Article https://doi.org/10.

1038/s41467-024-49096-1

A bispecific antibody exhibits broad

neutralization against SARS-CoV-2 Omicron
variants XBB.1.16, BQ.1.1 and sarbecoviruses
Received: 24 August 2023 Yingdan Wang1,2,6, Aihua Hao1,6, Ping Ji1,6, Yunping Ma1,3,6, Zhaoyong Zhang4,6,
Jiali Chen1, Qiyu Mao1, Xinyi Xiong4, Palizhati Rehati1, Yajie Wang1,
Accepted: 22 May 2024
Yanqun Wang 4, Yumei Wen1, Lu Lu 1, Zhenguo Chen 1, Jincun Zhao 4,5 ,
Fan Wu 3 , Jinghe Huang 1,2 & Lei Sun 1

Check for updates

1234567890():,;
1234567890():,;

The Omicron subvariants BQ.1.1, XBB.1.5, and XBB.1.16 of SARS-CoV-2 are

known for their adeptness at evading immune responses. Here, we isolate a
neutralizing antibody, 7F3, with the capacity to neutralize all tested SARS-CoV-
2 variants, including BQ.1.1, XBB.1.5, and XBB.1.16. 7F3 targets the receptor-
binding motif (RBM) region and exhibits broad binding to a panel of 37 RBD
mutant proteins. We develop the IgG-like bispecific antibody G7-Fc using 7F3
and the cross-neutralizing antibody GW01. G7-Fc demonstrates robust neu-
tralizing activity against all 28 tested SARS-CoV-2 variants and sarbecoviruses,
providing potent prophylaxis and therapeutic efficacy against XBB.1 infection
in both K18-ACE and BALB/c female mice. Cryo-EM structure analysis of the
G7-Fc in complex with the Omicron XBB spike (S) trimer reveals a trimer-dimer
conformation, with G7-Fc synergistically targeting two distinct RBD epitopes
and blocking ACE2 binding. Comparative analysis of 7F3 and LY-CoV1404
epitopes highlights a distinct and highly conserved epitope in the RBM region
bound by 7F3, facilitating neutralization of the immune-evasive Omicron var-
iant XBB.1.16. G7-Fc holds promise as a potential prophylactic countermeasure
against SARS-CoV-2, particularly against circulating and emerging variants.

Continuously, sub-lineages of Omicron spread across the world since (R346T, K444T, and N460K) on the receptor-binding domain (RBD) of
Omicron breakthrough in Nov. 2021, with increased transmissibility1 spike (S) protein. XBB.1.5 is a recombinant of the Omicron sub-lineages
and resistance to humoral immunity of convalescent and vaccine2,3. of BA.2.75 and BJ.1 with five more RBD substitutions (R346T,
BQ.1.1, XBB.1.5, and XBB.1.16 have been reported to have immune L368I, V445P, F486P, and F490S) than BA.2.756–8. XBB.1.16, another
evasion greater than those of earlier sub-lineages of Omicron1,4,5. BQ.1.1 descendent lineage of XBB with two additional mutations (E180A,
has evolved from the BA.5 sub-lineage with three additional mutations T478R) on the spike compared to XBB.1.5, has been spread rapidly

1
Key Laboratory of Medical Molecular Virology (MOE/NHC/CAMS), Shanghai Institute of Infectious Disease and Biosecurity, Shanghai Fifth People’s Hospital,
Institutes of Biomedical Sciences, School of Basic Medical Sciences, Fudan University, Shanghai, China. 2Shanghai Frontiers Science Center of Pathogenic
Microorganisms and Infection, School of Basic Medical Sciences, Fudan University, Shanghai, China. 3Shanghai Immune Therapy Institute, Shanghai Jiao Tong
University School of Medicine Affiliated Renji Hospital, Shanghai, China. 4State Key Laboratory of Respiratory Disease, National Clinical Research Center for
Respiratory Disease, Guangzhou Institute of Respiratory Health, the First Affiliated Hospital of Guangzhou Medical University, Guangzhou, Guangdong, China.
5
Shanghai Institute for Advanced Immunochemical Studies, School of Life Science and Technology, ShanghaiTech University, Shanghai, China. 6These
authors contributed equally: Yingdan Wang, Aihua Hao, Ping Ji, Yunping Ma, Zhaoyong Zhang. e-mail: [email protected]; [email protected];
[email protected]; [email protected]

Nature Communications | (2024)15:5127 1

Article https://doi.org/10.1038/s41467-024-49096-1

in 31 countries9. Due to the additional RBD mutations, BQ.1.1, XBB.1.5, XBB, XBB.1.5, or XBB.1.16 RBDs. We previously isolated a broadly NAb
and XBB.1.16 became resistant to a wide range of mAbs and antibody from a COVID-19 convalescent, named GW0124, which targeted the
cocktails, including those authorized for emergency clinical use, such outside of RBM and displayed broadly neutralization potency against
as LY-CoV1404 (also known as bebtelovimab) and Evusheld (also SARS-CoV-2 and SARS-CoV. However, it failed to bind BQ.1.1, XBB
known as tixagevimab and cilgavimab cocktail)2,4,10,11. (Fig. 1a). The binding affinities of 7F3, GW01, and LY-CoV1404 for the
Thousands of neutralizing antibodies targeting different epitopes RBDs of different SARS-CoV-2 strains accorded with the results of
of the spike protein have been identified and characterized12–19. The ELISA (Fig. 1b, Supplementary Fig. S2, and Supplementary Table S1).
majority of these antibodies recognize the RBD of the spike protein, To assess the potency and breadth of 7F3, we conducted neu-
while a small subset targets the NTD, SD1, SD2, or S2 stem-helix20. S3H3 tralization tests against a panel of 15 pseudotyped viruses expressing
antibody21 targeting SD1 as well as 12-16 and 12-19 antibodies22 target- the spike region of sarbecoviruses. Sarbecoviruses, encompassing a
ing the NTD-SD1 were able to neutralize XBB.1.5 and BQ.1.1. Antibodies group within the Betacoronavirus genus, share genetic affinities with
targeting RBD have proven effective in neutralizing SARS-CoV-2. the SARS-CoV and exhibit commonalities with other coronaviruses.
However, due to the viral evolution pressure and high genetic varia- Specifically, non-SARS-CoV-2 viruses, including SARS-CoV (SARS1),
bility of the RBD, most of the RBD antibodies failed to neutralize RS3367, and WIV1, are categorized as clade 1a sarbecoviruses, whereas
emerging SARS-CoV-2 variants of concern (VOCs), such as BQ.1.1, SARS-CoV-2 variants fall under clade 1b sarbecoviruses. 7F3 displayed a
XBB.1.5, and XBB.1.16. Only the cross-neutralizing antibodies targeting notable absence of neutralization activity against clade 1a sarbecov-
the conserved epitopes of sarbecoviruses in RBD, such as S309, SA55, iruses. However, it demonstrated potent neutralization against clade
and S2K146, remained effective for current SARS-CoV-2 VOCs10,11. The 1b sarbecoviruses, including SARS-CoV-2, Alpha, Beta, and Delta at sub-
S2 subunit is highly conserved, and the neutralization activities of S2 nanomolar level. Moreover, 7F3 exhibited broad spectrum neu-
stem-helix antibodies were less affected by the viral escape mutation23. tralization capacity against all the tested Omicron subvariants,
However, clinical application is unfortunately limited because of its including BA.1, BA.5, BF.7, and the highly resistant variants BQ.1.1,
low potency. XBB.1.5, XBB.1.16, and XBB.1.16.1 (Fig. 1c). In comparison, LY-CoV1404
With the emergence and rapid dissemination of SARS-CoV-2 var- exhibited high neutralization potency against various SARS-CoV-2
iants, the imperative for broadly neutralizing antibodies capable of strains with the exception of the Omicron subvariants BQ.1.1, XBB,
pan-sarbecovirus neutralization has intensified. Antibodies possessing XBB.1.5, XBB.1.16, and XBB.1.16.1 (Fig. 1c). GW01 exhibited high neu-
extensive neutralization across various sarbecoviruses are more likely tralization potency against the clade 1a sarbecoviruses, such as SARS-
to target conserved epitopes, rendering them more resilient against CoV, RS3367, and WIV1. It also neutralized various clade 1b SARS-CoV-2
immune evasion by swiftly emerging SARS-CoV-2 variants. Conse- variants except for the Omicron subvariants (Fig. 1c). An examination
quently, the development of novel pan-sarbecovirus antibodies exhi- of the combined breadth and potency of 7F3 and GW01 in a 1:1 ratio
biting broad and potent activity has become an urgent necessity for revealed no discernible enhancement in neutralizing SARS-CoV-2
both the prevention and treatment of coronavirus disease 2019 strains and Omicron subvariants (Fig. 1c). These results suggest that
(COVID-19). Moreover, considering that coronaviruses have incited 7F3 is a promising candidate for the development of broadly neu-
two pandemics and one epidemic within the past two decades invol- tralizing antibody therapies against SARS-CoV-2, particularly in the
ving SARS-CoV, SARS-CoV-2, and MERS-CoV, the demand for pan- context of the current surge in Omicron subvariants.
sarbecovirus antibodies becomes paramount in anticipation of
potential recurrences of coronavirus pandemics. Here, we identified an 7F3 blocks ACE2 binding and exhibits a robust ability to bind
RBM-binding antibody, designated 7F3, and demonstrated its unique RBD mutants
capability to neutralize various SARS-CoV-2 VOCs, including the newly In a bilayer interferometry (BLI) competition assay, 7F3, LY-CoV1404,
emerging Omicron subvariants BQ.1.1, XBB.1.5, and XBB.1.16. Impor- and GW01 impeded binding of the wild-type SARS-CoV-2 spike RBD to
tantly, G7-Fc, a bispecific antibody incorporating GW01 and 7F3 anti- angiotensin-converting enzyme 2 (ACE2) (Fig. 1d). In contrast, the
bodies, showed broad neutralization of SARS-CoV-2 variants and control antibody S309, which is a class III antibody and recognizes an
sarbecoviruses reported. G7-Fc prevented XBB.1 infection in mice and epitope outside the receptor-binding motif (RBM), exhibited no
bound two RBD sites: 7F3 bound a conserved epitope within the RBM, competition. These data indicate that 7F3 might bind within the RBM
and GW01 bound an epitope external to the RBM. These studies region or sterically block ACE2 from binding to the RBM.
highlight a conserved epitope within the RBM region targeted by 7F3 To map the epitope of 7F3 within the RBD, we used a panel of 37
that mediated the neutralization of the highly immune-evasive Omi- RBD mutant proteins bearing the circulating single or triple amino acid
cron variants BQ.1.1, XBB.1.5, and XBB.1.16. G7-Fc may represent a mutations and evaluated 7F3 binding to these mutants compared to
potential prophylactic countermeasure against SARS-CoV-2. the wild-type RBD. 7F3 retained binding to 100% of the tested RBD
mutants (Fig. 1e). In contrast, antibodies LY-CoV1404 and GW01 dis-
Results played diminished binding (below 85% of binding to the Wuhan-Hu-1
Isolation of a BQ.1.1 and XBB lineages neutralizing antibody 7F3 strain) to 9 and 5 RBD single mutants, respectively, and most RBD
from a COVID-19 convalescent individual triple mutants (Fig. 1e). These findings suggest that 7F3 targets an
We sorted and cultured memory B cells from a patient who had epitope within RBD that distinguishes it from other characterized
recovered from COVID-19 (Supplementary Fig. S1). After culture of the antibodies we investigated, which may enable its broad and potent
B cells for two weeks, we screened the supernatants of B cells for SARS- activity against circulating and emerging SARS-CoV-2 variants.
CoV-2 neutralization. We identified a SARS-CoV-2 neutralizing anti-
body (NAb), designated 7F3, which belongs to the IGHV3-49 and Construction and binding ability of bispecific antibody G7-Fc
IGKV3-11 immunoglobulin gene families. The antibody 7F3 exhibited We previously reported that IgG-like bispecific antibodies constructed
potent binding to the RBD proteins of the SARS-CoV-2 (Wuhan-Hu- by non-Omicron neutralizing antibodies were able to neutralize the
1 strain,GenBank: MN908947.3) wild-type (WT), Delta, BQ.1.1, XBB, Omicron variants BA.1 and BA.224,25. Therefore, we constructed bispe-
XBB.1.5, and XBB.1.16 variants, and weak binding to the BA.1 and BA.5 cific antibodies using 7F3 and GW01 to see if these two antibodies in
RBDs, as measured by enzyme-linked immunosorbent assay (ELISA, IgG-like structure format could enhance the breadth and potency of
Fig. 1a). In contrast, the control broadly neutralizing monoclonal 7F3. Two bispecific antibodies, GW01-7F3-Fc (G7-Fc) and 7F3-GW01-Fc
antibody LY-CoV1404 bound robustly only to the RBDs of wild- (7G-Fc), were successfully created as described presviously25 (Fig. 2a).
type (WT), Delta, BA.1, and BA.5 variants but did not engage the BQ.1.1, Briefly, the single-chain variable fragments (scFv) of antibodies were

Nature Communications | (2024)15:5127 2

Article https://doi.org/10.1038/s41467-024-49096-1

linked with a (Gly4Ser)4 linker, and subsequently fused to IgG1 configuration positions the single chain of 7F3 in the N terminal, fol-
Fc. The sequential arrangement of antibody G7-Fc, from N to C ter- lowed by the single chain of GW01. The formation of disulfide bonds
minus, is delineated as follows: GW01 VL-(Gly4Ser)3-GW01 VH-(Gly4- within the hinge region imparts a dimeric structure to the bispecific
Ser)4-7F3 VL-(Gly4Ser)3-7F3 VH-Hinge-CH2-CH3. In contrast, the 7G-Fc antibody. The single chain of bispecific antibodies was approximately

7F3 LY-CoV1404 GW01 WT

a
RBD binding (OD 405)

4 Delta
4 4
BA.1
3 3
3 BA.5
BF.7
2 2 2
BQ.1.1

1 1 1 XBB
XBB.1.5
0 0 0 XBB.1.16
0.01 0.1 1 10 100 1000 0.01 0.1 1 10 100 1000 0.01 0.1 1 10 100 1000

Antibody concentration (nM)

b
KD nM)
mAb ID
WT Delta BA.1 BA.5 BF.7 BQ.1.1 XBB XBB.1.5 XBB.1.16
7F3 <0.01 20.80 N.D. N.D. 1283 24.10 19.89 2.58 28.61
LY-CoV1404 <0.01 <0.01 <0.01 <0.01 <0.01 605.9 N.D. N.D. N.D.
GW01 0.42 <0.01 1732.00 74.79 33.01 561.6 N.D. 1663 <0.01

c e
IC50 nM) % of binding to WT RBD protein
Virus ID Mutants
7F3 LY-CoV1404 GW01 7F3+GW01 7F3 LY-CoV1404 GW01
WT 0.01781 0.00874 0.2159 0.3801 WT 1.00 1.00 1.00
B.1.1.7 0.04163 0.03898 0.2424 0.1733 A372T 0.99 1.02 1.03
B.1.351 0.03436 0.00758 0.2093 0.1554 F374A 0.97 0.82 0.77
B.1.617.2 0.02752 0.01944 0.1502 0.2779 T376A 0.98 0.94 0.97
BA.1 104.3 0.05272 >333.3 126.9 F377A 0.98 0.94 0.98
BA.5 62.520 0.03098 276.1 20.910 S383A 0.99 1.00 1.00
BF.7 23.130 0.03972 242.5 40.070 P384A 0.99 0.90 0.99
BQ.1.1 2.054 >333.3 >333.3 4.778 T385A 0.98 1.05 0.98
XBB 44.280 >333.3 >333.3 65.630 D427A 0.97 0.76 0.69
XBB.1.5 0.786 >333.3 >333.3 1.081 D428A 0.99 0.95 0.89
N439K 0.95 1.04 0.98
XBB.1.16 16.080 >333.3 >333.3 14.910
XBB.1.16.1 21.210 >333.3 >333.3 50.260
G446V 0.87 0.13 0.98
L452R 1.03 1.01 0.98
SARS-CoV >333.3 >333.3 0.843 2.719
I472V 1.01 1.08 0.94
RS3367 >333.3 >333.3 0.1042 0.1864
A475V 1.01 0.85 0.97
WIVI >333.3 >333.3 0.02446 0.01878 T478I 1.01 0.97 1.01
V483A 1.00 0.82 0.97
<333.3 <66.6 <6.66 <0.66 <0.066
E484A 0.86 0.64 0.67
G485A 1.01 0.86 0.95
F486A 0.99 1.03 1.04
N487A 1.02 0.91 0.99
F490L 1.02 0.85 0.98
Q493A 1.00 1.06 1.02
N501Y 1.01 0.90 1.01
d ACE2 + WT RBD + mAb Blockage (%)
G504N 1.00 0.89 1.01
Y508H 1.01 1.02 0.96
0
0.5 +S309 E516A 0.99 0.82 0.81
+VRC01 (Neg) 0
L517A 0.99 0.82 0.74
+7F3 40 NNL439-441AAA 1.06 0.99 0.87
Binding (nm)

0.4
+4I3 57 NYN448-4501AAA 0.93 0.21 0.70
0.3 +LY-Cov1404 91 GST476-478AAA 0.99 0.85 0.70
+ACE2+VRC01 (Pos) 100 NGV481-483AAA 0.93 0.72 0.63
0.2 NGV500-502AAA 0.96 1.07 0.96
YQP504-506AAA 1.04 0.20 0.60
0.1
<=0.85 <0.5
0 200 400 600
Time (s)

Fig. 1 | Isolation of BQ.1.1 and XBB.1.5 neutralizing antibody 7F3 from COVID-19 values. d Binding of 7F3 to the SARS-CoV-2 RBD in competition with ACE2 as
convalescent individuals. a Binding specificity of 7F3 to the RBDs of SARS-CoV-2, measured by BLI experiments. The HIV-1 antibody VRC01, severed as an IgG1
Delta, BA.1, BA.5, BF.7, BQ.1.1, XBB, XBB.1.5, and XBB.1.16 as measured in an ELISA. isotype-matched negative control, and a mixture of VRC01 and the ACE2 receptor
LY-CoV1404 and GW01 were used as controls. b Binding affinity KD values of 7F3 to served as a positive control. Blockage was calculated by the values of maximum
the receptor binding domains (RBDs) of SARS-CoV-2 variants as measured by BLI. binding. The maximum binding of positive control (ACE2) was set as 100%, and
LY-CoV1404 and GW01 were used as controls. c Neutralizing activities of 7F3, LY- negative control was set as 0%. e Binding of 7F3 to 33 RBD-Fc proteins containing
CoV1404, GW01, and the combination of 7F3 and GW01 at the ratio of 1:1 against single or triple mutations. GW01 and LY-CoV1404 were used as controls. The
pseudotyped SARS-CoV-2 subvariants as well as sarbecoviruses. The number dis- experiments in (c) were independently performed twice with similar results.
played in the box represents the half-maximal inhibitory concentration (IC50) Source data are provided as a Source Data file.

Nature Communications | (2024)15:5127 3

Article https://doi.org/10.1038/s41467-024-49096-1

85 kDa, while the full-length crosslinked bispecific antibodies were compared to the parental antibodies (Fig. 2d, Supplementary Fig. S3,
about 170 kDa after glutaraldehyde-mediated crosslinking, as visua- and Supplementary Table. S1). The bispecific antibody 7G-Fc, which
lized by SDS-PAGE (Fig. 2b). This mass is only about 10–20% larger than has a reverse orientation of GW01 and 7F3, was found to be less
the full-size IgG of 7F3 (150 kDa) and GW01 (130 kDa). efficient in binding Omicron RBD proteins than G7-Fc, with lower
Bispecific antibody G7-Fc strongly bound to the RBDs of Wuhan- binding affinity and bigger equilibrium dissociation constant (KD)
Hu-1 (WT), Delta, BA.1, BA.5, BQ.1.1, XBB.1.5, and XBB.1.16 (Fig. 2c) values (Fig. 2d). These results suggested that the structure of the
and exhibited a higher binding affinity towards these proteins bispecific antibody and the order of the parental antibodies impact

Nature Communications | (2024)15:5127 4

Article https://doi.org/10.1038/s41467-024-49096-1

Fig. 2 | Neutralization breadth and potency of bispecific antibody G7-Fc. and sarbecoviruses. LY-CoV1404 was used as a control. The number in the box
a Schematic diagrams of the structure of bispecific antibodies. b Cross-linking indicates IC50 values. Geometric mean (GM) IC50 was determined. f Neutralization
sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and SDS- of authentic viruses BQ.1 and XBB.1 by G7-Fc. The curves were fitted using nonlinear
PAGE analysis of G7-Fc, 7G-Fc, 7F3, GW01. Binding specificities (c), binding affinities regression (log [inhibitor] vs. normalized response, variable slope). The dots on the
KD values (d) of the bispecific antibodies G7-Fc and 7G-Fc to the RBD-His proteins curve represent the mean values of triplicate experiments, accompanied by bars
of WT, Delta, BA.1, BA.5, BF.7, BQ.1.1, XBB, XBB.1.5, XBB.1.16 as measured by ELISA indicating the standard error of the mean (SEM). The experiments in b, e were
and BLI experiments. ND represents not determined. e Neutralization breadth and independently performed twice with consistent results. Source data are provided in
potency of bispecific antibodies G7-Fc and 7G-Fc against the SARS-CoV-2 variants the Source Data file.

the binding affinity of the bispecific antibody to the RBDs of Omi- Cryo-EM structure of SARS-CoV-2 Omicron XBB S complexed
cron subvariants. with G7-Fc
To further understand the neutralization mechanism of G7-Fc, we
Neutralization activity of bispecific antibody G7-Fc determined the cryo-electron microscopy (cryo-EM) structure of the
To understand the breadth and potency of these bispecific anti- prefusion-stabilized SARS-CoV-2 Omicron XBB S ectodomain trimer
bodies, we performed a neutralization assay using a panel of twenty- complexed with G7-Fc in its full-length IgG form (XBB S/G7-Fc) (Sup-
eight pseudoviruses, including SARS-CoV-2 WT (Wuhan-Hu-1 strain), plementary Fig. S4a). Negative staining EM images showed that incu-
Alpha, Beta, Delta, twenty-one Omicron subvariants including BQ.1.1, bation of G7-Fc with S induced the formation of a trimer dimer
XBB.1.5, and XBB.1.16, as well as other sarbecoviruses such as SARS- consisting of two S trimers (Supplementary Fig. S4c). However, XBB S
CoV, WIV1, and RS3367. Remarkably, G7-Fc bispecific antibody trimers failed to form trimer dimers when incubated with G7-Fab
exhibited robust neutralization potency against 100% of the tested (Supplementary Fig. S4b, c). The whole complex was determined
pseudoviruses, with a geometric mean (GM) IC50 value of 0.13 nM to a resolution of 3.75 Å. To obtain detailed information about
(Fig. 2e), while the parental antibodies GW01, 7F3, and their com- the interactions between S and G7-Fc, local refinement focusing on
bination only neutralized 32.1%, 67.9%, and 89.3% with GM IC50 single trimers and RBD/scFvs was performed and improved the inter-
values of 0.85, 2.52, and 4.21 nM, respectively. Significantly, G7-Fc face region to 3.0 Å (Figs. 4a, e and Fig. 5a, Supplementary
outperformed the parent IgG and the mixture of IgG. G7-Fc potently Figs. S5 and S6).
neutralized BQ.1.1, XBB.1.5, and XBB.1.16 with IC50 values of 0.1143, The binding of G7-Fc induces S trimers to the state with all three
0.2173, and 0.4263 nM, respectively. The neutralization potency of RBDs up. Due to the distance limitation of the GS linkers, 7F3 and
7G-Fc, characterized by a reversed orientation of GW01 and 7F3, is GW01 from one arm of the G7-Fc bind to different RBDs, with 7F3
not comparable to that of G7-Fc. This observation corroborates the binding to RBM and GW01 binding to a non-RBM region. Two 7F3s are
notion that the arrangement of parental antibodies significantly close to each other but do not form close interactions. The Fc regions
influences both neutralizing activity and binding affinity, as pre- cross-link two S trimers, forming an unsymmetrical head-to-head
viously discussed. In contrast, LY-CoV1404 only neutralized 67.9% of dimer of trimers (Fig. 4a, Supplementary Fig. S7). 7F3 did not compete
the tested pseudoviruses. Importantly, it failed to neutralize BQ.1.1, with GW01 for RBD engagement (0%, Fig. 4b), implying that the 7F3
XBB.1.5, XBB.1.16, XBB.1.16.1, and BA.5.6.2 (Fig. 2e). Moreover, G7-Fc epitope likely differs from those of GW01 within the RBM. Antibodies
bispecific antibody demonstrated robust neutralization against the directed towards the RBD can be classified into four general categories
authentic viruses BQ.1 and XBB.1 in vitro (Fig. 2f). Taken together, (classes I to IV) based on their competition with the ACE2 and their
these data indicated that bispecific antibody G7-Fc is a sarbecovirus recognition of the up or down state of the three RBDs in spike27.
neutralizing antibody and potently neutralizes the Omicron sub- 7F3 strongly competed with the class II antibody BD2328 for Wuhan-Hu-
variants, especially the highly resistant variants BQ.1.1 and XBB.1.5, 1 (WT) RBD engagement (100%, Fig. 4b). In contrast, CR3022 showed
XBB.1.16. These results provide important insights into the devel- limited competition with 7F3 (17.4%). Moreover, CR302229, class IV,
opment of effective treatment options for COVID-19. exhibited strong competition to the WT RBD compared to GW01
(96.21%, Fig. 4c), whereas 7F3 and BD23 showed no competition or
Prophylactic and therapeutic efficacy of G7-Fc against XBB.1 limited competition with 7F3 (0% and 34.81%, respectively, Fig. 4c).
infection of mice These findings suggest a substantial epitope overlap exists between
To evaluate the in vivo prophylaxis and therapeutic efficacy of G7-Fc, 7F3 and BD23, and overlap between CR3022 and GW01 epitopes.
we administered the G7-Fc into K18-ACE2 transgenic mice and BALB/c Based on the Barnes Classification of antibodies, 7F3 is categorized as a
mice either before or after infection of XBB.1 (Fig. 3a). As XBB.1 tends class II antibody, whereas GW01 is classified as a class IV antibody.
to attach to the cells in the upper airways26, inhalation antibody ther- 7F3 binds to the RBM region and shares 26 epitope residues with
apy demonstrated predominant therapeutic efficacy in animal the ACE2 binding site (Fig. 4e, f, Supplementary Fig. S8b). The binding
study1,24. To determine the anti-viral activity of G7-Fc by different of 7F3 to RBD buries 976.4 Å2 surface area (Supplementary Table S3).
administration routes, we administered G7-Fc via intraperitoneal (i.p.) The interactions between 7F3 and RBD are mainly contributed by
route (200 μg/mouse) and intranasal (i.n.) route (20 μg/mouse) CDRH2, CDRH3, and CDRL3 (Fig. 5a). Residues Y449, Y453, L455, F456,
(Fig. 3a). Viral load in the lungs of K18-ACE2 transgenic mice was Y473, A475, G476, N477, K478, P479, A484, G485, S486, N487, C488,
detected 48 hours post-infection in the control group treated with Y489, S490, P491, L492, Q493, S494, Y495, G496, R498, Y501,
phosphate-buffered saline (PBS), with a value of 6.75 × Log10 focus- and H505 of RBD participate in the interactions with 7F3, forming 18
forming units (FFU)/g. Remarkably, no viruses were detected in any of pairs of hydrogen bonds (Fig. 5b, Supplementary Table S4). In addi-
the prophylactic or therapeutic groups regardless of administration tion, Y229 of CDRH3 packs against a hydrophobic pocket in the
routes (i.p. or i.n.) (Fig. 3b). Furthermore, there was no significant interface formed by L455, F456, Y473, and Y489 of RBD (Fig. 5b).
difference between the anti-viral activity of G7-Fc administered using Structure comparison reveals that the binding modes of XBB S/7F3 and
either i.p. or i.n. delivery routes (Fig. 3b). Similar findings were evident WT S/BD23 (PDB ID: 7BYR) are quite similar. Other than that, 7F3 was
in BALB/c mice, although a lower lung viral titer of 3.16 × Log10 FFU/g implicated in more interactions with RBD residues, and the E484A,
was recorded in the control PBS group (Fig. 3c). Our results demon- F486S, Q498R, N501Y, and Y505H mutations may interfere with the
strate that G7-Fc is highly efficacious in both the prevention and contacts between BD23 and SARS-CoV-2 variants (Supplemen-
treatment of XBB.1 infection, irrespective of administration route. tary Fig. S8).

Nature Communications | (2024)15:5127 5

Article https://doi.org/10.1038/s41467-024-49096-1

a G7-Fc G7-Fc

i.p.

-24 h 0h 24 h 48 h

Prevention (P) Treatment (T)

i.n.

XBB.1 titer

b
Log10(FFU/g)

K18-ACE2
mice P<0.0001 P<0.0001 P<0.0001 P<0.0001

2.18 Limit of detection

P T P T
PBS
Control i.p. i.n.

XBB.1 titer

c
Log10(FFU/g)

BALB/c
mice

P<0.0001 P<0.0001 P<0.0001 P<0.0001

Limit of detection

PBS P T P T
Control i.p. i.n.

Fig. 3 | In vivo prophylactic and therapeutic activity of the bispecific antibody Each data point represents an individual mouse within the respective groups. The
G7-Fc via different administration routes. a Schematic diagram of G7-Fc in the mean ± SEM of all data points are shown in (b, c). Statistical significance was ana-
prevention (abbreviation P) and treatment (abbreviation T) of XBB.1 infection. G7- lyzed using a two-sided t-test in (b, c) with Prism software (version 9, GraphPad
Fc was administrated intraperitoneally (abbreviation i.p.) or intranasally (abbre- Software), and p < 0.0001 is expressed as ***. Source data of b, c are provided in the
viation i.n.). b, c Viral titers (FFU, focus-forming units) of lung tissue of b K8-ACE2 Source Data file. The icon in a were created using Adobe Illustrator, and the
mice (n = 5) and c BALB/c mice (n = 4) were determined after 48 h post-infection. schematic diagram was created using Microsoft Office PowerPoint.

Nature Communications | (2024)15:5127 6

Article https://doi.org/10.1038/s41467-024-49096-1

Fig. 4 | Cryo-EM structures of the XBB S trimer in complex with the bispecific competition with bispecific antibody G7-Fc. A control IgG1 was used as a negative
antibody G7-Fc IgG. a G7-Fc binds to XBB S, forming a trimer dimer. Two per- control, and the addition of ACE2 was used as a positive control. e Close-up view of
pendicular views of XBB S/G7-Fc depict the surface of the trimer dimer and one of the interaction between G7-Fc and XBB S. The XBB S-RBD is displayed on a dark
the trimers, with GW01 in orange and 7F3 in dodger blue. b Binding of 7F3 to the gray surface. GW01 and 7F3 are shown as cartoons colored orange and dodger blue,
SARS-CoV-2 RBD in competition with CR3022, GW01, and BD23 was assessed using respectively. f Surface representation of RBD showing the buried binding site,
BLI. c Binding of GW01 to the SARS-CoV-2 RBD in competition with 7F3, BD23, and including GW01, 7F3, and the ACE2 binding site. Source data are provided as a
CR3022 was evaluated using BLI. d Binding of ACE2 to the RBDs of XBB trimer in Source Data file.

Nature Communications | (2024)15:5127 7

Article https://doi.org/10.1038/s41467-024-49096-1

Fig. 5 | The epitope of G7-Fc on XBB S. a The interactions between XBB S-RBD and e Conservation of the epitope of LY-CoV1404 according to the CoV-Spectrum of
7F3. b The detailed interactions between RBD and 7F3. The residues involved in GISAID database. f, g The detailed interactions between RBD and LY-CoV1404.
interactions are represented as sticks. c The detailed interactions between RBD and Source data are provided as a Source Data file.
GW01. d Mapping the epitope of G7-Fc by site-directed mutagenesis.

Nature Communications | (2024)15:5127 8

Article https://doi.org/10.1038/s41467-024-49096-1

The epitope of GW01 is located outside RBM. We previously the control antibody VRC01, an HIV-1 gp120 binding antibody, did not
identified the epitope of GW01 on BA.1 S-RBD25. GW01 interacts with affect the RBD/ACE2 interaction (Fig. 4d). Structural alignment of
BA.1 and XBB S-RBD in a very similar way. The contacts between the G7-Fc/RBD complex with the ACE2/RBD complex indicated that
GW01 and the XBB S-RBD are mainly contributed by CDRH3 by both 7F3 and GW01 were able to compete with ACE2 when binding to
forming hydrophobic interactions and 6 pairs of hydrogen bonds the RBD (Fig. 4f), which was consistent with the competition
(Supplementary Table S5). Y226, N233, Y234, and E235 of CDRH3 assay (Fig. 4d).
interact with XBB S-RBD through hydrogen bonds. Y226, V231, F232, These findings suggest that G7-Fc inhibits SARS-CoV-2 XBB variant
and Y234 of CDRH3 establish a large hydrophobic interface with infection by synergistically inducing the formation of trimer-dimers.
Y369, F374, F375, A376, F377, V407, A435, V503, and Y508 of RBD Both 7F3 and GW01 bind to conserved epitopes in XBB, enlarging the
(Fig. 5c). In addition, D155 of CDRH1 and D178 of CDRH2 are also interface area, thereby improving the affinity between the RBD and
involved in the interactions between GW01 and RBD (Fig. 5c). When single scFv and blocking the RBD from interacting with the ACE2
compared with BA.1 S-RBD/GW01 structure, fewer residues of XBB receptor. The structural arrangement of the bispecific antibody, par-
S-RBD are involved in the GW01 interaction, resulting in a reduction ticularly the Fc region, plays a critical role in facilitating G7-Fc binding
of interface area by 91.7 Å2 in the XBB S-RBD/GW01 structure (Sup- to the trimer.
plementary Table S3). This may explain why GW01 exhibits lower
binding affinities to XBB S-RBD than BA.1 S-RBD. A comparison of Discussion
XBB S/GW01 and WT S/CR3022 (PDB ID: 6W41) reveals that the The Omicron subvariants BQ.1.1, XBB.1.5, and XBB.1.16, were found to
CR3022 epitope, which primarily consists of L368-F392 residues, is be highly immune evasive variants of severe acute respiratory syn-
located near the bottom of the RBD. The S371F, S375F, T376A, and drome coronavirus 2 (SARS-CoV-2). All the antibodies targeting the
R408S mutations may cause the loss of binding affinity of CR3022 to RBM region, including potent antibody LY-CoV1404, failed to neu-
SARS-CoV-2 variants (Supplementary Fig. S9). tralize BQ.1.1, XBB.1.5, and XBB.1.16. In this study, we discovered an
To characterize the key epitope of G7-Fc, we analyzed the resi- RBM-specific antibody 7F3, which neutralized the highly immune-
dues involved in G7-Fc binding (Supplementary Fig. S8b, 9b) and evasive SARS-CoV-2 variants. G7-Fc, incorporating 7F3, showed broad
constructed eight XBB single mutants with crucial roles in G7-Fc neutralization of variants and sarbecoviruses, were reported and
binding. The K378A, Y473A, Y489A, Q493A, S494A, and Y501A exhibited potent prophylaxis and therapeutic efficacy against XBB.1
mutants dramatically decreased (>10 fold and <100 fold) or abol- infection of mice.
ished (>100 fold) G7-Fc neutralization (Fig. 5d). The K378 forms a Structural analysis of the XBB trimer bound by G7-Fc revealed that
hydrogen bond with GW01 (Fig. 5c), whereas the Y473, Y489, 7F3 interacts with 29 residues in the receptor-binding domain (RBD),
Q493,S494, and Y501 residues form hydrogen bonds with 7F3. The forming 18 hydrogen bonds and one hydrophobic interaction patch.
residues involved in the G7-Fc epitope exhibit a high degree of These 29 residues localize primarily to the RBM region, including
conservation, particularly with K378, Y473, Y489, Q493, Y449, L455-F456, Y473-K478, and A484-S494. To confirm the unique
S494, and Y501, showcasing over 95% conservation in variant binding capacity of 7F3, we constructed a series of RBD single mutants
sequences worldwide since Jan 2023 according to the CoV-Spectrum and triple mutants. Indeed, 7F3 demonstrated superior binding com-
of GISAID database (https://cov-spectrum.org/explore/World/ pared with LY-CoV1404, recognizing all RBD mutants, including suc-
AllSamples/AllTimes/variants?variantQuery=S%3A466R&) (Fig. 5d, cessive triple mutants in the RBM region (Fig. 1d). The 7F3 epitope is
Supplementary Table S6). The conserved binding epitope of 7F3 unique within RBD that distinguishes it from other characterized
conferred notable binding breadth to a panel of 37 RBD mutant antibodies. In contrast, the LY-CoV1404 epitope may focus on more
proteins with single or triple mutations (Fig. 1e). Despite the Y501A variable regions of the RBD that differ substantially between mutants.
mutation leading to escape from G7-Fc neutralization, an analysis of These findings suggest the notable breadth of G7-Fc can be attributed
global SARS-CoV-2 genomic databases since January 2023 revealed to the distinctive binding epitope and binding ability of 7F3.
only 6 reported sequences containing this 501A mutation. The The trimer dimer state of the spike induced by G7-Fc was
mutation rates of other residues in position Y501 worldwide are observed during the structural study of the XBB S/G7-Fc complex.
notably low (Supplementary Fig. S10b), indicating that escape Structural analysis of the XBB S/G7-Fc complex showed that GW01 only
mutations in position Y501 are uncommon. Importantly, this sug- binds to the up RBD and would clash with the adjacent down RBD
gests that G7-Fc can potently neutralize the vast majority of cur- (Supplementary Fig. S7a). Thus, 7F3 binding might induce RBD to the
rently circulating SARS-CoV-2 strains. Analysis of the structure of LY- up position, thus facilitating GW01 binding. This resembles the 2-step
CoV1404 complexed with WT RBD (PDB ID: 7MMO) showed that the models for antibody synergy25,30, where binding of the first antibody
epitope of LY-CoV1404 is of low conservation at R346, N440, K444- induces RBD into the up position and enhances the binding of the
G446, L452, and Q498 (Supplementary Fig. S10a and Fig. 5e), which second antibody. In addition, when the GW01 in G7-Fc was replaced
explains its decrease of binding affinity of SARS-CoV-2 variants. In with other SARS-CoV-2 mAbs, such as 4L12 and REGN10989, the trimer-
detail, the R346T, N440K, K444T, L452R, and Q498R mutations may dimer state of the spike was not induced by incubating 4L12-7F3-Fc or
result in the escape of BQ.1.1 (Fig. 5f). The V445P and G446S disrupt REGN10989-7F3-Fc with XBB S trimer, as observed in negative staining
the hydrophobic interface between V445 and L52, Y54, K60 of heavy EM images (Supplementary Fig. S4c). These results demonstrate that
chain and Y94 of light chain, which may contribute to the escape of 7F3 and GW01 work synergistically, inducing the crosslink of spike and
XBB, XBB.1.5, and XBB.1.16 (Fig. 5g). These results indicate that G7-Fc improving binding and neutralizing activity with the RBD.
bispecific antibody binds to two complementary epitopes across Taken together, our study discovered a specific RBM-binding
variants of the Omicron lineage of SARS-CoV-2, thereby explaining antibody 7F3, which neutralized the highly immune-evasive SARS-
the ability of G7-Fc to neutralize even the highly antigenically evasive CoV-2 variants BQ.1.1, XBB.1.5, and XBB.1.16. The conserved epitope
Omicron variants. within the RBM region bound by 7F3 may be critical for receptor
binding and viral entry, highlighting it as an attractive target site for
Neutralization mechanism of the bispecific antibody G7-Fc vaccine and drug development. Moreover, G7-Fc, incorporating 7F3,
To investigate the neutralization mechanism of the bispecific antibody exhibited excellent in vivo anti-XBB efficacy in both prevention and
G7-Fc, we performed a BLI competition assay and tested whether G7-Fc treatment therapy via different administration routes. G7-Fc may be
blocks RBD/ACE2 binding. G7-Fc prevented the RBD protein and the S used as a potential prophylactic countermeasure against SARS-
trimer of XBB binding to ACE2 protein in the competition assay, while CoV-2.

Nature Communications | (2024)15:5127 9

Article https://doi.org/10.1038/s41467-024-49096-1

Methods ELISA
Cell lines and participants The RBDs of SARS-CoV-2 variants spike protein (aa Arg319-Phe541)
HEK293T cells (ATCC, CRL-3216) and Huh-7 cells (National Collection with 8× histidine at the C-terminus was expressed by transfecting HEK
of Authenticated Cell Cultures, CSTR: 19375.09.3101HUMSCSP526) 293F cells. RBDs were coated at 1 μg/mL in Carb buffer (10 mM Na2CO3,
were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supple- 40 mM NaHCO3, pH 9.6) on a 96-well microtiter plate overnight at
mented with 10% fetal bovine serum (FBS), 1000 units/mL strepto- 4 °C. Plates were blocked with 5% non-fat powdered milk with 1% FBS in
mycin, and 1000 μg/mL penicillin at 37 °C in a 5% CO2 atmosphere. PBS. Antibodies were serially diluted in disruption buffer and trans-
HEK293F suspension cells (Thermo Fisher Scientific, R79007) were ferred into a PBST washed microtiter plate. After incubation for 1 h, the
cultured in serum-free SMM 293-TII Expression Medium (Sino Biolo- plate was washed, and goat-anti-human IgG antibodies conjugated
gical Inc.) at 37 °C in a 5% CO2 atmosphere with shaking at 120 rpm. with HRP (Jackson ImmunoResearch, Code No. 109-035-098, Lot.
Ethical approval of this study (YJ-2020-S021-01) was obtained from the 164726) were added (1:2500 dilution). Antibody binding was then
Ethics Committee of the Shanghai Public Health Clinical Center, developed using ABTS substrate solution, and absorbance was detec-
ensuring compliance with ethical standards for sample collection and ted at 405 nm. Absorbance Absorb data were collected by BioTek
antibody study. The participants involved in this study signed an Epoch 2.
informed consent form approved by the Investigational Review
Board (IRB). Antibody affinity by BLI
Binding affinities of antibodies with different RBD variants were per-
Isolation and characterization of SARS-CoV-2-specific mono- formed in Octet Ni-NTA (NTA) Biosensors (Sartorius). Recombinant
clonal antibodies RBD His-tag proteins were loaded on NTA Biosensors and binding
Peripheral blood mononuclear cells (PBMCs) from a 40-year-old kinetics were analyzed by following gradient dilution concentration of
COVID-19 patient, whose serum exhibited strong SARS-CoV-2 neu- Ab and monitoring association for 300 s. Then monitoring dissociation
tralizing activity, were collected on the date of discharge and stained while the sensor was immersed in the PBST buffer for 300 s. The
with CD19 (BD, Cat No. 341093, Clone No. SJ25C1), IgA (Jackson experiment was conducted on an Octet RED96 instrument (Sartorius)
ImmunoResearch Laboratories, Cat No. 109-135-011), IgD (BD, Cat No. and data was analyzed using a 1:2 binding model by FortéBio Data
555778, Clone No. IA6-2.), and IgM (Jackson ImmunoResearch Analysis 8.1 software.
Laboratories, Cat No. 709-116-073) antibodies. CD19+IgA−IgD−IgM−
memory B cells from COVID-19 patients were cultured in the presence ACE2 competition assay by BLI
of IL-21 (Invitrogen), IL-2 (Roche), and 3T3msCD40L feeder cells (NIH As for the ACE2 competition binding assay, 600 nM antibody was
reagent program) in the 384-well plates as previously described31. After incubated with 100 nM RBD-8his protein for 30 min earlier. Procedure
2 weeks, the culture supernatants were collected and screened for is similar to Ab competition assay. First, ACE2-Fc was captured by an
neutralization against SARS-CoV-2. Neutralizing activity was then AHC biosensor for 600 s. Second, after a baseline of 120 s in PBST
confirmed using RT-PCR for the immunoglobulin heavy chain (VH) and buffer, the sensor was blocked with IgG1 isotype control for 600 s.
light chain (VL) genes of those wells that exhibited neutralizing activ- Third, the sensor was soaked into the antibody-RBD pre-mixture for
ity. Finally, antibodies were expressed by HEK293F cells and purified 600 s after baseline again in the buffer. A mixture of ACE2-Fc, IgG
using protein G beads (Smart-Lifesciences) for further analysis. isotype, and RBD was as a positive control, and a mixture of isotype
mAb VRC01 and RBD was as a negative control. Percent competition
Production of pseudoviruses values were calculated based on the values of maximum binding. The
Pseudoviruses in this experiment were produced using a two-plasmid binding curve of ACE2-Fc, IgG isotype, and RBD was employed as a
system. The first plasmid pcDNA3.1-spike plasmid expressed the spike positive control and set as a 100% competition value, while the binding
gene, while the second plasmid pNL4-3.LucR−E− carrying the informa- curve of mAb VRC01 and RBD served as a negative control and was
tion necessary for single-cycle infectious pseudoviruses. Spike genes designated as a 0% value.
of VOCs and Omicron variants were synthesized (GenScript) and
inserted into the pcDNA3.1 vector. BA.2.75 and BA.5 subvariants plas- Antibody competition assay by BLI
mid were created by PCR using site-directed mutagenesis (Supple- Antibody competition binding assays were performed on the Octet
mentary Table S7). Primers were designed from the Quick-Change RED96 instrument as previously described14. Briefly, 100 nM RBD-8his
Primer design tool (Supplementary Table S8). Spike expression plas- protein and 600 nM second mAb were incubated for 30 mins in
mid and backbone plasmid were co-transfected into 70–80% confluent advance. The first mAb were captured on anti-human Fc (AHC) bio-
HEK293T cells using EZ Trans reagent (Life-iLab, China). After 28 h sensors at 20 μg/mL for 300 s, and then the biosensor was blocked by
transfection, the pseudoviruses were harvested, aliquoted, and stored immersing into isotype control IgG1 at 50 μg/mL. After blocking for
at −80 °C until use. 300 s, the biosensors were placed into a mixture of RBD-8his and
second mAb to detect the first mAb-binding RBD. The specific binding
Pseudovirus neutralization assay signal was normalized by subtracting the self-mAb control.
For the neutralization assay, 1 × 104 Huh-7 cells were plated to a 96-well
plate. The cells were cultured for 12–18 h at 37 °C in 5% CO2. Antibodies Construction expression and purification of Bispecific
were serially diluted and incubated with normalized pseudotyped antibodies
virus for 30 mins at 37 °C. After incubation, the supernatant of Huh-7 Bispecific antibodies G7-Fc and 7G-Fc utilizing the formats were con-
was removed, and then a 50 μL mixture of antibody and pseudovirus structed as previously described24. GW01 scFv and 7F3 scFv followed
was added to the 96-well plate. After infection for 24 h, 150 μL medium the order VL-(GGGGS)3-VH. The scFvs for each antibody were syn-
was supplied and cultured for another 48 h. The supernatant was thesized by Genscript company and connected with a (GGGGS)4 lin-
removed and Huh-7 cells were lysed with 50 μL lysis buffer for 30 mins. ker, followed by IgG1 Fc fragment (Hinge-CH2-CH3). Thus, the
For luciferase expression detection, 30 μL cell lysate was transferred sequence of G7-Fc is GW01VL-(GGGGS)3-GW01VH-(GGGGS)4-7F3VL-
into a new 96-well plate and 40 μL luciferase substrate (Promega) was (GGGGS)3-7F3VH-Hinge-CH2-CH3. While 7G-Fc was designed with
then added. The chemiluminescence signals were collected by Perki- 7F3 scFv in the N-terminal.
nElmer Ensight. The neutralization activity IC50 was calculated using HEK 293F cells (ATCC) were cultured in SMM 293-TII medium
GraphPad Prism9.0. (Sino Biological Inc.) and used to express antibodies. The Ab-encoding

Nature Communications | (2024)15:5127 10

Article https://doi.org/10.1038/s41467-024-49096-1

DNA plasmid was transiently transfected into HEK 293F cells using EZ on a Superose 6 increase 10/300 column (GE Healthcare). The peak
Trans (Life-iLab, China) with a 1:3 mass-to-volume ratio. After 3 days, fraction was concentrated to 0.5 mg/ml in 20 mM Tris, pH 8.0, and
cells were supplemented with fresh SMM 293-TII medium. On day five 200 mM NaCl for cryo-EM study.
post-transfection, the supernatants were collected, and the antibody
was affinity purified with protein G beads. The purity and molecular Negative staining EM sample preparation and imaging
weight of Ab were then future analyzed by SDS-PAGE. For XBB S, XBB S/G7-Fc, XBB S/G7-Fab, XBB S/7F3 IgG, XBB S/GW01
IgG, XBB S/7F3 IgG/GW01 IgG, XBB S/4L12-7F3-Fc or XBB S/
Neutralizing assay by focus reduction forming assays (FRNT) REGN10989-7F3-Fc, 5 µL samples (0.02 mg/mL) were loaded to the
FRNT was performed in a certified Biosafety level-3 lab, as previously glow-discharged carbon-coated grids (Electron Microscopy China) for
described32. In total, 50 μL of serially diluted antibodies were incu- 1 min. Then, the grids were blotted with filter paper and negatively
bated with 50 μL of BQ.1 or XBB authentic virus (180 focus forming stained with 2% (w/v) uranyl acetate, gently blotted, and air-dried.
unit, FFU) in 96-well microwell plates for 1 h at 37 °C. The mixtures Images were collected with Talos L120C TEM (Thermo Fisher Scien-
were added to 96-well plates seeded with Vero E6 cells (ATCC, Mana- tific) using a Ceta2 camera at a nominal magnification of 73,000×.
ssas, VA) and incubated for 1 h at 37 °C. Inoculums were then removed
before adding the overlay media (100 μL MEM containing 1.2% Car- Cryo-EM sample preparation
boxymethylcellulose, CMC). The plates were then incubated at 37 °C Three microlitre complex was applied on a freshly glow-discharged
for 24 h. Overlays were removed, and cells were fixed with 4% paraf- holey amorphous nickel-titanium alloy film supported by 400-mesh
ormaldehyde solution for 30 min. Cells were permeabilized with 0.2% gold grids, plunge frozen using the Vitrobot IV (FEI/Thermo Fisher
Triton X-100 and incubated with cross-reactive rabbit anti-SARS-CoV-N Scientific) with a blot force of −2 and 3.0 s blot time at 100% humid-
IgG (Sino Biological, Cat 40143-R001) for 1 h at room temperature (RT) ity and 4 °C.
before adding HRP-conjugated goat anti-rabbit IgG(H + L) antibody
(1:4000 dilution) (Jackson ImmunoResearch, West Grove, PA). Cells Cryo-EM data collection and image processing
were further incubated at RT. The reactions were developed with KPL Cryo-EM data were captured on a TITAN Krios G4 TEM (Thermo Fisher
TrueBlue Peroxidase substrates (Seracare Life Sciences Inc, Milford, Scientific) equipped with a Falcon 4i camera and a Selectris X Imaging
MA). An EliSpot reader (Cellular Technology Ltd., Shaker Heights, OH) filter (Thermo Fisher Scientific) setting to a slit width of 20 eV. Auto-
was used to calculate the numbers of SARS-CoV-2 foci. mated data collection was performed with EPU software at 300 kV in
AFIS mode at a nominal magnification of 105,000×, with a physical
Prophylactic and therapeutic efficacy of G7-Fc against XBB.1 pixel size of 1.19 Å and defocus values ranging from −1.0 μm to
infection of mice −3.0 μm. Each EER movie stack was dose-fractioned to 1737 frames
The animal study was approved by the Institutional Animal Care and with a total exposure dose of about 50 e−/Å2.
Use Committees of Affiliated First Hospital of Guangzhou Medical Movie stacks were imported to Relion3.1, then motion corrected
University (Approval Number: 20230615). 6-week-old ACE2 K-18 (binned 2 and dose weighted) by MotionCor233, and CTF estimated by
female mice (n = 5/group) and BALB/c female mice (n = 4/group) Gctf 34. All micrographs were imported to cryoSPARC35 v4.0.3 for par-
were used to evaluate the prophylactic and therapeutic efficacy of G7- ticle picking and 2D classification. A total of 1,650,604 good particles
Fc in the animal study. Mice were injected either intraperitoneally from 6332 selected micrographs were imported to Relion, and one
(200 μg/mouse) (i.p.) or intranasally (20 μg/mouse) (i.n.) with bispe- round of 3D classification resulted in 3.75 Å of trimer dimer. Local
cific antibody G7-Fc 24 h before or 24 h after intranasal infection with refinement is performed in cryoSPARC and results in 3.07 Å with a
105 FFU of XBB.1. Two days after XBB.1 infection, lung samples of mice mask around the top monomer and in 3.05 Å with a mask around the
were harvested for vial titration. The amount of virus per gram of lung bottom monomer. Then, the particles were expanded C3 symmetry in
tissue was determined after 48 h post-infection. Mice treated with PBS Relion, and further local 3D classification with a mask around RBD and
were used as controls. Fab was applied. The best class is exported into cryoSPARC, and local
refinement yields a resolution of 3.0 Å.
Expression and purification of RBD The resolution was estimated according to gold-standard
The coding sequence of RBDs (Arg319-Phe541) tagged with a Fourier shell correlation (FSC) 0.143 criterion. All the data processing
C-terminal 8×His tag was cloned into pSecTag mammalian expression procedures are summarized in Fig. S6. The maps were sharpened by
vector. The coding sequence of human ACE2(Met1-Ser740) was fused DeepEMhancer36, and handedness was corrected using UCSF
to IgG1 Fc and cloned into the pSecTag vector. The recombined RBDs Chimera37.
and hACE2-Fc were both expressed in HEK293F cells. The expression
plasmid was transiently transfected into cells. After 5 days, the medium Model building and refinement
was collected and filtered with a 0.45 μm filter. RBDs were purified by The initial model of XBB S was generated from Omicron S-
His affinity using Ni Sepharose while hACE2-Fc was purified by Protein FD01(PDB: 7WOW), while G7-Fc was predicted using swiss-model38.
G beads. Initial models were fitted into the maps using UCSF Chimera and then
manually adjusted using COOT39, and several rounds of real space
Expression and purification of SARS-CoV-2 XBB spike trimer refinement were performed using PHENIX40. The RBD domain bound
The XBB S HexaPro ectodomain was cloned into the pcDNA3.1 vector. with GW01 and 7F3 were refined was refined against the local refine-
The expression plasmid was transiently transfected into HEK293F cells ment map of RBD-G7-Fc and then docked back into the composite
using polyethyleneimine and cultivated for 72 h. XBB S trimers were map, and the whole model was refined against the composite map.
purified from supernatants by affinity column Histrap HP (GE Health- Model validation was performed using Molprobity. Figures were pre-
care). The protein was further purified by gel filtration chromato- pared using UCSF Chimera and UCSF ChimeraX41. The statistics of
graphy using a Superose 6 increase 10/300 column (GE Healthcare) in model refinement and data collection are listed in the Supplementary
20 mM Tris, pH 8.0, and 200 mM NaCl. Table S2.

Formation of XBB S/G7-Fc complex Reporting summary

The XBB S trimer was mixed with G7-Fc in a 1:1.3 molar ratio, incubated Further information on research design is available in the Nature
at 4 °C for 0.5 h, and further purified by gel filtration chromatography Portfolio Reporting Summary linked to this article.

Nature Communications | (2024)15:5127 11

Article https://doi.org/10.1038/s41467-024-49096-1

Data availability 21. Hong, Q. et al. Molecular basis of receptor binding and antibody
Coordinates and maps associated with data reported in this manu- neutralization of Omicron. Nature 604, 546–552 (2022).
script were deposited to the Electron Microscopy Data Bank (EMDB) 22. Liu, L. et al. Antibodies that neutralize all current SARS-CoV-2 var-
and Protein Data Bank (PDB) with accession numbers EMD-36423 and iants of concern by conformational locking. bioRxiv https://doi.org/
PDB 8JMM (XBB S trimer-dimer/G7-Fc), EMD-36321 and PDB 8JIN (local 10.1101/2023.04.08.536123 (2023).
refined map of XBB S/G7-Fc). Accession number of GW01 heavy chain 23. Zhou, P. et al. Broadly neutralizing anti-S2 antibodies protect
and light chain are OP480801.1 [https://www.ncbi.nlm.nih.gov/ against all three human betacoronaviruses that cause deadly dis-
nuccore/OP480801.1/] and OP480802.1 [https://www.ncbi.nlm.nih. ease. Immunity 56, 669–686.e667 (2023).
gov/nuccore/OP480802.1] in GenBank. Source data in this study are 24. Xiong, Q. et al. Close relatives of MERS-CoV in bats use ACE2 as
provided in the Supplementary Information/Source Data file. Source their functional receptors. Nature 612, 748–757 (2022).
data are provided with this paper. 25. Wang, Y. et al. Combating the SARS-CoV-2 Omicron (BA.1) and BA.2
with potent bispecific antibodies engineered from non-Omicron
References neutralizing antibodies. Cell Discov. 8, 104 (2022).
1. Qu, P. et al. Enhanced neutralization resistance of SARS-CoV-2 26. Vogel, L. What to know about Omicron XBB.1.5. Can. Med. Assoc. J.
Omicron subvariants BQ.1, BQ.1.1, BA.4.6, BF.7, and BA.2.75.2. Cell 195, E127–E128, (2023).
Host Microbe https://doi.org/10.1016/j.chom.2022.11.012 (2022). 27. Barnes, C. O. et al. SARS-CoV-2 neutralizing antibody structures
2. Planas, D. et al. Resistance of Omicron subvariants BA.2.75.2, BA.4.6 inform therapeutic strategies. Nature https://doi.org/10.1038/
and BQ.1.1 to neutralizing antibodies. bioRxiv https://doi.org/10. s41586-020-2852-1 (2020).
1101/2022.11.17.516888 (2022). 28. Cao, Y. et al. Potent neutralizing antibodies against SARS-CoV-2
3. Cao, Y. et al. BA.2.12.1, BA.4 and BA.5 escape antibodies elicited by identified by high-throughput single-cell sequencing of con-
Omicron infection. Nature 608, 593–602 (2022). valescent patients’ B cells. Cell 182, 73–84.e16 (2020).
4. Imai, M. et al. Efficacy of antiviral agents against Omicron sub- 29. Yuan, M. et al. A highly conserved cryptic epitope in the receptor
variants BQ.1.1 and XBB. N. Engl. J. Med. https://doi.org/10.1056/ binding domains of SARS-CoV-2 and SARS-CoV. Science 368,
NEJMc2214302 (2022). 630–633 (2020).
5. Kurhade, C. et al. Low neutralization of SARS-CoV-2 Omicron 30. Zhou, T. et al. Structural basis for potent antibody neutralization of
BA.2.75.2, BQ.1.1, and XBB.1 by parental mRNA vaccine or a BA.5- SARS-CoV-2 variants including B.1.1.529. Science 376,
bivalent booster. Nat. Med. https://doi.org/10.1038/s41591-022- eabn8897 (2022).
02162-x (2022). 31. Huang, J. et al. Isolation of human monoclonal antibodies from
6. WHO. Statement on the update of WHO’s working definitions and peripheral blood B cells. Nat. Protoc. 8, 1907–1915 (2013).
tracking system for SARS-CoV-2 variants of concern and variants of 32. Wang, Y. et al. Biparatopic antibody BA7208/7125 effectively neu-
interest. https://www.who.int/news/item/16-03-2023. tralizes SARS-CoV-2 variants including Omicron BA.1-BA.5. Cell
7. Chen, C. et al. CoV-Spectrum: analysis of globally shared SARS- Discov. 9, 3 (2023).
CoV-2 data to identify and characterize new variants. Bioinformatics 33. Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-
38, 1735–1737 (2021). induced motion for improved cryo-electron microscopy. Nat.
8. WHO. Tracking SARS-CoV-2 variants. https://www.who.int/en/ Methods 14, 331–332 (2017).
activities/tracking-SARS-CoV-2-variants/. 34. Rohou, A. & Grigorieff, N. CTFFIND4: fast and accurate defocus
9. WHO. Weekly epidemiological update on COVID-19 - 20 April 2023. estimation from electron micrographs. J. Struct. Biol. 192,
https://www.who.int/publications/m/item/weekly- 216–221 (2015).
epidemiological-update-on-covid-19---20-april-2023. 35. Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoS-
10. Cao, Y. et al. Imprinted SARS-CoV-2 humoral immunity induces PARC: algorithms for rapid unsupervised cryo-EM structure deter-
convergent Omicron RBD evolution. Nature 614, 521–529 (2023). mination. Nat. Methods 14, 290–296 (2017).
11. Wang, Q. et al. Alarming antibody evasion properties of rising SARS- 36. Sanchez-Garcia, R. et al. DeepEMhancer: a deep learning solution
CoV-2 BQ and XBB subvariants. Cell https://doi.org/10.1016/j.cell. for cryo-EM volume post-processing. Commun. Biol. 4, 874 (2021).
2022.12.018 (2022). 37. Pettersen, E. F. et al. UCSF chimera—a visualization system for
12. Pinto, D. et al. Cross-neutralization of SARS-CoV-2 by a human exploratory research and analysis. J. Comput. Chem. 25,
monoclonal SARS-CoV antibody. Nature 583, 290–295 (2020). 1605–1612 (2004).
13. Shi, R. et al. A human neutralizing antibody targets the receptor- 38. Waterhouse, A. et al. SWISS-MODEL: homology modelling of pro-
binding site of SARS-CoV-2. Nature 584, 120–124 (2020). tein structures and complexes. Nucleic Acids Res. 46,
14. Hansen, J. et al. Studies in humanized mice and convalescent W296–W303 (2018).
humans yield a SARS-CoV-2 antibody cocktail. Science 369, 39. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and
1010–1014 (2020). development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66,
15. Rappazzo, C. G. et al. Broad and potent activity against SARS-like 486–501 (2010).
viruses by an engineered human monoclonal antibody. Science 40. Adams, P. D. et al. PHENIX: a comprehensive Python-based system
371, 823–829 (2021). for macromolecular structure solution. Acta Crystallogr. D Biol.
16. Liu, L. et al. Potent neutralizing antibodies against multiple epitopes Crystallogr. 66, 213–221 (2010).
on SARS-CoV-2 spike. Nature 584, 450–456 (2020). 41. Pettersen, E. F. et al. UCSF ChimeraX: structure visualization for
17. Gruell, H. et al. Antibody-mediated neutralization of SARS-CoV-2. researchers, educators, and developers. Protein Sci. 30,
Immunity 55, 925–944 (2022). 70–82 (2021).
18. Yamin, R. et al. Fc-engineered antibody therapeutics with improved
anti-SARS-CoV-2 efficacy. Nature 599, 465–470 (2021). Acknowledgements
19. Winkler, E. S. et al. Human neutralizing antibodies against SARS- We thank the Center of Cryo-Electron Microscopy, Core Facility of
CoV-2 require intact Fc effector functions for optimal therapeutic Shanghai Medical College, Fudan University, for the support on cryo-EM
protection. Cell 184, 1804–1820.e1816 (2021). data collection. This work was supported by R&D Program of Guangzhou
20. Chen, Y. et al. Broadly neutralizing antibodies to SARS-CoV-2 and Laboratory (SRPG22-003 to L.S.), Shanghai Municipal Science and
other human coronaviruses. Nat. Rev. Immunol. 23, 189–199 (2023). Technology Major Project (ZD2021CY001 to J.H.), the Ministry of Science

Nature Communications | (2024)15:5127 12

Article https://doi.org/10.1038/s41467-024-49096-1

and Technology of China (2021YFC2302500 to L.S.), the National Nat- Correspondence and requests for materials should be addressed to
ural Science Foundation of China (32370995 and 31771008 to J.H. and Jincun Zhao, Fan Wu, Jinghe Huang or Lei Sun.
3197014 to Z.C.), National Key Research and Development Program of
China (2023YFE0118200 to F.W.), the “Shuguang Program” of Shanghai Peer review information Nature Communications thanks Daniela Fera,
Education Development Foundation and Shanghai Municipal Education Nancy Haigwood, and the other, anonymous, reviewer for their con-
Commission (22SG16 to F.W.), the National Science and Technology tribution to the peer review of this work. A peer review file is available.
Major Projects of China (2017ZX10202102 to J.H. and 2018ZX10301403
to F.W.), Shanghai Municipal Commission of Health and Family Planning Reprints and permissions information is available at
(2018BR08 to J.H.). http://www.nature.com/reprints

Author contributions Publisher’s note Springer Nature remains neutral with regard to jur-
J.H., L.S., and F.W. conceived and designed the experiments. J.H. and isdictional claims in published maps and institutional affiliations.
F.W. performed B cell sorting and antibody cloning. Y.D.W., J.P., Y.M.,
P.R., J.C., and L.L. constructed the bispecific antibodies and SARS-CoV-2 Open Access This article is licensed under a Creative Commons
pseudovirus mutants, purified antibodies, and performed neutralization Attribution 4.0 International License, which permits use, sharing,
assay, ELISA, and bilayer interferometry experiments. L.S., Z.C., A.H., adaptation, distribution and reproduction in any medium or format, as
Q.M., and Y.J.W. were responsible for the structural studies. J.Z., Z.Z., long as you give appropriate credit to the original author(s) and the
Y.Q.W., and X.X. were responsible for the authentic virus experiments. source, provide a link to the Creative Commons licence, and indicate if
Y.D.W. and J.H. created the experiment schematic diagram. J.H., Y.D.W., changes were made. The images or other third party material in this
Y.M., F.W., L.S., A.H., and Z.Z. analyzed the data. Y.M.W. supervised the article are included in the article’s Creative Commons licence, unless
project. J.H., L.S., F.W., Y.D.W., and A.H. wrote the paper. indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons licence and your intended
Competing interests use is not permitted by statutory regulation or exceeds the permitted
Patents about the G7-Fc and 7G-Fc bispecific antibodies in this study are use, you will need to obtain permission directly from the copyright
pending. The remaining authors declare no competing interests. holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/.
Additional information
Supplementary information The online version contains © The Author(s) 2024
supplementary material available at
https://doi.org/10.1038/s41467-024-49096-1.

Nature Communications | (2024)15:5127 13