CORONAVIRUS

A neutralizing human antibody binds to theN-terminal domain of the Spike protein of SARS-CoV-2Xiangyang Chi1*, Renhong Yan2*, Jun Zhang1*, Guanying Zhang1, Yuanyuan Zhang2, Meng Hao1,Zhe Zhang1, Pengfei Fan1, Yunzhu Dong1, Yilong Yang1, Zhengshan Chen1, Yingying Guo2,Jinlong Zhang1, Yaning Li3, Xiaohong Song1, Yi Chen1, Lu Xia2, Ling Fu1, Lihua Hou1, Junjie Xu1,Changming Yu1, Jianmin Li1†, Qiang Zhou2†, Wei Chen1†

Developing therapeutics against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) could beguided by the distribution of epitopes, not only on the receptor binding domain (RBD) of the Spike (S) proteinbut also across the full Spike (S) protein. We isolated and characterized monoclonal antibodies (mAbs)from 10 convalescent COVID-19 patients. Three mAbs showed neutralizing activities against authenticSARS-CoV-2. One mAb, named 4A8, exhibits high neutralization potency against both authentic andpseudotyped SARS-CoV-2 but does not bind the RBD. We defined the epitope of 4A8 as the N-terminaldomain (NTD) of the S protein by determining with cryo–eletronmicroscopy its structure in complex with theS protein to an overall resolution of 3.1 angstroms and local resolution of 3.3 angstroms for the 4A8-NTDinterface. This points to the NTD as a promising target for therapeutic mAbs against COVID-19.

The global outbreak of COVID-19 hasemerged as a severe threat to humanhealth (1–3). COVID-19 is caused by anovel coronavirus, the severe acute re-spiratory syndrome coronavirus 2 (SARS-

CoV-2), which is an enveloped, positive-strandRNA virus that causes symptoms such as cough,headache, dyspnea, myalgia, fever, and severepneumonia in humans (1, 3–5).SARS-CoV-2 is a member of the b corona-

virus genus,which also contains SARS-CoV andMERS-CoV, which caused epidemics in 2002and2012, respectively (6, 7). SARS-CoV-2 sharesabout 80% sequence identity to SARS-CoV anduses the same cellular receptor, angiotensin-converting enzyme 2 (ACE2) (8–16).The trimeric S protein decorates the sur-

face of coronavirus and plays a pivotal roleduring viral entry (17, 18). During infection,the S protein is cleaved into the N-terminalS1 subunit and C-terminal S2 subunit by hostproteases such as TMPRSS2 (18, 19) andchanges conformation from the prefusion tothe postfusion state (20). S1 and S2 comprisethe extracellular domain (ECD; 1 to 1208 aminoacids) and a single transmembrane helix andmediate receptor binding and membrane fu-sion, respectively (16). S1, which consists of theN-terminal domain (NTD) and the receptorbinding domain (RBD), is critical in determin-ing tissue tropism and host ranges (21, 22). TheRBD is responsible for binding to ACE2, where-

as the function of NTD is not well understood.In some coronaviruses, the NTD may recog-nize specific sugar moieties upon initial attach-ment and might play an important role inthe prefusion-to-postfusion transition of theS protein (23–26). The NTD of the MERS-CoVS protein can serve as a critical epitope for neu-tralizing antibodies (26).The SARS-CoV-2 S protein–targeting mono-

clonal antibodies (mAbs) with potent neutral-izing activity are a focus in the development oftherapeutic interventions for COVID-19 (27–29).Many studies reported the functions and struc-tures of SARS-CoV-2–neutralizing antibodiesthat target the RBD and inhibit the associa-tion between the S protein and ACE2 (28–34).The RBD-targeting antibodies, applied indi-vidually, might induce resistance mutationsin the virus (26). Antibodies that target non-RBD epitopesmight be added to antibody cock-tail therapeutics for SARS-CoV-2.We thus soughtto identify antibodies to different regions of theS protein and to the Nucleocapsid (N) protein.

ResultsIsolation of human mAbs from memory B cellsand plasma B cells

To isolate mAbs and analyze the humoralantibody responses to SARS-CoV-2, we col-lected plasma and peripheral blood mono-nuclear cells (PBMCs) from 10 Chinese patientswho had recovered fromSARS-CoV-2 infection.The age of donors ranges from 25 to 53 years.The interval from disease confirmation date toblood collection date ranged from23 to 29 daysfor patients 1 to 5 and 10 to 15 days for patients6 to 10 (table S1). We evaluated the titers ofbinding antibodies in plasma to differentfragments of the SARS-CoV-2 S protein—including the full ECD, S1, S2, and the RBD—and to the N protein. Plasma from all thepatients except donor 2 bound to all five

SARS-CoV-2 protein segments, whereas thatfrom donor 2 recognized S-ECD and S2 only(Fig. 1A). The neutralizing capacities of plasmaagainst authentic SARS-CoV-2andHIV-vectoredpseudotyped SARS-CoV-2 are correlated [cor-relation coefficient (r) = 0.6868, P < 0.05](Fig. 1B). These results indicate that humoralimmune responses were specifically elicitedfor all 10 patients during their natural infec-tion with SARS-CoV-2.To isolate S protein–specific mAbs, we first

sorted the immunoglobulin G–positive (IgG+)memory B cells from PBMCs of convalescentpatients 1 to 5 with flow cytometry, using S-ECDas the probe (Fig. 1C). The percentage of S-ECD–reactive IgG+ B cells ranges from 0.56 to 11%,as revealed with fluorescence activating cellsorting (FACS). To avoid losing B cells withlow copies of S-ECD–specific receptors on cellsurfaces, we sorted plasma B cells frommixedPBMCs derived from another five convalescentpatients (patients 6 to 10) without using S-ECDprotein as the probe in flow cytometry. The per-centage of plasma B cells in CD3-CD19+ B cellswas 12.8%, which is higher than the percentageofmemory B cells in CD3-CD19+ B cells (Fig. 1C).From the sorted B cells, we identified 9,

286, 43, 12, and 26 clones of single B cell frompatients 1 to 5, respectively, and 23 clones ofsingle B cell from the mixed PBMCs of patients 6to 10 (Fig. 1D). The distribution of the se-quenced heavy (IgH) gene families was com-parable among the 10 donors, with VH3 beingthe most commonly used VH gene, whereasdifferent donors displayed variable preferen-ces for the light chain (IgL) gene families (Fig.1D). The combination of V3 and J4, V3 andD3,and D3 and J4 were the most common usagefor the IgH gene family (fig. S1). The averagemutations of amino acids permAb frommem-ory B cells ranged from 17.50 to 48.04 for do-nors 1 to 5, respectively, whereas mAbs fromplasma B cells possessed an average of 13.99amino acidmutations for donors 6 to 10 (Fig. 1E).Human antibodies elicited through repeatedexposures to different antigens confer an av-erage of 26.46 amino acid mutations per Ab, aspreviously reported (35). These results indicatethat natural SARS-CoV-2 infection elicited highlevels of somatic hypermutation (SHM) inmem-ory B cells. The lengths of complementarity-determining region 3 (CDR3) for antibodieswere similar among the donors, with averagelengths of these CDR3 ranging from 13.9 to17.7 for VH and 9.3 to 10.1 for VL (Fig. 1F). TheCDR3 lengths of these mAbs were longer thanthat inantigen-specific immunereceptors (meansof 12.7 for VH and 6.5 for VL, respectively) re-ported previously (36).

Binding profiles of SARS-CoV-2 S protein–specific human mAbs

To screen for S protein–specific antibodies, wedetermined the binding specificity using enzyme-

RESEARCH

Chi et al., Science 369, 650–655 (2020) 7 August 2020 1 of 6

1Beijing Institute of Biotechnology, Academy of MilitaryMedical Sciences (AMMS), Beijing 100071, China. 2KeyLaboratory of Structural Biology of Zhejiang Province,Institute of Biology, Westlake Institute for Advanced Study,School of Life Sciences, Westlake University, Hangzhou310024, Zhejiang Province, China. 3Beijing AdvancedInnovation Center for Structural Biology, Tsinghua-PekingJoint Center for Life Sciences, School of Life Sciences,Tsinghua University, Beijing 100084, China.*These authors contributed equally to this work.†Corresponding author. Email: cw0226@foxmail.com (W.C.);zhouqiang@westlake.edu.cn (Q.Z.); lijmqz@126.com (J.L.)

on Novem

ber 1, 2020

http://science.sciencemag.org/

Dow

nloaded from

linked immunosorbent assay (ELISA) for the399 human mAbs sorted above. From donors 1to 5, respectively, 1, 16, 1, 3, and 9 S-ECD–specificmAbs were identified. A total of 35 S-ECD–specific mAbs were identified from donors 6 to10 (Fig. 2A). We further characterized domainspecificities of the 35 mAbs with different frag-ments of the S protein, including S1, S2, andRBD (Fig. 2A). The S-reactive mAbs are classi-fied into fourmajor groups on the basis of theirmedium effective concentration (EC50) values(Fig. 2A).Group 1 recognizes onlyS-ECD. Group 2recognizes S-ECD and S1, with subgroup 2AbindingS-ECD and S1 and subgroup 2B bindingS-ECD, S1, andRBD. Group 3 interacts with bothS1 and S2, where subgroup 3A targets the RBDand subgroup 3B fails to bind the RBD. Group 4recognizes S-ECD and S2. Only four mAbsrecognize the RBD among the 35 S-specificmAbs (Fig. 2, A and B).We performed a competition-binding assay

using ELISA for several representative mAbsto determine whether there are overlappingantigenic sites between different mAbs, withCR3022 being used as a positive control mAbthat reported to bind the SARS-CoV-2 RBD(Fig. 2C) (37). Among these mAbs, 4A8 ingroup 2A competed with 1M-1D2 in group2B. Another RBD-reactive mAb, 2M-10B11in group 2B, competed with CR3022, suggest-ing overlapped epitopes on RBD for thesetwo mAbs. These results indicate that anti-body responses elicited by natural SARS-CoV-2infection were diverse in epitope recognitionof S proteins.To characterize the diversity in gene usage

and affinitymaturation, the phylogenetic treesof these S-ECD–specific mAbs were analyzedon the basis of the amino acid sequences ofVHDJH and VLJL by using a neighbor-joiningmethod in MEGA7 Software (38). Results in-dicate that the VH gene usage is very diverseamong the 35 mAbs from 10 donors, withVH3-30 being the most frequently used germ-line gene. There was no particularly favoredVH gene identified among S1, S2, or RBD-reactive mAbs (Fig. 2D). The percentages ofheavy chain variable gene sequence identityranged from 40.9 to 97.6% in the 35 S-ECD–specific mAbs (fig. S2 and table S2).

Neutralizing activities of SARS-CoV-2 S–specifichuman mAbs

We first performed in vitro neutralizationstudies of the 35 S-ECD–specific mAbs usingauthentic SARS-CoV-2 in Vero-E6 cells (Fig. 3A).Of the 35 S-ECD–specific mAbs, only threemAbs neutralized authentic SARS-CoV-2. MAb1M-1D2, 4A8, and 0304-3H3 exhibitedmediumto high neutralizing capacity with EC50 of28, 0.61, and 0.04 mg/ml, respectively. As ex-pected, the RBD-targeting controlmAb, CR3022,failed to neutralize authentic SARS-CoV-2 (37).Moreover, although the CR3022-competing

mAb, 2M-10B11, bound to the SARS-CoV-2RBD with an EC50 of 5 ng/ml (Fig. 2A), it alsofailed to neutralize authentic SARS-CoV-2.These results suggest that binding affinities

of mAbs against RBD do not correlate fullywith the neutralizing abilities of mAbs. Tofurther investigate the inhibitory activity ofthe three authentic SARS-CoV-2–neutralizing

Chi et al., Science 369, 650–655 (2020) 7 August 2020 2 of 6

SA

RS

-CoV

-2 S

-EC

DC

D27

Don

or 5

Don

or 6

/7/8

/9/1

0D

onor

4D

onor

3D

onor

2D

onor

1

IgG B cells

CD38 B cells

12.8%

9.11%

11%

1.41%

0.56%

1.56%

DC FE

VHNt

VHaa

Vκ/λ

Nt

Vκ/λ

aa VH Vκ/λ

No.

ofm

utat

ions

from

ger

mlin

e

CD

R3

leng

th(a

a)

0

50

100

150

200

0

10

20

30

40

0

10

20

30

40

0

20

40

60

80

0

20

40

60

80

0

10

20

30

40

0

20

40

60

80

0

10

20

30

40

0

50

100

150

200

0

10

20

30

40

0

50

100

150

200

0

10

20

30

40

9 9

V5

V1V2V3V4

V6V7V8

+

+

+

286

H

H

H

H

H

H

287

44 43

13 12

2627

32 23

κ

κ

κ

κ

κ

κ

λ

λ

λ

λ

λ

λ

r = 0.6868P = 0.0283

SA

RS

-Co

V-2

bin

din

g t

iter

Donor ID

A B

0 500 1000 1500 2000 25000

500

1000

1500

Pseudotyped SARS-CoV-2NAb titer (EC50)

SA

RS

-CoV

-2N

Ab

tite

r(E

C50)

1

345678910

Plasma PlasmaDonor ID

2S-ECDS1S2RBDN

1 2 3 4 5 6 7 8 9 10102

103

104

105

Fig. 1. Isolation of antigen-specific mAbs from convalescent patients of SARS-CoV-2. (A) Reactions ofplasma to SARS-CoV-2 proteins. S-ECD (extracellular domain of S protein), S1, S2, RBD (receptor bindingdomain), and N protein were used in ELISA to test the binding of plasma. Plasma of heathy donors wereused as control, and cut-off values were calculated as optical density (OD) 450 of control × 2.1. Data wereshown with mean and SD of a representative experiment. (B) The correlations between the authenticSARS-CoV-2 neutralizing antibody (NAb) titers and the pseudotyped SARS-CoV-2 NAb titers in plasma.Neutralizing assays of plasma against authentic SARS-CoV-2 were performed by using Vero E6 cells, andneutralization against pseudotyped SARS-CoV-2 were determined by using ACE2-293T cells. The correlationswere calculated by means of Pearson correlation test in Graphpad 7.0. (C) Flow cytometry sorting fromPBMCs of 10 convalescent patients. (D) Distribution of V gene families in heavy and light chains of alldistinct clones (the total number is shown in the center of the pie charts) for each donor. (E) The number ofamino acid (AA) and total nucleotide (Nt) mutations from the germline of all clonal sequences identifiedin (D) is shown. (F) CDR3 amino acid lengths of VH and VL of all clonal sequences identified in (D).

RESEARCH | RESEARCH ARTICLEon N

ovember 1, 2020

http://science.sciencem

ag.org/D

ownloaded from

mAbs—4A8, 0304-3H3, and 1M-1D2—we testedtheRNA load of authentic SARS-CoV-2 in Vero-E6 cells treated with each mAb using real-timequantitative polymerase chain reaction (PCR)(Fig. 3B). Consistent with the cytopathic effect(CPE) assay results (Fig. 3A), mAbs 0304-3H3and 4A8 displayed higher inhibitory capacitiesthan did 1M-1D2 (Fig. 3B).We next performed luciferase reporter gene

assays for all 35 S-binding mAbs using HIV-vectored pseudotyped SARS-CoV-2 (39), amongwhich three mAbs exhibited neutralizing ac-

tivity against the pseudotyped virus (Fig. 3C).4A8 protected ACE2-293T cells with an EC50of 49 mg/ml. Although mAb 2M-10B11 and 9A1did not neutralize authentic SARS-CoV-2, 2M-10B11 protected against pseudotyped viruswithan EC50 of 170 mg/ml, and 9A1 provided weakprotection. To our surprise, neutralization by0304-3H3 and 1M-1D2 was not observed (Fig.3C). The inconsistency between the resultsfor pseudotyped SARS-CoV-2 compared withauthentic SARS-CoV-2 were also observed formAbs against MERS-CoV (40, 41) and may be

caused by the different presentation of S pro-tein resulted from the different environmentalfactors the viruses underwent, such as the cellsused for the neutralizing assays or for the pro-duction of the pseudotyped or authentic virions(42). On the basis of these results, 4A8 is a po-tential candidate for the treatment of SARS-CoV-2 because it displayed strong neutralizingcapacities against both authentic and pseudo-typed SARS-CoV-2.

Binding characterization of candidate mAbs

To determine the possible neutralizing mech-anism of themAbs, we determined the bindingaffinities of the fivemAbswith potential neutral-izing activity against different segments of the Sprotein—including the full S-ECD and domainsS1, S2, and RBD—using biolayer interferometry(BLI). All five testedmAbs bound to S-ECDwithhigh affinity; equilibrium dissociation constants(Kd) were less than 2.14 nM (Fig. 4A). 4A8 and1M-1D2 bound to S1 withKd of 92.7 and 108 nM,respectively, whereas 0304-3H3 and9A1 targetedS2 with Kd of 4.52 and <0.001 nM, respectively

Chi et al., Science 369, 650–655 (2020) 7 August 2020 3 of 6

0

1

2

3

4 9A1

O.D

450

-630

nm

1

2

3

44A8

1

2

3

41M-1D2

mAb concentration (µg/mL)

0

1

2

3

4 2M-10B11

0

1

2

3

4 0304-3H3

S-ECDS1S2RBD

0

1

2

3

4 CR3022

2M-1

3A3

2M-1

4B2

2M-1

4E5

2M-8

H10

0317

-A7

0317-A1

0317-C9

2M-9H1

9A1

2M-9F10

2M-12D70317-A3

2M-10B112M-2D41M-1D2

0304-4A102M

-2D

103

04-3

H3

0304

-4A

2

2M-1

4E4

2M-1

3D11

2M-2

G12

0304

-2F80317-B

10317-A20317-A8

2M-7E9

2M-8E7

2M-4G4

0317-C4

0317-A9

4A8

8D98D

210C

10VH

3-7

VDJ AA

VH3-30

VH3-9

VH3-48VH3-66VH3-23

VH

3-64

VH

4-39VH

4-39

VH

4-59

VH

4-61

VH

4-34

VH

3-11VH5-

51VH7-4

VH1-69

VH1-46

VH1-24

0

1

2

3

4 0304-4A10

0

1

2

3

4 2M-14B2

10110010-110-210-310-4 10110010-110-210-310-4

Group 2A 2B 3B 4

mAb0304-4A2 4A8

1M-1D2

2M-10B11

CR3022

0304-4A10

2M-14E4 9A1

2M-13D11

2M-13A3

2M-14E5

2A

0304-4A2 3 127 111 108 104 106 135 114 110 135 128

4A8 90 4 6 107 102 103 147 111 96 132 130

2B

1M-1D2 88 93 5 107 98 102 128 106 76 119 128

2M-10B11 84 95 128 6 13 92 123 103 75 110 99

CR3022 98 131 115 57 8 104 128 111 116 230 167

3B 0304-4A10 94 110 111 102 106 7 16 112 96 224 179

4

2M-14E4 76 102 119 101 101 91 12 116 86 165 107

9A1 87 90 119 102 98 102 116 6 6 117 120

2M-13D11 83 112 110 104 104 104 179 105 6 57 85

2M-13A3 104 112 114 105 102 104 256 109 62 13 28

2M-14E5 79 103 110 101 98 105 134 113 84 14 7

Detecting Antibody

Blo

ckin

g A

ntib

ody

ELISA binding (EC50 : ng/mL)A

C D

BGroup Donor mAb S-ECD S1 RBD S2

3 0304-2F8 1883 > > >5 0317-A3 3025 > > >5 0317-A9 5851 > > >5 0317-B1 9470 > > >5 0317-C4 7624 > > >

6/7/8/9/10 10C10 78 > > >4 0304-4A2 5 3 > >5 0317-A7 9 6 > >5 0317-A8 1702 1812 > >

6/7/8/9/10 4A8 5 8 > >1 1M-1D2 17 125 519 >2 2M-10B11 8 4 5 >2 2M-4G4 9102 > 164 >

3A 2 2M-14B2 1479 1700 1075 2102 2M-2D4 2138 9250 > 7492 2M-2G12 5715 7973 > 48382 2M-7E9 183 7899 > 862 2M-8E7 189 9783 > 704 0304-3H3 5 2094 > 34 0304-4A10 3 3875 > 35 0317-A1 119 4697 > 1392 2M-2D1 7177 > > 12332 2M-8H10 37 > > 232 2M-9F10 2959 > > 8632 2M-9H1 77 > > 192 2M-12D7 42 > > 132 2M-13A3 107 > > 242 2M-13D11 137 > > 92 2M-14E4 33 > > 92 2M-14E5 11 > > 75 0317-A2 1460 > > 20805 0317-C9 7965 > > 6429

6/7/8/9/10 8D2 49 > > 54896/7/8/9/10 8D9 2678 > > 6426/7/8/9/10 9A1 16 > > 5

1

2A

2B

3B

4

Fig. 2. Binding profiles of Spike protein–specific mAbs. (A) Heatmap showing the binding of mAbs todifferent types of spike proteins determined by using ELISA. The EC50 value for each S-mAb combination isshown, with dark red, orange, yellow, or white shading indicating high, intermediate, low, or no detectablebinding, respectively. EC50 values greater than 10,000 ng/ml are indicated (>). (B) Binding curves ofrepresentative mAbs. CR3022 is a control that was reported to bind SARS-CoV and SARS-CoV-2 RBD.Data were shown with mean and SD of a representative experiment. (C) Heatmap showing the competingbinding of some representative S-reactive mAbs assayed in ELISA. Numbers in the box indicate thepercentage binding of detecting mAb in the presence of the blocking antibody compared with the bindingof detecting mAb in the absence of the blocking antibody. The mAbs were considered competing if theinhibiting percentage is <30% (black boxes with white numbers). The mAbs were judged to noncompete forthe same site if the percentage is >70% (white boxes with red numbers). Gray boxes with black numbersindicate an intermediate phenotype (30 to ~70%). (D) Phylogenetic trees of all the S-specific mAbs.

Pseudotyped SARS-CoV-2

Authentic SARS-CoV-2A

B

C

Authentic SARS-CoV-2

020406080

100120140

mAb concentration (µg/mL)

Infe

ctio

n (%

)

4A8

1M-1D20304-3H3

9A12M-10B11

CR3022

4A8

1M-1D20304-3H3

9A12M-10B11

CR3022

4A8

1M-1D20304-3H3

Neu

traliz

atio

n (%

)

EC50

(µg/mL)0.61

280.04

IC50

(µg/mL)

EC50

(µg/mL)

0.39

49

-170

-

--

250.11

---

310210110010-110-210

310210110010-110-210

-310-410mAb concentration(µg/mL)

Neu

traliz

atio

n (%

)

mAb concentration (µg/mL)10-1 100 101 102 103

-40-20

020406080

100

-200

20406080

100120

Fig. 3. Neutralizing capacities of S-reactivemAbs. (A) Neutralization of S-reactive mAbs toauthentic SARS-CoV-2 in Vero-E6 cells. (B) Theauthentic SARS-CoV-2 virus RNA load wasdetermined in Vero-E6 cells treated with S-reactivemAbs by using quantitative PCR. Percent infectionwas calculated as the ratio of RNA load in mAb-treated wells to that in wells containing virusonly. (C) Neutralization of S-reactive mAbs againstHIV-vectored pseudotyped SARS-CoV-2 in ACE2-293T cells. Data were shown as mean ± SD of arepresentative experiment.

RESEARCH | RESEARCH ARTICLEon N

ovember 1, 2020

http://science.sciencem

ag.org/D

ownloaded from

(Fig. 4A, bottom). Moreover, 2M-10B11 boundthe RBD with Kd of 24.3 nM, which was ob-tained by using heterogeneous ligand modelowing to the avidity effects (Fig. 4A, bottom).To investigatewhether thesemAbs block the

binding of S protein to ACE2, we performedflow cytometry using human embryonic kid-ney (HEK) 293T cells expressing human ACE2.As expected, only 2M-10B11 among the fivemAbs and ACE2-Fc prevented S protein frombinding to ACE2. In the presence of 2M-10B11,only 0.52% of cells were double positive forIgG and S protein (Fig. 4B). CR3022, whichcompetes with 2M-10B11, did not block thebinding of S to ACE2. The control mAb 1A8,targeting the Marburg glycoprotein, did notinterfere with the binding either, and the 5.13%of double positives may be due to the non-specific binding of 1A8 to S protein. 4A8 alsofailed to interfere with the binding of the Sprotein to ACE2.

Cryo-EM structure of the complex between4A8 and S-ECD

The mAb 4A8 was overexpressed and purifiedby Protein A resin, and the S-ECD of SARS-CoV-2 was purified through M2 affinity resinand size exclusion chromatography (SEC). 4A8and S-ECD protein were mixed and incubatedat a stoichiometric ratio of ~1.2 to 1 for 1 hourand applied to SEC to remove excess proteins(fig. S3A). The fraction containing the complex

was concentrated for cryo–electron micros-copy (cryo-EM) sample preparation.To investigate the interactions between 4A8

and the S protein, we solved the cryo-EMstructure of the complex at an overall resolu-tion of 3.1 Å (Fig. 5 and movie S1). Details ofcryo-EM sample preparation, data collectionand processing, and model building can befound in in the supplementary materials,materials and methods (figs. S3 to S5). TheS protein exhibits asymmetric conformationssimilar to the previously reported structures(21, 22), with one of three RBDs in “up” con-formation and the other two RBDs in “down”conformation (Fig. 5).

Recognition of the NTD by 4A8

In the S protein–4A8 complex, each trimericS protein is bound with three solved 4A8Fabs, each of which interacts with one NTDof the S protein. Despite the different confor-mations of the three S protein protomers,the interface between 4A8 and each NTD isidentical (Fig. 5 and fig. S3I). The map qualityat the NTD-4A8 region was improved throughfocused refinement to a local resolution of 3.3 Å,enabling reliable analysis of the interactionsbetween the NTD and 4A8.Association with 4A8 appears to stabilize

the NTD epitope, which is invisible in thereported S protein structure alone (21, 22).Supported by the high resolution of NTD, we

were able to build the structural model forfive new loops for NTD, designated N1 (resi-dues 14 to 26), N2 (residues 67 to 79), N3 (resi-dues 141 to 156), N4 (residues 177 to 186), andN5 (residues 246 to 260), among which the N3and N5 loopsmediate the interaction with 4A8(fig. S5A). Besides, three new glycosylation sites(Asn17, Asn61, and Asn149) on the NTD are iden-tified in this structure (fig. S6).The heavy chain of 4A8 mainly participates

in binding to the NTD mainly through threecomplementarity-determining regions (CDRs),named CDR1 (residues 25 to 32), CDR2 (resi-dues 51 to 58), and CDR3 (residues 100 to 116)(Fig. 6A and fig. S5B). The interface is con-stituted by an extensive hydrophilic interac-tion network, and the buried surface area atthe 4A8-NTD interface is 832 Å2. Arg246 onthe N5 loop of the NTD represents one dock-ing site, which is stabilized by Trp258, simul-taneously interacting with Tyr27 and Glu31 of4A8 on CDR1 (Fig. 6B). On the N3 loop of theNTD, Lys150 and Lys147 respectively form saltbridges with Glu54 and Glu72 of 4A8 (Fig. 6C).Lys150 is also hydrogen (H)–bondedwith 4A8-Tyr111, while His146 forms a H-bond with 4A8-Thr30 (Fig. 6C). In addition to the hydrophilicinteractions, Trp152 and Tyr145 on the N3 loopof the NTD also interact with Val102, Pro106,and Phe109 on the CDR3 of 4A8 through hy-drophobic and/or p-p interactions (Fig. 6D).Additionally, the glycosylation site of Asn149 on

Chi et al., Science 369, 650–655 (2020) 7 August 2020 4 of 6

0 200 400 600 800100012000.0

0.5

1.0

1.5

2.0

4A8 S-ECD

0 200 400 600 80010001200

4A8 S115.63 nM7.81 nM

250 nM125 nM

31.25 nM62.5 nM

1000 nM500 nM

0 200 400 600 800100012000.0

0.5

1.0

1.5

2.0

1M-1D2 S-ECD

0 200 400 600 800100012000.0

0.5

1.0

1.5

2.0

0304-3H3 S-ECD

0 200 400 600 800100012000.0

0.2

0.4

0.6

0.8

1.0

0304-3H3 S2

0 200 400 600 800100012000.0

0.2

0.4

0.6

0.8

1.0

1M-1D2 S1

nm

Time (sec)

K = (0.996 ± 0.045) nM K = (2.14 ± 0.17) nM K = (2.04 ± 0.16) nM

0 200 400 600 800100012000.0

0.5

1.0

1.5

2.0

2M-10B11 S-ECDK = (0.342 ± 0.009) nM

K = (92.7 ± 5.7) nM K = (4.52 ± 0.33) nM K = (108 ± 8) nM

0 200 400 600 800100012000.0

0.5

1.0

1.5

2.0

9A1 S-ECDK < 0.001 nM

0 200 400 600 800100012000.0

0.2

0.4

0.6

0.8

1.0

2M-10B11 RBDK = (24.3 ± 0.3) nM

0 200 400 600 800100012000.0

0.2

0.4

0.6

0.8

1.0

9A1 S2K < 0.001 nM

0.0

0.2

0.4

0.6

0.8

1.0

hum

an Ig

G

A

B

Spike Protein

4A8 0304-3H3 1M-1D2

42.5%

2M-10B11

0.52%35.9%

CR3022

31.7%26.7%

1A8

5.13% 0%0%

9A1

39.5%

ACE2-Fc

0.18%

No mAbNo S proteinNo mAb

D

D

D

D D D D

DD D

Fig. 4. 4A8 did not block the binding of Spike protein to ACE2 receptor.(A) BLI sensorgrams and kinetics of mAbs binding to S proteins. Globalfitting curves are shown as black lines. The Kd were calculated by using a1:1 binding model in Data Analysis Software 9.0, except for 2M-10B11,which used a heterogeneous ligand model owing to avidity effect. (B) Thebinding of S protein to human ACE2-overexpressing 293T cells were deter-mined by means of flow cytometry. After the preincubation of S protein

with each indicated mAb, the mAb-S mixtures were added to the ACE2-expressing cells. Cells were stained with anti-human IgG fluoresceinisothiocyanate (mAb binding, x axis) and anti-His (S binding, y axis).Percentages of double-positive cells are shown. Control mAb CR3022 and1A8 were previously reported to bind SARS-CoV RBD and Marburgglycoprotein, respectively, and ACE2-Fc protein was a human ACE2 proteinconjugated with human Fc.

RESEARCH | RESEARCH ARTICLEon N

ovember 1, 2020

http://science.sciencem

ag.org/D

ownloaded from

the NTD is close to the 4A8-NTD interface, ofwhich N-glycansmight participate in the inter-actions on the interface (Fig. 6A and fig. S6).

Discussion

There is an urgent need for prophylactic andtherapeutic interventions for SARS-CoV-2 in-fections given the ongoing COVID-19 pandemic.Our work reveals that naturally occurringhuman SARS-CoV-2 mAbs isolated from theB cells of 10 recovered donors are diverse ingene usage and epitope recognition of S pro-tein. The majority of the isolated mAbs didnot recognize the RBD, and all the mAbs thatneutralize authentic SARS-CoV-2 failed to in-hibit the binding of S protein to ACE2. Theseunexpected results suggest the presence ofother important mechanisms for SARS-CoV-2neutralization in addition to suppressing theviral interaction with the receptor.The S1-targeting mAb 4A8 does not block

the interaction between ACE2 and S proteinbut exhibits high levels of neutralization againstboth authentic and pseudotyped SARS-CoV-2in vitro.Many neutralizing antibodies againstSARS-CoV-2 were reported to target the RBDof the S protein and block the binding be-tween RBD and ACE2 (28–30, 32–34). Ourresults show that 4A8 binds to the NTD ofS protein with potent neutralizing activity.Previous study has shown that mAb 7D10could bind to the NTD of S protein of MERS-CoV probably by inhibiting the RBD-DPP4binding and the prefusion-to-postfusion con-formational changeof S protein (26).Wealignedthe crystal structure of 7D10 in complex withthe NTD of S protein of MERS-CoV with ourcomplex structure and found that the inter-faces between the mAb and the NTDs are par-tially overlapped (fig. S7). 7D10 may inhibitthe interaction betweenMERS-CoV and DPP4through its light chain, which is close to the

RBD. In our complex, the light chain of 4A8 isaway from the RBD (fig. S7). Therefore, we spe-culate that 4A8mayneutralize SARS-CoV-2 byrestraining the conformational changes of theS protein. Furthermore, sequence alignmentof the S proteins from SARS-CoV-2, SARS-CoV,and MERS-CoV revealed varied NTD surfacesequences that are respectively recognized bydifferent mAbs (fig. S8).

This work reports a fully human neutraliz-ing mAb recognizing a vulnerable epitope ofNTD on S protein of SARS-CoV-2, functioningwith a mechanism that is independent of re-ceptor binding inhibition. Combination of 4A8with RBD-targeting antibodies may avoid theescaping mutations of the virus and serve aspromising “cocktail” therapeutics. The infor-mation obtained from these studies can be

Chi et al., Science 369, 650–655 (2020) 7 August 2020 5 of 6

Fig. 5. Cryo-EM structure of the 4A8 and S-ECD complex. The domain-colored cryo-EM map of the complex is shown on the left, and two perpendicular views ofthe overall structure are shown on the right. The heavy and light chains of 4A8 are colored blue and magenta, respectively. The NTDs of the trimeric S protein arecolored orange. The one “up” RBD and two “down” RBDs of trimeric S protein are colored green and cyan, respectively.

Fig. 6. Interactions between the NTD and 4A8. (A) Extensive hydrophilic interactions on the interfacebetween NTD and 4A8. Only one NTD-4A8 is shown. (B to D) Detailed analysis of the interface between NTDand 4A8. Polar interactions are indicated by red dashed lines. The residues involved in hydrophobicinteractions are presented as spheres.

RESEARCH | RESEARCH ARTICLEon N

ovember 1, 2020

http://science.sciencem

ag.org/D

ownloaded from

used for development of the structure-basedvaccine design against SARS-CoV-2.

REFERENCES AND NOTES

1. N. Zhu et al., N. Engl. J. Med. 382, 727–733 (2020).2. F. Wu et al., Nature 579, 265–269 (2020).3. C. Huang et al., Lancet 395, 497–506 (2020).4. P. Zhou et al., Nature 579, 270–273 (2020).5. C. C. Lai, T. P. Shih, W. C. Ko, H. J. Tang, P. R. Hsueh, Int. J.

Antimicrob. Agents 55, 105924 (2020).6. T. G. Ksiazek et al., N. Engl. J. Med. 348, 1953–1966 (2003).7. A. M. Zaki, S. van Boheemen, T. M. Bestebroer,

A. D. Osterhaus, R. A. Fouchier, N. Engl. J. Med. 367,1814–1820 (2012).

8. W. Li et al., Nature 426, 450–454 (2003).9. J. H. Kuhn, W. Li, H. Choe, M. Farzan, Cell. Mol. Life Sci. 61,

2738–2743 (2004).10. K. Kuba et al., Nat. Med. 11, 875–879 (2005).11. D. S. Dimitrov, Cell 115, 652–653 (2003).12. J. Lan et al., Nature 581, 215–220 (2020).13. Q. Wang et al., Cell 181, 894–904.e9 (2020).14. R. Yan et al., Science 367, 1444–1448 (2020).15. J. Shang et al., Nature 581, 221–224 (2020).16. M. Hoffmann et al., Cell 181, 271–280.e8 (2020).17. T. M. Gallagher, M. J. Buchmeier, Virology 279, 371–374

(2001).18. G. Simmons, P. Zmora, S. Gierer, A. Heurich, S. Pöhlmann,

Antiviral Res. 100, 605–614 (2013).19. S. Belouzard, V. C. Chu, G. R. Whittaker, Proc. Natl. Acad.

Sci. U.S.A. 106, 5871–5876 (2009).20. W. Song, M. Gui, X. Wang, Y. Xiang, PLOS Pathog. 14,

e1007236 (2018).21. D. Wrapp et al., Science 367, 1260–1263 (2020).22. A. C. Walls et al., Cell 181, 281–292.e6 (2020).23. C. Krempl, B. Schultze, H. Laude, G. Herrler, J. Virol. 71,

3285–3287 (1997).24. F. Künkel, G. Herrler, Virology 195, 195–202 (1993).25. G. Lu, Q. Wang, G. F. Gao, Trends Microbiol. 23, 468–478

(2015).

26. H. Zhou et al., Nat. Commun. 10, 3068 (2019).27. B. Ju et al., Potent human neutralizing antibodies elicited by

SARS-CoV-2 infection. bioRxiv 2020.03.21.990770 [Preprint]26 March 2020. https://doi.org/10.1101/2020.03.21.990770.

28. C. Wang et al., Nat. Commun. 11, 2251 (2020).29. D. Wrapp et al., Cell 181, 1004–1015.e15 (2020).30. X. Chen et al., Cell. Mol. Immunol. 17, 647–649 (2020).31. M. Yuan et al., Science 368, 630–633 (2020).32. Y. Wu et al., Science 368, 1274–1278 (2020).33. B. Ju et al., Nature 10.1038/s41586-020-2380-z (2020).34. Y. Cao et al., Cell 10.1016/j.cell.2020.05.025 (2020).35. A. Burkovitz, I. Sela-Culang, Y. Ofran, FEBS J. 281, 306–319 (2014).36. E. P. Rock, P. R. Sibbald, M. M. Davis, Y. H. Chien, J. Exp. Med.

179, 323–328 (1994).37. C. G. Price, P. W. Abrahams, Environ. Geochem. Health 16,

27–30 (1994).38. S. Kumar, G. Stecher, K. Tamura, Mol. Biol. Evol. 33, 1870–1874

(2016).39. Q. Li et al., Vaccine 35, 5172–5178 (2017).40. J. Xu et al., Emerg. Microbes Infect. 8, 841–856 (2019).41. L. Wang et al., J. Virol. 92, e02002-17 (2018).42. J. Shang et al., Proc. Natl. Acad. Sci. U.S.A. 117, 11727–11734

(2020).

ACKNOWLEDGMENTS

We thank the Cryo-EM Facility and Supercomputer Center ofWestlake University for providing cryo-EM and computationsupport, respectively. We thank the Beijing Institute ofMicrobiology and Epidemiology, Academy of Military MedicalSciences, China, for providing the SARS-CoV-2. We also thankT. Fang, T. Yu, P. Lv, and E. Ma for providing technical support.Funding: This work was funded by the National Key R&D Programof China (2020YFC0841400), the National Natural ScienceFoundation of China (projects 31971123, 81803429, 81703048,31900671, 81920108015, and 31930059), the Key R&DProgram of Zhejiang Province (2020C04001), the SARS-CoV-2emergency project of the Science and Technology Department ofZhejiang Province (2020C03129), the Leading Innovative andEntrepreneur Team Introduction Program of Hangzhou, theNational Science and Technology Major Project of the Ministry of

Science and Technology of China, (2018ZX10101003-005-007),and the Special Research Program of Novel CoronavirusPneumonia of Westlake University. Author contributions: W.C.,Q.Z., and J.L. conceived the project. X.C., R.Y., Ju.Z., G.Z., Y.Z.,Y.G., Y.L., L.X., M.H., Z.Z., P.F., Y.D., Z.C., Ji.Z., X.S., Y.C.,L.F., L.H., J.X., and C.Y. did the experiments. All authorscontributed to data analysis. X.C., R.Y., J.L., Q.Z., and W.C.wrote the manuscript. Competing interests: W.C., J.L., X.C.,Ju.Z., L.F., C.Y., J.X., L.H., G.Z., P.F., M.H., Y.D., X.S., Y.C., andJi.Z. are listed as inventors on a pending patent applicationfor mAb 4A8. The other authors declare that they have nocompeting interests. Data and materials availability: Atomiccoordinates and cryo-EM density maps of the S protein ofSARS-CoV-2 in complex bound with 4A8 (PDB: 7C2L; whole map:EMD-30276, antibody-epitope interface-focused refined map:EMD-30277) have been deposited to the Protein Data Bank(www.rcsb.org) and the Electron Microscopy Data Bank(www.ebi.ac.uk/pdbe/emdb), respectively. Antibody sequenceshave been deposited at GenBank (MTA 622682 to 622751). Thiswork is licensed under a Creative Commons Attribution 4.0International (CC BY 4.0) license, which permits unrestricted use,distribution, and reproduction in any medium, provided theoriginal work is properly cited. To view a copy of this license,visit https://creativecommons.org/licenses/by/4.0. This licensedoes not apply to figures/photos/artwork or other contentincluded in the article that is credited to a third party; obtainauthorization from the rights holder before using such material.

SUPPLEMENTARY MATERIALS

science.sciencemag.org/content/369/6504/650/suppl/DC1Materials and MethodsFigs. S1 to S7Tables S1 to S4References (43–66)Movie S1MDAR Reproducibility Checklist

8 May 2020; accepted 17 June 2020Published online 22 June 202010.1126/science.abc6952

Chi et al., Science 369, 650–655 (2020) 7 August 2020 6 of 6

RESEARCH | RESEARCH ARTICLEon N

ovember 1, 2020

http://science.sciencem

ag.org/D

ownloaded from

SARS-CoV-2A neutralizing human antibody binds to the N-terminal domain of the Spike protein of

Lihua Hou, Junjie Xu, Changming Yu, Jianmin Li, Qiang Zhou and Wei ChenDong, Yilong Yang, Zhengshan Chen, Yingying Guo, Jinlong Zhang, Yaning Li, Xiaohong Song, Yi Chen, Lu Xia, Ling Fu, Xiangyang Chi, Renhong Yan, Jun Zhang, Guanying Zhang, Yuanyuan Zhang, Meng Hao, Zhe Zhang, Pengfei Fan, Yunzhu

originally published online June 22, 2020DOI: 10.1126/science.abc6952 (6504), 650-655.369Science 

, this issue p. 650Sciencecombine with RBD-targeting antibodies in therapeutic cocktails.

electron microscopy revealed the epitope as the NTD. This NTD-targeting antibody may be useful to−he RBD. Cryoantibodies from 10 convalescent patients and identified an antibody that potently neutralizes the virus but does not bind t

isolated et al.N-terminal domain (NTD). In searching for neutralizing antibodies, there has been a focus on the RBD. Chi binding to host cells and membrane fusion, respectively. In addition to the receptor binding domain (RBD), S1 has anthe spike protein, a trimeric protein complex with each monomer comprising an S1 and an S2 domain that mediate

A key target for therapeutic antibodies against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) isHitting SARS-CoV-2 in a new spot

ARTICLE TOOLS http://science.sciencemag.org/content/369/6504/650

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2020/06/19/science.abc6952.DC1

CONTENTRELATED

http://stm.sciencemag.org/content/scitransmed/12/550/eabc3539.fullhttp://stm.sciencemag.org/content/scitransmed/12/554/eabc1126.fullhttp://stm.sciencemag.org/content/scitransmed/12/541/eabb5883.fullhttp://stm.sciencemag.org/content/scitransmed/12/549/eabb9401.full

REFERENCES

http://science.sciencemag.org/content/369/6504/650#BIBLThis article cites 65 articles, 12 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on Novem

ber 1, 2020

http://science.sciencemag.org/

Dow

nloaded from