A bispecific monomeric nanobody induces spike trimer dimers and neutralizes SARS-CoV-2 in vivo

Antibodies binding to the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike have therapeutic promise, but emerging variants show the potential for virus escape. This emphasizes the need for therapeutic molecules with distinct and novel neutralization mechanisms. Here we isolated a nanobody that interacts simultaneously with two RBDs from different spike trimers of SARS-CoV-2, rapidly inducing the formation of spike trimer-dimers leading to the loss of their ability to attach to the host cell receptor, ACE2. We show that this nanobody potently neutralizes SARS-CoV-2, including the B.1.351 variant, and cross-neutralizes SARS-CoV. Furthermore, we demonstrate the therapeutic potential of the nanobody against SARS-CoV-2 and the B.1.351 variant in a human ACE2 transgenic mouse model. This naturally elicited bispecific monomeric nanobody establishes a novel strategy for potent inactivation of viral antigens and represents a promising antiviral against emerging SARS-CoV-2 variants.


Introduction
Fu2 blocks the RBD from binding ACE2. 157 Structural alignment of an ACE2-RBD structure (PDB: 6LZG 18 ) to the Fu2-spike structure 158 (single RBD) shows that binding of ACE2 would be hindered by Fu2 bound to any of the two 159 interfaces (Fig. 2F). Fu2 bound to the interface-major would clash with ACE2 residues ~303-331

167
In the complex, Fu2's long CDR3 (G99 to Y120) plays a significant role in RBD binding. Here, 168 the Fu2 S107-R111 region adopts a b-strand conformation that binds the RBD T376-Y380 b-169 strand to give rise to an energetically favorable inter-chain extension of the RBD anti-parallel b-170 sheet (Figs. 2G and 2H). Interestingly, the Fu2 CDR1 (G26-Y32) and CDR2 (T52-S57) do not 171 participate in RBD binding, and none of the CDRs is directly involved in the formation of the 172 spike dimer.

Fu2 stabilizes the RBD in the 'up' conformation.
spike is mostly present in all-down (~60%) or 1-up (~40%) conformation 6 . Our Fu2-spike 177 structure shows the presence of two trimeric spike molecules, both in the 3-up conformation.  To test the breadth of neutralization, we evaluated cross-neutralization of SARS-CoV PSV (Fig. 189 3A). While Ty1 did not neutralize SARS-CoV, Fu2 and dimeric Fu2 constructs were 190 neutralizing, but less so than against SARS-CoV-2 PSV. Interestingly, the neutralization curves 191 displayed a much flatter slope as compared to those against SARS-CoV-2 PSV. It required 6.1 192 µg /ml of monomeric Fu2 to reach 50% neutralization in this assay, while 570 ng /ml of  and 57 ng/ml of the Fu2-dimer was sufficient for 50% neutralization.

194
To understand the molecular mechanism of cross-neutralization of SARS-CoV we 195 compared the RBD sequences and performed structural alignments (Figs. 3B, 3C and S4A). differences, suggesting that the divergent interface-minor of SARS-CoV RBD would not permit 198 Fu2 binding. From these analyses we conclude that Fu2 would be unlikely to induce SARS-CoV 199 trimer dimers.

200
To demonstrate neutralization of replication competent SARS-CoV-2 we 201 performed a plaque reduction neutralization test (PRNT) using infectious SARS-CoV-2, isolated 202 from the first Swedish COVID-19 patient 19 (Fig. 3D). Similar to neutralization of PSV, 203 monomeric Fu2 neutralized SARS-CoV-2 much better than Ty1, a dimeric construct of Fu2 204 provided a 2.8-fold improvement over the monomeric version of Fu2, while the heterodimer of 205 Ty1 and Fu2 showed a 50-fold improvement compared to monomeric Fu2 in this assay.  An important requirement for the development of antivirals is neutralization potential in vivo. To 220 address this, we used K18-hACE2 transgenic mice that express human ACE2 as a model for SARS-CoV-2 infection. To determine if Fu2 could reduce disease severity when administered in 222 early infection, we tested its ability to neutralize the virus and protect mice from signs of disease 223 in vivo (Fig. 4). When the Fu2-Ty1 heterodimer was administered prophylactically or 224 therapeutically, we noted that it slightly delayed the onset of disease (Fig. S7). A major issue 225 with using nanobodies therapeutically is their short half-life in vivo. To extend serum half-life, 226 we fused Fu2 to the nanobody Alb1 that binds mouse serum albumin 20 . Mice were challenged   250 Fu2 is also capable of neutralizing SARS-CoV, highlighting its specificity for a 251 more conserved epitope. However, the potency was significantly reduced for the monomeric  The RBD-specific nanobody Fu2 displays a particularly unique mode of virus 264 neutralization by rapidly inducing stable spike trimer dimers that are incapable of binding to 265 ACE2. In contrast to spike trimer dimers reported for the full-length spike that are based on S1-266 S1 interactions placing the trimers 'side-by-side' 30 , Fu2 induced a dimer in a 'head-to-head' 267 orientation with the C-termini at the distal ends. Targeting two epitopes increases the binding 268 surface area, but, more importantly, when locked in such a conformation the function of each spike is sterically blocked both by the nanobody and by the other spike in such a way that the 270 whole complex must dissociate for the spike to regain its function.

271
The interaction between Fu2 and the RBD is highly unusual, and for a monomer to 272 induce antigen dimerization in this manner, two copies of an antigen must be bound at different 273 epitopes with precisely the right spatial arrangement. This novel mode of antigen binding may be 274 uncommon in naturally elicited antibody repertoires but could be achievable by structure-based      (Table S1). reconstructions. In total 277,372 particles were retained for refinement steps and these particles 520 were processed with C1 symmetry (image processing scheme Fig. S5).

521
These particles were further processed using heterogeneous refinement that resulted in a 522 reconstruction with high-resolution structural features in the core of the spike. One round of 523 homogeneous refinement was followed by non-uniform refinement. All final reconstructions were analysed using 3D-FSC 37 and there were moderate anisotropy in the full map 525 reconstructions while the localized reconstruction displayed no significant anisotropy (Fig. S6A   526 and S6B). All CTF refinements were per-particle CTF refinements interspersed with global 527 aberration correction (beamtilt, trefoil, tetrafoil and spherical aberration). Please see Table S1 for 528 data collection and processing statistics and the respective cryo-EM data processing schemes.

529
For the Spike-Fu2 dimer interface we used a particle set with partial-signal subtraction of all 530 parts except for Fu2-RBD dimer interface (containing two RBDs and two Fu2). From this we 531 performed local reconstruction (non-uniform). The local reconstruction resulted in a map with 532 2.89Å overall resolution as compared to 3.2Å overall resolution for the full map.  Table S2 for refinement and validation were monitored daily until weight drop started, whereupon mice were monitored twice daily.

554
During the experiment, weight loss, changes in general health, breathing, body movement and 555 posture, piloerection and eye health were monitored. Mice were typically sacrificed when they 556 achieved 20% weight loss. However, some mice were sacrificed before losing 20% in body 557 weight, when movement was greatly impaired and/or they experienced difficulty breathing that 558 was considered to reach a severity level of 0.5 on Karolinska Institutet's veterinary plan for 559 monitoring animal health. The weight loss in response to infection was highly reproducible. In