Cytoplasmic protein-free mRNA induces stress granules by two G3BP1/2-dependent mechanisms

This article has been Reviewed by the following groups

Read the full article See related articles


Stress granules are cytoplasmic membraneless organelles that sequester proteins and non-translating mRNAs in response to various stressors. To assess the contributions of mRNA and RNA-binding proteins to stress granule formation, we use microinjection to deliver protein-free mRNA into the cytoplasm in a controlled manner. We demonstrate that mRNAs trigger stress granule formation through two mechanisms that are enhanced by the presence of G3BP1 and G3BP2. Low concentrations of in vitro transcribed mRNA activated protein kinase R (PKR), leading to phosphorylation and inhibition of the eukaryotic translation initiation factor eIF2α and stress granule formation. This was inhibited by replacing uridine with pseudouridine in the mRNA or by treating it with RNase III, which cleaves double-stranded RNA. High concentrations of mRNA triggered stress granule formation by a mechanism that was independent of PKR and enhanced by G3BP1/2, highlighting the importance of both protein-free mRNA and RNA-binding proteins in stress granule formation.

Graphical Abstract/Model


Microinjected mRNA induces stress granules in mammalian cells by two G3BP1/2-dependent mechanisms: one requires the stress-sensing protein kinase PKR to phosphorylate the translation initiation factor eIF2α, and the other is independent of PKR and phospho-eIF2α and acts when the cytoplasmic concentration of ribosome-free mRNA is increased acutely.

Article activity feed

  1. This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at

    The manuscript Cytoplasmic protein-free mRNA induces stress granules by two G3BP1/2-dependent mechanisms assesses the effects of microinjected mRNA in the formation of stress granules. They found that injecting low amounts of RNA triggers stress granule formation by activating PKR, and high concentrations of untranslating mRNAs induces stress granules through a mechanism independent of PKR, and partially in cells lacking G3BP proteins. This work provides strong support for previously established models of stress granule assembly. They use a novel method of microinjecting RNA which allowed them to directly study the effects of changes of concentration of protein free mRNAs in the cell, which has previously not been done to this extent. It demonstrates clearly that the formation of stress granules is dependent on the concentration of its components and can be triggered directly by increasing the cytoplasmic concentration of non-translating mRNAs.

    Major comments –

    The manuscript tells a fairly complete story, but I'd like to bring up a couple of alternative interpretations and ideas for the author's consideration that would make the conclusions more robust:

    1.      The authors interpret the results as acute increase of free mRNA saturates the capacity of RNA-binding proteins to prevent RNA-RNA interactions. There are other (or concurrent) interpretations of how the cell might be reacting to the acute increase of mRNAs. For example, does this increase of mRNA affect translation? The new mRNA may saturate the translation machinery leading to shutdown of translation without eIF2-alpha phosphorylation. The authors could explore this idea with a puromycin translation assay.

    2.      The authors' explanation of puromycin on page 19 is not completely correct. They state "puromycin treatment releases most mRNA from ribosomes…" Puromycin induces premature termination of translation, leading to the disassembly of the polysome but does not prevent continued loading of the initiation complex.

    Also in this paragraph, they hypothesize that "naked mRNAs will be bound quickly by RNA-binding proteins, reducing the effective concentrations of naked mRNAs…" How would this hypothesis explain how drugs that release mRNA from ribosomes cause stress granules without additional stressors and without phosphorylating eIF2-alpha (See for example, pateamine A in Dang et al, 2006, JCB)? As mentioned above, puromycin doesn't prevent loading of the initiation complex, while pateamine A does. It follows that the continued loading of initiation machinery and transiently elongating ribosomes in puromycin treatment keeps mRNAs from being incorporated into a stress granule. However, if that is what the authors mean they should make it clear that "RNA-binding proteins" includes the translation initiation complex. This makes exploring if additional mRNA is saturating the translation machinery even more interesting.

    Minor points –

    1.      Injecting large amounts of RNA, which contain dsRNA regions, might also trigger RNase L activation which can lead to the formation of G3BP positive puncta called RNase L bodies (described in Burke et al., 2020 JCB). The authors might want to rule out puncta formed in PKR deletion cells are RNase L bodies by performing a RNase L knockout.

    2.      The amount of mRNA in the cell line used here (page 8, line 14) has been published in Khong et al 2017 (Mol. Cell). Using this more recent data, the amount of mRNA post-injection would be higher. 

    3.      Some experiments were quantified using SG+ cells. I'm curious to see the quantification of stress granule area or ratio of stress granule area to cytoplasm area. I think this would be an excellent addition to the binary SG+ cells quantification method.

    4.      I find it curious that the total amount of RNA in the cell does not change that much according to oligo-d(T) staining even with 300 ng/uL (Figure 1b). Is this analysis within the linear range of poly-d(T) staining? I think the highest concentrations used in this manuscript (600 ng/uL) would cause a significant change and could be used as a positive control for the assay.

    5.      How many cells were micro-injected? I recommend adding that information to the methodology.

    Competing interests

    The authors declare that they have no competing interests.

  2. This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at

    Cytoplasmic protein-free mRNA induce stress granules by two G3BP1/2-dependent mechanisms

                In this manuscript, the authors demonstrate that microinjected mRNAs can promote stress granule formation in cells via two different mechanisms. When the mRNA concentration is low, the in vitro transcribed dsRNA can trigger a PKR-dependent mechanism that phosphorylates eIF2α, leading to translation repression and stress granule formation. However, at high mRNA concentration and when PKR is inactivated, the cells can activate a PKR-independent mechanism to assemble stress granules. Furthermore, they show that both the PKR-dependent and PKR-independent mechanisms are influenced by G3BP1/2. Overall, this manuscript is interesting and uses a new method of microinjection to provide substantial support for existing models of stress granule assembly.

    The manuscript made several conclusions, and some of them would require further work to make a robust conclusion.

    Major point:

    1.     In cells injected with high concentrations of mRNA, it is possible stress granules are forming due to mRNA translation being repressed by other eIF2α-independent mechanisms. This analysis would be stronger if the authors perform a SUnSET assay to measure changes in mRNA translation upon injecting a high concentration of mRNA.

    2.     In Fig. 1B, the authors used poly(A) staining to show that microinjection of mRNA did not affect the total mRNA amount in cells. One potential concern is that due to abundance of poly(A) signals at baseline level, the signal detection might be at saturation and no longer at the dynamic range to detect changes, if any. The authors could complement this by performing a Northern blot against poly(A) tails at the linear range of detection.

    3.     In their final conclusion, the authors propose that endogenous ribosome-free mRNAs released during ribosome run-off is insufficient to induce stress granule formation. They performed experiment by microinjecting 300ng/uL ftz mRNA and treating with puromycin in cells that cannot phosphorylate eIF2α. However, they have previously shown in Supple Fig. 2E that stress granules formation is restored without the need for puromycin in these cell lines. Thus, the experiment would have been more convincing if the authors performed this experiment with 75ng/uL ftz mRNA.

    4.     The authors used puromycin and harringtonine to induce translation shut-off and ribosome run-off. However, there are some caveats regarding the effects of these drugs. While puromycin is known to disassemble polysomes from mRNAs, it does not usually lead to stress granule formation (Kedersha et al., 2000, JCB). Similarly for harringtonine, while the drug traps a vacant 80S ribosome at the start of the transcript and leads to ribosome run-off for elongating ribosomes, it does not promote stress granule assembly (Fedorovskiy et al., 2023, Biochemistry (Moscow)). Therefore, it would strengthen the conclusion if the authors revise their interpretation of the data presented in Fig. 6 based on these caveats.

    5.     The authors could consider incorporating more parameters when quantifying stress granule assemblies. Apart from counting the number of stress granule positive cells, it would be informative to see if treatments alter stress granule size and number. For example, in the images shown in Fig. 5A and 6E, stress granules size looks smaller in mutant cells compared in WT cells.

    Minor point:

    1.     For one of the yellow arrows in Fig. 1F, the cell contains TIA-1+ assemblies too

    2.     Specify what is defined as docking of PB to SG? What's the range of distance that is defined as docking?

    3.     Typo in Fig. 2C, y-axis

    4.     Lacks p-eIF2α staining in Fig. 2D to show that shPKR is working, and ftz mRNA FISH in Fig. 2E to show its localization

    5.     Would prefer if the quantification of SG area in between shCtrl and shPKR in Fig. 2E is shown

    6.     Rescue the "no SG" phenotype in Fig. 2 by putting WT-PKR or OE WT p-eIF2α to show that it is a direct effect of the ISR.

    7.     Lacks ftz mRNA staining in Fig. 3D to show that ftz mRNA localizes to SG in absence of p-eIF2a activation

    8.     Wrong citation: (see quantification in Fig. 5A) à should be Fig. 5B

    9.     No SD for the WT injected with Puromycin column in Fig. 6C

    10. Missing total number of cells that is injected with RNA and quantified for each experiment

    Competing interests

    The authors declare that they have no competing interests.