Effective intervention strategies are urgently needed to control the COVID-19 pandemic. Human angiotensin-converting enzyme 2 (ACE2) is a membrane-bound carboxypeptidase that forms a dimer and serves as the cellular receptor for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). ACE2 is also a key negative regulator of the renin–angiotensin system that modulates vascular functions. We report here the properties of a trimeric ACE2 ectodomain variant, engineered using a structure-based approach. The trimeric ACE2 variant has a binding affinity of ~60 pM for the spike protein of SARS‑CoV‑2 (compared with 77 nM for monomeric ACE2 and 12–22 nM for dimeric ACE2 constructs), and its peptidase activity and the ability to block activation of angiotensin II receptor type 1 in the renin–angiotensin system are preserved. Moreover, the engineered ACE2 potently inhibits SARS‑CoV‑2 infection in cell culture. These results suggest that engineered, trimeric ACE2 may be a promising anti-SARS-CoV-2 agent for treating COVID-19.
The current COVID-19 pandemic, caused by SARS‑CoV‑2, has infected more than 29 million people worldwide, leading to over 900,000 deaths, with devastating socio-economic impacts. Effective intervention strategies are urgently needed to control the pandemic.
Since the outbreak of the virus, several therapeutic approaches have been evaluated in the hope of providing a viable treatment for COVID-19. First, convalescent sera from individuals recovered from the infection were used with encouraging results1,2,3, but also some drawbacks (for example, batch variations, possible blood-borne pathogens and blood-type matching). Second, patient-derived, potently neutralizing monoclonal antibodies have been isolated, which could provide a more powerful passive immunotherapy than convalescent sera4,5,6,7. Third, structure-guided design of peptide-based viral entry inhibitors has yielded promising leads in in vitro assays8,9; their efficacy requires further clinical evaluation. Fourth, known drugs or drug candidates, including remdesivir, favipiravir and ribavirin (viral RNA polymerase inhibitors); lopinavir and ritonavir (viral protease inhibitors); as well as hydroxychloroquine, corticosteroids and interferons (with more complicated antiviral mechanisms), have been repurposed as COVID-19 therapuetics10,11. Among them, remdesivir has received Emergency Use Authorizations from the US Food and Drug Administration, while hydroxychloroquine has been shown to be ineffective12. Finally, many new therapeutic candidates are in various stages of development (refs. 13,14; https://www.bio.org/policy/human-health/vaccines-biodefense/coronavirus/pipeline-tracker). While the infection resolves on its own in most asymptomatic and mild cases over time, COVID-19 in severe cases appears to progress in two phases—initial active viral replication in the respiratory system and subsequent excessive immune responses leading to multiple organ failure and possible death15. Thus, antivirals alone may be insufficient to change the course of disease progression for the population that needs intervention the most if administrated too late.
Human ACE2 is the cellular receptor for SARS-CoV-2 and binds the receptor-binding domain (RBD) of the spike (S) protein of the virus to promote viral entry into the host cells and initiate infection16,17. The interactions between soluble ACE2 and SARS-CoV-2 RBD have been studied extensively within a very short period of time18,19,20. The intact ACE2 is a type I membrane glycoprotein containing an extracellular ectodomain that has metallopeptidase activity. Its neck domain near the transmembrane anchor mediates dimerization20. ACE2 is also a key negative regulator of the renin–angiotensin system (RAS)—a major hormone system, conserved in mammals and some other vertebrate animals, for modulating vascular function21,22. The RAS controls extracellular fluid volume and blood pressure homeostasis by regulating the levels of renin and angiotensins in the circulation. Renin cleaves angiotensinogen to release angiotensin I (Ang I), which can be further processed by angiotensin-converting enzyme (ACE) into angiotensin II (Ang II)—a vasoconstrictive peptide that raises blood pressure and increases the extracellular fluid volume in the body by activating the Ang II receptors, including angiotensin II receptor type 1 (AT1R)23. ACE2 primarily converts Ang II to angiotensin-(1–7) (Ang 1–7), which is a vasodilator, thereby counter-balancing the effect of ACE/Ang II and playing critical roles in preventing hypertension and tissue damages24.
The protective roles of ACE2 in acute respiratory distress syndrome (ARDS) and acute lung injury have been demonstrated in animal models25,26,27. A recombinant soluble human ACE2 (rhACE2) was recently reported to block SARS-CoV-2 infection in cell culture and human organoids28, prompting a phase 2 clinical trial for use of rhACE2 as a treatment for patients with COVID-19 (NCT04335136). Thus, administration of exogenous ACE2 may be a promising therapeutic strategy for treating COVID-19, because it could not only block viral spread but also modulate the RAS to prevent organ injury. We therefore set out to design a series of ACE2 variants to enhance their binding affinity for SARS-CoV-2 S protein and their potency in blocking SARS-CoV-2 infection.
Structures of soluble ACE2 in complex with SARS-CoV-2 S protein trimer
To facilitate design of ACE2-based viral fusion inhibitors, we first determined, by cryo-EM, the structures of a monomeric soluble ACE2 (residues 18–615) in complex with a stabilized soluble SARS-CoV-2 S protein trimer (Extended Data Fig. 1; ref. 29). We prepared the complex by mixing the two proteins because the monomeric ACE2 dissociates from S trimer very rapidly30. After three-dimensional (3D) classification (Extended Data Figs. 2 and 3 and Table 1), we found four distinct classes that represent S trimer with different numbers of ACE2 bound: no ACE2 (in the one-RBD-up conformation); and with one to three ACE2s bound (Fig. 1). Consistent with previous findings with ACE2 binding to SARS-CoV S protein as well as a recent SARS-CoV-2 study31, ACE2 interacts with the RBD in its up conformation. While the N-terminal domain of S1 shifts outwards slightly, the S2 portion remains largely unchanged upon ACE2 binding, even when compared with our recently published structure of the full-length S protein in the closed prefusion conformation30. The structure of the complex with three ACE2s bound is not symmetrical, as the distances between the C termini of the three ACE2s (residue Tyr613) are 107 Å, 109 Å and 120 Å, respectively (Extended Data Fig. 4). This distance in the complex with two ACE2s bound is 110 Å. These observations suggest that there is a modest degree of freedom for the up conformation of RBD when ACE2 is bound. All the substrate binding sites of the bound ACE2s face away from the three-fold axis of the S trimer (Extended Data Fig. 4), incompatible with the structure of the full-length ACE2 dimer in complex with the amino acid transporter B0AT1, in which the two active sites of the two protomers are facing each other20. If the B0AT1-bound ACE2 dimer is indeed the form recognized by SARS-CoV-2, it appears that only one ACE2 protomer in the dimer can bind one RBD in an S trimer unless there are unexpectedly large structural rearrangements in either ACE2 or S.
Design of ACE2 variants to enhance its binding affinity to SARS-CoV-2 S trimer
Measurement of the binding kinetics of soluble monomeric ACE2 (ACE2615; Fig. 2) to the SARS-CoV-2 S trimer shows a relatively fast dissociation rate30, limiting its ability to compete with the membrane-bound ACE2 on the surface of a target cell. We therefore sought to enhance the effective affinity by creating a trimeric form of soluble ACE2. We fused a trimerization foldon tag, derived from bacteriophage T4 fibritin32, to the C-terminal end of the ACE2 peptidase domain (residue 615) through an 11-residue flexible linker, a construct we refer to as ACE2615-foldon (Fig. 2). We have also created another version (ACE2615-LL-foldon) with a slightly longer linker (LL) with 13 residues between ACE2 and the foldon tag to assess its impact on binding. To further strengthen the interaction between ACE2 and the RBD, we introduced mutations guided by high-resolution structures (Extended Data Fig. 5; ref. 18) at three different positions in the ACE2–RBD interface, namely T27, H34 and K353. Substitution with a bulky hydrophobic residue at each of these sites may enhance hydrophobic interactions between the two proteins and slow the dissociation (Extended Data Fig. 5). We designed five mutants, T27Y, T27W, H34W, K353Y and K353W, in the ACE2615-foldon background. Finally, for comparison, we also include two versions of dimeric forms, ACE2m615-Fc and ACE2740-Fc, both fused to an Fc domain of an immunoglobulin G (Fig. 2a). ACE2m615-Fc contains H374N and H378N mutations at its peptidase active site and ACE2740-Fc includes the neck domain that mediates dimerization in the full-length ACE2 (ref. 20).
To produce the soluble recombinant ACE2 and its variants, we transfected HEK293 cells with the expression constructs of the monomeric and trimeric forms containing a C-terminal His tag and purified the proteins by Ni-NTA (nitrilotriacetic acid) and gel filtration chromatography. The two dimeric forms were purified by protein G resin followed by gel filtration chromatography. While the monomeric and dimeric forms of soluble ACE2 were mostly secreted into cell medium, as judged by western blot, the trimeric ACE2615-foldon and its mutants were largely retained inside the cells. We therefore purified the secreted monomer and dimers from the cell supernatants and all the trimers from the cell lysates. Most proteins eluted from a size-exclusion column as a major symmetrical peak, regardless of their secretion status (Extended Data Fig. 6). The ACE2740-Fc protein containing the dimerizing neck domain appeared to aggregate substantially more than other constructs. The ACE2615-foldon K353W mutant aggregated completely, and we therefore did not pursue this construct any further. Only the fractions from the major peak for each construct were pooled and used for subsequent analyses. We have also compared the secreted ACE2615-foldon with the form purified from the cells and their biochemical properties are essentially identical (Extended Data Fig. 7).
Binding to SARS-CoV-2 soluble S trimer
We next measured binding of these recombinant ACE2 constructs to the stabilized soluble S trimer by bio-layer interferometry (BLI). As shown in Fig. 2b and Extended Data Fig. 8, the monomeric ACE2615 had a fast dissociation rate and a Kd of 77 nM, consistent with the measurement that we reported recently using surface plasmon resonance (ref. 30). The dimeric ACE2m615-FC bound slightly more tightly (Kd ~ 22 nM). The dimeric ACE2740-Fc also bound more strongly than did the monomer (Kd ~ 12 nM), although the dimer formed by the neck domain is not compatible with two ACE2 peptidase domains interacting with two distinct RBDs in a single S trimer (Extended Data Fig. 4). A possible explanation is that the neck domain-mediated dimerization is not very strong and that the ACE2 peptidase domains are much more flexible than what the full-length ACE2 structure has indicated20, particularly in the absence of B0AT1. The trimeric ACE2615-foldon interacted with the S trimer much more strongly than any of the monomeric or dimeric forms, with a Kd of 1.2 nM. ACE2615-LL-foldon with a longer linker between ACE2 and foldon showed an additional modest affinity enhancement (Kd ~ 0.62 nM). The two interface mutants, ACE2615-foldon-T27W and ACE2615-foldon-T27Y, bound substantially more tightly than did the trimeric wild-type ACE2, with Kd of ~60 and ~90 pM, respectively. While the H34W afforded a slight affinity increase, the K353Y mutation decreased affinity by more than 25-fold. Overall, these data show that our structure-guided design to increase the affinity of ACE2 to the S trimer, by trimerizing the receptor and by modifying the interface, has indeed been effective.
ACE2 peptidase activity and AT1R activation
We performed two independent assays to determine the enzymatic activity of these ACE2 constructs. First, we directly measured the peptidase activity using a synthetic peptide substrate that releases a free fluorophore upon ACE2 cleavage. In Fig. 3a, concentrations of all the proteins were normalized based on the number of active sites, and the fluorophore release was monitored continuously up to ~40 min. While all the trimeric forms showed essentially the same specific activity, the monomeric ACE2615 and the dimeric ACE2740-Fc had lower specific activities (Table 2 and Supplementary Table 1). The ACE2m615-Fc was inactive due to the mutations at the active site. Thus, all these ACE2 constructs with the wild-type sequence at the active site retained their wild-type peptidase activity.
To further support this conclusion, we next tested the ability of the ACE2 constructs to block Ang II-induced activation of AT1R. In Fig. 3b, an Ang II peptide was first directly incubated with various ACE2 proteins and then added to HEK293 cells transfected with an AT1R expression construct. Activation of AT1R was monitored by changes in the intracellular calcium concentration. When the digestion reaction was quenched by EDTA at time 0 as a control, all the mixtures with different ACE2 proteins could efficiently activate AT1R, suggesting that nothing in our protein preparations inhibited Ang II-mediated AT1R activation. In contrast, when the digestion was allowed to proceed for 40 min, all AEC2 constructs except for the inactive ACE2m615-Fc effectively blocked AT1R activation, presumably by converting Ang II to Ang 1–7, in agreement with the peptidase activity results.
Inhibition of SARS-CoV-2 infectivity in cell culture
We used three different assays to assess the neutralization potency of the ACE2 constructs in blocking SARS-CoV-2 infection. The circulating strain during the early days of the pandemic contained a D614 residue in its S protein, but it has subsequently been replaced by an emerging strain harboring a G614 substitution33. It has been difficult to generate pseudotyped viruses with the full-length S from the D614 strain34. We first used a murine leukemia virus (MLV)-based pseudovirus assay with a D614 S construct lacking 19 residues of the CT, which incorporates efficiently into pseudoviruses. In Fig. 4a, the monomeric ACE2615 showed the lowest potency with a half-maximal inhibitory concentration (IC50) value of 24.1 μg ml−1. The two dimeric forms, ACE2m615-Fc and ACE2740-Fc, and the trimeric mutant ACE2615-foldon-K353Y had greater potency, with IC50 values ranging from 1.2 to 6.3 μg ml−1. The two trimeric forms, ACE2615-foldon and ACE2615-LL-foldon, and the trimeric mutant ACE2615-foldon-H34W neutralized with even greater potency and an IC50 value around 0.6 μg ml−1. The most potent inhibitors were ACE2615-foldon-T27W and ACE2615-foldon-T27Y, which had IC50 values of 0.21 and 0.25 μg ml−1, respectively. Thus, the neutralization potency of these ACE2 constructs correlates strictly with their binding affinity, suggesting that the interaction between ACE2 and S is the principal determinant of neutralization of the virus pseudotyped with the CT-truncated S (D614). Neutralization by these ACE2 proteins in the human immunodeficiency virus (HIV)-based pseudovirus assay using a full-length S derived from the G614 circulating strain showed a very similar pattern, with ACE2615 the weakest and ACE2615-foldon-T27W and ACE2615-foldon-T27Y the most potent (Fig. 4b). Furthermore, when they were analyzed by a plaque assay with an authentic SARS-CoV-2 (Fig. 4c), the neutralization pattern was almost identical to that from the MLV-based assay (Table 2 and Supplementary Table 2). The IC50 values for ACE2615-foldon-T27W and ACE2615-foldon-T27Y were 0.08 and 0.14 μg ml−1, respectively. These results indicate that the engineered ACE2 constructs are very potent agents for blocking SARS-CoV-2 infection in cell culture.
A recombinant human ACE2, named APN01, is currently under evaluation as a treatment for COVID-19 in a phase 2 clinical trial (NCT04335136), primarily based on the favorable results from a previous phase 1 safety and tolerability trial (NCT00886353) in a small number of healthy individuals35, as well as on the recent evidence that the protein blocks SARS-CoV-2 infection effectively in vitro28. APN01 is a soluble ACE2 construct expressing residues 1–740 and probably dimerizes by the neck domain36, similar to ACE2740-Fc used in our study. We demonstrate here that our best trimeric ACE2 variant, ACE2615-foldon-T27W, has >200-fold-higher binding affinity for the soluble SARS-CoV-2 S trimer, and ~5-fold- and ~13-fold-higher neutralization potency against pseudoviruses and authentic viruses, respectively, than does ACE2740-Fc, while its peptidase activity and ability to block AT1R activation remain essentially unchanged.
Using a deep mutagenesis screening approach, a recent study has identified a dimeric ACE2 variant containing multiple mutations, which led to higher affinity binding to the RBD, but also a substantial loss in the catalytic activity (4–8-fold decrease), than the parental construct with the wild-type sequence37. One of the mutations from the mutagenesis screening is T27Y, coinciding with our structure-based design. Another study reported a trimeric ACE2 construct with the wild-type sequence38, similar to our ACE2615-foldon protein. Moreover, de novo protein design based on the ACE2 helix that interacts with the RBD without the catalytic domain has led to picomolar SARS-CoV-2 miniprotein inhibitors39. Our approach, however, distinguishes the S protein binding and the peptidase activity, which can be manipulated separately to maximize the therapeutic benefits of an ACE2 construct.
Although the molecular mechanism by which a soluble ACE2 blocks SARS-CoV-2 infection as a decoy receptor is obvious, its protective role against lung injury—a hallmark of severe COVID-19 cases—appears to be more complicated in humans than in animal models. ACE2 knockout mice have more severe ARDS symptoms than do wild-type mice, whereas ACE2 overexpression appears to be protective25. Moreover, administration of recombinant ACE2 reduces severity of lung injury in mice caused by respiratory syncytial virus or influenza virus26,27. In humans, rhACE2 was well tolerated with a short half-life35, but its infusion did not appear to ameliorate ARDS at least in a small number of patients40. A recent case report described using APN01 to treat a patient with severe COVID-19 with a very encouraging outcome41. The ongoing phase 2 clinical trial with expected 200 participants may provide much needed information regarding the therapeutic potential of recombinant ACE2 in the context of COVID-19. Structure-based modifications presented here may help augment therapeutic efficacy while suppressing adverse effects, if any.
The safety in humans of the foldon trimerization tag, derived from the bacteriophage T4 fibritin32, has been demonstrated by vaccine trials against HIV-1 and SARS-CoV-2 in clinical settings42,43. However, a dose in milligrams per kilogram of body weight of the trimeric ACE2 proteins as a therapeutic is likely much greater than that used as a vaccine (for example, 50–250 μg of foldon-stabilized HIV-1 gp140 protein per injection; ref. 42). If the foldon tag induces unacceptable levels of side effects at a high dose in animals or humans, other trimerization domains, such as those in abundant human collagens44, can be considered. Further improvements of these ACE2-based therapeutic candidates include modifications to enhance protein stability by introducing additional disulfide bonds (which may reduce the catalytic activity), to modulate peptidase activity by mutating residues in or near the active site and to increase its in vivo residence time in circulation, by strategies such as PEGylation45.
The structure of the membrane-bound ACE2 dimer formed by the neck domain is not compatible with a binding mode of two protomers interacting with two RBDs from a single S trimer, as depicted in Extended Data Fig. 4. Much stronger binding of ACE2740-Fc to the S trimer as well as greater neutralization potency than those of monomeric ACE2615 and even dimeric ACE2m615-Fc clearly indicate avidity, suggesting that the ectodomain of either ACE2 or SARS-CoV-2 S protein has much greater flexibility than the cryo-EM structures imply20,30,46,47. If the multivalency of the trimeric S protein on the surface of virion and the dimeric ACE2 on the host cells indeed plays an important role during viral attachment, the binding affinity for the virus to latch on to the target cells would be much stronger than the values measured using monomeric ACE2 (ref. 46). This multivalency may help explain the unexpected transmission efficiency of SARS-CoV-2 leading to a pandemic on a surprising scale and raise the hurdle for antivirals to effectively compete with ACE2 for S binding. Trimeric ACE2 variants exerting even greater avidity than the dimeric form on the host cells may have a competitive edge over other RBD-targeting inhibitors, such as monoclonal antibodies, with similar binding affinity.
HEK293T cells were purchased from ATCC; Expi293F from Thermo Fisher Scientific. Mycoplasma contamination has been routinely tested for in cell cultures in our laboratory.
Protein expression and purification
A synthetic gene encoding a human ACE2 fragment (residues 1–615) fused with a C-terminal 6xHis tag was generated by GenScript and cloned into pCMV-IRES-puro expression vector (Codex BioSolutions) to create the construct pACE2615. To construct a trimeric ACE2 variant, a DNA fragment encoding a foldon trimerization tag was inserted between the ACE2 fragment and the His tag by restriction digestion and DNA ligation to give the plasmid pACE2615-foldon. Site-specific mutations, guided by high-resolution crystal structures (Extended Data Fig. 5), were introduced to the ACE2615-foldon construct by PCR following standard protocols of site-directed mutagenesis. All the ACE2 variants were expressed in HEK293F cells by transient transfection using Opti-MEM (Gibco-Thermo Fisher Scientific). After incubation for 4 d at 37 °C with 5.5% CO2, the transfected cells were collected by centrifugation at 2,524g for 30 min.
For the monomeric ACE2615 protein, the cell supernatant was collected by centrifugation and loaded onto a column packed with Ni-NTA agarose beads (Qiagen). The column was washed with a buffer containing 20 mM Tris-HCl, pH 7.5, and 300 mM NaCl. The protein was eluted using a buffer containing 100 mM imidazole, and further purified by gel filtration chromatography on a Superdex 200 Increase 10/300 GL column (GE Healthcare).
To purify the dimeric ACE2615-Fc and ACE2740-Fc proteins, the cell supernatant was collected and loaded onto a column packed with GammaBind Plus Sepharose beads (GE Healthcare). The column was washed with PBS. The protein was eluted using 100 mM glycine (pH 2.5) and neutralized immediately with 2 M Tris-HCl (pH 8.0). The eluted protein was further purified by gel filtration chromatography on a Superdex 200 Increase 10/300 GL column.
For all the ACE2615-foldon variants, which were not secreted efficiently, the cell pellet was resuspended in the lysis buffer (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 1% NP40, 20 mM imidazole) and rocked gently for 1 h at 4 °C, followed by spinning at 17,554g for 30 min to remove cell debris. The supernatant was loaded onto a column packed with Ni-NTA agarose beads (Qiagen). The column was washed with a buffer containing 20 mM Tris-HCl, pH 7.5, 300 mM NaCl and 50 mM imidazole and the protein was eluted using a buffer containing 20 mM Tris-HCl, pH 7.5, 300 mM NaCl and 300 mM imidazole. The eluted protein was further purified by gel filtration chromatography on a Superdex 200 Increase 10/300 GL column.
To produce a stabilized ectodomain of SARS-CoV-2 S trimer protein, a synthetic gene (kindly provided by Dr. Dan Barouch), encoding residues 1−1208 with the furin cleavage site (residues 682–685) replaced by a ‘GGSG’ sequence, residues K986 and V987 substituted by prolines and addition of a foldon trimerization tag followed by a C-terminal 6xHis tag, was cloned into the vector pCMV-IRES-puro. The expression construct was transiently transfected in HEK293T cells using polyethylenimine (Polysciences). Protein was purified from cell supernatants using Ni-NTA resin (Qiagen), and the eluted fractions containing S protein were pooled, concentrated and further purified by gel filtration chromatography on a Superose 6 column (GE Healthcare).
Cryo-EM sample preparation and data collection
To prepare cryo grids, 3.5 μl of the freshly prepared mixture of the soluble S trimer and monomeric ACE2 (1:3 molar ratio) at ~1 mg ml−1 was applied to a 1.2/1.3 Quantifoil grid (Quantifoil Micro Tools), which had been glow discharged with a PELCO easiGlow Glow Discharge Cleaning system (Ted Pella) for 60 s at 15 mA. Grids were immediately plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific), and excess protein was blotted away using grade-595 filter paper (Ted Pella) with a blotting time of 4 s and a blotting force of −12 at 4 °C in 100% humidity. The grids were first screened for ice thickness and particle distribution using a Talos Arctica transmission electron microscope (Thermo Fisher Scientific), operated at 200 keV and equipped with a K3 direct electron detector (Gatan), at the Harvard Cryo-EM Center for Structural Biology. For data collection, images were acquired with selected grids using a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operated at 300 keV with a BioQuantum GIF/K3 direct electron detector. Automated data collection was carried out using SerialEM version 65 at a nominal magnification of ×105,000 and the K3 detector in counting mode (calibrated pixel size, 0.825 Å) at an exposure rate of ~14.8 electrons per physical pixel per second. Each movie had a total accumulated electron exposure of 50 e− /Å2 fractionated in 50 frames of 50 ms. Datasets were acquired using a defocus range of 1.5–2.6 μm.
Image processing, 3D reconstructions and model building
Drift correction for cryo-EM images was performed using MotionCor2 (ref. 48), and contrast transfer function (CTF) was estimated by CTFFIND4 (ref. 49) using motion-corrected sums without dose-weighting. Motion-corrected sums with dose-weighting were used for all image processing. CrYOLO50 was used for particle picking, and RELION3.0.8 (ref. 51) was used for two-dimensional (2D) classification, 3D classification and refinement. A total of 407,761 particles were extracted from 4,292 images. The selected particles were subjected to 2D classification, giving a total of 261,799 good particles. A low-resolution negative-stain reconstruction of the sample was low-pass-filtered to 40 Å and used as an initial model for 3D classification in C1 symmetry. One class containing 32,685 particles that appeared to represent the free S trimer with no ACE bound was further refined in C1 symmetry, giving a reconstruction at 3.6-Å resolution. Another major class with ~49% of the selected particles showing density for ACE2 was refined in C1 symmetry and subsequently subjected to CTF refinement, Bayesian polishing and particle subtraction by masking out the ACE2–RBD density, followed by 3D classification without alignment in six classes. Whole particles were re-extracted based on the six classes from the masked local classification and refined further, revealing different stoichiometry for ACE2 binding (one ACE2 per S trimer, two ACE2s per S trimer, and three ACE2s per S trimer). The three best maps representing each type of complex were chosen and further refined in C1 symmetry after CTF refinement and Bayesian polishing, leading to one reconstruction of the complex with one ACE2 bound at 3.6-Å resolution from 15,964 particles; another reconstruction of the complex with two ACE2s bound at 3.7-Å resolution from 13,515 particles; and a third reconstruction of the complex with three ACE2s bound at 3.4-Å resolution from 26,298 particles. Reported resolutions are based on the gold-standard Fourier shell correlation (FSC) using the 0.143 criterion. All density maps were corrected from the modulation transfer function of the K3 detector and then sharpened by applying a temperature factor that was estimated using post-processing in RELION. Local resolution was determined using RELION with half-reconstructions as input maps.
The initial templates for model building used the stabilized SARS-CoV-2 S ectodomain trimer structure (PDB 6VYB) and ACE2 from the ACE2–B0AT1 complex structure (PDB 6M17). Several rounds of manual building were performed in Coot52. Iteratively, refinement was performed in both Phenix53 (real space refinement) and ISOLDE54, and the Phenix refinement strategy included rigid body fit, minimization_global, local_grid_search and adp, with rotamer, Ramachandran and reference-model restraints, using PDB 6VYB and PDB 6M17 as the reference model. The refinement statistics are summarized in Table 1. Structural biology applications used in this project were compiled and configured by SBGrid55.
Binding assay by BLI
Binding of ACE2 variants to the soluble S trimer was measured using an Octet RED384 system (ForteBio). Each ACE2 protein was diluted using the running buffer (PBS, 0.005% Tween 20, 0.25 mg ml−1 BSA) and transferred to a 96-well plate. The soluble S protein was immobilized to Amine Reactive 2nd Generation (AR2G) biosensors (ForteBio), following a protocol recommended by the manufacturer. After equilibrating in the running buffer for 5 min, the sensors with immobilized S protein were dipped in the wells containing the ACE2 protein at various concentrations (1.852–150 nM for ACE2615; 0.926–75 nM for ACE2m615-Fc and ACE2740-Fc; 0.617–50 nM for all the ACE2615-foldon variants) for 5 min to measure the association rate. The sensors were then dipped in the running buffer for 10 min to determine the dissociation rate. Control sensors with no S protein were also dipped in the ACE2 solutions and the running buffer as references. Recorded sensorgrams with background subtracted from the references were analyzed using the software Octet Data Analysis HT version 11.1 (ForteBio). The curves for monomeric ACE2 were fit to a 1:1 binding model, while those for the oligomeric ACE2 variants were fit to a bivalent binding model.
ACE2 peptidase activity assay
The catalytic activity of the ACE2 variants was measured by detecting a free fluorophore 7-methoxycoumarin-4-acetic acid (MCA) released from a synthetic peptide substrate, using an ACE2 activity kit (BioVision). The ACE2615 and ACE2615-foldon variants were diluted to 0.25 µg ml−1 using the assay buffer from the kit. The ACE2m615-Fc and ACE2740-Fc proteins were diluted to 0.38 and 0.30 µg ml−1, respectively, to keep the same number of the active sites as other ACE2 variants. Then, 50 µl of diluted protein was set in the 96-well plate. Immediately before recording fluorescence signals, 50 µl of substrate diluted in the assay buffer, following a protocol recommended by the manufacturer, was added to each well. Fluorescence signals were recorded in a kinetic mode by a Flexstation 3 Multi Mode Microplate Reader (Molecular Devices). The specific activity was calculated as the amount of the released fluorophore divided by the reaction time and the amount of the ACE2 protein using the data within the initial linear phase, as described in the protocol provided by the manufacturer. To determine the initial linear phase, fluorescence signals were recorded with 1.25, 0.25 and 0.125 µg of ACE2615 protein, respectively, reaching maximum after the substrates were completely cleaved. Data from the first 2 min within the linear phase with signals less than 10% of the maximum were used for the calculation. The amount of released MCA was derived from the increase of the fluorescence signal divided by the slope of the MCA standard curve.
Inhibition of Ang II-induced AT1R activation
To treat the Ang II peptide with each ACE2 variant, 2 μl of ACE2 protein at 0.5 mg ml−1 was added to 198 μl of an assay buffer (1× PBS, 40 mM Tris-HCl, pH 6.8, 20 μM ZnCl2) containing 65 μM Ang II peptide. The reactions were incubated at 37 °C for 40 min, and then quenched by addition of 50 μl of 0.5 M EDTA. The final concentration of Ang II peptide was 52 μM. As a time 0 control, 198 μl of the assay buffer containing 65 μM Ang II was incubated with EDTA at 37 °C first, followed by addition of 2 μl of each ACE2 protein (0.5 mg ml−1).
Changes in the intracellular calcium concentration in AT1R-expressing cells when induced by Ang II peptide were measured to monitor the activation of the receptor. Briefly, HEK293 cells were transfected with pCMV-AT1R-IRES-Puro gene using Lipofectamine 3000 reagent (Thermo Fisher Scientifics). At approximately 24 h post-transfection, the cells were transferred into a 384-well black-clear plate at a density of 1.2 × 104 cells per well in 20 μl of culture medium. On day 4, 20 μl of 1× Non-Wash Calcium Dye solution (CB-80500-301, Codex BioSolutions) was added into each well. The cell plate was incubated at 37 °C in a CO2 incubator for 1 h. The pretreated ligands (Ang II peptide) at various concentrations (0.005–500 nM) were prepared in 1× HBSS (Hank’s balanced salt solution) with 20 mM HEPES (pH 7.46). Fluorescence intensity in each well was recorded on an FDSS 7000 (Hamamatsu Corporation) at the rate of 1 image per second (excitation 480 nM and emission 540 nM) and the baseline of each well was also recorded for 10 s. After the online addition of 10 μl of the prepared ligands (final concentration of 0.001–100 nM), the fluorescence intensity of each well was recorded at the rate of 1 image per second for an additional 170 s.
MLV-based pseudovirus assay
MLV particles (all plasmids of the MLV components were kindly provided by Dr. Gary Whittaker at Cornell University and Drs. Catherine Chen and Wei Zheng at the National Center for Advancing Translational Sciences, National Institutes of Health (NIH)), pseudotyped with a SARS-CoV-2 S protein construct, were generated in HEK293T cells, following a protocol described previously for SARS-CoV56,57. To enhance incorporation, the C-terminal 19 residues in the cytoplasmic tail of the SARS-CoV-2 S protein containing D614 were deleted. To prepare for infection, 7.5 × 103 Expi293F cells, stably transfected with a full-length human ACE2 expression construct, in 15 μl of culture medium were plated into a 384-well white-clear plate coated with poly-d-Lysine to enhance cell attachment. On day 2, 12.5 μl of SARS-CoV-2 MLV pseudoviruses were mixed with 5 μl of each ACE2 variant at different concentrations (0.001–300 μg ml−1) and incubated at 37 °C for 1 h. After the medium in each well containing the cells was removed, 17.5 μl of each ACE2–virus mixture was added. The plate was centrifuged at 54g for 15 min at 4 °C and an additional 7.5 μl of culture medium was then added. The total final volume in each well was 25 μl. The cells were then incubated at 37 °C for 42 h. Luciferase activities were measured with Firefly Luciferase Assay Kit (CB-80552-010, Codex BioSolutions). IC50 values were calculated based on curve fitting in GraphPad Prism.
HIV-based pseudovirus assay
Neutralization of HIV-based pseudovirus containing a full-length SARS-CoV-2 S protein was measured using a single-round infection assay in HEK293T/ACE2 target cells. Pseudotyped virus particles were produced in 293T/17 cells (ATCC) by cotransfection of a plasmid encoding codon-optimized SARS-CoV-2 full-length S containing G614, a packaging plasmid pCMV ΔR8.2 expressing HIV gag and pol, and a luciferase reporter plasmid pHRʹ CMV-Luc. All plasmids were kindly provided by B. Graham (NIH, Vaccine Research Center). The 293T cell line stably overexpressing the human ACE2 protein was created by the Farzan group at Scripps Research Institute. For neutralization assays, serial dilutions of the ACE2 constructs were performed in duplicate followed by addition of pseudoviruses. Plates were incubated for 1 h at 37 °C followed by addition of 293T/ACE2 target cells (1 × 104 per well). Wells containing cells and pseudoviruses without ACE2 proteins or cells alone were positive and negative infection controls, respectively. Assays were collected on day 3 using BrightGlo luciferase reagent (Promega) and luminescence detected with a Victor luminometer (PerkinElmer). IC50 values are reported as the ACE2 protein concentration that inhibited 50% of virus infection. All neutralization experiments were repeated twice with similar results.
Neutralization of authentic SARS-CoV-2
ACE2 variants were serially diluted in Dulbecco’s PBS (Gibco) using half-log dilutions starting at 31,579 ng ml−1. Dilutions were prepared in triplicate for each protein. Each dilution was incubated at 37 °C in 5% CO2 for 1 h with 1,000 plaque forming units per ml (PFU ml−1) of SARS-CoV-2 (isolate USA‐WA1/2020). Controls included DMEM (Gibco-Thermo Fisher Scientific) containing 2% fetal bovine serum (Gibco-Thermo Fisher Scientific) and antibiotic-antimycotic (Gibco-Thermo Fisher Scientific) only as a negative control, 1,000 PFU ml−1 SARS-CoV-2 (USA-WA1/2020) incubated with Dulbecco’s PBS and 1,000 PFU ml−1 SARS-CoV-2 incubated with DMEM. For each dilution or control, 200 μl was added to confluent monolayers of NR-596 Vero E6 cells in triplicate and incubated for 1 h at 37 °C and 5% CO2. The plates were gently rocked every 5–10 min to prevent monolayer drying. The monolayers were then overlaid with a 1:1 mixture of 2.5% Avicel RC-591 microcrystalline cellulose and carboxymethylcellulose sodium (DuPont Nutrition & Biosciences) and 2× MEM (Temin’s modification, Gibco-Thermo Fisher Scientific), supplemented with 2× antibiotic‐antimycotic, 2× GlutaMAX (Gibco-Thermo Fisher Scientific) and 10% fetal bovine serum. Plates were incubated at 37 °C and 5% CO2 for 2 d. The monolayers were fixed with 10% neutral buffered formalin and stained with 0.2% aqueous gentian violet (RICCA Chemicals) in 10% neutral buffered formalin for 30 min, followed by rinsing and plaque counting. The IC50 values were calculated using GraphPad Prism 8.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
The atomic structure coordinates and EM maps are deposited in the wwPDB and EMDB, respectively, under the accession numbers PDB 7KJ2 and EMD-22891 (SARS-CoV-2 S in complex with one ACE2), PDB 7KJ3 and EMD-22892 (SARS-CoV-2 S in complex with two ACE2s), PDB 7KJ4 and EMD-22893 (SARS-CoV-2 S in complex with three ACE2s), and PDB 7KJ5 and EMD-22894 (SARS-CoV-2 S, no ACE2). Source data are provided with this paper.
Duan, K. et al. Effectiveness of convalescent plasma therapy in severe COVID-19 patients. Proc. Natl Acad. Sci. USA 117, 9490–9496 (2020).
Bloch, E. M. et al. Deployment of convalescent plasma for the prevention and treatment of COVID-19. J. Clin. Invest. 130, 2757–2765 (2020).
Shen, C. et al. Treatment of 5 critically ill patients with COVID-19 with convalescent plasma. JAMA 323, 1582–1589 (2020).
Wec, A. Z. et al. Broad neutralization of SARS-related viruses by human monoclonal antibodies. Science 369, 731–736 (2020).
Shi, R. et al. A human neutralizing antibody targets the receptor binding site of SARS-CoV-2. Nature 584, 120–124 (2020).
Chi, X. et al. A neutralizing human antibody binds to the N-terminal domain of the spike protein of SARS-CoV-2. Science 369, 650–655 (2020).
Wu, Y. et al. A noncompeting pair of human neutralizing antibodies block COVID-19 virus binding to its receptor ACE2. Science 368, 1274–1278 (2020).
Xia, S. et al. Inhibition of SARS-CoV-2 (previously 2019-nCoV) infection by a highly potent pan-coronavirus fusion inhibitor targeting its spike protein that harbors a high capacity to mediate membrane fusion. Cell Res. 30, 343–355 (2020).
Xia, S. et al. Fusion mechanism of 2019-nCoV and fusion inhibitors targeting HR1 domain in spike protein. Cell Mol. Immunol. 17, 765–767 (2020).
Sanders, J. M., Monogue, M. L., Jodlowski, T. Z. & Cutrell, J. B. Pharmacologic treatments for coronavirus disease 2019 (COVID-19): a review. JAMA 323, 1824–1836 (2020).
Wu, R. et al. An update on current therapeutic drugs treating COVID-19. Curr. Pharmacol. Rep. https://doi.org/10.1007/s40495-020-00216-7 (2020).
Boulware, D. R. et al. A randomized trial of hydroxychloroquine as postexposure prophylaxis for Covid-19. N. Engl. J. Med. 383, 517–525 (2020).
Jin, Z. et al. Structure of Mpro from SARS-CoV-2 and discovery of its inhibitors. Nature 582, 289–293 (2020).
Gao, Y. et al. Structure of the RNA-dependent RNA polymerase from COVID-19 virus. Science 368, 779–782 (2020).
Polak, S. B., Van Gool, I. C., Cohen, D., von der Thüsen, J. H. & van Paassen, J. A systematic review of pathological findings in COVID-19: a pathophysiological timeline and possible mechanisms of disease progression. Mod. Pathol. 33, 2128–2138 (2020).
Zhou, P. et al. A pneumonia outbreak associated with a new coronavirus of probable bat origin. Nature 579, 270–273 (2020).
Hoffmann, M. et al. SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor. Cell 181, 271–280 (2020).
Lan, J. et al. Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor. Nature 581, 215–220 (2020).
Shang, J. et al. Structural basis of receptor recognition by SARS-CoV-2. Nature 581, 221–224 (2020).
Yan, R. et al. Structural basis for the recognition of SARS-CoV-2 by full-length human ACE2. Science 367, 1444–1448 (2020).
Tikellis, C., Bernardi, S. & Burns, W. C. Angiotensin-converting enzyme 2 is a key modulator of the renin–angiotensin system in cardiovascular and renal disease. Curr. Opin. Nephrol. Hypertens. 20, 62–68 (2011).
Clarke, N. E. & Turner, A. J. Angiotensin-converting enzyme 2: the first decade. Int J. Hypertens. 2012, 307315 (2012).
Dasgupta, C. & Zhang, L. Angiotensin II receptors and drug discovery in cardiovascular disease. Drug Disco. Today 16, 22–34 (2011).
Hemnes, A. R. et al. A potential therapeutic role for angiotensin-converting enzyme 2 in human pulmonary arterial hypertension. Eur. Respir. J. 51, 1702638 (2018).
Imai, Y. et al. Angiotensin-converting enzyme 2 protects from severe acute lung failure. Nature 436, 112–116 (2005).
Gu, H. et al. Angiotensin-converting enzyme 2 inhibits lung injury induced by respiratory syncytial virus. Sci. Rep. 6, 19840 (2016).
Zou, Z. et al. Angiotensin-converting enzyme 2 protects from lethal avian influenza A H5N1 infections. Nat. Commun. 5, 3594 (2014).
Monteil, V. et al. Inhibition of SARS-CoV-2 infections in engineered human tissues using clinical-grade soluble human ACE2. Cell 181, 905–913.e7 (2020).
Kirchdoerfer, R. N. et al. Stabilized coronavirus spikes are resistant to conformational changes induced by receptor recognition or proteolysis. Sci. Rep. 8, 15701 (2018).
Cai, Y. et al. Distinct conformational states of SARS-CoV-2 spike protein. Science 369, 1586–1592 (2020).
Zhou, T. et al. Cryo-EM structures of SARS-CoV-2 spike without and with ACE2 reveal a pH-dependent switch to mediate endosomal positioning of receptor-binding domains. Cell Host Microbe 28, 867–879.e5 (2020).
Meier, S., Guthe, S., Kiefhaber, T. & Grzesiek, S. Foldon, the natural trimerization domain of T4 fibritin, dissociates into a monomeric A-state form containing a stable β-hairpin: atomic details of trimer dissociation and local β-hairpin stability from residual dipolar couplings. J. Mol. Biol. 344, 1051–1069 (2004).
Korber, B. et al. Tracking changes in SARS-CoV-2 spike: evidence that D614G increases infectivity of the COVID-19 virus. Cell 182, 812–827.e19 (2020).
Johnson, M. C. et al. Optimized pseudotyping conditions for the SARS-COV-2 spike glycoprotein. J. Virol. 94, https://doi.org/10.1128/JVI.01062-20 (2020).
Haschke, M. et al. Pharmacokinetics and pharmacodynamics of recombinant human angiotensin-converting enzyme 2 in healthy human subjects. Clin. Pharmacokinet. 52, 783–792 (2013).
Poglitsch, M. et al. Recombinant expression and characterization of human and murine ACE2: species-specific activation of the alternative renin–angiotensin-system. Int. J. Hypertens. 2012, 428950 (2012).
Chan, K. K. et al. Engineering human ACE2 to optimize binding to the spike protein of SARS coronavirus 2. Science 369, 1261–1265 (2020).
Guo, L. et al. Engineered trimeric ACE2 binds viral spike protein and locks it in “Three-up” conformation to potently inhibit SARS-CoV-2 infection. Cell Res. 31, 98–100 (2020).
Cao, L. et al. De novo design of picomolar SARS-CoV-2 miniprotein inhibitors. Science 370, 426–431 (2020).
Khan, A. et al. A pilot clinical trial of recombinant human angiotensin-converting enzyme 2 in acute respiratory distress syndrome. Crit. Care 21, 234 (2017).
Zoufaly, A. et al. Human recombinant soluble ACE2 in severe COVID-19. Lancet Respir. Med. https://doi.org/10.1016/S2213-2600(20)30418-5 (2020).
Barouch, D. H. et al. Evaluation of a mosaic HIV-1 vaccine in a multicentre, randomised, double-blind, placebo-controlled, phase 1/2a clinical trial (APPROACH) and in rhesus monkeys (NHP 13-19). Lancet 392, 232–243 (2018).
Mulligan, M. J. et al. Phase 1/2 study of COVID-19 RNA vaccine BNT162b1 in adults. Nature 586, 589–593 (2020).
Sharma, U. et al. Structural basis of homo- and heterotrimerization of collagen I. Nat. Commun. 8, 14671 (2017).
Jevsevar, S., Kunstelj, M. & Porekar, V. G. PEGylation of therapeutic proteins. Biotechnol. J. 5, 113–128 (2010).
Wrapp, D. et al. Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation. Science 367, 1260–1263 (2020).
Walls, A. C. et al. Structure, function and antigenicity of the SARS-CoV-2 spike glycoprotein. Cell 181, 281–292 (2020).
Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).
Rohou, A. & Grigorieff, N. CTFFIND4: fast and accurate defocus estimation from electron micrographs. J. Struct. Biol. 192, 216–221 (2015).
Wagner, T. et al. SPHIRE-crYOLO is a fast and accurate fully automated particle picker for cryo-EM. Commun. Biol. 2, 218 (2019).
Scheres, S. H. RELION: implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 180, 519–530 (2012).
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010).
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010).
Croll, T. I. ISOLDE: a physically realistic environment for model building into low-resolution electron-density maps. Acta Crystallogr. D Struct. Biol. 74, 519–530 (2018).
Morin, A. et al. Collaboration gets the most out of software. Elife 2, e01456 (2013).
Millet, J. K. & Whittaker, G. R. Murine Leukemia Virus (MLV)-based coronavirus spike-pseudotyped particle production and infection. Bio Protoc. 6, e2035 (2016).
Chen, C. Z. et al. Identifying SARS-CoV-2 entry inhibitors through drug repurposing screens of SARS-S and MERS-S pseudotyped particles. ACS Pharmacol. Transl. Sci. 3, 1165–1175 (2020).
We thank G. Frey for generous advice; S. Sterling, R. Walsh Jr. and S. Rawson for technical support; K. Arnett for advice on the BLI experiments; and S. Harrison for critical reading of the manuscript. EM data were collected at the Harvard Cryo-EM Center for Structural Biology of Harvard Medical School. This work was supported by NIH grant nos. AI147884 (to B.C.), AI147884-01A1S1 (to B.C), AI141002 (to B.C.) and AI127193 (to B.C. and J. Chou); a COVID19 Award by Massachusetts Consortium on Pathogen Readiness (MassCPR; to B.C.); as well as a Fast grant by Emergent Ventures (to B.C.).
Boston Children’s Hospital has filed a patent application based on this work with B.C., T.X., J.Z. and Y.C. listed as coinventors. All other authors declare no competing interests.
Peer review information Nature Structural & Molecular Biology thanks Brandon DeKosky and Xinquan Wang for their contribution to the peer review of this work. Inês Chen was the primary editor on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended Data Fig. 1 Preparation of a stabilized soluble SARS-CoV-2 S trimer.
a, Schematic representation of the expression construct of the soluble SARS-CoV-2 S protein. Segments of S1 and S2 ectodomain (S2ecto) include: NTD, N-terminal domain; RBD, receptor-binding domain; S1/S2, S1/S2 cleavage site; S2’, S2’ cleavage site; FP, fusion peptide; HR1, heptad repeat 1; CH, central helix region; HR2, heptad repeat 2; and tree-like symbols for glycans. The S1/S2 cleavage site (RRAR) was mutated to GGSG. Two mutations K986P and V987P were introduced and a trimerization tag –foldon fused to the C-terminal end to stabilize the prefusion conformation29. A C-terminal histag was included for protein purification. b, The purified S protein was resolved by gel-filtration chromatography on a Superose 6 column and the pooled peak fractions were analyzed by Coomassie stained SDS-PAGE. The molecular weight standards include thyoglobulin (670 kDa), γ-globulin (158 kDa) and ovalbumin (44 kDa). The uncropped image for panel b is available as source data.
Extended Data Fig. 2 Cryo-EM analysis of the ACE2-S complexes.
Top, representative micrograph, and 2D averages of the cryo-EM particle images showing secondary structural features. Bottom, data processing workflow for structures of the free S trimer (no ACE2), S trimer with one ACE2 bound (1 ACE2), S trimer with two ACE2 bound (2 ACE2), S trimer with three ACE2 bound (3 ACE2), as indicated.
Extended Data Fig. 3 Analysis of the 3D reconstructions of the S-ACE2 complexes.
a, 3D reconstructions of the S trimer and its ACE2 complexes are colored according to local resolution estimated by RELION. Angular distribution of the cryo-EM particles used in the reconstruction is shown in the side view of the EM map. b, Gold standard FSC curves of the refined 3D reconstructions. c, Representative density in gray surface from the EM maps. No ACE2), the free S trimer; 1 ACE2, S trimer with one ACE2 bound; 2 ACE2, S trimer with two ACE2 bound; 3 ACE2, S trimer with three ACE2 bound.
Extended Data Fig. 4 ACE2-S interactions.
a, The top view of the S trimer in complex with two monomeric ACE2 molecules (in magenta and cyan, respectively) with the distance between the C-terminal ends of the ACE2s indicated. b, The top view of the S trimer in complex with three monomeric ACE2 molecules (in magenta, blue and cyan, respectively) with the distances between the C-terminal ends of the ACE2s indicated. c, Superposition of the structure of the full-length ACE2 (PDB 6M17) and the structure of S-3 ACE2 complex.
Extended Data Fig. 5 Design of mutations at the ACE2-RBD interface.
Top, the interface between ACE2 in ribbon diagram in magenta and RBD in stick model from the complex structure (PDB 6M0J) with T27, H34 and K353 from ACE2 indicated. Bottom, modeled T27W, H34W and K353W mutations that may enhance the hydrophobic interactions between ACE2 and RBD.
Extended Data Fig. 6 Purification of ACE2 variants.
The purified ACE2 proteins were resolved by gel-filtration chromatography on a Superdex 200 column. The molecular weight standards include thyoglobulin (670 kDa), γ-globulin (158 kDa) and ovalbumin (44 kDa). Inset, peak fractions were analyzed by Coomassie stained SDS-PAGE. The uncropped image is available as source data.
Extended Data Fig. 7 Comparison of secreted and cellular ACE2615-foldon protein.
a, The purified ACE2615-foldon protein either from cell supernatants (secrected) or cell lysates (cellular) was resolved by gel-filtration chromatography on a Superdex 200 column. Inset, peak fractions were analyzed by Coomassie stained SDS-PAGE. The uncropped image for panel a is available as source data. b, Binding of ACE2615-foldon to the stabilized soluble S trimer by bio-layer interferometry (BLI). The S protein was immobilized and subsequently dipped into the wells containing either secreted or cellular ACE2615-foldon at various concentrations (0.617–50 nM). The sensorgrams for the secreted ACE2615-foldon are in blue and those for the cellular ACE2615-foldon in red.
Extended Data Fig. 8 Binding of ACE2 variants to the stabilized soluble S trimer by bio-layer interferometry (BLI).
The S protein was immobilized and subsequently dipped into the wells containing ACE2 proteins at various concentrations (0.926–75 nM for ACE2615-Fc, 0.617–50 nM for the ACE2615-foldon variants). Binding kinetics was evaluated using a bivalent model for all oligomeric ACE2s. The sensorgrams are in black and the fits in red. All the experiments were repeated at least twice with essentially identical results. Binding constants derived from the BLI experiments are summarized in Supplementary Table 3.
Supplementary Tables 1–3.
Source Data Fig. 3
Source data for Fig. 3b.
Source Data Fig. 4
Source data for Fig. 4a–c.
Source Data Extended Data Fig. 1
Unmodified gel for Extended Data Fig. 1.
Source Data Extended Data Fig. 6
Unmodified gel for Extended Data Fig. 6.
Source Data Extended Data Fig. 7
Unmodified gel for Extended Data Fig. 7.
Rights and permissions
About this article
Cite this article
Xiao, T., Lu, J., Zhang, J. et al. A trimeric human angiotensin-converting enzyme 2 as an anti-SARS-CoV-2 agent. Nat Struct Mol Biol 28, 202–209 (2021). https://doi.org/10.1038/s41594-020-00549-3
This article is cited by
COVID-19 therapies: do we see substantial progress?
Cellular & Molecular Biology Letters (2022)
Structure genomics of SARS-CoV-2 and its Omicron variant: drug design templates for COVID-19
Acta Pharmacologica Sinica (2022)
Enhanced trimeric ACE2 exhibits potent prophylactic and therapeutic efficacy against the SARS-CoV-2 Delta and Omicron variants in vivo
Cell Research (2022)
Bioengineered angiotensin-converting-enzyme-2: a potential therapeutic option against SARS-CoV-2 infection
Journal of Human Hypertension (2022)
The SARS-CoV-2 spike reversibly samples an open-trimer conformation exposing novel epitopes
Nature Structural & Molecular Biology (2022)