Within the past decade, we have seen outbreaks of numerous viruses, including Middle East respiratory syndrome coronavirus (MERS-CoV)5, ZIKA virus6, Ebola virus7 and, at the end of 2019, SARS-CoV-2—which was first detected in Wuhan, Hubei province, China4, but rapidly developed into a pandemic. During the early phase of the SARS-CoV-2 outbreak, virus isolates were not available to health authorities and the scientific community, even though these isolates are urgently needed to generate diagnostic tools, to develop and assess antivirals and vaccines, and to establish appropriate in vivo models. The generation of the SARS-CoV-2 from chemically synthesized DNA could bypass the limited availability of virus isolates and would furthermore enable genetic modifications and functional characterization. However, although E. coli proved very useful for the cloning of many viral genomes, it has a number of disadvantages when used for the assembly and stable maintenance of full-length molecular clones of emerging RNA viruses, including coronaviruses.

Synthetic genomics is a field fuelled by the efforts to create a bacterial cell that is controlled by a synthetic genome8. Genome-wide reassembly of the approximately 1.1-megabase (Mb) genome of Mycoplasma was first attempted using E. coli as an intermediate host8; however, the maintenance of 100-kilobase (kb) DNA fragments appeared to be very difficult in this host. Therefore, the yeast S. cerevisiae was chosen to clone, assemble and mutagenize entire Mycoplasma genomes9,10. The rationale for using a yeast cloning system is the ability of yeast to recombine overlapping DNA fragments in vivo, which led to the development of a technique called transformation-associated recombination (TAR) cloning11.

More recently12,13, TAR cloning was successfully used for the assembly, genetic engineering and rescue of large DNA viruses such as cytomegalovirus and herpes simplex virus 1. For coronaviruses that belong to a family of positive-stranded RNA viruses termed Coronaviridae, the generation of full-length molecular clones has long been hampered by the large genome size (27–31 kb) and occasional instability of cloned DNA in E. coli. However, unconventional approaches—such as cloning in low-copy bacterial artificial chromosomes (BACs) or vaccinia virus, or cloning of subgenomic DNA fragments followed by in vitro ligation—were successful1,2,3, although each system has caveats that make the generation of recombinant coronavirus genomes cumbersome. Here we assessed the suitability of the yeast S. cerevisiae to assemble and maintain genomes of diverse RNA viruses to establish a rapid, stable and universal reverse-genetics pipeline for RNA viruses.

To generate a yeast-based reverse-genetics platform for RNA viruses, we first used mouse hepatitis virus (MHV) strain A59, which contains the gene for green fluorescent protein (MHV-GFP) and which has an established vaccinia virus-based reverse-genetics platform14,15. The overall strategy is shown in Fig. 1a. Viral RNA was prepared from MHV-GFP-infected mouse 17Cl-1 cells and used to amplify seven overlapping DNA fragments by reverse-transcription PCR (RT–PCR) that spanned the MHV-GFP genome from nucleotides 2024 to 29672. Fragments containing the 5′ and 3′ termini were PCR-amplified from the vaccinia virus-cloned genome to include a T7 RNA polymerase promoter directly upstream of the MHV-GFP 5′ end and a cleavage site (PacI) after the poly(A) sequence at the MHV-GFP 3′ end, which is required to produce RNA run-off transcripts using T7 RNA polymerase14. Overlap sequences for the TAR vector pVC604 were included in the primers that amplified the 5′- and 3′-terminal fragments (Supplementary Table 1). All DNA fragments were simultaneously transformed into S. cerevisiae (strain VL6-48N), and the resulting clones were screened for the correct assembly of the yeast artificial chromosome (YAC) containing the cloned MHV genome by multiplex PCRs that covered the junctions between recombined fragments. This screen revealed that more than 90% of the clones tested were positive, indicating that the assembly in yeast is highly efficient (Supplementary Fig. 1a). To rescue MHV-GFP, we randomly chose two clones, purified and linearized the YACs using PacI (Extended Data Table 1) and subjected the YACs to in vitro transcription using T7 RNA polymerase to generate capped viral genomic RNA. This RNA was transfected together with an in vitro-transcribed mRNA that encodes the MHV nucleocapsid (N) protein into BHK-MHV-N cells, which were then mixed with MHV-susceptible 17Cl-1 cells as previously described14. Cytopathogenic effects, virus-induced syncytia and GFP-expressing cells were readily detectable for both clones within 48 h, indicating the successful recovery of infectious virus (Fig. 1b). Finally, we assessed the replication kinetics of the recovered viruses, which were indistinguishable from the parental MHV-GFP line (Fig. 1c).

Fig. 1: Application of yeast-based TAR cloning to generate viral cDNA clones and the recovery of recombinant MHV-GFP.
figure 1

a, General workflow of TAR cloning and virus rescue. In-yeast genome reconstruction requires one-step delivery of overlapping DNA fragments that cover the viral genome and a TAR vector in yeast. Viral ORFs and the ORF for GFP are indicated. Transformed DNA fragments are assembled by homologous recombination in yeast to generate a YAC that contains the full-length viral cDNA sequence. In vitro production of infectious capped viral RNA starts with the isolation of the YAC, followed by plasmid linearization to provide a DNA template for run-off T7 RNA polymerase-based transcription. Virus rescue is initiated by electroporation of BHK-MHV-N cells, after which virus production and amplification is carried out by culturing the virus with susceptible cells. b, Recovery of infectious rMHV-GFP from yeast clones 1 and 2. Cell-culture supernatants—which contain viruses produced after virus rescue of two MHV-GFP YAC clones—were used to infect 17Cl-1 cells. At 48 h after infection, infected cells were visualized for GFP expression (top) and by bright-field microscopy (bottom). Mock represents 17Cl-1 cells inoculated with the supernatant from BHK-MHV-N cells electroporated without viral RNAs. Images are representative of two independent experiments. Scale bars, 100 μm. c, Replication kinetics of parental MHV-GFP and rMHV-GFP clones 1 and 2. L929 cells were infected (multiplicity of infection (MOI) = 0.1), and cell-culture supernatants were collected at the indicated time points after infection and titrated by plaque assay. PFU, plaque forming units. Data represent the mean ± s.d. of three independent biological experiments (n = 3). Statistical significance was determined by two-sided unpaired Student’s t-test without adjustments for multiple comparisons. NS, not significant. P values (from left to right): top, NS, P = 0.2905; NS, P = 0.3504; NS, P = 0.1817; NS, P = 0.9862; NS, P = 0.6738; bottom, NS, P = 0.0835; NS, P = 0.1400; NS, P = 0.2206; NS, P = 0.8020; NS, P = 0.5894.

Source data

To address whether the synthetic genomics platform can be applied to other coronaviruses and whether it can be used for rapid mutagenesis, we used a molecular BAC clone of MERS-CoV16. We PCR-amplified eight overlapping DNA fragments that covered the MERS-CoV genome (Extended Data Fig. 1a, Supplementary Fig. 1b and Supplementary Table 1). The 5′- and 3′-terminal fragments contained the T7 RNA polymerase promoter upstream of the MERS-CoV 5′ end and the restriction endonuclease cleavage site MluI downstream of the poly(A) sequence, and overlapping sequences with the TAR plasmid pVC604. To mutagenize the MERS-CoV clone, fragment 7 was divided into three overlapping PCR fragments to place the GFP gene in frame with a porcine teschovirus 2A element and open-reading frame 4a (ORF4a)16 (Extended Data Fig. 1a and Supplementary Table 1). Again, almost all YAC clones were successfully assembled (Supplementary Fig. 1b, c). Virus rescue from cloned DNA was performed as described previously16, resulting in recombinant (r)MERS-CoV and rMERS-CoV-GFP (Extended Data Fig. 1b). This demonstrates that the synthetic genomics platform is suitable to genetically modify coronavirus genomes. As expected, the replication kinetics of rMERS-CoV and rMERS-CoV-GFP were slightly reduced compared with the cell-culture-adapted MERS-CoV-EMC strain (Extended Data Fig. 1c).

Next, we thoroughly evaluated the stability of the cloned genomes, the range of applicability to other virus genomes and whether molecular clones can be generated from clinical samples. Yeast clones that contained YACs encoding MHV-GFP and MERS-CoV were passaged 15–17 times, and sequencing revealed that the genomes could be stably maintained (Extended Data Table 2). We further cloned several other coronaviruses (HCoV-229E2, HCoV-HKU1 (GenBank: NC_006577) and MERS-CoV-Riyadh-1734-2015 (GenBank: MN481979)) and viruses of other families, such as ZIKA virus (family Flaviviridae, GenBank: KX377337) and human respiratory syncytial virus (hRSV; family Pneumoviridae) (Table 1), which are known to be difficult to clone and stably maintain in E. coli. As shown in Supplementary Fig. 1d–h, cloning of these viral genomes in yeast was in all cases successful irrespectively of the virus source, the nucleic acid template or the number of DNA fragments. Of note, we cloned hRSV-B without any prior information on the virus genotype directly from a clinical sample (nasopharyngeal aspirate) by designing RSV consensus primers to amplify four overlapping DNA fragments (Supplementary Table 1) (sequence submitted to GenBank: MT107528). Collectively, these results demonstrate that the synthetic genomics platform provides the technical advance to rapidly generate molecular clones of diverse RNA viruses by using virus isolates, cloned DNA, synthetic DNA or clinical samples as starting material.

Table 1 RNA virus genomes cloned using the synthetic genomics platform

The detection of a new coronavirus in China at the end of 2019 prompted us to test the applicability of our synthetic genomics platform to reconstruct the virus based on the genome sequences released on 10–11 January 2020 (Fig. 2). We divided the genome into 12 overlapping DNA fragments (Fig. 3a, Extended Data Table 3, Supplementary Fig. 1i and Supplementary Table 1). In parallel, we aimed to generate a SARS-CoV-2 clone that expressed GFP, as this could facilitate the screening of antiviral compounds and be used to establish diagnostic assays (for example, virus neutralization assays). This was achieved by dividing fragment 11 into three subfragments (Fig. 3a, Supplementary Fig. 1j and Supplementary Table 1), and GFP was inserted in-frame of ORF7a, replacing nucleotides 40–282. We noticed that nucleotides 3–5 at the 5′ end of the reported SARS-CoV-2 sequence (5′-AUUAAAGG; GenBank MN996528.1; nucleotides that are different are highlighted in bold) differed from SARS-CoV (5′-AUAUUAGG; GenBank AY291315) and from the more closely related bat SARS-related CoVs ZXC21 and ZC45 (5′-AUAUUAGG)4,17,18 (Extended Data Fig. 2a, b). We therefore designed three 5′-end versions, and each version was combined with the remaining SARS-CoV-2 genome (constructs 1–3) or a corresponding SARS-CoV-2-GFP genome (constructs 4–6). Constructs 1 and 4 contained the 5′ end modified by three nucleotides according to the bat SARS-related CoVs (5′-AUAUUAGG), constructs 2 and 5 contained the 124 5′-terminal nucleotides of SARS-CoV, and constructs 3 and 6 contained the reported SARS-CoV-2 sequence (5′-AUUAAAGG; according to MN996528.1) (Extended Data Fig. 2a, b). Notably, differences between SARS-CoV-2 and SARS-CoV within the 5′-terminal 124 nucleotides are in agreement with the predicted RNA secondary structures (Extended Data Fig. 2b).

Fig. 2: Timeline of the reconstruction and recovery of rSARS-CoV-2 in relation to key events of the COVID-19 pandemic.
figure 2

Illustration of the rapidity of rSARS-CoV-2 reconstruction along with the timeline of key events of the COVID-19 pandemic. CDC, Center for Disease Control and Prevention; ICTV, International Committee on Taxonomy of Viruses; WHO, World Health Organization.

Fig. 3: Reconstruction, rescue and characterization of rSARS-CoV-2, rSARS-CoV-2-GFP and synSARS-CoV-2-GFP.
figure 3

a, Schematic representation of the SARS-CoV-2 genome organization and DNA fragments used to clone rSARS-CoV-2, rSARS-CoV-2-GFP and synSARS-CoV-2-GFP. Inserts show synthetic subfragments comprising fragments 5 (A–D) and 7 (Aa, Ab, B), and the fragments used to insert the GFP gene (fragments 13–15). b, Left, schematic of the experiment. Middle, rescue of rSARS-CoV-2 from yeast clones 1.1, 2.2 and 3.1. Supernatants (10−1, 10−2 and 10−3 ml) of cells infected with the indicated clones or mock-infected cells were transferred to Vero E6 cells to detect plaques (rSARS-CoV-2). Right, rescue of rSARS-CoV-2-GFP from yeast clones 4.1, 5.2 and 6.2. Supernatants (1 ml) from individual rescue experiments were transferred to Vero E6 cells to detect green fluorescence (rSARS-CoV-2-GFP). Mock, uninfected cells. Scale bars, 100 μm. c, Replication kinetics of rSARS-CoV-2 clones 1.1, 2.2, 3.1 (left) and rSARS-CoV-2-GFP clones 4.1, 5.2, 6.2 and synSARS-CoV-2-GFP (right) compared with the SARS-CoV-2 isolate. Vero E6 cells were infected (MOI = 0.01), and supernatants were collected at the indicated time points after infection and titrated (50% tissue culture infectious dose (TCID50) assay). Data represent the mean ± s.d. of three independent biological replicates. Statistical significance was determined for each clone against the SARS-CoV-2 isolate by two-sided unpaired Student’s t-test without adjustments for multiple comparisons. P values (from left to right): left, top, NS, P = 0.0851; NS, P = 0.1775; *P = 0.0107; NS, P = 0.0648; **P = 0.0013; *P = 0.0373; middle, NS, P = 0.0851; NS, P = 0.1713; *P = 0.0133; NS, P = 0.0535; NS, P = 0.0909; NS, P = 0.0632; bottom, NS, P = 0.1119; NS, P = 0.1641; NS, P = 0.0994; NS, P = 0.4921; NS, P = 0.3336; NS, P = 0.0790; right, top, NS, P = 0.0858; NS, P = 0.1429; *P = 0.0104; *P = 0.0466; **P = 0.0011; *P = 0.0287; second, NS, P = 0.0872; NS, P = 1360; *P = 0.0102; *P = 0.0461; **P = 0.0011; *P = 0.0282; third, NS, P = 0.4810; NS, P = 0.1758; *P = 0.0106; *P = 0.0478; **P = 0.0011; *P = 0.0287; bottom, NS, P = 0.3739; NS, P = 0.6817; *P = 0.0106; *P = 0.0473; **P = 0.0011 *P = 0.0285.

Source data

Fourteen synthetic DNA fragments were ordered as sequence-confirmed plasmids and all but fragments 5 and 7 were delivered (Extended Data Table 3, Supplementary Data 1). As we received SARS-CoV-2 viral RNA from an isolate of a Munich patient (BetaCoV/Germany/BavPat1/2020) at the same time, we amplified the regions of fragments 5 and 7 by RT–PCR (Supplementary Table 1). TAR cloning was immediately initiated, and for all six SARS-CoV-2 and SARS-CoV-2-GFP constructs we obtained correctly assembled molecular clones (Extended Data Fig. 3a and Supplementary Fig. 1i, j). Because sequence verification was not possible within this short time frame, we randomly selected two clones for each construct (Extended Data Fig. 3a), isolated the YAC DNA and performed in vitro transcription. The resulting RNAs were electroporated together with an mRNA that encodes the SARS-CoV-2 N protein into BHK-21 and, in parallel, into BHK-SARS-N cells that expressed the SARS-CoV N protein19 (Extended Data Fig. 3b). Electroporated cells were seeded on Vero E6 cells and two days later we observed green fluorescent signals in cells that received the GFP-encoding SARS-CoV-2 RNAs. Indeed, we could rescue infectious viruses for almost all rSARS-CoV-2 and rSARS-CoV-2-GFP clones (Extended Data Fig. 3b). As shown in Fig. 3b, for rSARS-CoV-2 clones 1.1, 2.2, and 3.1, plaques were readily detectable, demonstrating that infectious virus has been recovered irrespectively of the 5′-terminal sequences. Sequencing of the YACs and corresponding rescued viruses revealed that almost all DNA clones and viruses contained the correct sequence, except for some individual clones that contained mutations within fragments 5 and 7 that were probably introduced by RT–PCR (Extended Data Table 4). Nevertheless, we obtained at least one correct YAC clone for all constructs except for construct 6. To correct this, we reassembled construct 6 by replacing the RT–PCR-generated fragments 5 and 7 with four and three shorter synthetic double-stranded (ds)DNA fragments, respectively. The resulting molecular clone was used to rescue the synthetic SARS-CoV-2-GFP (synSARS-CoV-2-GFP) virus without any mutations exclusively from chemically synthesized DNA (Extended Data Fig. 4 and Extended Data Tables 3, 4).

Next we assessed the 5′ end of the recombinant viruses and the Munich virus isolate and confirmed the published 5′ end sequence of SARS-CoV-2 (5′-AUUAAAGG; GenBank MN996528.3). Full-length sequencing of the viral genomes and 5′ rapid amplification of cDNA end (5′-RACE) analysis of the recombinant viruses confirmed the identity of each virus, and showed that the 5′ end variant of each virus retained the cloned 5′ terminus (Extended Data Fig. 2a). This demonstrates that the 5′ ends of SARS-CoV and bat SARS-related CoVs ZXC21 and ZC45 are compatible with the replication machinery of SARS-CoV-2. Sequencing results also revealed the identity of leader–body junctions of SARS-CoV-2 subgenomic mRNAs, which are identical to those of SARS-CoV18 (Extended Data Fig. 2c–h). We also analysed rSARS-CoV-2 clone 3.1 for protein expression and demonstrated the presence of the SARS-CoV-2 nucleocapsid protein in dsRNA-positive cells (Extended Data Fig. 5b). The replication kinetics of rSARS-CoV-2 clone 3.1, which contains the authentic 5′ terminus, was indistinguishable from replication of the SARS-CoV-2 isolate, while clones 1.1 and 2.2 showed slightly reduced replication (Fig. 3c, left). All rSARS-CoV-GFP clones and synSARS-CoV-GFP displayed similar growth kinetics but they were significantly reduced compared with the SARS-CoV-2 isolate, suggesting that the insertion of GFP and/or the partial deletion of ORF7a affects replication (Fig. 3c, right and Extended Data Fig. 5d–f). Despite the reduced replication, green fluorescence was readily detectable and we demonstrated the use of the synSARS-CoV-GFP clone for antiviral drug screening by testing remdesivir, a promising compound for the treatment of COVID-1920 (Extended Data Fig. 5c). Similarly, the simple readout of green fluorescence greatly facilitates the demonstration of virus neutralization with human serum (Extended Data Fig. 5a).

Our results demonstrate the full functionality of the SARS-CoV-2 reverse-genetics system and we expect that this fast, robust and versatile synthetic genomics platform will provide new insights into the molecular biology and pathogenesis of a number of emerging RNA viruses. Although homologous recombination in yeast has already been used for the generation of a number of molecular virus clones in the past12,13,21,22, we present a thorough evaluation of the feasibility of this approach to rapidly generate full-length cDNAs for large RNA viruses that have a known history of instability in E. coli. We show that one main advantage of the TAR cloning system is that the viral genomes can be fragmented to at least 19 overlapping fragments and reassembled with remarkable efficacy. This facilitated the cloning and rescue of rSARS-CoV-2 and rSARS-CoV-2-GFP within one week. It should be noted that we see considerable potential to reduce the time of DNA synthesis. Currently, synthetic DNA fragments get routinely cloned in E. coli, which turned out to be problematic for SARS-CoV-2 fragments 5 and 7. We, however, used shorter synthetic dsDNA parts to assemble these fragments by TAR cloning and to generate the molecular clone synSARS-CoV-2-GFP by using exclusively chemically synthesized DNA, which is an additional proof of the superior cloning efficiency of yeast- versus E. coli-based systems.

The COVID-19 pandemic emphasizes the need for preparedness to rapidly respond to emerging virus threats. The rapidity of our synthetic genomics approach to generate SARS-CoV-2 and the applicability to other emerging RNA viruses make this system an attractive alternative to provide infectious virus samples to health authorities and diagnostic laboratories without the need of having access to clinical samples. As the COVID-19 pandemic is ongoing, we expect to see sequence variations and possibly phenotypic changes of the evolving SARS-CoV-2 virus in the human host. With this synthetic genomics platform, it is now possible to rapidly introduce such sequence variations into the infectious clone and to functionally characterize SARS-CoV-2 evolution in real time.


Cells and general culture conditions

Vero, Vero B4 and Vero B6 cells (all ATCC) were cultured in Dulbecco’s modified Eagle’s medium (DMEM); BHK-21, BHK-MHV-N (BHK-21 cells expressing the N protein of MHV strain A59)14, BHK-SARS-N (BHK-21 cells expressing the N protein of SARS)19, Huh-723, L92923 and mouse 17Cl-123 cells were grown in minimal essential medium (MEM). Both types of medium were supplemented with 10% fetal bovine serum, 1× non-essential amino acids, 100 units ml−1 penicillin and 100 μg ml−1 streptomycin. BHK-SARS-N cells were grown using MEM supplemented with 5% fetal bovine serum, 1× non-essential amino acids, 100 units ml−1 penicillin, and 100 μg ml−1 streptomycin, 500 μg ml−1 G418 and 10 μg ml−1 puromycin. BHK-MHV-N and BHK-SARS-N were treated with 1 μg ml−1 doxycyclin 24 h before electroporation. All cells were maintained at 37 °C and in a 5% CO2 atmosphere.

Cultured viruses

MHV-GFP14,15 and HCoV-229E2 were cultured in mouse 17Cl-1 and human Huh-7 cells, respectively. MERS-CoV-EMC24 was cultured in Vero B4 cells. HCoV-HKU1 strain Caen-1 (GenBank: NC_006577) was cultured in human airway epithelial cultures25. ZIKA virus strain PRVABC-59 (GenBank: KX377337) was provided by M. Alves and was cultured in Vero cells. SARS-CoV-2 (SARS-CoV-2/München-1.1/2020/929) was cultured in Vero E6 cells.

Bacterial and yeast strains

E. coli DH5α (Thermo Scientific) and TransforMax Epi300 (Epicentre) were used to propagate the pVC604 and pCC1BAC-His3 TAR vectors8, respectively. The bacteria were grown in lysogeny broth medium supplemented with the appropriate antibiotics at 37 °C overnight. E. coli Epi300 cells containing the different synthetic fragments of SARS-CoV-2 in pUC57 or pUC57mini were grown at 30 °C to decrease the risk of instability and/or toxicity. Saccharomyces cerevisiae VL6-48N (MATα trp1-Δ1 ura3-Δ1 ade2-101 his3-Δ200 lys2 met14 cir°) was used for all yeast transformation experiments26. Yeast cells were first grown in YPDA broth (Takara Bio), and transformed cells were plated on minimal synthetic defined (SD) agar without histidine (SD−His) (Takara Bio). S. cerevisiae VL6-48N-derived clones carrying different YACs were never streaked out together on the same agar dishes as mating switching and resulting recombination might occur at a very low frequency.

Generation of viral subgenomic fragments for TAR cloning using viral RNA, infectious cDNA clones and synthetic DNA

Table 1 displays the templates used to clone the different viral genomes into S. cerevisiae. In general, viral DNA fragments were obtained by RT–PCR of viral RNA extracted from viral strains, isolates and from clinical specimens, using the SuperScript IV One-Step RT–PCR System following the manufacturer’s instructions. Additionally, some fragments were PCR-amplified from vaccinia virus-cloned cDNA2,14, BAC-cloned cDNA16 and plasmid-cloned synthetic DNA (GenScript), using the CloneAmp HiFi PCR Premix according to the manufacturer’s instructions. Accessory sequences, that is, enhanced GFP and porcine teschovirus-1 2A (P2A) for the MERS-CoV-GFP construct, TurboGFP for SARS-CoV-2-GFP and T7 RNA polymerase promoter-hammerhead ribozyme and ribozyme-T7 terminator for human RSV-B, were amplified from plasmids.

For all coronaviruses, the fragment encompassing the viral 5′ untranslated regions (UTR) contained the T7 RNA polymerase promoter sequence immediately upstream of the 5′ end of the genome, and the fragment encompassing the 3′ end of the genome contained a unique restriction site (Extended Data Table 1) downstream of the poly(A) tail.

HCoV-HKU1 synthetic fragments 1–4 were provided individually cloned into pUC57 by GenScript (Extended Data Table 3). MERS-CoV-Riyadh-1734-2015 (GenBank: MN481979) fragments 1–8 were synthesized and cloned into pUC57 by GenScript (Extended Data Table 3), containing homologous regions to TAR vectors pVC604 and pCC1BAC-His3. Similarly, synthetic ZIKA virus fragment 6 cloned into pUC57 contained a hepatitis delta virus ribozyme sequence and pCC1BAC-his3 homology downstream of the viral 3′ UTR (Extended Data Table 3).

The SARS-CoV-2 synthetic DNA fragments were delivered cloned into pUC57 or pUC57mini by GenScript (Supplementary Data 1, Extended Data Table 3). Fragments 1.1, 1.2, 1.3 and 12 contained homologous sequences to pCC1BAC-His3. Each fragment was sequence verified using Sanger sequencing after plasmid isolation using QIAGEN Midiprep kit (QIAGEN). Fragments were released from the vector using the restriction enzymes described in Extended Data Table 3. Restricted fragments were subsequently gel-purified using standard methods27. DNA concentrations and purities of all fragments to be used for TAR cloning were determined using NanoDrop 2000/2000c Spectrophotometer (Thermo Scientific).

In-yeast cloning of viral genomes using TAR

In general, we used overlapping DNA fragments for TAR cloning with overlaps ranging from 45 to 500 bp. As all of our cloning experiments worked well, we did not assess whether the lengths of the overlap affected homologous recombination efficacy. The vectors pVC60411 and pCC1BAC-His38 were used for TAR cloning. These vectors were amplified by PCR using primers containing at least 45-bp overlaps to fragments encompassing the 5′ or 3′ ends of different viral genomes (Supplementary Table 1). Amplification was performed using KOD Hot Start DNA polymerase (Merck Millipore) according to the manufacturer’s instructions. Templates used for generating fragments for TAR cloning are shown in Table 1. TAR cloning was also used to reconstruct the full-length synthetic fragments 5 and 7 in yeast (Extended Data Fig. 4b, c).

Yeast transformation was done using the high-efficiency lithium acetate/SS carrier DNA/PEG method as described elsewhere28. In brief, yeast cells were grown in rich YPDA medium (Takara Bio) at 30 °C with agitation until an optical density at 600 nm of 1.0 was reached. Then, 3 ml of yeast culture was used per transformation event. DNA mixtures were prepared beforehand and contained 100–200 fmol of 3′ and 5′ open ends for all fragments. Transformation mixtures were plated onto SD−His plates (Takara Bio) and incubated at 30 °C for 48 h. Colonies were resuspended in 20 μl of SD−His broth, and DNA was extracted following the GC prep method29. Extracted DNA was used as template for screening by multiplex PCR using the QIAGEN Multiplex PCR kit (QIAGEN) according to the manufacturer’s instruction. One or two multiplex PCRs were designed to encompass different subsets of primer pairs, and cover all desired recombination junctions (Supplementary Table 1). Clones tested positive for all junctions were grown in SD−His until late logarithmic phase, and plasmids were extracted from 500 ml culture using the QIAGEN Maxiprep Kit (QIAGEN) with modifications. In brief, 10 ml of Buffer P1 was supplemented with 1 ml of zymolyase solution (10 mg ml−1 Zymolyase 100-T; 50 mM Tris-HCl pH 7.5; 50% (v/v) glycerol) and 100 μl of β-mercapthoethanol. The mixture was incubated for 1 h at 37 °C before the addition of buffer P2. The rest of the protocol followed the manufacturer’s instructions. DNA preparations were successfully used as templates to generate in vitro transcribed viral RNA even if they contained traces of yeast genomic DNA. In parallel, isolated YACs containing full-length synthetic fragments 5 and 7, as well as SARS-CoV-2 and SARS-CoV-2-GFP viral genomes, were successfully transformed into E. coli TransforMax Epi300 electrocompetent cells (Epicentre) (data not shown).

Stability testing of the YAC containing entire RNA virus genomes in yeast

The stability of viral genomes maintained as YACs in S. cerevisiae was tested for the clones containing MHV-GFP or MERS-CoV for 1 week. A single colony was grown in 20 ml of SD−His liquid medium, 1 ml aliquots were removed and expanded in fresh medium every 12 h. The generation time for each of the clones was estimated to range from 150 to 160 min. After 15–17 passages, each YAC clone was isolated and subjected to sequencing by MinION (Oxford Nanopore Technologies) to obtain the entire YAC sequence. Individual regions for which MinION sequencing did not reveal a clear sequence were resequenced by Sanger sequencing (Microsynth).

Virus rescue

The YAC containing viral cDNA was cleaved at the unique restriction site located downstream of the 3′ end poly(A) tail (Extended Data Table 1). In brief, 1–2 μg of phenol–chloroform-extracted and ethanol-precipitated restricted DNA was resolved in nuclease-free water and used for in vitro transcription using the T7 RiboMAX Large Scale RNA production system (Promega) with m7G(5′)ppp(5′)G cap provided as described previously2. Additionally, a similar protocol was performed on a PCR product of the N gene from corresponding coronaviruses, producing a capped mRNA that encodes the N protein. Then, 1–10 μg of in vitro transcribed viral RNA was electroporated together with 2 μg of the N gene transcript into BHK-21 cells and/or BHK-21 cells expressing the corresponding coronavirus N protein. Electroporated cells were co-cultured with susceptible mouse 17Cl-1, Vero B4 and Vero E6 cells to rescue rMHV-GFP (17Cl-1), rMERS-CoV and rMERS-CoV-GFP (Vero B4), and rSARS-CoV-2, rSARS-CoV-2-GFP and synSARS-CoV-2-GFP (VeroE6). Progeny viruses that were collected from the supernatant immediately after electroporation were termed passage 0 viruses and were used to produce stocks for subsequent analysis. Virus-infected cells were monitored, and images were acquired using an EVOS fluorescence microscope equipped with a 10× air objective. Brightness and contrast were adjusted using FIJI. Figures were assembled using the FigureJ plugin30.

All work involving the rescue and characterization of recombinant MERS-CoV, SARS-CoV and SARS-CoV-2 was performed in a biosafety level 3 laboratory at the Institute of Virology and Immunology, Mittelhäusern, Switzerland under appropriate safety measures with respect to personal and environmental protection.

Virus growth kinetics

In brief, 24 h before infection with MHV-GFP, L929 cells were seeded in a 24-well plate at a density of 3.6 × 105 cells per ml. Cells were washed once with PBS and inoculated with viruses (multiplicity of infection (MOI) = 0.1). After 2 h, the virus-containing supernatant was removed, and cells were washed three times with PBS and supplied with medium as described above. Cell-culture supernatants were collected at the indicated time points after infection. A similar protocol was used for MERS-CoV and MERS-CoV-GFP using Vero B4 cells (MOI = 0.01), and SARS-CoV-2 using Vero E6 cells (MOI = 0.01). Statistical significance was determined by two-sided unpaired Student’s t-test without adjustments for multiple comparisons.

Plaque assay and TCID50

MHV-GFP PFU ml−1 was determined by plaque assay in L929 cells as described previously14. In brief, 24 h before infection, L929 cells were seeded in a 24-well plate at a density of 3.6 × 105 cells per ml. Cells were washed with PBS and inoculated with viruses serially diluted in cell-culture medium at 1:10 dilution. Cells were washed with PBS 1 h after inoculation, and overlaid with 2% methylcellulose mixed at 1:1 with 2× DMEM supplemented with 20% fetal bovine serum, 200 units ml−1 penicillin and 200 μg ml−1 streptomycin. After 24 h of incubation, the overlay was removed and cells were fixed and stained with crystal violet.

The TCID50 assay was performed for MERS-CoV and MERS-CoV-GFP in Vero B4 cells and SARS-CoV-2 and SARS-CoV-2-GFP in Vero E6 cells. In brief, cells were seeded 24 h before infection in a 96-well plate at a density of 2 × 106 cells per plate. Viruses were serially diluted at 1:10 dilution from 10−1 to 10−8. After 72 h of incubation, the medium was removed and cells were fixed and stained with crystal violet. The TCID50 ml−1 titre was determined using the Spearman–Kaerber method31.

The PFU ml−1 of SARS-CoV-2 and SARS-CoV-2-GFP was determined by plaque assay using Vero E6 cells in a 6-well format. In brief, 24 h before infection, Vero E6 cells were seeded at a density of 2 × 106 cells per plate. At the time of infection, cells were washed with PBS and inoculated with viruses serially diluted in cell-culture medium at 1:10 dilution. Cells were washed with PBS 1 h after inoculation and overlaid with 2.4% Avicel mixed at 1:1 with 2× DMEM supplemented with 20% fetal bovine serum, 200 units ml−1 penicillin and 200 μg ml−1 streptomycin. After 48 h of incubation, the overlay was removed and cells were fixed and stained with crystal violet.

Sequencing and computational analysis

Full-length sequences of the SARS-CoV-2 and SARS-CoV-2-GFP cDNAs cloned in yeast were confirmed by Sanger sequencing (Microsynth). All other virus genomes cloned in yeast were confirmed using the Nanopore sequencer MinION from Oxford Nanopore Technologies according to standard protocols. The operating software MinKNOW performed data acquisition and real-time base calling, generating data as fast5 and/or fastq files. Subsequently, the Python command line qcat (Mozilla Public License 2.0., copyright 2018 Oxford Nanopore Technologies, v1.1.0, was run to demultiplex Nanopore reads from fastq files. Alignment of demultiplexed reads to reference sequences was carried out using the Minimap2 program32, producing a fasta file. Mutations of consensus sequences and regions for which the sequences were not clear were verified by Sanger sequencing (Microsynth).

rSARS-CoV-2 and SARS-CoV-2-GFP RNA was sequenced by next-generation sequencing using poly(A)-purified RNA. In brief, 1 × 106 Vero E6 cells were infected with rSARS-CoV-2 clones 1.1, 2.2, 3.1 and rSARS-CoV-2-GFP clones 4.1, 5.2, 6.2 (all passage 1) at an MOI = 0.001. Cellular RNA was prepared using NucleoSpin RNA Plus (Macherey-Nagel) according to the manufacturer’s recommendation. The quantity and quality of the extracted RNA was assessed using a Thermo Fisher Scientific Qubit 4.0 fluorometer with the Qubit RNA BR Assay Kit (Thermo Fisher Scientific, Q10211) and an Advanced Analytical Fragment Analyzer System using a Fragment Analyzer RNA Kit (Agilent, DNF-471), respectively. Sequencing libraries were produced using an Illumina TruSeq Stranded mRNA Library Prep kit (Illumina, 20020595) in combination with TruSeq RNA UD Indexes (Illumina, 20022371) according to Illumina’s guidelines. Pooled cDNA libraries were paired-end sequenced using an Illumina NovaSeq 6000 S Prime Reagent Kit (300 cycles; Illumina, 20027465) on an Illumina NovaSeq 6000 instrument, generating an average of 69 million reads per sample. The quality-control assessments, generation of libraries and sequencing run were all performed at the Next Generation Sequencing Platform, University of Bern, Switzerland. For analysis, the adaptor sequences were trimmed using TrimGalore software (v.0.6.5) and reads shorter than 20 nucleotides in length and/or with a Phred score of less than 20 were removed. Paired-end trimmed reads were mapped to the SARS-CoV-2 genome (GenBank accession MT108784; synthetic construct derived from SARS-2 BetaCoV/Wuhan/IVDC-HB-01/2019) using the Spliced Transcripts Alignment to a Reference (STAR) aligner (v.2.7.0a)33 with default parameters. Before mapping, STAR was also used to generate a genome index for SARS-CoV-2 with the parameters --genomeSAindexNbases 7 and --sjdbOverhang 149. SAMtools (v.1.10) was used to calculate mapped read depth from the resulting mapped read pairs at each position in the genome and subsequently visualized using a variety of software packages in R. Calculations were performed on UBELIX (, the HPC cluster at the University of Bern. Sequencing data have been deposited in the Sequence Read Archive (SRA) of the NCBI (

Apart from MinION and next-generation sequencing data handling, other sequence analyses were performed using Geneious Prime v.2019.2.3. Results from virus growth kinetics were analysed and graphically presented using GraphPad Prism v.8.3.0 for Windows. All figures were created with Adobe Illustrator and

Identification of leader–body junctions of viral mRNAs

To identify reads that mapped discontinuously to the SARS-CoV-2 genome and determine the location of potential transcription regulatory sites (TRS), we pooled reads that mapped to the viral genome as well as unmapped reads and searched for the sequence TTCTCTAAACGAAC (nucleotides 62–75 of MT108784; leader TRS is indicated in bold). We then filtered for reads that had at least 18 nucleotides 3′ of the aforementioned sequence and evaluated whether these reads were compatible with any of the SARS-CoV-2 mRNA sequences. Reads matching these criteria were used as input for the generation of a consensus sequence for each TRS site and analysed using a combination of SAMtools (v.1.10), R and the Integrative Genomics Viewer (IGV). Mapped read depth was also calculated for the discontinuously mapped reads as explained in the previous section.


Recombinant SARS-CoV-2 and SARS-CoV-2-GFP poly(A)-purified RNA used for next-generation sequencing was also used to determine the genome 5′ ends by 5′-RACE. M-MLV reverse transcription (Promega) was performed according to the manufacturer’s instructions using the gene-specific primer pWhSF-ORF1a-R18-655 (Supplementary Table 1) and 10 U RNase Inhibitor RNasin plus (Promega) per 25 μl reaction volume. Following reverse transcription, 1 μl RNase H (5 U μl−1, New England Biolabs) per 25 μl reaction was added, and the mixture was incubated at 37 °C for 20 min. The cDNA was immediately purified with the High Pure PCR product purification kit (Roche) according to the manufacturer’s instructions. A poly(A) tail was added to the cDNA with Terminal Transferase (New England Biolabs) according to the manufacturer’s instructions. Subsequently, a PCR reaction with the tailed cDNA was performed with the primer pair pWhSF-ORF1a-R18-655 and TagRACE_dT16 (Supplementary Table 1) using the HotStarTaq Master Mix (QIAGEN) according to the manufacturer’s instructions with a touchdown cycling protocol: 95 °C for 15 min; 15 cycles of 94 °C for 30 s, 65 °C touchdown to 50 °C for 1 min, 72 °C for 1 min; 25 cycles of 94 °C for 30 s, 50 °C for 1 min, 72 °C for 1 min. Subsequently, 1 μl of this reaction was used for a nested re-amplification with the primer pair pWhSF-5utr-R17-273 and TagRACE (Supplementary Table 1) in a final volume of 50 μl following the same cycling protocol as described above. The PCR fragment was purified using the NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel) according to the manufacturer’s instructions, and the purified PCR fragment was sent to Microsynth for Sanger sequencing with the primer pWhSF-5utr-R17-273 (Supplementary Table 1). Sequencing raw data were assessed using the SeqManTM II sequence analysis software (DNASTAR).

Remdesivir experiment

Remdesivir (MedChemExpress) was dissolved in DMSO and stored at −80 °C in 20 mM stock aliquots. One day before the experiment, Vero E6 cells were seeded in 24-well plates at a density of 8 × 104 cells per well. Cells were infected with synSARS-CoV-2-GFP (passage 1) at MOI = 0.01 or mock-infected as control. Innocula were removed at 1 h after infection, and replaced with medium containing remdesivir (0.2 μM or 2 μM) or the equivalent amount of DMSO. At 48 h after infection, cells were washed once with PBS and incubated in fresh PBS. Images were acquired using an EVOS fluorescence microscope equipped with a 10× air objective. Brightness and contrast were adjusted identically for each condition and their corresponding control using FIJI. Figures were assembled using the FigureJ plugin30.

Immunofluorescence assay

One day before infection, Vero E6 cells were seeded in a 12-well removable chamber glass slide (Ibidi) at a density of 4 × 104 cells per well. Cells were infected with rSARS-CoV-2 clone 3.1 (passage 2) or mock-infected as control. At 6 and 24 h after infection, cells were washed twice with PBS and fixed with 4% (v/v) neutral-buffered formalin. Cells were washed twice with PBS before permeabilization with 0.1% Triton X-100 and blocking with PBS supplemented with 50 mM NH4Cl, 0.1% (w/v) saponin and 2% (w/v) BSA (confocal buffer) for 60 min. Primary antibodies (anti-dsRNA, J2, English and Scientific Consulting, 10010500; and anti-SARS-CoV Nucleocapsid (N), Rockland, 200-401-50) and secondary antibodies (donkey anti-rabbit 594, Jackson ImmunoResearch 711-585-152; and donkey anti-mouse 488, Jackson ImmunoResearch 715-545-150) were diluted in confocal buffer. Slides were covered with 0.17-mm thick, high-performance (1.5H) glass coverslips and mounted using ProLong Diamond Antifade mountant containing 4′,6-diamidino-2-phenylindole (DAPI) (Thermo Fisher Scientific). Images were acquired using an EVOS FL Auto 2 Imaging System equipped with a coverslip-correct 40× air objective. Brightness and contrast were adjusted identically for each condition and their corresponding control using FIJI. Figures were assembled using the FigureJ plugin30.

Serum neutralization assay

One day before the experiment, Vero E6 cells were seeded in a 96-well clear-bottom, black plate at a density of 2 × 106 cells per well. Serum 2 has been described in another study34 as patient serum ID7 (convalescent human anti-SARS-CoV-2 serum). Serum 4 has been described previously as patient serum CSS 2 (convalescent human anti-SARS-CoV serum)35. Sera 1 and 3 were control sera. In brief, all sera were inactivated for 30 min at 56 °C and diluted at 1:10 in OptiMEM. A twofold serial dilution was performed in OptiMEM in a final volume of 50 μl in a separate 96-well plate (dilutions 1:10 to 1:1,280). Then, 50 μl of synSARS-CoV-2-GFP containing 250 TCID50 was added to the diluted sera. The serum–virus mixture was incubated at 37 °C for 60 min, and subsequently added to Vero E6 cells. After 1 h of incubation, supernatants were removed and replaced with medium as described above. At 48 h after infection, expression of GFP and cytopathogenic effects were monitored, and images were acquired using an EVOS fluorescence microscope equipped with a 10× air objective. Brightness and contrast were adjusted identically for each condition and their corresponding control using FIJI. Figures were assembled using the FigureJ plugin30.

Ethical statement

The authors are aware that this work contains aspects of Dual Use Research of Concern (DURC). The benefits were carefully balanced against the risks and the benefits outweigh the risks. Permission to generate and work with recombinant SARS-CoV-2 and SARS-CoV-2-GFP was granted by the Swiss Federal Office of Public Health (A131191/3) with consultation of the Federal Office for Environment, Federal Food Safety and Veterinary Office, and the Swiss Expert Committee for Biosafety.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.