Replication and single-cycle delivery of SARS-CoV-2 replicons

Description

Self-replicating RNAs, known as replicons, are model systems used to genetically probe numerous aspects of RNA virus life cycles without producing infectious virus (1)(2)(3)(4)(5)(6)(7). Replicons for positive-stranded RNA viruses are typically constructed using reverse genetics approaches to replace one or more viral structural proteins with selectable and reporter genes. When translated inside cells, replicon RNA produces viral gene products that establish RNA replication factories, with reporter genes providing readouts for replicon activity and selectable genes permitting selection of cells that stably harbor the replicon. Because key structural components of the virion are missing, replication proceeds without producing infectious virus. Replicon systems that do not require high containment laboratory settings have been invaluable as molecular virology and high-throughput drug development platforms, as perhaps best exemplified for hepatitis C virus (8,9).
Reverse genetics systems for SARS-CoV-2, the causative agent of COVID-19, have been developed for fully-infectious recombinant virus production (10)(11)(12)(13) and as replicon platforms (14)(15)(16)(17)(18)(19). In the latter case, trans-complementation of the deleted structural gene nucleocapsid (N) (18) or envelope (E) and Orf3a genes (19) can enable single-cycle infectious SARS-CoV-2 virion production that may reduce the need for high containment for a range of applications. While these systems permit spike-dependent replicon delivery, interrogating newly emerging spike variants requires replicon reengineering for each variant. In contrast, SARS-CoV-2 spike-pseudotyped lentiviruses (20)(21)(22) or chimeric rhabdoviruses (22,23) bearing spike(s) offer a rapid, plasmid-based means of virion production for spike directed studies (24), such as the characterization of neutralizing antibodies. There remains a need for experimental systems that harness the non-infectious advantages of replicons, while also enabling studies of virus entry and replication. In principle, combining a spike-deleted SARS-CoV-2 replicon with viral glycoprotein trans-complementation would achieve this goal. Such a system would enable isogenic studies of spike variants in a SARS-CoV-2 based platform, while also serving as an appropriately safe-guarded infection-based means of replicon launch that may be extended to additional cell types by using heterologous viral glycoproteins. Here we describe the construction, activity and single-cycle virion generation of spikedeleted SARS-CoV-2 replicons using viral glycoprotein transcomplementation to create replicon delivery particles (RDPs).

Spike-deleted replicon design and optimization of RNA production
The SARS-CoV-2 genome is thought to encode 16 nonstructural proteins (Nsp1-Nsp16) in two overlapping reading frames (Orf1a/1b), as well as four structural proteins-spike (S), Replication and single-cycle delivery of SARS-CoV-2 replicons (Page numbers not final at time of first release) 2 membrane (M), envelope (E), nucleocapsid (N)-and at least seven accessory proteins (3a, 3c, 6, 7a, 7b, 8, and 9b) expressed from subgenomic RNAs or alternative reading frames ( Fig.  1A) (25,26). We adopted a modular design to assemble a replicon consisting of all viral proteins with the exception of the primary structural glycoprotein spike (ΔS). The spike transcription-regulating sequence (TRS) was instead used to drive expression of a gene cassette consisting of neomycinresistance (NeoR) and a reporter gene (nuclear-localized monomeric NeonGreen or secreted Gaussia luciferase) separated by a T2A ribosome shift sequence (Fig. 1A). The plasmid encoding the replicon cDNA contains an upstream T7 promoter at the 5′ end for in vitro transcription and a self-cleaving hepatitis delta virus (HDV) ribozyme at the 3′ end, which cleaves after an encoded polyA sequence yielding an authentic terminus.
For replicon assembly, we employed a recently published RNA virus reverse genetics system in the yeast Saccharomyces cerevisiae (10). This system leverages transformation associated recombination (TAR) to accurately assemble numerous, large overlapping DNA fragments (27). Transformation of yeast with equimolar ratios of replicon cDNA fragments led to efficient replicon assembly as assessed by multiplex PCR (fig. S1A). We performed restriction enzyme digests of the resulting DNA to assess plasmid integrity and observed that yeast-derived plasmids were contaminated with yeast genomic DNA and did not reveal the expected NdeI digest pattern (Fig. 1B). In an alternative approach, we propagated yeast assembled plasmids in bacteria, which boosted plasmid purity, as demonstrated previously (10). However, the overall DNA yield was suboptimal in both instances and, in the case of bacterial propagation, often resulted in mutations in the coding region of the viral RNA-dependent RNA polymerase (RdRp). To improve plasmid purity and yield, we developed a method that relies on selective enzymatic digestion of contaminating yeast DNA followed by preferential amplification of the plasmid product. We first treated plasmid preparations with BamHI, which digests yeast genomic DNA but whose recognition sequence is absent in the replicon plasmid. We subsequently treated the DNA with Plasmid-Safe (PS) DNase to digest linear contaminating yeast DNA. We next turned to multiple displacement rolling circle amplification (RCA) using phi29 DNA polymerase (28) and random primers to amplify replicon plasmids from either yeast or sequence-verified bacterial clones. Due to its processivity (>70 kb per binding event), strand-displacement activity, and low error rate (<1E −6 ), phi29 polymerase can exponentially amplify long, circular DNA sequences with high fidelity under isothermal conditions and has been used to replicate bacterial artificial chromosomes, cosmids, mitochondrial DNA, and microbial genomes megabases in length (29)(30)(31). Indeed, amplification of replicon plasmids from yeast or bacteria yielded useful quantities of full length and intact replicon DNA (Fig. 1B) as verified by amplicon sequencing. We used the resulting DNA for T7 transcription reactions and observed apparently fulllength (~27.5 kb) replicon RNA in addition to shorter RNA products (Fig. 1C). To directly compare launch efficiency, mNeonGreen-encoding replicon RNAs were electroporated into BHK-21 cells together with additional in vitro transcribed N mRNA, previously shown to boost launch efficiency (10,15), and the percentage of cells expressing the mNeon-Green reporter was measured at 24 h post electroporation. We observed fewer than 1% of replicon positive cells using RNA from non-phi29 amplified templates, where bacteria propagated plasmids yielded the highest (0.8%) mNeonGreen reporter signal (Fig. 1D). In contrast, PS DNase-treated and phi29-amplified templates yielded approximately 20% mNeon-Green positive cells. The mNeonGreen signal was dependent on productive replication and transcription, as no signal was observed using replicon RNA harboring inactivating mutations in Nsp12, the viral RdRp (32, 33) (Fig. 1E, pol(-)). In summary, the most efficient means of reaching high replicon launch efficiency was achieved by making capped RNA from yeast plasmids that are PS digested and phi29-amplified prior to transcription (Fig. 1F).

Spike-deleted SARS-CoV-2 replicons are convenient and versatile assay platforms
To characterize the utility of spike-deleted replicons for antiviral compound evaluation and screening, host factor validation, and viral mutant phenotype assessment, we constructed a replicon encoding secreted Gaussia luciferase (Gluc) (Fig. 1A). RNA transcripts were electroporated into Huh-7.5 cells and cumulative reporter activity per 24 h period was monitored via luciferase quantification from culture supernatant. We detected robust luciferase activity by 24 h post-electroporation which steadily decreased over time ( Fig.  2A). Luciferase activity was dependent on viral replication, as no signal was observed for the pol(-) replicon mutant. Replicon-driven luciferase expression was also sensitive to remdesivir, a well-characterized inhibitor of the SARS-CoV-2 RdRp (34). As a luciferase-independent measure of replicon activity, we measured N subgenomic mRNA (sgmRNA) levels by qRT-PCR and observed RNA accumulation kinetics that paralleled reporter gene expression (Fig. 2B). This signal was distinct from co-electroporated N mRNA, needed for efficient replicon or virus launch from RNA (10), which was detectable in all measured samples as expected ( fig. S1B). These results show that SARS-CoV-2 replicons undergo robust and quantifiable RNA replication.
One of the most fruitful applications for replicons are as drug screening platforms. As a proof-of-concept, we treated cells bearing luciferase reporter replicons with compounds reported to have direct acting antiviral (DAA) activity at  (34), and masitinib, a proposed 3CLpro inhibitor (35). As a negative control, we used 27-hydroxycholesterol (27-HC), which was reported to have both anti-SARS-CoV-2 and anti-HCoV-OC43 activity by affecting viral entry, which we bypass by RNA transfection (36). We also tested a host-targeting agent (HTA), AM580, a retinoid derivative reported to have broad antiviral activity by disrupting sterol regulatory element binding protein (SREBP) lipid signaling (37). For remdesivir, we observed IC50 values and low cytotoxicity profiles similar to those previously reported for live virus ( Fig. 2C) (34). For masitinib, we observed inhibition with similar IC50 values to what was shown with infectious HCoV-OC43 and SARS-CoV-2 (35). However, cellular toxicity was observed at high concentrations ( Fig. 2D). For AM580, IC50 values were consistent with reported results, with low cytotoxicity at inhibitory concentrations ( Fig. 2E) (37). As expected, 27-HC showed no detectable inhibition (Fig. 2F). These representative results showcase the utility of SARS-CoV-2 replicons as scalable drug discovery platforms focused on intracellular viral replication events. We next tested if replicons could be used to identify or validate intracellular SARS-CoV-2 host factors. Transmembrane protein 41B (TMEM41B) was recently reported to be a critical intracellular host factor for multiple coronaviruses (38). To test whether SARS-CoV-2 replicon activity depended on TMEM41B function, we electroporated wild-type and TMEM41B KO Huh-7.5 cells with reporter replicon RNA. Consistent with results obtained with virus, TMEM41B KO resulted in a 22-fold decrease in reporter activity compared to wild-type Huh-7.5 cells (Fig. 2G), an effect similar in magnitude to remdesivir treatment. In contrast, an alphavirus replicon was not affected by TMEM41B ablation (Fig. 2H). TMEM41B reconstitution in KO cells led to an 11-fold rescue in replicon activity (Fig. 2I). These results demonstrate that SARS-CoV-2 replicons are sensitive to disruption of critical intracellular host factors.

SARS-CoV-2 replicons reveal viral determinants of interferon sensitivity
SARS-CoV-2 replicons can also be used to characterize viral mutants. Recent studies have highlighted the importance of Nsp1 as a suppressor of host translation (39)(40)(41)(42)(43)(44) consistent with prior studies on SARS-CoV (45,46). This Nsp1 activity can be ablated by two amino acid substitutions at positions 164 and 165 (47) that are important for association with the 40S ribosomal subunit (39,40). Nsp1 mutant (Nsp1 mut ) replicons might preserve translation and cell viability and hence prolong replicon activity or, given the proposed role of Nsp1 in blunting the innate immune response (14,39,48), such mutations might attenuate the replicon. We generated Nsp1 mut replicons and found that they performed similarly to wild-type versions in Huh-7.5 cells, could be inhibited by remdesivir (Fig. 3A), but exhibited less cellular toxicity (Fig.  3B). In contrast, viability of cells harboring wild-type replicons remained low even when replicons were launched in the presence of remdesivir (Fig. 3B). Taken together, these results suggest that the initial burst of SARS-CoV-2 Nsp1, expressed from transfected replicon RNA, is an important mediator of cell toxicity.
As SARS-CoV-2 Nsp1 activity is proposed to halt production of interferon-stimulated gene (ISG) products (39,40), we hypothesized that Nsp1 mut replicons would be more sensitive to interferon. Indeed, upon launch in Huh-7.5 cells, Nsp1 mutants were hypersensitive to IFNα and IFNβ compared to wild-type replicons (Fig. 3, C and D) while both replicons exhibited similar sensitivity to remdesivir (Fig. 3E). These results highlight the ability of Nsp1 to blunt the antiviral ISG response by inhibiting host translation, which contributes to cytotoxicity. While these features of Nsp1 mut replicons may be advantageous for DAA and HTA screening, the lack of Nsp1 functions may affect screening outcomes in different cell backgrounds, depending upon innate immune competence or other factors. Further, attempts to leverage the lower cytotoxicity of Nsp1 mut replicons to select for stable cell lines harboring non-cytopathic replicons, as done previously with HCoV-229E (49), have thus far been unsuccessful in Huh-7.5 and BHK-21 cells. Different cell backgrounds and multiple adaptive mutations may be necessary to achieve this goal.

SARS-CoV-2 replicons can be trans-complemented with spike to generate single-cycle virions
While BHK-21 and Huh-7.5 cells are permissive for replicon launch via electroporation, physiologically relevant lung cell lines and primary cell types are more challenging to transfect. Such cells also have intact RNA sensing and innate immune functions, which BHK-21 and Huh-7.5 cells lack (50,51). Electroporation or other methods of RNA delivery also preclude the study of normal viral entry processes, and the large quantities of transfected RNA do not mimic virus infection, where one or few RNA genomes initiate productive replication. As a more authentic route for replicon delivery, we attempted to package replicons as single-cycle virions that could infect cells in a spike-dependent manner but would not produce infectious progeny capable of further spread. We refer to these single-cycle virions as replicon delivery particles (RDPs). We transfected BHK-21 cells with a spike-expressing plasmid and 24 h later co-electroporated spike-deleted mNeonGreen replicon RNA and N mRNA (Fig. 4, A and B). The full-length native spike gene was human-codon optimized and designed to lack 5′UTR, leader, TRS, and 3′UTR viral sequences in order to minimize the possibility of recombination into the replicon (52,53). Given that Nsp1 hindered cell survival post electroporation, we hypothesized that Nsp1 mut replicons could be . We observed spike RDP titers of 10 3 -10 4 TCID50/ml in Huh-7.5 AT using unconcentrated stocks and up to 6.2 × 10 5 TCID50/ml upon concentration, with about 10-fold lower specific infectivity than virus (Fig.  4, E to G). This may indicate that RDPs have reduced spike density compared to virus.

Neutralization assays with RDPs recapitulate authentic SARS-CoV-2 antibody phenotypes
Spike RDPs could complement pseudovirus assays based on HIV-1 or vesicular stomatitis virus (VSV), with a SARS-CoV-2-based, and therefore more authentic, single cycle alternative. We generated Nsp1 mut Gluc RDPs trans-packaged with either the prototype spike (WA1/2020 isolate) or the B.1.351 variant of concern (54). We performed neutralization assays with these RDPs using two well-characterized human monoclonal antibodies, C135 and C144 capable of potently neutralizing SARS-CoV-2 or pseudoviruses expressing the prototype spike (55). C144 binding and neutralization is ablated by the spike E484K mutation, which was originally identified by antibody selection experiments using a VSV pseudovirus (56) and which later appeared in the B.1.351 variant (54). RDPs harboring the prototype spike were neutralized with both antibodies, while only C135, but not C144, neutralized B.1.351 spike RDPs (Fig. 5, A and B). Neutralization curves and relative IC50 values were comparable to those obtained with the respective SARS-CoV-2 isolates (Fig. 5, D to F) and with previous reports using pseudovirus assays (55,56) with the added advantage that RDP generation does not require deletions in the spike coding sequence (22).

Spike-deleted SARS-CoV-2 replicons can incorporate VSV glycoprotein
Efficient entry with spike RDPs appears to require high levels of ACE2 and TMPRSS2 in Huh-7.5 cells since RDP addition to Huh-7.5 or Vero cells not overexpressing these factors was unproductive. As numerous cell lines relevant for SARS-CoV-2 studies may have insufficient ACE2 or TMPRSS2 levels unless engineered to support viral entry (57), we tested whether spike-deleted replicons could be packaged with VSV-G, analogous to one-way lentiviral transduction. In such systems, VSV-G pseudotyping provides an efficient means of lentiviral vector entry (58) and for RDPs might provide an ACE2independent means of replicon delivery. We transfected BHK-21 cells with a VSV-G expression plasmid followed by co-electroporation with spike-deleted mNeonGreen replicon RNA and N mRNA (Fig. 6A). The resulting RDPs in the supernatant were concentrated and added to naïve Huh-7.5 cells. After 24 h, we observed mNeonGreen-positive cells in a VSV-G dependent manner (Fig. 6, B and C, P1). Importantly, there was no measurable signal in a second passage (Fig. 6, B and C, P2) and no VSV-G carryover was detected in the P1 cells ( fig. S3, A and B), suggesting that VSV-G RDPs possess single-cycle infectivity. In contrast to spike RDPs, producer (P0) VSV-G expressing BHK-21 cells had a significantly higher ratio of mNeonGreen compared to control cells when the Nsp1 mut replicon was used, suggestive of local VSV-G dependent spread. P1 cells also exhibited higher mNeonGreen and N protein positivity using Nsp1 mut RDPs compared to wild-type RDPs (Fig. 6, B and C, and fig. S3C). We observed VSV-G RDP titers between 1-8x10 3 TCID50/ml on Huh-7.5 cells. We then tested VSV-G RDP infectivity on additional cell types: African green monkey VeroE6 cells, human Caco2 intestinal epithelial cells (59,60) and both Calu3 and A549 human lung adenocarcinoma cells (61)(62)(63)(64)). Additionally, we tested RDP activity in two lines of human airway cells: normal human bronchial/tracheal epithelial cells (NHBE) and normal human lung fibroblasts (NHLF) (Fig. 6D). All cells were susceptible to infection with VSV-G RDPs and exhibited productive viral replication, as evident from mNeonGreen reporter expression. We pretreated NHBE, NHLF and A549 cells with 100 nM remdesivir or 100 pM IFNα and subsequently added Gluc VSV-G RDPs. As shown in Fig. 6E, both treatments inhibited Gluc reporter activity to levels approaching baseline.
Overall, these data demonstrate that VSV-G RDPs provide flexible means of replicon launch in primary cell contexts.
To further examine whether VSV-G RDPs have singlecycle infectivity using a more sensitive approach, we generated Gluc expressing VSV-G RDPs and infected naïve Huh-7.5 cells followed by Gluc measurements 24h later (P1). While the VSV-G RDP positive cells produced a robust Gluc signal, no signal was observed when the supernatant from P1 was concentrated and passaged onto naïve cells (P2) (Fig. 6F). Input

VSV-G trans-complementation of a SARS-CoV-2 replicon lacking all accessory genes
As the RDP results presented thus far have relied on spikedeleted replicons with or without mutations in Nsp1, we examined whether replicons harboring additional deletions could be packaged into RDPs. Previous work with SARS-CoV and SARS-CoV-2 has shown that M and E genes are required for virion morphogenesis and release (65,66).
On that basis, we tested VSV-G packaging of a replicon that lacked all accessory genes but retained genes encoding the structural M, E, and N proteins (ΔAcc) (fig. S4A). VSV-G dependent infectivity was observed on recipient Huh-7.5 cells using both mNeonGreen and Gluc ∆Acc versions at levels comparable to the spike-deleted (∆S) replicon ( fig. S4, B to D). Thus, accessory genes are dispensable for replicon pseudotyping with VSV-G. We highlight several features of SARS-CoV-2 RDPs. First, RDPs enable a rapid and isogenic means of generating singlecycle SARS-CoV-2 virions for spike-directed efforts, such as screening spike variants against neutralizing antibodies while eliminating the potential for spike mutations to arise, as is commonly encountered with cell culture passaged virus (67,68) and for screening antibodies against additional structural proteins. Second, single-cycle infectivity can be achieved with VSV-G, which may be beneficial for studies in systems where ACE2 receptor overexpression is infeasible. Finally, in contrast to replicon launch by RNA transfection, RDPs can be frozen, stored, and distributed, since they require no specialized equipment to use once generated. These attributes could eventually enable high-throughput drug screening without the need for high containment. Overall, these features highlight how SARS-CoV-2 spike-deleted replicons and RDPs provide a flexible and single-cycle infectious platform for future studies of this pandemic virus. the results described in this manuscript. CMR is a founder of Apath LLC, a Scientific Advisory Board member of Imvaq Therapeutics, Vir Biotechnology, and Arbutus Biopharma, and an advisor for Regulus Therapeutics and Pfizer. PDB has received compensation for consulting services to Pfizer. IRL is a paid consultant for Grit Bio. The remaining authors declare no competing interests. Data and materials availability: Data supporting the findings of this study are reported in the main text and figures and in supplementary figs. S1 to S4 and table S1. All reagents and materials generated in this study are available from the corresponding authors with a completed Materials Transfer Agreement to the Rockefeller University. Note that the biosafety containment level appropriate for use with these materials is subject to the policies of the requesting institution in addition to relevant regulatory bodies. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.