SARS-CoV-2 nucleocapsid protein inhibits the PKR-mediated integrated stress response through RNA-binding domain N2b

The nucleocapsid protein N of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) enwraps and condenses the viral genome for packaging but is also an antagonist of the innate antiviral defense. It suppresses the integrated stress response (ISR), purportedly by interacting with stress granule (SG) assembly factors G3BP1 and 2, and inhibits type I interferon responses. To elucidate its mode of action, we systematically deleted and over-expressed distinct regions and domains. We show that N via domain N2b blocks PKR-mediated ISR activation, as measured by suppression of ISR-induced translational arrest and SG formation. N2b mutations that prevent dsRNA binding abrogate these activities also when introduced in the intact N protein. Substitutions reported to block post-translation modifications of N or its interaction with G3BP1/2 did not have a detectable additive effect. In an encephalomyocarditis virus-based infection model, N2b - but not a derivative defective in RNA binding—prevented PKR activation, inhibited β-interferon expression and promoted virus replication. Apparently, SARS-CoV-2 N inhibits innate immunity by sequestering dsRNA to prevent activation of PKR and RIG-I-like receptors. Similar observations were made for the N protein of human coronavirus 229E, suggesting that this may be a general trait conserved among members of other orthocoronavirus (sub)genera.


Introduction
Vertebrate cells are provided with a diversity of interconnected sensors and effectors to timely detect and counter viral infection.Particularly dsRNA, an inevitable product of RNA and DNA virus replication, triggers a vigorous intracellular antiviral response [1].For example, binding of dsRNA by 2'-5'oligoadenylate (2-5A) synthase (OAS) leads to enzyme activation and production of 2-5A, which in turn activates RNase L to stall the synthesis of viral proteins through non-specific RNA degradation [2][3][4][5][6][7].Another dsRNA-activated pathway, the integrated stress response (ISR) with proteinase kinase R (PKR) as key sensor, entails inhibition of translation initiation and ultimately cell death.Upon dsRNA binding, PKR dimerizes, autophosphorylates and then proceeds to phosphorylate the alpha subunit of eukaryotic initiation factor eIF2, turning it into competitive inhibitor of guanine nucleotide exchange factor eIF2b.In consequence, the production of eIF2-GTP-(Met)tRNA i Met ternary complex is downregulated, recognition of the initiation codon is blocked and cap-dependent translation-initiation prevented [8][9][10][11].Polysome dissociation, ensuing activation of the stress response, results in an excess of stalled 48S preinitiation complexes, which are stored in cytoplasmic membraneless organelles called stress granules (SGs) [12][13][14].SGs are dynamic ribonucleoprotein assemblies that are formed through liquid-liquid phase separation with RNA binding protein Ras GTPase-activating protein-binding proteins 1 and 2 (G3BP1 and -2) functioning both as a molecular switches and main protein scaffolds together with T-cell-restricted intracellular antigen 1 (TIA1) and TIA1-related protein (TIAR) [15][16][17][18][19].The SGs thus serve as deposits from which mRNAs, poised for translation through association with critical components of the translation machinery (40S ribosomal subunits, eIF4E, eIF4G, eIF4A, eIF4B, Poly(A) binding protein, eIF3, and eIF2), can be rapidly retrieved.SGs, however, are also considered a coordinating hub for the activation of other antiviral defense mechanisms like those of RIG-I-like receptors (RLRs).Indeed, RIG-I and MDA5, RLR-regulating PKR-activating protein (PACT), RLR-modulating ubiquitin ligases TRIM25 and TRAF2, and polyubiquitin chains are all recruited to ISR-induced SGs [20][21][22][23].
The antiviral mechanisms elicited by dsRNA are highly effective, even to such an extent that all known mammalian viruses code for one or more antagonists [11].Coronaviruses, positive-stranded RNA viruses of exceptional genetic complexity, code for a universally conserved endonuclease (EndoU) that efficiently prevents simultaneous activation of host cell dsRNA sensors OAS, PKR and MDA5 through dsRNA decay [24][25][26].Illustrating the importance of EndoU, mutants defective for EndoU are severely attenuated in vivo, and their replication in cultured primary macrophages is restricted presumably due to high basal expression levels of host sensors in these cells [25].EndoU is derived by proteolytic processing of a large replicase polyprotein pp1ab, translated from the incoming viral genome, and essential to evade early innate and intrinsic antiviral host cell responses.Apparently, however, EndoU may not be sufficient to suppress dsRNA-mediated antiviral activities during later stages of the viral life cycle.
To express the open reading frames (ORFs) downstream of the replicase gene, CoVs produce a 3' co-terminal nested set of subgenomic mRNAs from which the structural proteins are translated in addition to a variety of so-called accessory non-structural proteins.Some of the latter also have been shown to counteract dsRNA-mediated antiviral host responses.For example, the ns4a protein of Middle East Respiratory Syndrome coronavirus (MERS-CoV) prevents

PLOS PATHOGENS
SARS-CoV-2 N protein inhibits the PKR-mediated stress response via RNA-binding domain N2b PKR-mediated stress by sequestering dsRNA [27,28], whereas the MERS-CoV ns4b protein is a phosphodiesterase (PDE) and antagonizes the OAS-RNase L pathway by enzymatically degrading 2-5A activators [29].Non-related PDEs, ns2 proteins, are found in members of the subgenus Embecovirus, including human coronavirus OC43 [30].Finally, a specific inhibitor of the ISR was found in gammacoronaviruses of cetaceans.The beluga whale coronavirus accessory protein 10 (BWCoV acP10) blocks p-eIF2-eIF2B association to allow continued formation of the ternary complex and unabated global translation even at high p-eIF2 levels that would otherwise cause translational arrest [31].
Coronavirus N proteins have a modular structure with N-terminal and C-terminal RNAbinding domains, called N1b and N2b respectively, bounded by largely disordered regions.Here, we took a reductionistic approach involving systematic deletion and individual transient expression of the different regions and domains.We show that the N2b domain is critical and sufficient to counter the ISR.Single site mutations in N2b that block dsRNA binding prevent PKR activation.This activity is not affected by posttranslational modifications elsewhere in the N protein that are known to regulate RNA binding nor dependent on physical interaction with G3BP1 and G2BP2.Using the encephalomyocarditis virus as a surrogate infection system we show that N2b domain prevents PKR-mediated activation of the ISR and suppresses IFNβ expression also in virus-infected cells.Our findings suggest that in addition to a function in replication and genome packaging, SARS-CoV-2 N functions as a classical antagonist of dsRNA-induced host defense.

SARS-CoV-2 N inhibits PKR-induced ISR
Transfection of cells with specific expression plasmids like pEGFP-N3 triggers the ISR through dsRNA-mediated PKR activation, resulting in translation arrest and the formation of SGs (Fig 1A) [27,[41][42][43].The dsRNA arises from spurious bidirectional transcription of plasmid sequences [44] and can be readily detected in transfected cells with dsRNA-specific antibodies [27].This phenomenon allows for a convenient method to identify potential viral ISR antagonists by transiently expressing these proteins genetically fused to enhanced green fluorescent protein (EGFP) [27,31].The expression levels of these fusion proteins, as judged by fluorescence microscopy, can then be compared to that of EGFP alone as an indicator for translation arrest and the prevention thereof.In addition, SG formation can be assessed by immunofluorescence analysis (IFA) by staining the transfected cells for established SG markers like G3BP1, G3BP2, eIF3 and TIA-1 (S1A Fig) .This procedure previously allowed us to identify several stress antagonists including MERS-CoV 4a and Beluga whale coronavirus AcP10 [27,31].
SARS-CoV-2-infected Vero E6-TMPRRS2 cells, identified by detection of dsRNA, were virtually devoid of SGs, suggestive of virus-induced suppression of the ISR (S2 Fig) .While probing SARS-CoV-2 proteins for a potential role in ISR inhibition, we noted that in transfected wildtype (wt) HeLa cells, the expression levels of the SARS-CoV-2 N-EGFP fusion protein were strongly increased as compared to the EGFP control.This pattern of enhanced expression in a sizeable population of transfected cells was similar to that observed for established ISR

PLOS PATHOGENS
SARS-CoV-2 N protein inhibits the PKR-mediated stress response via RNA-binding domain N2b  S3).Finally, Western blot analysis confirmed that N-EGFP was expressed to similar levels in HeLa wt and PKR KO cells.Levels of EGFP, however, were strongly increased (about 7-fold) in HeLa PKR KO  The N protein of another betacoronavirus, MERS-CoV, was previously noted to suppress SG formation too [35].Interestingly, we observed inhibition of eIF2α phosphorylation, translational arrest and SG formation also in cells expressing the N protein of human alphacoronavirus 229E (HCoV-229E) ( Figs 1E, 1F and S3).Apparently, N-mediated suppression of the ISR is not unique to betacoronaviruses but may well be a general trait conserved among members of other orthocoronavirus (sub)genera.

Domain N2b is essential and sufficient for suppression of PKR-induced SG formation
The N proteins from the four CoV genera share only 27 to 30% sequence identity but are conserved in their modular organization (Fig 2A) ([45,46] for a review see [47]).They are comprised of two ordered domains, the N-terminal domain (N1b also called NTD) and the Cterminal domain (N2b aka CTD) (in SARS-CoV-2 N, residues 49-175 and 248-365, respectively).N1b and N2b are separated and flanked by segments, predicted to be at least partially disordered (N1a, N2a and N3).N1b and N2b have been implicated in RNA binding and, in case of N2b, also in N dimerization.Indeed, in infected cells and upon heterologous expression, CoV N proteins form homodimers that, in turn, assemble into tetramers mediated by N3 [48].The N2a central spacer contains a serine-and arginine-rich (SR) region, immediately downstream of N1b, the regulated phosphorylation of which is deemed important for N Western-blot analysis for PKR, phosphorylated PKR (p-PKR), EGFP-fused proteins and β-actin.HeLa wt and PKR KO cells, mock-transfected or transfected with indicated plasmid were lysed at 24 hr.Of note, the larger yield of EGFP versus N-EGFP in either cell type is counter intuitive but can be explained from the fact that (i) this is an ensemble measurement (for the total transfected cell population) versus the analysis of individual cells by fluorescence microscopy and (ii) basal expression levels of the codon optimized EGFP prior to ISR activation will exceed those the N-EGFP fusion protein, which is non-codon optimized and three times larger in size (see also S4   function during different stages of the CoV replication cycle [49][50][51].The primary function of N is in virus assembly.It condenses newly produced gRNA into helical nucleocapsids, apparently with N2b controlling target selectivity [52,53], and then drives envelopment by binding to the viral membrane protein M via N3.However, N has several functions auxiliary to virion morphogenesis.At the very start of the infectious cycle, N is essential for the initiation of infection by the incoming gRNA through association of SR with the cytosol-exposed ubiquitin1 domain of replication organelle pore protein nsP3 [51,54].Moreover, purportedly relevant to ISR suppression, N binds to G3BP1 and 2 through segment N1a [38,40,55].
To determine the molecular basis for ISR suppression, we constructed a library of N mutants with partially disordered regions and domains either systematically deleted or expressed in isolation.As shown in Fig 2 , suppression was lost upon deletion of subdomain N2b.Moreover, expression of N2b in isolation caused a reduction in SGs to an extent similar to that of full-length N.The data therefore suggested that ISR inhibition and SG suppression as observed for the intact N protein is mediated at least in part by N2b.

N2b mutations that disrupt dsRNA binding abrogate suppression of SG formation
N2b binds both single and double stranded oligonucleotides, whether RNA or DNA [47,56,57].Hence, a possible mechanism for N2b to prevent PKR-induced activation of the ISR is by sequestering dsRNA.Crystal structural analysis revealed that N2b homodimers form a rectangular slab with wide faces of 45 Å × 35 Å in dimension [56].One is comprised of an interlaced inter-molecular four-stranded β-sheet and predominantly negatively charged, the other of two α-helical regions separated by a shallow central, positively charged groove thought to be the oligonucleotide binding site [47,56,58,59].Within the groove, there are several conserved positively charged residues with their surface-exposed side chains seemingly poised for interaction with nucleic acid, particularly Lys To test whether Lys 257 and Lys 261 are involved in RNA binding and whether such binding is important to counteract the ISR and suppress SG formation, they were replaced by Ala, separately and in combination.Indeed, dsRNA binding was either significantly reduced or lost beyond detection upon substitution of Lys Importantly, also in the intact N protein, Lys 261 Ala substitution abrogated inhibition of translational arrest as well as SG formation and so did the Lys 257 Ala mutation (Fig 3C and  3D).It is unknown why the latter mutation exerts a stronger effect in the context of the fulllength protein than in N2b.The combined data, however, do show that Lys 261 and Lys 257 are required for nucleic acid binding by N2b and suggest that this capacity to bind RNA is essential for SG suppression.Moreover, the observation that also for the intact N protein SG suppression was reduced to background levels by single site Lys 261 Ala and Lys 257 Ala substitutions suggests that inhibition of PKR-induced ISR by SARS-CoV-2 N critically relies on N2b and

PLOS PATHOGENS
SARS-CoV-2 N protein inhibits the PKR-mediated stress response via RNA-binding domain N2b

N2b-mediated inhibition of PKR-induced SG formation is not affected by posttranslational modifications of N
Previous studies identified N as a multifunctional protein involved in different aspects of the viral replication cycle beyond viral assembly.Its activity apparently is regulated by posttranscriptional modifications such as differential phosphorylation of the SR domain, which alters RNA binding affinity [61,62], and acetylation at Lys 375 , reportedly essential for liquid-liquid phase separation of N-RNA ribonucleoprotein complexes [63].Arginine methylation of SARS-CoV-2 nucleocapsid protein at Arg 95 and Arg 177 was reported to regulate RNA binding and its ability to suppress stress granule formation [33].
To test the importance of these posttranscriptional modifications on N2b-mediated inhibition of SG formation, we performed site-directed mutagenesis in the context of the intact N-EGFP fusion protein.Suppression of SG formation was not affected by Ala substitution of Arg 95 or Lys 375 indicating that methylation and acetylation of N is not essential (Fig 4A and 4B).
An N-EGFP derivative defective in phosphorylation of the SR segment was constructed by replacing 13 out of the 14 Ser residues by Ala.The resulting mutant, N-(13S>A)-EGFP, displayed a cellular distribution distinctively different from that of the parental wildtype protein.
Instead of an even distribution throughout the cytoplasm, the mutant protein clustered in what looked like large aggregates.These local deposits were enriched for G3BP2 but mostly devoid of eIF3 (Figs 4A and S5) and differed in size and appearance from typical SGs.Importantly, however, the expression levels of the N-13S>A mutant were like those of wildtype N, as judged by EGFP fluorescence intensity observed in IFA, and consistently higher than those of EGFP alone (Fig 4A).We interpret the findings to indicate that the SR mutations may cause the N protein to aggregate but they do not affect inhibition of ISR-induced translational arrest.The results confirm those of our systematic deletion analysis (Fig 2).

N2b-mediated inhibition of SG formation is not affected by disruption of the G3BP binding motif
A FxFG motif within N1a (residues 15-18), required for association with G3BP1 and 2 [40], was recently proposed to rewire the G3BP interactome to disrupt stress granules [38].To investigate a possible role of N-G3BP interaction in suppressing PKR-induced SGs, we mutated the N1a FxFG motif through Ala substitution of key residue Phe 17 (mut 1A) or by a combination of Ala substitions: Arg 14 and Ile 15 (mut 2A), Ile 15 , Phe 17 and Gly 18 (mut 3A), or Ile 15 , Thr 16 , Phe 17 and Gly 18 (mut 4A).In each case, binding to endogenous G3BP1 was no longer detectable by pull down assay (    4A and S6), the expression levels of these mutants were strongly reduced as compared to those of wt N and the amounts of coprecipitated G3BP1 were below the detection level (Fig 4C, 'HeLa wt').Unperturbed G3BP binding, however, was evident, when assessed in HeLa PKR KO cells (lane marked 'N2b mut') and lost upon concomitant mutation of the N1a FxFG G3BP binding motif (lane marked 'N1a+N2b mut') (Fig 4C).The findings indicate that SARS-CoV-2 N-G3BP interaction is not required for inhibition of the ISR, at least not upon induction of the ISR via the PKR signaling pathway under the conditions applied.Moreover, this interaction is also not sufficient to prevent PKR-induced formation of SGs or to promote their disassembly.
To study whether G3BP-binding might still affect SG formation, HeLa PKR KO cells were transfected to express wildtype N or mutant derivatives and subjected to arsenite/heme-regulated inhibitor kinase (HRI)-induced stress.This approach allowed us to study the importance of N-G3BP1 interaction more directly, i.e. without transfection induced PKR activation and N-mediated inhibition of PKR as complicating factors.Under these conditions, wt N-EGFP still inhibited SG formation as compared to EGFP alone albeit to a modest extent (Fig 5).This phenomenon may be ascribed to G3BP sequestration, because the observed reduction in the number of SG-producing cells was largely abrogated by mutations in the N1a ФXFG G3BP binding motif (

Suppression of PKR-induced ISR by coronavirus N proteins is mediated predominantly by N2b
To corroborate our observations, we measured N-mediated rescue of ISR-induced translational arrest also by flow cytometry.To this end, we used a cotransfection assay with red fluorescence protein (RFP) expressed from vector pcDNA-RFP serving as reporter [27].Cotransfection with ISR-inducing plasmid pEGFP-N3 strongly inhibited RFP production (Fig 6 Intriguingly, however, RFP expression levels were still higher than those observed for EGFP (p-value 0.0014) or N2b-K 257 A/K 261 A (p-value 0.0195).The data suggest that suppression of the PKR-induced ISR by coronavirus N proteins is mediated predominantly by the N2b domain though not exclusively.Apparently, N counteracts translational arrest also through other domains via alternative mechanisms, but this contribution only becomes detectable upon N2b inactivation.Saliently, however, disruption of N-G3BP interactions through substitutions in segment N1a did not have an additive effect when tested in combination with the N2b mutations (

PLOS PATHOGENS
SARS-CoV-2 N protein inhibits the PKR-mediated stress response via RNA-binding domain N2b

N2b-mediated suppression of the ISR and type I interferon response in virus-infected cells
Whereas the results provide conclusive evidence for ISR suppression upon transient expression of N in transfected cells, the question arises whether this phenomenon also occurs during natural infection and whether it is relevant for evasion of the innate host immune response.Unfortunately, the essential functions of N at multiple stages of the coronavirus life cycle, both during early replication as well was in genome packaging during virion assembly, precludes a  straightforward reverse genetics approach.Thus, the construction of recombinant SARS-CoV-2 mutants with an N protein defective in N2b RNA binding was deemed a nonviable option.We therefore resorted to a well-established alternative model based on recombinant encephalomyocarditis virus mutant EMCV-L zn , in which the autologous ISR antagonist (the leader protein L) is inactivated.The mutant virus is no longer able to counteract the ISR [64][65][66] and hence provides a convenient platform to identify and characterize ISR antagonists of other viruses.
Our initial experiments were performed with an EMCV mutant that coded for a chimeric polyprotein with N2b at its N-terminus, downstream of the first five residues of EMCV L zn that are important for efficient IRES-mediated translation [67] and two additional residues encoded by an engineered XhoI site.This virus, however, failed to suppress the ISR.We noted that in the fusion protein, two negatively charged residues, E 6 (from L) and E 8 (encoded by the XhoI sequence) are immediately upstream of N-terminal N2b residues K 248 and K 249 (residues 9 and 10 of the chimera) and proximal to critical N2b residues K 257 and K 261 S8 Fig) .Arguing that this might affect N2b RNA binding, we introduced into the EMCV-L zn genome an N-terminally extended N2b with the N2a region serving as a spacer (Figs 7A and S8).Indeed, in cells, infected with the resulting virus EMCV-L zn -N2aN2b wt , SG formation was suppressed to 30% of that caused by EMCV-Lzn, i.e. to levels similar to those reported for EMCV derivatives with L replaced by established PKR inhibitor MERS-CoV ns4a (Fig 7B and 7C).In accordance, in cells infected with the N2aN2b wt virus, levels of phosphorylated PKR were consistently low, comparable to those in wildtype EMCV-infected cells and, as calculated from band densities, 4.5 to 6-fold lower than those in cells infected with EMCV-L zn-.Thus, in its activity N2aN2b wt mimicked MERS-CoV ns4a which prevents PKR activation by sequestering dsRNA [27] but differed from AcP10 which inhibits the ISR at a level downstream of PKR [31].Like MERS-CoV ns4a and AcP10, N2aN2b restored replication efficacy of EMCV-L zn to near wildtype levels as based on expression of the viral capsid proteins (Fig 7D).
To test whether N2aN2b, like ns4a, also suppresses the type I interferon response, we tested for phosphorylation of IRF3 by Western blot analysis and measured the levels of IFN-β mRNA by qRT-PCR.Indeed, in cells infected with EMCV-L zn -N2aN2b wt IRF3 phosphorylation was inhibited (Fig 7D ) and IFN-β expression was reduced as compared to EMCV-L zn infected cells to levels observed in cells infected with EMCV wt or EMCV-L zn -MERS-ns4a (Fig 7E).In contrast, in cells infected with EMCV-L zn -N2aN2b-K 257 A+K 261 A, encoding an inactive N2b, there was no detectable suppression of SG formation, PKR and IRF3 were phosphorylated, and viral replication was delayed to a similar if not larger extent than in EMCV-L zn -infected cells (Fig 7B and 7D).EMCV-L zn -N2aN2b-K 257 A+K 261 A also lost the capacity to suppress the type I IFN response as detected by qRT-PCR, although this was noticeable only at 8 hr post infection (Fig 7E).This may be attributed to the considerable delay in virus replication and a consequential late onset of IFN induction.
We conclude that SARS-CoV-2 N2b functions as an antagonist of the ISR in EMCVinfected cells and can functionally replace the EMCV L protein by acting as a classical PKR antagonist and preventing activation of the ISR by sequestering dsRNA.This mode of action differs from that of beluga whale CoV AcP10 and Aichivirus L, which counteract the ISR by salvaging global translation through competitive inhibition of p-eIF2-eIF2B association [31].The data lend support to the notion that N through N2b may inhibit the ISR and SG formation also in the context of SARS-CoV-2-infected cells.

Discussion
To successfully establish infection, viruses must subdue the intracellular host defense long enough to complete their replication cycle and vigorously enough to curb intercellular immune signaling.The detection of dsRNA, an inevitable byproduct of both RNA and DNA virus replication is central to activation of the innate antiviral defense.The activation of RLRs must be averted that would otherwise culminate, through intra-and intercellular signaling, in the production of type I interferons and other proinflammatory cytokines.Also, while host cell protein synthesis should best be inhibited and kept at a minimum, efficient unperturbed production of viral proteins must be ensured by preventing translational suppression due to activation of the OAS/RNaseL pathway and PKR-dependent induction of the ISR.Recent publications implicated SARS-CoV-2 N protein as an inhibitor of SG formation and an RLR antagonist [33][34][35][36]68,69].Here we confirmed and extended these observations using a well-established plasmid-based expression system [27,31] to induce PKR-mediated ISR activation and to assess inhibition thereof by EGFP-tagged ISR antagonists.We conclusively demonstrate that SARS-CoV-2 N is endowed with the ability to inhibit PKR activation, translational arrest and ensuing SG formation.Moreover, we showed that the N protein of human alphacoronavirus 229E (subgenus Duvinacovirus) can also prevent translational arrest and suppresses SG formation to a similar extent and via a similar mode of action.These findings suggest that this property may be widely conserved among members of different orthocoronavirus (sub)genera.
To study the mechanism of ISR inhibition, we performed a systematic deletion and sitedirected mutational analysis of SARS-CoV-N.We found that inhibition of SG formation was lost upon deletion of N2b in accordance with findings of Zheng et al. [35].However, we crucially extend their observations by showing that N2b when expressed in isolation, is sufficient to impair transfection-induced ISR-mediated translational shut off as well as the formation of SGs.Apparently, N2b counteracts the ISR by binding dsRNA.Ala substitution of conserved Lys residues (Lys 257 and/or Lys 261 ) in the putative RNA-binding cleft of N2b that abolished dsRNA binding also abolished ISR inhibition.Importantly, these substitutions also caused the intact N proteins of SARS-CoV-2 and HCoV-229E to lose their activity as ISR antagonists.The observations thus strongly support a model in which coronavirus N proteins through their N2b domain prevent activation of the ISR by PKR by sequestering dsRNA, like the other viral PKR inhibitors IAV ns1 and MERS-CoV ns4a that were used as controls throughout.
The N protein plays a crucial role during the very early steps of the CoV infection cycle [56,[70][71][72] and is indispensable for virus morphogenesis.Evidently, this impedes a direct Dunnett post hoc test).(D) Western-blot analysis for PKR, phosphorylated PKR (p-PKR), IRF3, phosphorylated IRF3 (p-IRF3), EMCV capsid proteins and β-actin.HeLa cells, (mock-)infected with recombinant EMCVs at MOI 10, were lysed at 6 or 8 hpi.Images are representative of three independent experiments.(E) Suppression of type I interferon response in EMCV-infected cells by N2a2b but not N2aN2b-Mut.Infected HeLa cells (MOI 10) were lysed at 6 or 8 hpi.Total RNA was analyzed by RT-qPCR for IFN-β and actin.IFN-β levels were calculated as fold induction compared to levels in mock-infected cells, after correction for actin mRNA levels, and normalized with EMCV-L zn IFN-β levels set at 100%.Data represent means of three independent experiments.Standard deviations indicated by error bars; statistical significance compared to the results for EMCV-L zn or EMCV-L zn -N2aN2b infected cells calculated by two-way ANOVA with the Dunnett post hoc test; ** p<0.01; *** p<0.001; **** p<0.0001.https://doi.org/10.1371/journal.ppat.1011582.g007mutational reverse genetics approach to study the occurrence and relevance of N-mediated inhibition of the ISR in CoV-infected cells.However, by using a mutant EMCV as a platform, we showed that an extended protein N2aN2b can functionally replace the autologous enterovirus L protein and can salvage the EMCV replication defect caused by L inactivation, like we previously showed for MERS-CoV ns4a and BWCoV AcP10 [27,31].Like MERS-CoV ns4a, but unlike BWCoV AcP10 -which blocks the ISR downstream of PKR-, N2aN2b prevented PKR phosphorylation apparently via N2b-mediated RNA binding.The findings provide proof of principle that N2b can promote viral replication by preventing PKR and ensuing ISR activation, supporting the notion that N through its N2b domain may do so as well during CoV infection.
Adding relevance to our observations, selection for enhanced innate immune escape apparently favored SARS-CoV-2 variants with increased N expression [73].Moreover, a threenucleotide change in Gamma also present in Alpha and Omicron variants of concern created a new (cryptic) transcription regulating sequence to drive the synthesis of a novel subgenomic mRNA species from which a truncated N-protein, N*, may be translated corresponding with the C-terminus of the protein and initiating at in frame Met 210 [73,74].If N* is indeed expressed, our findings strongly support a role in dsRNA sequestration and innate antagonism.
The native coronavirus N proteins are subject to extensive posttranslational modifications in domains and segments other than N2b.Some have been implicated in liquid-liquid phase separation [63,75], RNA binding [33], SG formation and inhibition of innate immunity [33,63], prompting the question whether these modifications promote or hinder its function as an ISR antagonist.Reportedly, methylation of Arg 95 in the SARS-CoV-2 N1b segment by host protein arginine methyltransferases is required for RNA binding and for the inhibition of arsenite-induced formation of G3BP1-containing SGs [33].In our hands, however, an N mutant with Arg 95 Lys substitution, described to block N-mediated suppression of SGs [33], still prevented PKR-induced ISR-associated translational repression to identical extent as the wildtype protein and likewise prevented SG formation.In another recent study, Wang et al. [63] reported suppression of MAVS signaling by intact SARS-CoV-2 N protein but not by an N derivative from which N2b had been deleted.They attributed this to N-induced and N2bdependent liquid-liquid phase separation (LLPS) subject to acetylation of N3 residue Lys 375 .In accordance, we find that in EMCV-infected cells, N2aN2b, like MERS4a, suppresses β-interferon (IFN-β) expression and show that this activity is blocked by mutations that abolish N2bmediated RNA binding.In apparent contrast to the observations of Wang et al. [63], however, we find that acetylation of Lys 375 is not essential for RNA binding by N2b and that Lys 375 Arg substitution does not detectably affect inhibition of the PKR-induced ISR by the intact N protein.Our findings suggest that N, through the autonomous activity of the N2b domain, acts not only as a classical PKR inhibitor to thwart ISR activation but may also prevent RLR activation and thus block induction of the type I IFN response, whether directly by shielding dsRNA or indirectly by preventing SGs to act as platform for immune signaling.The interaction between N and SG scaffold proteins G3BP1 and -2 mediated through the N1a ФxFG motif [33,[37][38][39][40], was recently concluded to be the main determinant in SG disassembly [38].Our current findings would seem to be at odds with this view, but in accordance with observations by these authors, we did observe that functional disruption of the critical ФxFG motif abrogated inhibition of SG formation in arsenite-treated HeLa PKR KO cells.Our findings, however, would suggest that sequestration of the G3BPs to prevent assembly of SG or to promote their disassembly requires high expression levels of N. Importantly, N proteins with intact N1a domain but with mutations in N2b that abolish RNA binding failed to prevent SG formation in HeLa wt cells.Also, the 229E N protein lacks a ФxFG motif, yet inhibits formation of SGs via the same N2b-dependent mechanism as SARS-CoV-2 N. The findings may be reconciled, however, by assuming that sarbecovirus N proteins have been evolutionary selected to interfere with G3BP function, possibly even beyond the formation and function of SGs [76], through distinct mechanisms at multiple levels during different stages of the viral life cycle, mediated by different domains and regulated by distinct posttranslational modifications, and perhaps subject to protein distribution and availability of viral genomes for encapsidation.Such subtleties may be missed in over-expression experiments and overshadowed by the robust inhibition of the PKR-triggered ISR by N2b.Hence, in coronavirus-infected cells, N1amediated G3BP-N association [38,40,55] and N2b-mediated dsRNA binding [36] may well act in concert to hamper SG assembly and to impede recruitment and SG-facilitated activation of RLRs and PKR [20][21][22][23].Indeed, this would be consistent with observations by others [35,36] of SARS-CoV-2 N suppressing SG-associated RLR activation and inhibiting induction of type I IFNs by targeting G3BP1.The relative importance of dsRNA shielding, G3BP recruitment and other activities of N for suppression of innate immunity during coronavirus infection clearly deserves further study.

Plasmids for eukaryotic and prokaryotic expression
Eukaryotic expression plasmids were constructed by cloning PCR-amplified sequences, flanked by NheI and BamHI restriction sites, into NheI/BamHI digested vector pEGFP-N3 (ClonTech) such that the encoded viral proteins are C-terminally fused to enhanced green fluorescent protein (EGFP).pcDNA-RFP was purchased from Addgene.
All constructs were sequenced to confirm integrity.There were no major differences in transfection efficiency (S9 Fig)

Immunofluorescence assay
Cells were seeded onto 12 mm glass cover slips in 24-well plates (Corning Costar) at 5x10 4 cells, grown for 24 hr and transfected with 500 ng total DNA/well using Lipofectamine2000 (Invitrogen).At 24 hr post transfection, cells were either left untreated or treated with 500 μM sodium arsenite (NaAsO 2 , Riedel-de Hae ¨n) diluted in DMEM for 30 min at 37˚C and subsequently fixed in PBS + 3.7% paraformaldehyde (PFA).Vero E6-TMPRSS2 cells were infected with SARS-CoV-2 Wuhan (D614G) or Omicron BA.1 variants at a multiplicity of infection (MOI) of 5 TCID 50 /cell and HeLa-R19 cells were infected with recombinant EMCV viruses at an MOI of 10 TCID 50 /cell and incubated for times indicated in the text.Cells were fixed with paraformaldehyde (3.7% in PBS), incubated with PBS + 0.1% glycine for 10 min, permeabilized with 0.1% Triton X-100/PBS+s for 10 min at RT and blocked in blocking buffer (PBS + 1% BSA + 0.1% Tween-20) for 30min in a dark humidified chamber at 37˚C.The cells were then incubated in blocking buffer containing mouse anti-dsRNA (1:1000; English & Scientific Consulting), goat anti-eIF3η (1:200, SantaCruz) and rabbit anti-G3BP2 (1:200; Bethyl Laboratories) for 1h at room temperature (RT).After washing with PBS+0.1%Tween-20, cells were incubated with secondary antibody donkey anti-mouse Cy2 (1:100; The Jackson Laboratory), donkey anti-goat Alexa594 (1:200; Invitrogen), donkey anti-rabbit Alexa647 (1:200; Invitrogen) in block buffer for 1h at RT.The cells were then washed three times with PBS+0.1% Tween-20, once with distilled water and mounted on glass microscope slides in ProLong Diamond Antifade (Invitrogen) mounting medium.Cells were examined by conventional widefield (Olympus) and confocal immunofluorescence microscopy (Nikon A1R) in most cases also in a blinded fashion by a second independent observer.

Flowcytometry analysis
HeLa cells were seeded in a 12-well cluster (10 5 cells/well) and, after overnight incubation, transfected with the indicated plasmids (500 ng well; 250 ng/plasmid) using Lipofectamine 2000 (Invitrogen).At 24 hr post transfection, cells were either left untreated or treated with 500 μM sodium arsenite (NaAsO 2 , Riedel-de Hae ¨n) diluted in DMEM for 30 min at 37˚C.Cells were detached with trypsin, washed once PBS, and fixed in PBS, 3.7% PFA for 10 min.To stain for p-eIF2α, cells were permeabilized in ice-cold MeOH for 10min, washed twice in FACS buffer (PBS + 0.02% Na-azide and 1% BSA) and subsequently incubated with the primary antibody rabbit anti-p-eIF2α (1:100, Abcam) in FACS buffer for 45 min.The samples were washed twice in FACS buffer and incubated in the secondary antibody goat anti-rabbit Alexa 594 (1:100, Invitrogen) in FACS buffer for 45 min.Cells were washed twice with FACS buffer and stored at 4˚C.Fluorescence intensity was recorded with a CytoFLEX LX flow

PLOS PATHOGENS
SARS-CoV-2 N protein inhibits the PKR-mediated stress response via RNA-binding domain N2b cytometer (Beckman Coulter) and the data were analyzed by FlowJo v10 software (BD Biosciences).Samples were gated for live single cell populations and then gated for negative, low (L) and high (H) RFP expressing cells.For bar graphs [H/L] ratios were normalized with the ratio calculated for the relevant wildtype protein set at 100%.

Electrophoretic Mobility Shift Assay (EMSA)
To purify recombinant N2b proteins, E. coli BL21 cells (Sigma-Aldrich), transformed with pGEX2T-Hisx6-N2b or its mutated derivatives, were grown in LB medium containing ampicillin (50 μg/ml) at 37˚C until the optical density at 600 nm (OD 600nm ) reached 0.3.The temperature was then reduced to 18˚C, and when the OD 600nm reached 0.5, protein expression was induced with 0.5 mM IPTG.Following expression for 16 hrs, the cells were harvested by centrifugation and resuspended in lysis buffer (100 mM Tris-HCl pH = 8, 300 mM NaCl, 10 mM imidazole, 0.1% Triton X-100, 5% glycerol) complemented with lysozyme (0.25 mg/ml; Merk) and cOmplete Protease Inhibitor (Roche).The samples were sonicated and centrifuged at 15,000g for 45 min at 4˚C to pellet cell debris.The cleared lysates were passed through a 0.45-μm filter and incubated with 1 ml of Ni-NTA (nitrilotriacetic acid) resin (Thermo Scientific) at 4˚C overnight on a roller.The beads were washed twice with wash buffer (100 mM Tris-HCl (pH 8.0), 300 mM NaCl, 10mM Imidazole) and then eluted with a buffer consisting of 100 mM Tris-HCl, 300 mM NaCl, and 500 mM imidazole (pH 8).The eluted proteins were subjected to dialysis in 20 mM Tris-HCl, 150mM NaCl and flash-frozen in 20-μl aliquots.
For the EMSA, 1 μM of a 32-mer single stranded RNA oligonucleotides corresponding to SARS-CoV-2 (GenBank accession no.MZ558051.1)sequence 5 0 -CGAGGCCACGCGGAGU ACGAUCGAGGGUACAG-3 0 and a scrambled version thereof, 5'-GGCACGGAGUAUACCGG ACGAGCGGAACGGCU-3', each were mixed 1: 1 with their respective complimentary RNA oligonucleotides in denaturing buffer (10 mM NaH 2 PO 4 ·H 2 O, 50 mM NaCl, 1 mM EDTA, 0.01% NaN 3, pH 7.4 at 25˚C), incubated at 90˚C for 4 min and then allowed to anneal at RT for 30 min.For the ssRNA sample preparation, the oligonucleotide was diluted in denaturing buffer, incubated at 90˚C for 30 sec to destroy possible secondary structures and rapidly cooled on ice.Proteins were diluted in protein buffer (20 mM Tris-HCl (pH = 8), 150 mM NaCl) to a concentration of 100 ng/μl and mixed in 10X, 20X or 40X fold molar excess with 10 ng ssRNA or 10 ng dsRNA in binding buffer (20 mM Tris-HCl, 100 mM NaCl, 1 mM EDTA, 1mM TCEP, 0.02% Tween-20, pH 7.0 at 25˚C) in 5 μl reaction volumes.The samples were incubated for 30 min on ice, then supplemented with glycerol to a final concentration of 5% (v/v) and separated in ultrathin (10 ml; 75 x 50 mm) RNase-free 1% agarose gels in 0.5x TB buffer (45 mM Tris base, 45 mM boric acid, pH 8.2-8.5) at 200V.RNAs were stained by incubating the gels in 50 ml 2X Invitrogen SYBR Gold nucleic acid gel stain (ThermoFisher) in 0.5x TB buffer, diluted from a 10.000X concentrate, for 15 min under agitation and de-stained for 21 min with 0.5 X TB buffer refreshing the buffer 3 times.Stained RNA was visualized with the Gel Doc System (BioRad).

Co-immunoprecipitation (Co-IP) assay
HeLa R19 wt and PKR KO cells were seeded in 6-well clusters (4x10 5 cells/well) and, after a 16 hr recovery, transfected with the indicated plasmids.At 24 hrs post transfection, cells were washed once in PBS, released by trypsin and incubated in ice-cold lysis buffer (Tris-HCl pH 8.0, 50mM, NaCl 150mM, EDTA 1mM, NP40 1%, cOmplete Mini Protease Inhibitor Cocktail (Roche) and phosphatase inhibitor PhosSTOP (Roche)) for 30 min on ice.The cell lysates were cleared in a microcentrifuge for 10 min at 12000 rpm at 4˚C.Supernatants were harvested and incubated with 25 μl of preequilibrated GFP-Trap agarose beads (ChromoTek) for 1h at 4˚C with end-over-end rotation.Beads were then washed 3 times with wash buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA).Bound proteins were eluted in 80 μl 2x SDS-sample buffer, separated by SDS-PAGE in reducing 8% polyacrylamide gels and analyzed by Western blotting.

PLOS PATHOGENS
SARS-CoV-2 N protein inhibits the PKR-mediated stress response via RNA-binding domain N2b (40˚C for 10 sec).The experiments were carried out in triplex for each data point.The relative quantification of the IFN-β gene expression was determined using the 2 -ΔΔCt method [78], then the relative IFN-β mRNA levels were normalized to the EMCV-zn IFN-β mRNA level set as 100.

Fig 1 .
Fig 1. SARS-CoV-2 N inhibits PKR-induced ISR, preventing translation arrest and SG formation in HeLa and A549 cells.(A) Schematic representation of the PKR-induced ISR activation upon the pEGFP-N3 transfection in eukaryotic cells.Created with BioRender.com.(B) HeLa and A549 cells were transfected with expression vector pEGFP-N3 (EGFP) or pEGFP-N3 derivatives encoding SARS-CoV-2 N, MERS-4a, IAV-NS1 or BwCoV-AcP10 genetically fused to EGFP.Induction of the ISR via dsRNA-mediated PKR activation or suppression thereof was assessed by comparing EGFP fluorescence intensity and SG formation as detected by immunofluorescence staining for G3BP2 (see also S1A Fig).(left panel) Transfected HeLa cells; (immune)fluorescence microscopy images, representative results.Scale bars: 50 μm.Settings of image acquisition (laser intensity, exposure time) and processing conditions used throughout were chosen to avoid over-exposure in high-expressing cells.Note that under these conditions, EGFP expression is too low to be detected in a sizeable population of stressed and non-stressed cells (see cells (Fig 1D, right panel; S4 Fig).The combined findings identify SARS-CoV-2 N as an ISR antagonist that prevents translational arrest and ensuing SG formation by acting as a PKR inhibitor.

also
S1B Fig).(right panel) Bar graphs for Hela and A549 cells showing the percentage of EGFP expressing cells containing G3BP2-positive SGs.The results are representative of three independent experiments counting >200 cells per sample.Standard deviation indicated by error bars; *** p = 0.001, **** p< 0.001, ns = not significant (one-way ANOVA with Dunnett post hoc test).(C) HeLa PKR KO cells transfected as in B. Representative (immuno)fluorescence microscopy images are shown.(D) Fig). (E) Inhibition of SG formation and translational arrest by pEGFP-N3-expressed N proteins of MERS-CoV and HCoV-229E.(Immuno)fluorescence analysis (left panel), quantitative representation of the results and statistical analysis as in B. (F) Western-blot analysis for eIF2α, phosphorylated eIF2α (p-eIF2α), EGFP-fused proteins and β-actin.HeLa cells were transfected to express EGFP or EGFP-tagged N proteins of SARS-CoV-2 and HCoV 229E.Mock-transfected cells were either left untreated (mock) or treated with sodium arsenite (+Arsenite) to induce eIF2α phosphorylation.https://doi.org/10.1371/journal.ppat.1011582.g001PLOS PATHOGENS SARS-CoV-2 N protein inhibits the PKR-mediated stress response via RNA-binding domain N2b

Fig 2 .
Fig 2. Nucleocapsid domain N2b suppresses translational arrest and SG formation.(A) Schematic representation of truncated derivatives of SARS-CoV-2 N protein fused to EGFP (left panel).The proteins were overexpressed from pEGFP-N3-based vectors in HeLa cells and percentages of SG-positive cells were determined as in Fig 1A (right panel).The data are representative of three independent experiments with more than 200 cells counted for the presence of G3BP2-positive SGs per individual sample.Standard deviation indicated by error bars.For a statistical analysis of the results, see S1 Table.(B)Representative results from (immuno)fluorescence analysis of HeLa cells transfected to express the SARS-CoV-2 N protein derivatives.Scale bar: 50 μm.
Fig 2. Nucleocapsid domain N2b suppresses translational arrest and SG formation.(A) Schematic representation of truncated derivatives of SARS-CoV-2 N protein fused to EGFP (left panel).The proteins were overexpressed from pEGFP-N3-based vectors in HeLa cells and percentages of SG-positive cells were determined as in Fig 1A (right panel).The data are representative of three independent experiments with more than 200 cells counted for the presence of G3BP2-positive SGs per individual sample.Standard deviation indicated by error bars.For a statistical analysis of the results, see S1 Table.(B)Representative results from (immuno)fluorescence analysis of HeLa cells transfected to express the SARS-CoV-2 N protein derivatives.Scale bar: 50 μm.https://doi.org/10.1371/journal.ppat.1011582.g002 257 and Lys 261 (Fig3A).The binding by bacterially-expressed N2b of synthetic RNA oligonucleotides, whether single or double-stranded, SARS-CoV-2 derived (highly conserved 3' UTR stem-loop II motif, s2m;[60]) or a scrambled version thereof, can be readily demonstrated by electrophoretic mobility assay (EMSA) (Fig3B).

Fig 3 .
Fig 3. N2b mutations that disrupt dsRNA binding abrogate suppression of translational arrest and SG formation.(A) (left panel) Surface representation and (middle panel) cartoon representation of the SARS-CoV-2 N2b dimer (PDB ID code 7C22 [59] with monomers in light blue (chain A) and wheat (chain B).Side chains of mutated charged/polar residues labeled and shown in sticks (Chain A: Lys 257 and Lys 261 ; Chain B: Q 272 , Q 289 , R 276 and R 293 ).(right panel) N2b dimer surface representation colored according to calculated charge distributions, displaying the positively-charged central RNA-binding groove.Critical residues Lys 257 (single asterisk) and Lys 261 (double asterisk) marked for both monomers.Top view images at a forward 45˚tilt; generated with UCSF ChimeraX version 1.6.1 [79].(B) Electrophoretic mobility shift assays (EMSA).(left panel) EMSA with bacterially expressed N2b domain and single (ss) and double-stranded (ds) RNA oligonucleotides, designed after SARS-CoV-2 stem-loop II motif (s2m).Assays were performed with N2b in 10-or 20-fold molar excess.Non-bound RNA was included as a control (np, no protein).(middle panel) The effect of N2b amino acid substitutions on binding of s2m dsRNA or (right panel) a scrambled Fig 4C) yet suppression of SG formation was not affected (Figs 4A, 4B and S6).Conversely, N proteins with N2b mutations to abrogate dsRNA binding no longer blocked the ISR (Figs 3 and 4B and 4C) nor inhibited SG formation, even though N1a was still intact to bind G3BP1 to a similar extent as wildtype N (Fig 4C; for version thereof (right).EMSAs performed with N2b and mutant derivatives in 20-fold molar excess (middle and right panels).(C-D) Mutational analysis of SARS-CoV-2 N2b, full-length SARS-CoV-2 N and full-length HCoV-229E N. Select surface-exposed charged residues were substituted by Ala either individually or in combination as indicated and the effect on IRS-induced translational arrest (C) and SG formation (D) was analyzed in HeLa cells as in Fig 1.For a statistical analysis of the results, see S2 Table.

Fig 4 .
Fig 4. N-mediated suppression of SG formation and ISR-induced translational arrest is not affected by posttranslational modifications or G3BP1 interaction.(A) The effect on N-mediated ISR inhibition by mutations introduced to prevent posttranslational modifications (K 375 Q, K 375 R, R 95 K), to disrupt the G3BP1/2 binding site (1A-4A mutants) or to prevent phosphorylation of the SR element (13S>A) was tested in HeLa cells by (immune)fluorescence analysis as in Fig 1. Scale bar: 50 μm.(B).Quantification of the results in (A) representative of three independent experiments, with >200 cells counted per Fig 5B).SG inhibition seemed dependent on N-EGFP expression levels.The effect was more pronounced in cells with high EGFP fluorescence intensity (Fig 5B and 5C and S3 Table).Conversely, N2b-dependent loss of PKR-induced SGs in wildtype HeLa cells was already observed at very low levels of N (Fig 1B).
, top; see also S7 Fig).RFP expression was rescued in cells co-expressing N2b but not by its RNA-binding deficient derivative N2b-K 257 A/K 261 A. RFP expression was also rescued by the intact N proteins of either SARS-CoV-2 or HCoV-229E and to similar levels by SARS-CoV-2 N mutants defective for G3BP1/2 binding (N-R 14 A/I 15 A) or N1b methylation (N-R 95 K) (Fig 6).In contrast, inactivation of N2b through the K 257 A/K 261 A double mutation in the intact N protein of either SARS-CoV-2 or HCoV-229E significantly reduced RFP expression (Fig 6).

Fig 5 .
Fig 5.The effect of N-G3BP interaction on arsenite-induced SGs in HeLa PKR KO cells.HeLa PKR KO cells were transfected to express EGFP, SARS-CoV-2 N wt, N mutants defective in G3BP binding (1A-4A mut), N-K 257 A+K 261 A (N2b mut) or a mutant 'N1a + N2b mut' with substitutions in both the G3BP binding site (2A mut) and N2b (K 257 A+K 261 A).At 24 hrs post transfection, cells were sodium arsenite-treated to induce HRI-mediated ISR with ensuing formation of SGs.(A) Representative (immune)fluorescence images.Scale bar: 50 μm.(B) Quantification of the results based on three independent experiments by counting all cells with detectable EGFP fluorescence or (C) highly expressing cells exclusively.For the statistical analysis, see S3 Table.

Fig 6 .Fig 7 .
Fig 6.Quantitative assessment of N-mediated rescue of ISR-induced translational arrest.HeLa cells were transfected to express EGFP, SARS-CoV-2 N2b, the full-length N proteins of SARS-CoV-2 and HCoV-229E, and mutants thereof from pEGFP-N3-based expression vectors to induce PKR-activated ISR.The capacity of these proteins to rescue translational repression of red fluorescent protein (RFP) in trans or lack thereof was measured by flow cytometry at 24 hr post transfection (left panels) and fluorescence microscopy (see S7 Fig).Representative flow cytometry histograms shown.The transfected cells were divided into non-RFP-expressing, low (L) RFP-expressing and high (H) RFP-expressing populations (see dashed lines in histograms).For each mutant protein, the [H/L] ratio was calculated and normalized, with those of the corresponding wildtype proteins set at 100%.The bar graphs show mean values with standard deviations based on three independent experiments (unpaired t-test; ****, P<0.001; ns, nonsignificant) (right).https://doi.org/10.1371/journal.ppat.1011582.g006 SGs calculated from three biological triplicates, showing the decrease in % as compared to wt N or the mutant N1a+N2b N. Ordinary One-way ANOVA, Dunnett's multiple comparison test.(TIF)