trimer, potentially allowing for increased
recruitment and/or selection of specific clones
that can bind conserved regions of RBD ( 43 , 44 ).
By contrast, convalescent individuals were
primed against native, nonstabilized spike pro-
tein. Taken together, our data indicate robust
B cell memory to multiple components of the
spike protein as well as currently described
VOCs that continues to evolve and increase in
frequency over time.
Clonal evolution of variant-specific
memory B cells
We next asked what differences may underly
variant-binding versus nonbinding proper-
ties of memory B cells. Here, we focused on
the Beta B.1.351 variant RBD containing the
K417N, E484K, and N501Y mutations because
this variant resulted in the greatest loss of
binding relative to WT RBD (Fig. 3, E, G, and
H). We designed a sorting panel to identify
three populations of memory B cells with
different antigen-binding specificities: (i) mem-
ory B cells that bind full-length spike but not
RBD, (ii) memory B cells that bind full-length
spike and WT RBD but not B.1.351 variant RBD,
and (iii) memory B cells that bind full-length
spike and cross-bind both WT and B.1.351 var-
iant RBD (Fig. 4A and fig. S5A). Naïve B cells
were also sorted as a control. These populations
were isolated from eight SARS-CoV-2–naïve
Goelet al.,Science 374 , eabm0829 (2021) 3 December 2021 7of17
Naive Recovered
Naive Recovered
0.0
2.5
5.0
7.5
WT RBD+
RBD++
% mutated VH NT
Overlapping Clones
****
WT RBD+
18177
RBD++
576 4419
Clonal Overlap (50% mcf)
0
25
50
75
100
Pre−Im
mun
e
Naive
Reco
vere
d
% of WT RBD+
B.1.351+
Naive
B
Spike
+ RBD−WT RBD+RBD++Nai
ve B
Spik
e+ RBD−WT RBD+RBD++
1
10
100
D20 Score
Clonality
ns0.078
0.057
ns0.069
*
A
D
RBD+
7.24
RBD++
20.3
RBD-
72.1
Gated on Spike+ Memory B
RBD WT - APC
RBD B.1.351 - PE
Isolate B Cells Cell Sorting
Label SARS-CoV-2-Specific
Memory B Cells
N = 8 N = 4
3-4 Months Post-Vaccine
BCR Sequencing
IgH
RBD-WT RBD+RBD++
Spike+ Memory B
Naive
Recovere
d
Cross-BindingB.1.351
Prior COVID-19
BC
H
G
I
J
Overlapping Clones
N = 576
Higher SHM in:
WT RBD+
Equal
RBD++
: 23.6%
: 45%
: 31.4%
E
Naive Recovered
(^0510152005101520)
0.0
0.1
0.2
0.3
Density
SHM
Naiv
e B
Spike+ RBD−
WT RBD+RBD++Nai
ve B
Spik
e+ RBD−WT RBD+RBD++
0
4
8
12
% mutated VH NT
SHM
N=117,451
N=23,583
N=15,340
N=5,954
N=105,913
N=13,277
N=10,091
N=2,147
% mutated VH NT
F
Clonal Lineages with Binding Overlap
WT RBD+ RBD++
IGHV3-53
CARELGDFAFDIW
IGHV3-53
CARDLEVYGMDVW
IGHV3-33
CARDVNDTTMALGPLYFYGMDVW
IGHV3-53
CARDLDYYGMDVW
IGHV3-53
RARDYGDLYFDYW
IGHV4-59
CARNGGSGRIGLDKKWAFDIW
IGHV3−49
IGHV5−10−1
IGHV3−13
IGHV4−38−2
IGHV6−1
IGHV1−24
IGHV1−8
IGHV3−74
IGHV1−3
IGHV2−70
IGHV3−11
IGHV3−15
IGHV4−61
IGHV2−5
IGHV4−4
IGHV3−53
IGHV3−66
IGHV1−18
IGHV5−51
IGHV1−46
IGHV4−31
IGHV3−9
IGHV3−48
IGHV4−34
IGHV1−2
IGHV3−21
IGHV3−7
IGHV4−39
IGHV4−59
IGHV1−69
IGHV3−33
IGHV3−23|3−23D
IGHV3−30
Recovered
Binding
Binding
Spike+ RBD−
WT RBD+
RBD++
Prior COVID-19
No
Ye s
5
10
15
0
% of Clones (by Column)
GL GL GL GL GL GL
**
- ns
Fig. 4. Variant-binding memory B cell clones use distinct VH genes and
evolve through somatic hypermutation.(A) Experimental design for sorting
and sequencing SARS-CoV-2Ðspecific memory B cells. (B) Frequency of RBD++
(B.1.351 variant cross-binding) memory B cells as a percentage of total RBD+
cells. (C) Percentage of sequence copies occupied by the top 20 ranked
clones (D20) across naïve B cells and different antigen-binding memory B cell
populations. (D) Heatmap and hierarchical clustering of VH gene usage
frequencies in memory B cell clones across different antigen-binding populations.
Data are represented as the percentage of clones with the indicated VH gene
per column. (EandF) Somatic hypermutation (SHM) density plots (E) and
boxplots of individual clones across naïve B cells and different antigen-binding
memory B cell populations (F). Data are represented as the percent of
mutated VH nucleotides. Number of clones sampled for each population
is indicated. For (C) to (F), data were filtered on clones with productive
rearrangements and≥2 copies. (G) Venn diagram of clonal lineages that are
shared between WT RBD and RBD cross-binding (RBD++) populations. Data were
filtered on the basis of larger clones with≥50% mean copy number frequency
(mcf) in each sequencing library. (H) Example lineage trees of clones with
overlapping binding to WT and B.1.351 variant RBD. VH genes and CDR3
sequences are indicated. Numbers refer to mutations compared with the
preceding vertical node. Colors indicate binding specificity, black dots indicate
inferred nodes, and size is proportional to sequence copy number. GL, germline
sequence. (I) Classification of SHM within overlapping clones. Each clone
was defined as having higher (or equal) SHM in WT RBD binders or RBD++
cross-binders on the basis of average levels of SHM for all WT RBD versus RBD++
sequence variant copies within each lineage. (J) SHM levels within overlapping
clones. Data are represented as the percentage of mutated VH nucleotides
for WT RBD and RBD++sequence copies. Statistics were calculated using
unpaired nonparametric Wilcoxon test, with BH correction for multiple comparisons
in (C) and (F). Notches on boxplots in (F) and (J) indicate a 95% confidence
interval of the median. P< 0.05; P< 0.01; P< 0.001; ****P< 0.0001;
ns, not significant.
RESEARCH | RESEARCH ARTICLE