The RNA helicase Ddx52 functions as a growth switch in juvenile zebrafish

Animal development is an extremely robust process that generally continues without major delays, regardless of perturbations and obstacles (Mestek Boukhibar and Barkoulas, 2016; Waddington, 1957). Once initiated, the sequence of developmental processes proceeds with great momentum, like a train on a railway, until an organism's mature form is reached. Yet, some animals are known for their exceptional ability to suspend abruptly their transition to adulthood for extended periods. For instance, the monarch butterfly (Danaus plexippus) typically grows from an egg to a fertile adult in 1 month during the summer; but if an individual emerges in the fall, the butterfly will pause its development at juvenile stages for 6-8 months as it migrates several thousand miles to warmer temperatures (Reppert and de Roode, 2018). Although it is known that insect species can undergo this type of growth delay (Tatar and Yin, 2001), the question of whether vertebrates are able to stay physically active, yet pause their organismic growth at juvenile stages without detrimental consequences is still open. Discovery of such a ‘stay put’ switch and delineation of the enabling mechanisms in vertebrate models may provide crucial insights into the regulatory control of growth and aging in humans.

Laboratory zebrafish exhibit a well-defined developmental trajectory consisting of four major stages: embryo, larva, juvenile and adult (Parichy et al., 2009). Despite minor differences in developmental timing between strains and rearing conditions, most individuals readily reach adulthood within 3 months under regular laboratory settings. Researchers routinely stage developing zebrafish according to visible external body features, such as incremental body lengths and heights, appearance of scales and fin rays, and development of pigment patterns (McMenamin et al., 2016; Parichy et al., 2009; Singleman and Holtzman, 2014). This ease of staging and the consistency of zebrafish development during post-embryonic stages has enabled researchers to model the mechanisms of allometric and ontogenetic growth, pigment pattern formation, and the onset of aging in a vertebrate animal (Chopra et al., 2019; Daane et al., 2018; Harris et al., 2020; Iovine et al., 2005; Kujawski et al., 2014; Li et al., 2020; McMenamin et al., 2014). Furthermore, it is worth noting that zebrafish are poikilothermic and thus amenable to temperature-sensitive screens (Chen and Poss, 2017). This key feature offers a unique opportunity to study essential genes that may have previously unidentified functions at the late developmental stages or during adult tissue regeneration.

From a forward genetic screen targeting key regulators of tissue regeneration (Chen et al., 2015), we identified a zebrafish mutant with temperature-sensitive defects in juvenile growth (pan). Long-term tracking of juvenile-staged individuals after return to the permissive temperature indicated that the pan phenotypes are fully reversible. No differences in anatomy or reproductive capability were apparent between the developmentally paused and released pan fish and their respective heterozygous siblings at the adult stage. Through positional cloning and classical complementation assays, we determined that pan phenotypes arise from an alteration in the gene encoding DEAD-Box Helicase 52 (ddx52). This mutation causes a splicing error that leads to a deletion of 25 amino acids near the C terminus of the protein. We further determined that, as a probable RNA helicase, Ddx52 activity is specifically required for maintaining the level of 47S precursor ribosomal RNA in live animals, thereby affecting bulk RNA transcription in proliferating tissues. Moreover, either genetic silencing of the ddx52 ortholog in fruit flies or direct pharmacological inhibition of bulk RNA transcription in zebrafish and mice causes similar delays in growth. Thus, our study introduces a gene that is essential for post-embryonic growth in vertebrates and invertebrates, leading us to propose a conserved mechanism for modulating juvenile-to-adult transition trajectories.

As part of a systematic effort to identify key regulators of tissue regeneration, we performed a three-generation genetic screen for zebrafish mutants defective in tailfin regeneration (Chen et al., 2015). The screen was conducted on adult-stage fish under the restrictive temperature and aimed to isolate temperature-sensitive alleles for conditional studies (Johnson and Weston, 1995; Poss et al., 2002). Among the collection of regeneration-deficient mutants, we noticed that one, which we later named pan after Peter Pan, displayed impairments in both regeneration and growth. At 34°C, adult pan fish had greatly reduced fin growth at 7 days post-amputation (dpa) with complete penetrance (15/15; Fig. S1A). When kept at the permissive temperature (26°C), the pan regenerate length was normal or close to normal compared with heterozygous siblings (Fig. S1B), suggesting that pan is a bona fide temperature-sensitive (ts) allele. To determine whether pan affects specific phases of regeneration (i.e. wound healing, blastema formation or the outgrowth phase), we transferred animals from 26°C to 34°C at 0, 1, 2 or 4 dpa and examined the lengths of fin regenerates at 7 dpa. We found that transfer at any of the examined time points caused a significant reduction in regenerate length (Fig. S1C). Because the percentage reduction in fin length was less severe with later transfers (Fig. S1C), we concluded that pan activity is essential throughout all phases of regeneration. As tailfin regeneration is a complex morphogenetic process, we next examined whether pan is also required for simpler forms of repair, such as healing of an incision wound. As such, we made a straight cut in the inter-ray region of the tailfin tissue (Fig. S1D). This type of lesion only severs the inter-ray mesenchyme and epidermis, so the healing process takes just 2 days to complete in zebrafish (Chen et al., 2015). We found that both pan and pan/+ siblings exhibited complete healing of these wounds by 2 days post-incision (dpi) at 34°C (Fig. S1E). Thus, pan activity is dispensable for healing of a simple wound, which does not involve blastema formation. To determine whether pan is required for adult survival and tissue maintenance, we transferred fully developed animals at 3 months of age to 34°C for weekly monitoring. We found that both pan and pan/+ adults showed similar reductions in survival at the elevated temperature (survival was decreased by ∼30% after 21 days; Fig. S1F). Unlike previously reported sde1 or other mutants (Chen et al., 2015; Wang et al., 2019; Wills et al., 2008b), we detected no signs of atrophy in tailfin tissues by 21 days post-treatment (dpt), except for a loss of pigment stripe (n=12/15; Fig. S1G). Thus, we conclude that the pan locus does not have an active role in maintaining the integrity of adult tissues.

To determine whether the pan locus is generally required for tissue morphogenesis, we examined its effects during juvenile growth. We transferred 4-week-old animals to 34°C for long-term monitoring (Fig. 1A). Of note, defining specific developmental stages of zebrafish based on age is often not accurate. Thus, we assessed the progression of organismic growth by both fish length and physical features, as previously described (McMenamin et al., 2016; Parichy et al., 2009; Singleman and Holtzman, 2014). Intriguingly, we found that growth of pan mutants was much reduced compared with the growth of heterozygous siblings. By 28 dpt, 8-week-old pan mutants were on average 27% shorter by standard length (SL) and 42% shorter in terms of height at anterior of anal fin (HAA) (Parichy et al., 2009) (Fig. 1B,C). In contrast, only 9% differences in SL could be detected between pan mutants and their heterozygous siblings when maintained at 26°C (Fig. S1H,I). To determine whether the mechanism underlying pan-induced defects in juvenile growth is specific to this mutant, we examined the growth phenotype of another regeneration mutant, sde1, which carries a ts allele for the gene encoding the extracellular matrix component lamb1a (Chen et al., 2015). In contrast to pan, we detected no delay in juvenile growth in sde1 zebrafish, despite the expected evident signs of fin atrophy under the restrictive temperature (Chen et al., 2015) (Fig. 1B; bottom right). Thus, we conclude that the regulatory mechanisms governing tissue regeneration and juvenile growth are uncoupled in sde1, whereas juvenile growth is profoundly impacted by pan activity.

To determine whether pan is also required for late juvenile growth, we transferred 6-week-old animals to 34°C for 28 days. Again, only pan, but not sde1, produced a significant delay in late juvenile growth, according to SL and HAA measurements (Fig. 1D-F). The growth delay phenotype in pan also corresponded to delayed development of several physical adult characteristics, such as lack of adult barbels, immature pigment patterns, and absence or incomplete formation of calcified bone structures (Fig. 1G,H). Taken together, our data suggest that pan encodes a genetic determinant capable of attenuating growth at distinct juvenile stages. The effect of the pan mutation on growth at the restrictive temperature appears to be specific to the mutation, as we failed to detect similar phenotypes in other ts mutants within our collection (Chen et al., 2015; Wang et al., 2019; Whitehead et al., 2005).

Despite their reduced size and immature physical appearance, we detected no additional externally visible developmental anomalies or anatomical scaling defects (Chopra et al., 2019; Daane et al., 2018; Harris et al., 2020; Kujawski et al., 2014) in pan juveniles maintained at the restrictive temperature (Fig. 1). To identify long-term or possibly permanent defects in these individuals, we first inactivated pan for 28 days (from 4 to 8 weeks of age) and then monitored the growth-arrested fish at the permissive temperature for a 12-week period (Fig. 2A,B). Notably, we found that pan mutants readily resumed growth, eventually catching up with their heterozygous siblings by 20 weeks of age without notable phenotypic differences (Fig. 2B-D). To chart the growth trajectories of individual fish, we repeated the experiments and tracked each juvenile-staged fish over the entire recovery phase at the permissive temperature (one animal per 1.5-l tank; Fig. S2A,B). We found that restoration of pan activity accelerated the animal growth rate to a maximum of 2.4-fold over controls during a 4-week period (Fig. S2C). Thus, restoring pan activity is alone sufficient to jumpstart growth by raising the growth trajectory of an affected individual. To determine the maximum length of pause after which pan growth may resume, we inactivated pan for 56 consecutive days starting at 4 weeks of age. Remarkably, restoring pan activity in 3-month-old juveniles similarly boosted their growth, regardless of the long-term pause (Fig. S2D-F). Of note, key developmental features and growth remain coupled in pan mutants during the recovery phase, as determined by comparing the mutant phenotype with the normal table of postembryonic zebrafish development (Parichy et al., 2009) (SL and images of tailfin development and pigment pattern formation are shown in Fig. S3). To determine whether the growth-paused animals are able to acquire normal adult features after release, we performed two fertility assays on recovered pan individuals. We detected no significant differences in reproductive fitness, either in terms of male mating success rate or average number of viable progeny per mating event (Fig. 2E,F). Taken together, these results led us to conclude that pan may encode a switch-like regulator of growth with an activity that governs the juvenile-to-adult stage transition in zebrafish.

To identify the underlying mutation associated with the pan phenotype, we conducted whole-exome sequencing of genomic DNA from pan×pan/+ crosses (as described by Chen et al., 2015), followed by SNPtrack analysis of regions with high homozygosity (Leshchiner et al., 2012). This analysis yielded a single locus of 12.1 Mb on chromosome 21 (Fig. 3A). Phenotyping and genotyping of 757 adult animals from several mapping crosses revealed two closely linked single nucleotide polymorphisms (SNPs) in a 283 kb region (Fig. 3A). Next, cDNA sequencing of three candidate genes within this region (heatr6, ddx52 and traf4b; Fig. 3A) led us to focus on two of the candidates: HEAT repeat containing 6 (heatr6) and DEAD (Asp-Glu-Ala-Asp)-box Helicase 52 (ddx52). Notably, both genes remain uncharacterized in vertebrate systems. Within heatr6, we detected two non-synonymous mutations, each causing distinct amino acid changes (M76L and G530S, shown in Fig. S4A). In contrast, we did not identify any non-synonymous mutations in ddx52 exons. However, the cDNA displayed a splicing error in exons 13 and 14 that is expected to produce an in-frame deletion of 25 amino acids from the C-terminal domain of the putative protein (Fig. 3B,C). To determine which of these genes is responsible for the pan phenotype, we applied a CRISPR/Cas9-based approach (Talbot and Amacher, 2014) to generate stable deletion lines for complementation assays. We first established two deletion lines on a wild-type background for heatr6 (heatr6^del#1 and heatr6^del#2 in Fig. S4B-D). After crossing F1 heterozygous animals with pan, we detected no regeneration deficits at the restrictive temperature in any adult heart6^del/pan individuals (n=0/19 for the heatr6^del#1/pan and 0/35 for the heatr6^del#2/pan; Fig. S4E). Thus, heatr6 is not the functional gene in pan. Next, we generated a whole-gene deletion line of ddx52 (Fig. 3C,D). In sharp contrast to the results with heatr6-deletion lines, we found that each of the ddx52^del/+×pan/pan crosses consistently generated high ratios of progeny with regeneration defects at the restrictive temperature. We determined that adult animals with and without the defects invariably matched their predicted genotype (n=23/23 for pan/+ and 12/12 for ddx52^del/pan; Fig. 3E). Thus, the ddx52 null allele failed to complement pan. Based on high-resolution genetic and physical mapping, extensive Sanger sequencing of protein-coding exons, and complementation assays with three independently created stable deletion lines, we conclude that ddx52 is the gene responsible for pan phenotypes.

To determine whether ddx52 is required for early animal development, we examined progeny from several ddx52^del/+×ddx52^del/+ crosses. Consistent with Mendelian inheritance ratios, we found that about 25% of embryos displayed a shortened body length and a severe pericardial edema at 6 dpf (n=29/118 at 26°C; Fig. 3F), indicating that ddx52 has an essential role in supporting early embryonic development. Taken together, these findings suggest that the truncated form of the Ddx52 protein likely acts as a hypomorph at the restrictive temperature to impair adult tailfin regeneration and juvenile growth.

Because ddx52 activity appears to have a decisive role in regeneration and growth, we tested the idea that overexpression of ddx52 may be sufficient to promote organismic growth. By creating a stable transgenic line Tg(ubi:EGFP-2A-ddx52) with constitutive high expression of ddx52 in the whole body (19.3-fold; Fig. S5A-C), we determined that the body length is slightly increased at 8 dpf, but the difference becomes insignificant later in life compared with non-transgenic siblings (Fig. S5D). Thus, we speculated that ddx52 may either function in a switch-like manner or it might act synergistically with yet-to-be-identified components in order to regulate regeneration and growth. It is equally possible that rapid growth during juvenile stages may have physiological limits that are difficult if not impossible to surpass.

Because of its substantial influence on zebrafish growth, we hypothesized that the effect of ddx52 on growth might extend to other distantly related animal species. To examine this idea, we tested the impact of ddx52 silencing on the development of the fruit fly. Of note, the major protein domains encoded by ddx52 orthologs are conserved (Fig. 4A,B), despite the fact that invertebrates and vertebrates diverged 700 million years ago (Kumar et al., 2017; Wray, 2015). The fly ddx52 ortholog, CG5589, remains an uncharacterized gene (http://flybase.org/reports/FBgn0036754.html). To determine whether fly ddx52 similarly affects developmental trajectories, we acquired two independently created RNAi knockdown transgenic lines (UAS-CG5589^{RNAi BDSC} and UAS-CG5589^{RNAi VDRC}) to enable specific silencing of fly ddx52. To achieve global knockdown of ddx52 mRNA expression, we crossed the RNAi lines with a driver line that has whole-tissue expression of GAL4 (Perkins et al., 2015). The combination effectively downregulated the fly ddx52 level, as determined by RT-qPCR (decrease of 36% in UAS-CG5589^{RNAi BDSC}; Fig. 4C). As a control for non-specific effects of RNAi, we first determined that most animals (∼97%) emerging from act-GAL4×UAS-gfp^RNAi crosses reach adult stages with genotypes occurring at expected ratios (n=231; Fig. 4D,E). In contrast, both act-GAL4×UAS-CG5589^RNAi crosses consistently generated high ratios of progeny that failed to grow into adult flies. By 25 days post-fertilization, up to 56% of the progeny remained stuck either at larval or pupal stages (n=331 and 284 from three independent crosses; Fig. 4D,E). We then genotyped flies that had achieved adulthood and found that none of the act>UAS-CG5589^RNAi double transgenic animals was present in the adult pool (n=0/308; Fig. 4E). Thus, based on similar phenotypes from two independently created RNAi lines, we conclude that the previously uncharacterized fly ddx52 is required for the larva-to-adult stage transition during development. These findings are consistent with the notion that ddx52 has a profound and evolutionarily conserved role in supporting the progression of organismic growth.

Despite a lack of research on ddx52 in animal models, studies on the yeast homolog and other DEAD-box RNA helicase family members suggest it may have potential roles in ribosome biogenesis, ribosomal RNA (rRNA) synthesis and RNA editing (Khoshnevis et al., 2016; Linder and Jankowsky, 2011; Song et al., 1995; Venema et al., 1997). Specifically, mammalian Ddx52, along with other DEAD-box containing RNA helicases, affects the quantity of 47S precursor rRNA (pre-rRNA) in a cell culture system (Zhang et al., 2011). As the first product of the rRNA synthesis cascade, 47S pre-rRNA has a short lifespan and is rapidly processed into the mature rRNAs (i.e. 18S, 28S and 5.8S) once transcribed (Fig. 5A). Thus, one can examine the quantity of 47S pre-rRNA as a readout for the rates of its transcription and the overall rRNA synthesis (Cui and Tseng, 2004; Zhang et al., 2011). To determine whether ddx52 similarly regulates 47S pre-rRNA in live animals, we transiently inactivated Ddx52 activity in pan mutants and detected the removal of the 5′-external transcribed space (5′ETS) on 47S pre-rRNA by RT-qPCR (Bouffard et al., 2018) (Fig. 5A,B). Intriguingly, we found that a brief 6-h inactivation of Ddx52 in regenerating tailfin tissues markedly reduced the level of 47S pre-rRNA by 44% in the 4 dpa blastema (Fig. 5C). In contrast, the treatment had little or no effect on the 18S rRNA level, which presumably is much more stable in vivo (Fig. 5D). Notably, the Ddx52 influence on 47S pre-rRNA is likely to be specific and direct, as the release of the inactivation rapidly restored the pre-rRNA expression in less than 6 h (Fig. 5B,C).

Because up to 60% of total cellular RNA synthesis is devoted to rRNA transcription (Warner, 1999), we further hypothesized that Ddx52-mediated, transient blocking of the pre-rRNA synthesis should have a drastic impact on the level of bulk RNA transcription. To test this idea, we established a 5-ethynyl uridine (EU) incorporation assay (Jao and Salic, 2008) in regenerating tailfin tissues to detect dynamic changes in total RNA synthesis. We found that abruptly switching-off Ddx52 activity had an immediate influence on the number of EU-positive cells in the 4 dpa blastema (decrease of 45%; Fig. 5E,F). This impact is not generalizable to other ts mutants, as transient inactivation of lamb1a activity in sde1 mutants had no measurable effects on the level of 47S pre-rRNA or the number of EU-positive cells (Fig. 5C,F). As additional controls for specificity of the effect, we determined that transient inactivation of Ddx52 for 6 h had no effect on the levels of shha and lef1, i.e. specific messenger RNA species that are key for tissue regeneration (Fig. S6A,B); nor did it affect the level of DNA synthesis in proliferating tissues compared with mem mutants carrying a ts allele of a DNA polymerase subunit (Wang et al., 2019) (Fig. 5G,H). Taken together, these experiments using direct and temporal modulation of Ddx52 activity in intact and live animals led us to conclude that Ddx52 is constantly required for the synthesis of 47S pre-rRNA; thus, it maintains bulk RNA transcription in real time.

We then determined the spatiotemporal expression pattern of ddx52 in regenerating tailfin tissues by RT-qPCR and RNA in situ hybridization assays. We found that ddx52 is rapidly induced by 6 h post amputation (hpa) and remains upregulated until at least 48 hpa in the mesenchymal compartment (Fig. S6C,D). Intriguingly, the induction of ddx52 transcripts upon wounding appears to be a common occurrence in many regenerative tissues, including tailfin, heart, liver, muscle, lateral line and different glial cells (http://www.zfregeneration.org/) (Nieto-Arellano and Sánchez-Iranzo, 2018). In our system of tailfin amputation, we determined that the induction of ddx52 is not affected by injury level, wound healing perturbations, or Fgf and Wnt/β-Catenin signaling pathways (Fig. S6E-H). Thus, although we were unable to identify a specific wound signal responsible for ddx52 activation, stimulation of ddx52 function during regeneration seems to be a conserved phenomenon across distinct tissue types and organs.

Our finding that the main mechanism by which ddx52 affects regeneration and growth is through suppression of 47S pre-rRNA synthesis and subsequent reduction in protein translation (Fig. S6I,J) was predictable based on the expected functions of the gene. Yet, we were intrigued by the full reversibility of the growth pause caused by the acute suppression of RNA transcription. To test whether direct, pharmacological suppression of bulk RNA transcription could reproduce pan phenotypes during regeneration and growth, we treated wild-type animals with actinomycin D (Act D), a potent inhibitor of RNA transcription and common cancer chemotherapeutic agent (Bensaude, 2011; Liu et al., 2016). Of note, Act D may have multiple mechanisms of action; however, it is most effective at blocking the activity of RNA polymerase I, which only transcribes 47S pre-rRNA (Bensaude, 2011; Russell and Zomerdijk, 2006). To determine the effective dose of Act D for suppressing bulk RNA transcription, we performed the EU assay in regenerating tailfin tissues. Notably, we found that a single dose of 100 µg/ml Act D markedly reduced the number of EU-positive cells (decrease of 93%; Fig. 6A,B), mimicking the effect of pan on bulk RNA transcription (Fig. 5F). To test whether Act D treatment can reversibly pause tissue growth during regeneration, we intraperitoneally injected adult zebrafish upon tailfin amputation. As expected, the Act D-treated animals had a smaller regenerate at 4 dpa compared with vehicle-treated animals (Fig. 6C,D). Intriguingly, the effect of Act D on regenerates was transient and fully reversible, as we detected no morphological or size differences at 21 dpa when regenerates were compared with respective pre-injury fin structures (15/15; Fig. 6D-F). In contrast, when we treated animals with hydroxyurea (HU), a potent inhibitor of DNA replication, the treatment similarly reduced fin regenerate size at 4 dpa, yet many regenerates (∼93%) failed to restore the pre-injury shape and size by 21 dpa (Fig. 6D-F). These HU-induced changes in fin regenerates are likely to be permanent, as previously shown (Wang et al., 2019). To examine further the reversible effect of Act D on organismic growth, we intraperitoneally injected 11-week-old wild-type juveniles. As expected, the Act D-treated animals had significantly reduced growth at 21 dpt compared with vehicle-treated siblings (decrease of 8.4% in SL and 11% in HAA; Fig. 6G-J). Yet, consistent with the regeneration and pan phenotypes, the effect of Act D on growth was transient and fully reversible; we detected no morphological or length differences between vehicle- and Act D-treated fish at 49 dpt (Fig. 6H-J). Taken together, these results support the notion that direct targeting of 47S pre-rRNA or bulk RNA transcriptions may allow non-detrimental, reversible pausing of adult regeneration and juvenile growth. We further speculate that the RNA-based mechanism is at least somewhat specific, as targeting DNA replication machineries similarly affects tissue growth but fails to reproduce the reversible effect.

These findings led us to investigate whether the growth pausing mechanism may be generalizable to higher vertebrates, such as mammals. Therefore, we examined next the effect of Act D on a commonly used laboratory mouse strain (C57BL/6). We administered a series of 0.25 mg/kg doses to 4-week-old juvenile mice. As expected, the treatment drastically impaired growth of the animals at 11 dpt, as determined by a marked reduction in body length and tail length (Fig. 7A-C). Of note, we measured ‘tail length’ as an unambiguous readout reflecting the overall development of the vertebral column (Lee et al., 2017). To determine whether the treatment does indeed cause a developmental pause, we performed a high-resolution micro-CT scan of the growth-paused mice and their respective vehicle-treated siblings at 11 dpt. A significant reduction in skull length and size was observed (Fig. 7D-F), supportive of a bona fide developmental arrest. Intriguingly, yet consistent with our findings in pan and Act D-treated zebrafish, we found that the affected individuals readily resumed growth and development upon release from treatment, eventually catching up with the vehicle-treated animals by 46 dpt in a repeated experiment. We detected no significant differences in skull length, skull size, tail length, daily food intake, body length or body weight between vehicle- and Act D-treated animals at 46 dpt (Fig. 7D-K). To determine whether Act D may cause any minor developmental defects that may be missed in a gross examination, we further examined the micro-CT scan captured at 46 dpt. Notably, despite the fact that there was no marked difference in the femur length, the end-point scans revealed a slight yet statistically significant reduction in the tibia length of Act D-treated animals at 46 dpt (decreased by 3.4%; Fig. 7L-N). Taken together, the similar findings from zebrafish and mouse models suggest that short-term, direct suppression of 47S pre-RNA and/or bulk RNA transcription by Act D treatment has a profound but largely reversible impact on juvenile growth across different vertebrate species.

Because it encodes a putative RNA helicase, the effect of ddx52 on growth may be mediated in part through its direct influence on specific mRNA transcripts. Despite our observation of similar growth phenotypes upon pharmacological inhibition of bulk RNA transcription, it is possible that Ddx52 has important, specific mRNA targets or that it regulates aspects of rRNA metabolism other than the synthesis of pre-rRNA. Consistent with this notion, Bennett et al. (2018) showed that genetic deficiency of ddx27, another DEAD-box RNA helicase, similarly affects pre-rRNA synthesis as well as the translation of a specific subset of mRNA transcripts in zebrafish embryos. Notably, the ddx27 phenotypes appear to be tissue specific, predominantly affecting skeletal muscle growth and regeneration. This study and ours highlight the complexity faced when pinpointing the exact in vivo targets of an individual RNA helicase. Thus, we do not expect that ddx52 is a master regulator of animal development with an exclusive role on 47S pre-rRNA synthesis. Rather, our findings suggest that intricate control of RNA transcription or metabolism may offer a unique opportunity to safely suspend post-embryonic growth for an extended period.

Besides specific genetic factors, the growth of zebrafish may be affected by certain rearing conditions, including nutrient supply, oxygen levels, temperature, and aquarium density (Kamei et al., 2018; Kimmel et al., 1995; Wills et al., 2008a). Upon release from adverse, low-growth conditions, affected individuals may enter a catch-up period of rapid growth until the adult form is reached. Thus, it will be interesting to determine whether ddx52 or other DEAD-box RNA helicase family members (Zhang et al., 2011) might influence growth progression under known low-growth conditions, acting as a downstream effector of the environmental influences on animal growth and development. Of note, when manipulating Ddx52 activity, the progression of growth and development remained coupled in pan mutants (Fig. S3). This key feature is distinct from growth mutants that are defective in hormonal signaling, such as the gh1 and duox mutants (Chopra et al., 2019; McMenamin et al., 2013). In both of those mutants, growth is severely inhibited, but the animals still display many developmental characteristics of adults.

When juvenile monarch butterflies suspend their transition to a fully adult stage, their lifespan is quadrupled with respect to conspecifics that do not delay the transition (Herman and Tatar, 2001). Even though we detected no morphological or functional anomalies in released adult pan fish (Fig. 2), it is not known whether a 2-month delay in the developmental course will have any lasting impact on the animal's lifespan (lab-raised zebrafish have a generation time of 2-3 months, but may live up to 4 years). Moreover, we found that prolonged exposure to the restrictive temperature negatively affects the survival of treated animals, regardless of their genotype (Fig. S1F). Thus, we have limited our use of the pan mutant to examining long-term growth suspension. Nevertheless, it will be interesting to determine whether intermittent inactivation of Ddx52 or RNA transcription may further lengthen juvenile stages, thereby altering the subsequent adult life-course trajectory of affected individuals.

In this work, we describe one scenario in which animal growth may be suspended or delayed at juvenile stages. The pan mutants are generally as physically active as juvenile-staged individuals at the restrictive temperature. Despite some similarities, the animal state studied here is distinct from the state of ‘embryonic diapause’ (Hand et al., 2016; Hussein et al., 2020; Renfree and Fenelon, 2017). As a survival strategy triggered by harsh environmental conditions, such as temperature, drought, or scarcity of food, diapause in embryonic animals involves physical and metabolic inactivity, and the animals may remain completely dormant for extremely long periods of time. The mechanisms of diapause and diapause-like states are currently being examined, and an intriguing study from Romney et al. (2018) has demonstrated that direct perturbation of vitamin D signaling can induce a diapause-like state in embryos of killifish and zebrafish. By incubating fertilized embryos in dafadine A, an inhibitor of vitamin D3 synthesis, the robust development of zebrafish embryos could be suspended for 2 days (Romney et al., 2018). In another recent study from Hu et al. (2020), the authors identified a specific chromatin regulator, CBX7, that is essential for maintaining the level of metabolism and the diapause state in killifish. Additionally, inhibition of mTOR, the mammalian target of rapamycin, has been reported to induce a diapause-like state in pre-implanted mouse embryos and cultured blastocysts (Bulut-Karslioglu et al., 2016; Hussein et al., 2020). Intriguingly, the underlying pausing mechanism is tightly associated with a robust, global suppression of RNA transcription (Bulut-Karslioglu et al., 2016), reminiscent of the effects we saw with Ddx52 inactivation or Act D treatment. Thus, it will be interesting to determine whether the mechanism we described here – an acute, direct suppression of RNA transcription or pre-rRNA synthesis – affect developmental timing in different contexts.

In summary, we provide a proof of concept that the activity of a single gene can dictate post-embryonic growth of the whole animal in a vertebrate model. Our findings regarding ddx52 and its potent effect on RNA transcription highlight an evolutionarily conserved mechanism that could be hijacked to achieve robust, yet reversible control of ontogenetic growth in complex animals.

Animal experiments were approved by and performed in accordance with guidelines provided by the Institutional Animal Care and Utilization Committee (IACUC) at Academia Sinica. For zebrafish, the Ekkwill strain of zebrafish was maintained at four to six individuals per liter at 26°C. Animals at 4 weeks to 12 months of age were used for experiments. Of note, for experiments with juvenile animals, gender identity was not determined owing to lack of external sex characteristics. pan zebrafish are also referred to as pd335 (ddx52^pd335). Deletion lines for heatr6 and ddx52 are referred to as as58 (heatr6^del#1), as59 (heatr6^del#2) and as60 (ddx52^del). The ddx52 overexpression line is referred to as as64, Tg(ubi:EGFP-2A-ddx52). lamb1a (sde1), pola2 (mem) and fgf20a mutants (dob) were described previously (Chen et al., 2015; Wang et al., 2019; Whitehead et al., 2005). For fruit flies, Actin (act)-GAL4/CyO (B#4414), UAS-gfp^RNAi (B#41552), UAS-CG5589^{RNAi BDSC} (B#32334) and UAS-CG5589^{RNAi VDRC} (V#44322) were obtained from Bloomington Drosophila Stock Center (BDSC) or Vienna Drosophila Resource Center (VDRC). Fly stocks were maintained at 22-25°C on standard medium. To mimic the effect of a hypomorphic allele, offspring of GAL4 and RNAi were raised at 18°C, and the numbers of each genotype among adult flies, pupae and larvae were counted manually. For mice, 4-week-old C57BL/6JNarl mice were obtained from the National Laboratory Animal Center, Taiwan. The mice were maintained in a specific pathogen-free room with controlled temperature and moisture, and a 14-h light/10-h dark cycle.

For tailfin amputation, 50% of the tailfin was removed by a single cut with a razor blade. Animals were anesthetized in fish water containing tricaine (0.2 mg/ml) until fully sedated. Fin regenerate length was measured along the 2nd lateral bony ray from the dorsal side. For tailfin incision injuries, cuts were made under a dissecting microscope with a razor blade to the inter-ray region; the incision length was approximately two-thirds of the proximodistal length of the tailfin. To inactivate ddx52 activity in pan zebrafish, mutants were shifted to a constant temperature at 34°C for the time period indicated in each figure. Of note, when shifted to 34°C at 5 wpf, pan/+ and pan had similar survival rates at 9 wpf (wild type, 80%; pan/+, 73%; pan, 73%; n=30 each). Whole-animal and tailfin images were captured using a Leica M205 stereomicroscope with a high-resolution camera (Leica DFC7000T). Large-field, whole-animal images were stitched together using Image Composite Editor (Microsoft Research). Histology images were captured with either a compound microscope (Leica DM750) or a laser scanning confocal microscope (Leica SP8; HC PL APO 20×/0.75 CS2, WD 0.62 mm).

Mutagenesis, regeneration mutant screens, initial genetic and physical mapping were described previously (Chen et al., 2015). Primer sequences for detecting SNP37688707, SNP37971903 and SNP38867230 (shown in Fig. 3A) are listed in Table S1. SNP38867230 was used to detect the ddx52^pan allele in high resolution melt (HRM) assays.

Tg(ubi:EGFP-2A-ddx52)^as64 was generated with a transgenic construct consisting of a 3.5 kb ubi promoter (Mosimann et al., 2011), the EGFP-2A, and ddx52 cDNA sequences. The construct was flanked with I-SceI sites to facilitate transgenesis.

sgRNAs targeting the heatr6 and ddx52 genomic regions were designed using CHOPCHOP (http://chopchop.cbu.uib.no/) (Labun et al., 2016; Montague et al., 2014). Primer sequences used to generate sgRNA templates and detect deletions are listed in Table S1. sgRNAs were synthesized from PCR fragment templates using the MEGAshortscript T7 Transcription Kit (Invitrogen, AM1354), as described (Talbot and Amacher, 2014). sgRNAs (200-300 ng/μl in 2 μl each) targeting the heatr6 or ddx52 genomic regions were co-injected with 2 μl Cas9 protein (1 μg/μl; PNA Bio, CP01-200) into one-cell-stage zebrafish embryos. Genetic lesions were detected by PCR using the primers listed in Table S1. Positive F0 clutches were raised to adulthood and outcrossed with wild-type animals. To establish stable deletion lines, F1 animals were raised to adulthood for genotyping. Complementation assays were conducted using heterozygous F1 animals with respective deletions.

To determine the mating success rate, adult male animals were assessed four times at 2-week intervals by pairing with wild-type females. The ratio was determined as (number of spawned males/total number of examined males)×100%. Spawned eggs from each mating pair were collected, counted and recorded for viability. The number of successfully developed embryos was recorded as the viable progeny number. Results from four independent crosses were pooled for comparison.

For EU incorporation assays, 10 μl EU solution (Invitrogen, E10345; 10 mM in 50% DMSO in PBS) was intraperitoneally injected into each adult zebrafish 6 h prior to collection of tailfin regenerates. Tissues were fixed with 4% paraformaldehyde (Sigma-Aldrich, P6148) in PBS at 4°C overnight and processed as described, with some modifications (Jao and Salic, 2008). After fixation, fin tissues were washed twice with PBST (PBS with 0.1% Tween-20) for 5 min each time and transferred into methanol. Fin tissues were washed in methanol twice, for 10 min each time, and transferred into PBST. Next, tissues were incubated in freshly prepared Click-iT staining solution (1 mM CuSO₄, 50 mM ascorbic acid, 10 μM fluorescent azide and 100 mM Tris buffer, pH 8.0) for 1 h at room temperature before washing with PBST four times, 5 min each. Finally, fin tissues were mounted in Fluoromount G for imaging by confocal microscopy (Leica SP8; HC PL APO 20×/0.75 SC2, WD 0.62 mm).

5-Ethynyl-2′-deoxyuridine (EdU) incorporation assays were conducted as described (Wang et al., 2019). Briefly, 10 μl EdU solution (Molecular Probes, A10044; 10 mM in 50% DMSO in PBS) was injected intraperitoneally 1 h prior to collection of tailfin regenerates. Fin tissues were fixed in Carnoy's fixative (30% chloroform and 10% acetic acid in ethanol) for 3 h at room temperature, followed by methanol washes, 10 min each, and transfer into PBST. Next, tissues were incubated in freshly prepared Click-iT staining solution for 30 min at room temperature before four PBST washes of 5 min each. Fin tissues were mounted in Fluoromount G for imaging by confocal microscopy.

To manipulate the Wnt signaling pathway after a 50% tailfin amputation, zebrafish were incubated in fish water with 12 μM IWR-1 (Sigma-Aldrich, I0161-5G), 100 nM BIO (Sigma-Aldrich, B1686-5MG) or 0.02% DMSO for 6 h. Tailfin tissues were then collected for RNA extraction. To perturb the wound healing response after a 50% tailfin amputation, zebrafish were incubated with 100 μM Gd³⁺ (Sigma-Aldrich, G7532-5G) for 2 h. Animals were rinsed and kept in fish water for another 4 h before collecting tailfin tissues for RNA extraction.

To block bulk RNA transcription in zebrafish, Act D dissolved in 1% DMSO/PBS was intraperitoneally injected into adults (10 μl each; 100 μg/ml); a lower dose was injected into 11-week-old juveniles (5 μl each; 10 μg/ml) after anesthesia with tricaine (0.2 mg/ml) for 3-5 min. To block bulk RNA transcription in juvenile mice, Act D in 1% DMSO/PBS was intraperitoneally injected at a dosage of 0.25 mg/kg. To block DNA synthesis, live zebrafish were incubated in 20 mM hydroxyurea (Sigma-Aldrich, H8627) from 24 to 48 hpa as described (Wang et al., 2019).

Live zebrafish were incubated in 500 ml 0.01% ARS (Sigma-Aldrich, A5533) for 15 min at room temperature before rinsing for three times in fish water, 5 min each, according to a previously described protocol (Bensimon-Brito et al., 2016). Images of stained animals were acquired using a Leica M250 FA fluorescence stereomicroscope.

ISH, RT-qPCR and HRM assays were performed as described (Wang et al., 2019). To generate digoxigenin-labeled probes, shha (1.2 kb), lef1 (1.1 kb) and ddx52 (0.2 kb) cDNA sequences were cloned from zebrafish cDNA and then inserted into pCR2.1-TOPO vectors as synthesis templates (see Table S1). Primers used in RT-qPCR and HRM assays are listed in Table S1.

For whole-mount immunostaining of tenascin C, tailfin tissues were collected and fixed in Carnoy's fixative for 3 h at room temperature, followed by methanol washes for 10 min each and then transfer to PBST. Next, tissues were incubated in freshly prepared blocking buffer [10% heat inactivated newborn calf serum (Gibco, 26010066), 2% horse serum (Gibco, 16050130) and 0.1% DMSO in PBST] at 37°C for 30 min. Tissues were then incubated with antibodies against tenascin C (US Biological, T2550-23; 1:250) in blocking buffer without horse serum at 37°C for 4 h. After incubating with the secondary antibody (Invitrogen, A110037; 1:500) at 37°C for 30 min, tissues were then washed four times in PBST, 5 min each, and mounted in Fluoromount G for imaging by laser confocal microscopy (Leica SP8; HC PL APO 20×/0.75 SC2, WD 0.62 mm).

Body length was measured from the tip of the nose to the base of the tail; tail length was measured from the base of the tail to the tail tip. For micro-CT examinations, mice were euthanized before conducting two-dimensional imaging of whole mice and hind limbs (Skyscan 1076 system; Bruker) in the Taiwan Mouse Clinic, Academia Sinica. The scanning parameters were set as follows: image pixel size 9 μm, X-ray voltage 50 kV, X-ray current 140 μA, filter A1 0.5 mm, exposure 3300 ms, rotation step 0.8°, frame averaging 2, tomographic rotation 180°.

Statistical analyses were performed using GraphPad Prism 7. Data points from groups that had sufficient n number were tested for Gaussian distribution using the D'Agostino–Pearson test. If the data points were normally distributed, we used two-tailed Student's unpaired t-test to determine P-values. If the data were not normally distributed, we used non-parametric two-tailed Mann–Whitney U-test for comparison. Statistical parameters (i.e. exact n values, what n represents, s.e.m., s.d., etc.) are listed in the figure legends.

We thank Hsiao-Yuh Roan, Chung-Han Wang, and the Taiwan Zebrafish Core Facility (TZCAS; MOST 108-2319-B-400-002) for zebrafish care; Stephan Schneider, Jr-Kai Yu and Yi-Hsien Su for discussions; Chen laboratory members for comments on the manuscript; Marcus J. Calkins for English editing and comments.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199578.