SARS-CoV-2 ORF7a potently inhibits the antiviral effect of the host factor SERINC5

Serine Incorporator 5 (SERINC5), a cellular multipass transmembrane protein that is involved in sphingolipid and phosphatydilserine biogenesis, potently restricts a number of retroviruses, including Human Immunodeficiency Virus (HIV). SERINC5 is incorporated in the budding virions leading to the inhibition of virus infectivity. In turn, retroviruses, including HIV, encode factors that counteract the antiviral effect of SERINC5. While SERINC5 has been well studied in retroviruses, little is known about its role in other viral families. Due to the paucity of information regarding host factors targeting Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), we evaluated the effect of SERINC proteins on SARS-CoV-2 infection. Here, we show SERINC5 inhibits SARS-CoV-2 entry by blocking virus-cell fusion, and SARS-CoV-2 ORF7a counteracts the antiviral effect of SERINC5 by blocking the incorporation of over expressed SERINC5 in budding virions.

I n 2019, a novel pathogen associated with severe pneumonia was discovered and later identified to be a coronavirus, which was subsequently named Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) 1 and is responsible for the current pandemic. Disease hallmarks range from mild to severe, including loss of smell and taste, breathing difficulty, fever, malaise and in elderly patients or people with preexisting illnesses, death 2 .
An important determinant of virus infectivity is cellular entry 3,4 . Coronaviruses enter a target cell by first binding to a cell surface receptor followed by endosomal entry and fusion between the host and viral membranes 5 . In the case of SARS-CoV-2, the Spike (S) protein is critical for virus entry into human cells 5 . SARS-CoV-2 S is a transmembrane protein that forms a homotrimer on the surface of mature virions 6 . The S protein contains two subunits, S1 and S2. S1 interacts via its Receptor Binding Domain with the human angiotensin-converting enzyme 2 (ACE2) receptor on the surface of cells 5,7 , and S2 facilitates fusion between virus and target cell membrane. SARS-CoV-2 S contains a furin cleavage site at the S1/2 boundary that is processed in the producer cell during virus production 6 . Subsequent activation of SARS-CoV-2 S by the target cell proteases cathepsin B/L and type II membrane serine protease 2 (TMPRSS2) is critical for virus entry into the cell, as it permits fusion between the viral and cellular membranes 8,9 .
Viral entry, including that of SARS-CoV-2, is a major target for the development of intervention strategies 10,11 . However, the cellular host factors that regulate SARS-CoV-2 entry are not fully understood. Members of the Serine Incorporator (SERINC) protein family, which are thought to play an important role in sphingolipid and phosphatidylserine biogenesis 12 , have important antiviral functions against retroviruses including Human Immunodeficiency Virus 1 (HIV-1) and Murine Leukemia Virus (MLV) 13,14 . SERINC5, and to a lesser extent SERINC3, are incorporated in the budding virions and interfere with the fusion of the viral and cellular membranes 15,16 . In turn, HIV-1 encodes a viral protein that counteracts the antiviral effect of this host factor. Negative effector factor (Nef) counteracts SERINC5 by removing it from the plasma membrane and preventing SERINC5 incorporation in the budding virions 13,14 . Furthermore, SERINC5 restricts retrovirus infection in vivo 17 , which further highlights the importance of this host factor in virus restriction.
In this work, we elucidate the role of SERINC proteins in SARS-CoV-2 infection. We find that SERINC5 is expressed in pneumocytes and potently inhibits SARS-CoV-2 infection by binding to SARS-CoV-2 S, thereby blocking the fusion step during virus entry. In addition, we demonstrate that SARS-CoV-2 ORF7a, a SARS-CoV-2 accessory protein, counteracts the antiviral effect of SERINC5. SARS-CoV-2 ORF7a counteracts SER-INC5 by utilizing two distinct mechanisms; it blocks SERINC5 incorporation in budding virions and is incorporated in budding virions, forming a complex with SERINC5 and SARS-CoV-2 S as a means of blocking SERINC5-mediated restriction of virus infectivity. In summary, we determine the mechanism SERINC5 employs to block SARS-CoV-2 infection, and we identify a previously undescribed function of SARS-CoV-2 ORF7a that is conserved among coronaviruses.

Results
Transcriptional regulation of SERINC genes. To elucidate the role of SERINC genes during SARS-CoV-2 infection, we initially examined the levels of all the members of the SERINC protein family (SERINC1-5) in total lung tissue RNA from three donors and Calu-3 cells, a human pneumocyte cell line, as pneumocytes are natural targets of SARS-CoV-2 infection 18 . We performed RT-qPCR to determine the expression levels of the different members of the SERINC family and found that all SERINC genes were expressed in human lungs and Calu-3 cells except for SERINC4 (Fig. 1a, b). Thus, we excluded SERINC4 from further studies. ACE2 is an interferon stimulated gene (ISG) in lung epithelial cells, yet transcriptomic data from IFN treated peripheral blood mononuclear cells (PBMCs) never characterized ACE2 as an ISG 19 . Similarly, SERINC gene expression is not induced by type I IFN in PBMCs 13 . Therefore, we sought to determine if IFN-β upregulates SERINC gene transcription in Calu-3 cells. We treated Calu-3 cells with IFN-β (500 U/ml), harvested cells at different time points, and isolated RNA followed by RT-qPCR to determine changes in SERINC gene expression. We found that the transcription levels of all SERINC genes examined (SERINC1, 2, 3, and 5) were unaffected by the presence of IFN-β in Calu-3 cells (Fig. 1c). We also determined the effect of SARS-CoV-2 infection on the expression levels of SERINC genes. We infected Calu-3 cells with SARS-CoV-2 (5 MOI) (USA-WA1/2020 20 ) and harvested cells 4 and 6 h post infection (hpi). RNA was isolated and RT-qPCR was performed to determine viral RNA levels and possible changes in the expression levels of SERINC1, 2, 3, and 5. We found that SARS-CoV-2 infection had no effect on the transcription levels of SERINC genes (Fig. 1d, e). Therefore, we concluded that IFN-β and SARS-CoV-2 infection do not affect SERINC genes expression levels.
SERINC5 potently restricts SARS-CoV-2 S-mediated entry. SERINC5 exerts a potent antiretroviral effect by restricting HIV-1 entry, thereby decreasing virion infectivity 13,14 . We initially examined the role of SERINC proteins on SARS-CoV-2 entry by generating one-hit luciferase reporter SARS-CoV-2 S pseudoviruses using a replication-defective HIV-1 proviral plasmid (pHIV-1 NL ΔEnv-NanoLuc 21 ) in the presence of SERINC1, 2, 3 and 5. To determine if SERINC proteins are incorporated in SARS-CoV-2 S pseudoviruses, we concentrated SARS-CoV-2 S pseudoviruses by ultracentrifugation from the media of the transfected cells and performed western blots. We found that all SERINC proteins examined (SER-INC1, 2, 3, and 5) are packaged in SARS-CoV-2 S pseudovirions (Fig. 2a). We then infected Calu-3 and 293T-hACE2 cells using SARS-CoV-2 S pseudoviruses produced in the presence of the different SERINC genes and found that the presence of SERINC5 in SARS-CoV-2 S pseudoviruses resulted in a significant reduction in virus infectivity in both Calu-3 and 293T-hACE2 cells (Fig. 2b). SERINC3 reduced virus infectivity only in Calu-3 cells, albeit not as robustly as SERINC5, while SERINC1 and 2 had no effect (Fig. 2b). Thus, we concluded that SERINC5 blocks SARS-CoV-2 entry. To ensure that the SERINC5-mediated reduction in SARS-CoV-2 S-mediated entry is not due to the HIV-1 proviral packaging plasmid we used, we produced SARS-CoV-2 S pseudoviruses in the presence of SERINC3 and 5 using an MLV packaging plasmid and a luciferase reporter followed by infection of 293T-hACE2 cells. As a positive control, we used amphotropic MLV envelope pseudovirions, which are known to be restricted by SERINC5 17,22 . Similar to our studies with the lentiviral packaging system, we found that SARS-CoV-2 S pseudovirions produced in the presence of SERINC5 and the MLV packaging plasmid were significantly less infectious than those produced in the presence of empty vector (E.V.) (Supplementary Fig. 1), while SERINC3 had only a modest effect on virus infectivity ( Supplementary Fig. 1). Thus, our results show that SER-INC5, and to a lesser extent SERINC3, block SARS-CoV-2 S-mediated entry regardless of the packaging system we used.
While we clearly see that SERINC5 blocks SARS-CoV-2 S-mediated entry and is incorporated in SARS-CoV-2 S pseudoviruses, this does not necessarily reflect what happens with coronaviruses. Pseudoviruses generated using a lentiviral packaging plasmid, assemble at the plasma membrane 23 . On the other hand, coronaviruses, including SARS-CoV-2, assemble at the endoplasmatic reticulum Golgi intermediated compartment (ERGIC) 24 . Therefore, SERINC5 incorporation in pseudoviruses does not necessarily equate with SERINC5 incorporation in SARS-CoV-2 virions. To ascertain that SERINC5 is incorporated in SARS-CoV-2 virions, we employed two methodologies, using both virus-like particles (VLPs) and infectious virus. We initially evaluated SERINC5 incorporation in VLPs; VLPs generated by transfecting all SARS-CoV-2 structural proteins self-assemble at the ERGIC, similar to infectious SARS-CoV-2 virions 25,26 . Consequently, VLPs are an accurate model to study SARS-CoV-2 assembly. We cotransfected 293T cells with plasmids encoding SARS-CoV-2 N (Nucleoprotein), M (Membrane), E (Envelope), and S in the presence or absence of SERINC5. At 48 h post transfection, cells were lysed and VLPs in the culture media of the transfected cells were concentrated by ultracentrifugation. We then performed western blots and found that SERINC5 was incorporated in the purified VLPs (Fig. 2c). Therefore, we concluded that SERINC5 is incorporated in SARS-CoV-2 VLPs during assembly at the ERGIC. To determine whether SERINC3 and 5 are incorporated in infectious SARS-CoV-2 virions, 293T-hACE2 expressing either SERINC5 or SERINC3 were infected with SARS-CoV-2 (USA-WA1/2020 20 ). We purified virus from the culture media of the cells and performed western blots. We observed that SERINC5 was incorporated quite efficiently in SARS-CoV-2 virions, while SERINC3 was incorporated at lower levels (Fig. 2d). Unfortunately, we could not examine incorporation of endogenous SERINC3 and SERINC5 levels as the antibodies available are not suitable for western blots.
To determine the effect of incorporated SERINC3 and 5 on infectious SARS-CoV-2, we infected 293T-hACE2 and Calu-3 cells with equal amounts of SARS-CoV-2 (USA-WA1/2020 20 ) (normalized for RNA levels) isolated from the media of infected 293T-hACE2 cells expressing either E.V., SERINC3 or SERINC5. At 6 hpi RNA was isolated from the infected cells followed by RT-qPCR. We found that SERINC5 reduced SARS-CoV-2 infectivity, while SERINC3 had no effect (Fig. 2e). To ensure that the RNA levels detected were due to infection and not input virus inoculum, we also infected Mus dunni cells (mouse fibroblasts) with SARS-CoV-2, as mouse cells do not get infected by SARS-CoV-2 and any viral RNA detected will be from the input inoculum. As expected, very low levels of viral RNA were detected (Fig. 2e). In conclusion, our data using pseudoviruses, VLPs, and infectious SARS-CoV-2 demonstrate that SERINC5 restricts SARS-CoV-2 infection by blocking virus entry. Therefore, in this report we have focused on the role of SERINC5 on SARS-CoV-2 infection.
SERINC5 blocks SARS-CoV-2 entry of different Spike variants. It was previously shown that envelopes derived from different HIV-1 strains differ in their sensitivity to SERINC5-mediated restriction 16 . During the pandemic spread of SARS-CoV-2, a number of variants have emerged, which are associated with high transmissibility 27 27,29,30 . Therefore, it is possible that S proteins derived from different SARS-CoV-2 variants may differ in their susceptibility to SER-INC5 restriction. To determine the effect of SERINC5 on S proteins derived from different SARS-CoV-2 variants, we generated one-hit luciferase reporter SARS-CoV-2 S pseudoviruses in the presence or absence of SERINC5, using a replication-defective HIV-1 proviral plasmid described above and SARS-CoV-2 S derived from the Alpha (B.1.1.7), Beta (B.1.351), Gamma (P1), and Delta (B.1.617) variants respectively. Afterwards, Calu-3 cells were infected and luciferase levels were measured 48 hpi. We found that SERINC5 restricted infection of all viral particles regardless of the SARS-CoV-2 variant (Alpha, Beta, Gamma and Delta) from which the S protein was derived (Fig. 3). In  conclusion, our data show that SERINC5 restriction of SARS-CoV-2 S-mediated entry is consistent across the different SARS-CoV-2 variants.
SERINC5 inhibits SARS-CoV-2 by targeting virus-cell membrane fusion. SARS-CoV-2 entry into the cell is a complex process of events that involves the coordination of receptor interaction and proteolytic cleavages of the S protein that culminates with virus-cell membrane fusion. To determine which step of SARS-CoV-2 S-mediated entry is affected by SERINC5, we initially examined the effect of SERINC5 on SARS-CoV-2 S binding to the ACE2 receptor. 293T-hACE2 cells were incubated on ice for 1 h with FITC-labeled SARS-CoV-2 S pseudovirions produced in the presence or absence of SERINC5. Subsequently, cells were shifted for 1 h at 37°C and analyzed by flow cytometry. We found that SERINC5 incorporation did not affect the amount of virus bound to 293T-hACE2 cells (Fig. 4a). As negative control, we used 293T cells that did not express the ACE2 receptor and as expected we saw only minimal amounts of virus bound to the cells (Fig. 4a). Thus, we concluded that SERINC5 does not influence the SARS-CoV-2 S-ACE2 interaction. SARS-CoV-2 S activation by cellular proteases is an important step of virus entry 31 . Cathepsins found in the lysosomes and endosomes are critical for SARS-CoV-2 S activation and infection of 293T-hACE2 cells 5,31 . In order to facilitate SARS-CoV-2 fusion with the plasma membrane of the target cell, SARS-CoV-2 S is initially cleaved in the producer cells by furin at the S1/S2 cleavage site. This is followed by cleavage at the S2' site, which is located proximally to the S1/2 cleavage site, by cathepsins in endosomal compartments or by TMPRSS2 on the surface of target cell 8,31,32 . Previous reports have shown that cathepsin inhibitors potently reduce SARS-CoV-2 entry 5,33 . Thus, it is possible that SERINC5 blocks SARS-CoV-2 entry by interfering with the protease-mediated activation at S2' of SARS-CoV-2 S. To address this, we used a previously described SARS-CoV-2 S, in which the furin cleavage site has been mutated and thus cleavage only occurs by either cathepsins or TMPRSS2 at the S2' site (SARS-CoV-2 S-FKO -Furin KnockOut) 34 . We co-transfected 293T cells with SARS-CoV-2 S-FKO, the lentiviral packaging plasmid and either SERINC5 or E.V. At 48 h post transfection, pseudovirions produced in the presence of SERINC5 or E.V. were concentrated by ultracentrifugation and verified by western blot for SERINC5 incorporation ( Supplementary Fig. 2) followed by incubation with cathepsin L (10 μg/ml) as previously performed 35 . For our assay, we used a cathepsin L inhibitor (SID 26681509) as a negative control. Western blots were then performed to examine cathepsin L-mediated cleavage at S2' of SARS-CoV-2 S-FKO in the presence or absence of SERINC5. We found that SARS-CoV-2 S-FKO was cleaved by cathepsin L at similar levels both in the presence of SERINC5 or E.V. (Fig. 4b).
As expected, the use of a cathepsin L inhibitor decreased the levels of cleaved SARS-CoV-2S-FKO in the presence of cathepsin L (Fig. 4b). Thus, we concluded that SERINC5 has no effect on cathepsin L-mediated cleavage of SARS-CoV-2 S and does not interfere with the proteolytic activation of SARS-CoV-2 S.
As SERINC5 does not interfere with receptor binding or the proteolytic activation of SARS-CoV-2 S, we next investigated if SERINC5 interferes with the fusion step between the virus and the target cell membrane. To address the role of SERINC5 on SARS-CoV-2 S-mediated fusion, we took advantage of the sensitive beta-lactamase-Vpr (BlaM-Vpr) fusion assay, which involves packaging of BlaM-Vpr chimeric protein in virions followed by its delivery into the cytoplasm of target cells as a result of virus-cell fusion, and has been previously used to study coronavirus fusion 36 . The transfer of BlaM-Vpr to the cytosol of the target cell is detected by enzymatic cleavage of the CCF2 dye, a fluorescent substrate of BlaM, loaded in the target cells 37 . We generated BLaM-Vpr containing SARS-CoV-2 S pseudoviruses in the presence of either SERINC5 or E.V. Following normalization for p24 CA , 293T-hACE2 cells were infected with equal amounts of SARS-CoV-2 S pseudoviruses in the presence or absence of a neutralizing anti-SARS-CoV-2 S antibody and processed by flow cytometry. We found that the SARS-CoV-2 S-mediated fusion was decreased in the presence of SERINC5 (Fig. 4c). The addition of a neutralizing antibody abrogated SARS-CoV-2 S-mediated fusion (Fig. 4c). Therefore, we conclude that SERINC5 blocks SARS-CoV-2 entry by inhibiting SARS-CoV-2 S-mediated fusion.
SARS-CoV-2 ORF7a counteracts the antiviral function of SERINC5. Many retroviruses encode factors (HIV-1 Nef) that counteract SERINC5 by blocking its incorporation in budding virions 14,22,38 . However, viral antagonists of SERINC5 from other viral families have not been identified. Consequently, we examined the SARS-CoV-2 accessory proteins, as they are implicated in blocking host antiviral genes and are important for virus pathogenesis 39,40 . For our work we focused on ORF7a, a type I transmembrane protein of 121 amino acids that is very similar (85% sequence identity) with SARS-CoV ORF7a 40 . SARS-CoV ORF7a is found in the ER and Golgi as well as in purified virions 41 . Therefore, we hypothesized that SARS-CoV-2 ORF7a may be interfering with the antiviral function of SERINC5. To determine the effect of SARS-CoV-2 ORF7a on SERINC5mediated restriction of SARS-CoV-2 entry, we co-transfected 293T cells with the aforementioned HIV-1 proviral replication defective plasmid, SARS-CoV-2 S and SERINC5 along with different amounts of SARS-CoV-2 ORF7a. Virus was concentrated by ultracentrifugation from the media of the transfected cells and western blots were performed to verify the presence of virus ( Supplementary Fig. 3a). Viruses were used to infect Calu-3 and 293T-hACE2. At 48 hpi, we measured luciferase levels and found that SARS-CoV-2 ORF7a alleviated the antiviral effect of SER-INC5 on virus infectivity in a dose-dependent manner in Calu-3  cells and 293T-hACE2 cells (Fig. 5a). HIV Nef, a known viral antagonist of SERINC5, blocks the antiviral effect of SERINC5 by preventing its incorporation in nascent virions 13,14,17 . Therefore, we performed western blots on purified pseudovirions to determine the levels of SERINC5 incorporated. In the case of SARS-CoV-2 S pseudoviruses, SARS-CoV-2 ORF7a did not prevent the incorporation of SERINC5 in virions or affect the cellular levels of SERINC5 in the cell fractions of the transfected cells (Fig. 5b).
We also examined the effect of ORF7a on SERINC5 restriction in the context of SARS-CoV-2 infection. For these studies we used wild type (WT) SARS-CoV-2 (USA-WA1/2020) and SARS-CoV-2-eGFP virus, in which the ORF7a ORF in WT SARS-CoV-2 has been deleted and replaced with the eGFP gene 42 . In agreement with previous findings, both viruses replicate similarly in 293T-hACE2 cells ( Supplementary Fig. 3b) 43 . Initially, we generated SARS-CoV-2 (WT) and SARS-CoV-2-eGFP (ΔORF7a) in the presence or absence of SERINC5 by infecting 293T-hACE2 cells expressing either SERINC5 or E.V. To determine the effect of ORF7a on the incorporation of SERINC5 in infectious SARS-CoV-2, we performed western blots on purified WT and ΔORF7a virions produced in the presence or absence of SERINC5. When evaluating the virus fraction, we noticed that ORF7a is packaged in budding virions (Fig. 5c). In addition, unlike our findings with SARS-CoV-2 S pseudoviruses, we found that deletion of ORF7a resulted in more SERINC5 packaged inside the nascent virions (Fig. 5c). Therefore, we concluded that in the context of SARS-CoV-2 infection ORF7a prevents the incorporation of SERINC5 in budding virions. To determine the effect of ORF7a on SARS-CoV-2 infectivity, we performed a virus spread assay. 293T-hACE2 cells transfected with SERINC5 or E.V. were infected with WT or ΔORF7a virus. At 24, 36, and 48 hpi RNA was isolated from the infected cells followed by RT-qPCR. We found that compared to WT virus, replication of the ΔORF7a virus was reduced in the presence of SERINC5 (Fig. 5d). To further confirm the role of ORF7a-mediated counteraction of SERINC5 restriction, we complemented ORF7a in trans in infections with the ΔORF7a virus. 293T-hACE2 cells expressing SERINC5 were transfected with or without ORF7a followed by infection with the ΔORF7a virus. Virus containing culture supernatants and cells were collected 48 hpi followed by western blots. When looking at the virus fraction, we noticed that the addition of ORF7a in trans reduced the levels of SERINC5 incorporated in the budding ΔORF7a virions (Fig. 5e). Next, we used equal amounts of the purified virus (normalized for RNA levels) to infect 293T-hACE2 cells and infectivity was measured at 6 hpi by RT-qPCR. We found that trans complementation of ORF7a rescued the infectivity of ΔORF7a virus in the presence of SERINC5 (Fig. 5f). Thus, we conclude that ORF7a alleviates SERINC5 restriction of SARS-CoV-2.
Furthermore, we examined the effect of endogenous SERINC5 on SARS-CoV-2 particle infectivity by infecting cells with either SARS-CoV-2 WT or ΔORF7a produced in SERINC5-depleted cells. siRNA-mediated depletion of SERINC5 (siS5) was carried out in Calu-3 cells, which express endogenous SERINC5 (Fig. 1b); an siControl (siCon) served as a negative control. At 48 h after siRNA transfection, cells were infected with either WT or ΔORF7a virus. After knockdown verification ( Supplementary  Fig. 4a  qPCR. Calu-3 cells were infected with equal genome copies and harvested 6 hpi. To ensure that the viral RNA levels we detect at 6 h are not due to the inoculum, we infected Mus dunni cells in parallel. We performed RT-qPCR and found that viral RNA levels at 6 hpi were similar in Calu-3 cells infected with WT SARS-CoV-2 produced in cells treated with either siControl or siSERINC5 (Fig. 6a). On the other hand, viral RNA levels were increased in cells infected with ΔORF7a virions produced in cells depleted of SERINC5 compared to viral RNA levels from cells infected with ΔORF7a virions produced in cells treated with siControl (Fig. 6a). We also examined the effect of endogenous SERINC5 on SARS-CoV-2 spread using Calu-3 cells treated with either a SERINC5 (shS5-Calu-3) or Control (shCon-Calu-3) shRNA. After knockdown verification ( Supplementary Fig. 4b), shS5-Calu-3 and shCon-Calu-3 cells were infected with equal genome copies of SARS-CoV-2 WT or ΔORF7a viruses. RNA was isolated from the infected cells at 24 and 48 hpi followed by RT-qPCR analysis to measure the viral RNA levels. Unlike what we found in 293T-hACE2 cells ( Supplementary Fig. 3b), SARS-CoV-2 ΔORF7a virus replicated at considerably lower levels than SARS-CoV-2 WT virus in Calu-3 cells (Supplementary Fig. 4c). Therefore, we have abstained from making any direct comparisons between these two viruses in Calu-3 cells. In the case of SARS-CoV-2 WT, we found that virus replicated similarly in both SERINC5 expressing (shCon-Calu-3) and SERINC5 depleted (shS5-Calu3) cells (Fig. 6b). On the other hand, SARS-CoV-2 ΔORF7a replicated at higher levels in SERINC5 depleted (sh5-Calu-3) cells compared to SERINC5 expressing (shCon-Calu-3) cells (Fig. 6c). Thus, we conclude that endogenous SERINC5 mitigates SARS-CoV-2 infection in the absence of ORF7a.  Unfortunately, we could not verify at the protein level changes in endogenous SERINC5 incorporation in virions, as there is no antibody currently available for western blots.
Mapping the anti-SERINC5 domain of ORF7a. SARS-CoV-2 ORF7a is a type I transmembrane protein of 121 amino acids 40 . ORF7a contains an immunoglobulin-like (Ig-like) ectodomain with 7 β-sheets (residues 16-96), a transmembrane domain (residues 97-116) and a short C' terminal tail (residues 117-121) 47 . Recent reports have identified a number of SARS-CoV-2 ORF7a deletions in viruses isolated from infected patients [48][49][50] . Naturally occurring deletions in ORF7a may affect its ability to counteract SERINC5. For our work we decided to focus on the deletions that did not affect the signal peptide, transmembrane or cytoplasmic domains, as such deletions affect the localization and expression of ORF7a. Therefore, we examined previously described naturally occurring deletions of ORF7a, Δ9nt, Δ18nt, Δ57nt, and Δ96nt, all found in the ORF7a Ig-like ectodomain and isolated from clinical samples of infected patients 50 . First, we introduced the naturally occurring ORF7a deletions in a codon-optimized ORF7a expression plasmid. We then confirmed that our ORF7a deletion variants continued to localize in cellular membranes by isolating membrane fractions followed by western blots probing for ORF7a and for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) to verify the purity of our membrane fractions (Supplementary Fig. 6a). Next, we co-transfected 293T cells with SARS-CoV-2 S, pHIV-1 NL ΔEnv-NanoLuc, SERINC5 and either wild type SARS-CoV-2 ORF7a, ORF7a deletion variants (ORF7aΔ9nt, ORF7aΔ18nt, ORF7aΔ57nt, ORF7aΔ96nt) or E.V. Viruses from the culture media of the transfected cells were used to infect 293T-hACE2 cells and at 48 hpi luciferase levels were determined. We observed that all SARS-CoV-2 ORF7a naturally occurring deletion variants counteracted SERINC5 similar to wild type ORF7a (Fig. 8a). Thus, the naturally occurring deletions of SARS-CoV-2 ORF7a we examined do not affect its ability to counteract SERINC5. The Ig-like ectodomain of SARS-CoV ORF7a interacts and modulates the function of a number of host factors 51 . To further understand the role of the Ig-like ectodomain of SARS-CoV-2 ORF7a in counteracting SERINC5, we generated a series of deletion mutants, in which we deleted each of the 7 β-sheets. First, we verified that the different β-sheet deletions did not alter the association of ORF7a with cellular membranes, by purifying membrane fractions from 293T cells transfected with the different β-sheet deletion ORF7a constructs, followed by western blots probing for ORF7a and GAPDH. We found that the ORF7a βsheet deletions had no effect on the membrane association of ORF7a ( Supplementary Fig. 6b). To examine the role of the ORF7a β-sheets in counteracting SERINC5, we co-transfected 293T cells with SARS-CoV-2 S, pHIV-1 NL ΔEnv-NanoLuc, SERINC5 and either wild type ORF7a, ORF7a with β-sheet deletions or E.V. Cells were harvested 48 h post transfection, viruses from the culture media of the transfected cells were used to infect 293T-hACE2 cells, and luciferase levels were measured 48 hpi. We observed that all ORF7a β-sheet deletion mutants blocked the inhibitory effect of SERINC5 similar to wild type ORF7a (Fig. 8b). Therefore, the Ig-like ectodomain is not important for the anti-SERINC5 function of ORF7a.
We also examined the role of the SARS-CoV-2 ORF7a transmembrane domain (TM) in counteracting SERINC5. To determine the importance of the ORF7a TM in its anti-SERINC5 function, we generated a chimeric SARS-CoV-2 ORF7a, in which the transmembrane domain has been replaced with that of CD4 (ORF7aTM CD4 ). We initially confirmed that ORF7aTM CD4 still associated with cellular membranes, by purifying membrane fractions from 293T cells transfected with the ORF7aTM CD4 followed by western blots probing for ORF7a and GAPDH. We found that the CD4 TM did not alter the membrane association of ORF7a (Supplementary Fig. 6c). Next, we co-transfected 293T cells with SARS-CoV-2 S, pHIV-1 NL ΔEnv-NanoLuc, SERINC5, and either wild type ORF7a, ORF7aTM CD4 or E.V. Cells were harvested 48 hours post transfection, viruses from the culture media were used to infect 293T-hACE2, and luciferase levels were measured 48 hpi. We observed that ORF7aTM CD4 did not counteract the antiviral effect of SERINC5 (Fig. 8c). Furthermore, to determine if the ORF7a TM is important for the formation the ORF7a-Spike-SERINC5 complex, we cotransfected 293T cells with SARS-CoV-2 S and SERINC5 along with either SARS-COV-2 ORF7a or ORF7aTM CD4 and performed coIPs with anti-V5 (ORF7a). We found that ORF7aTM CD4 no longer interacted with SARS-CoV-2 S or SERINC5 (Fig. 8d). Therefore, the ORF7a TM is critical for its anti-SERINC5 function.
SERINC5 blocks SARS-CoV S-mediated entry and is offset by SARS-CoV ORF7a. SARS-CoV ORF7a and SARS-CoV-2 ORF7a share 85% sequence identity 40 . As SERINC5 counteracts SARS-CoV-2 mediated entry, we speculated that due to the high sequence homology between the S and ORF7a proteins of SARS- CoV and SARS-CoV-2, the antiviral function of SERINC5 may extend to SARS-CoV. To determine the role of SERINC5 on SARS-CoV entry and the anti-SERINC5 function of SARS-CoV ORF7a, we co-transfected 293T cells with SARS-CoV S and pHIV-1 NL ΔEnv-NanoLuc, along with either SERINC5, SARS-CoV ORF7a or E.V. Cells and virus from the media of the transfected cells were harvested and processed by western blots. We observed that SARS-CoV ORF7a is incorporated in purified pseudoviruses (Fig. 9a) in agreement with previous studies 41 . Furthermore, SERINC5, similar to what we saw with SARS-CoV-2 S pseudoviruses, is also incorporated in the virions and its incorporation is not affected by the presence of SARS-CoV ORF7a (Fig. 9a). Viruses were then used to infect 293T-hACE2 cells and luciferase levels were measured 48 hpi. We found that similar to SARS-CoV-2, SERINC5 potently blocked SARS-CoV Smediated entry (Fig. 9a) and SARS-CoV ORF7a counteracted the antiviral effect of SERINC5 (Fig. 9b). Hence, we concluded that the antiviral effect of SERINC5 is not just limited to SARS-CoV-2 but applies to other coronaviruses and that both SARS-CoV and SARS-CoV-2 ORF7a counteract SERINC5. ORF7a deletions and the β-sheets of the SARS-CoV-2 ORF7a ectodomain do not affect its anti-SERINC5 function. 293T-hACE2 cells were infected with SARS-CoV-2 S pseudoviruses produced in the presence or absence of SERINC5 along with a naturally occurring deletion variants of SARS-CoV-2 ORF7a or b β-sheet deletion mutants of SARS-CoV-2 ORF7a. c The SARS-CoV-2 ORF7a transmembrane domain is critical for its anti-SERINC5 function. 293T-hACE2 cells were infected with SARS-CoV-2 S pseudoviruses produced in the presence or absence of SERINC5 along with either wild type ORF7a or ORF7-CD4 transmembrane domain chimera (ORF7aTM CD4 ). For a, b, and c, luciferase levels were normalized to p24 CA levels. The percentage (%) of relative infectivity with respect to pseudovirus produced in the presence of empty vector is shown. All results are presented as mean ± SD from 3 independent experiments. Statistical analysis among E.V. and SERINC5 + E.V. conditions were performed by one-sample t test (two-tailed) while unpaired t-test (two-tailed) was used for comparisons among SERINC5 + E.V. and SERINC5 + ORF7a groups.

Discussion
In this report, we focused on identifying a novel host restriction factor, SERINC5, of SARS-CoV-2 entry. SERINC5 is an important antiretroviral host factor both in vivo and in vitro 13,14,17 that blocks retroviral entry of a diverse number of retroviruses including HIV-1 and MLV 13,14,22 .
We found that all members of the SERINC family, except SERINC4, are expressed in lung tissue and pneumocytes, natural targets of SARS-CoV-2 18 . In addition, using, SARS-CoV-2 S pseudoviruses, VLPs and infectious SARS-CoV-2, we showed that SERINC5 is incorporated inside viral particles and SERINC5 has the most robust effect on SARS-CoV-2 entry 13,14 . In our work, the appearance of SERINC5 in our western blots varies considerably, which we attribute to technical differences in the processing of the samples from experiment to experiment.
SARS-CoV-2 binds to the hACE2 receptor on the surface of the cells followed by activation of the S protein by target cell proteases resulting in viral-host membranes fusion and viral genome release inside the cell. Here we show that SERINC5 inhibits SARS-CoV-2 entry by targeting the virus-cell membrane fusion step, as is the case with retroviruses 15 . Thus, SERINC5 uses a conserved mechanism to exert its antiviral function against two disparate viral families. SARS-CoV-2 entry is dependent on the S protein. S proteins of different SARS-CoV-2 variants are quite polymorphic and show differences in entry efficiency depending on the cell line 27 . In contrast to HIV-1, for which not all HIV-1 envelopes are restricted by SERINC5 16 , SERINC5 blocked SARS-CoV-2 entry for all S variants tested. Thus, our findings suggest that despite the diversity in S protein among the different SARS-CoV-2 variants, susceptibility to SERINC5 is never lost. More studies are needed to further explore the effect of SERINC5 on S proteins derived from different SARS-CoV-2 variants.
Retroviruses encode genes that counteract the deleterious effect of SERINC5. In the case of HIV-1, Nef prevents the incorporation of SERINC5 into budding virions and thus blocks the deleterious effect of SERINC5 on virus infection 13,14,22 . No viral proteins, outside of retroviruses, have hitherto been identified to counteract SERINC5. We identified that the SARS-CoV-2 accessory protein ORF7a, a type I transmembrane protein that is highly homologous to SARS-CoV ORF7a 40 , is a viral antagonist of SERINC5. SARS-CoV ORF7a has many functions in infected cells including induction of apoptosis 52,53 , neutralization of Bst-2/tetherin 54 and cell cycle arrest at the G/G1 phase 55 among others. Using an array of functional and biochemical assays, we show that SARS-CoV-2 ORF7a co-localizes with and binds to SARS-CoV-2 S and SER-INC5. Supporting our findings, a recent large scale proteomics analysis that suggested that SARS-CoV-2 ORF7a interacts with factors involved with lipid metabolic processes 56 , which is the cellular function of SERINC5 12 . In addition, the SARS-CoV-2 ORF7a-S interaction mirrors the previously described SARS-CoV ORF7a-S interaction 41 . Nevertheless, the biophysical mechanism by which the formation of SARS-CoV-2 ORF7a/S/SERINC5 complex interferes with the antiviral function of SERINC5 remains to be elucidated. In the case of HIV-1, it is hypothesized that SERINC5 interacts with the HIV-1 envelope, resulting in conformational remodeling of the envelope and the inhibition of HIV-1 infection [57][58][59] . Therefore, it is possible that SERINC5 affects the structural conformation of SARS-CoV-2 S and ORF7a binds to SARS-CoV-2 S to block any such conformational changes on S. The fact that SARS-CoV-2 ORF7a, S and SERINC5 are found in both pseudoviruses as well as infectious SARS-CoV-2 virions suggests that this might be happening in the virions. Moreover, we were surprised to observe that in the context of SARS-CoV-2 infection, ORF7a blocked the incorporation of SERINC5 in budding virions, unlike what we observed with SARS-CoV-2 S pseudoviruses. This discrepancy could be due to the different sites of assembly of SARS-CoV-2 S pseudoviruses vs. infectious SARS-CoV-2 (plasma membrane vs ERGIC). In conclusion, our findings suggest that ORF7a counteracts SERINC5 in two ways; in the producer cells ORF7a prevents the incorporation of SERINC5 into SARS-CoV-2 virions and in the released virions ORF7a forms a complex with S and SERINC5 to block the antiviral effect of SERINC5 on virus-cell membrane fusion. This is in agreement with the idea that exclusion of SERINC5 from  virions is important but not adequate for alleviation of restriction of virus infection. Finally, the fact that SARS-CoV is also restricted by SERINC5 further emphasizes the importance of this host antiviral factor on coronavirus infection. In conclusion, the antiviral effect of SERINC5 on a diverse number of viruses and the presence of SERINC5 viral antagonists demonstrate that its interplay with viral proteins is an attractive target for the development of antiviral therapies.  60 were obtained from Paul Bates. The HIV-1 NL4-3 ΔEnv-NanoLuc, pCMV SARS-CoV-2 SΔ19 were obtained from Paul Bieniasz 21 . The pFB-luc construct has been previously described 61 . The following reagent was obtained through the NIH HIV Reagent Program, Division of AIDS, NIAID, NIH: HIV-1 YU2 Vpr β-lactamase expression vector (ARP-11444) contributed by Dr. Michael Miller (Merck Research Laboratories). SARS-CoV-2 E and N were PCR amplified from cDNA derived from genomic RNA of SARS-CoV-2 isolate USA-WA1/2020 (BEI Resources, NIAID, NIH, NR-52285) using primers 5ʹ-CCCAAGCTTATGTACTCATTCGTTTCG-3ʹ/5ʹ-CCGCTCGAGGACCAGA AGATCAGGAACTC-3ʹ for E and 5ʹ-CGCGGATCCGCCATGTCTGATAA TGGACCCCAAAATCAG-3ʹ/5ʹ-AAAAGCGGCCGCCGGCCTGAGTTGAGTC AGC-3ʹ for N followed by cloning into pCDNA3.1/myc-His A (Invitrogen) and pCDNA-V5/His TOPO (Invitrogen) respectively. Codon optimized SARS-CoV-2 ORF7a was PCR amplified from pLVX-EF1alpha-SARS-CoV-2-ORF7a-2xStrep-IRES-Puro plasmid, (Addgene, 141388, deposited by Nevan Krogan) using primers 5ʹ-CCCAAGCTTGCCGCCACCATGAAGATCA-3ʹ/5ʹ-CCGCTCGAGCGCT-CAGTCTTTCTTTTCAGTG-3ʹ followed by cloning into pCDNA-V5/His TOPO (Invitrogen). This codon optimized SARS-CoV-2 ORF7a (WT) in pCDNA-V5/His TOPO plasmid was used as a template to generate SARS-CoV-2 ORF7a deletion variants using the Phusion SDM kit (Thermo Fischer Scientific) and primers listed in Supplementary Table 1, and to obtain a SARS-CoV-2 ORF7a expression plasmid with a V5 tag at the Nʹ terminus, in which the V5 tag was placed immediately downstream of the signal peptide sequence followed by the removal of the C-terminal V5/His tag using the Phusion SDM kit and primers listed in Supplementary Table 1. SARS-CoV-2 ORF7aTM CD4 was created by replacing the transmembrane domain of ORF7a with that of CD4 using the NEBuilder HiFi DNA assembly kit (New England Biolabs) with the following primers (CD4TM_7a-F: 5'-GAGGTGCAAGAGATGGCCCTGATTGTGCTG-3'/ CD4TM_7a-R: 5ʹ-AGTCTTTCTTTTGAAGAAGATGCCTAGCCCAATG-3ʹ, 7a_delTM-F: 5ʹ-GGCATCTTCTTCAAAAGAAAGACTGAGCGCTC-3ʹ/ 7a_delTM-R: 5ʹ-AATCAGGGCCATCTCTTGCACCTCCTCTTG-3ʹ). pLVX-EF1a-EGFP-ERGIC53-IRES-Puromycin was acquired from Addgene (134859, deposited by David Andrews). pCMV6-TMED2-FLAG was acquired from Ori-Gene (RC206849). of viral RNA copies, 100 µl of culture supernatant was used for RNA isolation followed by cDNA synthesis and RT-qPCR. The following primers were used for SARS-CoV-2 S detection: 5ʹ-CCTACTAAATTAAATGATCTCTGCTTTACT-3ʹ/ 5ʹ-CAAGCTATAACGCAGCCTGTA-3ʹ.

Methods
SARS-CoV-2 infectious viruses containing SERINC3 or SERINC5 were generated by infecting 293T-hACE2 cells transfected with SERINC3 or SERINC5 with SARS-CoV-2 virus (isolate USA-WA1/2020). Cells and culture supernatants were harvested 48 hpi. Cells were lysed in 1× RIPA buffer and processed for immunoblotting (see Immunoblotting section). Culture supernatants were cleared of cell debris and aliquots were either stored at −80°C for future experiments or pelleted through a 30% sucrose cushion followed by lysis in 1× RIPA buffer for immunobotting.
Pseudovirus entry assays. For infections using the SARS-CoV-2 S MLV pseudotypes, 293T-hACE2 cells (2.5 × 10 4 cells/well) were seeded in a 24-well plate. Cells were infected the following day and lysed 48 hpi followed by measuring luminescence using the Steady-Glo luciferase assay system (Promega) per manufacturer's recommendation and Biostack4 (BioTek) luminometer, an automated plate reader. For SARS-CoV-2 S/CoV S HIV pseudotypes, Calu-3 or 293T-hACE2 cells (2.5 × 10 4 cells/well) were seeded in a 96-well plate. Cells were infected the next day and lysed 48 hpi followed by measuring luminescence using the Nano-Glo luciferase system (Promega) per manufacturer's recommendation and a Biostack4 (BioTek) luminometer. Infectivity was determined by normalizing the luciferase signals to virus levels as determined by western blots probing for MLV p30 CA or HIV p24 CA on culture supernatants (see Immunoblotting section).
Binding assay with FITC-labeled SARS-CoV-2 S pseudovirus. 293T-hACE2 or 293T cells (0.2 × 10 6 cells/well) were seeded in a 24-well plate. Concentrated SARS-CoV-2 S HIV pseudotyped viruses generated as mentioned above (see Pseudovirus production section) were labeled with FITC using EzLabel protein FITC Labeling Kit (BioVision) per manufacturer's recommendation. After p24 CA normalization, labeled virions were added to cells and incubated on ice for 1 h. Cells were shifted to 1 h at 37°C, washed 3× in ice cold 1× PBS, lifted using 100 μl of Versene (Gibco), fixed with 4% paraformaldehyde for 10 min at 4°C, and acquired on BD LSRFortessa using BD FACSDiva 8.0.2 followed by analysis using FlowJo version 10.8.0.
Cathepsin L treatment experiment. SARS-CoV-2 S FKO-FLAG pseudotyped HIV viruses were generated in the presence or absence of SERINC5-HA as mentioned above (see Pseudovirus production section). As previously described 35 , virus pellets were resuspended in 100 μl of 1× PBS with Ca 2+ and Mg 2+ , pH 5.5 and either treated with PBS (mock) or human Cathepsin L (10 μg/ml, Millipore Sigma, 219402) for 1 h at room temperature (RT). We used 20 μM of Cathepsin L inhibitor SID 26681509 (MedChemExpress) as a control. Virus supernatants were then resolved on 8% SDS-PAGE gels. SARS-CoV-2 S FKO was detected using rabbit anti-FLAG as described above (see Immunoblotting section).
BlaM-Vpr fusion assay. To generate BlaM-Vpr containing SARS-CoV-2 S HIV pseudotyped viruses in the presence or absence of SERINC5, transfections were performed as described above (see Pseudovirus production section) along with 2.5 μg of HIV-1 YU2 Vpr β-lactamase expression plasmid. Culture supernatants were harvested 24, 48, and 72 h post transfection, pooled, and pelleted by ultracentrifugation at 133,900 × g for 2.5 h at 4°C. Pseudoviruses were resuspended and stored at −80°C. Virion fusion assays were performed as previously described 36,37 with some modifications. Briefly, 293T-hACE2 cells (5 × 10 4 cells cells/well) were seeded in a 96-well plate a day prior to infection. Cells were infected with 150 ng of p24 CA equivalent virions per well, centrifuged at 257 × g for 2 h at 22°C and further incubated at 37°C for 1 h. When using a neutralizing antibody, cells were infected with virions pre-incubated for 1 h at 37°C in the presence of a monoclonal anti-SARS-CoV-2 Spike glycoprotein receptor binding domain, chimeric potent neutralizing antibody (1 µM) (BEI Resources, NIH, NIAID, NR-55410). Cells were washed in 200 μl of CO 2 -independent media (Gibco) by centrifuging at 365 × g for 5 min at RT and were resuspended in 100 μl of CO 2 -independent media followed by staining with 20 μl of 6× CCF2-AM substrate loading solution (Invitrogen, K1032) for 1 h at RT in the dark. Cells were then washed twice in 250 μl of development media [CO 2 -independent media containing 2.5 mM probenecid (Sigma)] by centrifuging at 365 × g for 5 min at RT, resuspended in 400 μl of development media and incubated at RT for 16 h in the dark. Subsequently, cells were washed, fixed and the change in emission fluorescence of CCF2-AM after cleavage by the BlaM-Vpr post-virion fusion was monitored via BD LSRFortessa using BD FACSDiva 8.0.2 followed by data analysis using FlowJo version 10.8.0.