ORF3c is expressed in SARS-CoV-2 infected cells and suppresses immune activation by inhibiting innate sensing

SARS-CoV-2 proteins are translated from subgenomic RNAs (sgRNAs). While most of these sgRNAs are monocistronic, some viral mRNAs encode more than one protein. For example, the ORF3a sgRNA also encodes ORF3c, an enigmatic 4l-amino acid peptide. Here, we show that ORF3c is expressed in SARS-CoV-2 infected cells and suppresses RIG-I- and MDA5-mediated immune activation and IFN-β induction. Mechanistic analyses revealed that ORF3c interacts with the signaling adaptor MAVS, induces its C-terminal cleavage and inhibits the interaction of RIG-I with MAVS. The immunosuppressive activity of ORF3c is conserved among members of the subgenus sarbecovirus, including SARS-CoV and coronaviruses isolated from bats. Notably, however, the SARS-CoV-2 delta and kappa variants harbor premature stop codons in ORF3c demonstrating that this reading frame is not essential for efficient viral replication in vivo and likely compensated by other viral proteins. In agreement with this, disruption of ORF3c did not significantly affect SARS-CoV-2 replication in CaCo-2 or CaLu-3 cells. In summary, we here identify ORF3c as an immune evasion factor of SARS-CoV-2 that suppresses innate sensing in infected cells.


39
Since the emergence of the COVID-19 pandemic, all canonical proteins of SARS-CoV-2 have 40 been extensively characterized for their expression, structure and function. In addition to its 41 prototypical genes, however, SARS-CoV-2 harbors several smaller open reading frames 42 (ORFs) that frequently overlap with other ORFs and may also contribute to efficient viral 43 replication. For example, the ORF3b peptide encoded by ORF3a subgenomic RNA (sgRNA) 44 was shown to suppress the induction of type I interferon (IFN) [1]. Intriguingly, naturally 45 occurring variants of ORF3b differ in their immunosuppressive activity and may be responsible 46 for phenotypic differences between SARS-CoV and SARS-CoV-2 [1]. Moreover, several short 47 upstream ORFs (uORFs) have been suggested to regulate translation of downstream genes such 48 as ORF7b [2]. Thus, non-canonical ORFs of SARS-CoV-2 may also be important determinants 49 of viral immune evasion, spread and/or pathogenicity. 50 Nevertheless, most of the cryptic ORFs of SARS-CoV-2 remain poorly characterized, and 51 several open questions remain: Do they encode proteins or are they merely a result of selection 52 pressures acting on overlapping reading frames? Do these ORFs exert any regulatory activity, 53 e.g. by modulating translation of downstream ORFs via leaky scanning or ribosomal re-54 initiation? Do they code for functional proteins that contribute to efficient immune evasion 55 and/or replication of SARS-CoV-2? Are the respective peptides or proteins immunogenic? 56 contrast, we found no evidence for antibodies against ORF3c in SARS-CoV-2 convalescent 90 sera (Fig S1) 91 To characterize the stability and potential activity of cryptic ORF3 peptides, we generated 92 expression vectors for the individual peptides harboring a C-terminal HA-tag (without codon-93 optimization). Apart from the ORF3a construct, ORF3c and ORF3d/ORF3d-2 code for stable 94 proteins that are readily detectable in transfected cells (Fig 1C). The remaining peptides were 95 independently of the receptor, it most likely targets a factor further downstream in the signaling 246 cascade. In line with this, co-immunoprecipitation experiments revealed an interaction of 247 ORF3c with the mitochondrial signaling adaptor MAVS. We found no evidence for an 248 interaction with other components of the sensing cascade (i.e. RIG-I, MDA5, or TBK1). 249 Notably, ORF3c failed to prevent IFNB1 promoter activation if MAVS itself was used as an 250 activator. This lack of inhibition is not merely the result of a saturation effect since MAVS 251 induced the IFNB1 promoter less efficiently than RIG-I (Fig 2). MAVS is targeted by proteins 252 from different viruses. For example, Influenza A Virus PB1-F2 inhibits innate sensing by 253 binding to MAVS [22] and decreasing the mitochondrial membrane potential [23]. Another 254 example is the NS3/4A serine protease of Hepatitis C virus (HCV), which cleaves MAVS, 255 thereby inhibiting downstream immune activation [24]. Similarly, SARS-CoV-2 ORF10 was 256 recently shown to induce the degradation of MAVS via mitophagy [25]. Further mechanistic 257 analyses revealed that ORF3c induces the proteolytic processing of MAVS, resulting in a C-258 terminal 9 kDa fragment of this signaling protein ( Fig 2G). Furthermore, ORF3c reduced the 259 interaction of MAVS with RIG-I ( Fig 2F). Thus, one possible mode of action is a competitive 260 binding of ORF3c and RIG-I (and possible MDA5) to MAVS. In the presence of ORF3c, the 261 CARD domain of MAVS may not be accessible and thus not be bound and activated by active 262 RIG-I or MDA5. Notably, however, co-immunoprecipitation experiments showed that the 263 CARD domain of MAVS is dispensable for an interaction of ORF3c with MAVS ( Fig 2E). 264 In line with a relevant role of ORF3c in viral replication, its immunosuppressive activity is 265 conserved in orthologs of other sarbecovirus species, including the SARS-CoV reference virus 266 Tor2. While all orthologs tested inhibited IFNB1 promoter activation, those of SARS-CoV Tor2 267 and batCoV BANAL20-52 were less active than their counterparts from SARS-CoV-2 Wuhan-268 Hu-1 and batCoV ZXC21. The reduced activity of BANAL-20-52 ORF3c can be ascribed to a 269 single amino acid change (L11P) distinguishing it from Wuhan-Hu-1 ORF3c (Fig 3B). While 270 L11 is largely conserved in the SARS-CoV-2 cluster, most of the viruses in the SARS-CoV 271 cluster (including Tor2) harbor a glutamine at this position ( Fig 3B) [3,5]. Thus, 272 polymorphisms at position 11 can affect the inhibitory activity of ORF3c. In addition to this, 273 our alanine scanning approach revealed that Leu2/Leu3 and Ile6/Leu7 are also contributing to 274 the immunosuppressive effect of ORF3c (Fig 3D). Most of these residues are conserved among 275 different sarbecoviruses. One notable exception is Ile6 (Fig 3B). Viruses from the SARS-CoV 276 cluster harbor a Valine at this residue. These include SARS-CoV Tor2 ORF3c, which was less 277 active than its SARS-CoV-2 counterpart. 278 ORF3c is not the only SARS-CoV-2 protein that interferes with RIG-I-and/or MDA-5-279 mediated immune activation. As already mentioned above, ORF10 suppresses innate sensing 280 by targeting MAVS [25]. Moreover, ORF3b, nucleocapsid, ORF6 and ORF8 have all been 281 shown to suppress IFN-β expression [1,12,20,26,27], highlighting the selection pressure 282 exerted by this pathway. The convergent evolution of viral proteins exerting overlapping 283 immune evasion activities may represent a backup mechanism that allows viral replication even 284 if one of the IFN-β suppressing proteins is lost. In line with this, SARS-CoV-2 variants 285 expressing a C-terminally truncated, inactive ORF6 protein have emerged several times during 286 the pandemic and spread via human-to-human transmission [12]. Similarly, natural SARS-287 CoV-2 variants lacking an intact ORF3c gene still efficiently spread in the human population. 288 In fact, more than 80% of the sequenced genomes of B.1.617.1, B.1.617.2 and B.1.630 harbor 289 premature stop codons at positions 5 and 15, respectively (Fig 4). We hypothesized that the loss 290 of ORF3c in these viruses may be compensated by the activity of ORF6. However, replication 291 kinetics in CaCo-2 and CaLu-3 cells revealed that loss of ORF3c does not affect viral 292 replication in the absence of ORF6 either (Fig 5). Intriguingly, IFN-β mRNA levels were not 293 increased upon infection with the SARS-CoV-2 double mutant lacking ORF3c and ORF6 (data 294 not shown). Thus, yet another viral inhibitor of IFN-β expression (e.g. ORF8 or N) may be able 295 to rescue efficient viral replication in this case.
Notably, the emergence of premature stop codons in small reading frames such as ORF3c may 297 also be tolerated or even be beneficial if they provide a fitness advantage by optimizing 298 overlapping reading frames. For example, the ORF3c Q5* mutation is accompanied by an S26L 299 change in ORF3a. However, experimental disruption of ORF3c without changing the amino 300 acid sequence of ORF3a (Fig 5A) or upon deletion of ORF3a ( Fig 5B) did not affect viral 301 replication in vitro either. 302 One intriguing observation is the emergence of a nucleotide change in a subfraction (~3%) of 303 B.1.617.2 viruses that reverts the stop codon at position 5 to a tyrosine (*5Y). Thus, it is 304 tempting to speculate that the loss of ORF3c was initially just carried along with mutations 305 elsewhere in the genome (e.g. in Spike) that conferred a major fitness advantage to the virus, 306 before ORF3c expression was reverted by another point mutation. 307 In summary, our study identifies ORF3c as an immune evasion factor of SARS-CoV-2 and 308 other sarbecoviruses. While an intact ORF3c gene is clearly dispensable for viral replication in 309 vitro and in vivo, the conservation of this short open reading frame and the pseudo-reversion of 310 premature stop codons suggests that it may still contribute to efficient viral replication in vivo. 311 The emergence of future SARS-CoV-2 variants may help to fully decipher the role of this 312 enigmatic ORF and its co-evolution with other viral genes. 313

Generation of expression plasmids 317
ORF3c genes were PCR-amplified using viral cDNA as a template and subsequently inserted 318 into a pCG expression vector co-expressing GFP via an IRES using unique XbaI  bovine growth hormone poly A signal and the cytomegalovirus promoter was also prepared by 352 PCR. The ten obtained DNA fragments were mixed and used for CPER. ORF3c mutations were 353 inserted in fragment 9/10 by site-directed overlap extension PCR with the primers listed in EV1. 354 To produce chimeric recombinant SARS-CoV-2, Tetracycline-inducible ACE2 and TMPRSS-355 expressing IFNAR1-deficient HEK293 (HEK293-C34) cells were transfected with the CPER 356 products using TransIT-LT1 (MirusBio, Cat#2300) according to the manufacturer's protocol. 357 One day post transfection, the culture medium was replaced with Dulbecco's modified Eagle's 358 medium (high glucose) containing 2% FCS, 1% PS and doxycycline. At 7 d post transfection, 359 the culture medium was harvested and centrifuged, and the supernatants were collected as the 360 seed virus. To remove the CPER products (i.e., any SARS-CoV-2 DNA), 1 ml of the seed virus 361 was treated with 2 μl TURBO DNase (Thermo Fisher Scientific, Cat# AM2238) and incubated 362 at 37°C for 1 h. Complete removal of the CPER products (i.e., SARS-CoV-2-related DNA) 363 from the seed virus was verified by PCR. To prepare virus stocks for 364 infection, VeroE6/TMPRSS2 cells (5,000,000 cells in a T-75 flask) were infected with 20-50 365 µl of the seed virus. One-hour post infection, the culture medium was replaced with DMEM 366 (low glucose) containing 2% FBS and 1% PS. Two to four days post infection, the culture 367 medium was harvested and centrifuged, and the supernatants were collected. Viral titers were 368 determined by TCID50. To verify the sequence of chimeric recombinant SARS-CoV-2, viral 369 RNA was extracted from the virus stocks using the QIAamp viral RNA mini kit (Qiagen,370 Cat#74136) and viral genomes were sequenced as described before [31]. 371 372 Tissue culture infectious dose (TCID50) 373 Viral titers were determined as the 50% tissue culture infectious dose. Briefly, one day before 374 infection, VeroE6/TMPRSS2 cells (10,000 cells) were seeded into 96-well plates. Cells were 375 inoculated with serially diluted virus stocks and incubated at 37°C. Four days later, cells were 376 checked microscopically for cytopathic effects (CPE), and TCID50/ml was calculated using the 377 Reed-Muench method. 378

SARS-CoV-2 replication kinetics in Vero E6, Caco-2 and Calu-3 cells 380
One day before infection with CPER-derived SARS-CoV-2 clones, Caco-2 cells (10,000 381 cells/well) or CaLu-3 cells (20,000 cells/well) were seeded into a 96-well plate. Cells were 382 infected with SARS-CoV-2 at an MOI of 0.1 and incubated at 37°C. One hour later, the infected 383 cells were washed and 180 μl of culture medium was added. The culture supernatants and cells 384 were harvested at the indicated timepoints and used for RT-qPCR to quantify the viral RNA 385 copy number. For replication kinetics of BAC-derived SARS-CoV-2 ΔORF6-YFP and SARS-386 CoV-2 ΔORF6-YFP ΔORF3c, Caco-2 cells (10,000/well) were seeded into a 96-well plate one 387 day prior to infection. Cells were infected in triplicates at an MOI of 0.1 & 0.01 for 1 hour at 388 37°C. After washing and addition of 100 µl fresh culture medium the plates were placed in an 389 Incucyte plate reader and images were taken at the indicated time points for up to 96 hours. 390 The 'Basic Analysis Mode' was applied to quantify virus growth as green area normalized to 391 phase area. Supernatants and cells were harvested at the indicated time points to determine 392 cytokine levels by Cytokine Array and RT-qPCR respectively.

LIPS assay 516
Luciferase fusion protein (10 6 RLU) in 50 µl Buffer A (50 mM Tris, 150 mM NaCl, 0.1% Triton 517 X-100, pH 7.5) and 1 µl sample serum in 49 µl Buffer A was added to 1.5 ml tubes and 518 incubated with shaking at 300 rpm for 1 h at room temperature. Pierce Protein A/G Magnetic 519 beads were added to each condition as a 30% suspension in PBS for an additional hour and 520 shaking at room temperature. Samples were placed on a magnetic rack, and supernatant was 521 removed after 1-minute incubation. Magnetic beads were washed twice with 150 µl Buffer A 522 followed by 2 washes with 150 µl PBS. Samples were transferred into a 96-well opaque nunc-523 plate (VWR), and 50 µl Coelenterazine (PJK Biotech, Cat# 102174) was added to each 524 condition. Samples were measured immediately on a TriStar² S LB 942 Multimode Reader 525 (Berthold Technologies) with an integration time of 0.1 seconds and a read height of 1 mm. 526 527 Statistical analyses were performed using GraphPad PRISM 9.4.1. For statistical testing 529 between two means P values were calculated using paired or unpaired Student's t test. For 530 comparison within one group, we used one-way analysis of variation (ANOVA) with Dunnett's 531 multiple comparison test and for comparison between two or more groups we used two-way 532 ANOVA with Sidak's multiple comparison test. Unless otherwise stated, data are shown as the 533 mean of at least three independent experiments ± SD. Significant differences are indicated as: 534 directly ("input") or upon pull-down using a Flag-specific antibody ("IP"). 718

F
HEK293T cells were transfected with plasmids expressing BFP, ORF3c, the C-terminal 719 part of YFP fused to RIG-I and/or the N-terminal part of YFP fused to MAVS. 24 h later, cells 720 were fixed, and YFP fluorescence was detected by flow cytometry as a reporter for MAVS-721 RIG-I interaction. Exemplary flow cytometry data are shown on the right. 722

G
HEK293T cells were transfected with increasing amounts of an expression plasmid for 723 SARS-CoV-2 ORF3c. One day post transfection, cells were lysed for Western blotting. ORF3c 724 was detected via an HA tag, and MAVS was detected with antiserum specific for the C-terminal 725 part of MAVS. GAPDH served as loading control. MAVS bands were quantified and the ratio 726 of the 9 kDa fragment to total MAVS was calculated. One exemplary blot is shown on the left. 727 Data information: In (B, C, F), data are presented as mean (±SD) of three independent 728 experiments. In (G), data are presented as mean (±SEM) of three independent experiments. 729 Multiple comparison within individual reporter assays were determined by one-way ANOVA 730 with Dunnett's test * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ % 0.001 and **** p ≤ 0.0001. Sidak's multiple comparison test; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ % 0.001 and **** p ≤ 0.0001. 752 for a viral protein of interest fused to Renilla luciferase. Subsequently, transfected cells are lysed and incubated with serum samples and magnetic beads. Antibodies against viral proteins 805 of interest will cross-link the luciferase-containing proteins with beads and allow magnet-806 assisted pull-down of both beads and luciferase activity. 807