The ACE2 receptor accelerates but is not biochemically required for SARS-CoV-2 membrane fusion

Abstract

The SARS-CoV-2 coronavirus infects human cells via the ACE2 receptor. Structural evidence suggests that ACE2 may not just serve as an attachment factor but also conformationally activate the SARS-CoV-2 spike protein for membrane fusion. Here, we test that hypothesis directly, using DNA-lipid tethering as a synthetic attachment factor in place of ACE2. We find that SARS-CoV-2 pseudovirus and virus-like particles are capable of membrane fusion without ACE2 if activated with an appropriate protease. Thus, ACE2 is not biochemically required for SARS-CoV-2 membrane fusion. However, addition of soluble ACE2 speeds up the fusion reaction. On a per-spike level, ACE2 appears to promote activation for fusion and then subsequent inactivation if an appropriate protease is not present. Kinetic analysis suggests at least two rate-limiting steps for SARS-CoV-2 membrane fusion, one of which is ACE2 dependent and one of which is not. Since ACE2 serves as a high-affinity attachment factor on human cells, the possibility to replace it with other factors implies a flatter fitness landscape for host adaptation by SARS-CoV-2 and future related coronaviruses.

Fig. 1. DNA-tethering and fusion of SARS-CoV-2 pseudoviruses and virus-like particles (VLP). The experiment design is schematized in panel (a), where DNA-functionalized viral particles are added to a microfluidic flow cell and allowed to bind to protein-free liposomes functionalized with complementary DNA. After unbound particles are washed away, fusion is initiated by addition of a soluble protease and monitored via lipid mixing, detected as fluorescence dequenching of Texas Red dye in the VLP or pseudoviral envelope. Representative images of a 10.1 × 9.6 μm sub-micrograph before and after lipid mixing are shown in panel (b) with a fusing particle outlined in magenta. The corresponding fluorescence intensity trace is plotted in panel (c).

Fig. 2. Cumulative distribution functions for fusion by different DNA-tethered SARS-CoV-2 spike particles. Lipid mixing is used as a surrogate for viral membrane fusion, and fusion kinetics are compared between HIV-based pseudoviruses displaying Wuhan or omicron spike versus virus-like particles displaying D614G N501Y spike. Bootstrapped 90% confidence intervals are plotted in dashed lines. The D614G/N501Y virus-like particles fused significantly faster than any of the Wuhan or omicron samples except Wuhan at 1000 μg per mL trypsin (p-values via 2-sample Kolmogorov–Smirnov test of 0.001 for omicron at 200 μg mL⁻¹, 0.04 for Wuhan at 200 μg mL⁻¹, 7 × 10⁻⁴ for omicron at 500 μg mL⁻¹, 7 × 10⁻⁴ for Wuhan at 500 μg mL⁻¹, 0.002 for omicron at 1000 μg mL⁻¹ and 0.11 for Wuhan at 1000 μg per mL trypsin).

Fig. 3. Cumulative distribution functions for fusion by DNA-tethered SARS-CoV-2 spike particles at different trypsin concentrations. Wuhan-pseudotyped particles are plotted in (a) at 200, 500, and 1000 μg per mL trypsin, and omicron-pseudotyped particles are plotted in (b) over the same concentration range. Bootstrapped 90% confidence intervals are plotted in dashed lines. Wuhan and omicron-pseudotyped particles fused at statistically indistinguishable rates at each trypsin concentration tested (p-values of 0.49 at 200 μg mL⁻¹, 0.84 at 500 μg mL⁻¹, and 0.09 at 1000 μg mL⁻¹, via 2-sample Kolmogorov–Smirnov test) and similarly across trypsin concentrations for each of omicron and Wuhan pseudovirus samples. Panel (c) shows the trypsin concentration extended over a million-fold range with no significant change in kinetics observed for Wuhan pseudovirus (all pairwise cumulative distribution functions non-significant via 2-sample Kolmogorov–Smirnov tests). Confidence intervals are omitted from this panel for visual clarity but are shown in Fig. S3.

Fig. 4. Addition of ACE2 speeds fusion of DNA-tethered pseudovirions. Wuhan-pseudotyped particles are plotted in (a) with and without 40 μg mL⁻¹ soluble human ACE2, and omicron-pseudotyped particles are plotted in (b). Rates of lipid mixing by DNA-tethered Wuhan pseudovirions with and without ACE2 are compared to rates of lipid mixing by Wuhan pseudovirions to Calu-3 plasma membrane vesicles (PMV) containing both ACE2 and TMPRSS2 in (c). Bootstrapped 90% confidence intervals are plotted in dashed lines. Because the longest waiting time measured for the plasma membrane vesicles was 224 s, cumulative distribution functions for the subset of DNA-tethered Wuhan pseudoviruses fusing in ≤224 s were replotted in (d) to account for any potential sampling bias. When compared by either method, fusion to PMV was significantly faster (p < 0.001, 2-sample Kolmogorov–Smirnov test).

Fig. 5. Prolonged exposure to ACE2 likely inactivates spike proteins. DNA-tethered pseudovirions were exposed to ACE2 either simultaneous to protease addition or a designated interval prior. Simultaneous exposure speeds membrane fusion, but prior exposure yields fusion kinetics indistinguishable from DNA-tethered pseudovirions exposed to protease without ACE2. This leads to a schema (panel a) where ACE2 binds to, activates, and ultimately inactivates individual spike proteins. Based on the data available, inactivation is fast relative to ACE2 dissociation and/or association. Panel (b) plots cumulative distribution curves for fusion when 40 μg mL⁻¹ ACE2 was added either simultaneous to or the designated interval before addition of 100 μg per mL trypsin. For comparison, fusion kinetics are also plotted for 200 μg per mL trypsin and 100 μg per mL trypsin with 120 μg mL⁻¹ ACE2. The only distributions that were significantly different were simultaneous addition of 40 μg mL⁻¹ ACE2 with either 100 μg mL⁻¹ or 200 μg per mL trypsin (p < 0.005 via Kolmogorov–Smirnov test).

Fig. 6. Gamma-function fits to fusion kinetics. Gamma-function fits were calculated for lipid-mixing CDFS of omicron and Wuhan pseudoviruses with and without ACE2. These fits were calculated either (a) varying N and tau independently, (b) fixing N globally across the data sets and varying tau, or (c) fixing tau and varying N. Fit parameters are listed in Table S2.