Immunization with Recombinant Accessory Protein-Deficient SARS-CoV-2 Protects against Lethal Challenge and Viral Transmission

ABSTRACT Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has led to a worldwide coronavirus disease 2019 (COVID-19) pandemic. Despite the high efficacy of the authorized vaccines, there may be uncertain and unknown side effects or disadvantages associated with current vaccination approaches. Live-attenuated vaccines (LAVs) have been shown to elicit robust and long-term protection by the induction of host innate and adaptive immune responses. In this study, we sought to verify an attenuation strategy by generating 3 double open reading frame (ORF)-deficient recombinant SARS-CoV-2s (rSARS-CoV-2s) simultaneously lacking two accessory ORF proteins (ORF3a/ORF6, ORF3a/ORF7a, and ORF3a/ORF7b). We report that these double ORF-deficient rSARS-CoV-2s have slower replication kinetics and reduced fitness in cultured cells compared with their parental wild-type (WT) counterpart. Importantly, these double ORF-deficient rSARS-CoV-2s showed attenuation in both K18 hACE2 transgenic mice and golden Syrian hamsters. A single intranasal dose vaccination induced high levels of neutralizing antibodies against SARS-CoV-2 and some variants of concern and activated viral component-specific T cell responses. Notably, double ORF-deficient rSARS-CoV-2s were able to protect, as determined by the inhibition of viral replication, shedding, and transmission, against challenge with SARS-CoV-2 in both K18 hACE2 mice and golden Syrian hamsters. Collectively, our results demonstrate the feasibility of implementing the double ORF-deficient strategy to develop safe, immunogenic, and protective LAVs to prevent SARS-CoV-2 infection and associated COVID-19. IMPORTANCE Live-attenuated vaccines (LAVs) are able to induce robust immune responses, including both humoral and cellular immunity, representing a very promising option to provide broad and long-term immunity. To develop LAVs for SARS-CoV-2, we engineered attenuated recombinant SARS-CoV-2 (rSARS-CoV-2) that simultaneously lacks the viral open reading frame 3a (ORF3a) in combination with either ORF6, ORF7a, or ORF7b (Δ3a/Δ6, Δ3a/Δ7a, and Δ3a/Δ7b, respectively) proteins. Among them, the rSARS-CoV-2 Δ3a/Δ7b was completely attenuated and able to provide 100% protection against an otherwise lethal challenge in K18 hACE2 transgenic mice. Moreover, the rSARS-CoV-2 Δ3a/Δ7b conferred protection against viral transmission between golden Syrian hamsters.

challenge with parental SARS-CoV-2. Similarly, hamsters vaccinated with the 3 double ORFdeficient rSARS-CoV-2s mounted robust levels of neutralizing antibodies without leading to any body weight changes. Importantly, upon challenge with the parental SARS-CoV-2, all vaccinated hamsters showed decreased viral replication, shedding, and transmission.
Together, our results demonstrate that the double ORF-deficient strategy, particularly, the D3a/D7b double deletion we have developed in this study, is promising for the development of safe, immunogenic, and protective LAVs for the prophylactic treatment against SARS-CoV-2 infection and associated COVID-19 disease.
Having validated the double ORF-deficient rSARS-CoV-2, we evaluated their plaque morphology in Vero E6 cells at 24, 48, 72, and 96 h postinfection (hpi) (Fig. 2A). All 3 . Fragments in the shuttle plasmids that contain a single deletion of the viral ORF6, ORF7a, and ORF7b were released using PpuMI and XhoI restriction enzymes and then were ligated to the PpuMI-XhoI-linearized vector backbone that contains the ORF3a deletion. The fragments in the resultant shuttle plasmids that contain a double deletion of ORF3a/ORF6, ORF3a/ORF7a, and ORF3a/ORF7b were released using BamHI and RsrII digestion and reassembled into the BAC that was linearized with the same restriction enzymes. PCR-positive BAC plasmid colonies were prepared and transfected into Vero E6 cells for virus rescue. (B) Confirmation of the double ORF-deficient rSARS-CoV-2 strain by RT-PCR amplification of ORF3a, ORF6, ORF7a, ORF7b, and N. (C) Deep sequencing analysis of the double ORF-deficient rSARS-CoV-2 P1 genome. Nonreference alleles present in less than 10% of reads are not shown. Amino acid changes respective to rSARS-CoV-2 WT are indicated. (D) Deep sequencing analysis of the double ORF-deficient rSARS-CoV-2 P10 genome. Nonreference alleles present in less than 10% of reads are not shown. Amino acid changes respective to rSARS-CoV-2 WT are indicated.
Characterization of double ORF-deficient rSARS-CoV-2 in K18 hACE2 transgenic mice. We next evaluated the replication and pathogenesis of the double ORF-deficient rSARS-CoV-2 in transgenic mice that express human angiotensin converting enzyme 2 (K18 hACE2), which are a common model to use to study SARS-CoV-2 pathogenesis (31,32). To that end, 5-week-old female K18 hACE2 transgenic mice were infected with the double ORF-deficient rSARS-CoV-2 strain using a dose of 2 Â 10 5 PFU/mouse (Fig. 3A). Mock-infected K18 hACE2 mice and mice infected with the same dose of rSARS-CoV2 WT were used as internal controls. The lungs of infected K18 hACE2 transgenic mice were excised, and gross pathological lesions were analyzed at 2 and 4 days postinfection (dpi) (Fig. 3B). At 2 dpi, the lungs of K18 hACE2 transgenic mice infected with rSARS-CoV-2 WT contained pathological lesions in ;35% of the total lung area, whereas the lungs of K18 hACE2 transgenic mice infected with the double ORF- Six plaques were randomly selected and measured using a standard ruler (millimeters, mm). Data are presented as mean 6 SD, and comparisons of diameter means between indicated groups were performed by one-way ANOVA. *, P , 0.05; **, P , 0.01; and ns, not significant. (C) Growth kinetics. Viral growth kinetics of the WT and double ORF-deficient rSARS-CoV-2 strains in Vero E6 (left) and A549-hACE2 (right) cells were measured in triplicate by plaque assay. Dotted lines indicate the limit of detection (LOD). Data are presented as mean 6 SEM, and viral titer means of the supernatant collected from the double ORF-deficient rSARS-CoV-2-infected cells are compared with that of the rSARS-CoV-2 WT-infected cells by one-way ANOVA. *, P , 0.05; and **, P , 0.01. Double Deletion Strategy To Generate Live-Attenuated SARS-CoV-2 Microbiology Spectrum deficient rSARS-CoV-2 strain had lesions in #25% of the total lung area with 19% for rSARS-CoV-2 D3a/D6, 23% for rSARS-CoV-2 D3a/D7a, and 12% for rSARS-CoV-2 D3a/ D7b. By 4 dpi, rSARS-CoV-2 WT had induced lesions in ;60% of the total lung area, while double ORF-deficient rSARS-CoV-2 had reduced lesions with ;40% for rSARS-CoV-2 D3a/D6, ;55% for rSARS-CoV-2 D3a/D7a, and ;40% for rSARS-CoV-2 D3a/D7b (Fig. 3C). We also evaluated viral replication of the double ORF-deficient rSARS-CoV-2 strain in the lungs and nasal turbinate of the infected K18 hACE2 transgenic mice. Compared with rSARS-CoV-2 WT, all double ORF-deficient rSARS-CoV-2s replicated to significantly lower titers in the lungs and nasal turbinates of infected mice at both 2 transgenic mice mock infected or infected (2 Â 10 5 PFU/mouse) with the indicated rSARS-CoV-2 strain at 2 and 4 dpi (n = 4/group). (C) Gross pathological lesion scoring on lung images in B using NIH ImageJ. Data are presented as mean 6 SD, and comparisons of means between indicated groups are analyzed by one-way ANOVA. *, P , 0.05; **, P , 0.01; and ns, not significant. (D) Viral titers in the clarified homogenate of lungs (left) and nasal turbinate (right) of K18 hACE2 transgenic mice infected in B at 2 and 4 dpi. The viral titers in the supernatant of the homogenate were determined in triplicate by plaque assay. Data are presented as mean 6 SEM, and comparisons of the means between indicated groups were analyzed by one-way ANOVA. *, P , 0.05; and **, P , 0.01. (E) Body weight changes in K18 hACE2 transgenic mice mock infected or infected (2 Â 10 5 PFU/mouse, n = 5/group) with the indicated WT or double ORF-deficient rSARS-CoV-2. (F) Survival curves of K18 hACE2 transgenic mice infected in E were calculated and plotted using daily observations for 21 days. The Kaplan-Meier survival analysis with a log rank (Mantel-Cox) test was applied to compare overall survival time. *, P , 0.05; and **, P , 0.01. (G) The IgG (sera) and IgA (BALF) against the full-length S glycoprotein in mice that survived in F were tested in triplicate by ELISA at 21 dpi. Sera and BALF collected from the two surviving K18 hACE2 transgenic mice infected with rSARS-CoV-2 WT (10 3 PFU/mouse, n = 5) for 21 days were included as a positive control. Data are presented as mean 6 SEM, and means of the double ORF-deficient rSARS-CoV-2 groups are compared with that of the rSARS-CoV-2 WT group by one-way ANOVA. *, P , 0.05; **, P , 0.01; and ns, not significant. (H) Splenocytes were isolated from the mice that survived in F at 21 dpi, and IFN-g -specific spot-forming cells (SFCs) were counted (duplicate) after stimulation with peptide pools of S1, S2, and N using flow cytometry. The splenocytes isolated from the two surviving K18 hACE2 mice infected with rSARS-CoV-2 WT (10 3 PFU/mouse, n = 5) for 21 days were included as a positive control. Data are presented as mean 6 SEM, and comparisons of the means between indicated groups are analyzed by one-way ANOVA. *, P , 0.05; **, P , 0.01; and ns, not significant.
Protection of rSARS-CoV-2 D3a/D7b-vaccinated K18 hACE2 transgenic mice against lethal challenge with SARS-CoV-2. Since infection with rSARS-CoV-2 D3a/ D7b was not lethal in K18 hACE2 transgenic mice ( Fig. 3E and F), but it induced robust humoral (Fig. 3G) and cellular ( Fig. 3H) immunity, we hypothesized that the K18 hACE2 transgenic mice vaccinated with rSARS-CoV-2 D3a/D7b would survive a lethal challenge with wild-type virus, confirming the feasibility of rSARS-CoV-2 D3a/D7b as an LAV strategy. To test this hypothesis, 5-week-old female K18 hACE2 transgenic mice were either mock vaccinated or rSARS-CoV-2 D3a/D7b vaccinated with 2 Â 10 5 PFU/ mouse and challenged intranasally with 10 5 PFU/mouse of rSARS-CoV-2 mCherryNluc at 21 days postvaccination (Fig. 4A). Viral replication was evaluated using a noninvasive in vivo imaging system (IVIS) in the whole organism (Nluc), as described (33). A strong Nluc signal in the lungs of mock-vaccinated mice was detected at 2 and 4 days postchallenge with rSARS-CoV-2 mCherryNluc, whereas no Nluc signal was detected in the lungs of K18 hACE2 transgenic mice vaccinated with rSARS-CoV-2 D3a/D7b ( Fig. 4B and C). In excised lungs, the expression of mCherry was detected readily in mock-vaccinated mice, while no mCherry was expressed in the lungs of mice vaccinated with rSARS-CoV-2 D3a/D7b (see Fig. S2A and S2B in the supplemental material). Moreover, pathological lesions on the lung surface were much more severe in mock-vaccinated mice than those in rSARS-CoV-2 D3a/D7b-vaccinated mice ( Fig. S2C and S2D). Supernatants of the lung and nasal turbinate homogenates from rSARS-CoV-2 D3a/D7b-vaccinated mice did not contain infectious virus at both 2 and 4 days postchallenge (Fig. 4D). Significantly decreased Nluc activity was further seen in the supernatants of the lung and nasal turbinate homogenates from rSARS-CoV-2 D3a/D7b-vaccinated K18 hACE2 transgenic mice at both 2 and 4 days postchallenge compared with that of mockvaccinate mice (Fig. 4E), which correlated with the in vivo imaging data. In the lungs of the mock-vaccinated mice, a significant production of IFN-a and IFN-g was induced after challenge of rSARS-CoV-2 mCherryNluc (Fig. S2E). In contrast, the IFN responses were not induced in the lungs of rSARS-CoV-2 D3a/D7b-vaccinated mice at either 2 or 4 days postchallenge, whereas an elevated production of tumor necrosis factor alpha (TNF-a), which is highly related to a protective Th17 response, was induced at 4 days postchallenge (Fig. S2E).
Double ORF-deficient rSARS-CoV-2 vaccination prevents viral replication and shedding in hamsters. Next, we sought to investigate the safety and protective efficacy of the double ORF-deficient rSARS-CoV-2 in hamsters, an animal model that  (34). Five-week-old female golden Syrian hamsters were mock vaccinated or vaccinated (4 Â 10 5 PFU) with the double ORF-deficient rSARS-CoV-2 strain to evaluate viral replication and were monitored daily for body weight (Fig. 5A). The left lung lobes of infected hamsters were collected at 2 and 4 dpi and sectioned to assess inflammation and immunopathology by using hematoxylin and eosin (H&E) staining. The bronchointerstitial pneumonia was focused primarily around bronchioles, terminal airways, and blood vessels, and the severity of pneumonia increased over time (Fig. 5B). However, infection with the double ORF-deficient mutant viruses resulted in a marked reduction in inflammation and reduced severity compared with animals infected with rSARS-CoV-2 WT (Fig. 5C). Furthermore, we also evaluated viral replication of the double ORF-deficient rSARS-CoV-2 strain in the lung and the nasal turbinate of the infected hamsters. Compared with rSARS-CoV-2 WT, all double ORF-deficient rSARS-CoV-2 strains replicated to significantly lower titers in the lungs and nasal turbinate of infected hamsters at both 2 and 4 dpi (Fig. 5D). In addition, infection with rSARS-CoV-2 WT led to an ;15% body weight loss by 6 dpi, and no changes in body weight were observed in hamsters infected with these double ORF-deficient rSARS-CoV-2s, whose body weight were comparable to mock-infected animals at all time points (Fig. 5E). After 21 days, all vaccinated hamsters were challenged with 2 Â 10 5 PFU of rSARS-CoV-2 mCherryNluc. To assess viral shedding and transmission, each challenge hamster (donor) was housed in the same cage with a susceptible contact hamster at 1 day postchallenge (Fig. 5A). The Nluc signal in all donor and contact hamsters was determined at 2 and 4 days postchallenge. In the donor hamsters, the Nluc signal was detected in the nasal turbinates and lungs of mock-vaccinated hamsters at both time points; to a lesser extent, the Nluc signal was also present in nasal turbinates, but not in the lungs, of the double ORF-deficient rSARS-CoV-2-vaccinated hamsters at 2 days postchallenge, and no Nluc was detected in all of these hamsters at 4 days postchallenge ( Fig. 5F and G). In contact hamsters, the Nluc signal was absent in any contact hamsters at 2 and 4 days postchallenge but was readily detectable in all hamsters in contact with mock-vaccinated hamsters at 4 days postchallenge ( Fig. 5F and G). After collecting lungs at 4 days postchallenge, we noted strong mCherry expression in the lungs of mock-vaccinated and rSARS-CoV-2 mCherryNluc-infected donor hamsters and their contacts, whereas mCherry fluorescence was significantly decreased in all vaccinated donor hamsters and their respective contacts (see Fig. S3A and S3B in the supplemental material). We next determined virus load in clarified supernatants of lung and nasal turbinate homogenates from both donor and contact hamsters. No detectable infectious virus was present in either tissue in any of the donor hamsters vaccinated with the double ORF-deficient rSARS-CoV-2 strain (Fig. 5H). All contact hamsters were free of virus, except for one contact hamster (;10 2 PFU/mL) that was cohoused with a hamster vaccinated with rSARS-CoV-2 D3a/D6 (Fig. 5H). We obtained equivalent results when following Nluc activity in the clarified supernatant of lung and nasal turbinate homogenates (Fig. 5I).
Double ORF-deficient rSARS-CoV-2 vaccination prevents transmission in hamsters. Since double ORF-deficient rSARS-CoV-2 vaccination significantly reduced viral replication and shedding after challenge, we sought to explore whether the double ORF-deficient rSARS-CoV-2 strain can provide protection against viral transmission between vaccinated contact hamsters and infected donors. Five-week-old female golden Syrian hamsters were vaccinated with the double ORF-deficient rSARS-CoV-2 strain (4 Â 10 5 PFU), and sera were collected at 18 days postvaccination. Vaccinated contact hamsters were housed with rSARS-CoV-2 mCherryNluc-infected donor hamsters at 21 days postvaccination, and all hamsters were analyzed by in vivo imaging and necropsy (Fig. 6A). Sera collected from the double ORF-deficient rSARS-CoV-2-vaccinated hamsters showed a high neutralizing potential against the SARS-CoV-2 WA1 strain and different variants of concern (VOCs) (alpha, a; beta, b; delta, d ; and omicron, ο) ( Fig. 6B and C). After rSARS-CoV-2 mCherryNluc-infected (2 Â 10 5 PFU) donor hamsters were cohoused with the double ORF-deficient rSARS-CoV-2vaccinated contact hamsters, the Nluc signal was readily detected in all donor hamsters at 2 and 4 dpi, whereas no detectable Nluc signal was observed in any of the contact animals at 2 dpi. At 4 dpi, high levels of Nluc signal were present in all mock-vaccinated contact hamsters, but signal was extremely low in all double ORF-deficient rSARS-CoV-2-vaccinated contact hamsters ( Fig. 6D and E). In the lungs excised at 4 dpi, mCherry expression was readily detected in all donor hamsters and mock-vaccinated contact hamsters but not in any of double ORF-deficient rSARS-CoV-2-vaccinated contact hamsters (see Fig. S4A and S4B in the supplemental material). When titrating infectious particles in the clarified lung and nasal turbinate homogenates, we found that no infectivity was present in tissues derived from any of the contacts cohoused with the hamsters vaccinated with the double ORF-deficient rSARS-CoV-2 strain (Fig. 6F). Consistent with the viral titer results, Nluc activity in the clarified lung and nasal turbinate homogenates was significantly decreased in all double ORF-deficient rSARS-CoV-2-vaccinated contact hamsters compared with that present seen in mock-vaccinated contacts (Fig. 6G).

DISCUSSION
LAVs, which can induce broad and long-term host immune responses and activate both innate and adaptive immunity, are attractive for prophylaxes against emerging Double Deletion Strategy To Generate Live-Attenuated SARS-CoV-2 Microbiology Spectrum and re-emerging pathogens that are undergoing rapid antigenic drift, such as SARS-CoV-2 (35). In this study, we used our established BAC-based reverse genetic system (30) to generate 3 potential SARS-CoV-2 LAV candidates by deleting two accessory protein ORFs simultaneously from the viral genome (ORF3a/ORF6, ORF3a/ORF7a, and ORF3a/ORF7b). After being passaged 9 times in Vero E6 cells, the three double ORF-deficient rSARS-CoV-2 strains did not revert, suggesting that the genomes of the double ORF-deficient SARS-CoV-2s are stable in vitro. Results also showed that double ORF-deficient rSARS-CoV-2s grow efficiently in Vero E6 cells, which is a cell line approved by the United States FDA for vaccine manufacture and production. K18 hACE2 transgenic mice were first developed for in vivo pathogenicity studies of SARS-CoV (36). Although K18 hACE2 transgenic mice display morbidity and mortality, including efficient replication in the upper and lower respiratory tract and brain, upon SARS-CoV-2 infection (37), it is still unknown how much this K18 hACE2 transgenic mouse model recapitulates a true infection in humans. Since the natural isolate of one of the first SARS-CoV-2 WA1 strains exhibits an MLD 50 of ;500 PFU (38), we favor the use of the K18 hACE2 transgenic mouse model for testing the safety of live-attenuated rSARS-CoV-2. Using K18 hACE2 transgenic mice, we show that our most attenuated double ORF-deficient virus, rSARS-CoV-2 D3a/D7b, has an MLD 50 greater than 2 Â 10 5 PFU, and rSARS-CoV-2 D3a/D6 showed a MLD 50 of ;2 Â 10 5 PFU, which is about 400fold higher than its WT counterpart, suggesting an adequate attenuation of these two rSARS-CoV-2 D3a/D6 and rSARS-CoV-2 D3a/D7b LAV candidates. Although all mice infected with rSARS-CoV-2 D3a/D7b (2 Â 10 5 PFU) succumbed to infection, they still showed a greater survival time than K18 hACE2 transgenic mice infected with rSARS-CoV-2 WT. In addition, we noticed that pathological lesions were observed on the lungs of the 3 double ORF-deficient rSARS-CoV-2-infected K18 hACE2 mice and hamsters at both 2 and 4 dpi, although they were reduced compared with those of the rSARS-CoV-2 WT. However, viral replication of all the double ORF-deficient rSARS-CoV-2s was significantly reduced compared with that of rSARS-CoV-2 WT.
In the case of rSARS-CoV-2 D3a/D7b-infected K18 hACE2 transgenic mice, 2 of the 5 infected mice had some minimal body weight loss before 6 dpi. Of note, the fully attenuated rSARS-CoV-2 D3a/D7b strain was able to provide, upon a single intranasal administration, a protective effect in K18 hACE2 transgenic mice challenged with a lethal dose of rSARS-CoV-2 mCherryNluc, and 100% of the rSARS-CoV-2 D3a/D7b-vaccinated K18 hACE2 transgenic mice survived the challenge with SARS-CoV-2, suggesting that even the most attenuated virus could confer complete protection against a lethal challenge of SARS-CoV-2. The S-specific IgA in the BALF may also contribute to this major protection. However, rSARS-CoV-2 D3a/D7b-vaccinated mice produced lower levels of IFN responses accompanied with significant decreases in chemokine production and a lower interleukin-6 (IL-6)/IL-10 ratio. These results are consistent with an attenuated cytokine storm, which may explain the reduction in tissue damage observed and the better control of the infection by mice. The later Th1 response observed at 4 dpi indicates that rSARS-CoV-2 D3a/D7b-vaccinated mice might attenuate the rapid evolution of the detrimental cytokine response observed in the mockvaccinated mice after challenge of rSARS-CoV-2 mCherryNluc.
Our hamster studies demonstrate that all 3 double ORF-deficient rSARS-CoV-2 strains are highly attenuated and that infection with 4 Â 10 5 PFU results in mild pathological lesions on the lungs without body weight loss, contrary to results with hamsters infected with rSARS-CoV-2 WT, where animals lost up to 15% of their initial body weight by day 6 and had severe pathological lesions, as described previously (30). Furthermore, virus replication in the double ORF-deficient rSARS-CoV-2-vaccinated hamsters after challenge was detected only transiently in the upper respiratory tract (nasal turbinate) at 2 but not at 4 dpi, which is very important for avoiding respiratory complications and pneumonia progression (39). Importantly, viral shedding was impeded in the double ORF-deficient rSARS-CoV-2-vaccinated hamsters as well, as suggested by a lack of viral replication at both 2 and 4 days postchallenge in susceptible contact hamsters. Moreover, vaccination with the double ORF-deficient rSARS-CoV-2s completely abrogated viral transmission from infected donor hamsters, as virus replication was completely undetectable in the vaccinated contact hamsters. The protection could be attributed to the high levels of neutralizing antibody found in sera, including neutralizing antibodies against different VOCs, as well as an activated T cell response. Interestingly, reduced virus replication was seen in infected donor hamsters, which were housed with double ORF-deficient rSARS-CoV-2-vaccinated contact hamsters. We speculate that the close proximity of the vaccinated contact hamster may allow the spread of virus-specific IgA through saliva, feces, and aerosols, which may neutralize virus replication in infected donors. A significant advantage of the double ORF deletion attenuation strategy proposed in our study over other LAVs is that the potential of viral reversion is much lower. Another advantage of our ORF-deficient LAV approach is that other viral S proteins could be expressed instead of the deleted ORF3a and/or ORF7b proteins for the development of bivalent and/or multivalent LAVs, respectively. Together, our data show that the three double ORF-deficient rSARS-CoV-2s, which express all viral structural proteins and yet lack two different accessory protein combinations, are attenuated to different degrees. All double ORF-deficient rSARS-CoV-2 strains induce robust innate and adaptive immune responses, including mucosal immunity, upon a single intranasal administration that protects against subsequent challenge with SARS-CoV-2. Based on the safety profile in K18 hACE2 mice and golden Syrian hamsters, we favor the use of the ORF3a/ORF7b-deficient rSARS-CoV-2 as a LAV for the treatment of SARS-CoV-2 infection. Although the double ORF-deficient rSARS-CoV-2 strain based on the WA1 strain backbone did not induce a comparable level of neutralization antibodies against Omicron compared with other VOCs, it is still reasonable to conclude that the ORF3a/ORF7b-deficient strategy is attractive for the development of SARS-CoV-2 LAVs, as it could be updated easily by substituting the viral S glycoprotein. Alternatively, it will be possible to develop bivalent LAVs by genetically engineering the ORF3a/7b rSARS-CoV-2 to encode additional S proteins from the locus of the viral ORF3a and/or ORF7b proteins. One of the limitations of our current study is the safety profile of the ORF3a/ORF7b-deficient rSARS-CoV-2 strain in immunocompromised hosts. Future experiments in immunocompromised animals are guaranteed to further demonstrate the safety profile of the rSARS-CoV-2 DORF3a/7b. Combining attenuation with high immunogenicity and feasibility of update makes the ORF3a/7b-deficient virus an attractive option for its use as safe and protective LAV candidate for prophylaxis of SARS-CoV-2 infection and associated COVID-19 disease. Moreover, the double ORF-deficient SARS-CoV-2 strain represents a valid viral surrogate that could be safely used under less restricted biosafety level 2 (BSL2) laboratories to accelerate research on this important human pathogen, including the identification and characterization of therapeutics or the generation of mutant viruses without gain-of-function safety concerns.

MATERIALS AND METHODS
Biosafety. All in vitro and in vivo experiments with infectious natural isolates or rSARS-CoV-2s were conducted under appropriate biosafety level 3 (BSL3) and animal BSL3 (ABSL3) laboratories, respectively, at the Texas Biomedical Research Institute. All experiments were approved by the Texas Biomed Institutional Biosafety (IBC) and Animal Care and Use Committees (IACUC).
Cells, peptides, proteins, antibodies, and viruses. African green monkey kidney epithelial cells (Vero E6, CRL-1586) were obtained from the American Type Culture Collection (ATCC; Bethesda, MD), and a A549 cell line expressing hACE2 and a Vero E6 cell line expressing hACE2 and TMPRSS2 (Vero AT) were obtained from BEI Resources (NR-53821 and NR-54970). Cells were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 5% (vol/vol) fetal bovine serum (FBS; VWR) and 1% penicillinstreptomycin (Corning). For Vero AT, cells were treated every other passage with 5 mg/mL of puromycin for selection of AT-expressing cells.
A set of 181 peptides spanning the complete S protein of the USA-WA1/2020 strain of SARS-CoV-2 and a set of 59 peptides spanning the complete N protein of USA-WA1/2020 of SARS-CoV-2 were obtained from BEI Resources (NR-52402 and NR-52404, respectively). These peptides are 13 to 20 amino acids long, with 10 overlapping amino acids. The S2 peptide pools contain 93 peptides representing the N-terminal half of the S protein (MFVFLVLLPL to AEHVNNSYE) and the S2 peptide pools contain 88 peptides representing the C-terminal half of the S protein (GAEHVNNSYE to VLKGVKLHYT). Peptides were dissolved in sterile water containing 10% dimethyl sulfoxide (DMSO).
Reverse genetics and generation of double ORF-deficient rSARS-CoV-2 strains. The BAC harboring the entire viral genome of SARS-CoV-2 USA-WA1/2020 strain (accession no. MN985325) was described previously (30). The double deletion of accessory ORF proteins was achieved in viral fragment 1 by inverse PCR using primer pairs containing a BsaI type IIS restriction endonuclease site. All the primer sequences are available upon request. Fragments containing the double deletion of accessory ORF proteins were reassembled into the BAC using BamHI and RsrII restriction endonucleases. Virus rescues were performed as described previously (30,40). Viral passage 1 (P1) stocks were generated in Vero E6 cells and then Double Deletion Strategy To Generate Live-Attenuated SARS-CoV-2 Microbiology Spectrum further concentrated with polyethylene glycol (System Biosciences; catalog no. LV825A-1) following the manufacturer's protocol, and then they were aliquoted, titrated, and stored at 280°C. Plaque assay and immunostaining. Confluent monolayers of Vero E6 cells (10 6 cells/well, 6-well plate format, triplicates) were infected with 10-fold serial diluted viral solutions for 1 h at 37°C. After viral adsorption, cells were overlaid with postinfection media containing 1% low melting agar and were incubated at 37°C. At desired times postinfection, cells were fixed overnight with 10% formaldehyde solution at 4°C. For immunostaining, cells were permeabilized with 0.5% (vol/vol) Triton X-100 in phosphate-buffered saline (PBS) for 15 min at room temperature and immunostained using the N protein 1C7C7 monoclonal antibody (1 mg/mL) and the Vectastain ABC kit (Vector Laboratories), following the manufacturers' instructions. After immunostaining, plates were photographed under a ChemiDoc system (Bio-Rad). Viral plaque diameters were determined using a ruler as a scale that was photographed together with the 6-well plates.
Growth kinetics. Vero E6 or A549-hACE2 cell monolayers (6-well plate, triplicate) were infected with viruses at an MOI of 0.01. After viral absorption for 1 h at 37°C, the viral supernatant was discarded and cells were washed 3 times with PBS. Then, 3 mL of postinfection media (DMEM containing 2% FBS and 1% penicillin-streptomycin-glutamine [PSG]) was added to each well. At the indicated time points (12,24,48,72, and 96h) postinfection, the supernatant (500 mL) was collected and stored at 280°C. All supernatants were finally subjected to viral titration by plaque assay.
RNA extraction and RT-PCR. Total RNA from mock-or virus-infected (MOI of 0.01) Vero E6 cells (10 6 cells/well, 6-well plate format) was extracted with TRIzol reagent (Thermo Fisher Scientific) according to the manufacturer's instructions. RT-PCR amplification of viral ORF3a, ORF6, ORF7a, ORF7b, ORF8 and N genes was performed using Super Script II reverse transcriptase (Thermo Fisher Scientific) and an expanded highfidelity PCR system (Sigma-Aldrich). Amplified DNA products were subjected to 1.0% agarose gel analysis. All primer sequences used for RT-PCR are available upon request.
Animal experiments. All animal protocols were approved by Texas Biomedical Research Institute IACUC. Five-week-old female K18 hACE2 transgenic mice were purchased from The Jackson Laboratory, and 5-week-old female golden Syrian hamsters were purchased from Charles River Laboratories. All the animals were maintained in the animal facility at Texas Biomedical Research Institute under specificpathogen-free conditions. Animal infection was performed by intranasal inoculation after animals were anesthetized following gaseous sedation in an isoflurane chamber.
Pathogenicity analysis of the double ORF-deficient rSARS-CoV-2 strain in K18 hACE2 transgenic mice. Five-week-old K18 hACE2 transgenic mice (n = 8) were infected intranasally (i.n.) (2 Â 10 5 PFU) with the double ORF-deficient rSARS-CoV-2 or rSARS-CoV-2 WT strain. At 2 and 4 dpi, four mice per group were sacrificed humanely to collect lungs and nasal turbinates, and gross images of lungs were taken using an iPhone 6s (Apple). Then, nasal turbinates and lungs were homogenized and processed as described previously (31). Tissue homogenates were centrifuged at 12,000 Â g for 5 min at 4°C, and the clarified supernatants were collected for further measurement of viral titers and cytokine and chemokine induction. Another set of 5-week-old K18 hACE2 transgenic mice (n = 5) were mock infected or infected (i.n.; 2 Â 10 5 PFU) with the double ORF-deficient rSARS-CoV-2 or rSARS-CoV-2 WT strain to evaluate body weight changes and survival rate daily for 21 days. The surviving mice were bled to collect serum to assess total IgG levels against the viral full-length S protein. Afterward, mice were sacrificed to collect spleens for an analysis of T cell responses in splenocytes.
Immunization with rSARS-CoV-2 D3a/D7b protected K18 hACE2 transgenic mice from lethal challenge of rSARS-CoV-2 mCherryNluc. Five-week-old K18 hACE2 transgenic mice (n = 8) were mock vaccinated or vaccinated with rSARS-CoV-2 D3a/D7b (i.n.; 2 Â 10 5 PFU). At 21 days postvaccination, all vaccinated mice were challenged i.n. with rSARS-CoV-2 mCherryNluc (10 5 PFU). At 2 and 4 days postchallenge, mice (n = 4) were anesthetized with isoflurane and imaged immediately under an in vivo imaging system (IVIS; AMI HTX) after being retro-orbitally injected with 100 mL of Nano-Glo luciferase substrate (Promega). The bioluminescence data acquisition and analysis were performed using the Aura program (Spectral Imaging Systems). Flux measurements were acquired from the region of interest. Then, lungs were excised to analyze mCherry expression under the IVIS, and brightfield images were taken using an iPhone 6s for pathological lesion analysis. Finally, nasal turbinates and lungs were homogenized as described above and the clarified supernatants of homogenate were collected for further measurement of viral titers, Nluc activity, and cytokine and chemokine induction. Another set of 5-week-old K18 hACE2 transgenic mice (n = 5) were mock vaccinated or vaccinated (i.n.; 2 Â 10 5 PFU) with rSARS-CoV-2 the absence of serum were included as internal controls. Neutralizing titer at 50% inhibition (NT 50 ) was calculated for each serum sample.
Evaluation of lung pathological lesions. Macroscopic pathology scoring was evaluated using ImageJ software to determine the percentage of the total surface area of the mouse lungs (dorsal and ventral view) affected by consolidation, congestion, and atelectasis, as described previously (48). Left lung lobes from hamsters were fixed in 10% formalin, processed, and stained with H&E to examine inflammation score and pulmonary pathology. The extent of bronchointerstitial pneumonia across the groups was quantified using HALO software v3.4 (Indica Labs).
Enzyme-linked immunosorbent assay (ELISA). ELISA plates were coated with 100 ng/well of the viral S protein in 50 mL of PBS overnight at 4°C. After being blocked by 2.5% BSA for 1 h at 25°C, the plates were washed 3 times with PBS containing 0.1% Tween 20 (PBST) and then incubated (triplicate) with 2-fold serial diluted samples at 25°C (starting dilution of 1:50 for serum and starting dilution of 1:2 for BALF). After 2 h of incubation, plates were washed 3 times with PBST and then incubated with a horseradish peroxidase (HRP)-conjugated goat anti-mouse secondary antibody (ThermoFisher Scientific; catalog no. 31430) at 25°C. After 1 h, plates were washed 3 times and then developed by adding 100 mL/well of 3,39,5,59-tetramethylbenzidine (TMB)-ELISA substrate (ThermoFisher Scientific; catalog no. 34029). Reactions were stopped after 10 min by adding 50 mL/well of 3 M H 2 SO 4 to all the wells. ELISA plates were evaluated with a plate reader (Bio-Tek) at an absorbance of 450 nm.
Statistical analysis. A two-tailed Student's t test was used to compare the mean between two groups, one-way analysis of variance (ANOVA) with post hoc Dunnett's multiple-comparison test (versus control) was executed for the mean comparisons between multiple groups and across time, and the Kaplan-Meier survival analysis with a log rank (Mantel-Cox) test was applied to compare overall survival time (GraphPad Prism v8.0). P values less than 0.05 (P , 0.05) were considered statistically significant (*, P , 0.05; **, P , 0.01; ns, P . 0.05).
Data availability. All the recombinant viruses and BAC-based reverse genetics described in the study are available online at the following website: https://www.txbiomed.org/services-2/virus-request/.