Stress granule-mediated sequestration of EGR1 mRNAs correlates with lomustine-induced cell death prevention Free

Stress granules (SGs) are dynamic cytoplasmic biocondensates that are formed in eukaryotic cells under various stress conditions, including heat shock, oxidative stress, viral infection and upon stresses induced by treatment with anticancer drugs (Pietras et al., 2021; Protter and Parker, 2016). Their formation is a part of cellular response programs that help cells cope with adverse conditions (Guzikowski et al., 2019). SGs are composed of RNAs and various proteins (Buchan and Parker, 2009). The RNA components in SGs mainly comprise mRNAs or ribosomal 18S RNA (rRNA) of the small 40S ribosomal subunit (Campos-Melo et al., 2021). The core protein components are ribosomal proteins of the 40S subunit, various RNA-binding proteins (RBPs) and eukaryotic translation initiation factors (eIFs), and other proteins that constitute stable or transient constituents of SGs (Protter and Parker, 2016). Dynamic association of specific proteins (such as signaling molecules) with SGs is proposed to regulate cellular metabolism (Ivanov et al., 2019).

SG formation is dynamic, in equilibrium with global mRNA translation and is regulated by two main molecular mechanisms controlling protein synthesis. The first mechanism involves the phosphorylation of serine 51 (Ser51) of the α subunit of eukaryotic initiation factor 2 (eIF2α) by stress-sensing kinases (Hofmann et al., 2021). The second mechanism interferes with the assembly of the trimeric translation initiation complex 4F (eIF4F) via inhibition of mammalian target of rapamycin (mTOR) activity or by the interference with eIF4A functions (Emara et al., 2012; Fujimura et al., 2012; Iwasaki et al., 2016). Both translation inhibition mechanisms lead to the inhibition of translation initiation, promoting polysome disassembly (Balagopal and Parker, 2009). In turn, untranslated ribonucleoprotein complexes containing messenger RNAs (mRNPs) decorated with specific translation factors and RBPs condensate into SGs via liquid–liquid phase separation. SG dynamics is also modulated by ATP-dependent processes via inactivation of eIF4A (Tauber et al., 2020). Through ATP-dependent RNA binding, eIF4A effectively reduces RNA condensation in vitro and restricts stress granule assembly in the cell.

Previously, we have demonstrated that some anticancer drugs, such as vinca alkaloids (vinorelbine, vincristine and vinblastine), activate a SG-mediated stress response program, which promotes the survival of stressed cancer cells by reprograming gene expression and inhibiting pro-apoptotic signaling pathways (Szaflarski et al., 2016). Such cytoprotection is achieved by assembling canonical SGs lacking specific signaling molecules. This approach opens a new window for a detailed study of their role in cancer cell survival during chemotherapy. Other studies suggest that SGs might significantly impact the efficacy of cancer therapy, possibly through manipulating their formation to withstand the effects of anticancer drugs (Kaehler et al., 2014; Li et al., 2023; Pietras et al., 2022).

Sequestration of proteins into SGs modifies cellular signaling pathways and determines cell fate (Kedersha et al., 2013). This phenomenon highlights the intricate relationship between cellular stress responses and the complex regulatory mechanisms governing cell survival or death. The interplay between SGs, mRNA translation and cellular signaling pathways in cell fate determination is an active area of research. One of the proposed mechanisms is the sequestration of specific signaling proteins within SGs, which in turn influence downstream events, such as the activation or suppression of crucial signal transduction cascades.

In cancer cells, SGs are suggested to play a crucial role in promoting the proliferation and the inhibition of apoptosis, both of which are beneficial for cancer progression (Arimoto et al., 2008; Tsai and Wei, 2010). In addition, SGs support the invasion and migration of cancer cells (Somasekharan et al., 2015). SG formation is associated with cancer therapy resistance, including resistance to chemotherapy and radiotherapy, which results in less effective therapy efficiency (Moeller et al., 2004). Theoretically, any anticancer drug that promotes eIF2α phosphorylation can also cause the formation of SGs. It is well documented that various eIF2α kinases are activated in response to different drugs. For example, sorafenib, which triggers endoplasmic reticulum stress, activates the kinase PERK (also known as EIF2AK3) (Adjibade et al., 2015). Lapatinib, a tyrosine kinase inhibitor, also activates PERK to phosphorylate eIF2α and promote SG formation (Adjibade et al., 2020). By contrast, 5-fluorouracil treatment activates the kinase PKR (also known as EIF2AK2) while bortezomib triggers the kinase HRI (also known as EIF2AK1), both leading to SG formation (Kaehler et al., 2014). The eIF2α phosphorylation is not the only pathway mediating SG formation in response to anticancer drugs. Previously, we have demonstrated that vinca alkaloids block mTOR activity, disrupting the eIF4F complex, which then facilitates the formation of SGs (Szaflarski et al., 2016).

Lomustine belongs to the nitrosourea family and is a bifunctional alkylating agent that causes the alkylation of guanine in the sixth position of the nucleotide and its conversion into O6- alkylguanine, which can form incorrect bonds with thymine during replication (Sebolt-Leopold and Scavone, 1992). Lomustine is mainly used to treat brain tumors (primary or metastatic) (Taal et al., 2014) and non-Hodgkin's lymphomas as the second-line option (Musolino et al., 2005). As a common side effect, lomustine causes delayed myelosuppression, resulting in thrombocytopenia, leukopenia and anemia (Jakobsen et al., 2018). Furthermore, long-term use of lomustine is associated with developing secondary malignancies (Vesper et al., 2009).

In this study, we found that lomustine promotes formation of SGs. By analyzing the transcriptome, we discovered that lomustine modulates the expression of the EGR1 gene. Further experiments revealed that EGR1 mRNA is sequestered into SGs upon lomustine treatment, reducing the available EGR1 mRNA pool and thus downregulating their translation. As the EGR1 protein promotes apoptosis (Wang et al., 2021), by reducing its expression in an SG-dependent manner, cancer cells become more resistant to lomustine treatment. Our findings suggest that SGs are part of a specific stress response pathway modulated by anti-tumor drugs. Such SG-mediated post-transcriptional modulation of specific transcripts contributes to chemotherapy resistance and can selectively fine-tune gene expression, cell survival and apoptosis.

To evaluate the cytotoxicity of lomustine in human osteosarcoma U2OS cells, we exposed them to increasing drug concentrations for 24 h. Cell viability was assessed to demonstrate the cytotoxic effects of lomustine, and we determined the IC₅₀ (inhibitory concentration 50%) value to be ∼750 µM (Fig. 1A). Based on the determined IC₅₀, we found an optimal concentration comprising minimal dose and maximal efficiency at inducing SG formation (Fig. 1B; Fig. S1). At a concentration of 200 µM, lomustine induced SG formation within 1 h (Fig. 1C). Lomustine-induced SGs are positive for the canonical SG markers G3BP1, eIF4G and eIF3b. It should be noted that eIF3b is absent in some SG subtypes, which are considered pro-apoptotic (Aulas et al., 2018; Fujimura et al., 2012). Hence, lomustine-induced SGs are canonical with proposed pro-survival properties. The correlation between the concentration of lomustine and the number of SGs is demonstrated in Fig. 1B. Based on the above data, 200 µM lomustine triggered SGs in 80% of cells. Compositionally, lomustine-induced SGs are positive for TRAF2, FXR1, HuR (also known as ELAVL1), eIF4E, RACK1, TIA-1 and TIAR (also known as TIAL1) (Fig. 1D,E). Some of these proteins have been previously reported to associate with SGs either transiently or permanently and might contribute to cell fate (Szaflarski et al., 2016). Additionally, we observed the uS6 protein (also known as RPS6) in lomustine-induced SGs, indicating the presence of small ribosomal 40S subunits (Fig. 1G). Altogether, these data demonstrate that lomustine induces the formation of canonical SGs.

N-acetylcysteine (NAC), a potent antioxidant, suppresses the formation of NaAsO₂-induced SGs, indicating that oxidative stress is a crucial factor in SG induction (Szaflarski et al., 2016). Like NaAsO₂, lomustine-induced SG formation is sensitive to NAC, suggesting that the generation of reactive oxygen species (ROS) contributes to SG formation (Fig. 1F).

Inhibition of protein synthesis is crucial for the formation of SGs (Mokas et al., 2009). A combination immunofluorescence with ribopuromycylation assay is a powerful tool to analyze both the translation activity and SG formation within a single cell. As expected, cells with higher translation activity did not form SGs; however, those presenting lower translation activity formed SGs (Fig. 2A). Cycloheximide (CHX), an elongation inhibitor, was used as a control for quantification (analyzed as heat maps and western blots in Fig. S3). This finding shows that lomustine promotes SGs through inhibition of translation. We validated this finding with a polysome profiling analysis, which revealed a dramatic reduction of polysomes and an increase in monosomes. These profiles are consistent with those seen with other stressors, such as NaAsO₂, which inhibits translation and triggers SG formation (Fig. 2B).

To probe the potential effect of lomustine on eIF2α phosphorylation, we performed ribopuromycylation assays in a set of human haploid HAP1 cells with knockouts of individual eIF2α kinases and cells edited to carry the eIF2α^Ser51Ala (S51A) variant, which cannot be phosphorylated (Aulas et al., 2018; David et al., 2012; Panas et al., 2015). Exposure of wild-type HAP1 cells to 200 µM lomustine for 1 h led to complete inhibition of protein synthesis (Fig. 2C, puromysin incorporation signal). However, in S51A or ΔHRI cells, translation inhibition was ablated. Knockouts of other eIF2α kinases, such as GCN2, PKR and PERK, did not affect inhibition of translation upon lomustine treatment. NaAsO₂ treatment, which activates HRI, was used as a control (McEwen et al., 2005). Production of ATF4, a factor expressed in response to eIF2α phosphorylation and which is crucial for activating the integrated stress response (ISR), mirrors the ribopuromycylation results (Fig. 2C).

By checking eIF2α phosphorylation status using antibodies specific to phosphorylated (P-)eIF2α, we confirmed that lomustine does not trigger eIF2α phosphorylation in HAP1 ΔHRI cells (Fig. 3A), in agreement with data of Fig. 2. Treatment with NaAsO₂ was used as a control (McEwen et al., 2005). As expected, neither lomustine nor NaAsO₂ promoted eIF2α phosphorylation in the mutant S51A HAP1 cells (Fig. 3A).

In addition to eIF2α phosphorylation, mTOR signaling regulates protein synthesis by controlling the formation of the eIF4F complex. Under optimal growth conditions, the mTORC1 kinase phosphorylates 4E-BPs to prevent them from binding to eIF4E and allows the formation of a stable eIF4F complex (Hay and Sonenberg, 2004). By contrast, dephosphorylated 4E-BPs bind to eIF4E, thus inhibiting protein biosynthesis. Compounds such as selenium or hydrogen peroxide lead to the arrest of protein biosynthesis by blocking the formation of the eIF4F initiation complex, and consequently promote SG formation (Emara et al., 2012; Fujimura et al., 2012). Our data suggest that lomustine does not promote the binding of the 4E-BP1 to the eIF4E, and thus does not inhibit eIF4F complex assembly (Fig. 3B). Therefore, the primary factor that promotes the inhibition of protein synthesis under exposure to lomustine is the phosphorylation of eIF2α.

Furthermore, lomustine fails to activate SG formation in S51A and ΔHRI HAP1 cells. By contrast, ΔGCN2, ΔPKR and ΔPERK cells formed SGs (Fig. 3C). Similarly, lomustine did not promote SG formation in S51A mutant mouse embryonic fibroblast cells (MEFs), whereas it induced SGs in control (parental) MEFs (Fig. S2). Lomustine also did not promote SG formation in human osteosarcoma U2OS cells with HRI knockout (U2OS ΔHRI) (Fig. 3D). Overall, all these findings confirm that lomustine activates SG formation through HRI-assisted phosphorylation of eIF2α.

We used a molecular array transcriptomic approach (Affymetrix) to investigate how lomustine and NaAsO₂ impact the transcriptome expression in a set of genetically edited HAP1 cells (ΔHRI, ΔGCN2, ΔPKR, ΔPERK and S51A). We treated individual cell lines with 200 µM lomustine or 100 µM NaAsO₂ for 2 h and measured the relative expression of genes (analyzing 34,662 transcripts in total). We then quantified the gene expression levels between individual cell lines exposed to the drug and the untreated cell line. Our analysis shows that some genes were markedly upregulated in response to lomustine and NaAsO₂ (Fig. 4A) compared to untreated parental cells (Fig. 4B).

The set of highly expressed mRNAs included EGR1, DDIT4, CHAC1, ATF3, JUN, FOSB, SESN2 and ULBP (also known as RAET1L). To confirm our results, we quantified specific transcripts using real-time quantitative RT-PCR (RT-qPCR) and observed high expression levels of corresponding mRNAs in cells exposed to lomustine and NaAsO₂ (Fig. 4C).

Early growth response protein 1 (EGR1) was one of the most highly expressed genes changed in response to treatments. In the case of the parental line, its expression was the highest after treating cells with lomustine, while it was in the second place among the 34,662 transcripts examined after NaAsO₂ administration. RT-qPCR against EGR1 mRNA confirmed this finding (Fig. 4C).

We applied fluorescence in situ hybridization (FISH) to determine whether highly expressed transcripts (EGR1 and DDIT4) associate with SGs. First, we induced SG formation in U2OS cells with a 200 µM lomustine or 200 µM NaAsO₂ treatment for 2 h. Then, cells were stained and processed with FISH using synthetic ssDNA oligomers complementary to the EGR1 or DDIT4 transcripts. The Cy3-labeled oligonucleotide (dT) probe was used as a control detecting all poly(A)-tail-containing transcripts. In addition to FISH probes, we used G3BP1 as a marker to visualize SGs.

Both lomustine and NaAsO₂ induced SG formation (Fig. 5, top left and right panels, respectively). We found that the EGR1 transcript was quantitatively colocalized with SGs whereas it was diffusely distributed throughout the cytosol in control untreated cells (Fig. 5, lower panel). Surprisingly, the second highly expressed transcript, which encodes DDIT4 protein, was not associated with either lomustine- and NaAsO₂-induced SGs. Signal from an oligo(dT) probe, used as a positive control, was strongly present in SGs, in agreement with published data (Khong et al., 2017). These findings confirmed that SGs could sequester mRNA transcripts (such as EGR1 mRNA) selectively and quantitatively.

We then investigated the impact of association of EGR1 transcripts with SGs on their relative translation. We utilized cell lines that could not form SGs despite exposure to stress. Specifically, two cell lines were used: (1) human S51A HAP1 cells, and (2) the mouse MEF-A/A cell line (eIF2α^Ser51Ala). Both cell lines carry a non-phosphorylatable variant of eIF2α, resulting in insensitivity to activation of any of eIF2α kinases. Parental (PAR) cell lines were employed as controls (MEF-S/S and HAP1-PAR). In addition, we analyzed ΔHRI, ΔGCN2, ΔPKR and ΔPERK HAP1 cell lines.

First, we examined the expression of EGR1 protein in all HAP1 cells (PAR, S51A, ΔHRI, ΔGCN₂, ΔPKR and ΔPERK) treated with lomustine or NaAsO₂ (or left untreated as controls). Our results showed that, compared to control, only cell lines incapable of forming SGs expressed higher EGR1 protein levels when activated by lomustine- or NaAsO₂-induced stress (Fig. 6A, HAP1-S51A and HAP1-ΔHRI cells). In contrast, in the lomustine- or NaAsO₂-treated lines capable of forming SGs, the protein levels were lower compared to that in the lines that did not form SGs and untreated cells, except for the HAP1-PAR line, where it was higher than in cells treated with any drugs. These results demonstrate that the presence of SGs reduces levels of EGR1 protein (Fig. 6A). This phenomenon is not due to an effect on protein stability, given that treatment with MG132, a known proteasome inhibitor, did not affect the decrease of EGR1 protein when cells were exposed to increasing concentrations of lomustine (Fig. S4).

We obtained similar results when examining the levels of EGR1 protein in MEFs (Fig. 6B). In MEF-A/A cells, the level of EGR1 protein production significantly exceeded that observed in MEF-S/S cells, which are capable of forming SGs (Fig. 6B, lower left). As a control, we also monitored the expression of ATF4 protein, whose translation is upregulated by eIF2α phosphorylation (Dey et al., 2010; Vattem and Wek, 2004). Indeed, ATF4 underwent translation exclusively in cells with P-eIF2α in response to drug-induced stress exposure (Fig. 6B).

Using U2OS cells, we investigated the relationship between EGR1 protein levels and the number of cells containing lomustine-induced SGs (Fig. 7). Fig. 7A presents immunofluorescence images of SGs in U2OS cells exposed to increasing concentrations of lomustine. Simultaneously, we analyzed the amount of EGR1 protein produced using western blotting (Fig. 7B). Upon quantification of both features, namely the percentage of cells with SGs and the amount of synthesized EGR1 protein, we concluded that the more SGs that are present, the less EGR1 protein is produced (Fig. 7C), which means that EGR1 production is controlled by translation efficiency and mRNA accessibility.

Previous studies have demonstrated that the EGR1 protein is pro-apoptotic. We used HAP1 ΔEGR1 cells to show that indeed EGR1 protein is not produced in these cells (Fig. 8A). Then, we tested the viability of cells in the presence of lomustine, and confirmed that HAP1 ΔEGR1 cells were less affected by lomustine than parental HAP1 cells (Fig. 8B). We also compared the viability of HAP1 S51A cells to HAP1 ΔEGR1 and parental cells (Fig. 8C). HAP1 S51A cells are the most sensitive cells, likely because they cannot form SGs and/or efficiently block mRNA translation. Altogether, our data suggest that SGs directly regulate expression levels of EGR1 by sequestration of EGR1 transcripts in response to stress (Fig. 8D).

In this study, we demonstrate the ability of lomustine to promote SG formation. In U2OS cells, this activity is observed after 1 h of exposure at 150–200 µM concentrations, which is comparable to what is seen in other studies on isolated cells involving lomustine (Shinwari et al., 2011; Staberg et al., 2017).

Lomustine-induced SGs are canonical SGs. The presence of eIF3b in the lomustine-activated SG hints at their protective and anti-apoptotic nature, as the absence of this protein promotes apoptosis (Aulas et al., 2018; Emara et al., 2012). The sensitivity of lomustine-activated SGs to NAC, suggests that SG activation is due, at least in part, to ROS production (Alluri et al., 2021; Aulas et al., 2018; Szaflarski et al., 2016), and lomustine is indeed known to generate ROS intracellularly (Behnisch-Cornwell et al., 2020). This also agrees with published literature where potent inducers of ROS, such as NaAsO₂ or nitric oxide, promote SG formation that is sensitive to NAC. NAC is a synergetic effector for anticancer treatment, especially in breast cancer, and can be used as a supplement during treatment of patients (Kwon, 2021; Monti et al., 2017). Indeed, NAC-mediated inhibition of the formation of lomustine-induced SGs might be beneficial given that EGR1 mRNA would then become available for translation, and synthesized EGR1 would trigger apoptosis.

SG formation is linked to protein synthesis. Indeed, in lomustine-treated cells that readily form SGs, global translational activity was found to be low (Fig. 2). This is due to lomustine activation of the HRI kinase to induce eIF2α phosphorylation (Fig. 3A). In turn, HRI-dependent phosphorylation in eIF2α is essential for SG formation (Fig. 3C,D). Lomustine does not affect the formation of the eIF4F initiation complex, suggesting that eIF2α phosphorylation is the sole pathway for activating SG formation (Fig. 3).

To determine the impact of lomustine treatment on gene expression, we employed microarray technology, which allowed us to assess the expression of 34,662 transcripts. We analyzed relative changes in expression in wild-type (PAR), eIF2α(S51A), ΔHRI, ΔGCN2, ΔPKR and ΔPERK HAP1 cells exposed to lomustine and NaAsO₂ (Fig. 4). We detected high expression levels of EGR1, DDIT4, CHAC1, ATF3, JUN, FOSB, SESN2 and ULBP1 mRNAs in treated cells. EGR1, ATF3, JUN and FOSB are stress-responsive transcription factors that regulate expression programs related to the proliferation of cancer cells (Lim et al., 1998). DDIT4 encodes a protein that is crucial for in regulating cellular metabolism by negatively regulating the mTOR pathway (Sofer et al., 2005). CHAC1 is a protein that plays a role in oxidative stress and response to nutrient deprivation and might also affect apoptosis processes (Crawford et al., 2015). SESN2 has been shown to regulate signaling pathways involved in the cellular stress response, including the AMP-activated protein kinase (AMPK) pathway, the mTOR pathway and the p53 tumor suppressor pathway (Rhee and Bae, 2015).

The EGR1 gene exhibits the most robust expression across all cell lines (Fig. 4). This gene encodes a transcription factor responsible for the early stress response of the cell. Previously published data suggest that EGR1 is the most upregulated gene following JNK activation, with JUN being an essential effector in EGR1 transcription regulation (Hoffmann et al., 2008). Our data also suggests exceptionally high expression of JUN, particularly after NaAsO₂- induced activation.

We probed the possible correlation between SGs and highly expressed transcripts after administration of lomustine and NaAsO₂, namely EGR1 and DDIT4 mRNAs. Interestingly, the EGR1 transcript was localized almost exclusively within SGs, unlike the DDIT4 transcript (Fig. 5). It should be noted that during heat shock, SGs have been shown to selectively exclude mRNAs encoding heat-shock proteins, making these transcripts available for translation, and thus helping cells to cope with heat shock (Kedersha and Anderson, 2002).

Further analysis revealed that the appearance of lomustine-induced SGs correlates with a decrease in EGR1 protein production (Fig. 7). We observed a similar effect of SGs upon reducing EGR1 protein production with NaAsO₂. We hypothesize that SGs regulate availability of specific transcripts for translation through sequestration, making such mRNAs unavailable to ribosomes.

Published data support the proapoptotic nature of the EGR1 protein (Boone et al., 2011; Wang et al., 2021). To examine the impact of EGR1 on cell survival, HAP1 cells with a knockout of the EGR1 gene were used. EGR1-deficient cells exhibited higher viability in the presence of lomustine than the parental cells capable of producing EGR1 (Fig. 8B). Importantly, HAP1 S51A cells, which cannot form SGs, exhibited a reduced survival rate (Fig. 8C). This suggests that lomustine-activated SGs have an anti-apoptotic nature.

Based on data, we propose an integrated model demonstrating how SGs increase the viability of cells exposed to lomustine. At low concentrations of the drug, EGR1 is highly expressed and might induce apoptosis. At higher concentrations of lomustine, SGs are efficiently formed. The EGR1 transcripts, which are highly expressed in cells, are then potently sequestered into SG, and their translation is inhibited (Fig. 8D). Cells demonstrate higher viability and survival because of such sequestration and downregulation of EGR1 expression.

Considering the many processes associated with multidrug resistance, we propose that SG-mediated sequestration of transcripts modulates the expression of various proteins, including EGR1. During chemotherapy treatments, SGs might directly modulate cell survival and apoptosis, thus making SGs a potential anticancer target.

The human osteosarcoma cell line (U2OS) was obtained from the American Type Culture Collection (ATCC^® HTB-96^TM). Mouse embryonic fibroblasts (MEFs, with or without the eIF2α-S51A mutation) were a gift from Oded Meyuhas, The Hebrew University of Jerusalem, Israel). The human-derived HAP1 (parental, S51A, ΔHRI, ΔGCN2, ΔPKR, ΔPERK, ΔEGR1 and ΔATF4) cell lines were purchased from Horizon Discovery, UK. The mutations in ΔHRI, ΔGCN2, ΔPKR and ΔPERK cell lines were confirmed by sequencing as presented in Fig. S5. Furthermore, parental, S51A, ΔHRI, ΔGCN2, ΔPKR and ΔPERK cell lines had already been used in another study, and their susceptibility for SG analysis was already confirmed (Aulas et al., 2018). U2OS and MEF cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM, Gibco #11995-065) supplemented with 10% fetal bovine serum (FBS, Thermo Fisher Scientific #10270106) and 1% penicillin-streptomycin cocktail (Sigma-Aldrich). HAP cell lines were cultured in Iscove's modified Dulbecco's medium (IMDM, Gibco #12440-053) supplemented with 10% FBS (Thermo Fisher Scientific #10270106) and 1% penicillin-streptomycin. All cell lines were maintained at 37°C with 5% CO₂. All cells were frequently tested for mycoplasma contamination using a MycoAlert Mycoplasma Detection Kit (Lonza, LT07-418).

5×10⁴ U2OS cells were grown on glass coverslips and incubated overnight under standard conditions. Cells were washed twice with PBS (Gibco, #70013-016), fixed in 4% paraformaldehyde, and permeabilized in 96% cold methanol (−20°C) (POCH, #621990110), for 15 min each, and washed with 1× PBS at room temperature. Then, samples were incubated with a blocking buffer of 5% horse serum (Sigma, #H1217) in PBS for 1 h at room temperature. Cells were incubated with primary antibodies (overnight) at 4°C. The following day, cells were washed with PBS and then incubated with secondary antibodies for 1 h in darkness. Secondary antibodies (Jackson Laboratories) were tagged with Cy2, Cy3 or Cy5. DAPI was used with the secondary antibodies to stain the nuclei. Cells were washed three times with 1× PBS (5 min each). Coverslips with cells were mounted in a polyvinyl mounting medium. Cells were imaged using an Olympus FV10i confocal laser scanning microscope. The images were analyzed and merged using Adobe Photoshop 2020.

This assay is based on the capability of puromycin incorporation into a nascent polypeptide chain during active protein biosynthesis. It is wildly used for the determination of the level of protein biosynthesis in cells (Aulas et al., 2018; Pietras et al., 2022; Szaflarski et al., 2016, 2022). Briefly, puromycin is added to the growing cells (5 µg/ml for 5 min before terminating experiments). Then, cells underwent either a western blot (WB) or immunofluorescence (IF) procedure. Puromycin was detected by antibodies (MABE343; 1:200 dilution for IF; 1:1000 dilution for WB, Millipore). Western blots were quantified using ImageJ (Schindelin et al., 2012). Immunofluorescence was visualized by heat maps (Olympus FV10i laser scanning microscope) and parametrized in ImageJ.

5×10⁴ U2OS cells grown on coverslips and incubated overnight under standard conditions. Cells were exposed to an appropriate drug for a defined time in 400 μl medium per single well in a 24-well plate. Then, drug-stressed cells were then fixed in 4% paraformaldehyde in PBS and permeabilized in 96% cold methanol (10 min each) and washed with 1× PBS for 5 min. To block samples, PerfectHyb™Plus Hybridization Buffer (Sigma-Aldrich, H7033) for 15 min at 52°C. Then cells were hybridized with Cy3-labeled FISH probes [synthetic oligo-dT40 labeled with Cy3 or Cy5; EGR1 rRNA, 5′-GGTGCAGGCTCCAGGGAAAAGCGGCCA–Cy3; DDIT4 rRNA,: 5′- TGTGTCCCCAATGCACAGCCACCCGCA–Cy3; and poly(A) rRNA, 5′-40×T-Cy3; Metabion International AG, Germany] for 1 h at 52°C. Then, samples were washed two times with 2× SSC, obtained from a water dilution of 20× UltraPure™ SSC (Gibco, #15557-044) for 5 min each (the first time with pre-wormed and the second time with room temperature buffer) and one time with PBS. Primary and secondary antibodies (see below) with DAPI were diluted in 5% horse serum and cells incubated (45 min each). Finally, coverslips with cells were washed twice with PBS and mounted in a polyvinyl mounting medium. Images were collected using an Olympus FV10i laser scanning microscope and then parametrically analyzed using Imaris software. For the colocalization analysis, the JACoP plugin for ImageJ software was used (Bolte and Cordelières, 2006).

4.5×10⁵ cells grown in 6-well plates overnight at 37°C with 5% CO₂ in supplemented medium. Then, the cells were washed with 1× PBS for 5 min. Total protein isolation was performed using the Minute™ Detergent Free protein extraction kit for cultured cells (Invent Biotechnologies, Plymouth, USA) according to the manufacturer's instructions. Total protein quantification was performed using the Pierce BCA protein assay (Thermo Fisher Scientific, #23227). Equal amounts of proteins were loaded on 4–20% TGX gel (Bio-Rad, Hercules, USA) for 30 min at 200 V. After electrophoresis, the gel was transferred onto nitrocellulose membranes using Trans-Blot^® Turbo™ system (Bio-Rad). For total protein visualization, the membrane was counterstained with Ponceau. The membrane was incubated with 5% nonfat dried milk (Cell Signaling, #9999) in Tris-buffered saline with Tween 20 (TBST) buffer (Sigma, #91414-100TAB) for 1 h at room temperature. Membranes were washed five times for 5 min with TBST to remove residual milk. Primary antibodies diluted in 5% horse serum were then applied and incubated overnight at 4°C. Secondary antibodies diluted in 5% horse serum were used for 1 h at room temperature. Membranes were also washed five times after each type of antibody. Finally, according to the manufacturer's instructions, proteins were detected using SuperSignal™ West Pico PLUS (Thermo Fisher Scientific) chemiluminescent substrate. Images were processed using an Amersham Imager 600 system and then quantified using ImageJ (Fiji) software (Schindelin et al., 2012). For the figures, the western blot images were processed in Adobe Photoshop. Here, all images were subjected to the Level tool. The input levels were designated automatically, and output levels were always equal to 150.

Anti-G3BP1 (sc-81940; 1:200 dilution for IF), anti-eIF4G (sc-11373; 1:200 dilution for IF), anti-eIF3b (sc-16377; 1:200 dilution for IF), anti-eIF4E (sc-9976; 1:200 dilution for IF), anti- TRAF2 (sc-2345, 1:200 dilution for IF), anti-FXR1 (sc-10554, 1:200 dilution for IF), anti-HuR (sc-5261; 1:200 dilution for IF), anti-TIAR (sc-1749; 1:200 dilution for IF), anti-TIA-1 (sc-1751; 1:1000 dilution for IF), anti-Rack1 (sc-17754; 1:200 dilution for IF) and anti-rpS6 (sc-74459; 1:200 dilution for IF) were purchased from Santa Cruz Biotechnology. Anti-G3BP1 (13057-2-AP; 1:200 dilution for IF) and anti-ATF4 (10835-1-AP; 1:1000 dilution for WB) were purchased from Protein Technology Group. Anti-EGR1 (#4154; 1:1000 dilution for WB), anti-total-4E-BP1 (#9452; 1:1000 dilution for WB) and anti-β-actin (#8457; 1:1000 dilution for WB), anti-β-tubulin (#2128; 1:1000 dilution for WB) and anti-total-eIF2α (#2103, 1:1000 dilution for WB) were purchased from Cell Signaling Technology. Anti-puromycin (MABE343; 1:200 dilution for IF; 1:1000 dilution for WB) was purchased from Millipore. Anti-P-eIF2α (ab32157; 1:1000 dilution for WB) was purchased from Abcam. The secondary antibodies for IF included Cy™2 AffiniPure donkey anti-mouse IgG (cat. 715-225-150), Cy™3 AffiniPure donkey anti-rabbit IgG (711-165- 152), and Alexa Fluor^® 647 AffiniPure bovine anti-goat IgG (805-605-180) and were purchased from Jackson ImmunoResearch.

Total RNA was isolated from HAP1 cells exposed to an appropriate drug for a defined time or left untreated (parental, S51A, ΔHRI, ΔGCN2, ΔPKR and ΔPERK) using Universal RNA Purification Kit (EUR_x, Poland). Then, the protocol, including in vitro transcription, biotin labeling and cDNA fragmentation, was performed using the Affymetrix GeneChip IVT express kit (Affymetrix, Santa Clara, USA). The biotin-labeled cDNA was hybridized with the Affymetrix Gene Chip Human Genome U219 microarrays and appropriate internal controls. The hybridization was performed in the AccuBlock^TM digital dry bath hybridization oven (Labnet International, Inc., Edison, USA) at 45°C for 16 h. Subsequently, the microarrays were washed and stained using the Affymetrix GeneAtlas Fluidics Station (Affymetrix, Santa Clara, USA). The microarrays were scanned using the imaging station of the GeneAtlas System. Initial analysis of the scanned microarrays was carried out with Affymetrix GeneAtlas operating software. The generated CEL files were further analyzed using the R statistical language and Bioconductor package with the relevant Bioconductor libraries.

At 24 h before drug treatment, cells were plated in 6-well plates at 4.5×10⁵ cells/well and cultured in their respective medium. The medium was refreshed 30 min before treatment and then cells were treated with 100 µmol/l NaAsO₂ for 1 h and 200 µmol/l lomustine for 2 h.

Total RNA was isolated according to the protocol for Universal RNA/miRNA Purification Kit (EURx, Gdańsk, Poland) and quantified using NanoDrop (Thermo Fisher Scientific, Waltham, USA). LunaScript RT^TM SuperMix (New England Biolabs, #10105340) generated cDNA according to the manufacturer's guidelines. RT-qPCR was carried out using Luna^® Universal qPCR Master Mix (New England Biolabs, #10103269) following the manufacturer's instructions. We have chosen glyceraldehyde-3-phosphate dehydrogenase (GAPDH), hypoxanthine phosphoribosyltransferase (HPRT), and alpha-2-macroglobulin (A2M) as reference genes. Expression level differences among samples were calculated using the ΔΔCt model. The primers for qPCR are listed in Table S1.

15–50% linear sucrose gradients containing 10 mM HEPES-KOH pH 7.4, 150 mM KCl, 10 mM MgCl₂, 0.1 mg/ml cycloheximide (Sigma, 01810-1G) and 1 mM DTT, were poured using Biocomp Gradient master. Cells were pretreated with 0.1 mg/ml of cycloheximide for 10 min before harvesting in lysis buffer composed of 10 mM HEPES-KOH pH 7.4, 150 mM KCl, 10 mM MgCl₂, 1 mM DTT, 0.1 mg/ml cycloheximide, 2% NP-50, 1× protease inhibitors (EDTA-free; Halt™ Phosphatase Inhibitor Cocktail, Thermo Fisher Scientific, 78420) and 1 µM Ac-DEVD-CHO (Sigma, A0835-1MG). Lysates were cleared by centrifugation (10,000 g for 10 min) and quantified by means of a Bradford assay. At least 1 mg of total protein was added on the top of each gradient. Polysomes were resolved by centrifugation for 2 h at 283,000 g in a SW40 rotor at 4°C. Gradients were fractionated using Biocomp fractionator.

2×10⁵ cells grown in 6-well plates overnight at 37°C with 5% CO₂ in supplemented medium. Cells were treated with drugs (NaAsO₂ for 1 h and lomustine for 2 h). The drug was washed with Hanks’ balanced salt solution (HBSS) and administrated in a fresh medium. The cells were left for 48 h in the incubator. After 2 days, cells were harvested, pelleted and resuspended in a 100 µl medium. 10 µl of cells were mixed with 10 µl trypan blue solution (0.4%). Then, the viability quantification was measured in Countess II Automated Cell Counter (Invitrogen™).

The CytoTox-Glo™ cytotoxicity assay was performed according to the manufacturer's instructions (Promega) with 10³ cells grown in a 96-well plate. Briefly, to each well 100 μl of AAF-Glo™ reagent was added, and the plate was incubated for 15 min at room temperature. For determination of living cells, the luminescence was measured, and 100 μl of lysis reagent with digitonin was added and incubated for 15 min. The luminescence was measured a second time to determine the total cells. The viability of cells was denoted as the percentage of living cells.

The method for 7-methyl GTP sepharose chromatography was as previously described (Ivanov et al., 2011). Torin used to treat cells was purchased from Selleckchem (S2827).

The mutations in HAP1 cells (ΔHRI, ΔGCN2, ΔPKR and ΔPERK) were verified using Sanger sequencing and are shown in Fig. S5. First, DNA from cells was isolated using a commercially available kit (Cell Culture DNA Purification Kit, EUR_X, Poland, cat. no. E3555). Then, gene fragments were amplified using standard PCR using primers demonstrated in Table S1 (Q5^® High-Fidelity DNA Polymerase, New England Biolabs, USA). PCR products were purified using a PCR/DNA clean-up purification kit (EURX, Poland, cat. no. E3520). Finally, pure PCR fragments were sequenced by oligo.pl (Poland), using primers shown in Table S1, and the sequence analysis was done using SnapGene software (https://www.snapgene.com/).

For RT-qPCR data, one-way ANOVA with Dunnett's post-hoc test was used to calculate statistical significance. A two-tailed paired t-test was used to calculate statistical value for cell viability data. For polysome quantification, an ordinary one-way ANOVA with Dunnett's post-hoc test was used. Statistical analysis was performed using GraphPad Prism 8.0.1 for Windows and is denoted as *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

We thank Dr Monika Świerczewska for the preparation of Fig. 8D using BioRender (https://www.biorender.com/).

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.261825.reviewer-comments.pdf