The Delta variant SARS-CoV-2 spike protein uniquely promotes aggregation of pseudotyped viral particles

Individuals infected with the SARS-CoV-2 Delta variant, lineage B.1.617.2, exhibit faster initial infection with a higher viral load than prior variants, and pseudotyped particles bearing the SARS-CoV-2 Delta variant spike protein induce a faster initial infection rate of target cells compared to those bearing other SARS-CoV-2 variant spikes. Here, we show that pseudotyped particles bearing the Delta variant spike form unique aggregates, as evidenced by negative stain and cryogenic electron microscopy (EM), flow cytometry, and nanoparticle tracking analysis. Viral particles pseudotyped with other SARS-CoV-2 spike variants do not show aggregation by any of these criteria. The contribution to infection kinetics of the Delta spike’s unique property to aggregate is discussed with respect to recent evidence for collective infection by other viruses. Irrespective of this intriguing possibility, spike-dependent aggregation is a new functional parameter of spike-expressing viral particles to evaluate in future spike protein variants.

single amino acid change from an aspartic acid to a glycine at amino acid 614 (D614G), that increased transmissibility [22] possibly by stabilizing the prefusion spike trimer [23], increasing the RBD up/receptor accessible conformation rate [24], and increasing the spike density on the virion [25]. This D614G variant, designated Pango lineage B.1 [26], rapidly supplanted the original Wuhan strain globally [27] by May 2020, and the D614G substitution is present in all subsequent variants [28]. Next, the VOC Alpha (lineage B.1.1.7) emerged in the UK and became dominant worldwide by early 2021, owing to additional substitution mutations occurring in the spike RBD that increased ACE2 receptor affinity, deletions in the NTD that increased immune escape, and a mutation in S2 that may enhance membrane fusion potential [29][30][31], collectively leading to about two-fold enhanced transmissibility of Alpha compared to other contemporaneously circulating variants [32].
Functional consequences of spike mutations can be tested using engineered mimics of enveloped viruses, called pseudotyped particles (PPs). PPs are produced by co-transfecting producer cells with plasmids encoding a capsid protein from a parental virus (typically a retrovirus or arbovirus capsid), the spike glycoprotein of interest, and a reporter gene that produces a fluorescent or luminescent protein signal upon host cell entry. The capsid core buds efficiently from the producer cell while incorporating the heterologous spike and encapsulating the reporter gene [45,46].
Recently, studies comparing the entry rate of PPs packaged with various SARS-CoV-2 spike variants into target cells showed that PPs bearing the Delta spike (Delta PPs) drove and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint markedly faster initial infection, and greater infection overall, than other variants [35,43]. This phenotype is similar to real-world reports of faster initial infection by Delta SARS-CoV-2 [37].
To search for an ultrastructural correlate to the increased initial infection rate of Delta Cryo-electron microscopy of Delta PP aggregates showed a variety of distances between the envelopes of PP, many consistent with the spike-to-spike interactions seen in negative staining.
Implications for Delta spike-mediated aggregation for the kinetics of initial viral infection are discussed.

Cell culture and Production of SARS-CoV-2 Variant Spike PPs:
HEK 293T cells were a gift from the laboratory of Dr. Gary Whittaker (Cornell University). EPI_ISL_6699757. To increase expression of the spike protein on the cell surface, the C-terminal and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint 19 amino acids were deleted [47,48]. 48 hours post-transfection, the supernatants were collected and centrifuged at 290 g for 7 min at 4 o C. The supernatants were then passed through a 0.45 µm syringe-tip filter [45] and stored on ice or 4 o C until use.

Negative stain transmission electron microscopy (TEM) and immunogold labeling:
PPs were stored on ice after harvest from producer cells and processed as follows for TEM within 4 hours of harvest or after overnight storage at 4 o C. All reagents were obtained from Electron Microscopy Sciences (EMS, Hatfield, PA) and all steps took place at room temperature unless otherwise specified. PPs, suspended by gentle trituration, were adhered to freshly glow discharged, formvar and carbon-coated, 300-mesh gold EM grids (EMS) by inverting grids for 2 min on 5 µl drops of culture medium supernatant containing PPs. Grids were rinsed by transferring the grids across 3 droplets of filtered DPBS (Dulbecco's phosphate buffered saline, Mg and Ca-free, Gibco, Paisley, UK). All solutions were filtered through a 0.22 µm syringe filter with RC membrane (EMS) before contact with the EM grid. For negative staining, grids were rinsed twice with filtered, distilled water and then placed on a drop of filtered 1% aqueous uranyl acetate for 1 minute, and then blotted to dryness with filter paper (Whatman #1).
To immunogold label the spike protein S1 subunit, grids with adhered PPs were floated on drops of blocking solution containing 2% BSA (Sigma, St. Louis, MO) in DPBS for 10 min.
Primary antibody to SARS-CoV-2 spike subunit 1 (Sino Biological, Beijing, Cat. No. 40589-T62), was diluted 1:25 in blocking solution and grids were transferred to drops of primary antibody for 30 minutes. Then, grids were transferred across two drops of blocking solution for 10 minutes before incubation with 10 nm gold-conjugated donkey-α-rabbit secondary antibody diluted 1:20 in blocking solution for 30 min. Grids were covered during incubation steps to prevent evaporation. Finally, grids were rinsed with 3 drops of DPBS and negative stained as described above. Grids were observed using a FEI Tecnai T20 transmission electron microscope (Thermo Fisher, Waltham, MA) operated at 200 kV, and images were acquired using a NanoSprint1200 CMOS detector (AMT Imaging, Woburn MA). Images were prepared for display using Photoshop 2022 (Adobe). The sizes of the Delta variant PP aggregates were measured and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint using Fiji. A freehand perimeter was traced around each aggregate to obtain area and size measurements.

Flow cytometry of PPs:
PPs at 4 hours, 24 hours and 7 days were labelled with 0.5 μM BODIPY FL (Thermo

Nanoparticle tracking analysis (NTA) of PPs:
The size and concentration of PPs at 4 hours, 24 hours, and 7 days were characterized by NTA using a NanoSight NS300 (Malvern Instruments Ltd, Malvern, UK) equipped with a 405nm laser. PPs were diluted 1:100 in PBS, loaded into 1 ml syringes, introduced into the sample chamber using a syringe pump and three 60 second videos were recorded per sample.
Acquisition and analysis settings were kept constant for all samples. Size distributions and concentrations were analyzed in NTA software, version 3.4. Aggregate percentages were calculated by summing the area under the size distribution curve for all species greater than the monomer, normalized by the total area under the curve.

Cryo Electron Microscopy:
7 µl of PPs at 4 or 24 hours were adhered to the top side of freshly glow discharged, 200-mesh Quantifoil gold R2/1 EM grids (EMS), back-blotted for 6-8 s after a 30-s pre-blotting time and immediately plunge frozen into liquid ethane using a GP-EM2 (Leica Microsystems, Wetzlar, Germany) with a chamber maintained at 20 °C and 95% humidity. Grids were screened using a FEI Tecnai T20 transmission electron microscope (Thermo Fisher, Waltham, MA) operated at 200 kV, and images were acquired using SerialEM [49] and direct electron detector K2 Summit (Gatan Inc., Pleasanton, CA) at a nominal magnification of 25,000x at a binned pixel size of 3.04 Å/px and about -3 µm defocus. At a dose rate of 8 e -/px/sec, 10 s exposures with and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint 0.2 s frames were collected. Each 50-frame-movie was motion corrected using SerialEM. To detect the size of the larger Delta variant PP aggregates, a medium magnification of 1700x and a binned pixel size of 88.7 Å/px was used.

PPs bearing the Delta spike variant cluster into aggregates:
Negative stain TEM was conducted to compare the structure of Delta PPs to those bearing other SARS-CoV-2 spike variants: D614G, Alpha, Delta AY.4.2, and Omicron BA.1. As a control, PPs produced in the absence of a spike, referred to as "Bald", were also examined. For each set of experiments, PPs were produced in parallel, maintained on ice or at 4 o C, handled under identical conditions, and prepared for TEM within 4 hours of harvest. By negative stain TEM, all spike variant PPs showed a population of ~120 nm diameter spherical particles ( Figure   1 A-D), the expected size of MLV-based particles [50]. PPs were abundant in the culture medium without concentration, producing a density on the surface of the EM grid of about five PPs per 60 µm 2 of EM grid surface. Also uniformly distributed on the grid surface were ~25 nm lipoprotein-like particles, sometimes closely associated with PPs, including the Bald PPs.  Figure 1D and Figure 2). To quantify the degree of and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint aggregation of the variant PPs, low magnification images were collected at random, and PPs counted and categorized as single PPs spaced more than 60 nm apart (beyond the length of two spike proteins), or in a cluster of two, three or more PPs. In three experimental replicates, D614G and Alpha PPs were present as single PPs about 75% of the time. Omicron BA.1 PPs (two experimental replicates) were also primarily present as single PPs (not quantified). Bald PPs aggregated at the same rate as D614G and Alpha, suggesting that these spike variants had a comparable and minimal effect on PP interactions that promote aggregation. was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022 The average Delta PP aggregate occupied 0.3 µm 2 and was 700 nm in its longest axis (57 Delta PP aggregates containing eight or more PPs measured in three experimental replicates, average area 0.30 ± 0.23 µm 2 , average major axis 708 ± 300 nm) ( Figure 1D and Figure 2). The number of PPs in an average sized aggregate was ~25 PPs, however this is likely an underestimate as PPs inside the aggregate were obscured from view, particularly in the case of  Figure 2D). and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105   (wblcdf(x,a,b)-wblcdf(7,a,b))/ (1-wblcdf(7,a,b)) where wblcdf is the Weibull Distribution and x is the PPs per aggregate, 7 is the truncation value, and a and b are the Weibull scale and shape parameters. Coefficients (with 95% confidence bounds) are a = 11.47 (8.35, 14.60) and b = 0.68 (0.57, 0.79). Goodness of fit measure Sum of Squared Error (SSE) = 0.009 and Standard Error of Regression (RMSE) = 0.021, and Resid. are the residuals or differences between the model fit and the data.

Delta spike variant PPs continue to aggregate in solution
The same samples of PPs that were prepared within 4 hours of harvest, were stored at 4°C overnight (still suspended in culture medium) and prepared again for negative stain TEM the next day. PPs of D614G, Alpha, Omicron BA.1, and Bald PPs remained predominantly singles and showed minimal aggregation. However, after overnight storage, the aggregates of Delta and Delta AY.4.2 PPs were much larger, less frequent, and required lengthy searching with the microscope to find. In parallel, there was a concomitant reduction in single Delta and Delta AY.4.2 PPs on the grid surface indicating that single PPs continued to become sequestered in aggregates while diffusing in the culture medium and aggregates themselves likely aggregated.
Aggregates at 24 hours were irregularly shaped, sometimes donut-shaped, and up to several microns in size (Figure 3). and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105

Flow cytometry and NTA confirm population of aggregated Delta Spike variant PPs
Since negative stain TEM permits observation of only those entities that adhere to the surface of the EM grid, we applied two alternative methods, flow cytometry and nanoparticle tracking analysis (NTA) which sample particles present in the entire solution volume, to corroborate the negative stain TEM observations. All variant PPs measured by flow cytometry and NTA within four hours of harvest from producer cells detected particles predominantly in the single PP range (not shown, n=2). However, after 24 hours stored at 4 o C, flow cytometry and NTA of PPs detected a significant population of larger particles in the Delta PPs that was not present in the other variant PPs, or the Bald PPs ( Figure 4A and C, n=2). Because NTA does and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint not count aggregates above 1 µm in size, the larger aggregates observed by negative stain TEM are not reflected in the NTA analysis. However, the abundance of single Delta PPs (peak ~150 nm), reduced by half compared to other PPs, is likely explained by their sequestration into aggregates ( Figure 4C). Despite these differences in sensitivity, a significant population of Delta PPs were observed in large aggregates by flow cytometry and NTA, that were absent in PPs bearing other variant spikes and Bald PPs. After storage of PPs at 4 o C for 7 days, flow cytometry was repeated and still minimal aggregation of D614G, Alpha, or Bald PPs was detected, suggesting that these variants and Bald PPs have less tendency to aggregate under these conditions ( Figure 4B, n=2). However, after 7 days there was evidence for some deterioration in the functionality of PP samples, as cell entry was very low (data not shown). and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022

Cryo-electron microscopy of Delta variant PPs
To further evaluate the interactions between spike variant PPs, we used cryo-EM which preserves PPs in hydrated vitreous ice, free from heavy metal stains and potential artifacts of  Figure 5C). There is no evidence to support a significant difference in aggregate sizes between Delta and Delta AY.4.2.
and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint  Unlike SARS-CoV-2 which buds into the ERGIC compartment during final assembly, retroviruses, including MLV, generally bud directly from the plasma membrane (but not always, see [51]). Thus, it is not known when or if Delta variant SARS-CoV-2 aggregates and this is currently being investigated. Delta SARS-CoV-2 could aggregate while budding into the ERGIC, during egress through alkalized lysosomal organelles [52], after egress in the extracellular milieu, or even on the target cell plasma membrane. The ability to aggregate may depend on the concentration of viral particles in each environment. The fact that Delta PP continue to aggregate (Figure 2, 4, and 5) is consistent with a mass action mechanism for Delta PP aggregation. Furthermore, to produce PPs, a 19 aa C-terminal truncated version of each variant spike was expressed, which has been shown to increase the amount of spike incorporated into the PP envelope and PP infectivity [47,48]. Truncations in the cytoplasmic tail could modify properties of the spike ectodomain structure and function [53,54]. Though each variant spike possessed the same truncation, it is not impossible that the truncation uniquely effected the Delta spike ectodomain, conferring the aggregation property.
Examples of Delta PPs with apparent spike tip interactions were observed by negative stain but these interactions did not extend to lateral aggregation of spike proteins on the and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. There are, however, three residues, E156, F157 and R158, in the NTD of Delta and Delta AY.4.2 that are uniquely and identically mutated: substitution E156G, and deletions at F157 and R158. It is possible that these three mutations in the NTD are sufficient to bestow the aggregation property, or perhaps the unique combination of mutations occurring on the Delta spike drive aggregation synergistically. and is also made available for use under a CC0 license. was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint The clustering of Delta PPs could account for the faster and larger initial infection observed in entry assays with the Delta PPs [35]. Because the number of spike trimers is larger on an aggregate comprising multiple PPs, and the cell surface contact area is larger for any collision between the aggregate and a target cell, the effective on-rate for aggregate binding should be larger, resulting in faster binding. Furthermore, the avidity of the aggregate to the target cell would be enhanced manyfold due to the multiple potential binding partners on a single contacting surface. Moreover, the increased dwell time at that contact area will allow for diffusional and conformational motions of proteins and lipids to increase the chance of membrane fusion, as these factors are important for avoiding hemifusion and promoting full fusion [55]. All these factors should lead to the relatively higher initial rate of PP entry into target cells from aggregated PPs. Whether or not aggregates could enable the simultaneous delivery of multiple copies of entry reporter genes to target cell is not clear, since the PPs need not display receptor and may not fuse to each other, even in an endosome, so one entry event may or may not mean copies of genes enter. Implicitly, there would be more overall binding events for unaggregated PPs, each at another site. However, if the probability for PP entry was low due to unbinding, then the factors discussed above to increase PP avidity would tend to increase overall fusion and its rate.
In summary, an ultrastructural analysis of retrovirus pseudotyped particles with coronavirus spike proteins from variants of SARS-CoV-2 led to a serendipitous discovery of significant aggregation when the Delta variant of the spike is expressed, but not upon expression of three other variants. Viral aggregation can impart fitness benefits by protecting virions from environmental hazards and by effecting simultaneous delivery of multiple viral genomes, or collective infection [56]. Notably, collective infection can favor initial infection in some contexts [57]. Likely, the size and number of virions per aggregate is important to increasing infectivity. Too large and it would effectively reduce infectious units below a threshold. Too few virions in an aggregate and the benefit of collective infection is not gained.
The unique property of the Delta spike to aggregate PPs may underlie the faster infection by and is also made available for use under a CC0 license.
was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 The copyright holder for this preprint (which this version posted April 8, 2022. ; https://doi.org/10.1101/2022.04.07.487415 doi: bioRxiv preprint Delta PPs. Furthermore, spike mediated aggregation could be part of the molecular mechanism by which Delta variant SARS-CoV-2 achieves increased transmissibility and faster infection with a higher viral load. The continued clustering over time indicates that the underlying factor for Delta PP clustering may be the spike tip interactions of the spike protein, which in turn may indicate an adhesivity of the viral surface recognized by the immune system thus altering the balance of host antiviral response towards inflammation.