Coronaviruses are of veterinary and medical importance and include highly pathogenic zoonotic viruses, such as SARS-CoV and MERS-CoV. They are known to efficiently evade early innate immune responses, manifesting in almost negligible expression of type-I interferons (IFN-I). This evasion strategy suggests an evolutionary conserved viral function that has evolved to prevent RNA-based sensing of infection in vertebrate hosts. Here we show that the coronavirus endonuclease (EndoU) activity is key to prevent early induction of double-stranded RNA (dsRNA) host cell responses. Replication of EndoU-deficient coronaviruses is greatly attenuated in vivo and severely restricted in primary cells even during the early phase of the infection. In macrophages we found immediate induction of IFN-I expression and RNase L-mediated breakdown of ribosomal RNA. Accordingly, EndoU-deficient viruses can retain replication only in cells that are deficient in IFN-I expression or sensing, and in cells lacking both RNase L and PKR. Collectively our results demonstrate that the coronavirus EndoU efficiently prevents simultaneous activation of host cell dsRNA sensors, such as Mda5, OAS and PKR. The localization of the EndoU activity at the site of viral RNA synthesis–within the replicase complex—suggests that coronaviruses have evolved a viral RNA decay pathway to evade early innate and intrinsic antiviral host cell responses. RESULTS Severe attenuation of EndoU-deficient coronaviruses Based on biochemical and structural information on coronavirus EndoU active-site residues[ 17 , 27 , 28 ], we constructed EndoU-deficient mutants of HCoV-229E and MHV (HCoV-229EH250A and MHVH277A) and assessed their replication characteristics in vitro and in vivo ( Fig 1A ). Replication of MHVH277A was reduced in L929 cells, but peak titers almost reached those of wild-type MHV-A59, confirming that the coronavirus EndoU activity is dispensable for virus replication in vitro ( Fig 1B )[ 25 , 26 ]. In sharp contrast, compared to wild-type MHV-A59, the EndoU-deficient MHVH277A was severely attenuated in vivo ( Fig 1C ). MHVH277A replication was not detectable in spleen and liver of C57BL/6 mice at two days post intraperitoneal infection with 500 plaque-forming units (pfu), demonstrating that the EndoU activity is required for efficient replication and spread in vivo . Notably, replication and spread of MHVH277A was partly restored in mice deficient for the IFN-I receptor (IFNAR), with viral titers of MHVH277A in the spleen and liver of IFNAR-deficient mice that did not reach those of MHV-A59. Interestingly, concerning the role of Mda5 and TLR7, which are known as main cytoplasmic and endosomal PRRs for coronaviral RNA, respectively, MHVH277A replication was not restored in Mda5-deficient, TLR7-deficient, or Mda5- and TLR7-deficient mice. This phenotype clearly differs from that described for coronaviruses that lack ribose-2’-O methyl-transferase (OMT) activity [ 9 ]. Thus, in experiments reported previously, we found that replication of OMT-deficient MHV (MHVD130A) is restored in mice that are deficient for Mda5 and TLR7, suggesting that lack of ribose-2’-O methylation is tolerated if these two RNA sensors are absent. The lack of any detectable replication of the EndoU-deficient MHVH277A in Mda5- and TLR7-deficient mice therefore indicates that Mda5- and TLR7-mediated IFN-I expression may not exclusively restrict MHVH277A replication and that other mechanisms contribute to the observed attenuation of MHVH277A replication. 	 EndoU-deficiency results in increased IFN-β expression 	The severe attenuation of MHVH277A growth in vivo prompted us to assess MHVH277A and HCoV-229EH250A replication in primary target cells. As shown in Fig 2A , MHVH277A replication in primary murine embryonic fibroblasts (MEFs) was comparable to that of MHV-A59 until 9–12 hours post infection (h.p.i.), but was significantly restricted later during the infection. Moreover, replication of MHVH277A was even more severely reduced in bone marrow-derived murine macrophages, and accompanied by early induction of IFN-β expression ( Fig 2B and 2C ). Notably, levels of IFN-β mRNA were only transiently (6 to 12 h.p.i.) elevated in MHVH277A compared to MHV-A59 infected macrophages, and declined along with viral titers and viral RNA during the late phase of infection. Likewise, replication of the EndoU-deficient HCoV-229EH250A was severely restricted in human blood-derived macrophages ( Fig 2D ). We observed significantly elevated IFN-I expression in a panel of human macrophages derived from seven individual donors after infection with HCoV-229EH250A compared to wild-type HCoV-229E infection ( Fig 2D ), consistent with reduced viral replication.
Next, we addressed if, and to what extent, the growth defects observed for EndoU-deficient coronaviruses correlate with the induction of IFN-β expression. Mda5 has been described as the main RNA sensor of coronavirus infection in murine macrophages [ 29 ]. Compared to wild-type macrophages, IFN-β expression was reduced in Mda5-deficient macrophages following MHV H277A infection, as shown by qRT-PCR and IFN-β ELISA ( Fig 3A and 3D ). Surprisingly, and again in contrast to the phenotype of the OMT-deficient MHV D130A [ 9 ], MHV H277A replication was not restored in Mda5-deficient macrophages ( Fig 3A ). Similarly, although IFN-β expression was likewise reduced in MAVS-deficient macrophages, MHV H277A replication was not restored in MAVS-deficient macrophages ( Fig 3B and 3D ). Even in IRF3/IRF5/IRF7 (IRF3/5/7 -/- ) triple-knockout macrophages, that display an almost negligible induction of IFN-β expression, MHV H277A replication was not fully restored ( Fig 3C and 3D ). IFN-β protein assessed by IFN-β ELISA was below detection in all three macrophage genotypes (Mda5 -/- , MAVS -/- , IRF3/5/7 -/- ; Fig 3D ). These results indicate that MHV H277A replication is either highly sensitive to already marginal amounts of IFN-I, or that other antiviral host cell responses may contribute to the attenuation of MHV H277A . EndoU-deficient coronaviruses display a pronounced sensitivity to IFN-I treatment 	In order to address the sensitivity of EndoU-deficient coronaviruses to IFN-I, we first assessed MHVH277A replication in IFNAR-deficient macrophages. As shown in Fig 4A , MHVH277A replication was partially restored to levels that almost reached those of wild-type MHV-A59 replication. Importantly, IFN-β expression was elevated in IFNAR-deficient macrophages that had been infected with MHVH277A compared to those infected with wild-type MHV-A59, demonstrating that increased expression of IFN-β can be uncoupled from attenuation of MHVH277A ( Fig 4B ). We also noted that IFN-β expression was delayed in IFNAR-deficient macrophages following MHVH277A infection compared to wild-type macrophages. This observation is in agreement with previous reports that suggested a macrophage-specific autocrine IFN-β priming loop in wild-type macrophages enhances cytokine and chemokine expression following MHV infection [ 30 ]. 	The severe attenuation of MHVH277A and HCoV-229EH250A replication in wild-type murine and human macrophages, respectively, precluded the use of primary macrophages to assess the sensitivity of EndoU-deficient coronaviruses to IFN-I pre-treatment. Therefore, we infected murine L929 cells and human MRC5 lung fibroblasts with MHVH277A and HCoV-229EH250A, respectively, and applied different dosages of IFN-I for 4 hours prior to infection. Compared to wild type MHV and HCoV-229E, respectively, both EndoU-deficient viruses indeed displayed a pronounced sensitivity to IFN-I pre-treatment ( Fig 4C ). Remarkably, MHVH277A displayed a sensitivity to IFN-I treatment that is comparable to that of the highly IFN-I sensitive OMT-deficient mutant MHVD130A ( S1 Fig ). However, compared to the OMT-deficient mutant MHVD130A the phenotype of the EndoU-deficient MHVH277A differs mainly in the lack of restoration of replication under conditions with strongly reduced IFN-I expression (e.g. in Mda5-/- macrophages; Fig 3A ), suggesting that other, most likely IFN-I inducible, antiviral effector mechanisms account for restriction of MHVH277A replication. Collectively, these results demonstrate that the coronavirus EndoU activity plays a pivotal role in innate immune evasion in the context of the IFN-I system. EndoU-deficient coronaviruses induce activation of the OAS/RNase L pathway
Coronavirus EndoU-deficiency results in a pronounced sensitivity to IFN-I treatment that is comparable to that of the highly IFN-I sensitive OMT-deficient mutant MHVD130A. 	However, replication of MHVH277A was not restored in Mda5-deficient macrophages. This observation prompted us to consider that replication of EndoU-deficient coronaviruses may activate additional dsRNA-triggered antiviral pathways. We therefore assessed the integrity of ribosomal RNA (rRNA), a marker for the activation of the OAS/RNase L pathway [ 31 ], during MHVH277A infection in primary murine macrophages. Indeed, the breakdown of rRNA in MHVH277A infected wild-type macrophages was readily detectable as early as 6–12 h.p.i., thus coinciding with the induction of IFN-β expression during the early phase of the infection (compare Figs 5A and 2C ). This finding is highly surprising since MHV-A59 encodes a PDE activity that has been shown to degrade 2’,5’-oligoadenylate messenger molecules essential for RNase L activation [ 14 ]. However, the PDE activity was apparently not sufficient to prevent RNase L activation in macrophages that had been infected with EndoU-deficient MHVH277A. To exclude that the lack of EndoU activity may directly impact on viral RNA synthesis and lead to reduced levels of subgenomic mRNAs, we assessed the level of genomic and subgenomic mRNA2 (encoding the PDE activity) by qRT-PCR. As shown in S2 Fig , genomic RNA and subgenomic mRNA2 were equally reduced in MHVH277A infected wild-type macrophages, suggesting that the lack of EndoU activity does not result in selective reduction of subgenomic mRNAs. Importantly, while breakdown of rRNA was also readily detectable in Mda5- and MAVS-deficient macrophages, rRNA remained intact in RNase L-deficient macrophages, demonstrating that infection of EndoU-deficient MHVH277A indeed results in the activation of the OAS-RNase L pathway and subsequent degradation of rRNA ( Fig 5A ; S2 Fig ).
A breakdown of rRNA was not detected in MHV H277A –infected IFNAR-deficient macrophages ( Fig 5A ) concurring with partial restoration of MHV H277A replication in these cells. Accordingly, and as previously published [ 32 , 33 ], the degree of RNase L activation correlates with levels of OAS expression and we noted indeed reduced baseline expression of OAS 1a, 2 and 3 in IFNAR-deficient compared to wild-type C57BL/6 macrophages ( S3 Fig ). Likewise, we did not observe breakdown of rRNA in L929 cells ( Fig 5B ). We therefore assessed the levels of OAS 1a, 2, and 3 expression in L929 with or without IFN-I treatment (12.5 U). As expected, expression of IFN-β was elevated in MHV H277A –, but not in MHV-A59-infected L929 cells, irrespectively of IFN-I pre-treatment ( S4 Fig ). Importantly however, expression of OAS 1a, 2 and 3 in L929 cells was significantly elevated following IFN-I treatment ( S4 Fig ), and accordingly, rRNA breakdown was readily detectable in IFN-I treated L929 cells that had been infected with MHV H277A ( Fig 5B ). This data provide evidence for a functional link between the observed pronounced IFN-I sensitivity of MHV H277A and restriction of MHV H277A replication by the OAS/RNase L pathway.
Discussion 	Coronaviruses have long been known to efficiently evade host innate immune responses during the early phase of the infection. However, a defined viral function accounting for the apparent lack of efficient sensing of coronavirus infection has remained elusive. Here we show that the highly conserved coronavirus EndoU activity within the viral RTC plays a major role in providing a first line of innate immune evasion during the early phase of coronavirus infection. We show that at least three dsRNA-triggered antiviral pathways are involved in restricting replication of EndoU-deficient coronaviruses ( Fig 8 ). First, infection with EndoU-deficient MHV and HCoV-229E results in rapid Mda5-mediated induction of IFN-β expression. Second, we observe breakdown of ribosomal RNA indicative of activation of the OAS/RNase L pathway that temporally coincides with IFN-β expression. Third, we show that efficient restriction of EndoU-deficient coronaviruses is furthermore dependent on PKR since restoration of EndoU-deficient MHVH277A replication required the absence of both, PKR and RNase L. Our data suggest that direct restriction of replication of EndoU-deficient coronaviruses is mediated by RNase L-mediated RNA degradation and inhibition of host cell translation through activation of PKR. In contrast, the effect of IFN-I appears to be indirect through the induction of ISG expression, that includes OAS/RNase L and PKR. Whether other ISGs may contribute to the restriction of EndoU-deficient coronavirus replication remains to be determined. Finally, we show that MHVH277A replication is associated with increased dsRNA levels during the early phase of the infection, providing a likely PAMP for the observed simultaneous activation of multiple cytoplasmic dsRNA-sensors in cells infected with EndoU-deficient coronaviruses. Reference & Source Information: https://journals.plos.org/ Read More on: