Neutralization of SARS‐CoV‐2 by highly potent, hyperthermostable, and mutation‐tolerant nanobodies

Abstract Monoclonal anti‐SARS‐CoV‐2 immunoglobulins represent a treatment option for COVID‐19. However, their production in mammalian cells is not scalable to meet the global demand. Single‐domain (VHH) antibodies (also called nanobodies) provide an alternative suitable for microbial production. Using alpaca immune libraries against the receptor‐binding domain (RBD) of the SARS‐CoV‐2 Spike protein, we isolated 45 infection‐blocking VHH antibodies. These include nanobodies that can withstand 95°C. The most effective VHH antibody neutralizes SARS‐CoV‐2 at 17–50 pM concentration (0.2–0.7 µg per liter), binds the open and closed states of the Spike, and shows a tight RBD interaction in the X‐ray and cryo‐EM structures. The best VHH trimers neutralize even at 40 ng per liter. We constructed nanobody tandems and identified nanobody monomers that tolerate the K417N/T, E484K, N501Y, and L452R immune‐escape mutations found in the Alpha, Beta, Gamma, Epsilon, Iota, and Delta/Kappa lineages. We also demonstrate neutralization of the Beta strain at low‐picomolar VHH concentrations. We further discovered VHH antibodies that enforce native folding of the RBD in the E. coli cytosol, where its folding normally fails. Such “fold‐promoting” nanobodies may allow for simplified production of vaccines and their adaptation to viral escape‐mutations.

The SARS-CoV-2 receptor-binding domain (RBD, residues 330-527) was expressed in the cytosol of E. coli SHuffle® Express as an N-terminal fusion with a His 14 -SUMO tag. Cells were lysed by sonication, the soluble lysate was obtained by ultracentrifugation and applied to a Ni-chelate column. Non-bound material was washed off with lysis buffer, and the fusion protein was eluted with buffer containing imidazole. The concentrated eluate was subjected to size exclusion chromatography (Superdex 200). The A 280 elution profile is shown in blue; the indicated fractions were analyzed by SDS-PAGE and Coomassie staining. Note that the RBD is trapped by GroEL in the IMAC eluate and that the majority of the RBD elutes within the void volume of the column. (A) The SARS-CoV-2 RBD was expressed in the cytosol of E. coli SHuffle® Express (by induction with IPTG; "-/+ IPTG") as an Re9F06-RBD fusion with an N-terminal His 14 -SUMO tag. Cells were lysed by sonication, the soluble lysate was obtained by ultracentrifugation and applied to a Ni-chelate column. Non-bound material was washed off with lysis buffer, and the Re9F06-RBD fusion was eluted by cleaving off the His 14 -SUMO tag with the Ulp1p protease. The indicated fractions were analyzed by SDS-PAGE and Coomassie staining. Note the strong Re9F06-RBD fusion band in the eluted fraction. The yield was approximately 10 mg Re9F06-RBD fusion per liter of culture.
(B) Binding of the Re9F06-RBD fusion to Ni-chelate beads as in panel A, but in a small (analytical) scale. Where indicated, 1 µM untagged Re5D06 was added to the binding reaction; the beads were washed with lysis buffer containing 1 M NaCl, and elution was done as in A (with the Ulp1p protease). Co-elution of Re5D06 (marked with a black asterisk) with the Re9F06-RBD fusion indicates a specific RBD·Re5D06 interaction. The endogenous E. coli proteins present in the soluble starting lysate can be regarded as negative controls for the binding reaction. Purification and crystallization of the Re9F06·RBD·Re5D06 complex.
(A) The Re9F06·RBD·Re5D06 complex was expressed in E. coli and purified as described in "Materials and Methods". The last affinity chromatography step entailed protease (Ulp1p)-mediated elution of the complex from an IMAC column, and further purification was performed by gel filtration on a Superdex 75 column. The elution profile (A 280 ) was recorded (upper panel), followed by analysis of the input material and eluate fractions by SDS-PAGE and Coomassie staining (lower panel). Fractions that were pooled and further concentrated for crystallization screening are indicated.
(B) 4 mg of the Re9F06·RBD·Re5D06 complex were treated for 1 hour (at 23ºC) with the indicated proteases (that vary broadly in their specificities) at complex:protease ratios of 1:2500; 1:500; 1:100 and 1:20 (w/w), respectively. Samples were mixed with SDS sample buffer, heated to 95ºC for 5 min, and 3 mg of each sample were analyzed by SDS-PAGE and Coomassie staining. "Input" marks the starting material; protease bands are labeled with an asterisk. Note that the complex is remarkably resistant to all but one of the tested proteases tested (Thermolysin), even at very high protease concentrations.
(C) Left: the Re9F06·RBD·Re5D06 complex was subjected to proteolysis as described in B. Samples were analyzed by SDS-PAGE, Coomassie staining and mass spectrometry (for "RBD band 1"/ "RBD band 2", after digestion with trypsin). Right: the N-terminal sequence of undigested RBD is shown, with the candidate-P1' sites of Thermolysin highlighted in red. Peptides identified in "RBD band 1" and "RBD band 2", respectively, are indicated.
(D) The purified Re9F06·RBD·Re5D06 complex was spiked with Thermolysin (at a 1:500 w/w ratio) and subjected to crystal screening at 20ºC. The micrograph shows the crystals obtained in a (scaled-up) 2.5-ml hanging drop of the first hit condition. See "Materials and Methods" for details.  (A) Overview of the Re9F06·RBD complex (PDB ID 7OLZ; Re5D06 was omitted for clarity), with the RBD shown as a green ribbon overlayed with its semitransparent surface. Re9F06 is depicted as a ribbon (blue) with the CDR regions highlighted in orange.
(B) Details of the Re9F06·RBD interaction interface, either depicting all interactions or interactions sorted by contact type. RBD and Re9F06 are shown as transparent ribbons, colored as in A, with interface side chains depicted in green (RBD) or blue (Re9F06). Blue marks nitrogen, oxygen is shown in red, a water molecule is depicted as a yellow sphere. Dashed lines link interacting atoms (distance ≤ 4 Å). Lines pointing onto backbones indicate contacts to carbonyl-carbons or amide groups. Note that the Re9F06·RBD interaction does not show the same degree of shape complementarity as the RBD·Re5D06 complex ( Fig 6E, S6). (A) Overview of the RBD·Re5D06 complex (PDB ID 7OLZ; Re9F06 was omitted for clarity), with the RBD shown as a green ribbon overlayed with its semitransparent surface. Re5D06 is depicted as a ribbon (magenta) with the CDR regions highlighted in orange, as in Fig 6E. (B) Details of the RBD·Re5D06 interaction interface, either depicting all interactions (equivalent to Fig 6E, with a few additional contacts) or interactions sorted by contact type. RBD and Re5D06 are shown as transparent ribbons colored as in A, with interface side chains depicted in green (RBD) or magenta (Re5D06). Blue marks nitrogen, oxygen is shown in red, a water molecule is depicted as a yellow sphere. Dashed lines link interacting atoms (distance ≤ 4 Å). Lines pointing onto backbones indicate contacts to carbonyl-carbons or amide groups. Note the intramolecular cation-π interactions of Re5D06. These contacts form an array critical for RBD interaction. RBD F486 ("hydrophobic interactions") is also poised to form a cation-π contact with Re5D06 R50.  Fig 6C) Compatibility of VHH Re5D06 with the "open" and "closed" Spike. Left: surface representation of the SARS-CoV-2 Spike in the "all-RBDs-up" (PDB ID 7A98; Benton et al., 2020) or "all-RBDs-down" conformation (PDB ID 6VXX; Walls et al., 2020). The RBD is shown in green, with the ACE2-binding surface depicted in brown.   Fig 6C) Sample and grid preparation for cryo-EM studies.

Right
(A) The purified pre-fusion-stabilized SARS-CoV-2 Spike ectodomain trimer ("HexaPro") was mixed with a 9-fold molar excess of VHH Re5D06 and passed over a Superose 6 column. The A 280 elution profile is shown (left panel), with the peak fraction of the HexaPro·Re5D06 complex highlighted in gray. This fraction was analyzed by SDS-PAGE and Coomassie staining (right panel).
(B) The peak eluate fraction was vitrified immediately after elution on Quantifoil UltrAuFoil R2/2 (200 mesh) grids. Grids were screened on a Glacios 200 kV TEM equipped with a Falcon III detector. The micrograph shows the atlas overview of the grid that was later used for data collection on a Titan Krios G2 microscope. Note the gradient of ice thickness across the grid.
(C) 15636 movies were collected on a Titan Krios G2 microscope equipped with an energy filter and a K3 detector. We aquired data from holes of a broader range of ice thickness, as illustrated by the shown micrograph: while particle side views dominated in thin ice (green circles), slightly thicker ice also accommodated Spike particles in the top/bottom view orientation.
(D) All selected particles were subjected to 2D classification in cryoSPARC. The images show representative 2D class averages. Note that the class averages include side views, edge-on views as well as top/bottom views at high resolution. Particle number, resolution and effective sample size ("ess", with a value of 1 indicating high certainty of the alignment) are shown for each class as reported by cryoSPARC.

Gradient of ice thickness Grid atlas
Spike top/bottom views Spike side views  Figure S9 (previous page; related to Fig 6C) Image processing workflow for cryo-EM maps 1-3.
The sequence of 3D classification and 3D refinement steps leading to maps 1-3 is shown. All high-resolution image processing was done in RELION (version 3.1.1). Volumes of 3D classes as well as unsharpened maps obtained in intermediate refinements are shown, along with the achieved resolution. The percentage of input particles left in each class is indicated. Classes and 3D refinements highlighted in salmon were used for downstream processing (with absolute particle numbers shown in green); 3D classes shown in gray were discarded. The 3D masks used for focused (masked) 3D classification or masked 3D refinement runs are shown in blue (semi-transparent surface). The final (post-processed) maps are shown in gold.
(A) Classification and refinement tree for the complete Spike·Re5D06 volume.
(B) Classification and refinement tree for the RBD·Re5D06 sub-volume. Map 3 was obtained by masking both the RBD·Re5D06 volume (body 02 mask) and the remaining volume (body 01 mask) and refining them separately as rigid bodies with partial signal subtraction and centering (as implemented in RELION's multi-body refinement; Nakane et al., 2018). Local Resolution: The locally filtered map was color-coded in relion_postprocess (RELION, version 3.1.1) according to the resolution shown in the scale bar. Even though the local resolution in the periphery of the complex (e.g. the RBD·Re5D06 modules) is lower in this "global" refinement, high-resolution reconstructions could be obtained for these sub-volumes through focused refinements with partial signal subtraction (see Appendix Fig S11 for an example). Angular distribution: The Euler angle distribution for the particles in the 3D reconstruction is shown for two views (top and side view of the Spike). The height of the bars and the coloring from blue to red represents the number of particles at the given orientation.

Higher-resolution reconstruction of the RBD·Re5D06 sub-volume (map 3/EMD-13105).
The RBD·Re5D06 module of the closed Spike protomer (see map 2) displayed the lowest relative flexibility and was thus chosen for high-resolution multi-body refinement in RELION (see Appendix Fig S9 for details).
(A) The crystal structure of the RBD·Re5D06 complex was manually docked into the sharpened cryo-EM density (shown as a gray mesh) and rigid body-refined in Phenix. The model is displayed as in Fig 6E. Note that the cryo-EM map agrees well with the crystal structure. The N-glycosylation at RBD Asn 343 is shown as yellow sticks.
(B) Detail views of the RBD·Re5D06 interface (left) and the RBD core (right) generated in Coot.
(C) Fourier shell correlation, local resolution and angular distribution plots for the RBD·Re5D06 sub-volume (map 3). See Appendix Fig S9 and S10A for details. Fig 7) Engineering of hyperthermostable Re5D06 (R) variants.

Figure S12 (related to
(A) VHH Re5D06 prior to engineering for thermostability. The panel shows a surface representation of Re5D06 (magenta) as in Fig 7D to illustrate a large cavity in the hydrophobic core of the VHH antibody. The electron density map revealed that a molecule of the crystallization additive NDSB-256 (a zwitterionic nondetergent sulfobetaine; Dimethylbenzylammonium Propane Sulfonate) filled this void with its hydrophobic benzyl moiety in the crystal structure of the Re9F06·RBD·Re5D06 complex. Two distinct orientations of the molecule's aliphatic part could be modeled and refined. (B) Point mutations introduced into Re5D06. The table lists the changes for the two Re5D06 variants discussed in this study (R15; R28), sorted by category. The Cα carbons of these altered residues are shown as spheres (color-coded as in the table) on the ribbon representation of Re5D06. The orientation is similar to that of Re5D06 in A.
All mutations indicated in red and blue stabilize the nanobody through intramolecular interactions. S53N and S54N (yellow) were found in other members of the Re5D06 sequence class and engage in alternative, possibly more stable intramolecular hydrogen bonding contacts. V93D (yellow) primarily serves to further minimize aggregation of Re5D06, but also improves surface packing. Y109H (yellow, a CDR3 mutation) was introduced to establish charge complementarity to RBD E484 while preserving the hydrophobic stacking with RBD F490 .  The cells were imaged by confocal laser scanning microscopy (CLSM). Note that Re6D06 failed to stain B.1.351. Contrary to that, Re7E02 and Re9C07 stained B.1.351 more brightly than wild-type SARS-CoV-2, and Re9H03 was unaffected. This suggests that sets of mutation-sensitive and mutation-tolerant nanobodies can be used to diagnose SARS-CoV-2 variants by straightforward immunostaining or ELISA.
(B) BLI experiment to assess binding of Re6D06 (at 20 nM concentration) to the wild-type RBD or the Beta/"South African" B.1.351 RBD variants. Consistent with A, Re6D06 rapidly associated with the wild-type RBD, binding it tightly, but it failed to bind the Beta/"South African" RBD mutant.
See Appendix Table S4 for information on naming SARS-CoV-2 variants. Statistics for the highest-resolution shell are shown in parentheses.  58°C yes 5000 n.d.