surface and then fixed onto a Vortex-Genie (Scientific Industries). Dechorionated embryos were irradiated with a blue or white LED light source (TaoTronics), during which the plate was vortexed to ensure that embryos were irradiated from all sides. Light intensity received by the embryos was measured as 3 mW/cm² (1.8 mW/cm² for white LEDs). For individual irradiations, the embryos were either mounted in an injection tray or in 1% low-melt agarose on a slide. Each embryo was irradiated using a Leica DM4500B upright compound microscope equipped with a mercury lamp, a GFP filter (ex: 470 nm, 40 nm bandpass) and a 20×/0.5 NA or a 63× water-immersion objective. To optically target the larval head, the fully open diaphragm was used to irradiate a region spanning the forebrain and the otic vesicle. For focal irradiation, an iris diaphragm was used to limit the targeted region to a 100 μm diameter. Light intensities from the mercury lamp were measured before and after the experiment, and they ranged from 900 to 2400 mW/cm².

Live embryos were mounted in 1% low-melting-point agarose on a glass slide. Embryos were imaged using a Leica M205FA microscope equipped with a SPOT Flex color camera and images were captured using SPOT software (Molecular Devices) or using an Olympus SZX16 microscope equipped with an Olympus DP80 dual-chip camera. Fluorescent images were captured with DP80 in monochrome mode. Pixel intensities were quantified using ImageJ, using the polygon tool on the bright-field image to circumscribe the region of interest. The average mCherry fluorescence intensity within the same region of the corresponding fluorescent micrograph was then calculated.

RNA was extracted from 10-15 embryos per condition at 10 hpf using the RNeasy Micro Kit (Qiagen). Reverse transcription, using olig(dT)₂₀ primers was performed with the SuperScript III First Strand Synthesis System (Invitrogen). The following primers were used to amplify M2 and normalized against eukaryotic translation elongation factor 1 alpha 1, like 1 (eef1a1l1), by standard PCR (55°C, 35 cycles for eef1a1l1 and 40 cycles for M2 amplification): eef1a1l1 forward, 5′-AGAAGGAAGCCGCTGAGATGG-3′; eef1a1l1 reverse, 5′-TCCGTTCTTGGAGATACCAGCC-3′; M2^H37A forward, 5′-GCCGCGAGTATCATTGGGAT-3′; M2^H37A reverse, 5′-TCAGGCACTCCTTCCGTAGA-3′.

Embryos were fixed at the desired timepoint in 4% (w/v) paraformaldehyde (PFA) in PBS at 4°C overnight and then stored in methanol. After rehydration, embryos were permeabilized with acetone and proteinase K, immunostained and imaged with a Leica DM4500B compound microscope equipped with a Retiga-SRV or Prime cMOS camera (QImaging) or Zeiss LSM 700 and LSM 800 confocal microscopes equipped with a MA-PMT. The following antibodies were used: mouse monoclonal anti-HuC/HuD (1:100 dilution, Molecular Probes, A-21271), mouse monoclonal anti-mCherry (1:500 dilution, Abcam, ab125096), rabbit polyclonal anti-active caspase 3 (1:100 dilution, BD Biosciences, BDB559565), rabbit polyclonal anti-influenza A virus M2 (1:100 dilution, GeneTex, GTX125951), goat anti-mouse and goat anti-rabbit antibodies conjugated to Alexa Fluor 488, 555 or 594 (1:200 dilution, Roche).

For M2^H37A and NTR experiments, embryos were raised in E3 medium supplemented with either 100 µg/ml rimantadine or 5 mM metronidazole, respectively. Embryos raised in E3 medium treated with an equivalent amount of DMSO vehicle were used as controls. The embryos were then dechorionated using forceps, fixed in 4% PFA at the desired timepoint, and stored in methanol. Embryos were rehydrated and permeabilized with diluted bleach solution containing 0.8% potassium hydroxide, 3% hydrogen peroxide and 0.2% Triton X-100. After bleaching, the embryos were washed with PBS supplemented with 0.2% Triton X-100 (PBS-X), dehydrated with methanol and stored overnight at −20°C. After rehydration, embryos were repeatedly washed with PBS-X and digested using proteinase K. Embryos were re-fixed in 4% PFA followed by treatment with a pre-chilled solution (2:1) of ethanol and acetic acid. Using the ApopTag Peroxidase In Situ Apoptosis Detection Kit (Millipore Sigma), the samples were incubated in equilibration buffer for 1 h and then TdT reaction buffer overnight at 37°C. Embryos were washed in diluted stop solution and PBS-X and then incubated with a blocking solution containing: 2% (w/v) blocking reagent (Roche), 0.15 M NaCl, 0.1 M maleic acid and 20% (v/v) sheep serum (Sigma #S2263). The embryos were incubated with horseradish peroxidase-conjugated anti-digoxigenin antibody overnight at 4°C and washed with PBS-X. Embryos were stained with fresh DAB mix (Metal Enhanced DAB Kit Thermo Fisher Scientific #34065), washed and fixed in 4% PFA. The embryos were stored and imaged in a solution of 90% glycerol in PBS.

Whole-mount in situ hybridization was performed according to standard protocols (Broadbent and Read, 1999). Zebrafish cDNA was prepared from RNA extracted from embryos as previously described (Payumo et al., 2015). T3-promoter containing PCR products were then amplified with the designated primers (T3 sequence underlined): elavl3, 5′-CACCTCACGCATCCTGGTAA-3′ and 5′-ATTAACCCTCACTAAAGGGATGGTCTTGAACGAGACCTGC-3′; GAVPO Probe 1, 5′-AAGAAAAACCGAAGTGCGCC-3′ and 5′-ATTAACCCTCACTAAAGGGATGTTTCATCTCGCACCGGAA-3′; GAVPO Probe 2, 5′-GCTCTGATTCTGTGCGACCT-3′ and 5′-ATTAACCCTCACTAAAGGGAATCAGCATGGGCTCAGTTGT-3′. RNA probes were in vitro transcribed from the PCR products using the MEGAscript T3 Transcription Kit (Ambion), substituting nucleotides from the kit with digoxigenin-UTP (Roche). For GAVPO staining, both probes were used simultaneously in separate tubes. Embryos were imaged using a Leica M205FA microscope equipped with a SPOT Flex color camera and images were captured using SPOT software (Molecular Devices).

Heterozygous Tg(elavl3:GAVPO;UAS:M2^H37A;myl7:mCherry) zygotes were co-injected with plasmid DNA encoding elavl3:mCherry-CAAX (25 pg/embryo) and Tol2 mRNA (50 pg/embryo). The resulting embryos were raised until 48 hpf, irradiated for 8 h and then fixed 12 h after irradiation in 4% PFA for 1 h at room temperature. The embryos were then mounted laterally in 1% low-melting agarose and imaged with a Zeiss LSM 700 confocal microscope equipped with a 63×/0.5 NA water-immersion objective. Z-stacks were generated from images taken at 2- to 5-μm intervals, using the following settings: 1024×1024 pixels, 8 speed, 4 averaging. Neurons were scored as intact when mCherry fluorescence was localized only to the plasma membrane and excluded from the cytoplasm. Counting was performed by an observer blinded to the experimental conditions. For spinal cord measurements, maximum intensity projections were made in ImageJ, and three measurements were made for each projection and averaged to determine the axonal tract width.

For all zebrafish experiments, at least two breeding tanks, each containing three or four males and three to five females from separate stocks, were set up to generate embryos. Embryos from each tank were randomly distributed across tested conditions, and unfertilized and developmentally abnormal embryos were removed prior to irradiation or compound treatment. No statistical methods were used to determine sample size per condition. Experimental statistics for the phenotypes reported in Figs 1-4 are provided in Tables S1-S4, respectively. χ² analyses were conducted for the phenotypic distributions reported in Figs S1 and S4. For the cell counting and axonal tract width measurements reported in Fig. 5, the scorer was blinded to treatment conditions. Values for individual fish are plotted, and each distribution was assessed using the Shapiro-Wilk test and determined to be non-normal. A Kruskal-Wallis test with a two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli was used to determine differences between all conditions. P-values were corrected for multiple comparisons testing by controlling the false discovery rate to 0.05.

We thank Shannon Linch for assistance with mating and genotyping. Imaging was supported in part by the University of Wyoming's Integrated Microscopy Core (NIH P20 GM121310).

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.183640.reviewer-comments.pdf