Uridine analogs prevent stress granule formation, not by blocking PKR recognition, but by inhibiting the synthesis of T7 RNA polymerase byproducts

Abstract

mRNA-based therapeutics are commonly produced through T7 RNA polymerase–mediated in vitro transcription. Introducing these exogenous RNAs into human cells activates an RNA sensor protein kinase R (PKR), which suppresses translation initiation and reduces their therapeutic effectiveness. Incorporating uridine analogs into these transcripts prevents PKR activation and translation shutdown, but the underlying mechanism remains unclear. Here, we demonstrate that treating T7 RNA polymerase–produced transcripts with RNase III, which selectively degrades double-stranded RNA (dsRNA), blocks PKR activation and downstream translation-inhibition events, including eIF2α phosphorylation and stress granule formation in human cells. Interestingly, dsRNAs generated with uridine analogs robustly induce eIF2α phosphorylation and stress granules to the same extent as dsRNA containing uridine. These findings indicate that uridine analogs do not prevent PKR from detecting dsRNA. Instead, we show that uridine analogs decrease the production of T7 RNA polymerase byproducts, including antisense RNA and dsRNA, which activate PKR and downstream stress responses. Finally, we demonstrate that higher amounts of exogenous RNA, lacking T7 RNA polymerase byproducts, can induce stress granules independently of PKR and phospho-eIF2α, but dependent on stress granule scaffold proteins G3BP1 and G3BP2. Together, our findings show that uridine analogs mitigate PKR signaling not by blocking mRNA–PKR interactions but by minimizing dsRNA byproducts from T7 polymerase transcription. Furthermore, stress granule formation in response to high levels of exogenous RNA can occur through a mechanism that does not depend on PKR but relies on G3BP1 and G3BP2. These insights clarify the role of uridine analogs in PKR activation and may inform future therapeutic RNA design.

INTRODUCTION

mRNA-based therapeutics have emerged as a key approach for treating diseases, including mRNA vaccines, where mRNAs encoding peptides of interest are synthesized in vitro and introduced into individuals. An inoculating peptide is then translated and primes the adaptive immune system for future infections (Sahin et al. 2014; Steinle et al. 2017; Park et al. 2021; Rohner et al. 2022). However, exogenous mRNAs generated by T7 RNA polymerase activate a cellular innate immune response, including protein kinase R (PKR) (Anderson et al. 2010; Nallagatla et al. 2011). How these exogenous mRNAs activate PKR remains poorly understood.

Upon binding to activators such as double-stranded RNA (dsRNA), PKR dimerizes and phosphorylates translation initiation factor 2α (eIF2α), a cofactor that brings the initiating tRNA to the ribosomes (Dar et al. 2005; Dey et al. 2005; Proud 2005). Phospho-eIF2α acts as a dominant negative repressor of eIF2B, the guanine exchange factor that is pivotal for successive rounds of translation initiation (Clemens 1994; Sonenberg and Hinnebusch 2009). The resulting block in global translation initiation causes widespread disassembly of polysomes from mRNAs and leads to the formation of cytoplasmic RNA–protein biomolecular condensates called stress granules, which concentrate mRNAs in a translationally stalled state (Kedersha et al. 1999, 2000; Panas et al. 2016; Khong et al. 2017; Ivanov et al. 2019; Mateju et al. 2020).

Stress granules can be induced without cellular stress by simply triggering PKR dimerization or by overexpressing either PKR or a phosphomimetic eIF2α mutant (S51D) (Kedersha et al. 1999; Corbet et al. 2022; Zuo et al. 2022), highlighting a direct relationship between PKR, phospho-eIF2α, and stress granules. Ribosome-free mRNAs have been proposed to potentiate stress granule formation by serving as scaffolds for RNA–RNA and RNA–RNA binding protein interactions (Bounedjah et al. 2014; Van Treeck et al. 2018; Guillén-Boixet et al. 2020; Yang et al. 2020; Campos-Melo et al. 2021; Parker et al. 2024). Together, these mechanisms decrease the translation of exogenous mRNAs produced by T7 RNA polymerase (Karikó et al. 2008; Anderson et al. 2010; Mahadevan et al. 2013), limiting the effectiveness of therapeutic mRNAs.

To avoid activation of the innate immune response, therapeutic mRNAs produced by T7 RNA polymerase are synthesized with uridine analogs, such as pseudouridine and N¹-methyl-pseudouridine. It is hypothesized that these uridine analog–containing mRNAs fail to activate PKR, which is thought to recognize uridine in foreign mRNAs (Karikó et al. 2012; Pardi et al. 2017; Morais et al. 2021). Indeed, incorporating uridine analogs into T7 RNA polymerase–generated RNAs reduces PKR activation and increases translational output in cells, indicating that these analogs mask or camouflage the mRNA from PKR detection (Karikó et al. 2008; Anderson et al. 2010; Andries et al. 2015).

T7 RNA polymerase is known to generate unintended antisense and dsRNA byproducts (Cazenave and Uhlenbeck 1994; Triana-Alonso et al. 1995; Arnaud-Barbe et al. 1998; Nacheva and Berzal-Herranz 2003; Gholamalipour et al. 2018), whose synthesis appears to be inhibited by uridine analog incorporation (Mu et al. 2018; Baiersdörfer et al. 2019). Given that dsRNAs, such as those produced during viral infection, are a potent activator of PKR (Balachandran et al. 2000; García et al. 2006; Lemaire et al. 2008; Gal-Ben-Ari et al. 2018; Hur 2019; Chen and Hur 2022), it is possible that uridine analogs prevent the production of PKR-activating dsRNA byproducts instead of camouflaging the mRNA from PKR detection.

In this study, we investigate whether uridine analogs camouflage exogenous RNAs from PKR detection by directly introducing in vitro–synthesized mRNAs into the mammalian cell cytoplasm by microinjection. Our data indicate that dsRNAs containing N¹-methyl-pseudouridine trigger eIF2α phosphorylation and stress granule formation to the same extent as dsRNAs with uridine, indicating that uridine analogs do not block PKR activation. Treating T7 RNA polymerase–generated mRNA with RNase III, which selectively degrades dsRNA, blocks PKR activation, as well as mRNA synthesized with uridine analogs. We demonstrate that both RNase III and uridine analogs block the production of antisense RNA and dsRNA byproducts by T7 RNA polymerase. Thus, the incorporation of uridine analogs indirectly inhibits PKR by preventing the production of dsRNA byproducts by T7 RNA polymerase that, when introduced into cells, activate PKR, trigger eIF2α phosphorylation, and lead to stress granule formation. Lastly, we demonstrate that even when PKR-activating byproducts are eliminated, exogenous RNA can still induce stress granule assembly at higher levels, independent of eIF2α phosphorylation and of the uridine modification state. Together, our results indicate that incorporating uridine analogs into therapeutic mRNAs may not be necessary to prevent PKR activation. However, therapeutic mRNAs can also trigger other innate immune responses, and the role of uridine analogs and dsRNA byproducts of T7 RNA polymerase in these contexts remains to be investigated. Our findings invite further optimization of exogenous RNAs used in mRNA therapies.

RESULTS

mRNAs produced by T7 RNA polymerase activate PKR and eIF2α phosphorylation

To examine the regulation of exogenous mRNAs inside cells, we directly introduced mRNAs produced using T7 RNA polymerase into the cytoplasm of human osteosarcoma (U2OS) cells by microinjection. We used two model mRNAs, fushi tarazu (ftz) and β-globin, which have been used extensively to study mRNA localization and metabolism (Luo and Reed 1999; Valencia et al. 2008; Mahadevan et al. 2013). To prevent degradation, the mRNAs were capped and polyadenylated (we estimate the poly(A) tail to be ∼100 nt long; Supplemental Fig. S1A). The mRNAs were coinjected with either Oregon Green– or FITC-conjugated 70 kDa dextran, which does not cross the nuclear pore and thus marks the microinjected subcellular compartment.

T7 RNA polymerase–generated transcripts activate PKR, whose primary substrate is eIF2α (Dar et al. 2005; Dey et al. 2005; Proud 2005), in cells (Nallagatla and Bevilacqua 2008; Anderson et al. 2010; Heinicke and Bevilacqua 2012; Nallagatla et al. 2013; Mayo et al. 2016; Safran et al. 2019; Kitano et al. 2024) and in animals (Anderson et al. 2010; Nallagatla et al. 2013). In line with this, microinjection of either ftz or β-globin mRNA significantly increased the intensity of phospho-eIF2α staining (Fig. 1A; Supplemental Fig. S1B, quantified in Fig. 1B), when compared to noninjected cells or cells injected with FITC-conjugated 70 kDa dextran (representative images in Supplemental Fig. S1C). To determine whether microinjected mRNA promotes eIF2α phosphorylation in a PKR-dependent manner, we depleted PKR with shRNAs (Fig. 1C). We found that ftz mRNA microinjection did not increase eIF2α phosphorylation in PKR-depleted cells but did in control-depleted cells (Fig. 1D, quantified in Fig. 1E). Together, we find that when T7 RNA polymerase–generated mRNA is microinjected into human cells, it induces eIF2α phosphorylation by PKR, consistent with previous findings (Anderson et al. 2010; Nallagatla et al. 2013).

FIGURE 1.

Microinjected T7 RNA polymerase–synthesized mRNA triggers eIF2α phosphorylation via PKR. (A,B) Wild-type U2OS cells were microinjected with ftz mRNA (300 ng/μL) and 70 kDa dextran conjugated to FITC. The FITC-dextran marks the microinjected cells and the injected compartment (cytoplasm). Cells were fixed 30 min postinjection and stained for phospho-eIF2α by immunofluorescence and DNA by DAPI. Representative images depicting a single field of view imaged for dextran (green)/DAPI (magenta) and phospho-eIF2α are shown in A: arrows, injected cells; asterisks, noninjected cells; and scale bars, 30 µm. Quantification of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view is shown in B. Cells were microinjected with either ftz or β-globin mRNAs or with FITC-dextran alone. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of three independent experiments consisting of at least 30–40 cells. (C) Wild-type U2OS cells were transduced with lentivirus-delivered shRNAs against PKR (two clones were used, PKR #1 and PKR #2) or a scrambled shRNA. After 48 h, cell lysates were collected and separated by SDS-PAGE and immunoprobed for PKR and α-tubulin, which served as a loading control. (D,E) Similar to (A,B), control- or PKR-depleted U2OS cells were microinjected with ftz mRNA (300 ng/μL) and 70 kDa dextran conjugated to FITC. Cells were incubated for 30 min, fixed, and stained for phospho-eIF2α by immunofluorescence and DNA by DAPI. Representative images are shown in D, with each row a single field of view: arrows, injected cells; asterisks, noninjected cells; and scale bars, 30 µm. Quantifications of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view for control and PKR-depleted U2OS cells are shown in E. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of at least three independent experiments consisting of at least 30–50 cells. Student's t-test was performed for (B) and (E): (***) P < 0.001; not significant (ns) P > 0.05.

T7 RNA polymerase–generated mRNAs trigger stress granule assembly

Microinjecting either ftz or β-globin mRNA led to the formation of G3BP1-positive foci in the cytoplasm, which were easily visible 30 min after injection (Fig. 2A), consistent with our previous findings (Mahadevan et al. 2013). Microinjected mRNAs also concentrated in these foci, as reported previously (Mahadevan et al. 2013). G3BP1 foci did not form when Oregon Green–conjugated 70 kDa dextran was injected alone (Fig. 2A), indicating that these structures depend on exogenous RNA. These foci also enriched other protein components of stress granules, such as TIA-1, Caprin-1, PABP, and FMRP (Supplemental Fig. S2A,B), as well as endogenous NORAD lncRNA (Fig. 2B), which was reported to be enriched in stress granules. In contrast, endogenous GAPDH mRNA (Fig. 2B), which does not accumulate in stress granules (Khong et al. 2017; Namkoong et al. 2018; Matheny et al. 2019), did not enrich in these structures. It is likely that many other endogenous mRNAs also localize to these foci, as they were visible by oligo(dT) staining (Supplemental Fig. S2C). Injected mRNA was diffuse throughout the cytoplasm immediately after microinjection (Supplemental Fig. S2D), suggesting that the foci were not caused by the injection of preformed clusters of ftz mRNA.

FIGURE 2.

Microinjected T7 RNA polymerase–synthesized mRNA triggers stress granule assembly. (A) Wild-type U2OS cells were microinjected with either ftz or β-globin mRNA (300 ng/μL) and 70 kDa dextran conjugated to Oregon Green (top, middle), or with dextran alone (bottom). The cells were fixed 30 min postinjection and stained for the injected mRNA by FISH, the stress granule-marker G3BP1 by immunofluorescence, and DNA by DAPI. Each row depicts a single field of view imaged for dextran (green)/DAPI (magenta), G3BP1, and mRNA: scale bar, 30 µm. Enlarged images of outlined regions are shown on the right: scale bar, 2 µm. (B) 30 min after microinjection of ftz mRNA (300 ng/μL), the cells were fixed and stained against ftz mRNA using FISH probes, endogenous NORAD lncRNA or GAPDH mRNA using single-molecule FISH (smFISH) probes, and DNA by DAPI. Each row is a single field of view for dextran (green)/DAPI (magenta), smFISH for GAPDH (top) or NORAD (bottom), and FISH for ftz mRNA: scale bar, 30 µm. Enlarged images of outlined regions are shown in boxed insets, and the merged images are shown on the right (GAPDH/NORAD smFISH in green and ftz FISH in magenta): scale bar, 2 µm. (C) Wild-type U2OS cells were microinjected with ftz mRNA (300 ng/μL), fixed after 30 min, and stained against ftz mRNAs by FISH, the stress granule-marker TIA1 and the P-body marker Dcp1a by immunofluorescence, and DNA by DAPI. The four images on the left are of a single field of view (scale bar, 30 µm), with the boxed areas enlarged on the right (scale bar, 2 µm). Shown in the overlay are ftz FISH (red), TIA-1 (green), and Dcp1 (blue). (D) Wild-type U2OS cells were microinjected with ftz mRNA (150 or 300 ng/μL), fixed 30 min postinjection, stained against poly(A)-RNA by using oligo(dT) probes, and quantified for whole-cell oligo(dT) intensities. Microinjected and noninjected cells within the same field of view were compared. Representative images are shown in Supplemental Figure S2C. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of three independent experiments consisting of at least 30–40 cells. Student's t-test was performed: not significant (ns) P > 0.05.

In addition to these large foci, we observed the appearance of smaller RNA-positive puncta that did not colocalize with stress granule markers, such as TIA-1, but instead with Dcp1a, a marker of processing bodies (P-bodies, Fig. 2C). These smaller puncta tended to be adjacent to TIA-1-positive RNA granules.

We estimate that microinjecting a solution of 150 and 300 ng/μL of ftz mRNA results in a 2%–4% and 4%–8% increase, respectively (see Materials and Methods for calculation), in the total amount of mRNA in U2OS cells, which is estimated to contain 400,000–800,000 mRNAs (Carter et al. 2005). Indeed, when we measure the increase in mRNA poly(A)-staining by oligo(dT) hybridization in cells microinjected with either 150 or 300 ng/μL, we do not observe a significant increase in whole-cell fluorescence (Fig. 2D; representative images in Supplemental Fig. S2C). Thus, microinjecting these exogenous mRNA amounts causes a minimal change in the total intracellular mRNA level.

Together, these data demonstrate that the microinjection of ftz and β-globin mRNA, produced by T7 RNA polymerase, induces the formation of cytosolic RNA granules that resemble stress granules in size and content, although their full composition may differ from stress granules formed in response to other stresses. In addition, some exogenous RNAs were enriched in smaller puncta that colocalized with a P-body marker.

Stress granule assembly in response to T7 RNA polymerase–generated mRNA requires PKR and phospho-eIF2α

Next, we tested whether the formation of stress granules by the microinjection of mRNA is dependent on PKR and eIF2α phosphorylation. We quantified stress granule formation in wild-type mouse embryonic fibroblasts (MEFs) or MEFs whose endogenous eIF2α contains the phosphodeficient mutation (S51A) (Kedersha et al. 1999; Scheuner et al. 2001, 2006). Sodium arsenite treatment, which triggers eIF2α phosphorylation by heme-regulatory inhibitor (HRI) kinase activation (Lu et al. 2001), did not induce eIF2α phosphorylation in eIF2α phosphodeficient cells in comparison to wild-type MEFs (Supplemental Fig. S3A). We observed that the microinjection of ftz mRNA did not induce observable stress granules in eIF2α S51A MEFs but did so in ∼90% of injected wild-type cells (Fig. 3A; quantified in Fig. 3B).

FIGURE 3.

Stress granules induced by T7 RNA polymerase–synthesized mRNA require PKR, eIF2α phosphorylation, and G3BP1/2. (A,B) Wild-type or mutant phosphodeficient eIF2α (S51A) MEF cells were microinjected with ftz mRNA (75 ng/μL) and 70 kDa dextran conjugated to FITC, fixed 30 min postinjection, and stained for ftz mRNA by FISH, the stress granule-marker G3BP1 by immunofluorescence, and DNA by DAPI. Representative images are shown in A, with each row being a single field of view: scale bar, 30 µm. Enlarged images of the outlined regions are shown on the right: scale bar, 2 µm. Quantification of the percentage of injected cells with visible stress granules is shown in B, where each column represents the average and standard error of six independent experiments, with each experiment composed of 30–50 cells. (C,D) Similar to (A,B), control- and PKR-depleted U2OS cells were microinjected with ftz mRNA (75 ng/μL), fixed 30 min postinjection, and stained for ftz mRNA and G3BP1. Representative images are shown in C: scale bar, 30 µm. Enlarged insets are shown on the right: scale bar, 2 µm. Quantification of the percentage of injected cells with visible stress granules is shown in D, where each column represents the average and standard error of three independent experiments, with each experiment composed of 30–50 cells. (E) Wild-type and G3BP1 and 2 double-knockout (ΔΔG3BP1/2) U2OS cells were microinjected with ftz mRNA (75 ng/μL) and 70 kDa dextran conjugated to FITC, fixed, and stained 30 min postinjection for stress granule-protein FMRP. Representative images are shown in Supplemental Figure S3B and quantified for the percentage of injected cells with visible FMRP⁺ granules, where each column represents the average and standard error of three independent experiments, with each experiment composed of 30–50 cells. (F) Similar to (E), wild-type and ΔΔG3BP1/2 U2OS cells were injected with ftz mRNA (300 ng/μL) and 70 kDa dextran conjugated to Oregon Green, fixed, and stained 30 min postinjection, and quantified for whole-cell phospho-eIF2α intensities, where microinjected and noninjected cells within the same field of view were compared. Representative images are shown in Supplemental Figure S3C. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of three independent experiments consisting of at least 30–40 cells. Student's t-test was performed for (B), (D), (E), and (F): (*) P < 0.05; (***) P < 0.001; not significant (ns) P > 0.05.

We found that ftz mRNA microinjection did not induce stress granules in PKR-depleted cells, compared to ∼40% of injected control-depleted cells (Fig. 3C; quantified in Fig. 3D). Together, these results indicate that stress granules induced by microinjected T7 RNA polymerase–generated mRNA require PKR-mediated eIF2α phosphorylation.

Next, we investigated whether microinjected mRNA–induced stress granule formation required G3BP1 and G3BP2. These two proteins serve as essential scaffolds for stress granules that are triggered by various stressors (Matsuki et al. 2013; Kedersha et al. 2016; Guillén-Boixet et al. 2020; Yang et al. 2020). Microinjection of ftz mRNA into ΔΔG3BP1/2 U2OS cells, which were generated previously by CRISPR/Cas9 gene editing (Kedersha et al. 2016), failed to induce stress granules, but did so in wild-type cells, as expected (quantified in Fig. 3E; see representative images in Supplemental Fig. S3B). Thus, we conclude that G3BP1 and/or G3BP2 are required for stress granule formation by microinjected mRNA synthesized by T7 RNA polymerase. Furthermore, mRNA microinjection induced eIF2α phosphorylation in both ΔΔG3BP1/2 and wild-type U2OS cells by immunofluorescence (quantified in Fig. 3F; see representative images in Supplemental Fig. S3C,D), indicating that G3BP1 and G3BP2 act downstream from PKR and eIF2α phosphorylation in promoting stress granule formation.

dsRNA byproducts from T7 RNA polymerase reactions trigger eIF2α phosphorylation and stress granule assembly

T7 RNA polymerase has been shown to generate unintended dsRNA byproducts by using its newly made transcript as a template to extend it into stem-loops (Cazenave and Uhlenbeck 1994; Triana-Alonso et al. 1995; Gholamalipour et al. 2018; Wu et al. 2020) or by using the coding DNA strand as a template to produce antisense transcripts that hybridize with sense transcripts (Konarska and Sharp 1989; Biebricher and Luce 1996; Mu et al. 2018). Since dsRNAs are potent activators of PKR (Kerr et al. 1974; Nanduri et al. 1998), we asked whether microinjected mRNAs triggered PKR and eIF2α phosphorylation due to dsRNA byproducts.

To test this, we first treated ftz mRNA produced with T7 RNA polymerase with the dsRNA-selective endonuclease RNase III (Lamontagne et al. 2001; Mu et al. 2018; Wu et al. 2020) prior to in vitro capping and polyadenylation. Although many RNAs contain double-stranded regions, such as stem-loop structures, they are not expected to be substrates of RNase III as it cleaves dsRNA with perfect complementarity that is at least 22 bp long. ftz mRNA also lacks predicted dsRNA structures of this length. Consistent with this, RNase III treatment did not have a noticeable effect on the migration of ftz mRNA on a gel (Fig. 4A), such as a smear or laddering, indicating that the enzyme was not cleaving within the transcript. Microinjecting RNase III–treated mRNA did not trigger eIF2α phosphorylation, resulting in levels comparable to noninjected cells (Fig. 4B; quantified in Fig. 4C), and did not induce stress granules. In contrast, mock-treated ftz mRNA induced eIF2α phosphorylation and stress granules in ∼50% of cells (Fig. 4D; quantified in Fig. 4E). Together, these results suggest that unintended dsRNA, generated in the T7 RNA polymerase reaction, triggers the PKR-dependent phosphorylation of eIF2α and stress granule assembly.

FIGURE 4.

RNase III treatment of T7 RNA polymerase–synthesized mRNAs inhibits eIF2α phosphorylation and stress granule formation after microinjection. (A) In vitro transcribed ftz RNA was treated with RNase III or buffer alone and then resolved on a 1% agarose gel. The migration of standards is indicated on the left. (B,C) Wild-type U2OS cells were microinjected with ftz mRNA (75 ng/μL), which was either treated or not with RNase III, and 70 kDa dextran conjugated to FITC. Cells were fixed 30 min postinjection and stained for phospho-eIF2α by immunofluorescence and DNA by DAPI. Representative images in B display a single field of view imaged for dextran/DAPI and phospho-eIF2α: arrows, injected cells; asterisks, noninjected cells; and scale bars, 30 µm. For either condition, quantification of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view is shown in C, as “−RNase III” and “+RNase III.” Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of three independent experiments consisting of 40–60 cells. (D,E) Similar to (B), wild-type U2OS cells were microinjected with RNase III–untreated or RNase III–treated ftz mRNA (75 ng/μL) and 70 kDa dextran conjugated to FITC, fixed 30 min postinjection, and stained for ftz mRNA by FISH, the stress granule-marker G3BP1 by immunofluorescence, and DNA by DAPI. Representative images are shown in D: scale bar, 30 µm. Enlarged images of the outlined regions are shown on the right: scale bar, 2 µm. Quantification of the percentage of injected cells with visible stress granules is shown in E, where each column represents the average and standard error of six independent experiments, with each experiment composed of 50–60 cells. Student's t-test was performed for (B) and (D): (***) P < 0.001; not significant (ns) P > 0.05.

Pseudouridine and N¹-methyl-pseudouridine inhibit eIF2α phosphorylation and stress granule assembly

mRNAs produced from T7 RNA polymerase reactions, where uridine triphosphate (UTP) is replaced with pseudouridine triphosphate (ΨTP) or N¹-methyl-pseudouridine triphosphate (m1ΨTP), do not stimulate PKR or interferon production when transfected into cells (Karikó et al. 2008; Anderson et al. 2010; Andries et al. 2015). Consistent with this, we found that in vitro transcribed ftz mRNA with ΨTP or m1ΨTP replacing UTP (Ψ-ftz and m1Ψ-ftz mRNA) did not trigger eIF2α phosphorylation (quantified in Fig. 5A; representative images in Supplemental Fig. S4A) nor stress granules after microinjection (Fig. 5B; quantified in Fig. 5C; also see Supplemental Fig. S4A). In contrast, the same mRNA synthesized with UTP (U-ftz) did so in ∼75% of injected cells. Treating U-ftz mRNAs with RNase III eliminated this response, failing to induce eIF2α phosphorylation (quantified in Fig. 5D; representative images in Supplemental Fig. S4B). RNase III–treated m1Ψ-ftz mRNAs induced comparable levels of eIF2α phosphorylation. Together, these results indicate that uridine analogs in the T7 RNA polymerase reaction inhibit PKR activation and stress granule formation, acting similarly to RNase III treatment.

FIGURE 5.

Uridine analog incorporation inhibits stress granule formation by T7 RNA polymerase–synthesized mRNAs, but not by camouflaging dsRNA from PKR. (A) Wild-type U2OS cells were microinjected with ftz mRNA (75 ng/μL) that was synthesized with uridines, pseudouridine (Ψ), or N¹-methyl-pseudouridine (m1Ψ) by T7 RNA polymerase. The cells were fixed 30 min postinjection and stained for phospho-eIF2α by immunofluorescence and DNA by DAPI. Quantification of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view is shown with each violin plot representing the median (thick dotted line) and quartiles (thin dotted lines) of three independent experiments consisting of at least 30–40 cells. Representative images are shown in Supplemental Figure S4A. (B,C) Similar to (A) except cells were stained for ftz mRNA by FISH and G3BP1 by immunofluorescence. The representative images, with each row depicting a single field of view for dextran/DAPI, G3BP1, and ftz mRNA FISH for either U-ftz, Ψ-ftz, or m1Ψ-ftz, are shown in B: scale bar, 30 µm. Enlarged images of the outlined regions are shown on the right: scale bar, 2 µm. Quantification of the percentage of injected cells with visible stress granules is shown in C, where each column represents the average and standard error of six independent experiments, with each experiment composed of 50–80 cells. (D) Wild-type U2OS cells were microinjected with RNase III–treated U- or m1Ψ-ftz mRNAs, fixed 30 min postinjection, and stained for phospho-eIF2α by immunofluorescence and DNA by DAPI. Whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view were quantified. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of three independent experiments consisting of at least 40–60 cells. Representative images are shown in Supplemental Figure S4B. (E–G) Wild-type U2OS cells were microinjected with dsRNA that contained either uridine or m1ψ on both strands (50 ng/μL), along with 70 kDa dextran conjugated to Oregon Green as the injection marker. The cells were fixed 30 min after injection and stained for phospho-eIF2α and stress granule-marker G3BP1 by immunofluorescence and DNA by DAPI. Representative single field of view images for dextran/DAPI, G3BP1, and phospho-eIF2α are shown in E: arrows, injected cells; asterisks, noninjected cells; scale bar, 30 µm. Enlarged images of the outlined regions are shown on the right: scale bar, 2 µm. Quantifications of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view are shown in F. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of three independent experiments consisting of 170–200 cells. Quantification of the percentage of injected cells with visible G3BP⁺ foci or granules, compared to that of noninjected cells, is shown in G, where each column represents the average and standard error of three independent experiments, with each experiment composed of 170–200 cells. Student's t-test was performed for (A), (C), (D), (F), and (G): (***) P < 0.001; not significant (ns) P > 0.05. (H) Various DNA or RNA samples, which were either pretreated with RNase III or buffer alone, were spotted and cross-linked to a nylon membrane and stained with either J2 antibody, which recognizes dsRNA, or methylene blue, which binds to nucleic acids. (I) Various ftz RNA samples, which were either pretreated with RNase III or buffer alone, were separated on a 1% denaturing agarose gel and transferred to a nylon membrane, which was probed with radiolabeled oligonucleotides that hybridize with ftz mRNA in the sense or antisense direction.

dsRNAs containing N¹-methyl-pseudouridine trigger eIF2α phosphorylation and stress granule assembly

The presence of nucleotide analogs, such as pseudouridine and N¹-methyl-pseudouridine, has been proposed to hide or camouflage exogenous mRNA from PKR detection, thereby preventing eIF2α phosphorylation (Anderson et al. 2010). To test this, we produced double-stranded m1Ψ-ftz RNA (m1Ψ-dsRNA) by synthesizing the sense and reverse complement strands in separate in vitro transcription reactions, treating them with RNase III to remove unintended dsRNA byproducts and then annealing them. Similarly, we generated double-stranded U-ftz RNA (U-dsRNA). As expected, microinjecting RNase III–treated U-dsRNA robustly induced both eIF2α phosphorylation (Fig. 5E; quantified in Fig. 5F) and stress granule formation (Fig. 5E; quantified in Fig. 5G). Strikingly, RNase III–treated m1Ψ-dsRNA triggered both responses to a similar extent as RNase III–treated U-dsRNA (Fig. 5E; quantified in Fig. 5F,G). Microinjecting the RNase III–treated reverse complement strand of ftz RNA alone did not trigger eIF2α phosphorylation (Supplemental Fig. S4C; representative images in Supplemental Fig. S4D) or stress granule formation (Supplemental Fig. S4D; quantified in Supplemental Fig. S4E). Therefore, uridine analogs can be detected by and activate PKR as well as uridine-containing mRNA; uridine analogs do not camouflage mRNA from PKR.

N¹-methyl-pseudouridine prevents the production of dsRNA and antisense RNA by T7 RNA polymerase

Given that uridine analogs can activate PKR, we asked how they influence the production of dsRNA byproducts in T7 RNA polymerase reactions. First, we probed T7 RNA polymerase products with the J2 antibody, which specifically recognizes dsRNA (Schönborn et al. 1991; Bonin et al. 2000). As expected, this antibody strongly recognized annealed dsRNA but not plasmid DNA (Fig. 5H). J2 signal was higher in ftz RNA produced by T7 RNA polymerase with UTP (“U”) compared to ftz RNA produced with m1ΨTP (“m1Ψ”) (Fig. 5H) and was reduced with RNase III treatment. The total amount of RNA in each sample was similar, as judged by methylene blue staining. This suggests that m1ΨTP inhibited dsRNA production.

To determine whether antisense transcripts are produced in the T7 RNA polymerase reaction, we performed northern blots using radiolabeled probes targeting antisense or sense transcripts. We detected antisense transcripts in RNA produced by T7 RNA polymerase with UTP (Fig. 5I), which was decreased in samples treated with RNase III. Notably, no antisense transcripts were found in RNA synthesized with m1ΨTP (Fig. 5I). The levels of sense transcripts remained relatively similar across all four samples. From these observations, we conclude that m1ΨTP inhibits the production of dsRNA and antisense transcripts by T7 RNA polymerase, which is consistent with published findings (Mu et al. 2018; Baiersdörfer et al. 2019).

Excess exogenous mRNAs trigger stress granules independently of PKR activation or eIF2α phosphorylation

During our study, we noticed that when RNase III–treated mRNAs were microinjected at higher concentrations, they triggered G3BP1-enriched stress granules without promoting eIF2α phosphorylation. For example, RNase III–treated ftz mRNA microinjected at 300 ng/µL, which is four times more concentrated than what was used in previous experiments examining stress granule formation (75 ng/µL), triggered stress granules in ∼50% of microinjected cells (Fig. 6A; representative images in Fig. 6B) without detectable increase in eIF2α phosphorylation (Fig. 6C; representative images in Supplemental Fig. S5A).

FIGURE 6.

Excess exogenous mRNA, which is free of dsRNA, triggers stress granules independently of PKR activation or eIF2α phosphorylation. (A–C) Wild-type U2OS cells were microinjected with ftz mRNA (300 ng/μL), which was either treated or not with RNase III, and 70 kDa dextran conjugated to Oregon Green, fixed 30 min postinjection, and stained for ftz mRNA by FISH, the stress granule-marker G3BP1 by immunofluorescence, and DNA by DAPI. The quantification of the percentage of injected cells with visible stress granules, compared to that of noninjected cells, is shown in A; each column represents the average and standard error of at least five independent experiments, with each experiment composed of 40–70 cells. The representative images, with each row depicting a single field of view imaged for dextran/DAPI, G3BP1, and ftz mRNA, are shown in B: scale bar, 30 µm. Enlarged images of the outlined regions are shown on the right: scale bar, 2 µm. These cells were also separately stained for G3BP1 and phospho-eIF2α, and the quantifications of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view are shown in C. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of at least three independent experiments, each consisting of 40–60 cells. The representative images of phospho-eIF2α and G3BP1 co-staining in cells injected with RNase III–untreated or RNase III–treated ftz mRNA are shown in Supplemental Figure S5A. (D–F) Similar to (A–C), control- or PKR-depleted U2OS cells were microinjected with ftz mRNA (300 ng/μL), which was either treated or not with RNase III, and 70 kDa dextran conjugated to Oregon Green. The cells were incubated for 30 min, fixed, and stained for phospho-eIF2α and G3BP1 by immunofluorescence and DNA by DAPI. The percentages of injected cells with visible stress granules, compared to noninjected cells, are shown in D; each column represents the average and standard error of at least three independent experiments, with each experiment composed of 50–80 cells. Representative images of control- and PKR-depleted cells injected with RNase III–treated ftz mRNA are shown in E: arrows, injected cells; asterisks, noninjected cells; and scale bars, 30 µm. Enlarged images of the outlined regions are shown on the right: scale bar, 2 µm. Quantifications of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view for control- and PKR-depleted U2OS cells injected with RNase III–treated ftz mRNAs are shown in F. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of at least three independent experiments consisting of 40–60 cells. The representative images of phospho-eIF2α and G3BP1 co-staining in control- or PKR-depleted cells injected with RNase III–untreated ftz mRNA are shown in Supplemental Figure S5B. (G,H) Similar to (A,B), wild-type or eIF2α S51A MEF cells were microinjected with RNase III–untreated or RNase III–treated ftz mRNA (600 ng/μL) and 70 kDa dextran conjugated to Oregon Green, fixed 30 min postinjection, and stained for ftz mRNA by FISH, G3BP1 by immunofluorescence, and DNA by DAPI. The percentages of injected cells, compared to noninjected cells, with visible stress granules after injecting RNase III–untreated or RNase III–treated ftz mRNAs are shown in G. Each column represents the average and standard error of at least three independent experiments, with each experiment comprising of 40–60 cells. The representative images of wild-type or eIF2α S51A MEF cells injected with RNase III–treated ftz mRNA are shown in H: scale bar, 30 µm. Enlarged images of the outlined regions are shown on the right: scale bar, 2 µm. The representative images of phospho-eIF2α and ftz mRNA co-staining in wild-type or eIF2α S51A mutant MEF cells injected with RNase III–untreated ftz mRNA are shown in Supplemental Figure S5C. (I,J) Similar to (D–F), wild-type or ΔΔG3BP1/2 U2OS cells were microinjected with RNase III–treated ftz mRNA (300 ng/μL), fixed after 30 min, and stained for stress granule proteins and phospho-eIF2α by immunofluorescence. The percentages of injected cells, compared to noninjected cells, with visible stress granules after injecting RNase III–treated ftz mRNAs are shown in I, wherein each column represents the average and standard error of at least four independent experiments, with each experiment composed of 40–60 cells. Quantifications of whole-cell phospho-eIF2α intensities of microinjected and noninjected cells within the same field of view for wild-type or ΔΔG3BP1/2 U2OS cells injected with RNase III–treated ftz mRNAs are shown in J. Each violin plot represents the median (thick dotted line) and quartiles (thin dotted lines) of at least three independent experiments consisting of 40–60 cells. The representative images of phospho-eIF2α and G3BP1/TIA-1 co-staining in wild-type or ΔΔG3BP1/2 U2OS cells are shown in Supplemental Figure S5G. Student's t-test was performed for (A), (C), (D), (F), (G), (I), (J): (*) P < 0.05; (**) P < 0.01; U2OS cells were microinjected with ftz mRNA (300 ng/μL) and 70 kDa dextran conjugated to Oregon Green, fixed 30 min postinjection, and co-stained for phosphor-eIF2α and either G3BP1 (C) or TIA-1 (D), by immunofluorescence: scale bar, 30 µm.

To test whether this phenomenon depends on PKR, we repeated this experiment in PKR-depleted cells. PKR depletion did not impact how efficiently high concentrations of RNase III–treated mRNA triggered stress granules, which were in ∼60% of both PKR-depleted and control cells (Fig. 6D, “+RNase III”; representative images in Fig. 6E). eIF2α phosphorylation in PKR-depleted and control cells was minimal and comparable to noninjected controls (Fig. 6F; representative images in Fig. 6E). These results indicate that microinjecting higher concentrations of RNase III–treated ftz mRNA triggers stress granule assembly independently of PKR and eIF2α phosphorylation. Microinjecting RNase III–untreated ftz mRNA triggered stress granule assembly in 60%–70% of PKR-depleted cells, compared to nearly 100% of control-depleted cells (Fig. 6D, “−RNase III”; representative images in Supplemental Fig. S5B). Thus, RNase III–untreated ftz mRNA triggers stress granules through both PKR-dependent and independent pathways; the former is likely mediated by dsRNA byproducts, while the latter is not.

To examine this further, we microinjected higher concentrations of RNase III–treated ftz mRNA into wild-type cells and cells with phosphodeficient mutant eIF2α (S51A). We observed that 40%–60% of microinjected cells formed stress granules in eIF2α (S51A) MEFs, which was comparable to wild-type cells (Fig. 6G, “+RNase III”; representative images in Fig. 6H). These results further confirm that stress granule formation in response to higher concentrations of RNase III–treated ftz mRNA is independent of eIF2α phosphorylation. Without RNase III treatment, microinjecting mRNA at higher levels induced stress granules in 30%–50% of eIF2α (S51A) MEFs compared to nearly 100% of wild-type MEFs (Fig. 6G, “−RNase III”; representative images in Supplemental Fig. S5C). Thus, we estimate that ∼30%–50% of stress granule formation without RNase III treatment results from an eIF2α phosphorylation-independent mechanism, while the remaining 50%–70% is caused by dsRNA byproduct-mediated eIF2α phosphorylation.

We reasoned that if these stress granules do not require PKR, they should form even when mRNA is synthesized with uridine analogs. Consistent with this, microinjecting Ψ-ftz mRNA (not treated with RNase III) at 300 ng/µL robustly triggered stress granules in ∼80% of injected U2OS cells (Supplemental Fig. S5D; quantified in Supplemental Fig. S5E), without inducing a significant increase in eIF2α phosphorylation compared to noninjected cells (Supplemental Fig. S5D; quantified in Supplemental Fig. S5F). This contrasts with experiments using U-ftz mRNA, which contains T7 RNA polymerase byproducts and therefore robustly triggers eIF2α phosphorylation and stress granule formation, as expected (Supplemental Fig. S5D,F). Thus, higher concentrations of exogenous mRNAs with uridine analogs induce stress granules without activating PKR or eIF2α phosphorylation. Overall, we conclude that a higher concentration of mRNA, free from T7 RNA polymerase byproducts, triggers stress granules independently of PKR and eIF2α phosphorylation.

G3BP1 and G3BP2 enhance stress granule formation induced by exogenous mRNA that are free of T7 RNA polymerase byproducts

We examined the role of G3BP1 and G3BP2 in PKR-independent stress granule formation caused by higher concentrations of exogenous mRNA. We found that microinjecting RNase III–treated ftz mRNA at higher concentrations induces stress granules in ∼50% of injected wild-type U2OS cells but in 0% of injected ΔΔG3BP1/2 U2OS cells (Fig. 6I; representative images in Supplemental Fig. S5G,H). Similar results were observed with RNase III–treated m1Ψ-ftz mRNA (Supplemental Fig. S5I). eIF2α phosphorylation levels remained comparable between noninjected cells and cells injected with RNase III–treated U-ftz (Fig. 6J; representative images in Supplemental Fig. S5G) and m1Ψ-ftz mRNA (Supplemental Fig. S5I) in both cell types, indicating that cells formed stress granules independently of eIF2α phosphorylation. Thus, G3BP1 and G3BP2 are required for efficient stress granule assembly in response to exogenous mRNA without dsRNA byproducts or PKR activation.

DISCUSSION

Our work clarifies the relationship between exogenous mRNAs and the PKR-mediated innate immune response. First, our data demonstrate that PKR recognizes and is activated by T7 RNA polymerase–generated mRNAs produced with uridine analogs; the presence of N¹-methyl-pseudouridine in dsRNA triggers PKR-dependent phosphorylation of eIF2α and stress granule assembly to the same extent as uridine-containing dsRNAs (Fig. 5F,G). Thus, uridine analogs do not camouflage mRNAs from PKR detection. Instead, our data show that uridine analogs inhibit the production of antisense RNA by T7 RNA polymerase (Fig. 5H,I), consistent with previously published work (Mu et al. 2018; Baiersdörfer et al. 2019). Supporting our findings, a recent study reported that pseudouridine-containing dsRNA robustly triggers PKR and eIF2α phosphorylation upon transfection into human cells (Drazkowska et al. 2025). We thus conclude that T7 RNA polymerase–generated mRNAs activate PKR in cells due to these dsRNA byproducts.

In addition to PKR, dsRNA sensor TLR3 and ssRNA sensors TLR7 and TLR8 are activated by RNA products generated by T7 RNA polymerase (Karikó et al. 2005, 2008; Kormann et al. 2011). Their activation and subsequent IFN upregulation are reduced when the RNA is synthesized with uridine analogs (Karikó et al. 2005, 2008; Kormann et al. 2011; Andries et al. 2015). It will be important to evaluate whether these receptors are also triggered by T7 RNA polymerase byproducts, whose production is inhibited by uridine analogs.

Uridine analogs are commonly used in the synthesis of therapeutic RNAs. It should be noted that uridine analogs promote ribosomal pausing during translation elongation and hinder protein production, as demonstrated with Ψ- and m1Ψ-incorporated transcripts (Svitkin et al. 2017; Eyler et al. 2019), and thus they may reduce the overall efficacy of mRNA therapeutics. Our findings suggest that removing dsRNA byproducts of in vitro transcription may be sufficient to bypass the PKR-mediated translation block, without the need to incorporate uridine analogs into the mRNA, as these could independently hinder translation. Further research is necessary to evaluate the costs and benefits of nucleotide analogs in mRNA therapeutic production.

Another important finding from our work is that exogenous RNAs can still induce stress granule formation even when the dsRNA-PKR pathway is disabled by PKR depletion, by a phosphodeficient mutation in endogenous eIF2α, or by RNase III treatment to remove dsRNA byproducts, albeit only at four- to sixfold higher mRNA concentrations. We obtained similar results with mRNAs containing uridine analogs. It seems unlikely that the PKR-independent stress granule formation is driven by other exogenous RNA sensors, such as TLR3, TLR7, TLR8, RIG-I, or MDA5. These sensors primarily operate at the level of transcriptional regulation, which can take hours to activate (Karikó et al. 2005, 2008; Kormann et al. 2011; Andries et al. 2015), in contrast to the minutes timescale of stress granule formation that we observe. Furthermore, since these granules require G3BP1/2, they are unlikely to be other RNA structures that form in a G3BP1/2-independent manner, such as RNase L-bodies (Burke et al. 2019, 2020) or dsRNA-induced foci (Corbet et al. 2022).

Alternatively, exogenous mRNA may serve as a weak multivalent platform to trigger stress granule formation with scaffolding proteins G3BP1 and G3BP2. This explains why ΔΔG3BP1/2 cells do not readily form stress granules when microinjected with exogenous RNA, regardless of whether T7 RNA polymerase byproducts are present. Indeed, it has been proposed that excess ribosome-free mRNA nucleates the formation of stress granules (Van Treeck and Parker 2018), supported by observations that introducing RNase into the cytosol, either by injection or permeabilization, dissolves existing stress granules and prevents further assembly (Guillén-Boixet et al. 2020; Sanders et al. 2020; Yang et al. 2020). Further investigation is needed to elucidate the factors causing PKR-independent stress granule formation in response to exogenous RNA and to assess its impact on translation.

MATERIALS AND METHODS

Plasmid constructs and primers

The MHC-ftz-Δi and βG-Δi reporter constructs in pcDNA3.0 were previously described (Palazzo et al. 2007; Akef et al. 2013). The reverse complement of the MHC-ftz-Δi sequence was synthesized in pJET1.2 by General Biosystems.

Antibodies

Antibodies used in this study were PKR (Abcam ab32052), P-eIF2α (Cell Signaling Technology 9721), eIF2α (Cell Signaling Technology 9722), G3BP1 (Abcam ab56574), Caprin (Proteintech 15112-1-AP), TIA-1 (Invitrogen MA5-26474), FMRP (Cell Signaling Technology 4317), PABP (Abcam ab21060), Dcp1a (Sigma-Aldrich WH0053802M6), and tubulin (Sigma-Aldrich T6199).

Cell culture and lentiviral-mediated shRNA protein depletion

Human osteosarcoma cells (U2OS), embryonic kidney 293T cells (HEK293T), and mouse embryonic fibroblasts (MEFs) were grown in DMEM (Wisent) supplemented with 10% fetal bovine serum (FBS) (Wisent) and 5% penicillin/streptomycin (Wisent). The lentiviral-delivered shRNA protein depletion was performed as previously described (Akef et al. 2013; Lee et al. 2020). Briefly, HEK293T was plated at 40% confluency on 60 mm dishes and transfected with the gene-specific shRNA pLKO.1 plasmid (Sigma-Aldrich), packaging plasmid (Δ8.9), and envelope (VSVG) vectors using Lipo293T DNA in vitro transfection reagent (SignaGen) according to the manufacturer's protocol. Viruses were harvested from the media 48 h later and added to U2OS cells pretreated with 8 μg/mL hexadimethrine bromide. U2OS cells were then selected with 2 μg/mL puromycin media for at least 4–6 days. Western blotting was used to determine the efficiency of PKR depletion. The shRNA constructs (Sigma-Aldrich) used in this study are as follows: “PKR #1” shRNA, TRCN0000197012 (CCGGGAGGCGAGAAACTAGACAAAGCTCGAGCTTTGTCTAGTTTCTCGCCTCTTTTTTG), “PKR #2” TRCN0000196400 (CCGGGCTGAACTTCTTCATGTATGTCTCGAGACATACATGAAGAAGTTCAGCTTTTTTG).

In vitro mRNA synthesis and RNase III treatment

MHC-ftz-Δi in pCDNA3.0 or pJET1.2 vectors were first digested with XhoI to create a linear DNA template. For pJET1.2 digest, the ∼600 bp product was purified from 1% agarose gel. RNA was synthesized in vitro using HiScribe T7 High Yield RNA Synthesis Kit (New England Biolabs) at 37°C for 4–5 h, with the linear DNA as a template for run-off transcription. Template DNA was subsequently removed from the RNA synthesis reaction with DNase I incubation at 37°C for 20 min. RNA products were then isolated via rapid column-based purification with PureLink RNA Extraction kit (Invitrogen). For RNase III treatment, ShortCut RNase III (New England Biolabs) was incubated with purified total synthesized RNA (1/1000–1/2000 units/µL for 100 ng/µL of total synthesized RNA) at 37°C for 30 min, following the manufacturer's protocol. Products of RNase III treatment were subsequently isolated again using PureLink RNA Extraction kit (Invitrogen). The RNA was capped with Vaccinia Capping System (New England Biolabs) and polyadenylated with Escherichia coli Poly(A) Polymerase (New England Biolabs), following the manufacturer's protocol. For pseudouridine-incorporated mRNA, all of the UTP in the reaction was replaced by Ψ-TP or m1Ψ-TP (TriLink). The in vitro synthesized mRNA was purified, evaluated via spectroscopy, and resolved on an agarose gel (1%).

In vitro RNA annealing reaction

Forward (fwd) and reverse complement (RC) of fushi tarazu in pJET1.2 were first digested by XhoI, gel-extracted, transcribed in vitro via HiScribe T7 Polymerase, subsequently treated with RNase III, and purified, as described above. Equimolar amounts of fwd and RC-ftz were then incubated together in annealing buffer (10 mM Tris, 50 mM NaCl, and 1 mM EDTA) at 95°C for 5 min. Afterward, the RNA samples were kept on the inactive heat block for 1 h, to allow for slow cooling to room temperature. The in vitro annealed RNA was then directly injected into cells.

Microinjection

For microinjection experiments, cells were plated on 22 × 22 mm coverslips (VWR) in 35 mm mammalian tissue culture dishes (Thermo Scientific) for 24 h prior to injection. Microinjections were performed as previously described (Gueroussov et al. 2010). RNA was microinjected at indicated concentrations with 70 kDa FITC- or Oregon Green–conjugated dextran (1/4th–1/5th volume) in injection buffer (140 mM KCl, 10 mM HEPES, pH 7.4). Injected cells were incubated at 37°C for the indicated times and then fixed with 4% paraformaldehyde in 1× PBS for 30 min at 4°C.

Estimation of injected RNA molecule quantity

Fushi tarazu mRNA containing a 7-methylguanylate cap and a poly(A) tail of average 100 bp is ∼230,000 g/mol (230 kDa), according to ribonucleotide monophosphate molecular masses described by Thermo Fisher. We estimate that per microinjection, the change in U2OS cell volume is nearly 10%. Together, 300 ng/µL represents ∼45,000 ftz mRNA molecules (4%–8% of the total number of transcripts in the cell), and 75 ng/µL represents ∼11,000 mRNA molecules (1%–2% of the total number of transcripts in the cell).

Fluorescence in situ hybridization (FISH) and oligo(dT) staining

For mRNA staining, cells were washed twice in 1× sodium saline citrate (SSC; 150 mM NaCl, 15 mM sodium citrate, pH 7.0) buffer supplemented with 60% formamide. Cells were then treated with hybridization buffer (60% formamide, 100 mg/mL dextran sulfate, 1 mg/mL yeast tRNA, 5 mM vanadyl riboside complex, and 1× SSC) containing 200 nM Alexa 546–conjugated ssDNA probe (Integrated DNA Technologies) for 18–24 h at 37°C. Subsequently, coverslips were washed with 1× SSC supplemented with 60% formamide, and the coverslips were mounted on DAPI. The probe oligonucleotide sequences included anti-ftz (GTCGAGCCTGCCTTTGTCATCGTCGTCCTTGATAGTCACAACAGCCGGGACAACACCCAT) and anti-β-Globin (CTTCATCCACGTTCACCTTCGCCCCACAGGGCAGTAACGGCAGACTTCTCCTCAGGAGTCA). For poly(A) mRNA staining, cells were washed in 1× SSC supplemented with 25% formamide. Cells were treated with poly(A) hybridization buffer (100 mg/mL dextran sulfate, 1 mg/mL yeast tRNA, 5 mM vanadyl riboside complex, and 1× SSC) containing 200 nM Alexa 546–conjugated ssDNA 60mer dT oligonucleotide (Integrated DNA Technologies).

The FISH signal is reflective of the presence of reporter RNA for two reasons. Firstly, the staining is internally controlled between microinjected and noninjected cells. If the FISH probe hybridizations were not specific, we would expect to see RNA FISH staining in both injected and noninjected cells. This difference can be seen in several images in the manuscript (e.g., in Supplemental Fig. S2A,B, compare the injected cells, which have bright FISH signals, and noninjected cells, which have no detectable FISH signal). Secondly, α-amanitin-treatment of cells microinjected with plasmids containing the ftz gene abolishes RNA FISH staining when compared to control-treated cells (Gueroussov et al. 2010). Thus, the detection of ftz RNAs by FISH probes depends on the presence of target RNAs.

Stellaris single-molecule (sm)FISH

Stellaris smFISH experiments were performed as previously described (Wu and Palazzo 2020). To summarize, following fixation with 4% paraformaldehyde (Electron Microscopy Sciences), coverslips were submerged in methanol at −20°C for 30 min and then rehydrated twice in 1× PBS, each time incubating for 15 min. Cells were then washed twice in 1× SSC buffer (150 mM NaCl, 15 mM NaCitrate, pH 7.1) with 10% formamide and incubated with 100 μL of hybridization buffer (100 mg/mL dextran sulfate, 10%–15% formamide, and 2× SSC) containing 125 nM Stellaris probes (LGC Biosearch Technologies) for 48 h.

Immunofluorescence

Following fixation with 4% paraformaldehyde, the coverslips were washed with 1× PBS twice and then incubated with 0.1% Triton X-100 in 1× PBS for 15 min. Subsequently, the coverslips were washed with 1× PBS twice and incubated with either anti-TIA-1 antibodies (mouse monoclonal, 1:200 dilution, Invitrogen), anti-G3BP1 (mouse monoclonal, 1:500 dilution, Abcam), anti-FMRP (rabbit polyclonal, 1:50 dilution, Cell Signaling Technology), anti-PABP (rabbit polyclonal, 1:200 dilution, Abcam), anti-Caprin (rabbit polyclonal, 1:200 dilution, Proteintech) anti-Dcp1a (mouse monoclonal, 1:200 dilution, Sigma-Aldrich), and anti-P-eIF2α (rabbit polyclonal, 1:200 dilution, Cell Signaling Technology) antibodies for 1 h. Antibody incubation was supplemented with 0.1 mg/mL RNase-free BSA and 0.1% Triton X-100 in 1× PBS. In between antibody incubations, coverslips were washed by being placed cell-side down onto 0.5 mL droplets of 1× PBS on parafilm and allowed to incubate at room temperature three times for 5 min. Lastly, the coverslips were mounted onto glass coverslips using DAPI Fluoromount-G stain mounting solution (Southern Biotech).

Microscopy and image analysis

Cells were imaged using an epi-fluorescence microscope with a 60× objective (Nikon). Image analysis of stress granule presence and eIF2α phosphorylation levels was performed using Nikon Imaging Software (NIS) Elements Advanced Research (Nikon). For stress granule assembly analysis, microinjected cells containing one or more distinct foci or granules were identified as stress-granule-positive cells. For eIF2α phosphorylation analysis, whole-cell intensity for P-eIF2α staining was measured for both microinjected and noninjected cells and normalized to cell-absent background intensity per field of view.

Immunoblotting

For immunoblot analysis, U2OS cells were washed twice with ice-cold 1× PBS, then lysed with sample buffer containing 50 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, 0.1% bromophenol blue, and 1% β-mercaptoethanol, and preheated at 95°C. The whole-cell lysates were subsequently further heated at 95°C for 5 min and then rested on ice for 5 min. The lysate was passed through a 25-gauge needle to shear genomic DNA. Whole-cell lysates were separated by SDS-PAGE on 6%–15% polyacrylamide gels, transferred to nitrocellulose membranes, and probed with primary antibodies against PKR (rabbit monoclonal, 1:3000 dilution, Abcam), phospho-eIF2α (rabbit polyclonal, 1:3000 dilution, Cell Signaling Technology), eIF2α (rabbit polyclonal, 1:3000 dilution, Cell Signaling Technology), and tubulin (mouse monoclonal DM1a, 1:1000 dilution, Sigma-Aldrich). Chemiluminescence luminol reagent (Pierce) and the VersaDoc system (Bio-Rad) were used to visualize the blots.

RNA electrophoresis and northern blotting

³²P-labeled probes were generated by phosphorylating desalted IDT primers (“antisense”: TACGTCCTGCAGTAAGCTCGCTTTCTTGCTGTCCCAATTTCTATTAAACTCGA; “sense”: TCGAGTTTAATAGAAATTGGGACAGCAAGAAAGCGAGCTTACTGCAGGACGTA) with [γ³²P]ATP (Revvity BLU035C001MC) using T4 polynucleotide kinase (NEB M0201S). Radiolabeled probes were purified using the MicroSpin G-25 Columns (Cytiva 27532501).

Purified in vitro transcribed RNA was resuspended in RNA loading dye (NEB B0363S) and heated at 70°C for 10 min followed by rapid cooling on ice. Four micrograms of total RNA was loaded on a MOPS–formaldehyde agarose gel (1% agarose, 1× MOPS, 1.3% formaldehyde). Electrophoresis was carried out in 1× MOPS running buffer at 100 V for 1.0 h (or until bromophenol blue reached the center of the gel). Gel was then equilibrated in 10× SSC buffer for 20 min. RNA was transferred onto a charged nylon membrane (Cytiva Amersham Hybond-N+) in 10× SSC buffer via capillary transfer. Transferred RNA was cross-linked to the membrane using UV light (UV Stratalinker Model 1800; auto-cross-link setting). Membranes were incubated with 50 mL prehybridization buffer (5× SSC, 20 mM Na₂HPO₄ pH 7.2, 7% SDS, 2× Denhardt's solution [Invitrogen] and sheared salmon sperm DNA (1 mg per blot]) for at least 1.5 h at 30°C in a hybridization oven prior to the addition of radiolabeled probes. Membranes were incubated with radiolabeled hybridization solution overnight at 30°C. Membranes were washed three times with 50 mL wash buffer (3× SSC, 25 mM Na₂HPO₄ pH 7.2, and 5% SDS) for 20 min each at 30°C. Membranes were exposed on phosphor screen overnight. Images were taken with a Typhoon FLA 9000 imager.

Detection of dsRNA by RNA dot blot analysis

After in vitro transcription from T7 RNA polymerase and subsequent RNase III treatment, the RNA samples were diluted in nuclease-free water to a final concentration of 200 ng/μL in 5 μL aliquots (total loading mass of 1 μg per sample) and then spotted on a positively charged nylon membrane (Invitrogen, 0.45 μm pore size). The spotted RNAs were cross-linked to the nylon membrane using UV light (UV Stratalinker Model 1800; 125 m J/cm² at 254 nM). After cross-linking, the membrane was washed in 10 mL of Tris-buffered saline solution with 0.1% Tween-20 (1× TBS-T) for 5 min at room temperature and then blocked in 10 mL of 1× TBS-T with 0.5% BSA for 1 h at room temperature. For detecting dsRNA, the membrane was incubated in 10 mL of 1× TBS-T with 0.5% BSA and J2 anti-dsRNA antibody (mouse monoclonal, Scicons 10010200) diluted to 1:5000 at 4°C overnight. The membrane was subsequently washed three times for 5 min each in 10 mL of 1× TBS-T with 0.5% BSA at room temperature and then incubated with horseradish peroxidase (HRP)–conjugated anti-mouse secondary antibody diluted to 1:5000 in 1× TBS-T with 0.5% BSA for 1 h at room temperature. After washing the membranes three times for 5 min each in 10 mL TBS-T at room temperature, the blots were visualized by chemiluminescence luminol reagent (Pierce) and the VersaDoc system (Bio-Rad). After chemiluminescent imaging, the membrane was transferred to 10 mL of methylene blue staining buffer (0.2% methylene blue in 0.4 M sodium acetate and 0.4 M acetic acid) for 5 min, washed with dH₂O for 5 min, and imaged on the VersaDoc.

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.