key: cord-0808273-og4k3fc7
authors: Konno, Yoriyuki; Kimura, Izumi; Uriu, Keiya; Fukushi, Masaya; Irie, Takashi; Koyanagi, Yoshio; Sauter, Daniel; Gifford, Robert J.; Nakagawa, So; Sato, Kei
title: SARS-CoV-2 ORF3b is a potent interferon antagonist whose activity is increased by a naturally occurring elongation variant
date: 2020-09-04
journal: Cell Rep
DOI: 10.1016/j.celrep.2020.108185
sha: 133e3ef3150027918a36e5cfa5a018d1dfc7e57f
doc_id: 808273
cord_uid: og4k3fc7

One of the features distinguishing SARS-CoV-2 from its more pathogenic counterpart SARS-CoV is the presence of premature stop codons in its ORF3b gene. Here, we show that SARS-CoV-2 ORF3b is a potent interferon antagonist, suppressing the induction of type I interferon more efficiently than its SARS-CoV ortholog. Phylogenetic analyses and functional assays reveal that SARS-CoV-2-related viruses from bats and pangolins also encode truncated ORF3b gene products with strong anti-interferon activity. Furthermore, analyses of approximately 17,000 SARS-CoV-2 sequences identify a natural variant, in which a longer ORF3b reading frame was reconstituted. This variant was isolated from two patients with severe disease and further increased the ability of ORF3b to suppress interferon induction. Thus, our findings not only help to explain the poor interferon response in COVID-19 patients, but also describe the emergence of natural SARS-CoV-2 quasispecies with an extended ORF3b gene that may potentially affect COVID-19 pathogenesis.

the SARS-CoV-2 outbreak has originated from cross-species coronavirus 82 transmission from these mammals to humans, the exact origin remains to be 83 determined (Andersen et al., 2020). 84 One prominent feature that distinguishes COVID-19 from SARS in terms 85 of immune responses is the poor induction of a type I interferon (IFN-I In this study, we therefore aimed to characterize the viral factor(s) determining 91

immune activation upon SARS-CoV-2 infection. We particularly focused on 92 differences in putative viral IFN-I antagonists and revealed that the ORF3b gene 93 products of SARS-CoV-2 and SARS-CoV not only differ considerably in their length, 94

but also in their ability to antagonize type I IFN. Furthermore, we demonstrate that 95 the potent anti-IFN-I activity of SARS-CoV-2 ORF3b is also found in related viruses 96 from bats and pangolins. Mutational analyses revealed that the length of the 97 C-terminus determines the efficacy of IFN antagonism by ORF3b. Finally, we 98 describe a natural SARS-CoV-2 variant with further increased ORF3b-mediated 99

anti-IFN-I activity that emerged during the current COVID-19 pandemic. 100 J o u r n a l P r e -p r o o f

To determine virological differences between SARS-CoV-2 and SARS-CoV, we set 103 out to compare the sequences of diverse Sarbecoviruses. Consistent with recent 104

reports (Lam et al., 2020; Zhou et al., 2020c), Sarbecoviruses clustered into two 105 groups, SARS-CoV-2-related and SARS-CoV-related viruses (Figures 1A and S1 ; 106 the sequences used are listed in Table S1 ( Figure 1B) , we hypothesized that the antagonistic activity of ORF3b against IFN-I 120 also differs between these two viruses. To test this hypothesis, we monitored 121 human IFNB1 promoter activity in the presence of ORF3b of SARS-CoV-2 122 (Wuhan-Hu-1) and SARS-CoV (Tor2) using a luciferase reporter assay. The 123

influenza A virus (IAV) non-structural protein 1 (NS1) served as positive control 124 (Garcia-Sastre et al., 1998; Krug et al., 2003) . As shown in Figure 1C , all three viral 125 proteins dose-dependently suppressed the activation of the IFNB1 promoter upon 126

Sendai virus (SeV) infection. Notably, the antagonistic activity of SARS-CoV-2 127

ORF3b was slightly, but significantly higher than that of SARS-CoV ORF3b ( Figure  128 1C, bottom (Figures 1A and 2A) . Since the lengths of ORF3b 141 proteins in SARS-CoV-2-related viruses including those from bats and pangolins 142

were on average shorter than those from SARS-CoV and related viruses ( Figure  143 1B), we next analyzed the variation of the ORF3b length in diverse Sarbecoviruses. 144

As shown in Figure 2B , the vast majority of SARS- the length of ORF3b is highly variable in SARS-CoV-related bat viruses ( Figure 2B  152 and Table S2 ). Only 2 out of the 54 ORF3b proteins of SARS-CoV-related bat 153 viruses (3.7%) are 154 amino acids in length, 50% of them express a 114 amino 154 acid ORF3b ( Figure 2B and Table S2 ). Similarly, only 174 five out of the ten ORF3b proteins of SARS-CoV-related viruses from bats (Rs4231, 175 YNLF34C, Shaanxi2011, Rm1 and F46) exhibited significant anti-IFN-I effects, and 176 all these ORF3b proteins were shorter than 114 amino acids ( Figure 2C ). Although 177 three additional ORF3b proteins of SARS-CoV-related viruses from bats (HKU3-2, 178 GX2013 and Yunnan2011) were shorter than 39 amino acids in length, they did not 179 exhibit anti-IFN-I activity, most likely because of their poor expression and/or 180 stability (Figures 2C and S2B) . Altogether, these findings suggest that the 181 C-terminal region (residues 115-154) attenuate the anti-IFN-I activity of ORF3b. at its C-terminus ("L115+NLS") ( Figure 2E) . Furthermore, we also attached the 198 c-Myc NLS to the C-terminus of Rs4231 ORF3b ( Figure 2E ). As expected, 199 SARS-CoV Tor2 ORF3b L115* as well as Rs4231 wild-type (WT) mainly localized 200 to the cytosol, while the two mutants harboring the c-Myc NLS were localized to 201 similar levels in both the cytosol and the nucleus ( Figure 2F ). Reporter assays 202

showed that the SARS-CoV Tor2 L115* mutant exhibits significantly higher 203

anti-IFN-I activity than WT SARS-CoV Tor2 ORF3b although both are expressed at 204 similar levels ( Figure 2G) . Moreover, the anti-IFN-I activity of both Tor2 L115* and 205 Rs4231 ORF3b was attenuated by the addition of an NLS (Figure 2G ), suggesting 206 that cytosolic localization of ORF3b is important to exhibit anti-IFN-I activity. 207

Consistent with the biochemical assays ( Figures 2D and 2F) , immunofluorescence 208 microscopy showed that WT SARS-CoV-2 ORF3b (Wuhan-Hu-1) as well as the 209 SARS-CoV ORF3b L115* mutant are mainly localized in the cytosol, while WT 210 SARS-CoV ORF3b (Tor2) and the Tor2 L115+NLS mutant reside in both the cytosol 211 and the nucleus ( Figure 2H ). In parallel, we monitored the subcellular localization of 212 IRF3, since this transcription factor is a key regulator of IFNB1 expression 213

[reviewed in (Park and Iwasaki, 2020)] that has previously been shown to be 214 ORF3b and the SARS-CoV ORF3b L115* mutant, but less so by WT SARS-CoV 217

ORF3b and its L115+NLS mutant ( Figure 2H ). Collectively, these findings 218

demonstrate that the C-terminal region of SARS-CoV ORF3b attenuates its 219

anti-IFN-I activity by impairing its ability to prevent the translocation of IRF3 into the 220 nucleus. CoV-GLUE webtool (http://cov-glue.cvr.gla.ac.uk) revealed that the ORF3b gene is 254 highly conserved (Table S4 and Figure S2D) . Notably, however, we detected two 255 viral sequences (GISAID accession IDs: EPI_ISL_422564 and EPI_ISL_422565), in 256 which the ORF3b gene was extended due to the loss of the first premature stop 257 codon (*23Q) (Figures 3B, bottom and S3C and Table S4 ). (Figure 3B , bottom; see also Figure S2D ). IFNβ reporter assays 270 revealed that the Ecuador variant ORF3b exhibits significantly higher anti-IFN-I 271 activity than the parental SARS-CoV-2 ORF3b ( Figure 3D ). 272

Since we found that SARS-CoV-2 ORF3b hampers the nuclear 273 translocation of IRF3 (Figure 2H) , we investigated the ability of WT SARS-CoV-2 274

ORF3b, the Ecuador variant ORF3b, as well as SARS-CoV ORF3b and IAV NS1 to 275 suppress IRF3-driven gene expression. Luciferase reporter assays revealed that 276 the inhibitory activity of WT SARS-CoV-2 ORF3b was significantly higher than that 277 of SARS-CoV ORF3b ( Figure 3E) . Importantly, the Ecuador variant ORF3b was 278 even more effective and suppressed IRF3-driven gene expression as efficiently as 279 IAV NS1 ( Figure 3E) for recombination of ORF3b between the lineages of SARS-CoV-2 and SARS-CoV. 310

Notably, phenotypic differences in the ability of ORF3b to suppress IFN-I responses 311

may also be associated with the likelihood of successful zoonotic transmission of 312

Sarbecoviruses to humans since many IFN-stimulated genes are antagonized in a 313 species-specific manner. While more than 50 SARS-CoV-related viruses were The full-length sequences (~30,000 bp) of SARS-CoV-2 (Wuhan-Hu-1 as a 613

representative), SARS-CoV-2-related viruses from bats (n=4) and pangolins (n=4), 614 SARS-CoV (n=190), SARS-CoV-related viruses from civets (n=3) and bats (n=54), 615

and outgroup viruses (n=2; BM48-31 and BtKY72) were analyzed. Accession 616 number, strain name, and host of each virus are indicated for each branch. Note 617 that the branches including SARS-CoV (n=190) and SARS-CoV-related viruses 618 from civets (n=3) were collapsed for better visualization. The uncollapsed tree is 619

shown in Figure S1 , and the sequences used are summarized in Table S1 . independent experiments is shown. Note that ORF3b and NS1 were run on 644 separate blots for better visualization. Figure S2A shows them on the same blot 645 with high and low exposure. For qRT-PCR, the expression levels of endogenous 646 IFNB1 and GAPDH were quantified. For the luciferase assay (C) and real-time 647 QUANTIFICATION AND STATISTICAL ANALYSIS 910

Data analyses were performed using Prism 7 (GraphPad Software). The data are 911

presented as averages ± SEM. Statistically significant differences were determined 912 by Student's t test. Statistical details can be found directly in the figures or in the 913 corresponding figure legends. 914 

mean values of three independent experiments with SEM are shown, 648 and statistically significant differences (P < 0.05) compared to the SeV-infected 649 empty vector-transfected cells (#) and the same amount of the SARS

ORF3b-transfected cells (*) are shown. E, empty vector

See also Figures S1 and S2 and Table S1

Figure 2. C-terminal truncations increase the IFN-antagonistic activity of 654 ORF3b 655 (A) Maximum likelihood phylogenetic tree of Sarbecovirus ORF3b. The ORF3b 656 sequences of SARS-CoV-2 (Wuhan-Hu-1)

SARS-CoV (Tor2, GZ0402, GZ02, Urbani, BJ02

SARS-CoV-related viruses from civets (civet007) and bats 660 (Rs7327, YN2013, Rs4231, YNLF34C, Shaanxi2011, Rm1, F46, HKU3-2, GX2013 661 and Yunnan2011), and two outgroup viruses

The ORF3b sequences of all SARS-CoV-related viruses are summarized in Table 663 S2, and the ORF3b sequences used in this study are summarized in Table S3

Bootstrap value; *, 665 >70%. 666 (B) Proportion of the ORF3b lengths in each Sarbecovirus. The distribution of 667 different lengths of ORF3b in each viral group is summarized in pie charts

The number at the pie charts give the protein length indicated, and the numbers in 670 bold indicate the most prevalent protein length for each viral group

Note that the amino acid sequences of ZXC21 and ZC45 679 are identical. An uncropped dot blot is shown in Figure S2B. 680 (E) Subcellular localization of Sarbecovirus ORF3b. Cell lysates of the HEK293 681 cells transfected with a plasmid expressing HA-tagged Sarbecovirus ORF3b were 682 separated into cytosolic and nuclear fractions as described in the Methods section. 683 The percentage of ORF3b protein localized in the nucleus (top, n=4) and a 684 representative Western blot (bottom) are shown. TUBA and LMNA were used for as 685 controls for cytosolic and nuclear proteins

Cell lysates of the HEK293 cells transfected with 692 a plasmid expressing HA-tagged ORF3b mutants were separated into cytosol or 693 nuclear fractions as described in the Methods section. The percentage of ORF3b 694 protein localized in the nucleus (top, n=4) and a representative Western blot 695 (bottom) are shown. TUBA and LMNA were used for as controls for cytosolic and 696 nuclear proteins. (G) HEK293 cells were cotransfected with a plasmid expressing 697 the

SeV was inoculated at MOI 10. 24 h post infection, cells 699 were harvested for Western blotting (top) and luciferase assay (bottom)

Subcellular localization of ORF3b and IRF3. HeLa cells were transfected with 701 the indicated plasmids expressing HA-ORF3b and were infected with SeV as 702 described in the Methods section. Representative figures are shown. Scale bar, 10 703 µm. The white circles in the panels of

One 706 representative blot out of three independent experiments is shown. For the 707 luciferase assay, the value of the SeV-infected empty vector-transfected cells was 708 set to 100%. The mean values of three independent experiments with SEM are 709 shown, and statistically significant differences

SeV-infected empty vector-transfected cells (#) are shown. In (D), red asterisks 711 indicate statistically significant differences

In (G), blue and green asterisks indicate 713 statistically significant differences (P < 0.05) compared the same amount of either 714

Tor2 ORF3b L115*-transfected cells or Rs4231 ORF3b WT

See also Figure S2 and Tables S2 and S3

Enhanced anti-IFN-I upon reconstitution of the cryptic SARS-CoV-2 719 ORF3b 720 (A) Schemes illustrating the genomic regions encoding ORF2, ORF3a, ORF3b and 721 ORF4 of SARS-CoV-2 and SARS-CoV. Open squares with dotted red lines indicate 722 a cryptic ORF3b reading frame in SARS-CoV-2 that is similar to SARS-CoV ORF3b 723 (see also Figure S3A). Asterisks indicate stop codons in the ORF3b frame. 724 (B) SARS-CoV-2

726 119* and 155*) are shown. Asterisks indicate the stop codons in the original ORF3b 727 frame. (Bottom) A natural ORF3b variant detected in two sequences deposited in 728 GISAID (GISAI accession IDs: EPI_ISL_422564 and EPI_ISL_422565; herein 729 designated an "Ecuador variant") are shown. The corresponding nucleotide and 730 amino acid sequences are shown in Figure S3C

HEK293 cells were 732 cotransfected with two different amounts of plasmids expressing the indicated 733 HA-tagged SARS-CoV-2 ORF3b derivatives (WT, 57*, 79*, 119* and 155*; 50 and 734 100 ng) and p125Luc (500 ng). 24 h post transfection, SeV was inoculated at MOI 735 10

Enhanced anti-IFN-I activity of an Ecuador variant ORF3b. HEK293 cells were 738 cotransfected with two different amounts of plasmids expressing HA-tagged 739 "Ecuador variant" ORF3b or parental SARS-CoV-2 ORF3b (50 and 100 ng) and 740 p125Luc (500 ng). 24 h post transfection

Enhanced inhibition of the IRF3-mediated IFN-I activation by the Ecuador 744 variant ORF3b. HEK293 cells were cotransfected with two different amounts of 745 plasmids expressing the indicated HA-tagged viral proteins (50 and 100 ng) and 746 p55C1B-Luc (500 ng). 24 h post transfection

For the luciferase assay, the value of the SeV-infected empty 752 vector-transfected cells was set to 100%. The mean values of three independent 753 experiments with SEM are shown, and statistically significant differences (P < 0.05) 754 compared to the SeV-infected empty vector-transfected cells (#) and the same 755 amount of the SARS-CoV-2 ORF3b WT

See also Figures S2 and S3 and Table S4

were transfected with using a FuGENE HD transfection reagent (Promega) 833 according to the manufacturer's protocol. For luciferase reporter assay, cells were 834 cotransfected with 500 ng of either p125Luc (expressing firefly luciferase driven by 835 human IFNB1 promoter; kindly provided by Dr

1993) and the pCAGGS-based HA-tagged expression 838 plasmid (the amounts are indicated in the figure legends). A549 cells (100,000 839 cells) were electroporated with 500 ng of the pCAGGS-based HA-tagged 840 expression plasmid using a Neon transfection system (Thermo Fisher Scientific) 841 according to the manufacturer's protocol (1200 V; 30 ms; 2 times pulse)

2018) was inoculated into the transfected cells at 844 multiplicity of infection (MOI) 10 (for HEK293 and A549 cells) or 5 (for HeLa cells)

Briefly, 50 µl of cell 849 lysate was applied to a 96-well plate (Nunc), and the firefly luciferase activity was 850 measured using a PicaGene BrillianStar-LT luciferase assay system (Toyo-b-net), 851 and the input for the luciferase assay was normalized by using a CellTiter-Glo 2.0 852 assay kit (Promega) following the manufacturers' instructions

Subcellular fractionation was performed using Nuclear/cytosol fractionation kit 857 (Biovision) according to the manufacturer's procedure

Transfected cells were lysed with 1x SDS sample buffer (62.5 mM Tris-HCl

5% 2-mercaptoethanol and 0.0025% bromophenol blue)

2018) using an HRP-conjugated rat 864 anti-HA monoclonal antibody (clone 3F10; Roche), a mouse anti-alpha-tubulin 865 (TUBA) monoclonal antibody (clone DM1A; Sigma-Aldrich); a rabbit anti-lamin A/C 866 (LMNA) polyclonal antibody (Cell Signaling Technology); an HRP-conjugated horse 867 anti-mouse IgG antibody (Cell Signaling Technology); and an HRP-conjugated goat 868 anti-rabbit IgG antibody (Cell Signaling Technology). Dot blotting was performed 869 using a Bio-Dot microfiltration apparatus

45-µm membranes (Merck) were used for Western blotting, 873 while nitrocellulose 0.20-µm membranes (Bio-Rad) were used for dot blotting

Thermo Fisher Scientific). cDNA was 878 synthesized using SuperScript III reverse transcriptase (Thermo Fisher Scientific) 879 and oligo(dT)12-18 primer (Thermo Fisher Scientific). Real-time RT-PCR was 880 performed as previously described

Immunofluorescence Staining

Twenty-four h post infection, cells were fixed with formaldehyde, 888 permeabilized with Triton X-100, and then stained using an FITC-conjugated 889 anti-HA antibody (clone 3F10; Roche); a rabbit anti-IRF3 polyclonal antibody 890 (Abcam); and an Alexa 546-conjugated anti-rabbit IgG antibody (Thermo Fisher 891 Scientific)

Antipode Moutant with DAPI (Thermo Fisher Scientific) and observed using an 893 FV-1000D confocal microscope

CoV-GLUE 896 To survey the ORF3b derivatives in pandemic SARS-CoV-2 sequences, we used 897 the viral sequences deposited in GISAID

We constructed a phylogenetic tree 903 using RAxML-NG version 0.9.0 (Kozlov et al., 2019) with a TPM3uf substitution 904 model (Figure S2D). We also detected the two SARS-CoV-2 sequences (GISAID 905 accession IDs: EPI_ISL_422564 and EPI_ISL_422565, collected in Quito, Ecuador) 906 possessing the V163T/T164N substitutions in ORF3a

All viral genome sequences used in this study and the respective GenBank or 797 GISAID (https://www.gisaid.org) accession numbers are summarized in Table S1 . 798 We first aligned the viral genomes using the L-INS-i program of MAFFT version 799 7.453 (Katoh and Standley, 2013). Based on the multiple sequence alignment and 800 the gene annotation of SARS-CoV, we extracted the region of the ORF3b gene. We 801 then constructed phylogenetic trees using the full-length genomes (Figures 1A and 802 S1) and ORF3b genes (Figure 2A ). We generated a maximum likelihood based 803 phylogenetic tree using RAxML-NG version 0. Table S3 ) and the cryptic 813 SARS-CoV ORF3b-like sequence in SARS-CoV-2 [Wuhan-Hu-1 (GenBank 814 accession no. NC_045512.2), nucleotides 25814-26281, see also Figure S3A ) was 815 synthesized by a gene synthesis service (Fasmac). The ORF3b derivatives were 816 generated by PCR using PrimeSTAR GXL DNA polymerase (Takara), the 817 synthesized ORFs as templates, and the primers listed in Table S5 . The HA-tagged 818Ecuador variant ORF3b (GISAID accession IDs: EPI_ISL_422564 and  819 EPI_ISL_422565, which corresponds to the S23Q/L24M mutant of SARS-CoV-2 820Wuhan-Hu-1 ORF3b *57; see also Figure S3C ) was generated by overlap 821 extension PCR by using PrimeSTAR GXL DNA polymerase (Takara), the 822 SARS-CoV-2 ORF3b 155* as the template, and the primers listed in