Molecular interaction and inhibition of SARS-CoV-2 binding to the ACE2 receptor


S1 subunit specifically binds to purified ACE2 receptors

As SARS-CoV-2 binding to ACE2 receptors is thought to play a key role in the first binding step at the cellular membrane3, we first used FD-curve-based AFM to evaluate at the single-molecule level the binding strength of the interaction established between the glycosylated S1 subunit and ACE2 receptors on model surfaces (Fig. 2a). To mimic cell-surface receptors in vitro, ACE2 receptors were covalently immobilized onto gold surfaces coated with OH- and COOH-terminated alkanethiols using carbodiimide conjugation (see Methods). These model surfaces were imaged by AFM, and the thickness of the grafted layer was validated by a scratching experiment, revealing a deposited layer of 6.1 ± 0.4 nm (mean ± S.D., N = 3) (see Methods and Supplementary Fig. 1). To study the interaction between the S1 subunit and the immobilized ACE2 receptors, we covalently grafted either the purified full S1 subunit or RBD only to the free end of a long polyethylene glycol (PEG)24 spacer attached to the AFM tip7,8,9. To investigate the properties of the binding complex, force–distance (FD) curves were recorded by repeatedly approaching and withdrawing the S1 subunit or RBD- functionalized tip from the ACE2 model surface (Fig. 2a, b). Specific adhesion events were observed on 4–5% of the retraction FD curves at rupture distances >15 nm, which corresponds to the extension of the PEG linker (Fig. 2c and Supplementary Fig. 2), and is in line with studies carried out for other virus–cell-surface receptor systems8,10,11,12. To confirm the specificity of these interactions, we conducted additional independent control experiments using (i) an AFM tip only functionalized with the PEG linker or (ii) toward OH-/COOH-terminated alkanethiol surfaces missing the receptor. The binding frequency observed during those control experiments is significantly lower, thereby confirming the specificity of the S1 subunit/RBD–ACE2 complexes under our experimental conditions (Fig. 2c).

Fig. 2: Probing S-glycoprotein binding to the ACE2 host receptor on model surface.

a Binding of S-glycoprotein subunit (S1 or RBD) is probed on an ACE2-coated surface. b Retraction part of four force–distance curves showing either nonadhesive or specific adhesive curves. c Box plot of specific binding probabilities (BP) measured by AFM between the functionalized tip (S1, RBD, or PEG) and the grafted surface (ACE2 or OH-/COOH-terminated alkanethiol (bare surface)). One data point belongs to the BP from one map acquired at 1 µm/s retraction speed. The square in the box indicates mean, the colored box indicates the 25th and 75th percentiles, and the whiskers indicate the highest and the lowest values of the results. The line in the box indicates median. N = 12 (S1, RBD), 18 (PEG), and 9 (S1, RBD vs. bare surface) maps examined over 4 (S1, RBD), 6 (PEG), and 3 (S1, RBD vs. bare surface) independent experiments. d Bell–Evans model describing a virus-receptor bond as a two-state model. The bound state is separated from the unbound state by a single energy barrier located at distance xu. koff and kon represent the dissociation and association rate, respectively. e, f Dynamic force spectroscopy (DFS) plot showing the distribution of the rupture forces as a function of their loading rate (LR) measured either between the S1 subunit and the ACE2 receptor (N = 1052 data points) (e) or between the RBD and the ACE2 receptor (N = 1490 data points) (f). The error bar indicates s.d. of the mean value for a single interaction (0–200 pN). The solid line represents the fit of the data with the Bell–Evans fit. Experiments were reproduced at least four times with independent tips and samples. g, h The BP is plotted as a function of the contact time for S1 subunit and RBD on ACE2 model surfaces, and data points were fitted using a least-squares fit of a monoexponential growth. One data point belongs to the BP from one map acquired at 1 µm/s retraction speed for the different contact times. Experiments were reproduced three times with independent tips and samples. P values were determined by two-sample t test in Origin. The error bar indicates s.d. of the mean value. Source data are provided as a Source Data file.

Exploring the dynamics of S1 subunit–ACE2 interaction

Single-molecule force-probing techniques, such as FD-based AFM, measure the strength of a bond under an externally applied force, enabling to get insights into the binding free-energy landscape. According to the Bell–Evans model13,14, an external force stressing a bond reduces the activation-energy barrier toward dissociation and, hence, reduces the lifetime of the ligand-receptor pair15 (Fig. 2d). The model also predicts that far-from-equilibrium, the binding strength of the ligand-receptor bond is proportional to the logarithm of the loading rate (LR), which describes the force applied on the bond over time. To investigate the kinetics of the probed complex, FD curves were recorded at various retraction rates and contact times (Fig. 2e–h). Dynamic force spectroscopy (DFS) plots were obtained for both S1 subunit (Fig. 2e) and RBD (Fig. 2f) binding toward immobilized ACE2 receptors. In each case, the unbinding force increases linearly with the logarithm of the LR, as observed earlier for other virus-receptor bonds8,10,11,16,17. To determine whether single- or multiple-bond rupture between S1/RBD and ACE2 is taking place, bond strengths (every single gray data point in Fig. 2e, f) were analyzed through distinct discrete ranges of LRs, plotted as force histograms and further fitted with multipeak Gaussian distribution, as established previously11,16 (Supplementary Figs. 3 and 4). Using this distribution, we are able to determine the most probable unbinding force of each force peak (maximum of rupture force distribution; black dots plotted over mean LR of this range in Fig. 2e, f), and can determine if single or multiple interactions were taking place. The presence of multiple parallel unbinding events is first observed in the distribution of rupture forces with the presence of multiple Gaussian fits. The histograms show that most probably only single interactions were taking place; thus, the Bell–Evans model15 was used to fit the data enabling to interpret the binding complex as a simple two-state model, in which the bound state is separated from the unbound state by a single energy barrier (Fig. 2d). From the slope of the fit, we estimated the length scale of the energy barrier (xu). We obtained very close values, xu = 0.81 ± 0.05 nm and 0.79 ± 0.04 nm for both the S1 subunit and RBD, showing that we are probing similar bonds (Fig. 2e, f). The kinetic off-rate (koff) or dissociation rate is obtained from the intercept of the fit (at LR = 0) yielding koff values of 0.008 ± 0.005 s−1 and 0.009 ± 0.006 s−1 for S1 subunit and RBD, respectively. These values are in good agreement with reported values obtained by surface plasmon resonance for the S glycoprotein (koff = 0.003 s−1)18 and the RBD subunit (koff = 0.008 s−1) binding to ACE2 receptors19.

Assuming that the receptor-bond complex can be approximated by a pseudo-first-order kinetics, we also estimated the kinetic on-rate (kon) from our single-molecule force spectroscopy experiments11 (Fig. 2g, h). This association rate is extracted from the binding probability (BP) measured at various contact times, and depends on the effective concentration described as the number of binding partners (ligand + receptor) within an effective volume Veff accessible under free-equilibrium interaction. Veff can be approximated by a half-sphere with a radius including the linker, the viral glycoprotein (S1 subunit or RBD) and the ACE2 receptor. For both the S1 subunit and RBD, we observed that the binding frequency increased exponentially with contact time, and we extracted an interaction time of ~0.250 ms, leading to a kon of 6.4 × 104 M−1 s−1 and 8.0 × 104 M−1 s−1, respectively. Finally, the dissociation constant KD is calculated as the ratio between the koff and the kon, yielding values around ~120 nM for both complexes. This value corresponds to a high-affinity interaction, confirming the specificity of the complexes established by SARS-CoV-2 with the ACE2 cell-surface receptor, which in turn results in a long lifetime of the virus attachment to the cell surface. Other interaction studies between SARS-CoV (80% sequence homology to SARS-CoV-2) and ACE2 reported specific, high-affinity association values also in the nM range20. For comparison, a variety of examples for low- as well as high-affinity interactions between other virus-receptor pairs are summarized in Dimitrov et al.21 and include influenza A—SA (mM) or HIV-1—CD4 (nM) interactions. For single-molecule interactions, the bond lifetime τ can be directly related to the inverse kinetic off-rate (τ = koff−1), resulting here in a τ of 125 ms for the S1 subunit and 111 ms for the RBD, respectively. Of course, at the virion level, the overall bond lifetime will increase with the multivalence of the interaction. By definition, high-affinity interaction has a long lifetime as the dissociation constant KD is defined as the ratio between koff and kon. For high-affinity interactions, the KD is in the nM range, leading to koff « kon and therefore maintaining the interaction in its bond state for very long times, making the development of anti-binding molecules targeting this interaction more difficult. Finally, we also used optical biolayer interferometry (BLI) to confirm the kinetic parameters characterizing this interaction, and obtained very close affinities in the same nM range as AFM experiments (Supplementary Fig. 5). Taken together, our in vitro experiments confirm that SARS-CoV-2 binding to the ACE2 receptors is mediated by the RBD–ACE2 interface as our experimental conditions did not highlight any significant difference between S1 subunit and RBD binding.

Validation of the interaction on living cells

Next, we wanted to investigate whether the interaction probed on isolated receptors is also established in physiologically relevant condition. To this end, we performed binding assays on living A549 cells (human adenocarcinoma alveolar basal epithelial cells). While this cell line is widely used as a type II pulmonary epithelial cell model, it has been shown recently that those cells are incompatible with SARS-CoV-2 infection22. Interestingly, ACE2 expression positively correlated with the differentiation state of epithelia. Although undifferentiated cells (cultured at low confluency) only express little ACE2, overexpression of ACE2 in undifferentiated A549 cells facilitated virus entry23. We transiently transfected ACE2–eGFP in A549 cells (A549–ACE2) and probed S1-subunit binding to those cells as well as to A549 cells (serving as internal control) (Fig. 3a and Supplementary Fig. 6). Confocal images showed ACE2–eGFP receptors homogeneously distributed in small domains at the surface of A549 cells (Fig. 3b). Guided by fluorescence (Fig. 3c), we chose areas in which both cell types, i.e., transfected (A549–ACE2, green fluorescence) and nontransfected (A549, no fluorescence) cells, were in proximity to one another. Having both A549 cell types in one image area served as a direct control to evaluate whether interactions measured by the functionalized tip were indeed due to specific binding to fluorescent ACE2–eGFP receptors, and to evaluate the extent of other types of interactions (Fig. 3c–e). In such area, we simultaneously recorded a height image (Fig. 3d) and the corresponding adhesion map (Fig. 3e), which were reconstructed from FD curves recorded for each topographic pixel. The retraction part of FD curves showed specific adhesion events mainly on A549–ACE2 cells, with a significantly higher BP (Fig. 3f), as exemplified with the presented adhesion map that shows 20.1% of adhesive pixels on the A549–ACE2 cell versus 13.5% on the control cells (Fig. 3e and Supplementary Fig. 7). Specific binding forces (and corresponding LR) were extracted from force vs. time curves recorded on A549–ACE2 cells (Fig. 3g) and overlaid on the DFS plot obtained on purified ACE2 receptors (Fig. 3h). To explore a wide range of LR, we probed the interaction at various frequencies and amplitudes (see Methods). We observed a very good alignment between the data obtained on purified receptors and on living cells confirming the physiological relevance of our results obtained on model surfaces.

Fig. 3: Probing S-glycoprotein binding to the ACE2 host receptor on living cells.
figure3

a Binding of S-glycoprotein subunit 1 (S1) is probed on A549 and A549–ACE2 cells. b Confocal microscopy (z stack) of A549–ACE2–eGFP (green) cell transduced with plasma membrane BFP (blue). c Overlay of eGFP and DIC images of a mixed culture of A549 and A549–ACE2–eGFP cells. d, e Force–distance (FD)-based AFM topography image (d) and the corresponding adhesion map (e) in the specified area in (c). The frequency of adhesion events is indicated. f Box plot of the binding probability between S1 and A549 cells (gray) or A549–ACE2 cells (green) without and after injection of cyclic RGD (cRGD, checked boxes) or sialic acid (SA, dashed boxes), respectively. The square in the box indicates mean, the colored box indicates the 25th and 75th percentiles, and the whiskers indicate the highest and the lowest values of the results. The line in the box indicates median. g Force versus time curves showing either a nonadhesive curve (bottom) or specific adhesive curves acquired at different LRs (LR1–LR3). h DFS plot showing the distribution or the rupture forces measured either between the S1 subunit and the ACE2 on model surfaces (black dots, extracted from Fig. 2e), and between the S1 subunit and ACE2-overexpressing A549 cells acquired at three different LRs (blue and red dots) (N = 403). Blue dots belong to a data set acquired in fast-force volume mode, with a retraction velocity of 20 µm s1 (LR1). Red dots belong to data sets acquired in peak force tapping mode with 0.125 kHz peak force frequency and 375-nm amplitude (LR2) or at 0.25 kHz and 750 nm (LR3), respectively. The error bar indicates s.d. of the mean value. Histograms of force distribution on A549–ACE2 cells for LR1–LR3 are shown on the side. For experiments without injection of cRGD or SA, data are representative of at least N = 11 cells from N = 6 independent experiments. The data for blocking experiments with cRGD or SA were acquired for at least N = 4 cells from N = 2 independent experiments. P values were determined by two-sample t test in Origin. Source data are provided as a Source Data file.

S1 subunit binding to the cell involves other receptors

Our FD-based AFM experiments performed on living cells put in evidence that the S1 subunit interacts even on control cells with a frequency ≈10% although the expression level of ACE2 should be very low as the cells are not differentiated. Nevertheless, some evidence pointed out that human CoV S glycoproteins possess sialic acid (SA)-binding sites and in particular to 9-O-acetyl-sialogycans24, and that integrins could also be a receptor for the SARS-CoV-2 (ref. 25), which possesses a RGD motif close to the ACE2-binding site. To evaluate whether these other receptors could be involved during the early binding steps to the cell surface, we performed additional experiments by injecting 9-O-acetyl-sialogycans to block interaction with cell-surface SA, or added cyclo-RGD (cRGD) to compete with the interactions with integrins. After SA injection, the binding frequency was reduced on A549 cells down to ~7% and to ~10% on ACE2-transfected cells (Fig. 3f and Supplementary Fig. 8). For integrins, injection of cRGD only reduces the binding frequency of ~1–2% on both cell types, which is in good agreement with the fact that integrins are mostly expressed on the bottom of the cell26. Altogether, these data obtained on cells by AFM represent to date the best evidence that S1–ACE2 complex is established in physiologically relevant conditions and underlines the complex situation with multiple cell-surface receptors accounting for the whole interaction.

Inhibition of S1-subunit binding using ACE2-derived peptides

Human recombinant soluble ACE2 (hrsACE2) is currently being considered for treatment of COVID-19 (refs. 27,28). However, ACE2 is involved in many key cellular processes, such as blood-pressure regulation and other cardiovascular functions. Therefore, hrsACE2 treatment could lead to dysregulation of those vital processes and subsequently cause deleterious side effects for treated patients. To avoid any interference of the ACE2 homeostasis, we wanted to test whether small ACE2-derived peptides can also interfere with SARS-CoV-2 binding, by blocking binding sites on the S glycoprotein. To this end, we synthetized four different peptides (sequences provided in Supplementary Fig. 9), which have been selected to mimic the regions of ACE2 that interact with the S1 subunit as determined by the crystal structure29, and we tested their binding inhibition properties using our single-molecule force spectroscopy approach (Fig. 4a, b). We first measured the BP between the S1 subunit and the ACE2 in the absence of peptide (0 µM), with a contact time of 250 ms, as reference, and then injected our ACE-derived peptides at three different concentrations (1, 10, and 100 µM). For the four peptides, we observed a progressive reduction of the BP as a function of the concentration confirming a specific inhibition. In addition, for each peptide, we noticed a reduction of >50% of the probed interactions already for the 1–10 µM concentration, suggesting a 50% inhibitory concentration (IC50) in the µM range. The [22–44] peptide shows the highest inhibition of the S1–ACE2 complex formation with a measured reduction in the BP of ~76%. The [22–57] peptide shows a similar inhibition potential (~74%), suggesting that the additional amino acids do not influence the overall affinity of the peptide for the S1 subunit, as also confirmed by molecular dynamics (MD) simulations showing that although the peptide 22–57 is longer, less H bonds are established between the peptide and the RBD domain (Supplementary Fig. 10). Overall, these results are in good agreement with the structural insights because these peptides are derived from the N-terminal helix of the ACE2 and therefore form with the RBD interface an important network of hydrophilic interactions (including nine hydrogen bonds and a salt bridge). Within the ACE2–RBD complex, the [351−357] fragment is also part of a “hot binding spot” that results in our test by a good score with a reduction of ~60% of the initial specific BP. Finally, the [22–44–g–351–357] peptide was also synthetized and tested based on the fact that in the crystal structure, the distance between S44 and L351 is close enough to be filled by a single amino acid. A glycine residue was added between the two fragments because the two ACE2 fragments have opposite directionality, and glycine has a high propensity to form reverse turns. Nevertheless, under our experimental conditions, we did not notice any strong improvement in the binding inhibition. Altogether, our in vitro assays at the single-molecule level provide direct evidence that ACE2-derived peptides are strong candidates to potentially inhibit SARS-CoV-2 binding to ACE2 receptors (Fig. 4c).

Fig. 4: Anti-binding effects of ACE2-derived peptides on S1-subunit binding.
figure4

a Efficiency of blocking peptides is evaluated by measuring the binding probability of the interaction between the S1 subunit and ACE2 receptor on model surface before and after incubation of the functionalized AFM tip with the four different peptides at increasing concentration (1–100 µM). b Histograms, with the corresponding data points overlaid in dark gray, showing the binding probability without peptide (0 µM) and upon incubation with 1, 10, or 100 µM of ACE2-derived peptides ([22–44], [22–57], [22–44–g–351–357], and [351–357]). The binding probability measured with a polyethylene glycol (PEG) tip enables to evaluate the nonspecific binding level. The prediction of the structure of the ACE2-derived peptides is shown in the inset. The structure of the peptides is based on the structure of the peptide in the crystal structure (PDB ID: 6m0j). For the [22–44–g–351–357] peptide, its structure was generated using homology modeling41. The error bar indicates s.d. of the mean value. c Graph showing the reduction of the binding probability. Control with ddH2O is provided in the inset showing that repetitive measurements do not result in a similar decrease of the binding probability. Data are representative of at least N = 3 independent experiments (tips and sample) per peptide concentration. P value was determined by two-sample t test in Origin. The error bar indicates s.d. of the mean value. Source data are provided as a Source Data file.

ACE2-derived peptide blocks specific binding to living cells

Finally, we tested whether the [22−57]-binding inhibition peptide could also prevent S1-subunit binding in the cellular context (Fig. 5). The interaction between the S1 subunit and the confluent layer of a coculture of A549 and A549–ACE2 cells was probed before and after addition of the peptide at 100 µM. Before injection, cells overexpressing the ACE2 receptors (A549–ACE2) show higher BP (9.4 ± 1.6% vs. 19.4 ± 7.3%, for A549 and A549–ACE2, respectively) (mean ± S.D., N = 4) (Fig. 5a–d), in good agreement with our previous observation (Fig. 3f). After injection of the [22−57] ACE2-derived peptide, we observed a significant decrease of the BP on both cell types (Fig. 5e, f). In particular, the BP on A549–ACE2 cells significantly drops (~70%), reaching a level close to the one of the control cells. Taking into account that undifferentiated A549 cells express little ACE2 and are poorly infected by CoV23, this result supports the biological relevance of our ACE2-derived peptide acting as potential inhibitor capable of efficiently blocking SARS-CoV-2 binding.

Fig. 5: Blocking of S1-subunit binding using ACE2-derived peptide on living cells.
figure5

a Box plot showing that the reduction of binding probability measured the S1-subunit-derivatized tip and a mixed culture of A549 and A549–ACE2 cells upon injection of the [22–57] ACE2-derived peptide. The square in the box indicates mean, the colored box indicates the 25th and 75th percentiles, and the whiskers indicate the highest and the lowest values of the results. b Overlay of eGFP and DIC images of a mixed culture of A549 and A549–ACE2–eGFP cells. FD-based AFM topography images (c, e) and the corresponding adhesion map (d, f) recorded in the specified area in (b) (scanned with a scan angle) before (c, d) and after (e, f) incubation of the tip with the [22–57] ACE2-derived peptide. The frequency of adhesion events is indicated. Data are representative of at least N = 4 cells from N = 2 independent experiments. P values were determined by two-sample t test in Origin. Source data are provided as a Source Data file.

In conclusion, we investigated the interaction established between the SARS-CoV-2 S glycoprotein and the ACE2 receptor using single-molecule force spectroscopy. We demonstrated a specific binding mechanism between the S1 subunit and the ACE2 receptor. By comparing the binding of the S1 subunit and the RBD toward the ACE2 receptor, our experiment evidenced that both domains interact with the same kinetic and thermodynamic properties toward the ACE2 receptor, highlighting that SARS-CoV-2 binding to ACE2 is dominated by the RBD/ACE2 interface. Our measurements show that under our physiologically relevant conditions, the RBD binds the ACE2 receptor with an intrinsic high affinity (~120 nM), which could even be further stabilized at the whole-virus level, thanks to possible multivalent bonds between the S-glycoprotein trimer and ACE2 dimer.

Based on the available crystal structures of the molecular complex, we examined how several ACE2-derived peptide fragments could interfere with the S1–ACE2 complex formation. While all tested peptides show binding inhibition properties, peptides mimicking the N-terminal helix of the ACE2 receptor show the best results. Both [22–44] and [22–57] peptides exhibit an anti-binding activity with IC50 in the µM range, resulting in a >70% decrease in the BP observed by AFM on purified receptor and >70% on living cells. On the cellular model, we observed that the BP drops to the level of the control cells (undifferentiated A549 cells) that are poorly infected by CoV23. Therefore, those peptides appear as strong therapeutic candidates against the SARS-CoV-2 infection.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *