bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022.

The copyright holder for this

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

1 Combating the SARS-CoV-2 Omicron variant with non-Omicron

2 neutralizing antibodies
3 Yingdan Wang1#, Xiang Zhang1#, Jiangyan Liu1#, Yanqun Wang2#, Wuqiang Zhan1, Mei Liu1,

4 Meng Zhang1, Qimin Wang1, Qianying Liu1, Tongyu Zhu1, Yumei Wen1, Zhenguo Chen1*,

5 Jincun Zhao2,3*, Fan Wu1*, Lei Sun1*, Jinghe Huang1*

7 1
Key Laboratory of Medical Molecular Virology (MOE/NHC/CAMS) and Shanghai Institute

8 of Infectious Disease and Biosecurity, the Fifth People's Hospital of Shanghai, Shanghai

9 Public Health Clinical Center, Institutes of Biomedical Sciences, School of Basic Medical

10 Sciences, Fudan University, Shanghai 200032, China.

2State
11 Key Laboratory of Respiratory Disease, National Clinical Research Center for

12 Respiratory Disease, Guangzhou Institute of Respiratory Health, the First Affiliated

13 Hospital of Guangzhou Medical University, Guangzhou, Guangdong 510182, China;

14 3Institute of Infectious Disease, Guangzhou Eighth People's Hospital of Guangzhou

15 Medical University, Guangzhou, Guangdong 510060, China.

16
#These
17 authors contributed equally.
18 *Correspondence to: Jinghehuang@fudan.edu.cn, LLSun@fudan.edu.cn,
19 wufan@fudan.edu.cn, zhaojincun@gird.cn or ZhenguoChen@fudan.edu.cn.
20
21 Keywords: COVID-19; SARS-CoV-2; Neutralizing Ab; Omicron Variant
22
23
bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

24 Abstract

25 The highly mutated and transmissible Omicron variant has provoked serious
26 concerns over its decreased sensitivity to the current coronavirus disease 2019
27 (COVID-19) vaccines and evasion from most anti-severe acute respiratory
28 syndrome coronavirus 2 (SARS-CoV-2) neutralizing antibodies (NAbs). In this
29 study, we explored the possibility of combatting the Omicron variant by
30 constructing bispecific antibodies based on non-Omicron NAbs. We engineered
31 ten IgG-like bispecific antibodies with non-Omicron NAbs named GW01, 16L9,
32 4L12, and REGN10987 by fusing the single-chain variable fragments (scFvs)
33 of two antibodies through a linker and then connecting them to the Fc region of
34 IgG1. Surprisingly, eight out of ten bispecific antibodies showed high binding
35 affinity to the Omicron receptor-binding domain (RBD) and exhibited extreme
36 breadth and potency against pseudotyped SARS-CoV-2 variants of concern
37 (VOCs) including Omicron, as well as authentic Omicron(+R346K) variants. Six
38 bispecific antibodies containing the cross-NAb GW01 neutralized Omicron
39 variant and retained their abilities to neutralize other sarbecoviruses. Bispecific
40 antibodies inhibited Omicron infection by binding to the ACE2 binding site. A
41 cryo-electron microscopy (cryo-EM) structure study of the representative
42 bispecific antibody FD01 in complex with the Omicron spike (S) revealed 5
43 distinct trimers and one unique bi-trimer conformation. The structure and
44 mapping analyses of 34 Omicron S variant single mutants elucidated that two
45 scFvs of the bispecific antibody synergistically induced the RBD-down
46 conformation into 3-RBD-up conformation, enlarged the interface area,
47 accommodated the S371L mutation, improved the affinity between a single IgG
48 and the Omicron RBD, and hindered ACE2 binding by forming bi-trimer
49 conformation. Our study offers an important foundation for anti-Omicron NAb
50 design. Engineering bispecific antibodies based on non-Omicron NAbs may
51 provide an efficient solution to combat the Omicron variant.

52
bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

53 INTRODUCTION

54 Omicron, first identified in South Africa and reported to the WHO at the end of
55 Nov. 2021, became the dominant severe acute respiratory syndrome
56 coronavirus 2 (SARS-CoV-2) variant globally in Jan. 2022. Omicron is the most
57 severely altered version of SARS-CoV-2, with more than 30 mutations in the
58 spike protein, sixteen of which are in the receptor-binding domain (RBD).
59 Omicron variants extensively escape neutralization by sera from vaccinated or
60 convalescent individuals 1-10. Moreover, most neutralizing antibodies (NAbs),
61 including many clinical-stage monoclonal antibodies (mAbs), have completely
62 lost their neutralization potency against Omicron 1,3,4,11. Therefore, there is an
63 urgent need to explore and develop countermeasures against the Omicron
64 variant.

65 In this study, we used four NAbs named GW01, 16L9, 4L12, and REGN10987,
66 which failed to bind or neutralize the Omicron variant, to engineered full-length
67 IgG-like bispecific antibodies. We surprisingly found that these bispecific
68 antibodies could neutralize all the VOCs, including Omicron and authentic
69 Omicron(+R346K), while the parental antibody cocktail showed no
70 neutralization against Omicron. Cryo-EM structure study showed six dynamic
71 states of the Omicron S trimer upon bispecific antibody binding, including a
72 novel bi-trimer conformation, within which RBDs were all in “up” conformations.
73 This bi-trimer is critical for inhibiting ACE2 binding and explains the superiority
74 of the bispecific antibody. These novel bispecific antibodies are strong
75 candidates for the treatment and prevention of infection with the Omicron
76 variant and VOCs and other sarbecoviruses that may cause future emerging or
77 reemerging coronavirus diseases.

78

79 RESULTS

80 Isolation of three non-Omicron neutralizing antibodies from COVID-19

 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

81 convalescent individuals
82 We sorted and cultured SARS-CoV-2 S-specific memory B cells from two
83 recovered coronavirus disease 2019 (COVID-19) patients and discovered three
84 anti-SARS-CoV-2 NAbs, designated GW01, 4L12, and 16L9. The germlines
85 and CDR3 of these antibodies are listed in Table S1. All three antibodies
86 showed strong binding to the RBD of SARS-CoV-2 (Fig. 1A). However, they
87 had no or weak binding to the S trimer or S RBD of the Omicron variant (Fig.
88 1A).

89 GW01, 4L12, and 16L9 potently neutralized SARS-CoV-2 and the VOCs Alpha,
90 Beta, Gamma, and Delta, but they failed to neutralize the Omicron variant (Fig.
91 1B). A panel of control NAbs failed to neutralize the Omicron except S309. S309
92 neutralized Omicron to a similar degree as previous reports12,13. GW01 was a
93 cross-NAb that was able to neutralize SARS-CoV and the SARS-related
94 coronaviruses (SARSr-CoVs) RS3367 and WIV1. GW01 antibodies showed no
95 competition with 4L12, 16L9 or the control antibody REGN10987 in binding the
96 RBD (Fig. 1C), indicating that GW01 binds to an epitope different from that
97 bound by 4L12, 16L9, and REGN10987.
98

99 Binding and neutralization of the Omicron variant by bispecific antibodies

100 We constructed bispecific antibodies targeting different epitopes in the RBD
101 using GW01 in combination with 16L9, 4L12, and REGN10987, and explored
102 their possibilities to neutralize Omicron variant. We linked the single-chain
103 variable fragments (scFvs) of the parental antibodies with a (Gly4Ser)4 linker
104 and then fused them to a hinge-CH2-CH3 fragment of human immunoglobulin
105 (hIgG1 Fc) to generate a single gene-encoded IgG-like bispecific antibody (Fig.
106 2A). For example, the sequence order of the GW01-16L9 (FD01) bispecific
107 antibody was as follows: GW01 VL-(Gly4Ser)3-GW01 VH-(Gly4Ser)4-16L9 VL-
108 (Gly4Ser)3-16L9 VH-hinge-CH2-CH3. SDS–PAGE results showed that the size
109 of the single chain of two representative bispecific antibodies, GW01-16L9 and
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

110 16L9-GW01, was approximately 100 kDa and that the purity was >95% (Fig.
111 2B). Crosslinking bispecific antibodies with glutaraldehyde revealed that the full
112 size of the bispecific antibodies was approximately 200 kDa, which was 10%
113 larger than that of the parental antibodies (180 kDa, Fig. 2B).

114 We constructed ten bispecific antibodies and tested their binding abilities to the
115 RBD or S trimer of SARS-CoV-2 and the Omicron variant. Eight bispecific
116 antibodies, FD01 (GW01-16L9), 16L9-GW01, GW01-REGN10987,
117 REGN10987-GW01, GW01-4L12, 4L12-GW01, 4L12-REGN10987, and 4L12-
118 16L9, not only strongly bound to the RBD of SARS-CoV-2 and the S trimer and
119 RBD proteins of the Omicron variant (Fig. 2C) but also showed high binding
120 affinity to these proteins (Fig. 2D). These results indicated that the structure of
121 the bispecific antibody increased the parental antibody binding affinity to the
122 Omicron RBD.

123 To understand the breadth of these bispecific antibodies, we performed a

124 neutralization assay using SARS-CoV-2 pseudoviruses, including Alpha, Beta,
125 Gamma, Delta, and Omicron variants, and the sarbecoviruses SARS-CoV,
126 WIV1 and RS3367. Surprisingly, these eight bispecific antibodies potently
127 neutralized the Omicron variant with IC50 values from 39.7 to 548 ng/ml. Six
128 bispecific antibodies containing the cross-NAb GW01 strongly neutralized all
129 the tested VOCs and sarbecoviruses (Fig. 2E). FD01 and GW01-REGN10987
130 were the best broadly NAbs, with geometric mean (GM) IC50 values of 25.4
131 and 12.1 ng/ml, respectively. 4L12-REGN10987 and 4L12-16L9 strongly
132 neutralized all the tested VOCs with GM IC50 values of 6.9 and 6.8 ng/ml,
133 respectively (Fig. 2E). However, the parental NAb combinations showed no
134 neutralization against the Omicron variant (Fig. 2E). Taken together, these data
135 indicated that bispecific antibodies consisting of non-Omicron NAbs efficiently
136 neutralize the Omicron variant in a way that is different from the antibody
137 cocktail.

138 To confirm the neutralization efficacy of the bispecific antibodies, we performed

 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

139 plaque reduction neutralization assays with an authentic Omicron variant

140 containing the R346K mutation, which escapes more SARS-CoV-2 NAbs than
141 the Omicron variant3. All five representative bispecific antibodies efficiently
142 neutralized the live Omicron variant (Fig. 2F), confirming that the bispecific
143 antibodies composed of non-Omicron NAbs are able to neutralize the Omicron
144 variant.
145

146 Bispecific antibodies bind to the ACE2-binding site

147 Competition assays were then performed to evaluate the abilities of the
148 bispecific antibodies to inhibit the binding of SARS-CoV-2 RBD to the
149 recombinant ACE2 protein (Fig. 2G). All five representative bispecific
150 antibodies prevented RBD binding to ACE2 protein, while the control antibody
151 S309 did not affect the S/ACE2 interaction. FD01 showed a strong inhibitory
152 effect against ACE2 binding. These results indicated that bispecific antibodies
153 inhibit Omicron variant infection by occupying the ACE2-binding site on the
154 RBD.
155

156 Snapshots of Omicron S-FD01 structures determined by cryo-EM

157 To further investigate the neutralization mechanism of the bispecific antibodies,
158 we chose FD01 (GW01-16L9) as a representative antibody for structural study.
159 Local refinement focused on the RBD and ScFvs improved the interface region
160 to 3.51 Å resolution and allowed us to unambiguously build the RBD and scFvs
161 (Table S2). We determined the cryo-EM structure of the prefusion stabilized
162 SARS-CoV-2 Omicron S ectodomain trimer in complex with the bispecific
163 antibody FD01 (Omicron S-FD01), revealing 6 states of the complex: In the
164 state 1 Omicron S-FD01 structure (~26% of the particles), only one 16L9 binds
165 to the “up” RBD, and the two down RBDs have no antibody binding (up-down-
166 down RBDs, 1 scFv, 3.47 Å). In state 2 (~38%), the bispecific antibody FD01
167 (GW01-16L9) binds to a widely open RBD, the so-called “wide_up” state, 16L9
168 binds to an “up” RBD, and the third RBD remains in a down state (wide_up-up-
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

169 down RBDs, 3 scFvs, 3.70 Å). In both state 3 (~8%) and state 4 (~15%), two
170 FD01 (GW01-16L9) bind separately to two “wide_up” state RBDs, and the third
171 RBD represents the half-up conformation in state 3 (wide_up-wide_up-half_up
172 RBDs, 4 scFvs, 3.91 Å) and the up conformation in state 4 (wide_up-wide_up-
173 up RBDs, 4 scFvs, 3.47 Å), without antibody binding. In state 5 (~4%), all RBDs
174 are in the “wide_up” state, each bound with an FD01 (GW01-16L9) (all wide_up
175 RBDs, 6 scFvs, 3.87 Å). In the final state 6 (~4% of the particles), two state 5
176 trimers are connected by three bispecific FD01 antibodies, forming a bi-trimer
177 structure (bi-trimer, all wide_up RBDs, 12 scFvs, 6.11 Å) (Fig. 3, Fig. S2 and
178 S3).

179

180 Collaborative binding mechanism of FD01 bispecific antibody

181 These six cryo-EM structures represent the conformational transitions of the
182 Omicron S trimer upon FD01 binding. To simplify the presentation, three
183 protomers of a spike trimer are clockwisely defined as 1, 2, and 3 (Fig. 3, Fig.
184 4). The apo spike trimer (state 0) includes one regular up RBD and two down
185 RBDs. First, 16L9 binds to the “up” RBD (RBD-1) of the apo spike trimers and
186 forms the state 1 confirmation. After that, GW01, connected with 16L9, binds to
187 RBD-1, inducing it into a wide_up state (via an ~13 Å outward motion) and
188 pushing RBD-2 to flip from the “down” to the “up” state, making enough space
189 to accommodate the first bispecific antibody FD01 (16L9 and GW01) on RBD-
190 1 and the 16L9 of the second FD01 on RBD-2. A slight ~3 Å inward motion of
191 RBD-3 is induced by the neighboring RBD-2 to stabilize this state (state 2).
192 Then, in state 3, the “up” RBD-2 opens up further to the “wide_up” state,
193 allowing both 16L9 and GW01 of the second FD01to bind RBD-2. In addition,
194 RBD-3 is pushed up to a “half_up” state. In state 4, the “half_up” RBD-3 opens
195 up to the regular “up” state and is ready for 16L9 binding. Following that, in state
196 5, the third FD01 binds to RBD-3 and induces RBD-3 to adopt the “wide_up”
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

197 state. Finally, two trimers in state 5 form a bi-trimer induced by three pairs of Fc
198 regions from six antibodies.
199 Thus, the six states represent the continuous conformational transitions starting
200 from the first 16L9 binding to three FD01 binding and the final bi-trimer
201 formation, which inhibits the ACE2 binding by aggregating virions. In addition to
202 the motion of RBDs, the N-terminal domains (NTDs) of the trimer are also
203 moved or rotated following the motion of their neighboring RBDs.
204

205 FD01 targets two conserved epitopes

206 16L9 and GW01 bind to two different sites of one RBD (Fig. 5A). The epitope
207 of 16L9 almost overlaps with the receptor-binding motif (RBM), while GW01
208 binds outside the RBM. The binding of 16L9 and the RBD buries a 1061 Å2
209 surface area, and a total of 20 residues from RBD are involved. The interaction
210 between 16L9 and RBD is largely driven by extensive hydrophilic and
211 hydrophobic interactions between CDRH1, CDRH2, CDRH3 and CDRL1 of
212 16L9 and RBD (Fig. 5C). Residues D405, T415, D420, Y421, Y453, L455, F456,
213 Y473, A475, N477, Y489, R493, P500, Y501, and H505 of the RBD are involved
214 in this interaction, forming 14 pairs of hydrogen bonds and 3 patches of
215 hydrophobic interactions (Fig. 5C). In addition, the hydrogen bond between S96
216 of CDRL3 and R403 from the RBD and the salt bridge between E52 of CDRL2
217 and R493 from the RBD further enhance the interaction (Fig. 5C).
218 Coincidentally, residues Y453, A475, Y489, R493, T500, Y501, and H505 of the
219 Omicron RBD are important for ACE2 recognition and binding14. The
220 neutralization activity test showed that 16L9 alone was able to broadly
221 neutralize SARS-CoV-2 and SARS-CoV-2 variants. This result implies that
222 16L9 targets the conserved residues of the RBM, which is needed for receptor
223 binding.
224 GW01 interacts with another novel conserved epitope beyond the binding site
225 of 16L9. The binding site of GW01 and the RBD has a buried surface area of
226 668.2 Å. The interaction between GW01 and the RBD is mainly contributed by
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

227 CDRH3. The long loop (226-YGPPDVFNY-234) of CDRH3 engages with Y369,
228 F374, T376, F377, Y508, and V503 from the RBD, forming 3 patches of
229 hydrophobic interactions and 2 pairs of hydrogen bonds. D155 on CDRH1 and
230 N53 on CDRL2 are also involved in the interaction by forming hydrogen bonds
231 between V503 and N370, respectively (Fig. 5D).
232 Interestingly, the simultaneous binding of 16L9 and GW01 with the RBD
233 introduces additional interactions. Hydrogen bonds are formed between N178
234 and N196 of GW01 and S26, Y93, S96 and N98 of 16L9 (Fig. 5B), which further
235 enhances the interaction between FD01 and S.
236 Structural alignment of the 16L9-GW01-RBD complex with the ACE2–RBD
237 complex indicated that both 16L9 and GW01 were able to compete with ACE2
238 when binding to the RBD (Fig. 5E), which is consistent to the competition assay.
239

240 Bispecific antibodies accommodated the mutations in the Omicron

241 variant

242 We constructed 34 single mutants of the Omicron variant to identify the key
243 residues that mediate resistance to GW01, 16L9, 4L12, and REGN0987. The
244 S371L mutation, which was found to stabilize the Omicron into a single-RBD-
245 down conformation12, greatly decreased the neutralization activities of GW01,
246 4L12, and REGN10987 (Table 1). The S375F mutation decreased the
247 neutralization activity of GW01 by 16-fold and resulted in resistance to
248 REGN10987. The K417N mutation resulted in complete resistance to 16L9
249 (>1000-fold). All six tested bispecific antibodies showed only a slight decrease
250 (12.4- to 25.5-fold) or no change in neutralization activity against the S371L
251 mutant. Therefore, bispecific antibodies bind to the ACE2-binding site of the
252 RBD and accommodate the S371L mutation of the Omicron variant, resulting
253 in extraordinary breadth for these bispecific antibodies.

254

255 DISCUSSION
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

256 The recently emerged SARS-CoV-2 Omicron strain raised unprecedented

257 global concern about invalidation of most FDA-approved antibody drugs,
258 including LY-CoV555, LY-CoV016, REGN10933, REGN10987, AZD8895 and
259 AZD106115. Previous studies have shown that combining two NAbs that target
260 different neutralizing epitopes in the SARS-CoV-2 S protein increased
261 therapeutic and prophylactic efficacy. However, the most potent anti-SARS-
262 CoV-2 antibody cocktail, RENG10933/REGN10987, failed to neutralize the
263 Omicron variant. Zhou tested 10 antibody combinations against Omicron and
264 three antibody combinations with increasing neutralization contained a VH1-58
265 derived anti-Omicron NAb that bound RBD in the “up” position. Cyro-EM
266 structure showed that the antibody combination with improved neutralization
267 (B1-182.1/A19-46.1) synergistically induced the 3-RBD-up conformation12.
268 However, this antibody combination approach is not an ideal solution because
269 few anti-Omicron NAbs are available.
270 Using the NAbs that failed to neutralize the Omicron variant, we constructed a
271 serial of novel bispecific antibodies that capable of neutralizing all SARA-CoV-
272 2 variants of concern (VOCs), including the Omicron strain. Interestingly, single
273 IgG parental antibodies or the combination of parental antibodies failed to
274 neutralize Omicron, although the neutralization activity against SARS-CoV-2 or
275 SARS-CoV-2 Alpha, Beta, Gamma, and Delta was remarkable. Thus, the
276 effective neutralization of Omicron by bispecific antibodies may be due to the
277 construction of this bispecific antibody.
278 The structure of FD01 (GW01-16L9) bispecific antibody gives a hit at this
279 collaborative binding mechanism. First, one 16L9 scFv binds to the exposed
280 epitope of the “up” state RBD-1 in the apo Omicron S trimer as a trigger. Then,
281 the 20 aa GS linker between 16L9 and GW01 guides GW01 to its targeting
282 epitope of RBD-1, pushing RBD-1 more open, which unlocks the “down” state
283 of RBD-2 and induces it to adopt the up state. Thus, another 16L9 scFv could
284 easily catch the “up” state RBD-2. The same triggering process would occur on
285 RBD-2 and RBD-3, allowing the binding of the second and the third FD01 to
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

286 RBD-2 and RBD-3, respectively. GW01 Fab alone could not trigger binding to
287 the apo state of the Omicron trimer without the help of 16L9 and the GS linker
288 guider. Although 16L9 can bind to the first “up” state RBD, it lacks the ability to
289 release other “down” state RBDs for further binding (Fig. S5). In summary, the
290 two scFvs of the GW01-16L9 bispecific antibody have collaborative roles in the
291 neutralization process. The neutralization mechanism of FD01 may be
292 mediated by the unique engineering of the combination of two antibodies into
293 one, which induces RBD-up conformation, enlarges the interface area,
294 improves the affinity of a single IgG and the RBD, stabilizes the interaction by
295 additional interactions between the two antibodies, forms bi-trimer which
296 hinders the ACE2 binding by aggregating virions, and therefore blocks the
297 infection of SARS-CoV-2 Omicron variant.
298 Taken together, the unique construction of bispecific antibodies enables non-
299 Omicron NAbs to neutralize Omicron variant. Our approach can rescue the
300 majority of the SARS-CoV-2 antibodies, such as REGN10987, to overcome the
301 resistance of Omicron and prepare for future SARS-CoV-2 variants.

302

303 ACKNOWLEDGMENTS

304 We thank Center of Cryo-Electron Microscopy, Fudan University for the

305 supports on cryo-EM data collection. This work was supported by the National
306 Natural Science Foundation of China (31771008 to JH and 81900729 to LS),
307 the National Major Science and Technology Projects of China
308 (2017ZX10202102 to JH and 2018ZX10301403 to FW), Shanghai Municipal
309 Health Commission (2018BR08 to JH), the Chinese Academy of Medical
310 Sciences (2019PT350002 to JH).

311

312 AUTHOR CONTRIBUTIONS

313 JH, LS, and FW conceived and designed the experiments. JH, ML, and FW
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

314 performed B cell sorting and antibody cloning. YDW and YJ constructed the
315 bispecific antibodies and performed neutralization assay, ELISA, bilayer
316 interferometry experiments. YDW, YJ, and QW constructed and expressed
317 SARS-CoV-2 pseudovirus mutants and purification antibodies. LS, XZ, WZ, and
318 ZC performed the structural studies. JZ and YQW were responsible for the
319 authentic virus experiments. JH, YDW, YJ, FW, LS, XZ, WZ, JZ, and YQW
320 analyzed the data. YW and TZ supervised the project. JH, LS, FW, YDW, and
321 XZ wrote the manuscript.

322

323 COMPETING INTERESTS

324 Patents about the bispecific antibodies in this study are pending.

325
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

326
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

A
GW01 16L9 4L12 REGN10987(Control)
Binding (OD 405)

Antibody concentration (μg/mL)

Ab / Virus ID
IC50 (ng/ml) % of GM Median
WT Alpha Beta Gamma Delta Omicron SARS-CoV WIVI RS3367 Neut IC50 IC50
GW01 23.1 15.4 51.0 89.1 50.2 >10,000 56.8 2.3 1.4 89 18.7 36.6
16L9 4.1 9.3 412 5.4 7.0 >10,000 >10,000 >10,000 >10,000 56 14.2 7.0
4L12 4.5 6.4 12.0 1.3 19.2 >10,000 >10,000 >10,000 >10,000 56 6.1 6.4
S309 30.8 86.6 41.1 60.0 63.5 192 24.1 113 295 100 73.7 63.5
REGN10987 2.3 1.7 5.7 38.5 32.9 >10,000 >10,000 >10,000 >10,000 56 7.8 5.7
CC12.1 17.3 301 >10,000 24.5 20.3 >10,000 >10,000 >10,000 >10,000 44 40.1 22.4
REGN10989 5.7 0.4 >10,000 >10,000 2.9 >10,000 >10,000 >10,000 >10,000 33 1.8 2.9
4A8 196 2269 >10,000 >10,000 1224 >10,000 >10,000 >10,000 >10,000 33 817 1224

C mAb1=GW01
st
1 A b = 4 I3 mAb2
1 .2
4L12

16L9
0 .9
R E G N 10987

0 .6

0 .3

0 .0
0 100 200 300

mAb1=4L12
st
1 Ab=4L12
mAb2
0 .9 G W 01
Binding (nm)

R E G N 10987

0 .6 16L9

0 .3

0 .0
0 100 200 300

mAb1=16L9
st
1 Ab=16L9 mAb2
0 .6 4L12

G W 01

0 .4 R E G N 10987

0 .2

0 .0

-0 .2
0 100 200 300

Time (s)

327

328
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

329 Figure 1. Isolation of three non-Omicron neutralizing antibodies from

330 COVID-19 convalescent individuals.

331 (A) Binding of GW01, 16L9, and 4L12 to the SARS-CoV-2 RBD, Omicron RBD
332 and trimer in an ELISA. REGN10987 was used as a control. (B) Neutralizing
333 activities of GW01, 16L9, and 4L12 were determined against pseudotyped
334 SARS-CoV-2 and its variants Alpha, Beta, Gamma, Delta, and Omicron, as well
335 as sarbecoviruses. REGN10987 was used as a control. (C) Binding of 4L12,
336 16L9, and RGN10987 to the SARS-CoV-2 RBD in competition with GW01, as
337 measured by bilayer interferometry experiments.

338
339
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

A B SDS-PAGE Cross-linking SDS-PAGE

GW01 16L9- GW01 16L9-
-16L9 GW01 GW01 16L9 M M -16L9 GW01 GW01 16L9
kDa
180
130
100
70

55
40
Bispecific Ab
35

25

C FD01(GW01-16L9) GW01-REGN10987 GW01-4L12 4L12-REGN10987 4L12-16L9

Binding (OD 405)

16L9-GW01 REGN10987-GW01 4L12-GW01 REGN10987-4L12 16L9-4L12

Antibody concentration (μg/mL)

KD (nM)
mAb/
D bispecific Ab WT RBD
Omicron Omicron G C o p y o f C o p y o f 2 0 2 2 0 1 1 3 A C E 2 C o m p e t it io n S A R S 2 - R B D
trimer RBD
No competition
ACE2 + RBD
S 309+RBD
GW01 1.28 Escape Escape
16L9 1.15 Escape Escape 0 .5 +S309 S A R S 2 -R B D a n d V R C 0 1

4L12 1.49 Escape Escape S A R S 2 -R B D a n d B 4 + 1 0

Binding (nm)

10987 0.85 Escape Escape 0 .4 +IgG1 S A R S 2 -R B D a n d B 4 3

FD01(GW01-16L9) 0.43 8.8 2.5 +4L12-REGN10987
16L9-GW01 0.52 7.5 3.9 0 .3
+4L12-16L9 S A R S 2 - R B D a n d B 1 + 1 0
GW01-REGN10987 0.34 2.3 1.8
0 .2 +GW01-REGN10987
REGN10987-GW01 0.36 3.6 2.5 S A R S 2 -R B D a n d B 1 4
+GW01-4L12
GW01-4L12 0.81 8.3 3.7 S A R S 2 -R B D a n d B 1 3
4L12-GW01 0.24 2.9 4.1
0 .1
+FD01(GW01-16L9)
4L12-REGN10987 0.79 2.8 4.2 0 .0
+IgG1+ACE2 S A R S 2 - R B D a n d A C E 2 -F c
REGN10987-4L12 0.45 Escape Escape 0 200 400 600

4L12-16L9 1.8 4.2 12.3

Time (s)
16L9-4L12 4.4 Escape Escape

E F
Pseudoviruses Authentic Omicron(+R346K)
IC50 (ng/ml) % of GM Median Ab IC50 (μg/ml)
Ab / Virus ID
WT Alpha Beta Gamma Delta Omicron SARS-CoV WIVI RS3367 Neut IC50 IC50 FD01(GW01-16L9) 1.285
S309 (control) 30.8 86.6 41.1 60.0 63.5 192 24.1 113 295 100 73.7 63.5 GW01-REGN10987 0.6487
FD01(GW01-16L9) 20.3 36.2 33.9 6.5 88.4 69.6 53.8 19.8 4.2 100 25.4 33.9 GW01-4L12 2.319
16L9-GW01 47.4 25.5 109 6.4 79.0 39.7 434 161 20.7 100 53.9 47.4 4L12-REGN10987 1.451
GW01-REGN10987 5.4 30.8 3.9 2.8 12.9 59.0 79.8 10.6 4.9 100 12.1 10.6 4L12-16L9 0.6125
REGN10987-GW01 5.8 4.5 4.9 5.8 9.9 668 1526 215 6.0 100 27.8 6.0
GW01-4L12 22.4 159 43.0 7.6 95.3 548 47.5 25.8 2.1 100 37.8 43.0
4L12-GW01 69.9 137 32.1 31.1 123 183 400 69.5 14.0 78 75.9 69.9
4L12-REGN10987 2.0 6.7 1.5 4.5 4.7 245 >10,000 >10,000 >10,000 78 6.9 4.6
REGN10987-4L12 0.4 2.5 1.6 1.6 2.2 >10,000 >10,000 >10,000 >10,000 56 1.4 1.6
4L12-16L9 0.6 7.7 5.4 3.0 8.5 158 >10,000 >10,000 >10,000 78 6.8 6.6
16L9-4L12 6283 >10,000 6976 4919 >10,000 >10,000 >10,000 >10,000 >10,000 33 5996 6283
GW01+16L9 mix 5.8 41.7 496 398 55.8 >10,000 217 212 27.4 89 87.3 134.1
GW01+REGN10987 mix 57.4 11.8 5.3 6.9 43.7 >10,000 299 76.5 26.2 89 30.0 34.9
GW01+4L12 mix 14.9 106 153 3.6 56.6 >10,000 442 62.3 16.8 89 46.7 59.4
4L12+REGN10987 mix 64.5 11.8 9.9 10.5 17.2 >10,000 >10,000 >10,000 >10,000 56 16.8 11.8
4L12+16L9 mix 57.3 46.2 164 36.6 31.3 >10,000 >10,000 >10,000 >10,000 56 54.8 46.2

340
341
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

342 Figure 2. Binding and neutralization of the Omicron variant by bispecific

343 antibodies.

344 (A) Schematic diagrams showing the structures of bispecific antibodies. (B)
345 SDS–PAGE and cross-linking SDS–PAGE gels showing the sizes of the
346 representative bispecific antibodies and their parental antibodies. (C) Binding
347 specificities of the bispecific antibodies to the SARS-CoV-2 RBD-his, Omicron
348 trimer-his, or RBD-his protein. (D) Binding affinities of the bispecific antibodies
349 to the SARS-CoV-2 RBD-his, Omicron trimer-his, or RBD-his protein were
350 measured by bilayer interferometry experiments. (E) Neutralization by bispecific
351 antibodies and combinations of parental antibodies against the VOCs, including
352 Omicron variant, and sarbecoviruses. (F) Neutralization of five representative
353 bispecific antibodies against authentic Omicron(+R346K) variants. (G) Five
354 representative bispecific antibodies bind to the ACE2-binding site and block
355 RBD binding to ACE2. Binding of ACE2 to the SARS-CoV-2 RBD in competition
356 with bispecific antibodies (red), S309 (blue), control IgG1 (green), and
357 IgG1+ACE2 (black).

358

359

360

361

362

363

364

365

366

367

368

369

370
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

371

Bi-trimer
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

372 Figure 3. Cryo-EM structures of the Omicron S trimer in complex with

373 the bispecific antibody FD01.
374 The bispecific antibody FD01 binds to Omicron S trimers in six states. Two
375 perpendicular views of Omicron S-FD01 are shown in surface representation,
376 with 16L9 ScFv in lime and GW01 ScFv in cornflower blue.
377
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

378

Bi-trimer
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

379 Figure 4. Conformation transitions of Omicron S-FD01 in all states.

380 State 0 inside the red dashed box is a hypothetical apo state structure. Small
381 triangles inside the black dashed circle indicate the up/half_up/down states of
382 three RBDs in the trimer.
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

383
384
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

385 Figure 5. Two conserved epitopes recognized by FD01. (A) Close-up view
386 of the interaction between FD01 and Omicron; the Omicron RBD is displayed
387 in pink in surface representation. 16L9 and GW01 are shown as cartoons
388 colored green and medium-blue, respectively. (B) The interface between 16L9
389 and GW01. (C-D) The interaction of 16L9 (C) and GW01 (D). The residues
390 involved in interactions are represented as sticks. Polar interactions are
391 indicated as dotted lines. (E) Ribbon diagrams of FD01 and ACE2 (PDBID:
392 7T9L) bound to the Omicron RBD.
393
394
395
396
397
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

398 Table 1. Neutralization by GW01, 16L9, 4L12, REGN10987, and six

399 bispecific antibodies against 34 strains containing single mutations
Table1 Neutralization of GW01, 16L9, 4L12, REGN10987, and six bispecific antibodies against 34
400 present within the Omicron variant.
single mutants within Omicron variant.

IC50 fold change vs. WT

Virus ID REGN 16L9- GW01- GW01- 4L12- 4L12-
Spike GW01 16L9 4L12 FD01
10987 GW01 REGN10987 4L12 REGN10987 16L9
WT 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
A67V 0.7 0.8 4.4 1.1 2.8 2.1 0.8 3.2 3.3 31.3
Δ69-70 0.6 1.3 2.5 1.9 2.3 2.6 4.9 3.9 4.5 2.8
T95I 1.4 1.9 2.5 1.1 1.3 2.0 1.9 2.4 3.3 0.8
G142D 3.2 4.8 0.7 3.4 2.5 1.2 21.1 1.6 0.8 5.0
NTD Δ143-145 0.1 2.4 0.9 0.7 1.3 0.2 0.2 1.3 0.6 1.6
Δ211 0.5 0.8 9.5 5.2 3.8 0.5 3.3 0.8 0.4 1.4
L212I 0.6 1.1 1.3 2.1 3.7 2.1 1.5 7.0 5.3 5.0
ins214EPE 6.4 3.2 6.7 0.9 3.9 1.1 1.9 3.1 5.3 0.3
G339D 1.0 0.7 4.1 0.5 1.6 0.7 1.1 0.8 32.8 3.1
S371I 80.0 1.3 116.5 74.1 12.4 14.4 2.6 1.2 25.4 15.6
S373P 4.8 6.4 2.8 5.9 0.9 0.1 6.0 10.7 0.3 2.8
S375F 16.0 3.2 n.t. >1000 1.2 n.t. n.t. n.t. n.t. n.t.
K417N 3.9 >1000 0.3 1.7 1.7 1.6 4.3 3.8 5.8 4.4
N440K 2.3 7.3 3.5 >1000 3.8 1.1 1.1 2.6 0.9 1.0
G446S 1.1 1.1 9.1 >1000 1.8 1.6 0.5 2.5 0.9 0.8
S477N 4.1 1.4 16.0 0.9 2.2 2.1 3.7 8.9 3.1 4.1
RBD
T478K 1.5 2.3 4.5 1.0 2.2 2.7 2.8 5.5 2.4 1.6
E484A 0.5 3.0 2.5 1.4 4.2 1.3 1.4 3.9 1.0 0.9
RBM
Q493R 1.0 0.6 8.1 0.9 1.9 1.9 4.6 8.3 0.7 1.1
G496S 1.6 3.5 9.4 4.7 0.8 2.7 2.4 5.8 2.7 1.4
Q498R 0.4 0.8 4.8 0.9 2.5 0.5 2.2 0.6 0.4 1.4
N501Y 0.6 2.5 20.3 3.2 2.9 2.0 3.3 6.9 1.0 2.1
Y505H 0.2 1.0 19.3 0.2 7.9 2.9 6.4 1.7 0.7 0.9
SD1 T547K 0.6 6.4 4.8 0.2 2.5 0.8 0.6 2.1 0.9 0.4
D614G 2.4 1.8 4.6 1.5 2.3 1.1 2.3 3.6 3.4 1.2
H655Y 4.0 2.1 6.2 0.6 0.7 3.1 4.6 9.1 5.6 2.7
SD2
N679K 3.2 1.9 8.6 0.8 2.5 6.8 6.6 14.0 3.5 4.3
P681H 4.8 3.5 3.8 1.4 3.7 0.9 0.5 0.6 0.7 1.0
N764K 1.5 1.3 7.2 1.1 2.0 5.0 11.4 10.3 4.0 3.3
FP D796Y 6.4 0.7 8.2 0.5 1.9 5.0 2.2 8.4 4.0 3.3
N856K 9.5 1.9 8.1 0.7 0.7 1.3 1.5 4.9 1.6 1.5
Q954H 0.7 0.8 3.0 1.1 2.8 2.2 2.7 4.5 1.5 1.2
HR1 N969K 9.5 8.7 n.t. 5.9 2.4 n.t. n.t. n.t. n.t. n.t.
L981F 0.5 1.7 0.7 0.2 2.5 0.2 1.6 0.4 1.7 5.3

401 Fold change is defined as the IC50 of the mutant/the IC50 of the WT. Mutants that resulted
402 in fold change values between 10-50 are highlighted in yellow, and those with values >50
403 are highlighted in red. The S375F and N969K pseudoviruses were not available to some
404 of the antibodies and are labeled as not tested “n.t.”.
405

406

407
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

408 Materials and Methods

409 Cell lines, proteins, viruses and plasmids

410 The human primary embryonic kidney cell lines (HEK293T) and 293T-hACE2
411 cells were cultured in DMEM medium with 10% fetal bovine serum (FBS). RBD-
412 his proteins of SARS-CoV-2 and Omicron variant were purchased from Sino
413 Biological. Genes of bispecific antibodies were synthesized by Genscript. The
414 authentic Omicron (B.1.1.529) with R346K mutation used in this study were
415 isolated from COVID-19 patients in Guangzhou, passaged, and titered on Vero
416 E6 cells. African green monkey kidney-derived Vero E6 cells were grown in
417 Dulbecco's modified Eagle's medium (DMEM; Gibco, Grand Island, NY, USA)
418 supplemented with 10% fetal bovine serum (FBS). All work with authentic
419 SARS-CoV-2 was conducted at the Guangzhou Customs Technology Center
420 Biosafety Level 3 (BSL-3) Laboratory.

421 Production of Pseudoviruses

422 S genes of SARS-CoV-2 (NC_045512), Alpha (containing 69–70 and 144

423 deletions and N501Y, A570D, D614G, P681H, T716I, S982A, and D1118H
424 substitutions), Beta (containing D80A, D215G, 241-243 deletions and K417N,
425 E484K, N501Y, D614G and A701V substitutions), Gamma (containing
426 L18F,T20N,P26S,D138Y,R190S,K417T,E484K,N501Y,D614G.H655Y,T1027I,
427 and V1176F substitutions), Delta (containing T19R, 157-158 deletions and
428 L452R, T478K, D614G, P681R and D950N substitutions), and Omicron
429 (containing A67V, 69-70del, T95I, G142D, 143-145del, N211I, 212del,
430 ins215EPE, G339D, S371L, S373P, S375F, K417N, N440K, G446S, S477N,
431 T478K, E484A, Q493R, G496S, Q498R, N501Y, Y505H, T547K, D614G,
432 H655Y, N679K, P681H, N764K, D796Y, N856K, Q954H, N969K, L981F
433 substitutions), SARS-CoV, bat SARSr-CoVs (WIV1 and Rs3367) were
434 synthesized by BGI and constructed in pcDNA3.1 vector. Pseudoviruses were
435 generated by co-transfection of 293T cells with an env-deficient HIV backbone
436 pNL4-3.Luc.R-E- backbone and a spike expressing vector16,17. Fifty additional
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

437 spike variants carrying currently circulating single-point mutations were

438 constructed by site-directed mutagenesis.

439 Neutralization assay

440 The neutralization activities of mAbs and bispecific antibodies were determined
441 using a single-round pseudovirus infection of 293T-hACE2 cells. 10 μl of 5-fold
442 serially diluted antibody was incubated with 40 μl of pseudovirus in 96-well plate
443 at 37 °C for 1 h. 104 293T-hACE2 cells were then added to the mixture and
444 cultured for 48 h at 37 °C. Cells were lysed and firefly luciferase activity were
445 developed with a luciferase assay system (Promega) and detected on a
446 luminometer (Perkin Elmer). The IC50s of NAbs were
447 calculated using the GraphPad Prism 7.04 software (La Jolla, CA, USA).

448 Memory B-cell staining, sorting and antibody cloning

449 CD19+IgA−IgD−IgM− primary B cells were sorted out from peripheral blood
450 mononuclear cells (PBMC) of recovered patients of COVID-19 and expanded
451 in vitro in MEM medium with 10% FBS in the presence of irradiated 3T3-
452 msCD40L feeder cells, IL-2 and IL-21 as previously described 18. After fifteen
453 days of incubation, supernatants were screened for neutralization against
454 SARS-CoV-2. From the wells with SARS-CoV-2 neutralization activities, the
455 variable regions of the antibody (VH and VL) genes were amplified by RT–PCR.
456 mAbs were expressed as human IgG1 by HEK293F cells and purified using a
457 protein G column (Smart-Lifesciences).

458 ELISA

459 2 μg/ml of SARS-CoV-2 RBD-his, Omicron trimer-his, and Omicron RBD-his

460 protein were coated overnight at 4 °C in a 96-well plate (MaxiSorp Nunc-
461 immuno, Thermo Scientific, USA). Wells were blocked with 5% non-fat milk
462 (Biofroxx, Germany) in PBS for 1 hour at room temperature, followed by
463 incubation with 5-fold serially diluted mAb in disruption buffer (PBS, 5% FBS,
464 2% BSA, and 1% Tween-20) for 1 hour at room temperature. After 3 washing
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

465 steps with PBS 0.05% Tween 20 (PBS-T), 1:2500 diluted HRP-conjugated goat
466 anti-human IgG antibody (Jackson Immuno Research Laboratories, USA) was
467 added for 1 hour at room temperature. Plates were washed three times with
468 0.2% Tween-20 in PBS and developed using ABST (Thermo Scientific, USA)
469 for 30 minutes. Absorbance at 405 nm was read on a Multiskan FC plate reader
470 (Thermo Scientific, USA).

471 Biolayer interferometry (BLI) binding assay and competition assay

472 Experiments were carried out on a FortéBio OctetRED96 instrument. The

473 kinetics of monoclonal antibody binding to SARS-CoV-2 RBD-his, Omicron
474 RBD-his, or trimer-his proteins was measured using anti-human IgG (AHC)
475 biosensors. 10 μg/ml of mAbs were immobilized on biosensors for 200s. After
476 a 120 sec stabilization step with 0.02% PBST (PBS with 0.02% Tween),
477 biosensors were moved into 6 μg/ml of RBD-his or trimer-his proteins for the
478 300 sec association step. Then biosensors were moved into 0.02% PBST to
479 detect dissociation for 300 sec. The buffer control binding was subtracted to
480 deduct nonspecific binding. Kon, Koff, and KD were calculated by FortéBio Data
481 Analysis software (Version 8.1) using 1:1 binding and a global fitting model.

482 For the Ab competition assay, 10 μg/ml of mAbs 1 were immobilized on the anti-
483 human IgG (AHC) biosensors for 200s. After wash with 0.02% PBST for 120s
484 to reach baseline, biosensors were moved into 50 μg/ml of IgG1 isotype control
485 for 200s and then moved into SARS-CoV-2 RBD at 6 μg/ml for 300s. After wash
486 with 0.02% PBST for 120s, biosensors were moved into 10 μg/ml of mAb2 for
487 600s to detect the association between mAb2 and SARS-CoV-2 RBD.

488 For ACE2 competition, biosensors were moved into 20 μg/ml of ACE2-Fc for
489 600s. After baseline, wash, and blocking steps, biosensors were moved into
490 pre-mix of 600 nM of mAb and 100 nM SARS-CoV-2 RBD for 600s. A mixture
491 of ACE2-Fc and SARS-CoV-2 RBD was used as a positive control, while the
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

492 mixture of IgG1 isotype control and SARS-CoV-2 RBD was used as a negative
493 control.

494 Focus reduction neutralization test

495 SARS-CoV-2 Omicron(+R346K) focus reduction neutralization test (FRNT) was

496 performed in a certified Biosafety level 3 lab. Fifty microliters antibody were
497 serially diluted, mixed with 50 μl of SARS-CoV-2 (100 focus forming unit, FFU)
498 in 96-well microwell plates and incubated for 1 hour at 37°C. Mixtures were then
499 transferred to 96-well plates seeded with Vero E6 cells (ATCC, Manassas, VA)
500 for 1 hour at 37°C to allow virus entry. Inoculums were then removed before
501 adding the overlay media (100 μl MEM containing 1.2%
502 Carboxymethylcellulose, CMC). After 24-hour post infection, the overlay was
503 discarded and the cell monolayer was fixed with 4% paraformaldehyde solution
504 for 2 hours at RT. After permeabilized with 0.2% Triton X-100 for 20 min at room
505 temperature, the plates were sequentially stained with cross-reactive rabbit
506 anti-SARS-CoV-2 N IgG (Cat. No.: 40143-T62, Sino Biological Inc) as the
507 primary antibody and HRP-conjugated goat anti-rabbit IgG(H+L) (No.: 109-035-
508 088, Jackson ImmunoResearch) as the secondary antibody in 37°C for 1 hour
509 respectively. The reactions were developed with KPL TrueBlue Peroxidase
510 substrates. The numbers of SARS-CoV-2 foci were calculated using CTL
511 ImmunoSpot S6 Ultra reader (Cellular Technology Ltd). Neutralizing activity
512 was defined as the ratio of inhibition of SARS-CoV-2 focus comparing diluted
513 antibody to control.
514 Construction and expression of bispecific NAbs

515 Genes of a bispecific Ab consisting of the scFv of GW01 and scFv of 16L9,
516 REGN10987 or 4L12 were synthesized and codon-optimized by GenScript. The
517 bispecific antibody sequence alignment was as follows: variable light chain (VL)
518 and variable heavy chain (VH) of mAb 1 or mAb 2 were linked with a (Gly4Ser)3
519 linker. VL-VH of mAb 1 and VL-VH of mAb 2 were linked with a (Gly4Ser)4 linker
520 and then fused to the expression vector with hinge-CH2-CH3 fragment of
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

521 human immunoglobulin (hIgG1 Fc). FD01 bispecific antibody sequence order
522 was as follows: GW01 VL-(Gly4Ser)3-GW01 VH-(Gly4Ser)4-16L9 VL-(Gly4Ser)3-
523 16L9 VH-hinge-CH2-CH3.

524 293F cells were transiently transfected with bispecific Abs plasmid. After 6 days
525 of culture at 37°C in a 5% CO2 incubator, supernatant was collected and filtered.
526 Bispecific antibodies were purified with protein G colume (Smart-Lifesciences)
527 and stored in PBS at -80°C.

528 SDS-PAGE and cross-linking SDS-PAGE of bispecific antibodies

529 The purity and molecular weight of bispecific antibodies were then analyzed by
530 SDS-PAGE and cross-linking SDS-PAGE. Briefly, 5 μg of bispecific antibodies
531 were mixed with 5x SDS-loading sample buffer containing 10% β-
532 mercaptoethanol. The samples were heated for 10 min at 100°C and were then
533 loaded on an SDS gradient gel (4–20% Precast Protein Improve Gels,
534 Genscript Biotech Corporation). The gel was run for 120 min at 120 V, and
535 Coomassie staining was performed.

536 Extent of dimer was investigated by cross-linking of bispecific antibodies with

537 glutaraldehyde (Sigma-Aldrich). Briefly, 5 μg of antibodies were diluted in 25 μl
538 of PBS in the presence of a 2.7 µM of glutaraldehyde cross-linker. The mixture
539 was incubated at RT for 5 minutes, and then glutaraldehyde was quenched by
540 adding 1 M Tris-HCl buffer (pH 8.0) to a final concentration of 40 mM. After
541 mixing with 5x SDS-loading sample, the protein samples were loaded on a 4-
542 20% SDS gradient gel. The gel was run for 180 min at 120 V and confirmed by
543 Coomassie staining.

544 Expression and purification of SARS-CoV-2 Omicron Spike

545 The Human codon gene encoding SARS-CoV-2 Omicron S ectodomain was
546 purchased from GeneScript. The expression plasmid of Omicron S 6P
547 substitution19 was constructed and transfected into suspension HEK293F using
548 polyethlenimine. After 72 hours, the supernatants were harvested and filtered
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

549 for affinity purification by Histrap HP (GE). The protein was then further purified
550 by gel filtration using Superose 6 increase 10/300 column (GE Healthcare) in
551 20 mM Tris pH 8.0, 200 mM NaCl.
552 Cryo-EM sample preparation
553 Purified SARS-CoV-2 Omicron S at 1.554 mg/mL was mixed with FD01
554 antibody by a molar ratio of 1:1.5 incubated for 10 min on ice before application
555 onto a freshly glow-discharged holey amorphous nickel-titanium alloy film
556 supported by 400 mesh gold grids 20. The sample was plunged freezing in liquid
557 ethane using Vitrobot IV (FEI/Thermo Fisher Scientific), with 2 s blot time and -
558 3 blot force and 10 s wait time.
559 Cryo-EM data collection and image processing
560 Cryo-EM data were collected on a Titan Krios microscope (Thermo Fisher)
561 operated at 300 kV, equipped with a K3 summit direct detector (Gatan) and a
562 GIF quantum energy filter (Gatan) setting to a slit width of 20 eV. Automated
563 data acquisition was carried out with SerialEM software21 through beam-image
564 shift method22.
565 Movies were taken in the super-resolution mode at a nominal magnification
566 81,000×, corresponding to a physical pixel size of 1.064 Å, and a defocus range
567 from −1.2 μm to −2.5 μm. Each movie stack was dose-fractionated to 40 frames
568 with a total exposure dose of about 58 e−/Å2 and exposure time of 3s.
569 All the data processing was carried out using either modules on, or through,
570 RELION v3.023 and cryoSPARC24. A total of 4,363 movie stacks was binned
571 2 × 2, dose weighted, and motion corrected using MotionCor2 25 within RELION.
572 Parameters of contrast transfer function (CTF) were estimated by using Gctf 26.
573 All micrographs then were manually selected for further particle picking upon
574 ice condition, defocus range and estimated resolution.
575 Remaining 3,817 good images were imported into cryoSPARC for further
576 patched CTF-estimating, blob-picking and 2D classification. From 2D
577 classification, bi-trimer and trimer particles were observed. Several good 2D
578 classes of these two kind particles were used as templates for template-picking
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

579 separately. After 2D classification of particles from template-picking was

580 finished, all good particles from blob-picking and template-picking were merged
581 and deduplicated, subsequently being exported back to RELION through pyem
582 package 27.
583 For bi-trimer map, 1,003,956 particles were extracted at a box-size of 540 and
584 rescaled to 180, then carried on 2 round of 3D classification with a soft circular
585 mask of 480 Å in diameter in RELION. Only good classes were selected,
586 yielding 166,441 clean particles. These particles were re-extracted unbinned
587 (1.064 Å/pixel) and auto-refined without applying symmetry, yielding a map at
588 6.11 Å.
589 For trimer map, 1,003,956 particles were extracted at a box-size of 320 and
590 rescaled to 160, then carried on 1 round of 3D classification with a soft circular
591 mask of 220 Å in diameter in RELION. Three classes with different conformation
592 change on trimer RBDs were selected separately for another round of 3D
593 classification. Particles in different states were auto-refined, CTF-refined and
594 polished separately. Some density of RBDs and Fabs in some state were not
595 well-resolved, so we carried out no-alignment 3D classification with NTD-RBD-
596 Fabs (NRF) mask to improve those regions. Finally, we got 5 states of Omicron
597 S-FD01 trimer.
598 To get clear interfaces of RBD with Fabs, we did local-refinement focused on
599 that region. We first selected all good 3D-classes with relatively complete RBD
600 and FD01 density within all states. We auto-refined these particles with a C3-
601 aligned reference, but the auto-refinement procedure was not applied any
602 symmetry. Then particles were expanded with C3 symmetry and further
603 subtracted with one NRF mask. After no-alignment 3D-classification, 249,122
604 particles with complete NRF density were selected out, exported to cryoSPARC
605 and carried out local refinement, yielding a local-refined map at 3.51 Å.
606 The reported resolutions above are based on the gold-standard Fourier shell
607 correlation (FSC) 0.143 criterion. All the visualization and evaluation of 3D
608 density maps were performed with UCSF Chimera 28 and ChimeraX29. The
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

609 above procedures of data processing are summarized in Fig. S3 and Fig. S4.
610 These sharpened maps were generated by DeepEMhancer 30 and then “vop
611 zflip” to get the correct handedness in UCSF Chimera for subsequent model
612 building and analysis.
613 Model building and refinement
614 For model building of SARS-CoV-2 Omicron S FD01 complex, the SARS-CoV-
615 2 Omicron S trimer model and the antibody model generated by swiss-model 31
616 were fitted into the map using UCSF Chimera and then manually adjusted with
617 COOT 32. Several iterative rounds of real-space refinement were further carried
618 out in PHENIX 33. The RBD bounded with 16L9 and GW01 was refined against
619 the local refinement map and then docked back into global refinement trimer
620 and bi-trimer maps. Model validation was performed using phenix.MolProbity.
621 Figures were prepared using UCSF Chimera and UCSF ChimeraX29.

622 Data and materials availability: The cryo-EM map and the coordinates of
623 SARS-CoV-2 Omicron S complexed with FD01 have been deposited to the
624 Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) with
625 accession numbers EMD-32655 and PDB 7WOQ (state 1), EMD-32656 and
626 PDB 7WOR (state 2), EMD-32657 and PDB 7WOS (state 3), EMD-32659 and
627 PDB 7WOU (state 4), EMD-32660 and PDB 7WOV (state 5), EMD-32661 and
628 PDB 7WOW (state 6), EMD-32654 and PDB 7WOP (NTD-RBD-GW01-16L9
629 local refinement).

630

631 REFERENCES

632 1 Cameroni, E. et al. Broadly neutralizing antibodies overcome SARS-CoV-2 Omicron

633 antigenic shift. Nature, doi:10.1038/s41586-021-04386-2 (2021).
634 2 Cele, S. et al. Omicron extensively but incompletely escapes Pfizer BNT162b2
635 neutralization. Nature, doi:10.1038/s41586-021-04387-1 (2021).
636 3 Liu, L. et al. Striking Antibody Evasion Manifested by the Omicron Variant of SARS-CoV-
637 2. Nature, doi:10.1038/s41586-021-04388-0 (2021).
638 4 Planas, D. et al. Considerable escape of SARS-CoV-2 Omicron to antibody neutralization.
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

639 Nature, doi:10.1038/s41586-021-04389-z (2021).

640 5 Rössler, A. et al. SARS-CoV-2 Omicron Variant Neutralization in Serum from Vaccinated
641 and Convalescent Persons. doi:10.1056/NEJMc2119236 (2022).
642 6 Carreño, J. M. et al. Activity of convalescent and vaccine serum against SARS-CoV-2
643 Omicron. Nature, doi:10.1038/s41586-022-04399-5 (2021).
644 7 Garcia-Beltran, W. F. et al. mRNA-based COVID-19 vaccine boosters induce neutralizing
645 immunity against SARS-CoV-2 Omicron variant. Cell, doi:10.1016/j.cell.2021.12.033
646 (2022).
647 8 Dejnirattisai, W. et al. Reduced neutralisation of SARS-CoV-2 omicron B.1.1.529 variant
648 by post-immunisation serum. Lancet 399, 234-236, doi:10.1016/s0140-6736(21)02844-0
649 (2022).
650 9 Wang, Y. et al. Resistance of SARS-CoV-2 Omicron Variant to Convalescent and
651 CoronaVac Vaccine Plasma. Emerg Microbes Infect, 1-8,
652 doi:10.1080/22221751.2022.2027219 (2022).
653 10 Wang, Y. et al. The significant immune escape of pseudotyped SARS-CoV-2 variant
654 Omicron. Emerg Microbes Infect 11, 1-5, doi:10.1080/22221751.2021.2017757 (2022).
655 11 Dejnirattisai, W. et al. Omicron-B.1.1.529 leads to widespread escape from neutralizing
656 antibody responses. bioRxiv, doi:10.1101/2021.12.03.471045 (2021).
657 12 Zhou, T. et al. Structural basis for potent antibody neutralization of SARS-CoV-2 variants
658 including B.1.1.529. 2021.2012.2027.474307, doi:10.1101/2021.12.27.474307 %J bioRxiv
659 (2021).
660 13 VanBlargan, L. A. et al. An infectious SARS-CoV-2 B.1.1.529 Omicron virus escapes
661 neutralization by therapeutic monoclonal antibodies. Nature Medicine,
662 doi:10.1038/s41591-021-01678-y (2022).
663 14 Mannar, D. et al. Structural analysis of receptor binding domain mutations in SARS-CoV-
664 2 variants of concern that modulate ACE2 and antibody binding. Cell Rep 37, 110156,
665 doi:10.1016/j.celrep.2021.110156 (2021).
666 15 Cao, Y. et al. Omicron escapes the majority of existing SARS-CoV-2 neutralizing
667 antibodies. Nature, doi:10.1038/s41586-021-04385-3 (2021).
668 16 Ju, B. et al. Human neutralizing antibodies elicited by SARS-CoV-2 infection. Nature 584,
669 115-119, doi:10.1038/s41586-020-2380-z (2020).
670 17 Wu, F. et al. Evaluating the Association of Clinical Characteristics With Neutralizing
671 Antibody Levels in Patients Who Have Recovered From Mild COVID-19 in Shanghai, China.
672 JAMA internal medicine 180, 1356-1362, doi:10.1001/jamainternmed.2020.4616 (2020).
673 18 Huang, J. et al. Isolation of human monoclonal antibodies from peripheral blood B cells.
674 Nat Protoc 8, 1907-1915, doi:10.1038/nprot.2013.117 (2013).
675 19 Hsieh, C. L. et al. Structure-based design of prefusion-stabilized SARS-CoV-2 spikes.
676 Science 369, 1501-1505, doi:10.1126/science.abd0826 (2020).
677 20 Huang, X. et al. Amorphous nickel titanium alloy film: A new choice for cryo electron
678 microscopy sample preparation. Prog Biophys Mol Biol 156, 3-13,
679 doi:10.1016/j.pbiomolbio.2020.07.009 (2020).
680 21 Mastronarde, D. N. Automated electron microscope tomography using robust prediction
681 of specimen movements. J Struct Biol 152, 36-51, doi:10.1016/j.jsb.2005.07.007 (2005).
682 22 Wu, C. et al. High-quality, high-throughput cryo-electron microscopy data collection via
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

683 beam tilt and astigmatism-free beam-image shift. Journal of structural biology 208,
684 107396, doi:10.1016/j.jsb.2019.09.013 (2019).
685 23 Zivanov, J. et al. New tools for automated high-resolution cryo-EM structure
686 determination in RELION-3. Elife 7, doi:10.7554/eLife.42166 (2018).
687 24 Punjani, A. et al. M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure
688 determination. Nature methods 14, 290-296, doi:10.1038/nmeth.4169 (2017).
689 25 Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for
690 improved cryo-electron microscopy. Nat Methods 14, 331-332, doi:10.1038/nmeth.4193
691 (2017).
692 26 Zhang, K. Gctf: Real-time CTF determination and correction. J Struct Biol 193, 1-12,
693 doi:10.1016/j.jsb.2015.11.003 (2016).
694 27 Asarnow, D., Palovcak, E., Cheng, Y. UCSF pyem v0.5. Zenodo
695 https://doi.org/10.5281/zenodo.3576630 (2019).
696 doi:https://doi.org/10.5281/zenodo.3576630 (2019).
697 28 Pettersen, E. F. et al. UCSF Chimera--a visualization system for exploratory research and
698 analysis. J Comput Chem 25, 1605-1612, doi:10.1002/jcc.20084 (2004).
699 29 Pettersen, E. F. et al. UCSF ChimeraX: Structure visualization for researchers, educators,
700 and developers. Protein Sci 30, 70-82, doi:10.1002/pro.3943 (2021).
701 30 Sanchez-Garcia, R. et al. DeepEMhancer: a deep learning solution for cryo-EM volume
702 post-processing. Commun Biol 4, 874, doi:10.1038/s42003-021-02399-1 (2021).
703 31 Waterhouse, A. et al. SWISS-MODEL: homology modelling of protein structures and
704 complexes. Nucleic Acids Res 46, W296-W303, doi:10.1093/nar/gky427 (2018).
705 32 Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot.
706 Acta Crystallogr D Biol Crystallogr 66, 486-501, doi:10.1107/S0907444910007493 (2010).
707 33 Afonine, P. V. et al. Real-space refinement in PHENIX for cryo-EM and crystallography.
708 Acta Crystallogr D Struct Biol 74, 531-544, doi:10.1107/S2059798318006551 (2018).
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

709 Table S1. The germline and CDRH3 sequences of GW01, 4L12, and 16L9.

710
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

711 Table S2. Cryo-EM data collection and refinement statistics.

State State State State 4 State State Local
1 2 3 5 6 refine
NRF
Data collection
and processing
Magnification 81,000
Voltage (kV) 300
Electron exposure 58
(e–/Å2) -1.2 to -2.5
Defocus range (μm) 1.064
Pixel size (Å) 1,003,956
Initial particles (no.) C1
Symmetry imposed 194,02 62,04 74,41 141,577 39,29 71,568 249,122
Final particles (no.) 6 0 5 4
Map resolution (Å) 3.47 3.70 3.91 3.47 3.87 6.11 3.51

Refinement

R.m.s. deviations 0.003 0.003 0.003 0.003 0.003 0.002 0.002

Bond lengths 0.539 0.506 0.499 0.557 0.527 0.437 0.524
(Å)
Bond angles (°)
Validation
MolProbity 2.52 2.48 2.46 2.50 2.53 2.41 2.88
score 9.18 8.06 8.38 8.44 8.14 7.67 10.97
Clashscore 5.14 5.34 5.30 5.39 5.97 5.33 9.21
Rotamer outlier
(%)
Ramachandran
plot 91.75 92.00 92.74 91.84 91.67 93.20 88.62
Favored (%) 7.95 7.90 7.14 8.13 8.17 6.80 11.38
Allowed (%) 0.30 0.11 0.13 0.03 0.16 0.00 0.00
Disallowed (%)
EMDB 32655 3265 3265 32659 3266 32661 32654
6 7 0
PDB 7WOQ 7WO 7WO 7WOU 7WO 7WOW 7WOP
R S V
712
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

SARS-CoV-2 Omicron Omicron

RBD Trimer RBD
1 .0 1 .0

0 .8 0 .8

GW01 0 .6 Escape 0 .6
Escape
0 .4 0 .4

0 .2 0 .2

0 .0 0 .0
0 200 400 600 0 200 400 600
1 .0 1 .0

0 .8 0 .8

16L9 Escape
0 .6

0 .4
0 .6

0 .4
Escape
0 .2 0 .2

0 .0 0 .0
0 200 400 600 0 200 400 600
1 .0 1 .0

4L12 Escape
0 .8 0 .8

0 .6 Escape 0 .6

0 .4 0 .4

0 .2 0 .2

0 .0 0 .0
0 200 400 600 0 200 400 600
1 .0 1 .0

REGN10987
0 .8 0 .8

0 .6

0 .4
Escape 0 .6

0 .4
Escape
0 .2 0 .2

0 .0 0 .0
0 200 400 600 0 200 400 600

FD01
(GW01-16L9)
Binding (nm)

16L9-GW01

GW01-REGN10987

REGN10987-GW01

GW01-4L12

4L12-GW01

4L12-REGN10987

1 .0 1 .0

0 .8 0 .8

REGN10987-4L12
0 .6

0 .4
Escape 0 .6

0 .4
Escape
0 .2 0 .2

0 .0 0 .0
0 200 400 600 0 200 400 600

4L12-16L9

1 .0 1 .0

0 .8 0 .8

16L9-4L12 0 .6

0 .4
Escape 0 .6
Escape
0 .4

0 .2 0 .2

0 .0 0 .0
0 200 400 600 0 200 400 600

713 Figure S1. Binding affinities of GW01, 16L9, 4L12, REGN10987, and ten
714 bispecific antibodies to SARS-CoV-2 RBD-his, Omicron trimer-his and
715 Omicron RBD-his measured by bilayer interferometry experiments.
716 Antibodies were immobilized on anti-human IgG (AHC) biosensors and then
717 tested for their binding abilities to the target proteins.
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

718

719 Figure S2. Cryo-EM data collection and processing of FD01 bound SARS-
720 CoV-2 Omicron S. (A) Representative electron micrograph and 2D
721 classification results of FD01 bound SARS-CoV-2 S. (B) The reconstruction
722 map of the complex structures at six states. (C) The local-refined map of the
723 NRF region. (D) Gold-standard Fourier shell correlation curves generated in
724 RELION for structures of six states. The 0.143 cut-off is indicated by a
725 horizontal dashed line. (E) Gold-standard Fourier shell correlation curves
726 generated in cryoSPARC for local-refined map.
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

727 Figure S3. Data processing flowchart of FD01 bound SARS-CoV-2

728 Omicron S trimer. Particles number above cyan line is used for particle
729 counting statistics.
730
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

731

732 Figure S4. Data processing flowchart of local refinement of RBD-FD01.

733

734

735

736

737

738

739
 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.30.478305; this version posted January 31, 2022. The copyright holder for this
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

740

741 Figure S5. Hypothesis of binding features when Omicron S trimer meets
742 with mAbs of 16L9 or GW01.
743
744