N1-methylpseudouridine mRNA modification enhances efficiency and specificity of gene overexpression by preventing Prkra-mediated global translation repression

Abstract

In vitro transcribed messenger RNA (IVT mRNA) has emerged as a pivotal tool in mRNA-based therapies and has been extensively employed in gene function studies and genetic tool applications. However, the IVT process generates double-stranded RNA (dsRNA) by-products that are recognized by dsRNA sensors, triggering innate immune responses. In this study, we comprehensively analyzed the detrimental effects of dsRNA by-products on early zebrafish embryos, revealing that these by-products induce cell necrosis and delay maternal–zygotic transition (MZT) by reducing global translation efficiency via Prkra (Protein Activator Of Interferon Induced Protein Kinase; also called PACT in mammals), a dsRNA sensor recently identified in pluripotent cells. Importantly, we demonstrate that N1-methylpseudouridine (m1Ψ) modification of IVT mRNAs effectively mitigates these adverse effects, as m1Ψ-modified dsRNAs exhibit significantly lower binding affinity to the Prkra dimer. Our findings underscore a previously overlooked challenge in the use of IVT mRNA in early embryos and offer a robust solution to enhance the fidelity of mRNA applications. Furthermore, we elucidate that m1Ψ modification minimizes the dsRNA-induced stress response in pluripotent cells through a distinct mechanism.

Graphical Abstract

Graphical Abstract.

Introduction

In vitro transcribed messenger RNA (IVT mRNA) is extensively utilized in gene gain-of-function studies and serves as a fundamental component in mRNA-based therapies, with mRNA vaccines against SARS-CoV-2 being the most notable example [1, 2]. IVT process accompanies the production of double-stranded RNA (dsRNA) by-products [3, 4], which are recognized as virus-like non-self molecules by pattern-recognition receptors (PRRs). These include endosomal toll-like receptors such as TLR3, as well as cytosolic sensors like retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated 5 (MDA5) [5–7]. Activation of these PRRs triggers signaling pathways that elevate the expression of interferons (IFNs) and pro-inflammatory cytokines. They initiate outside-in JAK/STAT cascades that upregulate numerous genes, including interferon-stimulated genes (ISGs) involved in antiviral responses [8, 9]. Furthermore, dsRNA can activate an integrated stress response (ISR) by binding to and activating protein kinase R (PKR, EIF2AK2), which inhibits global protein translation through the phosphorylation of the translation initiation factor eIF2α [10, 11]. Uridine modifications, such as pseudouridine (Ψ) and 2-thiouridine (S2U), effectively diminish the recognition of dsRNA by sensors including RIG-I, PKR, and TLR3 [12–16]. These strategic alterations enhance the potential for vaccination while minimizing associated inflammatory responses. Notably, N1-methylpseudouridine (m1Ψ) modification of mRNAs elicits an even lower immune response than Ψ while supporting enhanced translation activity [17].

These conclusions are primarily drawn from studies on differentiated cells. However, our recent findings reveal that in pluripotent stem cells, dsRNA is detected by the dimerized dsRNA-binding protein Prkra (Protein Activator Of Interferon Induced Protein Kinase; also called PACT in mammals), circumventing the canonical dsRNA sensors and PKR–eIF2α pathway [18]. Prkra is activated by dsRNA through a highly ordered binding mechanism. Prkra–dsRNA complex leads to a global translation inhibition by the sequestration of translation machinery components, a process facilitated by dimerized dsRNA-binding domain 3s (dsRBD3s) [18]. The impact of this recently identified dsRNA-induced stress mechanism on the utilization of IVT mRNA in early embryos and stem cells remains unexplored. Additionally, it is still unclear whether and how m1Ψ mRNA modification affects the Prkra-mediated stress response in undifferentiated cells.

The use of unmodified IVT mRNA in developmental biology has a long-standing history. The first mRNA injection into Xenopus oocytes dates back to 1983 [19], while for zebrafish, the first mRNA overexpression record was published in 1994 [20]. Since then, mRNA-based gene overexpression has become a standard approach for gene gain-of-function studies and rescuing gene loss-of-function phenotypes. More importantly, genetic tools for transgenesis and genome editing depend highly on IVT mRNAs encoding exogenous functional proteins like Tol2 transposase, Cas9, etc. [21–23]. However, because the side effects of IVT mRNA are often interpreted as the function of its encoded protein, the risk of introducing IVT mRNA in early embryos has been largely ignored for decades.

In this study, we examined the detrimental effects of dsRNA by-products generated during the IVT of mRNA intended for gene overexpression and genetic tool application. These dsRNA by-products often lead to extensive cell necrosis, delay of maternal–zygotic transition (MZT) and the emergence of misleading phenotypes, which can hinder the functional rescue by wild-type mRNA and reduce the effectiveness of mRNA-based tools in genome editing and transgenesis. Importantly, we found that m1Ψ modification significantly mitigates the toxic effects of dsRNA in early embryos. Mechanistically, m1Ψ modification markedly decreases the binding of Prkra to dsRNA, thereby alleviating the activation of Prkra-induced global translation repression in the early embryos. Our findings provide a robust solution for achieving accurate gene overexpression in early zebrafish embryos, enhancing the fidelity and reliability of mRNA-based genetic tools in this model system. Additionally, we present evidence and the underlying mechanism that m1Ψ modification minimizes the dsRNA-induced stress response in pluripotent cells.

Materials and methods

Resource of all materials were listed in Supplementary Table S1.

Animals

Zebrafish AB strain, as well as nanog and prkra mutant lines, were kept in standard breeding systems (Haisheng) at a temperature of 28°C. Adult fish were fed with live brine shrimp larva twice a day at 10 am and 4 pm, respectively. Fish at the age of 6–12 months were selected for mating. All experiments adhered to the ARRIVE guidelines and were approved by the Ethics Committee for Animal Research of the School of Life Sciences at Shandong University (permit number SYDWLL-2023-069).

Generation of maternal and zygotic mutants in zebrafish

To rapidly generate prkra maternal and zygotic mutants (MZprkra), we utilized a CRISPant strategy. Five single-guide RNAs (sgRNAs) (Supplementary Table S2) targeting the open reading frames (ORFs) of prkra were designed using the online software CRISPRscan (https://www.crisprscan.org/) [24]. These sgRNAs were synthesized via standard IVT and purification [24]. The sgRNAs (100 pg/embryo in total) and Cas9 protein (200 pg/embryo) were injected into fertilized eggs. To assess the editing efficiency, injected embryos at 10 h postfertilization (hpf) were homogenized for genomic DNA preparation in 50 mM NaOH for 20 minutes (min) at 95°C, followed by neutralization with 1/10 volume of 1 M Tris–HCl (pH 7.5). Then, the DNA fragment flanking the sgRNA sites were amplified by polymerase chain reaction (PCR) and subjected to Sanger sequencing. Next, the sequencing chromatograms were uploaded to http://shinyapps.datacurators.nl/tide/ to estimate the editing efficiency following the TIDE algorithm. The CRISPant F0 embryos were then raised to adulthood, and the resultant male and female fish were crossed to produce MZprkra mutant embryos. These embryos were either raised to produce a stable mutant line or subjected directly to experimental assessment. They were genotyped by reverse transcription PCR (RT-PCR) and Sanger sequencing to confirm the loss of Prkra function.

Mnanog mutant embryos were generated following a previously established procedure [25, 26]. sgRNAs targeting nanog ORF were cloned into pU6-sgRNA vectors and then the sgRNA expression cassettes were excised and tandemly aligned into a pDest-3sgRNA vector via Golden Gate ligation. This vector was purified and coinjected with I-Sce I meganuclease for transgenesis. The GFP positive offspring were collected to examine the characteristic nanog mutant phenotype.

mRNA and dsRNA synthesis

The gfp reporter mRNAs and mRNAs analyzed in Figs 1, 2, 4, 5 and Supplementary Figs S1 and S5 were transcribed from the pCS2 vector containing the SV40 3′ UTR. To examine how m1Ψ affects translation efficiency, we omitted the SV40 3′ UTR to avoid possible interference of m1Ψ on cytoplasmic polyadenylation. Consequently, the mRNAs used in Figs 6 and 7 were truncated in 3′UTR and directly polyadenylated with the Poly(A) Tailing Kit (Thermo Fisher, AM1350).

Figure 1.

dsRNA in IVT mRNA triggers unspecific toxic effects in zebrafish embryos. (A) Time-lapse imaging of cas9 IVT mRNA-injected embryos, highlighting yolk cell rupture during gastrulation. (B, B’, C, and C’) α-Tubulin antibody staining (B and B’), and phalloidin staining (C and C’). Asterisks mark the cytoskeleton-absent “holes” on the surface of the yolk cell. (D) Aberrant phenotypes of cas9 mRNA-injected embryos at 36 hpf. (E) Phenotypic analysis of cas9 mRNA or Cas9 protein-injected embryos with injection dose per embryo indicated. (F) Western blot analysis to determine Cas9 protein level in mRNA and protein-injected groups. (G) The nucleotide sequence of 1 bp deletion in cas9 mRNA and the resultant amino acid sequence. (H) Mutant cas9 mRNA failed to generate Cas9 protein. (I) Phenotypic analysis of wild-type and -1 bp mutant cas9 mRNA injected embryos. (J) Dot blot analysis with J2 antibody of three IVT mRNAs with or without cellulose chromatography. (K) Quantification of dot blot from three biological replicates. (L) Phenotypic analysis after injecting 200 pg/embryo IVT mRNAs. (M) Effects of cas9 mRNAs (200 pg/embryo) on survival of zebrafish embryos. Values are means ± SDs; **** in M, P< 0.0001, Log-rank (Mantel-Cox) test. For others, **P< 0.01, ***P< 0.001, Student’s t-test; scale bars: 200 μm.

Figure 2.

Differential dose of dsRNA by-products in 29 IVT RNAs. (A) Acridine-orange staining of 29 randomly selected IVT RNA preparations revealed marked variability in dsRNA content. (B) Dot-blot analysis with the dsRNA-specific monoclonal antibody J2 corroborated these differences. (C) Quantification of the AO and J2 signals showed a strong positive correlation. Relative dsRNA levels were calculated as dsRNA dose in a certain IVT mRNA divided by the mean dsRNA dose of all IVT mRNAs. (D) Dot plot showing a significant negative correlation between dsRNA levels and normal ratio of IVT mRNA-injected embryos (50 pg/embryo). Relative dsRNA levels in panel (D) were calculated as the mean of relative dsRNA levels measured by AO staining and J2 antibody dot blot for each IVT mRNA. Dashed lines in panels (C) and (D) indicate the 95% confidence interval, while the solid lines designate the fitted linear regression line. r represent Pearson correlation coefficient. (E) Microinjection of 200 pg IVT mRNA per embryo induced widespread toxicity. (F) Reducing the dose to 50 pg/embryo markedly attenuated the toxic phenotype. Numbers on top of each column designate embryos analyzed.

Figure 4.

The toxic effects of IVT mRNA are dependent on the Prkra-mediated dsRNA sensing and global translation repression. (A) Representative phenotypes of embryos treated with translation inhibitors, CHX and puromycin. (B) Quantitative phenotypic analysis of embryos injected with dsRNA (30 pg/embryo) or treated with CHX (5 μg/ml) or puromycin (8 μg/ml). The numbers above each stacked column represent the total number of embryos analyzed. (C) dsRNA injection and treatment with CHX and puromycin cause cell necrosis in 10 hpf embryos, as revealed by PI staining. (D) Quantification of fluorescence intensity of PI in individual embryo. (E) Puromycin incorporation assay showing translation inhibitory effects of dsRNA in wild-type and MZprkra embryos. (F) Quantification of incorporated puromycin levels. (G) Ribosome profile analysis via sucrose gradient ultracentrifugation. Notably, dsRNA treatment reduces the polysome fraction in wild-type embryos but not in MZprkra embryos. (H) Phenotypic analysis of dsRNA-injected wild-type and MZprkra embryos. (I) Comparison of cell necrosis between wild-type and MZprkra embryos following dsRNA injection. (J) Co-injecting 200 pg/embryo of individual IVT mRNAs with 100 pg/embryo of gfp reporter mRNA reveals varying levels of GFP translation. Values shown in columns are means ± SDs. ^nsP > 0.05, *P< 0.05, ***P < 0.001, ****P < 0.0001, Student’s t-test; scale bars: 1 mm for panels (A) and (J); 200 μm for panel (C).

Figure 5.

dsRNA by-products cause experimental artifacts and reduce the efficiency of genetic tools. (A) Typical ISH result of goosecoid in Nodal-overexpressed embryos at 6 hpf. (B) Phenotypic analysis shows that dsRNA-positive nanog IVT mRNA (200 pg/embryo for panels A–D) has an inhibiting effect of Nodal activation. (C) Phenotypes of CKO nanog maternal mutant, with or without injecting nanog IVT mRNAs. (D) dsRNA-removed nanog mRNA, but not the unpurified counterpart, can rescue the defective phenotype of its maternal mutant. (E) Representative phenotypes of 36-hpf slc45a2 CRISPants. (F) Phenotypic analysis of slc45a2 CRISPants (200 pg/embryo cas9 mRNA and 100 pg/embryo slc45a2 sgRNA for panels E–J). (G) Structure and sgRNA target site of slc45a2 locus. (H and I) Sequencing chromatograph after editing by coinjecting slc45a2 sgRNA and unpurified (H) or dsRNA-removed (I) cas9 IVT mRNA. (J) Quantification of editing efficiency by Tide algorithm. (K) Representative fluorescence patterns of 3 dpf transgenic F0 larva (200 pg/embryo Tol2 TP mRNA plus 10 pg/embryo plasmid for panels K–O). (L) Quantification of fluorescence pattern after injecting unpurified and dsRNA-removed Tol2 TP mRNA. (M) The experimental design of transposon excision assay. (N) dsRNA removal significantly increased excision efficiency. (O) The percentage of GFP-positive offspring significantly increases after injecting dsRNA-removed Tol2 TP mRNA. Values shown in columns are means ± SDs. *P< 0.05, **P < 0.01, ***P < 0.001, Student’s t-test; scale bars: 200 μm for panels (A) and (D); 500 μm for upper row in panel (E), 100 μm for the lower row; 1 mm for panels (C) and (K). *P< 0.05, **P< 0.01, ***P< 0.001, Student’s t-test.

Figure 6.

m1Ψ modification abolishes the dsRNA stress response and enhances the efficiency of mRNA-based genetic tools. (A) Puromycin incorporation assay demonstrating the loss of translation inhibition by m1Ψ-modified dsRNA in both wild-type and MZprkra backgrounds (30 pg/embryo dsRNA for panels A–D). (B) Quantification of puromycin incorporation into newly synthesized proteins. (C) Phenotypic analysis of wild-type and MZprkra embryos injected with m1Ψ or UTP dsRNA. (D) PI staining analysis showing significantly reduced cell necrosis induced by m1Ψ-modified dsRNAs. (E) Comparison of the translation-repressive effects of UTP and m1Ψ-modified cas9, Tol2 TP, and cre mRNAs injected at 200 pg/embryo. (F) Quantification of puromycin incorporation from panel (E). (G) Phenotypic analysis of embryos injected with m1Ψ and unmodified IVT mRNAs. (H) Significantly reduced cell necrosis in embryos injected with m1Ψ-modified IVT mRNAs. (I) Elevated genome editing efficiency observed when injecting m1Ψ-modified cas9 mRNA together with sgRNA targeting slc45a2 (200 pg/embryo cas9 mRNA and 100 pg/embryo slc45a2 sgRNA). (J) Improved excision efficiency associated with the use of m1Ψ-modified Tol2 TP mRNA (200 pg/embryo Tol2 TP mRNA plus 10 pg/embryo plasmid). (K) Increased transgenic GFP expression following the injection of m1Ψ Tol2 TP mRNA. (L) Experimental design for detecting recombination events using Cre. (M) PCR products from detection primers appear only when cre mRNA is co-injected. (N) Quantification of the recombination rate by quantitative PCR (qPCR) (200 pg/embryo cre mRNA plus 10 pg/embryo plasmid). Values are expressed as means ± SD. ^nsP > 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, Student’s t-test.

Figure 7.

Reduced affinity of m1Ψ dsRNAs for Prkra dimer and enhanced translation of m1Ψ IVT mRNAs in early zebrafish embryos. (A) EMSA analysis of the binding patterns of UTP dsRNA and m1Ψ dsRNA to recombinant Prkra protein. (B) Binding curves showing the interaction of Prkra with UTP dsRNA and m1Ψ dsRNA. The vertical dashed lines indicate the dissociation constants for UTP and m1Ψ dsRNAs with the Prkra protein. (C) RNA pulldown assay demonstrates a significantly weakened interaction between m1Ψ dsRNA and eIF2α, both in the presence and absence of exogenous Prkra protein. (D) Western blot analysis of Cas9 produced from UTP-containing versus m1Ψ-substituted cas9 mRNAs (200 pg/embryo for panels D–H); densitometric quantification is shown on the right. (E–H) Translation efficiencies of unmodified and m1Ψ-modified mRNAs encoding GFP, and myc-tagged p53, Il1b, and Pax2a. For each transcript, the left sub-panel presents a representative blot, and the right sub-panel displays the corresponding quantitative analysis. (I) Dose-dependent effects of UTP cas9 mRNA and m1Ψ cas9 mRNA on the translation of gfp reporter mRNA and their own translation. Values are expressed as means ± SD; **P < 0.01, ***P < 0.001, Student’s t-test.

For generating standard pCS2-derived mRNAs in Figs 1, 2, 4, 5 and Supplementary Figs S1 and S5, coding sequences of the genes used in this study were either PCR-amplified or directly synthesized, then subcloned into the pCS2 vector using a modified Gibson Ligation method [27]. Recombinant vectors were linearized by Not I digestion, followed by purification with a PCR cleanup kit. The IVT reactions were performed with either the mMESSAGE mMACHINE SP6 or T7 kit (Cat# AM1344 or Cat# AM1340), according to the manufacturer’s instructions. Synthesized mRNAs were extracted with phenol–chloroform–isopentanol (25:24:1, v/v) followed by extraction with chloroform–isopentanol (24:1, v/v). The RNAs in the aqueous phase were then precipitated in 50% ice-cold isopropanol. After centrifugation at 12 000 × g at 4°C for 15 min, the RNA pellet was rinsed with 70% cold ethanol and resuspended in 15–20 μl of RNase-free water. mRNA quality (A₂₆₀/A₂₈₀ ratio) was checked with microvolume UV–Vis spectrophotometer (ALLSHENG, Nano-300). High-quality mRNAs with A260/A280 ratio between 2.0 and 2.2 were used for further experiments. A step-by-step protocol is provided in Supplementary Data.

The mRNA templates for Figs 6 and 7 were PCR-amplified from the corresponding pCS2 plasmids using the primers pCS2-f (5′-ACGTCGGAGCAAGCTTGATT-3′) and pCS2-r (5′-AATGTATCTTATCATGTCTG-3′). The m1Ψ-modified mRNAs were synthesized similarly as the UTP mRNAs except that rUTP was replaced by m1Ψ (TriLink, N-1081) in the rNTP mix (10 mM rATP, 10 mM m1Ψ, 10 mM rCTP, 2 mM rGTP, and 8 mM cap analog). Then the purified mRNAs were subjected to poly(A) tailing reaction following the procedure of the Poly(A) Tailing Kit (ThermoFisher, AM1350). A step-by-step protocol is provided in Supplementary Data.

To generate dsRNAs, DNA templates were PCR-amplified from plasmids containing zebrafish codon-optimized cas9 sequence [23] with T7 and SP6 promoter sequences incorporated into the 5′ or 3′ ends via primers (see Supplementary Table S2). Sense and antisense RNAs were synthesized separately and then annealed at equal amount. The program for dsRNA annealing is as follows: 95°C for 5 min, ramping down at 0.1°C/s to 50°C, and then quickly cooling to 4°C.

dsRNA removal by cellulose chromatography

Fifty micrograms of mRNAs were incubated with 50 mg of cellulose (Sigma–Aldrich, C6288) for 30 min in a buffer containing 10 mM HEPES (pH 7.2), 0.1 mM EDTA, 125 mM NaCl, and 16% (v/v) ethanol. The mixture was then centrifuged at 12 000 × g at 4°C, and the supernatant was collected. To precipitate RNAs from the supernatant, 1/10 volume of 4 M sodium acetate (pH 4.0) and 2.5 volume of absolute ethanol were added. After thorough vortexing, the mixture was incubated in a −20°C refrigerator for 30 min, followed by centrifugation at 12 000 × g for 15 min at 4°C. The resulting precipitate was rinsed with 70% ethanol and dissolved in RNase-free water. A step-by-step protocol is provided in Supplementary Data.

dsRNA detection

dsRNAs were detected following three strategies: dot blot by J2 antibody, acridine orange (AO) staining, and RT-PCR.

For dot blot analysis using a dsRNA-specific J2 antibody (Scicons, 10010500), equal amount of RNAs (200 ng) were spotted onto the Nylon membrane (Amersham, S952) using a 10 μl of pipet tip and allowed to dry at room temperature. The membrane was then incubated with J2 antibody solution (1:2000, diluted in tris-buffered saline with tween® 20 [TBST]) at 4°C overnight, followed by washing with TBST three times every 10 min. The J2 antibody bound to the membrane was further detected by anti-mouse IgG-HRP diluted at 1:5000 in TBST. After rinsing the membrane every 10 min with TBST for three times, the dsRNA was visualized using ECL reagent (Meilunbio, MA0186) and imaged using a chemiluminescence imaging system (Sagecreation).

For AO staining, IVT RNAs (400 ng) were separated on a 1.2% agarose gel prepared in 1× TBE and run at 120 V for 30 min. The gel was then incubated, protected from light, in AO staining solution (10 μg/ml AO in 50 mM Tris–HCl, pH 6.5) for 15 min with gentle agitation. Excess dye was removed with a 5-min wash in deionized water. Fluorescence images were acquired on an iBright 1500 imaging system (Thermo Fisher Scientific) using the following excitation/emission settings: for single-stranded RNA (ssRNA) detection 460 nm/650 nm; for dsRNA detection 502 nm/525 nm [28].

dsRNAs were also detected by amplifying the antisense strand of the cas9 mRNA sequence using RT-PCR. Briefly, sense primers were designed to recognize the antisense RNA strands and were employed for complementary DNA (cDNA) synthesis. The amount of antisense cas9 RNAs was quantified by qRT-PCR. In parallel, an antisense primer was utilized to synthesize cDNA specifically representing cas9 mRNAs, which was subsequently quantified by qRT-PCR. The relative amount of antisense RNAs serves as an indicator of dsRNA levels, calculated as the ratio of antisense RNAs to mRNAs. The primer design and sequences are shown in Supplementary Fig. S1N and listed in Supplementary Table S2.

Bright-field live-imaging

Embryos were dechorionated by pronase at a concentration of 1 mg/ml in Ringer’s solution for 10 min. They were then mounted in 1.5% low-melting agarose on a cover glass-bottomed Petri dish, and their position was adjusted before agarose solidification. After adding 1/3× Ringer’s buffer, samples were loaded onto an inverse microscope (JiangNan, XD-202) for time-lapse recording. To acquire differential interference contrast (DIC) image, dechorionated embryos were mounted in 1.5% low-melting agarose on a concavity slide and photographed under a Leica upright microscope (Leica, DM2500) equipped with DIC apparatus. The videos were generated and further edited by Adobe Premiere Pro CC software.

Western blot

Embryos were dechorionated with their yolk cells manually removed. The resulting embryonic cells were collected in 1.5 ml centrifuge tubes and homogenized in a lysis buffer (100 mM NaCl, 0.5% NP-40, 5 mM EDTA, and 10 mM Tris–HCl pH 7.5 with protease inhibitors). After clearing by centrifugation at 12 000 × g for 10 min, cell lysates were boiled in 1× Laemmli buffer (10% SDS, 0.5% bromophenol blue, 50% glycerol, 5% β-mercaptoethanol, and 250 mM Tris–HCl, pH 6.8). Proteins were then separated by SDS–PAGE and transferred to a nitrocellulose membrane. The primary antibodies used for western blot can be found in the Key Resource Table.

qRT-PCR

Total RNA was extracted at the appropriate developmental stages, and cDNA synthesis was performed using the TransScript First-Strand cDNA Synthesis Kit (Transgene) according to the manufacturer’s instructions. qPCR was conducted with 2× M5 HiPer SYBR premix (MF787-01, Mei5) on a Q2000 real-time PCR system (LongGene). Primers used for gene expression analysis are listed in Supplementary Table S2. Relative expression levels were calculated using the 2^−ΔΔCt method, with the expression of gapdh serving as the loading control.

Microinjection and cell transplantation

The injection needles were prepared using Narishigi needle puller (Puller PC-100). The melt-sealed end of the needle was clipped by tweezers, creating an open tip at a diameter of 5 μm. IVT mRNAs, dsRNAs, or other reagents were dissolved in deionized RNase-free water at appropriate concentrations, and then loaded into the needle from the tip by the “Fill” function of the PLI-100A Picoliter Microinjector and then injected into the yolk of 1-cell stage zebrafish embryos at a volume of 1–2 nl per embryo.

Cell transplantation experiment was performed as described previously [29]. Briefly, wild-type embryos were first labeled with Alexa647-dextran by microinjection at 1-cell stage. At high stage, cells in the blastodisc were aspirated and introduced into the marginal region of stage-matched unlabeled recipients. At 7 hpf, chimeric embryos were mounted in 1% low melting agarose on a cover slip for confocal imaging.

Tol2 transposon excision assay

100 pg of Tol2 TP IVT mRNA and 10 pg of pT2AL200R150G plasmid were co-injected into each 1-cell stage embryo. The injected embryos were raised to the shield stage and then lysed for DNA extraction using a genomic DNA extraction kit (Sangon, B518251). The extracted DNA was subsequently used as a template for nested PCR amplification. The primers used in this assay are listed in Supplementary Table S2. M13R and Tol2 excision assay AS were employed for the first round of nested PCR, while M13F and M13R were used in the second round. The pre-amplification was set for 15 cycles. The PCR product from the pre-amplification was then diluted at 1:10 and used as a template for the second round of qPCR analysis. GAPDH was used as a loading control in the qPCR analysis.

Quantification of genome editing efficiency

The sgRNAs targeting the slc45a2 ORF and the third intron of rbm24a are listed in Supplementary Table S2. DNA templates for the sgRNAs were synthesized by fill-in PCR following the guidelines on the CRISPRScan website. After IVT, purified sgRNAs were injected along with zebrafish codon-optimized cas9 mRNA into the fertilized eggs. Embryos at 10 hpf were homogenized in 50 mM NaOH for 20 min at 95°C, followed by the addition of 1/10 volume of 1 M Tris–HCl (pH 7.5). The DNA fragment flanking the sgRNA site was PCR amplified and sequenced. Sanger sequencing chromatogram files were submitted to http://shinyapps.datacurators.nl/tide/ to estimate the efficiency of the sgRNA.

Phalloidin, propidium iodide, and Hoechst staining

For phalloidin staining, embryos were fixed with freshly prepared 4% paraformaldehyde and washed in PBST for three cycles of 10 min each. The embryos were then incubated with fluorescein-labeled (FITC) phalloidin for 30 min and washed again in PBST for three cycles of 10 min each. They were subsequently analyzed under a confocal microscope (Zeiss LSM880), and Z-stack projections were obtained using ZEN software.

For propidium iodide (PI) and Hoechst staining, live embryos, either uninjected or dsRNA-injected, were raised in E3 buffer containing 1 μg/ml PI and 1 μg/ml Hoechst33342. The embryos were subjected to fluorescence time-lapse imaging under an inverse microscope or mounted in 1.5% low-melting agarose and imaged using a Zess confocal microscope.

Embryos at 8 hpf were dissociated by trypsin–EDTA treatment (0.25% m/v) at 37°C for 15 min with gentle pipetting. The dissociated cells from uninjected or dsRNA-injected embryos were washed and stained with 1 μg/ml PI and 1 μg/ml Hoechst33342 in E3 buffer. They were then subjected to time-lapse analysis using a Leica inverse microscope equipped with the Thunder function.

PI staining and quantification of cell necrosis

The number of PI-stained nuclei and the intensity of PI staining of each embryo can be used as cell necrosis indicators. 1 μg/ml of PI in E3 buffer was used to stain dsRNA-injected embryos until 7–8 hpf. The red fluorescent nuclei of each embryo were counted manually or the integrated optical density of the red fluorescence of each embryo was measured by ImageJ.

Electron microscopy

Embryos were fixed in 2.5% glutaraldehyde buffered in phosphate-buffered saline (PBS) for 1–4 h at 4°C. After washing in PBS five times for 20 min each, the embryos were further fixed in 1% osmium tetroxide. Following rinsing in PBS three times for 15 min each, the embryos were dehydrated through a graded ethanol series and subjected to two treatments with absolute acetone. They were then infiltrated with Epon 812 resin and polymerized using a temperature-rising program (12 h at 36°C, 12 h at 50°C, and 24 h at 60°C). The embedded samples were sectioned to a thickness of 100 nm and stained with 2% uranyl acetate and 3% lead citrate. Ultrathin sections were prepared using an ultramicrotome (PowerTome-XL) equipped with a glass knife and subsequently visualized under a focused ion beam scanning electron microscope (ZEISS, Crossbeam 550).

JB4 sectioning

Embryos were fixed overnight at 4°C in PBS-buffered 4% paraformaldehyde. They were embedded in JB-4 resin following the manufacturer’s guide (Sigma, EM0100-1KT). 2 μm sections were obtained using an ultramicrotome (PowerTome-XL) equipped with a glass knife and stained with hematoxylin and hydrophilic eosin Y.

In situ hybridization

In situ hybridization (ISH) was performed according to previously described protocols [30]. Templates for synthesizing prkra probes were generated through PCR amplification. Digoxigenin-labeled probes were synthesized via IVT using the DIG-labeling mix (Roche) with SP6 RNA polymerase. Following treatment with DNase I to remove template DNA, the labeled probes were directly precipitated with 50% (v/v) isopropanol.

Puromycin incorporation assay

Embryos were exposed to 50 μg/ml puromycin from 2 to 4 hpf. Subsequently, they were collected after the removal of the yolk cell and lysed in protein immunoprecipitation (IP) buffer containing 10 mM Tris–HCl (pH 8), 150 mM NaCl, 1% NP-40, 1 mM EDTA, 10 μg/ml aprotinin, 10 μg/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride (PMSF). Samples were mixed with 1× Laemmli buffer (10% SDS, 0.5% bromophenol blue, 50% glycerol, 5% β-mercaptoethanol, and 250 mM Tris�HCl, pH 6.8), followed by denaturation at at 95°C for 5 min. Puromycin-conjugated proteins were then visualized via western blot analysis using a puromycin-specific antibody (EMD Millipore, Cat# MABE343).

Ribosome profile analysis

Sucrose was dissolved in polysome buffer (50 mM Tris–HCl, pH 7.0, 100 mM NaCl, 5 mM MgCl₂, 100 μg/ml CHX) at concentrations of 10% and 60%. A sucrose density gradient was established using a 41Ti centrifuge tube on a Gradient Station (BIOCOMP) and allowed to reach equilibrium. Embryos at 5 hpf were harvested from both control and dsRNA-injected groups. Chorions were enzymatically digested using pronase, and the yolk cells were manually removed. The yolk-removed embryos were then exposed to 50 μg/ml cycloheximide (CHX) for 10 min before being homogenized on ice in lysis buffer containing 1% Triton X-100, 100 μg/ml CHX, 300 U/ml RNase inhibitor, 10 μg/ml aprotinin, 10 μg/ml leupeptin, and 1 mM PMSF. The lysate was then centrifuged at 12 000 × g for 10 min at 4°C. The resulting supernatant was carefully layered on top of the sucrose density gradient and subjected to ultracentrifugation at 240 000 × g for 2 h using an ultracentrifuge (Beckman, XPN-100). Subsequently, the centrifuge tube was transferred to an automated density fractionation system (BIOCOMP) for fraction separation, and the absorbance of each fraction was measured at 260 nm.

RNA pulldown

The detailed protocol is provided in the Pierce^™ Magnetic RNA-Protein Pulldown Kit manual. Initially, 50 μl of streptavidin magnetic beads were transferred to RNase-free centrifuge tubes. These tubes were then placed on a magnetic separation stand for 30 s to facilitate separation, after which the supernatant was removed. The beads were washed twice with 20 mM Tris–HCl, pH 7.5 (50 μl for each wash). Subsequently, 1× RNA Capture Buffer was added to the beads, which were vortexed and incubated with 50 pmol of biotin-labeled RNAs under gentle rotation at room temperature for 20 min. After discarding the supernatant, the beads were washed twice again with 20 mM Tris–HCl, pH 7.5.

Concurrently, 400 de-yolked embryos were lysed in 400 μl of protein lysis buffer (10 mM Tris–HCl, pH 8, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 10 μg/ml aprotinin, 10 μg/ml leupeptin, and 1 mM PMSF). The resulting lysate was centrifuged at 12 000 × g for 10 min at 4°C, and 130 μl of the cleared supernatant was diluted to a total volume of 400 μl in 1× protein–RNA binding buffer and 15% glycerol. This mixture was then incubated with the beads at 4°C under rotation for 1 h. Following the incubation, the beads were washed twice with 100 μl of 1× washing buffer with gentle mixing. Finally, the beads were incubated with 50 μl of elution buffer at 37°C for 20 min. The eluate was subsequently analyzed by western blot.

Recombinant protein expression and purification

The pET-Duet plasmid was linearized using Kpn I and Nco I restriction enzymes. A 6× His tag sequence was incorporated at the 3′ end of the coding regions for prkra. The Rosetta strain of Escherichiacoli, transformed with the plasmids, was cultured overnight at 37°C in LB medium and then inoculated into 1 L of fresh LB for amplification until an optical density (OD) of 0.6 was reached. To induce protein expression, isopropyl β-D-1-thiogalactopyranoside (IPTG, final concentration of 1 mM) was introduced into the culture and incubated for an additional 26 h at 20°C with agitation. The bacteria were harvested by centrifugation at 2800 × g for 10 min at 4°C, and the pellets were resuspended in 10% glycerol before undergoing a second centrifugation at 1850 × g for 20 min at 4°C. The cell pellets were weighed and subsequently resuspended in AP buffer (50 mM Tris–HCl, pH 7.2, 400 mM NaCl, 40 mM imidazole, 10% glycerol, and 0.02% NP-40). The cells were homogenized using lysozyme solution (1 mM dithiothreitol [DTT], 1 mM PMSF, and 1 mg/ml lysozyme in AP buffer), followed by sonication with an ultrasonic homogenizer (JY92-IIN) set to a cycle of 5 s ON and 10 s OFF for a total of 4 min at a power setting of 20%. After centrifugation at 24 000 g for 1 hour at 4°C, the supernatant was filtered through a 0.45 μm membrane. In parallel, a Ni column was prepared, washed twice with ddH2O, and then once with AP buffer. The cleared supernatant was added to the column and incubated at 4°C for 1 h to facilitate protein binding. The column was drained and washed twice with buffer A (1 mM DTT and 1 mM PMSF in AP buffer), followed by the addition of cold buffer B (2 mM ATP-Na, 1 mM MgAc, 1 mM DTT, and 1 mM PMSF in AP buffer) and incubation at 4°C for 10 min to activate the protein. Following two washes with buffer C (1 mM DTT in AP buffer), proteins were eluted slowly using the elution buffer (50 mM Tris–HCl, pH 7.2, 400 mM NaCl, 400 mM imidazole, 10% glycerol, and 0.02% NP-40). For long-term storage, glycerol was added to the protein solution to achieve a final concentration of 50%, and the samples were stored at −80°C.

Electrophoretic mobility shift assay

An 85 bp dsRNA sequence was derived from the Vpr gene of HIV-1, with the sequence: GCAUGGCUUAGGUCAACAUAUCUAUGAAACUUAUGGGGAUACUUGGUCAGGAGUGGAAGCAUAAUAAGAAUUCUGCAACAACUGC. Two forward ssRNA strands were chemically synthesized by Sangon, either with or without m1Ψ modification. Both synthesized ssRNAs were labeled with fluorescein arbitrary moniker (FAM) at the 5′ end and with biotin at the 3′ end. The reverse ssRNA strands were in vitro transcribed with or without m1Ψ modification. The m1Ψ modified and unmodified forward and reverse ssRNAs were then annealed to produce fluorophore-labeled m1Ψ dsRNAs or UTP dsRNAs. The dsRNAs and Prkra protein were mixed in 1× binding buffer (10× storage: 80% glycerol, 100 mM MgCl₂, 5 M NaCl, 0.5 M EDTA, 1 M DTT, 1 M Tris–HCl, pH 8.0) and incubated at room temperature for 15 min. Subsequently, the protein–RNA mixture was electrophoresed using a 6% native PAGE gel at 80 V for 2 h in 1× running buffer (25 mM Tris and 0.5 M glycine). Finally, gel imaging was performed using a full-spectrum gel imager (Invitrogen, iBright 1500).

Quantification and statistical analysis

All experiments were performed at least in duplicate with similar results. Data plotting and statistical tests were performed using GraphPad Prism 9. Statistical information is described in each figure legend.

Results

dsRNAs in IVT mRNAs induce toxic effect on zebrafish early embryos

Injection of IVT mRNAs into zebrafish embryos frequently causes toxicity, as indicated by massive deaths and malformed embryos. Although early cleavage was not affected (Supplementary Video S1), cas9 mRNA-injected embryos (200 pg/embryo) exhibited general developmental delay at the onset of epiboly movement (Fig. 1A and Supplementary Video S2). More than one-fifth of injected embryos lysed during the gastrula stage, probably due to disrupted cortical F-actin and microtubule cytoskeletons in the cortical layer of the yolk cell (Fig. 1B, B', C, and C'). Injected embryos that survived at 36 hpf exhibited developmental defects, such as tail truncation, split axis, abnormality or absence of eyes, malformed brain structures, and cardiac edema (Fig. 1D). We categorized the phenotypes as follows: normal (embryos exhibiting typical morphology without noticeable defects), mild (embryos displaying a slightly shortened body axis, with or without accompanying pericardial edema), moderate (embryos characterized by a significantly shortened body length, with or without ocular defects), and severe (embryos with complete absence of the trunk and tail, as well as head deformities or a split body axis).

Interestingly, these aberrant phenotypes depended on the dose of cas9 IVT mRNAs but were not related to the Cas9 protein level (Fig. 1E). The Cas9 protein levels in the protein-injected group were 4–80-fold higher than those in the mRNA-injected group (Fig. 1F). However, such high levels of Cas9 protein seldom resulted in malformed embryos (Fig. 1E). Thus, components in the IVT mRNAs per se should cause these toxic effects. Consistently, cas9 mRNA with one base pair (bp) deletion in the ORF did not produce functional Cas9 protein (Fig. 1G and H) but had similar toxicity when 400 pg were injected into the embryo (Fig. 1I). We also compared the toxicity of mRNAs encoding Tol2 TP or Nanog with their frame-shifted counterparts and found that they elicited comparable teratogenic effects on early embryos (Supplementary Fig. S1A–C).

IVT often generates dsRNA by-products via 3′-end extension and antisense strand transcription [28, 31–33]. In addition, the toxic phenotypes resulting from IVT mRNAs are analogous to the malformations induced by dsRNAs [18, 34]. Thus, we speculated that the dsRNA contaminants in mRNA might cause the toxic effect. A noticeable phenomenon is that the toxic effects varied between different batches of IVT mRNAs (Supplementary Fig. S1D). By dot blot analysis using a dsRNA-specific J2 antibody, we found that cas9, Tol2-TP, and nanog mRNAs with substantial toxicity showed 2.6–3.6-fold more dsRNA contamination (Supplementary Fig. S1E and F). Using AO staining, we found marked batch-to-batch variation in the amount of dsRNA by-products and in the resulting toxicity, even when the same RNA (sense strand of 978-nucleotide dsRNA sequence shown in Supplementary Table S3) was transcribed (Supplementary Fig. S1G and H). Moreover, elevating the template concentration during IVT of this RNA further increased dsRNA contamination and exacerbated toxicity (Supplementary Fig. S1I and J). IVT mRNAs with varying sequences exhibited differing levels of toxicity, which appears to be positively correlated with the levels of dsRNAs (Supplementary Fig. S1K–M). Strand-specific RT-PCR (40 amplification cycles) revealed reverse-complementary RNA species in both the highly and the weakly toxic cas9 mRNA preparations, providing additional evidence for dsRNA contaminants in the IVT products (Supplementary Fig. S1N and O). qRT-PCR also revealed that weakly toxic IVT cas9 mRNAs had significantly lower levels of reverse-complementary RNAs (Supplementary Fig. S1P).

Next, we used cellulose chromatography to reduce dsRNA by-products in IVT mRNAs [35], resulting in a 60%–80% decrease in dsRNA levels (Fig. 1J and K, and Supplementary Fig. S1Q). As expected, injection of purified mRNAs (200 pg/embryo) significantly reduced toxic effects (Fig. 1L) and the removal of dsRNAs also led to a well-improved survival rate (Fig. 1M).

To determine how commonly the problem occurs, we constructed an unbiased RNA panel by sub-cloning the coding sequences of 29 zebrafish genes that ranked among the top 35 hits by “relevance” in the NCBI Gene database (Supplementary Fig. S2). All transcripts were synthesized in vitro without a 5′ cap to ensure that any phenotypic effects arose from the RNA itself, particularly its dsRNA by-products, rather than from the encoded proteins. As expected, uncapped gfp mRNA produced no detectable GFP (Supplementary Fig. S3A and B), and uncapped bmp4 mRNA failed to induce the ventralized phenotype observed with capped bmp4 mRNA (Supplementary Fig. S3C and D). dsRNA by-products in these 29 IVT mRNAs were determined with both AO staining and J2 dot-blotting, each revealing template-dependent variability in dsRNA formation (Fig. 2A and B). Quantitative analysis showed a strong positive correlation between the J2 dot-blot and AO signals (Fig. 2C), confirming that either assay reliably measures dsRNA contaminants. The percentage of normal embryos showed a significant negative correlation with the amount of dsRNA present in each IVT preparation (Fig. 2D), suggesting dsRNAs contribute to the aberrant phenotypes. Microinjection of 200 pg/embryo revealed that nearly every IVT RNA produced overall developmental abnormalities (Fig. 2E). In contrast, when the injection dose was lowered to 50 pg/embryo, sufficient for effective gene overexpression in most cases, 75.9% RNAs no longer exhibited pronounced toxicity (normal ratio > 80%, Fig. 2F). Our data show that IVT–mRNA injection invariably elicits a dsRNA-mediated stress response in zebrafish embryos, and that the intensity of this response is modulated by both the specific mRNA construct and the injection dose.

dsRNAs elicit severe cell necrosis and delay the maternal–zygotic transition

The dsRNA-induced toxic phenotypes are possibly associated with cell death. As the toxicity of dsRNA is sequence independent, we next used a previously described 1 kb dsRNA from cas9 ORF in the following analysis [18]. In sections made from dsRNA-injected embryos at 8 hpf, we frequently observed enlarged and fragmented cells, reminiscent of cell necrosis (Fig. 3A and B). Consistent with the frequent membrane rupture, F-actin lining the cell membrane was remarkably reduced in swollen cells after dsRNA injection (Fig. 3C and D). To further confirm the presence of necrosis, we stained dsRNA-injected (50 pg/embryo) live embryos with PI and observed red fluorescent nuclei gradually emerging during development (Fig. 3E and F, Supplementary Video S3), suggesting the loss of membrane integrity, a hallmark for necrosis. However, these gradually emerged PI-stained nuclei never appeared in uninjected controls. In cell transplantation experiments, large swollen and broken cells were also present in gastrula embryos (Fig. 3G and H). The process of cell membrane rupture was also captured in isolated single cells (Fig. 3I and Supplementary Video S4). We also compared the ultrastructure between dsRNA-stimulated and uninjected cells by transmission electron microscopy (TEM) at 8 hpf. Swollen cells caused by the injection of dsRNAs exhibited active mitophagy (Supplementary Fig. S4), suggesting a potential self-rescue response to mitigate the stress caused by dsRNA stimulation.

Figure 3.

dsRNAs induce cell necrosis and delay MZT. (A and B) JB4 sections of uninjected and dsRNA-injected embryos at 8 hpf stained by HE. (C and D) Phalloidin staining to visualize cortical F-actin in uninjected (C) and dsRNA stimulated (D) cells at 8 hpf. (E and F) The emerging dots are PI-stained nuclei in a time-lapse analysis of a dsRNA-injected embryo. (G) Procedure of labeling, transplantation, and imaging. (H) dsRNA-injected cells exhibit swelling and fragmentation in chimera embryos. (I) Time-lapse analysis showing the rupture process of single isolated cells from embryos injected with dsRNA. (J) Quantitative reverse transcriptionq PCR (RT-PCR) analysis of eight maternal genes’ expression in uninjected, IVT mRNA-injected (100 pg/embryo), and dsRNA-injected (30 pg/embryo) embryos. (K) qRT-PCR analysis of the transcript level of eight zygotically expressed genes at 6 hpf. (L) Northern blot analysis demonstrates the accumulation of mature miR-430 in uninjected embryos and in embryos injected with dsRNA (30 pg/embryo). The upper panel presents the miR-430 hybridization signal, whereas the lower panel displays the ethidium-stained total RNA for each lane, confirming equal loading. Values shown in columns are means ± SDs. ^nsP > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, Student’s t-test; scale bars: 50 μm for panels (A) and (B); 20 μm for panels (C) and (D); 100 μm for panel (E), 10 μm for panel (F), 50 μm for panel (H), and 20 μm for panel (I).

Given the significant developmental delay observed following dsRNA injection, we investigated whether dsRNA interfered with MZT, also called zygotic genome activation (ZGA) or mid-blastula transition (MBT), representing a crucial event during early development. In this critical period, a substantial proportion of maternal transcripts are typically degraded, while zygotic transcripts rapidly accumulate from the activated genome. To determine the effects of dsRNAs on these processes, we assessed the expression of eight maternal transcripts, btg4, casd1, cldnd, lipg, ripor3, slc35f2, acad1, and cd82b, which typically undergo rapid degradation following MZT in uninjected embryos [36]. qRT-PCR analysis revealed significantly reduced degradation of these transcripts after dsRNA stimulation, either from annealed dsRNAs or IVT cas9 mRNAs (Fig. 3J). In addition, we examined the expression of eight zygotic genes, including apoeb, cxcr4b, dusp6, fgfr4, grhl3, nnr, vent, and wnt11 [36]. Injection of annealed dsRNAs (30 pg/embryo) inhibited the transcriptional activation of all these genes at 6 hpf, and injection of dsRNA-positive IVT mRNAs (200 pg/embryo) also showed a similar trend, though to a lesser extent (Fig. 3K). We also monitored the appearance of mature miR-430, the key microRNA that drives degradation of maternal transcripts during the MZT. In uninjected embryos, the mature miR-430 species (21–23 nt) became detectable at 4 hpf. In contrast, embryos injected with 30 pg dsRNA per embryo exhibited a marked decrease in both the amount and the accumulation rate of mature miR-430 (Fig. 3L), indicating a substantial delay in miR-430 biogenesis. These findings suggest a notable delay in the MZT following dsRNA stimulation.

Global translation inhibition mediated by Prkra is responsible for the toxic effects of IVT mRNA

Given that dsRNA stimulation can induce global translation inhibition [18], we sought to determine whether defects in protein synthesis are sufficient to elicit nonspecific toxic phenotypes. For this purpose, we treated zebrafish embryos with the translation inhibitors CHX and puromycin, which inhibit translation elongation, for comparison with the effects of dsRNA injection. Our results indicated that the translation repression induced by these two compounds mimicked the effects of dsRNAs on overall phenotypes and necrosis when administered at titrated concentrations (Fig. 4A–D).

In early embryos and embryonic stem cells, the dsRNA sensor involved in regulating global translation efficiency is Prkra, rather than PKR [18]. To investigate whether mRNA injection induces translation blockage in a Prkra-dependent manner, we examined the effects of IVT mRNAs in wild-type and MZprkra embryos. The MZprkra embryos were generated by crossing homozygous mutants carrying two complex mutations (Supplementary Fig. S5A and B). One mutant has -5, -8, and -14 bp deletions at the first three sgRNA target sites, with the initial -5 bp deletion disrupting the exon–intron junction (Supplementary Fig. S5B). The other mutant allele contains a large 235 bp deletion flanked by the second and third sgRNA target sequences, as well as -6 and -9 bp small deletions at the first and fourth sgRNA sites (Supplementary Fig. S5B). Both mutations prevent the production of a functional Prkra protein (Supplementary Fig. S5C) and lead to nonsense-mediated decay (NMD) (Supplementary Fig. S5D). By injecting 100 and 300 pg cas9 IVT mRNAs into wild-type and MZprkra embryos, we found that unlike wild-type embryos, where ∼50%–80% translation inhibition was observed, translation inhibition was significantly alleviated in prkra mutants (Fig. 4E and F). The ribosome profile analysis through ultracentrifugation using 5-hpf embryos revealed that the administration of 100 pg/embryo of cas9 IVT mRNAs reduced the polysome fraction in wild-type embryos, but not in MZprkra embryos (Fig. 4G). This finding aligns with the observation that injecting dsRNA-positive IVT mRNAs into MZprkra embryos produced minimal aberrant phenotypes and necrosis (Fig. 4H and I). These results thus demonstrate that dsRNA–Prkra axis and the resultant translation inhibition is responsible for the toxic effects of IVT mRNAs.

To probe the breadth of the translation-inhibitory effect exerted by IVT RNAs, we co-injected zebrafish embryos with 100 pg of gfp reporter mRNA and 200 pg of each of 29 randomly selected IVT transcripts as shown in Supplementary Fig. S2. 72.4% IVT RNAs produced a marked reduction in GFP fluorescence (Fig. 4J). The extent of inhibition showed a strong positive correlation with the amount of dsRNA by-product associated with each transcript (see Fig. 2A and B).

mRNA overexpression leads to experimental artifacts and compromise the efficiency of mRNA tools

The above results led us to realize that the dsRNA-induced stress response may impact the effectiveness and reliability of mRNA overexpression. Here, we present two examples to substantiate this assertion. The first is a co-injection experiment. We activated Nodal signaling by injecting dsRNA-removed squint-myc (sqt-myc) mRNA, which encodes a myc-tagged Nodal ligand. This resulted in ubiquitous expression of Nodal target gene goosecoid (gsc). When we co-injected nanog mRNA along with sqt-myc, the ectopic induction of gsc was significantly inhibited (Fig. 5A and B). This observation seemingly indicates a suppressive effect of Nanog protein on Nodal signaling. However, after removal of dsRNAs by cellulose chromatography, the purified nanog mRNA did not exhibit such an effect (Fig. 5B), indicating that the above phenomenon only represents a stress effect from dsRNAs. In fact, the dsRNAs in nanog IVT mRNAs significantly reduced global translation efficiency and thus the production of Squint-myc protein, while the cellulose purified nanog mRNA did not (Supplementary Fig. S6A).

The adverse effect of dsRNA contaminants was also shown in a rescue experiment. We generated nanog maternal mutant (Mnanog) embryos via an oocyte-specific conditional knockout approach [25]. One founder female fish produced 65% Mnanog mutants, which displayed severe epiboly defect and a dorsalized-like phenotype (Fig. 5C). The dsRNA-removed nanog mRNA can efficiently rescue these mutant phenotypes. Instead, injecting the dsRNA-positive nanog mRNA failed to rescue but exaggerated the lethal phenotypes (Fig. 5C and D). These results thus demonstrate a considerable risk in using dsRNA-positive IVT mRNAs for gene functional studies in zebrafish embryos.

We also suspected that this stress response likely influences the effectiveness of mRNA tools. To test this, we first compared genome editing efficiencies by injecting the original and dsRNA-removed cas9 mRNAs. By targeting slc45a2, which encodes a transporter protein that mediates melanin synthesis, we observe a remarkable increase in embryos with pigment loss phenotype when purified cas9 mRNA was used (Fig. 5E and F). Consistently, we observed a 1.5-fold higher mutation rate after injecting dsRNA-removed cas9 mRNA than the unpurified control (Fig. 5G–J). Similar improvement was observed when targeting the last intron of the rbm24a gene [37] (Supplementary Fig. S6B–E). In accordant with the increased efficiency, purified cas9 mRNAs produced three folds more Cas9 protein than the unpurified control (Supplementary Fig. S6F and G).

To test the effect of dsRNA by-products on the efficiency of transgenesis, we performed a proof-of-principle test using the pT2AL200R150G plasmid containing a ubiquitously expressed GFP reporter [38]. When this vector was coinjected with Tol2 transposase (Tol2 TP) mRNAs, we observed an remarkably improvement of transgenic GFP expression at 3 dpf after coinjecting dsRNA-removed Tol2 TP mRNA (Fig. 5K and L). Consistently, by transient embryonic excision assay [38], we detected a 4-fold more increase in excision events after injecting dsRNA-purified Tol2 TP mRNA (Fig. 5M and N). Also, the proportion of GFP-positive F1 offspring increased by almost one fold when using dsRNA-removed Tol2 TP mRNA (Fig. 5O).

m1Ψ modification mitigates the toxic side-effects of IVT mRNAs

The removal of dsRNAs present a viable approach to mitigate the toxicity associated with IVT mRNAs; however, the efficacy and reproducibility of the cellulose chromatography method for dsRNA removal are often constrained. Moreover, the yield of mRNAs is frequently below 50% following the complex and cumbersome cellulose purification process (see “Materials and methods” section), rendering this method less than optimal. Injecting unpurified IVT mRNAs into MZprkra zebrafish can effectively mitigate or prevent toxicity; however, mutations in prkra may adversely impact zebrafish health and introduce potential risks when interpreting experimental results derived from this genetic background. Consequently, we sought to address the issue of gene overexpression by developing other strategies, with mRNA modification emerging as a critical option. Given that m1Ψ modification has been demonstrated to inhibit the activation of innate immune responses in differentiated cells [17], we hypothesized that this modification could similarly circumvent stress responses in undifferentiated cells and early embryos.

Therefore, we first compared the translation-inhibiting effects of m1Ψ-modified dsRNAs with its UTP analog and found that m1Ψ modification significantly diminished translation inhibition in wild-type embryos, as revealed by puromycin incorporation experiment (Fig. 6A and B). As expected, neither UTP nor m1Ψ dsRNA could inhibit global translation in MZprkra background (Fig. 6A and B), which is consistent with the phenotypic and cell necrosis assay that m1Ψ dsRNAs failed to induce evident malformation and cell necrosis in either wild-type or MZprkra background (Fig. 6C and D). We next synthesized five dsRNAs with 0%, 25%, 50%, 75%, or 100% m1Ψ substitution and coinjected them with gfp mRNA into zebrafish embryos at 200 pg/embryo. dsRNA with ≤50% m1Ψ substitution had little effect on relieving translation inhibition, as evidenced by nearly absent GFP fluorescence, and produced toxicity levels comparable to the unmodified control based on the frequency of aberrant phenotypes (Supplementary Fig. S7A and B). In contrast, dsRNAs containing 75% or 100% m1Ψ markedly diminished both translational repression and developmental toxicity (Supplementary Fig. S7A–D). These results suggest that m1Ψ modification may enable evasion from the surveillance of dsRNA by its sensor, Prkra, in the early embryo.

As such, we anticipated that this modification can eliminate the side effects of IVT RNAs. Before examining this, we first analyzed the effects of m1Ψ incorporation on mRNA and dsRNA production. To assess the incorporation efficiency of m1Ψ, we evaluated three representative sequences from tp53, fgf8a, and myod1 transcripts. Our results showed that m1Ψ incorporation led to a modest but significant reduction in RNA yield (Supplementary Fig. S8A). Furthermore, we observed increased electrophoretic mobility of m1Ψ-modified RNA on native gels (Supplementary Fig. S8B), likely due to increased base stacking interactions that promotes a more compact RNA conformation [39]. To evaluate how m1Ψ incorporation influences dsRNA formation, we quantified dsRNA in three IVT RNAs synthesized with either canonical UTP or m1Ψ by AO staining. We did not use a J2-antibody dot blot to quantify the modified dsRNAs, because the incorporation of m1Ψ does not affect AO staining but greatly diminishes the J2 antibody’s detection efficiency (Supplementary Fig. S8C–E). Contrary to previous reports [28], m1Ψ incorporation produced a modest yet reproducible increase in dsRNA contaminants (Supplementary Fig. S8F). We attribute this discrepancy to a possible sequence-specific effects of the DNA templates employed in our study. Although m1Ψ-modified RNAs contain a higher proportion of dsRNA contaminants, they trigger markedly less toxicity than their unmodified UTP counterparts when injecting at 200 pg/embryo (Supplementary Fig. S8G). This observation further rules out the idea that the blunted stress response is simply a secondary effect of altered dsRNA levels; rather, it points to a direct, intrinsic impact of the m1Ψ modification itself.

We next synthesized capped cas9, Tol2 TP, and cre mRNAs using either UTP or m1Ψ, without cellulose purification. When injected at 300 pg/embryo, unmodified mRNAs significantly suppressed protein synthesis to ∼50%, while m1Ψ-modified mRNAs at comparable levels exhibited no such an effect (Fig. 6E and F). Consistently, high doses of UTP mRNAs induced significant toxic effects and a high incidence of lethal malformations. In stark contrast, the introduction of m1Ψ modification resulted in nearly normal development (Fig. 6G). This observation aligns with the phenomenon whereby only UTP mRNAs triggered pronounced cell necrosis, while m1Ψ-modified mRNAs did not exhibit such effects (Fig. 6H).

Hence, it can be expected that m1Ψ modification may help to increase the efficiency of mRNA-based genetic tools. We first assessed the genome editing efficiency by coinjecting UTP or m1Ψ-modified cas9 mRNAs with sgRNA targeting the slc45a2 gene, finding that the efficiency increased by about one fold with m1Ψ-modified cas9 mRNA (Fig. 6I). Similarly, when comparing the excision efficiency of Tol2 transposase translated from UTP and m1Ψ-modified mRNAs, we observed a more than five folds increase with m1Ψ-modified mRNAs (Fig. 6J), which was associated with an improved transgenic GFP expression in 3 dpf embryos (Fig. 6K). The cre mRNA represents another frequently used tool for removing loxP-flanked sequences. By co-injecting a loxP-containing plasmid with cre mRNA, with or without m1Ψ modification, we found a >4-fold increase in cutting efficiency by Cre protein translated from the modified cre mRNAs (Fig. 6L–N). These results indicate that the use of m1Ψ-modified IVT mRNAs can effectively mitigate the toxic side effects, while also achieving a much more robust activity in the application of genetic tools.

m1Ψ-modified dsRNA exhibits reduced binding affinity to the dsRNA sensor Prkra to mitigate translation inhibition

The above findings indicate that m1Ψ modification can effectively evade translation repression mediated by Prkra. To elucidate the underlying molecular mechanism, we hypothesized that m1Ψ modification might influence the binding affinity of the Prkra dimer to dsRNA. To investigate this, we synthesized fluorophore-labeled 85 bp AU-rich dsRNAs derived from sequences of the HIV genome (see Supplementary Table S3), either with or without m1Ψ modification. We then conducted electrophoretic mobility shift assays (EMSAs) to evaluate their binding characteristics to the Prkra dimer. Remarkably, the m1Ψ modification significantly weakened the binding affinity with the Prkra dimer (Fig. 7A). Additionally, the typical laddering pattern resulted from the interaction of Prkra dimer with dsRNA was disrupted at high molecular weight level (Fig. 7A), indicating that the highly ordered binding of the Prkra dimer is compromised following m1Ψ modification of dsRNA. Analysis of the binding curves revealed that the dissociation constant (K_d) for the m1Ψ-modified dsRNA was reduced to ∼1/6 of that for unmodified UTP dsRNA (Fig. 7B). However, the Hill coefficient (h) and maximum specific binding (B_max) remained largely unchanged post-modification (1.005 versus 1.023), with the h value ∼2 for both UTP dsRNA and m1Ψ-modified mRNA (2.125 versus 2.087) (Fig. 7B). This suggests that binding of the Prkra dimer promotes further dimer binding to dsRNA, and this feature was not affected by m1Ψ modification.

We subsequently assessed the efficiency of dsRNA in sequestering the eIF2 complex through a dsRNA pull-down assay [18]. As anticipated, m1Ψ modification significantly diminished the sequestration of eIF2α by dsRNA (Fig. 7C). The addition of recombinant Prkra protein to the pull-down assay significantly enhanced the capture of eIF2α by dsRNA [18]. Notably, the amount of eIF2α pulled down by m1Ψ-modified dsRNA is significantly reduced compared to that captured by the unmodified control in the presence of recombinant Prkra protein (Fig. 7C). These results suggest that the modified dsRNA by-products may be not able to affect translation efficiency. Thus, the m1Ψ-modified IVT mRNAs likely display improved translation rates. Indeed, when injecting at 500 pg/embryo, the Cas9 protein translated from m1Ψ-modified mRNA was three times higher than that from the unmodified control, as revealed by western blot analysis (Fig. 7D). Consistent findings were obtained with IVT mRNAs encoding GFP, p53, Il1b, and Pax2a: in every case, the m1Ψ-modified transcripts exhibited significantly higher translation efficiency than their unmodified controls (Fig. 7E–H). To further evaluate the effect of mRNA dosage on the production of the protein of interest, we injected unmodified and m1Ψ-modified cas9 IVT mRNAs at increasing doses, alongside with gfp mRNA at a constant dose of 30 pg/embryo. We observed that as the dose of UTP cas9 mRNA increased, the amount of Cas9 protein initially rose but then declined when cas9 mRNA was injected at extremely high doses (Fig. 7I, left panel). In line with this trend, the translation efficiency of the co-injected gfp mRNA began to decrease to 64% when UTP cas9 mRNA was injected at 320 pg/embryo, further dropping dramatically to 20% at a dose of 640 pg/embryo (Fig. 7I, left panel). Remarkably, for m1Ψ-modified cas9 mRNA, the level of translated Cas9 protein consistently increased with the amount of mRNA injected, even at extremely high doses. Simultaneously, the translation rate of co-injected gfp mRNA remained unaffected (Fig. 7I, right panel). These findings illustrate that m1Ψ-modified dsRNA by-products generated during IVT exhibit significantly reduced binding affinity to Prkra, leading to decreased activation of Prkra-mediated sequestration of translation initiation factors. We can thus conclude that m1Ψ-modified mRNAs are more suitable for applications in mRNA-based gene overexpression in zebrafish early embryos.

Discussion

Our findings underscore the critical impact of dsRNA by-products generated during the IVT of mRNA, which induce global translation inhibition and misleading phenotypes in early zebrafish embryos. This study builds on our recent work, wherein we identified the dsRNA-binding protein Prkra as a mediator of translation repression, bypassing the canonical PKR–eIF2α pathway [18]. Here, we deliver a thorough analysis of the detrimental effects of dsRNA by-products on gene overexpression studies and the application of genetic tools such as genome editing and transgenesis in the zebrafish model system, highlighting that dsRNA by-products induce cell necrosis and MZT delay (see the Graphical Abstract). Furthermore, we provided compelling evidence that m1Ψ modification of dsRNAs significantly reduces the binding and activation of Prkra. This discovery offers a robust solution to mitigate the adverse effects of dsRNA by-products, addressing a long-standing technical challenge in the early embryo.

In differentiated cells, achieving a complete absence of immunogenicity for mRNA requires both the modification of the RNA and the removal of dsRNA by-products [40]. Modification of uridine is effective in reducing the affinity of dsRNAs to their sensors, such as RIG-I, PKR, and TLR3, as well as diminishing the affinity of ssRNA to TLR7 and TLR8 [12–16, 41–44]. This reduction in affinity leads to decreased activation of downstream IFNs and proinflammatory cytokines. However, m1Ψ-modified dsRNA can still partially activate the IFN response via RIG-I, mimicking the action of RIG-I agonist 5′ppp-hpRNA [40]. Therefore, the elimination of any dsRNA contaminants is still crucial for the effective application of mRNAs in differentiated cells.

Several strategies have emerged to minimize or eliminate dsRNA by-products in IVT mRNAs. One promising approach involves the engineering of T7 RNA polymerase to reduce dsRNA formation while maintaining high yields and purity of mRNAs. The double mutant T7 RNA polymerase (G47A + 884G) developed by Moderna exemplifies this strategy [3]. Indeed, a commercially available T7 variant (Yeasen 10629ES10) can significantly reduce the dsRNA by-product dose and thus alleviated the toxic effects (Supplementary Fig. S9). Additionally, the covalent immobilization of T7 RNA polymerase and templates onto magnetic beads during IVT reactions has shown to effectively decrease dsRNA by-product levels [45]. Further optimization of the IVT reaction conditions can also play a critical role. For instance, limiting UTP concentrations has been found to lower dsRNA levels, while restricting ATP concentrations paradoxically increases dsRNA formation [46]. Moreover, the presence of loose mRNA secondary structures and an excess of unpaired uridine bases can lead to increased dsRNA production during IVT [33]. This indicates that rNTP concentrations and sequence design are vital considerations for minimizing dsRNA contamination. This aligns with our observations that dsRNA production is influenced by the sequences of the templates and may also be related to the concentration of the template (Supplementary Fig. S1I and J). After the IVT process, several purification methods, such as high-performance liquid chromatography (HPLC) and cellulose chromatography, can be employed to remove dsRNA by-products [4, 35]. Treatment with RNase III to cut dsRNA into small fragments has also showed an effect to reduce the immune-stimulatory response of mRNAs [47].

In contrast to differentiated cells, it appears that m1Ψ modification alone is sufficient to circumvent the stress response in early zebrafish embryos, thus negating the need for further dsRNA removal. This phenomenon may be attributed to the low expression levels of RIG-I in early embryos, rendering it largely ineffective. This is consistent with the observation that dsRNA has a very low capacity to activate IFN gene expression in undifferentiated cells, a characteristic highly conserved across species, including zebrafish and mouse early embryos, mouse embryonic stem cells (mESCs), mouse teratocarcinoma cell lines, as well as human embryonic stem cells (hESCs) and induced pluripotent stem cells (hiPSCs) [18, 48, 49]. In contrast to RIG-I, the stem cell-specific dsRNA sensor Prkra does not recognize the triphosphate structure at the 5′ end of dsRNAs; instead, it is only able to recognize the conformation of dsRNA helices through its dsRBDs [18], thereby making m1Ψ modification of dsRNA sufficient to evade its detection.

Indeed, our findings demonstrate a significant reduction in the binding affinity between m1Ψ-modified dsRNA and Prkra, decreasing to approximately one-sixth that of UTP dsRNA. Additionally, the m1Ψ modification disrupts the laddering pattern observed in EMSA assays, indicating an alteration in the ordered arrangement of Prkra dimers on modified dsRNAs. As a result, m1Ψ-modified dsRNA fails to effectively capture the eIF2 complex in the presence of Prkra, underscoring its diminished interaction with key translational components. Consequently, the injection of m1Ψ-modified IVT mRNA allows for the administration of high doses without eliciting stress responses, even in contexts where dsRNA contamination is a concern. It was recently reported that the presence of m1Ψ may introduce a risk of translational frameshifting, resulting in truncated or misfolded protein products [50]. Nevertheless, given the relatively low proportion of erroneous translations, the impact on the overall efficacy of gene overexpression may be minimal. Actually, our western blot analysis did not reveal the presence of any low molecular weight bands associated with translational frameshifting in m1Ψ-modified cas9 mRNAs. This demonstrates an overall fidelity of translation in this case.

In summary, dsRNA by-products in IVT-synthesized mRNAs induce toxicity in early zebrafish embryos by activating Prkra dimer-mediated translation repression. The m1Ψ modification disrupts the recognition of dsRNA by the Prkra dimer, thereby alleviating the stress response. We believe a dual strategy, m1Ψ substitution combined with a dsRNA-minimizing RNA polymerase, as the most reliable route to high-quality IVT mRNA for gain-of-function studies. This consideration is especially critical when several mRNAs are co-injected: the cumulative dose often reaches several hundred pg/embryo, a range in which residual dsRNA may not be neglected, and elicit marked stress response. Given that the function of Prkra is conserved across species, it is conceivable that m1Ψ-modified mRNAs in mammalian stem cells can similarly evade this distinct innate immune surveillance, potentially facilitating more specific and robust gene overexpression.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

China patent 202310609693.3 has been authorized to M.S., T.L., and A.J.C. for the application of zebrafish embryos for dsRNA by-product detection.

Funding

This work was supported by National Natural Science Foundation of China 32370860, 32450630, 32170816, and 31871451 to M.S., Program of Outstanding Middle-aged and Young Scholars of Shandong University to M.S., the Intramural Joint Program Fund of the State Key Laboratory of Microbial Technology (SKLMTIJP-2025-01). Funding to pay the Open Access publication charges for this article was provided by the National Natural Science Foundation of China.

Data availability

Any information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author, Ming Shao (shaoming@sdu.edu.cn). The image and numerical source data are available at https://data.mendeley.com/datasets/yhyk89n944/2.