SARS-CoV-2 variant-specific differences in inhibiting the effects of the PKR-activated integrated stress response

Highlights • SARS-CoV-2 triggers the integrated stress response through activation of PKR.• Omicron induces more stress granule formation than Delta or Ancestral SARS-CoV-2.• ISR inhibition negatively affects SARS-CoV-2 replication.


Introduction
Coronaviruses are a family of enveloped, positive-sensed, singlestranded RNA viruses within the Nidovirales order.They infect a variety of animals like birds, bats and other mammals including humans (Masters, 2006).Since its emergence in November 2019, the betacoronavirus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has quickly spread around the world causing coronavirus disease 2019 (COVID-19) (Gorbalenya et al., 2020;Huang et al., 2020;Hu et al., 2021).SARS-CoV-2 has mutated over time, giving rise to several variants associated with higher infectivity and increased escape from neutralizing antibodies (Lippi et al., 2022;Abdool Karim and de Oliveira, 2021).This includes Delta (B.1.617.2) which was first described in India in October 2020 and Omicron (B.1.1.529)which was first described in South Africa in November 2021 (Planas et al., 2021;Thakur and Ratho, 2022).The offspring of Omicron subvariants is still dominating in the population to this date (Chakraborty et al., 2023).Viral infections induce a variety of stress stimuli within the host cells which can trigger stress sensors and thereby activate the integrated stress response (ISR).Activation of the ISR-pathway allows the cells to respond to the stress, either by resolving it or by inducing apoptosis (McCormick and Khaperskyy, 2017;Pakos-Zebrucka et al., 2016).The ISR is activated by phosphorylation of the alpha subunit of eukaryotic initiation factor 2 (eIF2α) by four distinct kinases each sensing a different type of stress.Heme regulated eIF2α-kinase (HRI, EIF2AK1) gets activated during oxidative stress.dsRNA-activated protein kinase R (PKR, EIF2AK2) senses double-stranded RNA (dsRNA), PKR-like kinase (PERK, EIF2AK3) detects misfolded proteins in the endoplasmic reticulum and general control nonderepressible 2 (GCN2, EIF2AK4) senses amino acid availability and responds to a lack of amino acids as well as glucose deprivation (Pakos-Zebrucka et al., 2016;Harding et al., 2003;Donnelly et al., 2013).Phosphorylation of eIF2α, by the above-mentioned kinases, triggers a signaling pathway leading to a reduction in global translational levels and an increased expression of a few selected genes (stress-responsive transcripts) that enable the cell to recover from the stress and terminate the ISR (Wek, 2018).As a result of the translational shutdown, stalled initiation complexes containing silenced mRNAs, small ribosomal subunits, translation initiation factors Uninfected and SARS-CoV-2 infected cells were treated with 1 mM sodium arsenite (SA), an ISR-inducing chemical, for 1 h, as positive control.Cell lysates were collected at 24 hpi and analyzed for levels of p-eIF2α.Calnexin was used as a loading control.(B) Protein levels and phosphorylation of PKR in SARS-CoV-2 infected A549-hACE2 cells.Cells were infected SARS-CoV-2.Cell lysates were collected at 24 hpi and analyzed for expression and phosphorylation levels of PKR.b-actin was used as a loading control.(C) Quantification of (B).Protein levels were normalized to cellular levels of b-actin.The data are presented as the fold change in relation to the protein levels of uninfected cells.Data are represented as mean ± SEM. n = 3. (D) G3BP1 staining in SARS-CoV-2 infected A549-hACE2 cells.Cells were infected with SARS-CoV-2 and fixed after 24 h.Uninfected cells that were either untreated or treated with 1 mM SA were used as a control.The cells were stained with antibodies against SARS-CoV-2 N protein (magenta) and G3BP1 (green) and with DAPI (blue).Imaging was performed with confocal microscopy (20X).Scale bars 20 mm.(E) Localization of the viral N protein within SGs.A549-hACE2 cells were fixed 24 h hpi and stained with antibodies against SARS-CoV-2 N protein (magenta) and G3BP1 (green) and with DAPI (blue).Arrows show colocalization of viral proteins with G3BP1.Imaging was performed with confocal microscopy (20X).Scale bars 20 mm.*, p < 0.05; **, p < 0.005.and other RNA-binding proteins accumulate within the cytoplasm to form stress granules (SGs) (Kedersha et al., 2002;Kimball et al., 2003).These structures are thought to function as a storage for the initiation complexes until translation initiation is restored.SGs also provide certain antiviral functions, for example by recruiting and activating antiviral proteins or by sequestering viral factors (McCormick and Khaperskyy, 2017;Onomoto et al., 2012).
Viruses are dependent on the host cell translation machinery for their own protein production and therefore many viruses have developed strategies to interfere with the ISR and SG formation (Poblete-Durán et al., 2016).A common approach, also employed by other coronaviruses like MERS-CoV and infectious bronchitis virus (IBV), is to inhibit the eIF2α-kinase PKR (Gao, 2021;Rabouw et al., 2016).Other viruses inhibit eIF2α-phosphorylation or cleave G3BP, a protein that is essential for SG formation (Linero et al., 2011;White et al., 2007).The SARS-CoV-2 nucleocapsid protein (N protein) has been shown to inhibit SG formation via binding to G3BP1 and G3BP2 (Luo et al., 2021;Zheng et al., 2021;Liu et al., 2022).However, many questions about the activation and role of the ISR during SARS-CoV-2 infection remain open.
Here, we show that the ISR is activated during SARS-CoV-2 infection via activation of PKR, but despite phosphorylation of eIF2α only limited SG formation or expression of stress-responsive proteins is observed, indicating SARS-CoV-2 mediated active inhibition of the ISR pathway.Chemical stimulation of the stress kinases shows viral resistance especially towards stress mediated through PKR but also a general inhibition of stress-responsive protein expression.Finally, we show specific differences in the activation and inhibition of the ISR pathway between the ancestral strain and the Delta and Omicron variants, highlighting SARS-CoV-2 sub lineage differences in manipulation of cell responses.

SARS-CoV-2 infection activates the integrated stress response through PKR but does not trigger widespread SG formation
To investigate whether the ISR gets activated during SARS-CoV-2 infection, levels of phosphorylated eIF2α (p-eIF2α) were measured in Vero E6 cells, an African green monkey kidney cell line, and in A549-hACE2 cells, a human lung epithelial cell line overexpressing human ACE2.Cells treated with sodium arsenite, a strong inducer of eIF2αphosphorylation, were used as a control.At 24 h post-infection (hpi), phosphorylation of eIF2α was detected in both cell lines (Fig. 1A).The eIF2α-kinase PKR is commonly activated upon infections with various viruses (Eiermann et al., 2020).To analyze its activation during SARS-CoV-2 infection, we examined PKR phosphorylation status and expression levels in A549-hACE2 cells.As shown in an earlier study (Li et al., 2021), strong phosphorylation of PKR was observed (Fig. 1B and  C).
To assess whether the eIF2α-phosphorylation was associated to formation of SGs, infected cells were stained for the SG marker G3BP1 and inspected via immunofluorescence.As reported by others (Zheng et al., 2021), the vast majority of SARS-CoV-2 infected cells were negative for SGs (Fig. 1D, Suppl.Fig. 1).The few infected cells positive for SGs showed lower levels of N protein and co-localization of N protein with G3BP within SGs (Fig. 1E).
Together, these results show that while the ISR is activated during SARS-CoV-2 infection, as indicated by phosphorylation of eIF2α and PKR, this does not trigger widespread SG formation.

SARS-CoV-2 infected cells are resistant to oxidative and PKRmediated stress
The SARS-CoV-2 N protein has been shown to inhibit SG formation by binding to the SG-nucleating protein G3BP (Luo et al., 2021;Zheng et al., 2021).As this occurs downstream of eIF2α-phosphorylation the N protein can therefore potentially inhibit all types of ISR-activating stress signals.To study the ability of SARS-CoV-2 to inhibit SG formation in more detail, uninfected and infected Vero E6 cells were exposed to the stress-inducers sodium arsenite, Poly(I:C), thapsigargin, or starvation, and then examined for SG formation (Fig. 2A).As previously reported, SG formation was inhibited in SARS-CoV-2 infected cells exposed to sodium arsenite, an inducer of oxidative stress (Luo et al., 2021;Zheng et al., 2021) (Fig. 2B).A reduction in SG formation in infected compared to uninfected cells was also observed after Poly(I:C) treatment, which triggers activation of PKR (Fig. 2C).The ER stress inducer thapsigargin and starvation both induced SG-formation in infected cells but not in uninfected cells (Fig. 2D and 2E).Together, this suggests that SARS-CoV-2 can inhibit SG-formation triggered by oxidative stress and dsRNA, but infected cells remain susceptible and even sensitized towards SG formation induced via ER stress and starvation.
To better understand the effects of different stress stimuli on SARS-CoV2 infectivity, we investigated the infection rate in Vero E6 cells when adding the virus together with sodium arsenite, Poly(I:C) or thapsigargin or with starvation medium (Fig. 2A).All treatments negatively affected the infection rate, and the strongest effect was observed for thapsigargin, which almost completely inhibited infection when administered with the virus (Fig. 2F).To analyze the effect of the treatments on viral replication, infected cells were first infected, and at 18 h hpi treated with the different stress inducers for 6 h before collection of supernatants at 24 hpi.Viral titers were then analyzed in the supernatants (Fig. 2A).In contrast to the infection rate, viral titers were not significantly affected by the different treatments (Fig. 2G, Suppl.Fig. 2A).This indicates that while SARS-CoV-2 is susceptible towards the cellular stress response during the initial infection step, progeny virus production does not decrease significantly in infected cells when stress responses are induced.

The SARS-CoV-2 ancestral strain, delta, and omicron BA.1 show differences in ISR activation and SG formation
The SARS-CoV-2 variants Delta and Omicron have accumulated large numbers of mutations throughout their genome (Hodcroft, 2021).While mutations of the spike protein are known to impact receptor binding efficiency and neutralizing antibody escape, little is known about possible differences in intracellular virus-host interactions between different SARS-CoV-2 variants.To investigate possible variant specific differences in suppressing SG formation, ancestral SARS-CoV-2, Delta, and Omicron BA.1 infected Vero E6 cells were analyzed for SG formation at different time points post-infection.The SARS-CoV-2 ancestral strain and Delta showed a similar pattern with a low prevalence of SGs throughout the infection (Fig. 3A).However, Omicron induced significantly higher levels of SG-positive cells than the other variants at 24 hpi and 48 hpi (Fig. 3A).This difference was even more pronounced in A549-hACE2 cells, where around 5 % of ancestral strain or Delta infected cells were SG-positive compared to almost 60 % of Omicron infected cells (Fig. 3B, Suppl.Fig. 3A).
As we observed higher frequency of SG-positive cells after Omicroninfection, we next analyzed for possible variant-specific differences in capacity to suppress SG-formation.All three variants efficiently suppressed sodium arsenite-induced SG formation in both cell types (Fig. 3C), suggesting a common mechanism to inhibit oxidative stress mediated SG-formation among SARS-CoV-2 variants.The N protein has been suggested to be important for SARS-CoV-2 mediated inhibition of SGs (Luo et al., 2021).In line with this, lower N protein levels were observed in ancestral strain and Delta infected SG-positive, compared to SG-negative, cells (Suppl.Fig. 3A and B).In contrast, no difference in N protein levels were observed between SG-positive and SG-negative omicron-infected cells (Suppl.Fig. 3A and B) which suggests that the Omicron N protein might be less effective in inhibiting SG formation.This could explain why SGs appear more frequently in Omicron-infected cells and may contribute to the decrease in progeny virus production we observed when comparing viral replication of Omicron with the ancestral strain and Delta in A549-hACE2 cells (Fig. 3D).
To further investigate if the observed differences in infectioninduced SG formation among the three variants were due to differences in ISR activation, we next analyzed levels of PKR and eIF2α phosphorylation in infected A549-hACE2 cells.Clear differences were noted for the three variants: Levels of p-elF2a were highest in Omicroninfected cells, and higher levels were observed in cells infected with the ancestral SARS-CoV-2 compared to Delta infected cells (Fig. 3E and F).
Omicron and the ancestral strain induced similar levels of p-PKR while the levels were clearly lower in Delta-infected cells (Fig. 3E and F).
While focusing primarily on PKR and eIF2α phosphorylation, we also noticed that Omicron has an N protein pattern which is distinct from the other two variants.For both the ancestral strain and Delta N proteins two separate bands of similar intensity can be detected.In contrast, for the Omicron N protein it is mainly the lower band that can be detected whereas the upper one gives no or only a very weak signal (Fig. 3E).

Effects of stress inducers on infection rate and progeny virus production for ancestral SARS-CoV-2, delta, and Omicron
After observing differences in ISR activation and SG formation between the SARS-CoV-2 variants, we next compared their capacity to withstand stress stimuli.To this end, we analyzed the infection rate of the ancestral strain and the Delta and Omicron variants in A549-hACE2 cells treated with different stress inducers concurrently with the infection (Fig. 4A).As observed in ancestral strain-infected Vero E6 cells (Fig. 2E), thapsigargin treatment almost completely blocked infection with all three variants in A549-hACE2 cells (Fig. 4A).In A549-hACE2 cells, significant reduction in infectivity was also observed with Poly(I: C) treatment for all variants and SA treatment for the ancestral strain and Delta, while the effect of starvation was small to negligible (Fig. 4A).Analysis of viral titers following post treatments with stress inducers (added to the cells at 18 hpi) showed resistance to Poly(I:C) and susceptibility to thapsigargin by all three variants, but responses to SA and starvation differ between Omicron and the other two variants (Fig 4B, Suppl.Fig. 2B).

Activation of the ISR contributes to translational arrest of infected cells and is beneficial for SARS-CoV-2 replication
During activation of the ISR, eIF2α-phosphorylation leads to a global Cells infected with different SARS-CoV-2 variants were fixed at 6, 24 or 48 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n ≥ 3. (B) SG formation in SARS-CoV-2 infected A549-hACE2 cells.Cells infected with different SARS-CoV-2 variants were fixed at 24 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n ≥ 3. (C) SG formation after sodium arsenite challenge in SARS-CoV-2 infected Vero E6 and A549-hACE2 cells.Cells were infected with different SARS-CoV-2 variants and treated with 1 mM sodium arsenite for 1 h.Cells were fixed at 24 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM.  and 3E) triggers a translational shutdown, infected A549-hACE2 cells were exposed to puromycin followed by analysis of total levels of newly produced proteins (Fig. 5A and B).At 24 h after infection with ancestral SARS-CoV-2, Delta, or Omicron, global host cell protein production was clearly inhibited (Fig 5A and B).Treatment with sodium arsenite reduced host cell protein production even more (Fig 5A and B), suggesting that despite already high phosphorylation of eIF2α infected cells remain susceptible to increased translational arrest mediated by additional ISR activation.Further analysis by immunofluorescence microscopy showed that translational levels were reduced as early as 6 hpi in infected Vero E6 cells (suppl.Fig 5A and B).
To further investigate the extent to which ISR activation contributes to the translational arrest after SARS-CoV-2 infection (Fig. 5A), infected cells were treated with the ISR inhibitor ISRIB, which targets eIF2B and thereby inhibits p-eIF2α-mediated translational shutdown (Rabouw et al., 2019).ISRIB treatment restored translation in infected cells to some extent (Fig. 5A and B, Suppl.Fig. 5A and B), but the impact was significant only in omicron-infected cells (Fig. 5A and B).ISRIB treatment also reduced SG formation at 6 hpi (Suppl.Fig. 5C), indicating that the ISR plays a role in formation of SGs and triggering translational shutdown at early stages of infection.In uninfected cells treated with sodium arsenite, translational levels were only partly restored.Sodium arsenite highly activates the ISR, possibly exceeding the threshold of eIF2α-phosphorylation which ISRIB can counter (Rabouw et al., 2019).
To analyze the impact of ISR activation on viral replication for the three variants, we compared viral titers of untreated and ISRIB treated SARS-CoV-2 infected A549-hACE2 cells (Fig. 5C).Overall, inhibiting the ISR had little effect on viral replication though a small decrease in viral titers could be observed for all three variants.Intriguingly, this indicates that the translational inhibition resulting from ISR activation during SARS-CoV-2 infection does not negatively affect virus production.

Expression of ATF4 and CHOP is suppressed in SARS-CoV-2 infected cells
While phosphorylation of eIF2α leads to a reduction in global translational levels, selected stress-responsive transcripts are still translated and even upregulated under these conditions (Pakos- Zebrucka et al., 2016).This includes the transcription factor ATF4, a master regulator that controls the expression of a variety of proteins that in turn enable the cell to recover from stress, promote cell survival and eventually terminate the ISR via dephosphorylation of eIF2α (Novoa et al., 2003).However, if the stress is too intense or long-lasting, ATF4 can also induce apoptosis, for example through upregulation of the transcription factor CHOP (Gil and Esteban, 2000;Teske et al., 2013).Since SARS-CoV-2 infection induced eIF2α phosphorylation (Fig. 1A), we analyzed whether this also triggered the expression of ATF4 and its downstream effector CHOP.Interestingly, neither ATF4 nor CHOP could be detected in ancestral strain-infected A549-hACE2 cells and even SA-treatment failed to induce ATF4 production in infected cells (Fig. 6A  and B).Treatment with thapsigargin induced strong expression of ATF4 and CHOP in uninfected and infected cells, but the levels of both were clearly reduced in infected cells despite high p-eIF2α levels, suggesting a viral mechanism inhibiting the ISR downstream of eIF2α (Fig. 6A).ISRIB treatment had a similar impact as SARS-CoV-2 infection on thapsigargin-induced production of ATF4 and CHOP, reducing the expression of both proteins in uninfected cells and further decreasing their levels in infected cells to an almost undetectable level (Fig. 6C and  D).Comparison of the SARS-CoV-2 variants showed even stronger inhibition of ATF4 and CHOP expression by Delta and Omicron variants, suggesting that ISR inhibition is a feature shared by all SARS-CoV-2 variants (Fig. 6E and F).

Discussion
Viral infections disrupt cellular homeostasis, which in turn can trigger stress responses.Here we show that SARS-CoV-2, via activation of the eIF2α-kinase PKR, activates the ISR.When investigating downstream effectors of the ISR, we found that the production of ATF4 and CHOP is inhibited in infected cells.Further, despite strong eIF2α phosphorylation and a reduction in cellular translational levels, SG prevalence remains mostly low in infected cells.When comparing the ancestral SARS-CoV-2 and the two variants Delta and we noted that Delta induces lower levels of p-PKR and p-eIF2α, and Omicron induces stronger SG formation, suggesting variant specific differences.
A key player of the ISR is eIF2 which, when phosphorylated at Ser51 in its alpha subunit, induces an arrest in cap-dependent translation (Anderson and Kedersha, 2002).Four kinases, each activated by a specific type of stress, are known to phosphorylate eIF2α and three of them (PKR, PERK and GCN2) are known to be activated during viral infections (Donnelly et al., 2013).Here, we observed strong phosphorylation of PKR, a kinase that detects double-stranded RNA and hence often is activated during viral infections (Dauber and Wolff, 2009).While many viruses have developed strategies to block its function, we detected high levels of p-eIF2α in SARS-CoV-2 infected cells, indicating PKR-mediated activation of the ISR.This is consistent with Li and colleagues, who reported phosphorylation of PKR and eIF2α in infected cells (Li et al., 2021).SARS-CoV mediated PKR and eIF2α phosphorylation have been described by Krähling et al. (2009).For SARS-CoV, it was shown that in addition to PKR, PERK significantly contributes to eIF2α phosphorylation in infected cells (Krähling et al., 2009).Coronavirus glycoproteins can induce rearrangements of the ER membrane, forming double-membraned vesicles, inducing ER stress and subsequent PERK activation (Fung et al., 2014;Hagemeijer et al., 2014).In line with this, another study reported PERK activation during early SARS-CoV-2 infection, showing that other eIF2α-kinases could also be involved in ISR activation (Rosa-Fernandes et al., 2021).
Phosphorylation of eIF2α and the subsequent reduction in global translational levels normally trigger the formation of SGs.Nevertheless, despite high levels of p-eIF2α and a strong translational shutdown, most cells infected with SARS-CoV-2 remained negative for SGs, showing that SARS-CoV-2 counteracts this cellular response.The few SG-positive cells detected showed a common pattern of low N protein levels, in line with previous findings that the SARS-CoV-2 N protein localizes to SGs through liquid-liquid phase separation and can dissolve SGs by binding to G3BP (Luo et al., 2021;Wang et al., 2021).This may also explain why most SGs were found during early ancestral SARS-CoV-2 infection before We observed that SARS-CoV-2 efficiently suppressed SG formation when stimulated with sodium arsenite or Poly(I:C), but upon challenge with ER stress or starvation infected cells showed an increased susceptibility towards SG formation.Why SGs form more frequently during ER stress and starvation is unclear but since both types of stress affect the cellular protein turnover, they might also influence N protein production or modification and therefore its capability to interact with G3BP.High levels of ATF4, the effector that regulates eIF2α dephosphorylation via production of GADD34, may partly explain why uninfected cells are free of SGs after treatment with thapsigargin.
As shown previously (Al-Beltagi et al., 2021), we noted that thapsigargin-treatment efficiently inhibited SARS-CoV-2 infection.Furthermore, both thapsigargin and sodium arsenite had a strong negative effect on viral replication in A549-hACE2 cells.In contrast, Poly(I:C), while reducing the infection rate, barely affected progeny virus production.
Poly(I:C) mimics viral infection and triggers activation of PKR as well as RIG-I (Palchetti et al., 2015).Since PKR is already highly activated during SARS-CoV-2 infection, additional Poly(I:C) treatment might not be able to activate PKR any further and the virus likely has developed mechanisms to cope with this kinase´s other antiviral effects.Moreover, SARS-CoV-2 was shown to suppress RIG-I activity and downstream IFN signaling (Chen et al., 2021;Oh and Shin, 2021).Thapsigargin and sodium arsenite also trigger additional pathways besides the ISR that can negatively affect viral replication including the unfolded protein response and mTOR signaling (Lindner et al., 2020;Bernstam and Nriagu, 2000).Lacking a targeted response against those pathways may Fig. 6.Suppression of ATF4 and CHOP production in SARS-CoV-2 infected A549-hACE2 cells.(A) Uninfected and ancestral strain-infected cells were treated with sodium arsenite (SA), thapsigargin (T) or remained untreated.At 24 hpi, cell lysates were collected and analyzed for expression of ATF4, CHOP and p-eIF2α via western blot.Calnexin was used as a loading control.(B) Quantification of (A).Levels of ATF4 and CHOP were normalized against cellular levels of calnexin.The data are represented as the fold change in relation to the protein levels of untreated, uninfected cells and are represented as mean ± SEM. n = 3. (C) Uninfected and ancestral strain-infected cells were treated with T and ISRIB (I) or remained untreated.At 24 hpi, cell lysates were collected and analyzed for expression of ATF4 via western blot.Calnexin was used as a loading control.(D) Quantification of (C).Levels of ATF4 and CHOP were normalized against cellular levels of calnexin.The data are represented as the fold change in relation to the protein levels of untreated, uninfected cells and are represented as mean ± SEM. n = 3. (E) Uninfected cells and cells infected with different SARS-CoV-2 variants were treated with T or remained untreated.At 24 hpi, cell lysates were collected and analyzed for expression of ATF4 via western blot.Calnexin was used as a loading control.(F) Quantification of (E).Levels of ATF4 and CHOP were normalized against cellular levels of calnexin.The data are represented as the fold change in relation to the protein levels of untreated, uninfected cells and are represented as mean ± SEM. n = 3. *, p < 0.05; **, p < 0.005; ***, p < 0.0005; ****, p < 0.0001.
W. Christ et al. explain why the virus is more sensitive towards these treatments.
An analysis of the translational activity during late-stage SARS-CoV-2 infection revealed a strong suppression of translation in infected cells.While the high levels of p-eIF2α present during the infection might suggest ISR activation as the mechanism behind the translational arrest, the fact that ISRIB treatment largely failed to restore translation indicated it is not.Instead, the bulk of the translational arrest might be mediated by nsp1, an important SARS-CoV-2 virulence factor known to suppress host gene expression by binding to the entry channel of ribosomes and blocking translation of host mRNAs while favoring viral mRNAs (Banerjee et al., 2020;Schubert et al., 2020;Thoms et al., 2020).Worth noting is, however, that ISRIB treatment reduced SG formation at early time points of SARS-CoV-2 infection, indicating that ISR activation plays a more significant role in triggering translational arrest early during the infection.It is also noticeable that while p-eIF2α levels of uninfected, sodium arsenite-treated A549-hACE2-cells and SARS-CoV-2 infected cells were similar, sodium arsenite had a stronger effect on translational levels than infection per se.
During ISR activation, the binding of p-eIF2α to eIF2B leads to a reduction in availability of the eIF2-GTP-tRNAiMet-ternary complex, which is an essential factor for translation initiation (Costa-Mattioli and Walter, 2020;Jackson et al., 2010;Kenner et al., 2019).It has been recently shown that the Beluga whale coronavirus SW1 employs a class IV ISR inhibitor which blocks the interaction between eIF2α and eIF2B to avert translational arrest in the presence of high p-eIF2α concentrations (Rabouw et al., 2020;Aloise et al., 2022).Possibly, SARS-CoV-2 could similarly utilize a class IV ISR inhibitor to assure the availability of sufficient ternary complex levels for its own protein production.In line with this, we found that levels of a key regulator of the ISR, the transcription factor ATF4, as well as of its effector protein CHOP, were both suppressed in SARS-CoV-2 infected cells.Both proteins are normally upregulated and preferentially translated during p-eIF2α mediated translational arrest (Pakos-Zebrucka et al., 2016).Under prolonged or intensive stress conditions ATF4 and CHOP promote apoptosis (Teske et al., 2013).which is a common mechanism for elimination of infected cells.Consequently, many viruses have developed anti-apoptosis mechanisms, for instance upregulating the anti-apoptotic factor BCL-2 and affecting DR5 expression (Solà-Riera et al., 2020;Du, 2009;Solà-Riera, 2019).Employing a class IV ISR inhibitor could therefore be a strategy to simultaneously avoid a lack of ternary complex as well as avert apoptosis during persistent ISR signaling.Since ISR activation is one of the host cells' ways to fight off viral infections, it is perhaps surprising that inhibition of the pathway did not enhance SARS-CoV-2 replication but instead lead to a slight decrease in viral titers.This suggests that SARS-CoV-2 infection copes so well with ISR activation that inhibiting the pathway either has no effect on viral replication or that it actually hampers it conceivably by benefiting host cell translation more than viral translation.
When comparing ancestral SARS-CoV-2 with the Delta and Omicron variants, we observed several variant-specific differences in regard to ISR activation and SG formation.All variants triggered ISR activation, but Delta induced significantly less phosphorylation of PKR and eIF2α than the ancestral strain and Omicron.It has been suggested that the SARS-CoV-2 N protein can inhibit PKR activation by sequestering dsRNA (Zheng et al., 2021;Aloise et al., 2022) and adaptations in the Delta N protein might make this process more efficient.Omicron on the other hand induces similar levels of p-PKR but increased levels of p-eIF2α, indicating stronger ISR activation, possibly through additional activation of PERK or GCN2.The high concentrations of p-eIF2α could partly explain why Omicron responds differently to ISR inhibition.ISRIB treatment restored translational levels more efficiently in Omicron infected cells than in cells infected with ancestral SARS-CoV-2 or Delta, suggesting that the ISR contributes more to inducing translational arrest than with the other variants.In contrast to ancestral SARS-CoV-2 and the Delta variant, Omicron strongly induced SG formation, especially at late time points of the infection.The elevated levels of p-eIF2α in Omicron-infected cells could contribute to this effect.It is possible that strong ISR activation and the following SG formation may partly explain why Omicron infection leads to lower progeny virus production than the ancestral strain or Delta in A549-hACE2 cells, however, this needs to be further studied before any clear conclusions can be drawn.
Intriguingly, unlike in ancestral SARS-CoV-2 or Delta-infected cells, almost no difference in N protein levels could be observed between Omicron-infected SG-negative and SG-positive cells.This suggests that the Omicron N protein might be less efficient in SG suppression than the N proteins of the other variants.The Omicron N protein contains several mutations within the intrinsically disordered region of its N-terminus, which contributes to the protein´s ability to induce phase separation, the process through which the N protein likely localizes to SGs (Zachrdla et al., 2022).Interestingly, we found that one of the two N protein bands usually detected in cells infected with ancestral or Delta SARS-CoV-2 via Western Blot is lacking in Omicron infected cells, which underlines the difference between the variants regarding their N protein expression patterns.Further supporting the possible involvement of the N protein, Barh and colleagues recently reported that the binding of the Omicron N protein to G3BP is weaker than that of ancestral SARS-CoV-2 or Delta N proteins (Barh et al., 2022).
To summarize, we report that while SARS-CoV-2 strongly activates the ISR through PKR leading to phosphorylation of eIF2α, the prevalence of SGs in infected cells remains low and no expression of stressresponsive proteins is observed.Treatments with ISRIB and different stress kinase inducers shows SARS-CoV-2 resistance especially towards PKR activation and indicates a balance where ISR activation resulting from SARS-CoV-2 infection appears to have no negative effect on the virus, but where overt stimulation of the stress kinases will result in inhibition of viral replication.Finally, we show variant-specific differences in SARS-CoV-2 activation of the ISR, particularly in that the Omicron variant causes significantly more formation of stress granules compared to the ancestral strain and Delta.

Immunofluorescence microscopy
Cells were fixed for 20 min with 4 % (vol/vol) paraformaldehyde in PBS.Before the staining, the cells were permeabilized by treatment with 0.5 % (vol/vol) triton X in PBS for 5 min and blocked with 0.5 % (wt/ vol) bovine serum albumin (BSA) in PBS for 30 min.Primary antibodies against the SARS nucleocapsid protein (Ziegler et al., 2005) (rabbit, 1:200), G3BP (abcam, ab181149) (mouse, 1:500) and puromycin (Merck, MABE343) (mouse, 1:500) were diluted in the blocking solution and incubated on the cells for 1 h.After washing, the samples were incubated for 1 h with the respective secondary antibodies, goat anti-mouse alexa-488 or goat anti-rabbit alexa-647, and DAPI diluted 1:1000 in blocking solution.Imaging was done on a Nikon A1R+ confocal microscope with NIS-Elements C software.The images are representative images from at least three independent experiments.For the manual quantifications, 100 to 200 cells on 5 to 10 random microscope frames were counted for each biological replicate.For the quantification of SGs in infected cells, only cells that were positive for the N protein were included in the counting.Measuring of N protein intensity by calculating the CTCF was performed with Fiji/ImageJ version 2.0.0 software.The acquisition settings were kept constant between all conditions to ensure comparability.

Treatments
To induce stress and SG formation, the cells were treated with 1 mM sodium arsenite (sigma) for 1 h, 5 mM thapsigargin (Merck; T9033) for 6 h, HBSS (+2 % HEPES + 2 % FBS) for 6 h to starve the cells.For treatment with Poly(I:C), cells were transfected with 2.5 mg Poly(I:C) using Lipofectamine LTX with PLUS reagent (Thermo Fisher,15,338,100).When analyzing the effect of different stress inducers on viral replication, the cells were treated with 0.1 mM sodium arsenite, 5 mM thapsigargin or HBSS (+2 % HEPES + 2 % FBS) for 6 h.No clear toxicity was observed for any of the conditions as evaluated by trypan blue staining (data not shown).When analyzing the effect of different stress inducers on infectivity, the cells were treated with 10 mM sodium arsenite, 0.5 mM thapsigargin or medium containing 50 % HBSS (+2 % HEPES + 2 % FBS) for 24 h or transfected with 0.25 mg Poly(I:C).For the puromycin labeling, the cells were treated with 20 mM puromycin for 10 min and washed twice with PBS before sample collection.Infected cells were treated with 200 nM ISRIB (Merck; SML0843) throughout the infection.No clear toxicity was observed for ISRIB treatment as evaluated by trypan blue staining (data not shown).For uninfected cells, 200 nM ISRIB was added together with the sodium arsenite.

End point dilution assay
Vero E6 cells in 96-well plates were treated with supernatants from infected cells in a dilution series of 10 − 1 to 10 − 8 with eight wells per dilution.After 5 days the cells were analyzed for cytopathic effect (CPE) with a light microscope to detect infection.Wells that were positive for CPE were marked and the number of 50 % tissue culture infectious doses (TCID 50 ) per milliliter was calculated using the Spearman-Karber method.

Statistical analysis
Error bars show standard deviations.t tests (Fig. 1 and suppl.

Fig. 1 .
Fig. 1.ISR activation and SG formation in ancestral SARS-CoV-2 infected cells.(A) eIF2α-phosphorylation in SARS-CoV-2 infected Vero E6 and A549-hACE2 cells.Uninfected and SARS-CoV-2 infected cells were treated with 1 mM sodium arsenite (SA), an ISR-inducing chemical, for 1 h, as positive control.Cell lysates were collected at 24 hpi and analyzed for levels of p-eIF2α.Calnexin was used as a loading control.(B) Protein levels and phosphorylation of PKR in SARS-CoV-2 infected A549-hACE2 cells.Cells were infected SARS-CoV-2.Cell lysates were collected at 24 hpi and analyzed for expression and phosphorylation levels of PKR.b-actin was used as a loading control.(C) Quantification of (B).Protein levels were normalized to cellular levels of b-actin.The data are presented as the fold change in relation to the protein levels of uninfected cells.Data are represented as mean ± SEM. n = 3. (D) G3BP1 staining in SARS-CoV-2 infected A549-hACE2 cells.Cells were infected with SARS-CoV-2 and fixed after 24 h.Uninfected cells that were either untreated or treated with 1 mM SA were used as a control.The cells were stained with antibodies against SARS-CoV-2 N protein (magenta) and G3BP1 (green) and with DAPI (blue).Imaging was performed with confocal microscopy (20X).Scale bars 20 mm.(E) Localization of the viral N protein within SGs.A549-hACE2 cells were fixed 24 h hpi and stained with antibodies against SARS-CoV-2 N protein (magenta) and G3BP1 (green) and with DAPI (blue).Arrows show colocalization of viral proteins with G3BP1.Imaging was performed with confocal microscopy (20X).Scale bars 20 mm.*, p < 0.05; **, p < 0.005.

Fig. 2 .
Fig. 2. The effect of different stress stimuli on SG formation during ancestral SARS-CoV-2 infection and on viral replication in Vero E6 cells.(A) Experimental setup for Figs.2B -G.Uninfected and ancestral SARS-CoV-2 infected cells were (B) treated with 1 mM sodium arsenite for 1 h, (C) transfected with 2.5 mg Poly(I:C) overnight, (D) treated with 5 mM thapsigargin for 5 h or (E) starved for 5 h and fixed at 24 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n ≥ 3. (F) Effect of different stress inducers on infectivity.SARS-CoV-2 infected cells were treated with sodium arsenite, Poly(I:C), thapsigargin or starved (see materials and methods) and fixed at 24 hpi They were stained for SARS-CoV-2 N protein.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells was determined and normalized against the infection rate of untreated cells.Data are represented as mean ± SEM. n = 3. (G) Effect of different stress inducers on viral replication.SARS-CoV-2 infected cells were treated with sodium arsenite, Poly(I:C), thapsigargin or starved (see materials and methods) and 6 h after treatment the viral titers were determined.Data are represented as mean ± SEM. n ≥ 3. *, p < 0.05; **, p < 0.005; ****, p < 0.0001.

Fig. 3 .
Fig. 3. ISR activation and SG formation during infection with different SARS-CoV-2 variants (A) Time kinetics of SG formation in SARS-CoV-2 infected Vero E6 cells.Cells infected with different SARS-CoV-2 variants were fixed at 6, 24 or 48 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n ≥ 3. (B) SG formation in SARS-CoV-2 infected A549-hACE2 cells.Cells infected with different SARS-CoV-2 variants were fixed at 24 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n ≥ 3. (C) SG formation after sodium arsenite challenge in SARS-CoV-2 infected Vero E6 and A549-hACE2 cells.Cells were infected with different SARS-CoV-2 variants and treated with 1 mM sodium arsenite for 1 h.Cells were fixed at 24 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n = 3. (D) Replication of different SARS-CoV-2 variants in A549-hACE2 cells.Cells were infected with different SARS-CoV-2 variants and at 24 hpi the viral titers in the supernatant were determined.Data are represented as mean ± SEM. n = 4. (E) eIF2αand PKR-phosphorylation in SARS-CoV-2 infected A549-hACE2 cells.Cells were infected with ancestral (A), Delta (D), or Omicron (O) SARS-CoV-2.Lysates were collected at 24 hpi and analyzed for levels of p-eIF2α and p-PKR.Calnexin was used as a loading control.(F) Quantification of (D).Protein levels were normalized to cellular levels of b-actin.The data are presented as the fold change in relation to the protein levels of cells infected with ancestral SARS-CoV-2.Data are represented as mean ± SEM. n = 3. *, p < 0.05; **, p < 0.005; ***, p < 0.0005; ****, p < 0.0001.
Fig. 3. ISR activation and SG formation during infection with different SARS-CoV-2 variants (A) Time kinetics of SG formation in SARS-CoV-2 infected Vero E6 cells.Cells infected with different SARS-CoV-2 variants were fixed at 6, 24 or 48 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n ≥ 3. (B) SG formation in SARS-CoV-2 infected A549-hACE2 cells.Cells infected with different SARS-CoV-2 variants were fixed at 24 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n ≥ 3. (C) SG formation after sodium arsenite challenge in SARS-CoV-2 infected Vero E6 and A549-hACE2 cells.Cells were infected with different SARS-CoV-2 variants and treated with 1 mM sodium arsenite for 1 h.Cells were fixed at 24 hpi.They were stained for SARS-CoV-2 N protein and G3BP.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells showing SGs was determined.Data are represented as mean ± SEM. n = 3. (D) Replication of different SARS-CoV-2 variants in A549-hACE2 cells.Cells were infected with different SARS-CoV-2 variants and at 24 hpi the viral titers in the supernatant were determined.Data are represented as mean ± SEM. n = 4. (E) eIF2αand PKR-phosphorylation in SARS-CoV-2 infected A549-hACE2 cells.Cells were infected with ancestral (A), Delta (D), or Omicron (O) SARS-CoV-2.Lysates were collected at 24 hpi and analyzed for levels of p-eIF2α and p-PKR.Calnexin was used as a loading control.(F) Quantification of (D).Protein levels were normalized to cellular levels of b-actin.The data are presented as the fold change in relation to the protein levels of cells infected with ancestral SARS-CoV-2.Data are represented as mean ± SEM. n = 3. *, p < 0.05; **, p < 0.005; ***, p < 0.0005; ****, p < 0.0001.

Fig. 4 .
Fig. 4. The effect of different stress stimuli on infectivity and replication of ancestral SARS-CoV-2, Delta, and Omicron.A) Effect of different stress inducers on infectivity.SARS-CoV-2 infected A549-hACE2 cells were treated with sodium arsenite, Poly(I:C), thapsigargin or starved as described in Fig. 2 and fixed at 24 hpi.They were stained for SARS-CoV-2 N protein.For the analysis using fluorescence microscopy, ≥5 images of a total of >100 cells were taken at random positions and the number of infected cells was determined and normalized against the infection rate of untreated cells.Data are represented as mean ± SEM. n = 3. (B) Effect of different stress inducers on viral replication.SARS-CoV-2 infected A549-hACE2 cells were treated with sodium arsenite, Poly(I:C), thapsigargin or starved as described in Fig. 2. 6h after treatment the viral titers were determined and normalized against the titers in untreated cells.Data are represented as mean ± SEM. n = 3. *, p < 0.05; **, p < 0.005; ***, p < 0.0005; ****, p < 0.0001.

Fig. 5 .
Fig. 5.The effect of ISR activation during SARS-CoV-2 infection on global translational levels and viral replication in A549-hACE2 cells.(A) Puromycin incorporation in A549-hACE2 cells infected with SARS-CoV-2 variants.Uninfected and infected cells were treated with sodium arsenite and/or ISRIB.Before sample collection at 24 hpi, the cells were treated with puromycin to visualize translational levels.(B) Quantification of (A).Puromycin levels were normalized to cellular levels of b-actin.The data are presented as the fold change in relation to the puromycin levels of untreated, uninfected cells and are represented as mean ± SEM. n ≥ 3. (C) ISRIB treatment decreases viral replication.SARS-CoV-2 infected cells were treated with ISRIB, and the viral titers of the supernatant collected at 24 hpi were determined.The data are represented as mean ± SEM. n = 3. *, p < 0.05; **, p < 0.005.