5' Half of specific tRNAs feeds back to promote corresponding tRNA gene transcription in vertebrate embryos.

5′tRFls are small transfer RNA (tRNA) fragments derived from 5′ half of mature tRNAs. However, it is unknown whether 5′tRFls could feed back to regulate tRNA biogenesis. Here, we show that 5′tRFlGly/GCC and 5′tRFlGlu/CTC function to promote transcription of corresponding tRNA genes and are essential for vertebrate early embryogenesis. During zebrafish embryogenesis, dynamics of 5′tRFlGly/GCC and 5′tRFlGlu/CTC levels correlates with that of tRNAGly/GCC and tRNAGlu/CTC levels. Morpholino-mediated knockdown of 5′tRFlGly/GCC or 5′tRFlGlu/CTC down-regulates tRNAGly/GCC or tRNAGlu/CTC levels, respectively, and causes embryonic lethality that is efficiently rescued by coinjection of properly refolded corresponding tRNA. In zebrafish embryos, tRNA:DNA and 5′tRFl:DNA hybrids commonly exist on the template strand of tRNA genes. Mechanistically, unstable 5′tRFl:DNA hybrid may prevent the formation of transcriptionally inhibitory stable tRNA:DNA hybrids on the same tRNA loci so as to facilitate tRNA gene transcription. The uncovered mechanism may be implicated in other physiological and pathological processes.

INTRODUCTION

tRNAs, approximately 70 to 90 nucleotides (nt) long, are well known to deliver amino acids to ribosomes and decode codons in mRNAs during protein synthesis. Each amino acid is decoded by one or several tRNAs, called isoacceptors, each with a unique anticodon, and one isoacceptor may have several different isodecoders. According to GtRNAdb, the human genome contains 429 tRNA genes, and the zebrafish genome has 8676 tRNA genes, which means that each tRNA gene has multiple copies (). tRNA genes usually reside in clusters of repeats in the genome (). Each tRNA gene is a transcription unit including an upstream sequence and two internal promoter regions (A and B boxes) within the tRNA coding sequence. TFIIIC complex recognizes and binds the internal promoters and then recruits TFIIIB complex, which is followed by recruitment of RNA polymerase III (PolIII) to the upstream sequence to initiate transcription (). It is known that the abundance of a particular tRNA isoacceptor is generally related to its gene copy number. An interesting question is how tRNA genes maintain high-level expression during rapid cell proliferation, e.g., during early embryogenesis.

In the past 10 years, increasing evidence showed that tRNAs could be processed to generate fragments [tRNA-derived fragments (tRNAFs)] of different lengths under normal cellular processes or stress conditions (–). Cleavage of a mature tRNA at the anticodon loop gives rise to long 5′tRNAF (5′ half, 5′tRFl) or long 3′tRNAF (3′ half, 3′tRFl) with a length of 30 to 37 nt, and further cleavages at more sites such as lateral T loop and D loop result in smaller fragments of 18 to 21 nt, called short tRNAFs or tsRNAs (). tRNAFs may inhibit protein synthesis in diverse ways (–) or facilitate mRNA translation (, ), to reduce or enhance the stability of target mRNAs (, ) and even to repress PolII–catalyzed gene transcription probably by remodeling chromatins (, ). However, nothing is known whether tRNAFs could take part in transcription control of tRNA genes.

tRNAFs have been shown to be implicated in cell stress response (, , ), tumorigenesis (, , –), and neurological diseases (–). tRNAFs, mostly 5′tRFls, are found in 0- to 1-hour and 7- to 8-hour Drosophila embryos (), in 24-hour zebrafish embryos (), and in mouse oocytes and zygotes (, ), but their necessity for embryonic development has not been investigated. tRNAFs are also found in mouse sperms (, , , , ). Mouse zygotes injected with tRNAFs isolated from sperms of males that were fed with low-protein diet or high-fat diet show altered expression of some genes at the two-cell stage () or develop impaired glucose tolerance during postnatal growth (), respectively, which leads to the hypothesis that sperm-derived tRNAFs may transmit epigenetic information to offspring.

In our study, we intended to explore the role of tRNAFs in early embryonic development and underlying mechanisms. We disclosed that 5′tRFlGly/GCC and 5′tRFlGlu/CTC can competitively form 5′tRFl:DNA hybrids to prevent the formation of transcriptionally inhibitory long tRNA:DNA hybrids on the template strand of the corresponding tRNA genes, which ultimately promotes tRNA gene transcription and ensures normal embryonic development. Our finding provides an unprecedented insight into how tRNA genes maintain high-level transcription.

RESULTS

tRNAFs expressed in a dynamic level and pattern during zebrafish early embryogenesis

The zebrafish embryo develops fast with heartbeat starting at 24 hours postfertilization (hpf) (). The first phase of embryonic development heavily relies on maternal transcripts and proteins until the major zygotic genome activation (ZGA) occurs around the 1k-cell (1kc) stage (3 hpf) (). So far, little is known about the zygotic activation of tRNA genes and dynamics of tRNAFs during zebrafish embryogenesis. To investigate dynamics of tRNAFs, we performed deep sequencing of 18- to 40-nt RNAs extracted from zebrafish mature eggs and embryos at 1-cell (1c) (0.2 hpf), 256-cell (256c) (2.5 hpf), sphere (4 hpf), shield (6 hpf), or 24 hpf stages. Two pretreatment steps were incorporated to remove some “roadblock” modifications and to make both ends of tRNAFs more compatible with adaptor ligation (Fig. 1A and fig. S1A) (–), which notably improved detection of tRNAs by reverse transcription polymerase chain reaction (RT-PCR) (fig. S1B). Deep sequencing results showed that a certain proportion (~5%) of those small RNAs were tRNAFs in eggs as well as in 1c and 256c stage embryos, which is suggestive of maternal origin (Fig. 1B). From the sphere stage onward, the proportion of tRNAFs steadily increased to 9% at the shield stage and then slightly fell down at 24 hpf (Fig. 1B). Among all detected small RNAs, microRNA (miRNA) was not a major type during early stages since only miR-430 expressed at relatively high levels from the sphere to the shield stage (Fig. 1B and fig. S1C), which might be ascribed to the explosive expression of miR-430 for clearing maternal mRNAs as demonstrated before (, ). However, maternally expressed ribosomal RNA (rRNA)–derived fragments constituted the dominant class among small RNAs (Fig. 1B and fig. S1, D and E) (, ), the proportion of which decreased as the development went on. Among tRNAFs, approximately 80% belonged to 5′tRFls, whereas any of the other types accounted for less than 15% (Fig. 1C). The length distribution of tRNAFs showed that tRNAs with a high percentage of G:C pairs in the T stem generated much more 5′tRNAFs than 3′tRNAFs, whereas tRNAs with 100% of G:C pairs in the D stem mainly gave rise to 3′tRNAFs (fig. S2), suggesting that the abundance of a tRNAF is related to its sequence and secondary structure. As demonstrated by previous studies, the abundance/stability of tRNAFs may also be influenced by base modifications (, ). It is widely accepted that tRFls are generated by Angiogenin-catalyzed cleavage on the anticodon loop in vertebrates (, ). It remains unknown whether tRNAF biogenesis involves other mechanisms such as transcriptional abortion/pretermination.

Fig. 1.

Identification and expression analysis of tRNAFs.

(A) Principle of small RNA pretreatments for library construction. Only tRNAFs are shown. P, phosphate group; cP, 2′3′-cyclic phosphate group; OH, hydroxyl group; ATP, adenosine 5′-triphosphate. (B) Proportion of different types of small RNAs in zebrafish embryos. Small RNAs were aligned to different databases, and the unknown types were mapped to the genome. lncRNA, long noncoding RNA. (C) Percentages of different types of tRNAFs based on their cleavage sites, showing the dominance of long 5′tRFls. (D) Percentages of different 5′tRFls based on their original isoacceptors or isodecoders. (E) Percentages of full-length tRNAs. (F) RT-qPCR verification of 5′tRFlGly/GCC and 5′tRFlGlu/CTC expression levels from 1c to shield stage. Expression levels at later stages were normalized to that at the 1c stage. (G and H) Northern blot verification of 5′tRFlGly/GCC (G) and 5′tRFlGlu/CTC (H) expression levels from 1c to shield stage. Left: GelSafe staining of urea polyacrylamide gel electrophoresis (PAGE) gel with indicated full-length tRNA and 5′tRFl positions. Middle: Northern blot detected by locked nucleic acid (LNA) probe against tRNA/5′tRFlGly/GCC (G) or tRNA/5′tRFlGlu/CTC (H). Right: Average 5′tRFl level with ±SD from three independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 (Student’s t test, two-tailed).

Tracing back to the origin of 5′tRFls, we found that 84 to 87% of 5′tRFls at any examined stages were derived from only five tRNA isotypes, i.e., tRNAGly/GCC, tRNAGlu/CTC, tRNAGlu/TTC, tRNAHis/GTG, and tRNAAsp/GTC (Fig. 1D). To understand the relationship between 5′tRFl and full-length tRNA levels, we then performed deep sequencing of 60- to 120-nt RNAs isolated from embryos at different stages. Results showed that levels of tRNAs varied greatly with isoacceptors and isodecoders (Fig. 1E). tRNAHis/GTG, tRNAGly/GCC, tRNAGlu/TTC, tRNAAsp/GTC, tRNALys/CTT, and tRNAGlu/CTC together accounted for 89.2 to 96.2% of total tRNAs. In general, tRNA and 5′tRFl expression levels correlated with gene copy number (fig. S3, A and B). Similar to tRNA expression in other systems (–), however, some tRNA genes showed severely biased expression levels in zebrafish embryos. For example, 75 to 86% of total tRNAs at any examined developmental stages were tRNAHis/GTG that has 511 gene copies in the genome (Fig. 1E), which contrasted to tRNALys/TTT that has 643 gene copies but only contributed to 0.02 to 0.29% of total tRNAs. We currently do not know why and how this expression bias happens. Nevertheless, we observed that, in general, the 5′tRFl level positively correlates with the full-length tRNA level (fig. S3C).

Since 5′tRFlGly/GCC and 5′tRFlGlu/CTC were the most abundant 5′tRFls from egg to the shield stage (Fig. 1D), which were also abundant in mouse mature sperm and zygotes (, , , , ), we focused on these two 5′tRFls in subsequent studies. We examined the dynamics of these two 5′tRFls by RT–quantitative PCR (qPCR) using specific primers (fig. S4) and by Northern blotting. Results showed that compared to the level at the 1c stage, the level of 5′tRFlGly/GCC or 5′tRFlGlu/CTC decreased at the 256c stage and then increased from the sphere stage onward (Fig. 1, F to H). It is likely that maternal 5′tRFlGly/GCC and 5′tRFlGlu/CTC gradually degraded, and they were newly generated to meet the demand for development after ZGA.

Antisense morpholinos may not bind to and inhibit function of native full-length tRNAs

To look into 5′tRFlGly/GCC and 5′tRFlGlu/CTC function, we attempted to adopt an antisense morpholino (MO) knockdown approach that has been successfully used for studying miRNA function (, , ). The MOs, named as 5′tRFl-MOs, were designed to target a 25-nt sequence close to the 5′ end of tRNA (Fig. 2A and fig. S5). As shown in the illustration, 5′tRFl and 5′ end of the corresponding full-length tRNA share the same sequence, and we worried about the binding of antisense MO to full-length tRNA in vivo. To test this possibility, we added 5′tRFl-MOs and several control MOs (fig. S5) to the rabbit reticulocyte lysate (RRL) in vitro translation system (Fig. 2B), which contains rabbit tRNAGly/GCC and tRNAGlu/CTC identical to the zebrafish counterparts in sequence. We speculated that binding of MO with a full-length tRNA would prohibit its function due to structural change. However, we found that green fluorescent protein (GFP) protein translated from added GFP mRNA, which contains 19 GGC (Gly) codons (decoded by the GCC anticodon) and 15 GAG (Glu) codons (decoded by the CTC anticodon), showed no obvious changes after addition of 5′tRFlGly/GCC-MO, 5′tRFlGlu/CTC-MO, or any control MOs (Fig. 2B), suggesting that antisense MOs may not interfere with the activity of full-length tRNA. Besides, Northern blot results disclosed that the level of full-length tRNAGly/GCC or tRNAGlu/CTC in the in vitro translation system remained stable in the presence of any MOs (Fig. 2C). Thus, MOs may not interfere with the function and stability of full-length tRNAs, at least in in vitro translation system.

Fig. 2.

Specificity of 5′tRFl complementary antisense MOs.

(A) Illustration of 5′tRFl-MO target region. (B) 5′tRFlGly/GCC and 5′tRFlGlu/CTC MOs did not affect translation activity in the RRL in vitro translation system. Left: The experimental procedure. The synthesized and exogenous proteins were immunoblotted with anti-GFP and anti-mCherry antibody, respectively. Right top: Western blot results with short exposure (S. exp) or long exposure (L. exp). Bottom: The relative GFP/mCherry ratio quantified from band intensity. Data are shown in averages with ±SD from three independent experiments. ns, nonsignificant with P > 0.05 (Student’s t test, two-tailed). (C) Northern blot results of tRNAGlu/CTC (top) and tRNAGly/GCC (bottom) from the RRL system with the addition of different MOs using antisense LNA probes. (D) Experimental procedure for detection of the tRNA/MO complex by Northern blotting. (E and F) Estimation of MO binding capacity in vivo with tRNAGlu/CTC (E) or tRNAGly/GCC (F). Each blot was subjected to short or/and long exposure after hybridization. S, 1.2 ng of in vitro synthesized full-length tRNA and 1.2 ng of 5′tRFl-mimetic that were directly loaded onto gel; S + MO, preannealed synthetic tRNA (1.2 ng)/5′tRFl (1.2 ng) and MO (1.5 ng); other lanes in right, MO injection at 10 ng per embryo. The possible composition of each band after GelSafe staining is indicated by arrows. Note that 5′tRFlGlu/CTC-MO/tRNAGlu/CTC was undetectable, suggesting that they do not bind to each other. Weak 5′tRFlGly/GCC-MO/tRNAGly/GCC complex (indicated by a blue arrow) was detected and is quantified in the bar graph on the left. 5′tRFlGly/GCC/5′tRFlGly/GCC-MO or 5′tRFlGlu/CTC/5′tRFlGlu/CTC-MO complex is indicated by a red arrow.

Next, we tested whether MO binds to mature tRNA and 5′tRFl in vitro and in vivo. The in vitro binding assay indicated that 5′tRFlGly/GCC-MO preferred binding to 5′tRFlGly/GCC-mimetic (fig. S6A), whereas 5′tRFlGlu/CTC bound to 5′tRFlGlu/CTC and full-length tRNAGlu/CTC without obvious preference (fig. S6B). For in vivo assay, we injected 5′tRFlGlu/CTC-MO or 5′tRFlGly/GCC-MO into 1c stage embryos at different doses and harvested RNAs at the 256c stage by neutral phenol:chloroform extraction to perform Northern blotting (Fig. 2D). In vitro transcribed tRNAGlu/CTC or tRNAGly/GCC, corresponding synthetic 5′tRFl-mimetic, and MO were mixed, denatured, and renatured, and the mix was used as a control. We found that MO was able to efficiently bind to corresponding predenatured tRNA or 5′tRFl, causing a shift of the band (first lane in Fig. 2, E and F). However, tRNAGlu/CTC bound by 5′tRFlGlu/CTC-MO was not detected in embryos (Fig. 2E, left); only a minor proportion of tRNAGly/GCC was bound by 5′tRFlGly/GCC-MO, ranging from 1.1 to 4.9% depending on MO dosage (Fig. 2F, left). Similarly, 3′tRGlu/CTC-MO, 3′tRGly/GCC-MO, or standard control (Std)–MO did not bind to endogenous full-length tRNAGlu/CTC or tRNAGly/GCC in embryos (Fig. 2, E and F, right). Therefore, MOs could be used to knock down 5′tRFls without much interference with mature tRNAs in vivo.

Locked nucleic acid (LNA)–based antisense oligos were previously used to knock down tRNAFs in mouse zygotes (, ). We found that injection of synthesized antisense 5′tRFlGly/GCC-LNA (αGly-LNA) or 5′tRFlGlu/CTC-LNA (αGlu-LNA) (fig. S5) into zebrafish embryos allowed binding of approximately 50% tRNAGly/GCC or 20% tRNAGlu/CTC (fig. S6, C and D), respectively. The higher binding affinity of LNA oligos with mature tRNAs suggests that LNA oligos are not as appropriate as MOs, at least in zebrafish embryos.

5′tRFlGly/GCC and 5′tRFlGlu/CTC are required for early embryonic development

Then, we performed knockdown of 5′tRFlGly/GCC or 5′tRFlGlu/CTC in zebrafish embryos. Injection of 10 ng of 5′tRFlGly/GCC-MO or 5′tRFlGlu/CTC-MO into 1c stage embryos did not cause morphological defects before the sphere stage (4 hpf) (Fig. 3A). 5′tRFlGly/GCC morphants then started to show developmental delay around 4.7 hpf and were arrested before 6 hpf, while 5′tRFlGlu/CTC morphants started to show delayed development around 7 hpf and died about 10 hpf (Fig. 3, A and C). Injection with a higher dose (20 ng) caused a similar effect, whereas a lower dose (1.25 ng) could delay the appearance of anomaly (Fig. 3C). p53 mutant embryos injected with 5′tRFlGly/GCC-MO or 5′tRFlGlu/CTC-MO still abnormally developed (fig. S7), implying that defects are unlikely due to the nonspecific activation of p53 apoptotic pathway by MOs (). In comparison, injection of mismatched control MOs (5′tRFlGly/GCC-cMO and 5′tRFlGlu/CTC-cMO), Std-MO, or MO targeting the 3′ end of tRNAGly/GCC or tRNAGlu/CTC (fig. S5) at 10 or 20 ng in wild-type embryos did not cause visible morphological changes (Fig. 3, A, B, and D). These results together suggest that knockdown effects are 5′tRFl specific rather than general toxicity and that 5′tRFlGly/GCC and 5′tRFlGlu/CTC are essential for survival of embryos. The observation that embryonic abnormality in 5′tRFlGly/GCC morphants appeared earlier than in 5′tRFlGlu/CTC morphants implies that the threshold for normal embryonic development may be different between 5′tRFlGly/GCC and 5′tRFlGlu/CTC and may also depend on developmental stages.

Fig. 3.

Knockdown of 5′tRFlGlu/CTC or 5′tRFlGly/GCC leads to early embryonic lethality.

(A) Knockdown effect of 5′tRFlGlu/CTC and 5′tRFlGly/GCC. The binding region of MO in 5′tRFl is illustrated on top. All embryos were laterally viewed. The ratio of representative embryos was indicated on the right bottom. Note that 5′tRFlGly/GCC morphants could not develop beyond the shield stage, and most 5′tRFlGlu/CTC morphants died right after the bud stage. (B) Mismatch control MOs induced no obvious phenotype. The complementarity of cMO to 5′tRFl is illustrated on top. (C) Dose-dependent effects of 5′tRFlGlu/CTC-MO (top) and 5′tRFlGly/GCC-MO (bottom) on embryonic development. Normal embryos were those that had no obvious morphological defects. n, embryo number at the start of observation. (D) Injection of 3′tRGlu/CTC-MO or 3′tRGly/GCC-MO leads to no obvious phenotype. The binding region of 3′tR-MOs is illustrated on top. (E) Puromycin-incorporated nascent proteins are reduced in 5′tRFlGlu/CTC or 5′tRFlGly/GCC morphants. Left: Experimental procedure. Purified GFP protein injected at 6 ng per embryo served as an internal control. Right top: Western blot results at indicated stages using anti-puromycin or anti-GFP antibody. Molecular weight markers are shown on the right side. CHX, cycloheximide, a translation inhibitor. Right bottom: Relative puromycin/GFP ratio normalized to that in the Std-MO group. Data are shown in averages with ±SD from three independent experiments. ns, nonsignificant with P > 0.05; ***P < 0.001; ****P < 0.0001 (Student’s t test, two-tailed). Scale bars, 100 μm (for zebrafish embryos).

We noted that injection of 5′tRFlGlu/CTC-MO did not interrupt cell fate specification and embryonic patterning (fig. S8). We next tested whether protein synthesis in morphants was influenced. Puromycin could be incorporated into nascent protein and detected by Western blot (). Puromycin incorporation assay results indicated that, compared to the control group, 5′tRFlGly/GCC morphants and 5′tRFlGlu/CTC morphants had reduced amounts of nascent proteins at the sphere or shield stage (Fig. 3E), respectively, which happened earlier than the appearance of morphological changes. This observation differs from the in vitro translation assay result (Fig. 2B). Thus, we guess that knockdown of 5′tRFlGly/GCC or 5′tRFlGlu/CTC in embryos may interfere in biogenesis of some translation machinery components such as tRNAs.

5′tRFlGlu/CTC and 5′tRFlGly/GCC promote transcription of matching tRNA genes

Then, we set out to investigate the dynamics of tRNA levels and potential implication of 5′tRFls in tRNA biogenesis during zebrafish embryogenesis. tRNA is a group of highly expressed “housekeeping” noncoding RNAs, and it could be easily seen and quantified on polyacrylamide gels (Fig. 4A). When the same amount of total RNAs was loaded on the gel, levels of 5S rRNA and 5.8S rRNA remained stable during early embryogenesis; in contrast, the level of total tRNAs exhibited a marked increase from the sphere to the shield stage, although not much changes happened from the 1c to the sphere stage (Fig. 4A). We examined tRNAGly/GCC and tRNAGlu/CTC levels by RT-qPCR with specific primers and by Northern blotting using specific LNA probes. Using ALKB (DNA oxidative demethylase) pretreated RNA samples for RT-qPCR analysis, we found that levels of full-length tRNAGly/GCC and tRNAGlu/CTC were relatively stable from the 1c to sphere stage but increased markedly from the sphere to the shield stage (Fig. 4B), which are very similar to the expression pattern of total tRNAs (Fig. 4A). Northern blotting revealed steady increase in tRNAGly/GCC and tRNAGlu/CTC levels from the 1c to the shield stage (Fig. 4C). The difference between RT-qPCR and Northern blot results might arise from cross-hybridization of shorter LNA probes (fig. S5, A and B) with other tRNAs of some sequence homology and from interference of reverse transcription efficiency by the remaining base modifications of tRNAs and variations of primer binding sequences (fig. S5, A and B).

Fig. 4.

Loss of 5′tRFlGlu/CTC and 5′tRFlGly/GCC impairs normal tRNA gene transcription.

(A) Dynamics of total tRNAs during early embryogenesis. (B and C) Dynamics of full-length tRNAGly/GCC and tRNAGlu/CTC levels during early embryogenesis. Their levels were detected by both RT-qPCR (B) and Northern blot (C). The RT-qPCR values were normalized to that at the 1c stage. (C) Left: Northern blot and gel staining results. (C) Right: Relative level of tRNAGly/GCC and tRNAGlu/CTC based on the band intensity on the blot that was normalized to the average intensity of 5S and 5.8S rRNAs on gel. (D and E) Knockdown of 5′tRFlGly/GCC and 5′tRFlGlu/CTC down-regulates the level of related tRNA. Embryos at the 1c stage were injected with 10 ng of MO and collected at indicated stages for RT-qPCR (D) or Northern blot (E) as described in (B) and (C). (F) Effect of 5′tRFlGly/GCC-MO injection on the expression of different tRNAGly isoacceptors/isodecoders (id) with divergent sequences. Left: Sequence comparison among different tRNAGly in the 5′tRFlGly/GCC-MO targeting sequence. The identical nucleotides are indicated by “-”. Note that tRNAGly/GCC-id1 and tRNAGly/GCC-id2 have different nucleotides in the remaining part (as shown by the primers in table S3) of tRNA sequence. Right: Expression levels of indicated tRNAGly. 1c stage embryos were injected with 10 ng of MO per embryo and collected at the sphere stage for RT-qPCR analysis. (G) ChIP-qPCR results showing a reduction of Polr3a binding to tRNAGlu/CTC genes in 5′tRFlGlu/CTC morphants. 1c stage embryos were injected with 10 ng of 5′tRFlGlu/CTC-MO or 5′tRFlGlu/CTC-cMO and harvested at 6 to 8 hpf for ChIP assay using Polr3a antibody or immunoglobulin G (IgG). The immunoprecipitated DNA was used for amplifying the indicated loci using specific primers. Data are shown as averages with ±SD from three independent experiments. ns, nonsignificant with P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 (Student’s t test, two-tailed).

We tested whether 5′tRFlGly/GCC or 5′tRFlGlu/CTC knockdown could affect the amount of full-length tRNAGly/GCC and tRNAGlu/CTC. Both RT-qPCR and Northern blot results showed that knockdown of 5′tRFlGly/GCC resulted in a remarkable reduction of tRNAGly/GCC level at the sphere stage but had no effect on tRNAGlu/CTC and tRNAGlu/TTC levels (Fig. 4, D, left and E, top) and that knockdown of 5′tRFlGlu/CTC caused a significant decrease in both tRNAGlu/CTC and tRNAGlu/TTC, which share 84% identity in the MO-targeting sequence, with unchanged tRNAGly/GCC level at the shield stage (Fig. 4, D, right and E, bottom). Given that there are tRNAGly isoacceptors and isodecoders with divergent sequences (see Fig. 4F for examples), we further tested whether knockdown by 5′tRFlGly/GCC-MO could affect the expression levels of different isoacceptors/isodecoders. Result revealed that only those with a high sequence identity to the target sequence of 5′tRFlGly/GCC-MO showed a significantly reduced expression level in morphants (Fig. 4F, right).

To check whether the reduction of tRNA was caused by transcription inhibition, we then focused on 5′tRFlGlu/CTC to test its influence on transcription of tRNAGlu/CTC genes in zebrafish embryos. tRNA genes are transcribed by RNA PolIII. According to the GtRNAdb, the zebrafish genome encodes more than 300 tRNAGlu/CTC genes, most of which cluster on several chromosomes. We chose six separately located tRNAGlu/CTC loci and a tRNALeu/CAG locus on chromosome 18 to perform chromatin immunoprecipitation (ChIP) assay using an antibody against Polr3a, the large subunit of PolIII. As shown in Fig. 4G, all of the tRNA loci could be enriched in the Polr3a complex from 6 to 8 hpf embryos, whereas the promoter of β2-actin (actb2), a protein-coding gene transcribed by RNA PolII, was not enriched. When 5′tRFlGlu/CTC was knocked down in embryos, the Polr3a occupancy was reduced on five tRNAGlu/CTC loci (chr1-1, chr1-3, chr7-2, chr16-2, and chr21-1) but remained unchanged on the chr16-1 tRNAGlu/CTC locus and the chr18 tRNALeu/CAG locus (Fig. 4G). This observation suggests a positive role of 5′tRFlGlu/CTC in transcription of most tRNAGlu/CTC genes.

To perform a gain-of-function study for 5′tRFls, we synthesized 5′tRFl-mimetics, RNA oligos that do not contain “native” modifications. Embryos injected with 5′tRFlGly/GCC or 5′tRFlGlu/CTC at the 1c stage showed no obvious anomaly during early embryogenesis (Fig. 5A). As detected by RT-qPCR and Northern blot analyses, however, injection of 5′tRFlGlu/CTC-mimetic resulted in significant increase in tRNAGlu/CTC and tRNAGlu/TTC levels at the 256c stage (Fig. 5, B and C). Injection of 5′tRFlGly/GCC-mimetic induced a weak increase in tRNAGly/GCC level (Fig. 5, B and C). These results support the idea that 5′tRFlGly/GCC and 5′tRFlGlu/CTC positively regulate the synthesis of tRNAGly/GCC and tRNAGlu/CTC in embryos.

Fig. 5.

5′tRFlGlu/CTC and 5′tRFlGly/GCC promote related tRNA gene transcription.

(A) Injection of 5′tRFlGlu/CTC or 5′tRFlGly/GCC-mimetics leads to no obvious defects till 36 hpf. Scale bar, 100 μm. (B and C) 5′tRFlGly/GCC and 5′tRFlGlu/CTC positively regulate related tRNA levels. Embryos at the 1c stage were injected with 0.6 pmol of RNA oligo and collected at the 256c stage for RT-qPCR (B) or Northern blot (C) as described in Fig. 4 (B and C). (D) Illustration of in vitro tRNA gene transcription system (ivTGTS). The upstream sequences (red line) and transcription sequences (green box) of three tRNAGlu/CTC loci were used to construct reporter plasmids by each fusing to the artificial reporter sequence F30-Broccoli (F30-Br). TSS, transcription start site; TTS, transcription termination site. (E) Effectiveness test of F30-Broccoli reporter transcription in ivTGTS by RT-PCR. Presynthesized GFP mRNA was added to serve as an exogenous reference. (F and G) Effect of 5′tRFlGlu/CTC-mimetic (F) or 5′tRFlGlu/CTC-MO (G) on F30-Broccoli reporter transcription in ivTGTS. Doses (per reaction): 1 μg of DNA mix; MO, 1 μg; RNA oligos, 60 pmol. For all experiments in this figure, data are shown in average with ±SD from three independent experiments. ns, nonsignificant with P > 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 (Student’s t test, two-tailed).

In view of the large amount of maternal tRNAs that may lead to underestimate the effect of 5′tRFls on tRNA gene transcription, we attempted to develop an in vitro tRNA transcription system for verifying the effect of 5′tRFlGlu/CTC on tRNAGlu/CTC gene transcription. This system consisted of HeLa nuclear lysate () and three reporter constructs. The reporter constructs were made by individually fusing the upstream regulatory sequence plus tRNAGlu/CTC coding sequence of chr1-1, chr16-2, or chr21-1 to a 90-nt F30-Broccoli aptamer reporter (Fig. 5D) (). This reporter sequence does not exist in vertebrate genomes, and thus, it allows faithful RT-PCR amplification of the fusion transcript (Fig. 5E). We found that addition of 5′tRFlGlu/CTC-mimetic to this system significantly increased the F30-Broccoli transcript level (Fig. 5F). In the presence of 5′tRFlGlu/CTC-MO, however, the reporter RNA level was decreased compared to the addition of 5′tRFlGlu/CTC-cMO (Fig. 5G), suggesting that 5′tRFlGlu/CTC-MO interferes with the function of endogenous 5′tRFlGlu/CTC in the nuclear lysate. Since zebrafish tRNAGly/GCC genes are tandemly repeated or intercalated on chromosomes, we were unable to identify well-separated tRNAGly/GCC genes to perform ChIP assay or to establish an in vitro transcription system.

Exogenous specific tRNAs can functionally rescue the phenotype of 5′tRFl morphants

Assuming that the embryonic lethality of 5′tRFl morphants is ascribed to insufficient amount of corresponding tRNAs, we attempt to rescue the morphants using cognate tRNAs. Although 5′tRFl-MOs do not bind efficiently with native mature tRNAs in vivo (Fig. 2, E and F), we still cautiously chose tRNAGly/GCC and tRNAGlu/CTC from other species, which contain base variants on the MO complementary sequence (Fig. 6A and fig. S9A). For example, the yeast (Sc) tRNAGly/GCC has 7 bases mismatching to the 5′tRFlGly/GCC-MO targeting sequence, and the yeast Sc-tRNAGlu/CTC has 13 bases mismatching to the 5′tRFlGlu/CTC-MO targeting sequence (Fig. 6A). Individual tRNA genes were in vitro transcribed using T7 RNA polymerase; the product was purified on denaturing polyacrylamide gel electrophoresis (PAGE) gel and refolded to form active structures (Fig. 6B). 1c stage embryos were first injected with 5′tRFl-MO and then injected with renatured tRNAGly/GCC. As shown in Fig. 6C, the Sc-tRNAGly/GCC antisense probe did not hybridize with endogenous Dr-tRNAGly/GCC or Dr-5′tRFlGly/GCC in embryos, and coinjection of Sc-tRNAGly/GCC with 5′tRFlGly/GCC-MO did not result in an increase in tRNAGly/GCC/5′tRFlGly/GCC-MO complex in amount, which suggests that 5′tRFlGly/GCC-MO does not bind to Sc-tRNAGly/GCC. We observed that injection with synthesized Sc-tRNAGly/GCC or Dr-tRNAGly/GCC did not affect embryonic development (fig. S9B). As demonstrated before, embryos injected with 10 ng of 5′tRFlGly/GCC-MO (about 1242 fmol per embryo) died before the completion of gastrulation (10 hpf) (Fig. 6D, second column). Coinjection of these morphants with 800 pg (about 35 fmol per embryo) of Sc-tRNAGly/GCC allowed 11 of 32 of them to develop to segmentation stage (22 hpf) albeit with some defects (extensive cell death in the head) (Fig. 6D, fourth column). In contrast, synthetic Dr-tRNAGly/GCC failed to rescue embryos injected with 10 ng of 5′tRFlGly/GCC-MO. Embryos injected with 1.25 ng of 5′tRFlGly/GCC-MO could develop to 24 hpf with a shorter body axis and massive cell death, which could be well rescued by Sc-tRNAGly/GCC and slightly rescued by Dr-tRNAGly/GCC (fig. S9C). It is known that sequence and posttranscriptional modifications of a tRNA play important roles in forming stable, functional tertiary structure (). Compared to renatured synthetic Sc-tRNAGly/GCC, renatured synthetic Dr-tRNAGly/GCC that is distant to Sc-tRNAGly/GCC in sequence (Fig. 6A) has little functional activity, which is likely ascribed to the lack of posttranscriptional modifications for forming and maintaining a proper tertiary structure.

Fig. 6.

Supply of exogenous tRNAs alleviates knockdown effect of 5′tRFlGly/GCC or 5′tRFlGlu/CTC in zebrafish embryos.

(A) Sequence comparison between zebrafish (Danio rerio, Dr) and yeast (Saccharomyces cerevisiae, Sc) tRNAGly/GCC or tRNAGlu/CTC. The mismatched bases are boxed. The MO-targeted sequences are indicated by arrows. (B) Procedures for in vitro tRNA transcription using the T7 RNA polymerase and subsequent purification and refolding. RT, room temperature. (C) Disability of 5′tRFlGly/GCC-MO binding to Sc-tRNAGly/GCC in zebrafish embryos as detected by Northern blotting. Ten nanograms of 5′tRFlGly/GCC-MO and/or 800 pg of tRNAGly/GCC (per embryo) was injected into 1c stage embryos, and embryos were harvested at 2.5 hpf for extraction of total RNAs. Synthetic tRNAGly/GCC was directly loaded onto the gel, serving as molecular markers. RNAs on blots were probed using digoxigenin-labeled antisense Sc-tRNAGly/GCC oligo (left bottom) or antisense Dr-tRNAGly/GCC-LNA oligo (right bottom). The endogenous Dr-tRNAGly/GCC/5′tRFlGly/GCC-MO complex is indicated by a red arrow (right bottom). (D) Sc-tRNAGly/GCC rescues the lethal phenotype of 5′tRFl Gly/GCC morphants. (E) Procedure for isolation of native Dr-tRNAGlu/CTC and Sc-tRNAGlu/CTC. The starting material was purchased yeast crude tRNA extracts or total tRNAs extracted from zebrafish embryos. Biotin-labeled antisense Sc-tRNAGlu/CTC or Dr-tRNAGlu/CTC oligo was used to pull-down specific tRNAGlu/CTC molecules, which were enriched by streptavidin beads. The isolated tRNAGlu/CTC molecules were subjected to gel separation, gel cut (boxed), elution, and refolding. (F) Effective rescue of 5′tRFlGlu/CTC morphants by native Dr-tRNAGlu/CTC but not Sc-tRNAGlu/CTC. 5′tRFlGlu/CTC-MO (1.25 ng per embryo) alone or together with purified native Dr- or Sc-tRNAGlu/CTC (1.6 ng per embryo) was injected into 1c stage embryos, followed by morphological observation at 24 hpf. The ratio of embryos with the representative morphology is indicated. Scale bars, 100 μm (for embryos).

To perform rescue experiment for 5′tRFlGlu/CTC morphants, we chose tRNAGlu/CTC genes of zebrafish and seven other species (fig. S9A) for in vitro transcription using T7 RNA polymerase. None of these synthetic tRNAGlu/CTC species rescued 5′tRFlGlu/CTC morphants, which was also likely due to the lack of proper modifications or/and failure of recognition by the zebrafish glutamyl-prolyl-tRNA synthetase 1 (Eprs1). Then, we turned to native zebrafish and yeast tRNAGlu/CTC. We used a biotin-tagged antisense oligo to isolate native tRNAGlu/CTC from zebrafish embryos and commercial yeast tRNA crude extract, followed by gel purification and renaturation (Fig. 6E). Embryos injected with either of renatured native tRNAGlu/CTC developed normally (fig. S9B). As shown in Fig. 6F, 87.8% (43 of 49) of embryos injected with 1.25 ng of 5′tRFlGlu/CTC (about 155 fmol per embryo) alone developed to 24 hpf with extensive cell death and arrested thereafter (Fig. 6F). When the morphants were subjected to the second injection with 1.6 ng of native zebrafish tRNAGlu/CTC (about 70 fmol per embryo), 87% (40 of 46) of them had normal morphology at 24 hpf, suggesting a good rescue effect (Fig. 6F). However, coinjection of purified native Sc-tRNAGlu/CTC did not show any rescue effect, probably because of its incompatibility with the zebrafish glutamyl-prolyl-tRNA synthetase or inappropriate modifications. The above data together indicate that knockdown effect of 5′tRFlGly/GCC or 5′tRFlGlu/CTC in zebrafish embryos could be alleviated by input of a single exogenous tRNA species, which is consistent with the idea that these 5′tRFls function to promote transcription of their matching tRNA genes.

Sense R-loops widely exist in tRNA gene loci in zebrafish embryos

R-loop is a triple-stranded structure consisting of an RNA:DNA hybrid and an unpaired single-stranded DNA (ssDNA) (). During transcription, an RNA:DNA hybrid forming between a nascent RNA and the DNA’s transcribed/template/noncoding strand may prevent transcription (). One study found that mutation of ribonuclease H (RNase H) in yeast, an RNase specifically digesting the RNA in R loops, resulted in higher levels of nascent tRNAs, indicating that R loops that formed on tRNA gene loci may repress the tRNA gene transcription (). We performed ssDRIP-seq, a method for single-strand DNA ligation–based library construction after RNA:DNA hybrid immunoprecipitation combined with next-generation sequencing (), to genome-widely identify R loops in zebrafish wild-type embryos at 256c, sphere, and shield stages, each with a biological replicate (Fig. 7A). We obtained 21.2 million to 29.3 million reads with an average length of about 210 base pairs (bp) from each library. If the genomic DNA was first digested with RNase H before DRIP assay, then the R loop signals became inconspicuous (fig. S10A), suggesting that the detected signals are reliable. We defined the sense R loop (sR-loop) as RNA:DNA hybrid between RNA and the DNA’s template strand and the antisense R loop (aR-loop) as the hybrid between RNA and the DNA’s coding strand (Fig. 7B). Results showed that a large number of R loops without bias between sR-loops and aR-loops existed in protein-coding genes, and both remained relatively constant from the 256c to the shield stage (Fig. 7C). R loop signals were clearly detected at several isolate tRNAGlu/CTC, tRNAGlu/TTC, and tRNAGly/GCC loci (fig. S10B); however, sR-loop signals were predominant over aR-loop signals at all tested stages (Fig. 7D), indicating that the template strands of tRNA genes are more likely to be bound by RNAs. Besides, 56 and 86% of tRNA genes showed a decrease in sR-loop level from the 256c to the sphere stage and from sphere to shield stage, respectively (Fig. 7E), implying that RNAs bound to the template strands of tRNA genes could be removed as embryos develop further. When looking into individual tRNA loci, we found that the sR-loop levels in almost all of tRNAGly/GCC loci and in most of tRNAGlu/CTC loci decreased from the 256c to the shield stages (Fig. 7F). The sR-loops were obviously enriched in the region from the transcription start site to the transcription termination site of tRNA genes (fig. S10C), suggesting occupancy of their whole transcribing regions by RNAs. It appears that RNA:DNA hybrid, which forms between the template strand of a tRNA gene and an RNA, tends to disassociate from ZGA to the gastrulation onset so as to allow its more efficient transcription.

Fig. 7.

tRNAs and 5′tRFls form sR-loops on tRNA loci in zebrafish embryos.

(A) Illustration of ssDNA and RNA isolation from R-loops. (B) Illustration of sR-loop and aR-loop. The template (t-) and coding (c-) strands of DNA are depicted with different color lines. (C and D) Levels of R-loops in all protein-coding genes (C) and tRNA loci (D). Data are shown as means ± SD of two replicates. (E) The number of tRNA genes with altered sR-loop signals between two indicated stages. (F) The relative sR-loop signal changes on all tRNAGly/GCC and tRNAGlu/CTC loci during development. Data are shown as means ± SD. Wilcoxon matched-pairs signed-rank test was used with significance levels: ns, nonsignificant; ****P < 0.0001. (G) Detection of DNA-bound tRNAs and 5′tRFls by PCR using RNA template isolated from RNA:DNA hybrids in embryos at indicated stages. PCR was run for 25 to 34 cycles. See also fig. S4 for primers and specificity. (H) Quantification of DNA-bound specific tRNAs and 5′tRFls by RT-qPCR using RNA template isolated from RNA:DNA hybrids in embryos. qPCR results of RNA were normalized to Chr1-1 amplification signal of input genome, and signals of all samples were normalized to that at the shield stage. Ctr, control sample for which antibody (Ab) was replaced by IgG. (I) 5′tRFlGlu/CTC binds to tRNA genes during in vitro transcription. Left: Procedure for pulling down biotin-5′tRFlGlu/CTC–bound reporter DNA (see also Fig. 5D) with primer positions indicated. Right: qPCR result from three independent experiments (means ± SD). Biotin-GFP-r1 served as a negative control. Significant levels (Student’s t test, two-tailed) (H and I): *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

tRNAGlu/CTC and tRNAGly/GCC and their 5′tRFls bind to corresponding tRNA genes

sR-loops are widely formed in tRNA genes, but an important question is what RNAs are there. Since the formation of RNA:DNA hybrids relies on complementary sequences, we guessed that RNAs that bound to the template strand of tRNA genes are either tRNAs or their derivatives. Then, we purified RNAs from RNA:DNA hybrids immunoprecipitated from genomic DNAs with S9.6 antibody and used these RNAs as template to amplify full-length tRNAGlu/CTC and tRNAGly/GCC as well as their 5′tRFl derivatives with specific primers (see illustration in Fig. 7A and primer specificity in fig. S4). Results showed that all of these full-length tRNAs and their 5′tRFls were detected from the RNA:DNA precipitates obtained from 1kc and shield stage embryos (Fig. 7G), strongly suggesting physical association of both full-length tRNAs and their 5′tRFls with tDNA in vivo. Using DRIP RNAs at different developmental stages as templates, RT-qPCR analysis revealed that both DNA-bound tRNAGlu/CTC/tRNAGly/GCC and DNA-bound 5′tRFlGlu/CTC/5′tRFlGly/GCC decreased as embryos developed (Fig. 7H).

To further verify whether 5′tRFls bind to DNA, we performed a DNA pull-down assay in the in vitro transcription system (Fig. 5D) with addition of biotin-labeled 5′tRFlGlu/CTC-mimetic or GFP-r1 RNA (as a control). Biotin-RNA:DNA hybrids were pulled down using streptavidin beads, and the F30-Broccoli DNA sequence on plasmids was amplified by qPCR (Fig. 7I, left). Result revealed that the F30-Broccoli DNA was significantly enriched from the biotin-5′tRFlGlu/CTC pulled-down RNA:DNA hybrids (Fig. 7I, right), confirming that 5′tRFlGlu/CTC is capable of binding the template strand of tRNAGlu/CTC genes during transcription.

5′tRFlGlu/CTC and 5′tRFlGly/GCC prevent binding of corresponding tRNAs to DNA

To further test whether 5′tRFls prevent full-length tRNAs from binding to tRNA genes during transcription, we quantified the dynamic levels of RNA-bound tRNA gene loci or DNA-bound tRNAs in 5′tRFl morphants using ssDNA or RNA strand isolated from DRIP products (Fig. 8A). We found that, in 5′tRFlGlu/CTC morphants, RNA-bound DNA levels on the tRNAGlu/CTC loci chr1-1, chr1-3, chr7-1, chr7-2, and chr4 “EYER” (a region consisting of concatemerically located tRNAGlu/TTC, tRNATyr/GUA, tRNAGlu/CTC, and tRNAArg/ACG genes) were significantly up-regulated (Fig. 8B). Although levels of total full-length tRNAGlu/CTC and tRNAGly/GCC were significantly reduced in corresponding morphants (Fig. 8, C and E), DNA-bound tRNAGlu/CTC and tRNAGly/GCC levels increased more than three- and sixfolds, respectively, when compared to those in control morphants (Fig. 8, D and F). Thus, we propose that the presence of 5′tRFls can prevent homologous full-length tRNAs from forming transcriptionally inhibitory RNA:DNA hybrids on the same tRNA genes.

Fig. 8.

Effect of 5′tRFl knockdown on the formation of RNA:DNA hybrids on tRNA loci in zebrafish embryos.

(A) Experimental procedures. MO and cMO doses: 10 ng per embryo. (B) Increased RNA-bound DNA template strand levels on tRNAGlu/CTC genes in 5′tRFlGlu/CTC morphants. ssDNAs isolated from RNA:DNA hybrids in shield stage embryos were used as templates to amplify indicated tRNAGlu/CTC loci. (C and E) Decrease in tRNA and 5′tRFl levels in morphants. Total RNA extracted by neutral phenol:chloroform was used as template for RT-qPCR. (D and F) Increase in DNA-bound tRNA levels accompanied with decrease in DNA-bound 5′tRFls in morphants. RNA isolated from RNA:DNA hybrids was used as template for RT-qPCR. In (B), amplification signal of each tRNA locus was normalized to that in 5% genomic DNA input. In (C) to (F), all qRT-PCR signals were normalized to the level of Chr1-1 tRNAGlu/CTC locus from DNA input. Data are shown as means ± SD. Student’s t test (two-tailed). ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001. Stages of analyzed embryos: Shield stage in (C) and (D) and sphere stage in (E) and (F).

Given an inhibitory role of tRNA:DNA hybrids in tRNA gene transcription, we hypothesize that removal of tRNAs and 5′tRFls on RNA:DNA hybrids may enhance tRNA gene transcription. To test this idea, we overexpressed human RNASEH1 (hRNASEH1) with addition of nuclear localization signal (NLS) and GFP reporter (Fig. 9A), which is expected to degrade the RNA component of RNA:DNA hybrids (). Wild-type or Tg(Chr21-1:tRNA transgenic embryos injected with 300 pg of NLS-hRNASEH1-GFP mRNA developed normally and showed a significant up-regulation of the F30-Broccoli reporter expression (Fig. 9B) and several tRNA gene expressions including tRNAGlu/CTC and tRNAGly/GCC (Fig. 9C). This result supports the idea that tRNA bound to DNA at tDNA loci represses tDNA transcription.

Fig. 9.

tRNA gene transcription is regulated by tRNAs and 5′tRFls.

(A to C) RNASEH1 overexpression elevates endogenous tRNA levels. Wild-type (WT) (A and C) or Tg(Chr.21-1:tRNA (B) embryos were injected with 300 pg of NLS-hRNASEH1-GFP [composition shown on top of (A)] or GFP mRNA at the 1c stage and immunostained for GFP observation at indicated stages (A) or at the 256c stage for examining expression levels of F30-Broccoli (B) or several tRNAs by RT-qPCR (C), which were normalized to 18S rRNA levels. Note that NLS-hRNASEH1-GFP was enriched in nuclei at 4 hpf. (D to F) Length-dependent regulation of tRNA gene transcription. Three tRNAGlu/CTC oligos with different lengths (D) were synthesized, and they have different minimum free energy (ΔG37°C) and melting temperature (Tm) for DNA:RNA hybrids (E). 1c stage Tg(Chr.21-1:tRNA transgenic embryos were injected with individual RNA oligos at 60 fmol per embryo and harvested at the 256c stage for examining F30-Broccoli expression level by RT-qPCR analysis (F). The reporter expression level was normalized to 18S rRNA level. For data in (B), (C), and (F), the average with ±SD from three independent experiments is given. ns, nonsignificant with P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001 (Student’s t test, two-tailed). (G) Working model. Left: Scenario of tRNA gene (tDNA) transcription in wild-type zebrafish embryos. The major wave of zygotic gene activation for protein-coding genes occurs at the 1kc stage. Before the sphere stage, tDNA transcription may be slow because of tRNA association with the template strand. After the sphere stage, increasing amount of 5′tRFls, possibly with the aid of other factors, takes over the tRNA for binding with the template strand. Right: Scenario of tDNA transcription in embryos depleted of 5′tRFl in which full-length tRNA forms stable tRNA:DNA hybrids on the template strand and thus hinders tDNA transcription. DAPI, 4′,6-diamidino-2-phenylindole.

The stability of RNA:DNA hybrids relies on the length of RNA (). We estimated the minimum free energy (ΔG37°C) for DNA:RNA hybridization and the melting temperature (Tm) of DNA:RNA hybrid for 32-nt 5′tRFlGlu/CTC, 50-nt 5′tRFlGlu/CTC, and 75-nt full-length tRNAGlu/CTC (Fig. 9, D and E). The ΔG37°C decreases with increasing length; however, Tm increases from 73.5°C for the 32-bp hybrid to 83.4°C for the 50-bp hybrid, while the 75-bp hybrid has a Tm of 84°C. It is apparent that longer DNA:RNA hybrids of more than 50 bp long should be more difficult to be dissociated. Then, we synthesized those three RNA oligos and injected them individually into 1c stage Tg(Chr21-1:tRNA transgenic embryos and analyzed the tRNA expression level by RT-qPCR at the 256c stage. Result showed that the reporter expression level was significantly enhanced by 32-nt 5′tRFlGlu/CTC, unaffected by 50-nt 5′tRFlGlu/CTC, and significantly reduced by 75-nt full-length tRNAGlu/CTC (Fig. 9F), confirming the transcription-facilitating effect of 5′tRFl.

DISCUSSION

On the basis of our data primarily from 5′tRFlGlu/CTC and 5′tRFlGly/GCC studies, we could establish a model about 5′tRFls dynamics and function in zebrafish embryos (Fig. 9G). In zebrafish wild-type embryos (Fig. 9G, left), tRNA transcription occurs in two phases, which are separated around the sphere stage (4 hpf). In phase 1, maternal tRNA and its 5′tRFls are present in low amount, tRNA transcription is low probably because a nascent tRNA can bind the template strand of a tRNA gene to attenuate transcription, and 5′tRFls are not enough to competitively bind the same template strand. In phase 2, occurring after the sphere stage, biogenesis of more 5′tRFls promotes tRNA gene transcription with gradually increasing levels; the increasing amount of 5′tRFls allow their high frequent binding to the template strand of the corresponding tRNA gene, which prevents binding of the longer tRNA and so enhances transcription. When 5′tRFls are depleted (Fig. 9G, right), e.g., using antisense MO oligos, lack of competitive binding of 5′tRFls gives opportunity to the nascent tRNAs for binding the template strand, which frustrates transcription by next RNA PolIII transcription machinery and consequently reduces tRNA abundance, and insufficient amount of tRNAs for protein synthesis causes defective embryonic development.

We note that tRNA:DNA hybrids exist in the vast majority of tRNA genes in zebrafish embryos, suggesting that transcription control of tRNA genes by competition of a tRNA and its 5′tRFls may work for other tRNA genes in addition to tRNAGlu/CTC and tRNAGly/GCC. Some studies suggest that local RNA secondary structures are formed as they get transcribed, and the L-shaped tertiary structure is formed only when the 3′ sequence gets transcribed (, ). We hypothesize that, during tRNA gene transcription, the 5′ leader sequence of a tRNA intermediate temporally binds to the template strand due to sequence complementarity, and this binding would impair local RNA secondary structure formation; when the template strand is occupied by a shorter 5′tRFl, the 5′ leader sequence of the tRNA intermediate would not bind to the template strand so that the formation of long tRNA:DNA hybrids is prohibited because of the formation of local secondary structures. However, other mechanisms may be implicated in the function of 5′tRFls. For instance, a 5′tRFl may bind to and present an unknown protein or protein complex to the nascent tRNA intermediate, which dissociates the tRNA intermediate from the DNA template strand; afterward, the 5′tRFl is dissociated from the DNA template strand by specific factors such as components of the RNA PolIII transcription machinery.

We have shown that knockdown of either 5′tRFlGly/GCC or 5′tRFlGlu/CTC caused embryonic lethality in zebrafish embryo (Fig. 3). Previous studies revealed the presence of abundant 5′tRFlGly/GCC and 5′tRFlGlu/CTC in mouse mature sperm and zygotes (, , , , ), and it is likely that they also play an important role in mouse embryonic development. In our in vitro transcription system that contained HeLa cell nuclear lysate, addition of 5′tRFlGlu/CTC-mimetic significantly increased the tRNA reporter transcription (Fig. 5, D and F), which suggests a conserved function of 5′tRFls in promoting tRNA gene transcription in mammalian cells. In this study, a mechanistic insight into 5′tRFl function is learned mainly from 5′tRFlGly/GCC and 5′tRFlGlu/CTC, the most abundant 5′tRFls in zebrafish early embryos, along with a few other 5′tRFls in some assays. Given that there are many 5′tRFl species and their abundance varies with cell types, it is quite possible that some 5′tRFl species may function in distinct mechanisms depending on cellular context.

tRNAFs have been demonstrated to play a role in various biological and pathological processes. Notably, up-regulated expression of tRNAs has been reported in breast cancer (, , ) and multiple myeloma (), and up-regulation of some 5′tRFls has also been shown in prostate cancer (, ). It is conceivable that high proliferation of tumors requires higher levels of tRNAs for vigorous protein synthesis in general. The mechanism of 5′tRFl-promoted tRNA gene transcription may be adopted to ensure efficient transcription of tRNA genes in some types of tumors.

MATERIALS AND METHODS

Animals

Zebrafish Tübingen (TU) line and tp53 mutant line () are maintained in the Meng laboratory. The laboratory animal facility has been accredited by AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care International), and the IACUC (Institutional Animal Care and Use Committee) of Tsinghua University approved all animal protocols used in this study.

Organisms as source for materials used in experiments

Human HeLa cells (gift of Y.-G. Chen, Tsinghua University) were cultured in Dulbecco’s Modified Eagle’s Medium with 10% fetal bovine serum, penicillin (100 U/ml), and streptomycin (100 μg/ml). Culture conditions were maintained at 37°C, with 5% CO2 and humidity. HeLa cells were used as the source materials of the active component in in vitro transcription assay.

Animal maintenance

Fish were maintained at 28.5°C. Naturally spawned fish embryos were collected at the 1c stage and cultured in Holtfreter’s solution. The developmental stages were visually inspected and pictured under a stereomicroscope (Leica).

Oligonucleotides and microinjection

MOs were synthesized from Gene Tools LLC, and all of them were stored at −80°C in a concentration of 20 μg/μl (see table S1). Before microinjection, MOs were heated at 65°C for at least 10 min (95°C for 1 min for 3′tR-MOs), snapped on ice, and then diluted to the required concentration. Short RNA oligos/mimetics were synthesized from GenScript Biotech, and all of them were stored at −80°C in a concentration of 0.6 mM (table S2). Digoxigenin-labeled LNA-DNA chimeric oligos were synthesized from GenScript Biotech and stored at −20°C. DNA oligos, including primers, were synthesized by GENEWIZ (table S3). Biotinylated RNA oligos were synthesized using a 5’EndTag kit (Vector Laboratories, MB-9001), which was performed according to the manual. Biotin (Long Arm) maleimide (Vector Laboratories, SP-1501-12) was used for labeling a long-armed biotin group to the 5′ end of RNA oligos (GFP-r1 or 5′tRFlGlu/CTC-mimetic), which was then purified by phenol:chloroform extraction. mRNAs for GFP and NLS-hRNASEH1-GFP were in vitro synthesized using T7 polymerase. The coding sequence of hRNASEH1 was amplified from HeLa cell RNAs using a pair of specific primers. MO or RNA was injected into the yolk of zebrafish embryos at the 1c stage. The injection doses are given in the figures or figure legends.

In situ hybridization and immunofluorescence

Zebrafish embryos that reached the desired stages were fixed in 4% paraformaldehyde. Digoxigenin-labeled antisense RNA probes were used for in situ hybridization. In situ hybridization and immunofluorescence in zebrafish embryos were performed essentially as before (). Anti-GFP antibody (1:1000 dilution; Santa Cruz Biotechnology, sc-9996) and anti–β-catenin antibody (1:1000 dilution; Sigma-Aldrich, C2206-1ML) were used for immunofluorescence. Embryos were observed by confocal microscopy with a ZEISS LSM710 META microscope.

Puromycin incorporation assay

Puromycin incorporation assay was modified from previously described protocols (). 1c stage embryos were injected with 6 ng of purified glutathione S-transferase–NLS-GFP protein in addition to MO. When the embryos developed to a desired stage, they were treated with puromycin (50 μg/ml; Solarbio) in Holtfreter’s solution for 60 min (sphere stage) or 30 min (shield stage). Thereafter, the embryos were washed and dechorionated in Holtfreter’s solution once and then lysed in TNE buffer [10 mM tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, and 1% Triton X-100] for 10 min on ice. The lysate was cleared by centrifugation, and the supernatant was heated for 10 min at 95°C after addition of SDS loading buffer. The newly synthesized proteins with incorporated puromycin and reference GFP protein were detected by Western blotting. Anti-puromycin (1:5000; Millipore, MABE343) and anti-GFP (1:5000; Easybio, BE2002) antibodies were diluted in 5% nonfat milk. Signals were detected by reaction with Super-ECL (Solarbio) and exposed to x-ray film. The gray values of puromycin and GFP signals were calculated by ImageJ software (National Institutes of Health).

ALKB protein preparation and tRNA pretreatment

The wild-type bacterial ALKB-expressing construct pET28-wtALKB was a gift from C. Yi. Its two mutant forms, ALKB (D135S) and ALKB (D135S and L118V), were made by PCR-mediated mutation using mutated primer pairs (see table S3). The constructs expressed a recombinant ALKB protein with an N-terminal His-tag. Expression and purification of wild-type and mutant ALKB protein were performed according to the previous study () with some modifications. Plasmids were transformed into Escherichia coli BL21 competent cells (Transgen), and positive colonies were incubated in 3 to 4 ml of LB medium overnight for activation. Then, the culture was transferred to 200 ml of LB medium for large-scale incubation at 37°C. Until optical density at 600 nm (OD600) reached 0.6 to 0.8, 1 mM isopropyl-β-d-thiogalactopyranoside (Amresco) was added to the culture to induce the recombinant protein expression at 16° to 18°C. After overnight expression at low temperature, the bacteria were spun down and resuspended in lysis buffer [10 mM tris-HCl (pH 7.4), 300 mM NaCl, 5% glycerol, 2 mM CaCl2, 10 mM MgCl2, 10 mM β-mercaptoethanol, and freshly added 1 mM phenylmethylsulfonyl fluoride (PMSF)]. Bacterial cells were lysed by ultrasonication, and then the supernatant was kept after centrifugation. His-tagged ALKB protein was captured by incubating the supernatant with Ni–nitrilotriacetic acid beads (QIAGEN) for 1 hour at 4°C on a vertical rotator. The captured protein was washed sequentially with lysis buffers containing 1 M NaCl or 20 mM imidazole (Sigma-Aldrich). Last, the ALKB protein was eluted from beads using the elution buffer that contains 250 mM imidazole. The eluted protein was verified by electrophoresis and Coomassie Blue staining.

To remove the potential contaminants from the bacteria, the eluted protein was further purified by ion exchange chromatography. RESOURCE S cation exchange chromatography column (GE) was loaded on AKTA Purifier 10 (GE) and washed with solution A [20 mM MES (pH 6.0)] until OD280 was proximately 0. Then, the purified ALKB protein was diluted with 10 volumes of solution A, followed by filtration with a 0.22-μm filter. Samples were loaded onto and captured by the column. Then, by setting a 100% gradient with solution B [20 mM MES (pH 6.0) and 1 M NaCl], fractions were collected from the first peak till the last peak. Proteins of all fractions were detected by Coomassie Blue staining. Fractions that were enriched for ALKB protein were combined and concentrated by an Amicon centrifugal filter (Millipore; 10 kDa), and the solution was changed to 2× storage buffer [40 mM tris-HCl (pH 8.0), 400 mM NaCl, and 4 mM dithiothreitol (DTT)]. An equal volume of glycerol was added to the solution and thoroughly mixed on a rotator at 4°C. Aliquoted wild-type and mutant ALKB proteins were snap-frozen in liquid nitrogen and then stored at −80°C.

Before RT-qPCR, total RNA was extracted from wild-type embryos or morphants, followed by sequential treatment with 100 mM tris-HCl (pH 9.0) buffer and T4 Polynucleotide Kinase (PNK) for 30 min each time. Then, 5 μg of purified RNA was treated by ALKB protein mixture [wtALKB, 160 pmol; ALKB (D135S), 200 pmol; and ALKB (D135S/L118V), 200 pmol] in reaction buffer [300 mM KCl, 2 mM MgCl2, 50 μM (NH4)2Fe(SO4)2, 300 μM α-ketoglutarate, 2 mM vitamin C, bovine serum albumin (50 μg/ml), 50 mM MES (pH 5.0), 40 U of RNase inhibitor, and 1× protease inhibitor] for 2 hours at room temperature. Thereafter, RNA was purified by phenol:chloroform extraction, followed by E. coli polyadenylate [poly(A)] polymerase treatment for poly(A) tail addition. Last, 1 μg of RNA was used for reverse transcription using SuperScript IV reverse transcriptase (Thermo Fisher Scientific, 18090050) at 55°C. The acquired cDNA was used as template for tRNA RT-qPCR using tRNA-specific primers and TransStart Top Green qPCR Super Mix (Transgen). For activity verification of ALKB protein mix, the same amount of RNA was used for pretreatment, but 10 mM EDTA was added to the ALKB reaction buffer, which would chelate the Fe2+ ion required for ALKB activity. Last, both reactions, with or without EDTA, were purified and treated as described previously.

Small RNA library construction and sequencing

About 50 to 100 zebrafish eggs or embryos at a desired stage were collected to extract total RNA and dechorionated by pronase digestion. Embryos were then transferred to a 1.5-ml Eppendorf tube and lysed in 1 ml of TRIzol reagent (Thermo Fisher Scientific, 15596018). The lysate was centrifuged at 12,000 rpm for 10 min, and the supernatant was collected and mixed with 200 μl of chloroform in a fresh tube. After centrifugation at 12,000 rpm for 10 min at 4°C, the supernatant was transferred to a fresh tube with addition of equal volume of isopropanol and incubated at room temperature for 15 min, followed by centrifugation at 12,000 rpm for 10 to 20 min. The RNA pellet was washed with 75% ethanol once and stored as pellet in 75% ethanol at −80°C until all samples were collected. When needed, the pellets were dried and dissolved in nuclease-free water.

Two pretreatment steps were included: First, purified total RNAs were treated with T4 PNK (NEB, M0201S) with adenosine 5′-triphosphate for 30 min at 37°C and then purified by phenol:chloroform extraction. The aim of this step is to remove the 2,3′-cyclic phosphate group from the 3′ end of 5′tRFls and add the phosphate group to the 5′ end of 3′tRFls. In the second step, total RNAs were treated with ALKB protein mix as mentioned above. Purified total RNAs were then subjected to small RNA library preparation.

Small RNA libraries, using 1 μg of total RNA each, were prepared using the NEBNext Multiplex Small RNA Library Prep Set for Illumina (NEB) with modification. In the first step, T4 RNA ligase 2 truncated KQ (NEB, M0373L) was used for 3′ adaptor ligation. The amplified DNA library from each sample was first concentrated by DNA concentrator spin column (Zymo Research, D4003), and then bands of 135 to 160 bp, about the length of 15- to 40-nt RNA ligated with both full adaptors, were manually selected on native PAGE gel. DNA bands were eluted from the gel piece with 0.3 M NaOAc (pH 4.5), followed by precipitation with addition of 2.5 volumes of ethanol and 1 μl of GlycoBlue. The concentration and quality of purified DNA libraries were checked by Bioanalyzer 2100 (Agilent Technologies). Then, the libraries were loaded on a NovaSeq 6000 sequencer (Illumina) for sequencing, and only reads from R1 were left.

For data processing, reads were quality-checked by FastQC, and adaptors were cut off by Cutadapt (), and only reads that lay between 18 and 40 nt were kept. The reads of each sample were first aligned to the GRCz11 zebrafish reference genome and ensemble cDNA reference sequences using Bowtie2 (). The aligned sequences were further sequentially aligned to the tRNA database (GtRNAdb and GRCz11), miRNA database (miRbase), 45S rRNA (), Piwi-interacting RNA (piRNA) database (IsopiRBank), small nucleolar RNA (snoRNA) (snOPY), tandem repeats (TRDB), and transposable element (Dfam).

tRNA library construction and sequencing

Total RNAs isolated from zebrafish embryos were dissolved in 10% PEG8000 and 0.5 M NaCl, placed on ice for 30 min, and then centrifuged at 14,000 rpm for 15 min at 4°C. The supernatant containing low–molecular weight (LMW) RNAs was collected and precipitated by addition of ethanol. The LMW RNAs were separated by 8% 8 M urea PAGE gel, and then the bands of 60 to 120 nt lengths were cut and crunched into small pieces. RNAs were eluted from the gel by overnight incubation with 0.3 M NaOAc on a vertical rotator at 4°C. Then, the supernatant was precipitated at −20°C after the addition of 2 μl GlycoBlue and 2.5 volumes of ethanol. After centrifugation at 12,000 rpm for 30 min and washing by 75% ethanol, the RNA pellet was dissolved in RNase-free water.

The pretreatment steps for purified tRNAs were the same with the procedure for tRNA RT-qPCR except for poly(A) tail addition. RNAs were first incubated in 100 mM tris-HCl (pH 9.0) at 37°C for 30 min then treated with T4 PNK to repair ends at 37°C for 30 min. Modifications on tRNAs were removed by ALKB treatment for 2 hours at room temperature. Purification was needed after each pretreatment step. RNAs (100 ng) were used for RNA library preparation using VAHTS Small RNA Library Prep Kit for Illumina (Vazyme, NR801). The libraries were first concentrated by DNA concentrator spin column (Zymo), followed by separation on 6% native PAGE gel. Bands of 180 to 240 bp lengths (i.e., 60- to 120-nt RNA ligated with both adaptors) were cut, and the DNA was eluted from the gel. The libraries were sequenced with the strategy of PE150 on an Illumina NovaSeq (Novogene) machine. Sequencing data were analyzed on Galaxy platform (https://usegalaxy.org/). Adaptor trimming and quality control were performed by FastP and Cutadapt (), and then 60- to 100-nt-length reads were preserved for downstream analysis. Reads were first mapped to genome (GRCz11) by Bowtie2 with very-sensitive-local mode (-D 20 -R 3 -N 0 -L 20 -i S,1,0.50). Then, the aligned reads were further mapped to tRNA database (GtRNAdb and GRCz11), 5S rRNAs, 5.8S rRNAs, and mitochondrion genome sequentially. The counts of reads aligned to one tRNA isoacceptor/isodecoder were calculated and normalized to all reads mapped to genome for downstream analysis.

Small RNA preparation and qPCR

Total RNAs were extracted and pretreated as mentioned above for small RNA sequencing. After treating with poly(A) polymerase, total RNA was then used for synthesizing the first strand cDNA with GoScript RT Mix Oligo(dT) (Promega, A2791). The reverse transcription process took place at 55°C. cDNA was diluted to 100 μl with double-distilled H2O (ddH2O), and then the same aliquot of each sample was used for qPCR.

For testing the specificity of 5′tRFl primers, we designed and synthesized ssDNA oligos (GENEWIZ; see table S3) that mimic the first-strand cDNAs of full-length tRNA and 5′tRFls. These oligos were diluted to 2 nM, 0.4 nM, 80 pM, and 16 pM with ddH2O and used as PCR templates.

Northern blotting and electrophoretic mobility shift assay

Procedures of Northern blot refer to a nonradioactive protocol () with some modifications. Native PAGE gel was prepared by only 1× tris-borate EDTA (TBE) buffer and 19:1 acrylamide/bis-acrylamide solution (Sigma-Aldrich, A2917), while for denature gels, urea was added to 8 M final concentration. Denature gels need to prerun for at least 30 min to warm up the gels. Total RNA mixed with 50% formamide was heat-denatured at 95°C for 5 min before loading. Gel was run at 200 V until bromophenol blue reached the bottom of the gel and then stained with GelSafe dye (Yph-bio, EP106) in 0.5× TBE buffer. RNA was wet-transferred to nylon membrane (GE, RPN303B) in 400 mA for 1 hour with 0.5× TBE buffer. Membrane was cross-linked by ultraviolet and then blocked with UltraHyb (Thermo Fisher Scientific, AM8670) at 42°C for 30 to 60 min. Digoxigenin-labeled LNA-DNA probe was denatured at 95°C for 1 min and then snapped on ice. Suitable probe concentration was 1 nM for αGlu-LNA and 0.1 nM for αGly-LNA. Probe was added to the block reagent and then hybridized with membrane for 12 to 14 hours at 42°C. After hybridization, the membrane was washed at 42°C as follows: low stringent buffer (2× SSC and 0.1% SDS) for 15 min twice, high stringent buffer (0.2× SSC and 0.1% SDS) for 5 min twice, and wash buffer (1× SSC) for 10 min. Then, the membrane was blocked with 1% block reagent (Roche, 11096176001) in MABT (Maleic acid buffer with Tween 20) at room temperature for 3 hours. Anti-digoxigenin POD (peroxidase) (1:3000; Roche, 11207733910) was added and incubated at room temperature for 1 hour or at 4°C overnight. The membrane was washed by MABT for 15 min for four times, then reacted with ECL-plus (Solarbio), and exposed to x-ray film. T7-transcribed tRNAs and/or synthesized 5′tRFl-mimetics were loaded as molecular weight markers and controls.

Total RNAs for electrophoretic mobility shift assay (EMSA) experiment were extracted by neutral phenol:chloroform as acidic phenol and TRIzol may prefer single-stranded nucleic acids. Embryos were dechorionated by pronase at indicated stages and then lysed in genome extraction buffer [200 mM NaCl, 10 mM tris-HCl (pH 8.0), 10 mM EDTA, 0.5% SDS, and freshly added Proteinase K (200 μg/ml)] immediately. The lysate was incubated at 37°C for 2 hours and then extracted by 1 volume of phenol:chloroform (pH 7.8) and chloroform sequentially. Total nucleic acids were precipitated by NaOAc and ethanol. For EMSA, RNA should not be heat-denatured before loading, and the loading buffer contained glycerol instead of formamide. For band shift indicator, T7-transcribed tRNAs or synthesized 5′tRFl-mimetics were incubated with antisense MOs or LNA oligos at 95°C for 1 min and then slowly cooled to room temperature to form hybrids.

tRNA purification and refolding

T7 promoter sequence was simply added to the upstream of the zebrafish or the yeast tRNAGly/GCC sequence, and PCR-amplified templates were used for in vitro transcription (Thermo Fisher Scientific, AM1334). Transcription of tRNAGlu/CTC was much more complex as its first base is T, which is not preferred by T7 polymerase. Therefore, hammer head ribozyme and hepatitis delta virus ribozyme sequences were added to the upstream and downstream of the tRNAGlu/CTC sequence, respectively, to enhance the transcription product yield and to maintain a right 3′ end (). Both T7-transcribed tRNAGly/GCC and tRNAGlu/CTC were separated and purified from denature PAGE gel. tRNAGlu/CTC needed further treatment with T4 PNK to remove the 3′ end phosphate group.

Native tRNAs were enriched according to a published paper (). Biotin-labeled antisense DNA oligos were linked to streptavidin beads (Thermo Fisher Scientific, 11206D). Zebrafish total RNAs or yeast crude tRNA extract (100 mg; Solarbio, T8630) were added to the beads and incubated in 1× TMA buffer [10 mM tris-HCl (pH 7.6), 0.9 M tetramethylammonium chloride, and 0.1 mM EDTA] at 35°C for 10 min, then supernatant was discarded, and beads were washed several times with 10 mM tris-HCl (pH 7.6).The bead-bound tRNAs were eluted by incubation with 10 mM tris-HCl (pH 7.6) at 65°C for 5 min, and then the supernatant was removed immediately to a new tube. Native tRNAs were purified from denature PAGE gel.

Purified T7-transcribed and native tRNAs were refolded before microinjection. tRNAs were diluted in RNase-free water to a desired concentration, then incubated at 80°C for 2 min, and followed by incubation at 60°C for 2 min. An equal volume of 20 mM MgCl2 was added and mixed thoroughly. tRNAs were then incubated at room temperature for 5 min and then placed on ice for 30 min.

In vitro translation assay

Flexi Rabbit Reticulocyte Lysate System (Promega, L4540) was used for in vitro translation assay according to the manual. Each reaction was assembled as follows: 9.9 μl of RRL, 0.3 μl of 1 mM complete amino acid mixture (Promega, L4461), 0.42 μl of 3 M KCl, 0.39 μl of RNase inhibitor, 1 μl of GFP mRNA (500 ng/μl), 0.5 μl of purified NLS-2xFlag-mCherry-HisTag protein (3 mg/ml), and 1.49 μl of nuclease-free water. A premix was made each time for multiple samples. When needed, 120 nmol of RNA oligo or 2 μg of MO was added to each reaction. The reaction took place at 30°C for 1 hour. To stop the reaction, SDS loading buffer was added to denature the sample at 95°C for 10 min. The translation products were detected with regular Western blot procedure using anti-GFP (1:10,000 dilution; Easybio, BE2002) and anti-mCherry antibody (1:5000 dilution; Easybio, BE2027).

DNA:RNA immunoprecipitation

The DRIP-seq experiments and data processing procedures were essentially similar to ssDRIP-seq (, ). Embryos were collected at desired stages with two replicates and lysed in genome extraction buffer [200 mM NaCl, 10 mM tris-HCl (pH 8.0), 10 mM EDTA, 0.5% SDS, and freshly added Proteinase K (200 μg/ml)] at 37°C for 8 to 12 hours. Then, 1/2 volume of 3 M potassium acetate solution (pH 5.2) was added to the lysate and mixed gently, followed by incubation on ice for 20 min. Following extraction with equal volume of phenol:chloroform:isoamyl alcohol [25:24:1 (pH 7.8)] and chloroform sequentially, genomic DNA, which included RNA:DNA hybrids, was precipitated by addition of equal volume of isopropanol for 10 min at room temperature. The precipitate was washed by 70% ethanol for once and then air-dried.

Genomic DNA fragmentation was performed by digestion at 37°C for 4 to 6 hours with a set of restriction enzymes, including Alu I, Mbo I, Dde I, and Mse I (NEB, R0137V, R0147V, R0175V, and R0525V). For the negative control sample, genomic DNA was digested with RNase H (NEB, M0297S) at 37°C overnight in advance. After digestion, the fragmented genomic DNA was purified by phenol:chloroform extraction and quantified by Qubit 3.0 (Invitrogen). To immunoprecipitate RNA:DNA hybrids, 5 μg of genomic DNA was diluted to 440 μl with nuclease-free water, then to it, 50 μl of 10× DRIP binding buffer [100 mM NaPO4 (pH 7.0), 1.4 M NaCl, and 0.5% Triton X-100] and 10 μg of S9.6 antibody (American Type Culture Collection, HB-8730) were added. The mixture was incubated on a vertical rotator for 12 to 14 hours at 4°C. After addition with 50 μl of Protein G Dynabeads (Invitrogen, 10004D), the mixture was incubated for 4 hours further. Then, beads with its conjugates were spun down and washed by 1× DRIP binding buffer four times at room temperature. Elution buffer [50 mM tris (pH 8.0), 10 mM EDTA, and 0.5% SDS] and Proteinase K were added to the beads and incubated for 40 min at 55°C. DNA was purified by phenol:chloroform:isoamyl alcohol [25:24:1 (pH 7.8)] extraction and then subjected to library construction using the Accel-NGS 1S Plus DNA Library Kit (Swift Biosciences) or DRIP-DNA qPCR. The libraries were sequenced on a HiSeq X Ten sequencer (Illumina).

For ssDRIP-seq data processing, the adaptors were trimmed off by Cutadapt (). Low-quality bases and the first seven bases of each end were discarded from all reads. In addition, only reads longer than 50 bp were kept. Then, the zebrafish GRCz11 reference genome was used for reads alignment, which was accomplished by Bowtie2 () with default parameters. SAMtools () was used to remove reads with low mapping quality, and Picard tools (http://broadinstitute.github.io/picard) were used to remove duplicated reads.

Mapped reads were split into forward and reverse R loop reads by SAMtools according to their orientation on chromosomes. Forward reads mean that the ssDNAs in the R loop structure locate on the Watson strand, which makes it also named as wR-loop; on the contrary, reverse reads, also called cR-loop, indicate ssDNAs located on the Crick strand. The aligned reads files (BAM) were converted to normalized coverage files (bigWig) with 5-bp bins using bamCoverage from deepTools (). Normalization was performed using bamCoverage from deepTools, with read coverage normalized to 1× sequencing depth (also known as reads per genomic content). Integrative Genomic Viewer () was used for visualization. MACS2 () was used to identify peaks in individual replicates, and peaks of <100 bp were discarded to reduce the noise during data analysis. Metaplots were generated with deepTools and R scripts. Normalized R loop signal on genes and differential R loop genes were identified by DESeq2 ().

For isolating DRIP–pull-down RNA, the procedure was almost the same as that for DRIP-DNA described above, but diethyl pyrocarbonate–treated buffers and RNase-free tubes and tips were used. For semi-qPCR, 10% input material of each sample was kept before DRIP. For DRIP-RNA qPCR, several parts of the starting material (containing 5 μg of genomic DNA, equal to 100% input) were kept for RNA input and DNA input. The DNA input was treated by RNase A (Thermo Fisher Scientific, EN0531) overnight at 37°C, and then the DNA strands were precipitated by ethanol. After DRIP, the DRIP product and RNA input were incubated with deoxyribonuclease I (DNase I) (NEB, M0303S) for 20 min at 37°C, and then the RNA strands were extracted by TRIzol, followed by treatment with T4 PNK and E. coli poly(A) polymerase sequentially. cDNA of both RNA input and DRIP product was synthesized by GoScript RT Mix Oligo(dT) and used as template to amplify full-length tRNAs and 5′tRFls by semi-qPCR/qPCR using specific primer pairs. The signals from DRIP and RNA input samples were further normalized to DNA input signals, which were reflected by qPCR with DRIP1-1 primer pairs (see table S3).

Chromatin immunoprecipitation

ChIP was performed on the basis of a previous report with some modifications (). Briefly, 1000 to 1500 embryos at 6 to 8 hpf were collected for each sample. After dechorionation, these embryos were cross-linked by treatment with 1% formaldehyde in Holtfreter’s solution for 10 min at room temperature. Reaction was quenched by addition of 0.125 M glycine for 5 min. After yolk removal, associated cells were spun down and collected in 1.5-ml Eppendorf tube. Cells were then lysed in 2 ml of SDS lysis buffer [1% SDS, 10 mM EDTA, 50 mM tris-HCl (pH 8.0), and 1× protease inhibitor cocktail] on ice for 10 min, followed by ultrasonication to fragmentate chromatins. After clearance by centrifugation, 90 μl of lysate was kept as DNA input, and the rest of the supernatant was divided to two equal 900-μl aliquots, and each aliquot was diluted 10 times with ChIP dilution buffer [0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM tris-HCl (pH 8.0), 167 mM NaCl, and 1× protease inhibitor cocktail]. Ten micrograms of anti-POLR3A antibody (Abcam, ab96328) or immunoglobulin G (IgG) (Cell Signaling Technology, 2729S) was added to the lysate, followed by incubation on a vertical rotator for 12 to 16 hours at 4°C. Magna ChIP Protein A + G magnetic beads (20 μl per sample; Merck, 16-663) were washed with the ChIP dilution buffer three times and then added to the mixture with incubation for 2 hours at 4°C. Bead/antibody/chromatin complexes were washed once sequentially in the following buffers: low-salt buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM tris-HCl (pH 8.0), and 150 mM NaCl], high-salt buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM tris-HCl (pH 8.0), and 500 mM NaCl], LiCl buffer [0.25 M LiCl, 1% NP-40, 1% deoxychloride, 1 mM EDTA, and 10 mM tris-HCl (pH 8.0)], and TE buffer [10 mM tris-HCl (pH 8.0) and 1 mM EDTA] twice. Each wash step lasted for 4 min on a vertical rotator. Last, beads were resuspended in elution buffer (1% SDS and 50 mM NaHCO3) with Proteinase K (0.2 mg/ml), followed by incubation at 65°C for at least 6 hours for reverse cross-link. Then, DNA was extracted by phenol:chloroform and precipitated, followed by qPCR. For qPCR, the same volume of ChIP product from each sample was used, and qPCR was performed with ChIP-specific primers (see table S3) and TransStart Top Green qPCR Super Mix (Transgen). The ChIP-qPCR signals were normalized to that of DNA input.

In vitro tRNA reporter transcription assay and transgenic fish line

HeLa nuclear lysate was prepared on the basis of previously reported methods (, ). Briefly, HeLa cells were cultured on three 15-cm dishes until they were 90% confluent. After washing by phosphate-buffered saline buffer once, all liquid was aspirated, and cells were scrapped from the plate and then transferred to a 1.5-ml Eppendorf tube. The packed cell volume (PCV) was equal to the volume of the cell pellet after centrifugation at 100×g for 5 min. The cell pellet was resuspended in 3× PCV of hypotonic buffer [10 mM Hepes (pH 7.9), 1.5 mM MgCl2, 10 mM KCl, 0.2 mM PMSF, and 0.5 mM DTT] twice and then incubated in hypotonic buffer after the second resuspension for 10 min at 4°C. Cells swelled in hypotonic buffer, which made their volume become larger. Using a chilled Dounce homogenizer, the cytoplasmic membrane was broken by squeezing, which could be checked by staining with trypan blue. Then, the nuclei were precipitated by centrifugation at 1500×g for 5 min, and the volume of the pellet was called as packed nuclear volume (PNV). Afterward, 0.5× PNV of low-salt buffer [20 mM Hepes (pH 7.9), 1.5 mM MgCl2, 20 mM KCl, 0.2 mM EDTA, 25% glycerol, 0.2 mM PMSF, and 0.5 mM DTT] was added to the nuclei pellet with gentle mixing, and 0.5× PNV of high-salt buffer [20 mM Hepes (pH 7.9), 1.5 mM MgCl2, 1.4 M KCl, 0.2 mM EDTA, 25% glycerol, 0.2 mM PMSF, and 0.5 mM DTT] was added to the suspension with quick inversion and subsequent rotation for 30 min at 4°C. Next, the lysate was centrifuged at 4°C for 15 min at 18,000×g. The supernatant was loaded on a chilled mini-centricon (Millipore, UFC500396) and centrifuged at 4°C for 50 min at 14,000×g. Next, every 45 μl of remaining extract was dialyzed in mini-dialysis Slide-A-Lyzers (Thermo Fisher Scientific, 69572) in dialysis buffer [20 mM Hepes (pH 7.9), 100 mM KCl, 0.2 mM EDTA, 20% glycerol, 0.2 mM PMSF, and 0.5 mM DTT] for 2 hours at 4°C. Last, the extract was aliquoted to a small volume, snap-frozen in liquid nitrogen, and then stored at −80°C.

To make tRNAGlu/CTC gene transcription reporter constructs, primers were designed to amplify the upstream sequence together with a tRNAGlu/CTC gene (e.g., chr1-1, chr16-2, or chr21-1), and the products were cloned into the pST–F30-Broccoli plasmid (table S3). When combining these three constructs with equal molar ratio, the plasmid mixture was called as pST-Chr.X-tRNAGlu/CTC–F30-Broccoli, which served as the template for in vitro transcription experiment. F30-Broccoli sequence was synthesized according to the published paper ().

An in vitro transcription reaction was assembled as follows: 30% HeLa nuclear extract, 12 mM Hepes, 12% glycerol, 0.3 mM DTT, 0.12 mM EDTA, 60 mM KCl, 12 mM MgCl2, 600 μM nucleoside triphosphate each, 1× protease inhibitor cocktail (Roche), 20 U of RNase inhibitor, 40 pg of GFP mRNA, 1 μg of plasmid mix, and, if needed, 1 μg of MO. Reaction started just after the addition of nuclear extract, and so, it was added last. GFP mRNA served as reference RNA, which was used to estimate the relative transcription level of the reporter F30-Broccoli RNA. Reaction was incubated at 30°C for 1 hour. To stop reaction, 1 ml of TRIzol reagent was added, followed by RNA purification. The purified RNA was used as template for specific reverse transcription, and the F30-Broccoli reporter RNA, GFP mRNA, and human U6 small nuclear RNA (snRNA) were detected by PCR using specific primers (see table S3). The qPCR signals of F30-Broccoli RNA were normalized to that of GFP mRNA.

For performing DNA pull-down assay in in vitro transcription system, transcription reaction system was assembled almost the same as that described above, but the volume was tripled. A premix without RNA oligos was made in advance, and then it was divided into two equal parts, to each was added 200 ng of Biotin-labeled 5′tRFlGlu/CTC-mimetic or GFP-r1 RNA oligo. Reactions were mixed gently and thoroughly and then incubated at 30°C for 1 hour. During this period, 50 μl per sample of Dynabeads M-280 streptavidin (Thermo Fisher Scientific, 11206D) was prepared by sequential washes with 1× B&W (Binding and washing) buffer, solution A buffer, and solution B buffer according to the manual. The beads were further washed two to three times in 1× Dignam buffer without glycerol (12 mM Hepes, 0.3 mM DTT, 0.12 mM EDTA, 60 mM KCl, and 12 mM MgCl2). The washed beads were added to each reaction, followed by mixing thoroughly and incubation at room temperature for 30 min. The RNA-plasmid complex was captured and washed five times by 1× Dignam buffer without glycerol, and then the plasmids were extracted by phenol:chloroform. Plasmids were detected by qPCR using F30-Broccoli–specific primers.

For transgenic fish line establishment, Tg(Chr.21-1:tRNA plasmid was coinjected with Tol2 transposase mRNA into 1c stage embryos. Injected embryos were grown up, and their transgenic offspring were identified by PCR using F30-Broccoli–specific primers. F1 male transgenic fish were mated to wild-type female, and their embryos were collected for experiments. For RT-qPCR, purified total RNAs were first denatured at 80°C for 5 min and then reverse-transcribed by specific primers (100 pmol of F30Brocc-reverse and 100 fmol of 18S-RT-as2 or 5.8S-rRNA-as per reaction) at 55°C for 1 hour by GoScript reverse transcriptase (Promega).

Quantification and statistical analysis

All data came from at least three independent experiments and are shown as means ± SD unless otherwise stated. Significance of differences between the means of most experiments was analyzed using Student’s t test (two-tailed). The Wilcoxon matched-pairs signed-rank test was used for DRIP-seq data. All significance levels are indicated by the following: not significant (ns); *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.