The Spike D614G mutation increases SARS-CoV-2 infection of multiple human cell types
Isolating the SARS-CoV-2 Spike D614G substitution from the ORF1b P314L variant it travels with, through pseudotyped lentiviral particles and a trans-complementation assay with replication-competent virus, shows that the single Spike change raises entry efficiency across human lung, liver and colon cell lines without altering S1 affinity for ACE2.
- Senior authors
- Benjamin R tenOever; Neville E Sanjana
- Correspondence
- Benjamin R tenOever; Neville E Sanjana
Research areas & themes
Citation
Daniloski Z, Jordan TX, Ilmain JK, Guo X, Bhabha G, tenOever BR, Sanjana NE. The Spike D614G mutation increases SARS-CoV-2 infection of multiple human cell types. eLife. 2021. Volume 10, article e65365.
DOI 10.7554/elife.65365. PMID 33570490. PMCID PMC7891930.
One-sentence contribution
Isolating the SARS-CoV-2 Spike D614G substitution from the ORF1b P314L variant it travels with, through pseudotyped lentiviral particles and a trans-complementation assay with replication-competent virus, shows that the single Spike change raises entry efficiency across human lung, liver and colon cell lines without altering S1 affinity for ACE2.
Executive summary
By mid-2020 a SARS-CoV-2 Spike variant carrying glycine rather than aspartate at position 614 had become dominant worldwide, but because that change is in linkage disequilibrium with an ORF1b P314L variant within the same clade, population genetics could not attribute any phenotype to Spike alone, and groups disagreed about whether the rise reflected selection or a founder effect. This study separates the two by building the substitution into a human-codon-optimised Spike coding sequence and testing it in isolation. EGFP lentiviral particles pseudotyped with either D614 or G614 Spike were used to transduce four human cell lines, two with endogenous ACE2 and two engineered to overexpress it, across four viral doses. G614 particles transduced more cells at every dose in every line, by 1.3 to 2.4-fold in the endogenous ACE2 lines and 1.5 to 7.7-fold in the overexpressing lines, with only a small opposing difference in particle titre. Bio-layer interferometry found comparable binding of D614 and G614 Spike S1 subunit to immobilised human ACE2, so affinity of that subunit does not account for the difference. Western blotting showed the G614 Spike to be more resistant to proteolytic cleavage both in transfected cells and on virions, without a difference in Spike incorporation. A trans-complementation assay, co-transfecting ACE2 and one Spike variant into HEK293T cells before infection with replication-competent virus, reproduced the increase.
Scientific context
The first sequenced SARS-CoV-2 isolate and most January and February 2020 sequences carried aspartate at Spike position 614. From February onward the glycine variant rose, and in a survey of 22,103 GISAID genomes in early June 2020 the authors found roughly 72 percent carrying G614. Whether this reflected a functional advantage was contested, with some groups proposing increased transmissibility under positive selection and others arguing that the available evidence did not support selection. The interpretive obstacle the paper identifies is specific. In the A2a clade the Spike change is in linkage disequilibrium with an ORF1b P314L variant, so any association measured in sequence databases cannot be assigned to Spike. Two clinical datasets, from Sheffield and from the University of Washington, had reported roughly threefold higher viral RNA by diagnostic PCR in patients carrying G614. Separately, most SARS-CoV-2 vaccines then in development were built on the original D614 Spike sequence, which made the functional status of the substitution relevant beyond virology.
Central question
Does the Spike D614G substitution by itself, separated from the linked ORF1b P314L variant, alter the efficiency with which SARS-CoV-2 enters human cells, and if so through what step.
Experimental strategy
The design is built around the linkage problem. Because natural isolates cannot supply an isogenic comparison, the substitution is introduced by site-directed mutagenesis into a cloned Spike coding sequence, so the only difference between the two arms is the single codon.
Pseudotyped lentiviral particles then isolate entry from every other stage of the viral life cycle. Because the particles carry an EGFP reporter and cannot replicate, the flow cytometry readout at three days counts successful entry and integration events rather than rounds of spread, so a difference cannot be attributed to downstream replication or to the linked polymerase variant. Four cell lines spanning lung, liver and colon, two relying on endogenous ACE2 and two overexpressing it, test whether any effect depends on receptor abundance or tissue of origin. Four viral volumes per line guard against a dose-specific artefact, a no-pseudotype arm establishes the background, and quantitative PCR of particle RNA checks that the two preparations were comparably titred.
Mechanism is then approached at the two steps the Spike protein performs. Bio-layer interferometry with purified S1 subunit and immobilised human ACE2 asks whether receptor affinity differs. Western blotting with a C-terminal C9 tag, which allows full-length Spike and the S2 and S2 prime fragments to be resolved on the same blot, asks whether proteolytic processing differs, and doing this both in transfected cells and on purified virions distinguishes cleavage during production from the state of the delivered particle. Normalising total Spike to the p24 capsid protein separately tests whether the variants differ in how much Spike each particle carries, which is an alternative explanation for a transduction difference.
Finally, because pseudotypes may not report the behaviour of Spike in its native context, a trans-complementation assay tests the substitution against replication-competent virus without requiring a reverse genetics system. Co-transfecting ACE2 together with one Spike variant into cells that are otherwise poorly infectable means that only transfected cells are readily infected, so the supplied Spike variant shapes the outcome. Low multiplicities and early time points were used because the authors reasoned that higher doses and longer infections would mask the contribution of the transfected Spike.
Key findings
-
The G614 variant rose to roughly 72 percent of 22,103 surveyed genomes by early June 2020 (Figure 1a). Across 56 countries, G614 prevalence showed a small but significant positive correlation with case-fatality rate (r = 0.29, p = 0.04) (Figure 1b), which the authors present as a smaller effect than a previously reported correlation computed on a roughly tenfold smaller dataset. This is an ecological correlation and is not offered as evidence of causation.
-
Reanalysis of published Sheffield and University of Washington patient PCR data showed a roughly 5 threshold cycle offset between the two sites, attributable to methodological differences, but a consistent difference between variants within each site of 1.6 and 1.8 threshold cycles (Figure 1c and 1d). The authors read the consistency as suggesting a biological difference.
-
G614-pseudotyped particles transduce human cells more efficiently than D614 particles, at every dose in all four lines (Figure 2c). Fold increases were 1.4 to 1.9 in Calu-3, 1.3 to 2.4 in Caco-2, 1.8 to 4.6 in A549-ACE2 and 1.5 to 7.7 in Huh7.5-ACE2 (Figure 2d). Particles lacking an attachment protein gave negligible transduction.
-
The difference is not a titre artefact. Viral RNA measured with two independent primer sets differed by about 7 percent in favour of D614, which if anything means the transduction advantage of G614 is slightly underestimated (Figure 2 supplement 2).
-
ACE2 binding by the S1 subunit is comparable between variants. Bio-layer interferometry fit best to a 2 to 1 heterogeneous binding model, giving KD1 of 8.45 nanomolar and KD2 of 127 nanomolar for D614 and KD1 of 18.0 nanomolar and KD2 of 92.7 nanomolar for G614 (Figure 3b, 3c and Table 1). The authors read this as indicating the transduction phenotype is independent of S1 affinity for ACE2.
-
G614 Spike is more resistant to proteolytic cleavage. In transfected HEK293FT cells the ratio of cleaved to full-length Spike was roughly 2.5-fold lower for G614 (Figure 3d to 3f), and on purified pseudotyped particles it was roughly 1.4-fold lower (Figure 3g to 3i). No significant difference was found in Spike incorporation into particles when total Spike was normalised to p24 (Figure 3 supplement 1).
-
The substitution alters predicted MHC binding for at least one epitope, with predicted affinity for HLA-A 02 01 shifting from 58 to 221 nanomolar (Figure 3 supplement 2). This is computational prediction, not measured immunogenicity.
-
The effect holds for replication-competent virus. In the trans-complementation assay, cells supplied with G614 Spike showed significantly more infection than those supplied with D614 Spike at 12, 18 and 24 hours after infection, at multiplicities of both 0.01 and 0.1 (Figure 4b to 4d). Untransfected cells showed minimal infection, under 1 percent in most cases.
Mechanistic model
The study does not establish a definitive mechanism, and the authors describe the cleavage result as suggesting a possible mechanism rather than demonstrating one.
What the data constrain is the following. The entry advantage of G614 is a property of Spike alone, since it is reproduced when the substitution is the only difference between two otherwise identical constructs and when the ORF1b variant is absent entirely. It does not arise from higher affinity of the S1 subunit for ACE2 under the conditions measured, and it does not arise from more Spike per particle. It is accompanied by less proteolytic cleavage of Spike, both during production in cells and on the assembled particle.
The model the authors propose is that reduced cleavage leaves a greater fraction of functional Spike, uncleaved and retaining the receptor-binding domain, on each newly assembled virion, which would raise the probability of productive attachment. They are careful about the boundaries of this proposal. They note that other groups working with full-length Spike trimers reported that G614 shifts the trimer toward an ACE2-binding competent conformation, which they suggest means trimers behave differently from the S1 monomers measured here, so the absence of an affinity difference in their assay does not exclude a conformational contribution. They also note that two studies using isogenic replication-competent viruses found no difference in Spike cleavage or incorporation, and that reported differences across the field may partly reflect the choice of pseudoviral system, producer cell handling and transfection reagents, which can affect protease activity. They state that future work is required to establish whether G614 Spike is processed differently on live virus, and cite a report that G614 may additionally be processed by elastase-2.
Conceptual or technical advance
The trans-complementation assay is the methodological contribution. It allows a Spike variant to be tested against replication-competent SARS-CoV-2 in an isogenic fashion without a reverse genetics system, by supplying both the receptor and the Spike variant to cells that are otherwise poorly permissive, which puts such experiments within reach of laboratories that cannot build recombinant coronaviruses. The concordance the authors report between the pseudotyped lentiviral result and the replication-competent result also supports the broader use of pseudotypes for studying Spike variants, which they note explicitly as a practical implication. Conceptually, the work separates a phenotype from a linked genetic marker, converting a contested population-genetic observation into a measurable property of one amino acid.
Relationship to the broader research program
This study shares its first author, its two corresponding laboratories and much of its cell line panel, including A549-ACE2, Huh7.5-ACE2, Caco-2 and Calu-3, with the genome-scale CRISPR screen published by the same collaboration in Cell in 2021. The tenOever laboratory contributed the work with replication-competent virus while the Sanjana laboratory contributed the pseudotype and genetic engineering components, and the two are joint corresponding authors.
Category 3 synthesis. Read alongside the CRISPR screen, which found that RAB7A loss reduces surface ACE2 and that entry depends on endosomal machinery, this paper addresses the complementary side of the same entry step from the viral rather than the host direction. Whether receptor availability and Spike cleavage state interact is a question the two papers raise together and neither answers.
Related publications
- Daniloski and colleagues, 2021, Cell, companion. Same first author and same two corresponding laboratories, sharing the engineered A549-ACE2 and Huh7.5-ACE2 lines and the Caco-2 and Calu-3 panel, addressing host requirements for entry where this paper addresses a viral determinant of it.
- Korber and colleagues, 2020, predecessor. Source of the Sheffield patient PCR data reanalysed here and of the argument about G614 spread.
- Wagner and colleagues, 2020, predecessor. Source of the University of Washington patient PCR data reanalysed here.
- Yurkovetskiy and colleagues, 2020, companion. Reported that G614 shifts full-length Spike trimers toward an ACE2-binding competent conformation, a result the authors discuss as complementary to and possibly reconciling with their monomer binding data.
- Plante and colleagues, 2020, and Hou and colleagues, 2020, companion. Isogenic replication-competent SARS-CoV-2 comparisons that found no difference in Spike cleavage or incorporation, discussed here as discrepant with the pseudotype result.
- Shang and colleagues, 2020, methodological foundation. Source of the human-codon-optimised Spike coding sequence used for mutagenesis.
Limitations and boundaries
All experiments are in transformed human cell lines, two of which overexpress ACE2 at non-physiological levels, and the largest fold effects were seen in those overexpressing lines, so the magnitude of the effect is system dependent. There is no primary airway tissue, organoid or animal component, and nothing in the study addresses transmissibility between hosts or clinical severity, which the authors note remains uncertain and uncorrelated with variant status in the two clinical studies they cite. The binding measurement used purified S1 subunit monomers rather than full-length trimers, and the authors themselves identify this as the likely reason their result differs from a study using trimers, so the absence of an affinity difference is bounded to that construct. The cleavage measurement is made on transfected cells and on lentiviral particles, not on authentic SARS-CoV-2 virions, and the authors state that whether the same holds on live virus requires further work, noting that two isogenic virus studies found no such difference. The trans-complementation assay depends on transient overexpression of both ACE2 and Spike, so the ratio of supplied Spike to virus-encoded Spike is not controlled, and readouts are restricted to 12 to 24 hours at low multiplicity because later times were expected to mask the effect. The correlation between G614 prevalence and case-fatality rate is ecological, drawn across countries, and subject to every confounder that applies to such comparisons. The MHC binding result is prediction from an algorithm with no experimental immunology. Only one isolate, USA-WA1/2020, was used for the live virus work, and the study predates the variants of concern that followed.
Audience summaries
25 words
A single amino acid change in the SARS-CoV-2 Spike protein, tested in isolation, made the virus enter human lung, liver and colon cells more efficiently.
75 words
Early in the pandemic a Spike variant called D614G took over globally, but it always travelled with a second mutation, so its own effect was unclear. Building the change into Spike on its own showed that particles carrying it entered four types of human cell more efficiently, and the same held for live virus supplied with the variant. Receptor binding by the outer Spike fragment was unchanged, but the variant Spike was cut less readily by host proteases.
150 words
The SARS-CoV-2 Spike D614G substitution sits in linkage disequilibrium with an ORF1b P314L variant, so its phenotype cannot be read from sequence databases. Introducing it alone into a codon-optimised Spike, EGFP lentiviral particles pseudotyped with G614 transduced Caco-2 and Calu-3 cells 1.3 to 2.4-fold better and A549-ACE2 and Huh7.5-ACE2 cells 1.5 to 7.7-fold better than D614 particles, across four doses, with particle RNA content differing by only about 7 percent in the opposite direction. Bio-layer interferometry with purified S1 subunit showed comparable ACE2 binding for both variants, while western blotting showed roughly 2.5-fold less cleavage of G614 Spike in transfected cells and roughly 1.4-fold less on virions, with no difference in Spike incorporation. A trans-complementation assay, supplying ACE2 and one Spike variant to cells before infection with replication-competent virus, reproduced the advantage at 12, 18 and 24 hours. The cleavage explanation is proposed, not established.
Documented publication relationships
- Identification of Required Host Factors for SARS-CoV-2 Infection in Human Cells — companion.
- Identification of Required Host Factors for SARS-CoV-2 Infection in Human Cells — companion.