Article
Broad and Potent Neutralizing Antibodies Recognize
the Silent Face of the HIV Envelope
Graphical Abstract
Highlights
d
Silentface (SF) bNAbs can reach substantial neutralization
breadth and potency
d
Defined structure of a SF bNAb bound to an Env trimer
d
Binding of SF bNAb resulted in trimer asymmetry in the V3
region
d
SF antibodies have
in vivo
activity and potential for
clinical use
Authors
Till Schoofs, Christopher O. Barnes,
Nina Suh-Toma, ..., Florian Klein,
Michel C. Nussenzweig,
Pamela J. Bjorkman
Correspondence
nussen@rockefeller.edu (M.C.N.),
bjorkman@caltech.edu (P.J.B.)
In Brief
VRC-PG05 was the only donor-derived
antibody against the silentface (SF) of
HIV-1 envelope described to date.
Schoofs et al. identify the antibody SF12
and its relatives, which recognize the
center of the SF with a different angle and
more extensive protein recognition than
VRC-PG05, thereby achieving substantial
neutralizing ability and potential for
clinical use.
Schoofs et al., 2019, Immunity
50
, 1513–1529
June 18, 2019
ª
2019 The Authors. Published by Elsevier Inc.
https://doi.org/10.1016/j.immuni.2019.04.014
Immunity
Article
Broad and Potent Neutralizing Antibodies
Recognize the Silent Face of the HIV Envelope
Till Schoofs,
1
,
2
,
3
,
13
,
14
Christopher O. Barnes,
4
,
13
Nina Suh-Toma,
4
,
5
Jovana Golijanin,
1
Philipp Schommers,
2
,
3
Henning Gruell,
2
,
3
Anthony P. West, Jr.,
4
Franziska Bach,
2
Yu Erica Lee,
4
Lilian Nogueira,
1
Ivelin S. Georgiev,
6
,
7
Robert T. Bailer,
8
Julie Czartoski,
9
John R. Mascola,
8
Michael S. Seaman,
10
M. Juliana McElrath,
9
Nicole A. Doria-Rose,
8
Florian Klein,
2
,
3
,
11
Michel C. Nussenzweig,
1
,
12
,
*
and Pamela J. Bjorkman
4
,
15
,
*
1
Laboratory of Molecular Immunology, The Rockefeller University, New York, NY 10065, USA
2
Laboratory of Experimental Immunology, Institute of Virology, Faculty of Medicine and University Hospital of Cologne, University of Cologne,
50931 Cologne, Germany
3
German Center for Infection Research, partner site Bonn-Cologne, 50931 Cologne, Germany
4
Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA 91125, USA
5
Westridge High School, 324 Madeline Drive, Pasadena, CA 91105, USA
6
Vanderbilt Vaccine Center, Department of Pathology, Microbiology and Immunology, Vanderbilt University Medical Center, Nashville,
TN 37232, USA
7
Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN 37232, USA
8
Vaccine Research Center, National Institute of Allergy and Infectious Diseases, NIH, Bethesda, MD 20892, USA
9
Vaccine and Infectious Disease Division, Fred Hutchinson Cancer Research Center, Seattle, WA 98109, USA
10
Center for Virology and Vaccine Research, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA 02215, USA
11
Center for Molecular Medicine Cologne (CMMC), University of Cologne, 50931 Cologne, Germany
12
Howard Hughes Medical Institute, The Rockefeller University, New York, NY 10065, USA
13
These authors contributed equally
14
Present address: GSK Vaccines, 1300 Wavre, Belgium
15
Lead Contact
*Correspondence:
nussen@rockefeller.edu
(M.C.N.),
bjorkman@caltech.edu
(P.J.B.)
https://doi.org/10.1016/j.immuni.2019.04.014
SUMMARY
Broadly neutralizing antibodies (bNAbs) against
HIV-1 envelope (Env) inform vaccine design and
are potential therapeutic agents. We identified
SF12 and related bNAbs with up to 62% neutraliza-
tion breadth from an HIV-infected donor. SF12
recognized a glycan-dominated epitope on Env’s
silent face and was potent against clade AE vi-
ruses, which are poorly covered by V3-glycan
bNAbs. A 3.3A
̊
cryo-EM structure of a SF12-Env
trimer complex showed additional contacts to Env
protein residues by SF12 compared with VRC-
PG05, the only other known donor-derived silent-
face antibody, explaining SF12’s increased neutral-
ization breadth, potency, and resistance to Env
mutation routes. Asymmetric binding of SF12 was
associated with distinct N-glycan conformations
across Env protomers, demonstrating intra-Env
glycan heterogeneity. Administrating SF12 to HIV-
1-infected humanized mice suppressed viremia
and selected for viruses lacking the N448
gp120
glycan. Effective bNAbs can therefore be raised
against HIV-1 Env’s silent face, suggesting their po-
tential for HIV-1 prevention, therapy, and vaccine
development.
INTRODUCTION
Neutralizing antibodies (NAbs) play a key role in antiviral immu-
nity and are the correlate of protection of most available vaccines
(
Burton, 2002; Plotkin, 2010
). The HIV-1 envelope glycoprotein
(Env) is the only potential target for NAbs on the surface of the vi-
rus (
Burton and Hangartner, 2016; Kwong and Mascola, 2018;
Wibmer et al., 2015
). Env is a trimeric spike composed of
gp120/gp41 heterodimers that has evolved a plethora of immune
escape mechanisms to evade antibody recognition. These
include instability of the trimer, sparsity of spikes on the virion
surface, high sequence divergence across strains, and epitope
masking through its extensive glycan shield (
Burton and Hang-
artner, 2016; Haynes, 2015; Klein and Bjorkman, 2010; Kwong
and Mascola, 2018
).
Consequently, effective humoral responses to HIV-1 typically
only emerge several years after infection and only in a subset
of HIV-1-infected individuals (
Gray et al., 2011; Landais et al.,
2016; Mikell et al., 2011; Rusert et al., 2016; Tomaras et al.,
2011
). Although
50% of chronically HIV-1-infected individuals
develop some degree of cross-clade serum neutralization, only
a small fraction of individuals mounts outstandingly broad and
potent antibody responses against the virus (
Doria-Rose et al.,
2010; Hraber et al., 2014; Landais et al., 2016; Rusert et al.,
2016; Sather et al., 2009; Simek et al., 2009
). The development
and use of single B cell antibody cloning revealed that this activ-
ity can usually be attributed to one or a combination of broadly
neutralizing antibodies (bNAbs) that target HIV-1 Env (
Scheid
Immunity
50
, 1513–1529, June 18, 2019
ª
2019 The Authors. Published by Elsevier Inc.
1513
This is an open access article under the CC BY license (
http://creativecommons.org/licenses/by/4.0/
).
et al., 2009a, 2009b, 2011; Walker et al., 2009, 2010; Wu
et al., 2010
).
NAbs are proposed to interfere with viral entry in a variety of
ways, including blocking rece
ptor engagement, preventing
membrane fusion, and enhancing decay of Env spikes (
Bur-
ton and Hangartner, 2016
). Pre-clinical and recent human
studies have highlighted the potential of bNAbs for HIV-1
therapy and prevention (
Bar et al., 2016; Bar-On et al.,
2018; Barouch et al., 2013; Caskey et al., 2015, 2017; Gautam
et al., 2016; Julg et al., 2017; Lynch et al., 2015; Mendoza
et al., 2018; Scheid et al., 2016; Shingai et al., 2013
). More-
over, structural insights into mechanisms of bNAb binding
have been key to designing novel immunogens and strate-
gies for vaccination (
Escolano et al., 2017; Jardine et al.,
2013; Kwong and Mascola, 2018; McGuire et al., 2013;
Sanders et al., 2013; Ward and Wilson, 2017
). However, there
remain a number of challenges to the elicitation and clinical
use of bNAbs. For vaccine efforts, these include unusual
structural features of bNAbs such as large insertions/dele-
tions, and/or unusual complementarity determining region
(CDR) lengths as well as extensive somatic hypermutation
(SHM), all of which are rare features in the human repertoire
(
Burton and Hangartner, 2016; Haynes and Burton, 2017;
Sok and Burton, 2018
). For clinical use of bNAbs, viral
coverage gaps, manufacturability, and pre-existing bNAb
resistance represent potential problems (
Escolano et al.,
2017; Gruell and Klein, 2018; Sok and Burton, 2018
). Thus,
there is a continuing need to identify bNAbs that may be
more readily elicited by vaccination and that are suitable for
clinical use.
Although many bNAbs have been characterized, their tar-
gets, or ‘‘sites of vulnerability’’, on the HIV-1 Env spike appear
to be limited (
Burton and Hangartner, 2016; Kwong and Mas-
cola, 2018; Sok and Burton, 2018; Ward and Wilson, 2017;
West et al., 2014; Wibmer et al., 2015
). Numerous monoclonal
antibodies recognize the CD4-binding site, the V3-glycan
patch, the V2-apex, the membrane proximal external region
(MPER), and several epitopes encompassing the gp120-gp41
interface (
Burton and Hangartner, 2016; Kwong and Mascola,
2018; Sok and Burton, 2018; Ward and Wilson, 2017; Wibmer
et al., 2015
). In contrast, VRC-PG05 is the only donor-derived
antibody isolated to date that binds to the highly glycosylated
‘‘silent face’’ of gp120 (
Zhou et al., 2018
). However, VRC-PG05
neutralized only 27% of tested HIV-1 strains and had a rela-
tively high mean IC
50
of 0.8
m
g/mL, leaving uncertain the po-
tential usefulness of this epitope for vaccine design, therapy,
or prevention.
Here, we describe silent face (SF) bNAbs targeting a VRC-
PG05-related epitope that cover up to 62% of evaluated strains
with a mean IC
50
of 0.20
m
g/mL. To characterize the binding
mechanism of the new antibodies, we determined the 3.1 A
̊
crys-
tal of the unbound SF12 Fab and a 3.3 A
̊
cryo-EM structure SF12
Fab bound to the clade B B41 Env trimer. We found that SF12
binds the center of the Env silent face with a different orientation
and set of contacts than VRC-PG05. The overall breadth and
potency achieved by SF12 suggests that the silent face is an
additional target for vaccine design and that antibodies to this
site may be clinically useful as a complement to other avail-
able bNAbs.
RESULTS
Isolation of an Antibody Family from Donor 27845 by B
Cell Culture and BG505 Sorting
Donor 27845 was diagnosed with HIV-1 in 1985 and followed in a
cohort of long-term non-progressors at the Fred Hutchinson
Cancer Research Center from 1998–2006. Apart from an inter-
ventional study during which the subject started and stopped
anti-retroviral therapy (ART) at set intervals from 1998–2001,
the subject has been off ART (
Figure 1
A). The individual’s purified
immunoglobulin G (IgG) isolated from a 2005 time point was
evaluated for neutralization against a 12-virus panel representa-
tive of the global epidemic (
deCamp et al., 2014
)(
Figure 1
B) and
found to be both broad and potent with a coverage of 92% and
an average median inhibitory concentration (IC
50
) of 92.3
m
g/mL
(
Figure 1
B). To inform potential antibody isolation strategies,
neutralization fingerprinting of the subject’s IgG was performed,
but the results were inconclusive due to borderline prediction
confidence scores (
Doria-Rose et al., 2017
)(
Figure 1
B).
Based on the fingerprinting results, we employed an unbiased
B cell microculture approach for antibody cloning (
Doria-Rose
et al., 2015; Huang et al., 2013
). From a starting number of
4.4
3
10
4
memory B cells, we identified seven B cells, six of
which were members of a single clone, that showed potent
anti-HIV-1 neutralizing activity against two indicator strains.
Subsequent single B cell sorting using fluorescently labeled
BG505.SOSIP.664 native-like Env trimers (
Sanders et al., 2013
)
yielded two additional members of this antibody family, one of
which was identical to an antibody obtained in the B cell culture.
The members of the clone utilized V
H
4-59*01 and V
K
3-20*01
heavy and light chain variable gene segments and included
CDRH3s and CDRL3s of 23 and 6 amino acids, respectively
(
Table S1
). V
H
gene segment mutation frequencies ranged
from 17%–25% of nucleotides (21%–39% amino acids), and
V
K
gene segment mutation frequencies ranged from 15%–21%
of nucleotides (20%–29% amino acids), intermediate rates
of SHM for HIV-1 bNAbs (
Table S1
). Based on heavy chain
sequences, the family segregated into three phylogenetic
branches (
Figure 1
C), with the SF5/SF12 branch showing a
three-nucleotide CDRH2 insertion.
When tested on two representative panels of 20 (f61 panel)
and 12 (global panel) viruses (
deCamp et al., 2014; Doria-
Rose et al., 2017
), members of the V
H
4-59 clone showed diverse
levels of activity and breadth (
Figure 1
D). Two closely related
members of the V
H
4-59 clone that were the most active, SF5
and SF12, were then evaluated against a 119-virus panel repre-
sentative of all major circulating HIV-1 clades (
Figure 1
E;
Table
S2
)(
Freund et al., 2017; Mouquet et al., 2012
). SF5 and SF12
neutralized 58% and 62% of viruses in this larger panel, with
geometric mean IC
50s
of 0.25 and 0.20
m
g/mL, respectively.
Notably, SF12 neutralized all of 18 tested clade AE viruses
across the three panels and showed a pattern of neutralizing
activity that differed from previously described bNAbs (
Fig-
ure 1
F). Overall, the antibody clone recapitulated the majority
of the polyclonal IgG neutralization activity, with the potency
correlation between isolated monoclonal antibodies (mAbs)
and donor 27845’s IgG resembling those of other elite neutral-
izers from whom we previously isolated bNAbs (
Freund et al.,
2017; Scheid et al., 2011
).
1514
Immunity
50
, 1513–1529, June 18, 2019
We next evaluated potential autoreactivity and polyreactivity
of SF12 and SF5 using HEp-2 staining (
Haynes et al., 2005
)
and a baculovirus-based polyreactivity assay (
Ho
̈
tzel et al.,
2012
), respectively (
Figure S1
). In contrast to bNAbs with known
autoreactive and polyreactive properties such as 2F5 and 4E10
(
Haynes et al., 2005
), we found minimal to no autoreactivity or
AB
C
D
E
F
Figure 1. Isolation of Antibody Family from Donor 27845 by B Cell Culture and BG505 Sorting
(A) ViralloadandCD4
+
T cellcounts ofHIV-1-infected subject27845over time.Arrows indicatetime points ofB cell microculture andBG505.SOSIP.664bait-sorting.
(B) Neutralization data of donor 27845’s serum IgG in 2005 against a 12-virus cross-clade panel (global) and a 20-virus fingerprinting panel (f61). Sh
own are
median inhibitory concentrations (IC
50
)in
m
g/mL. On the right, fingerprinting analysis of f61 serum neutralization. Neutralization testing performed in duplicates,
average shown.
(C) Maximum-likelihood phylogenetic tree of heavy chain sequences of newly isolated antibody family. MC = Antibodies isolated by B cell microcultur
e, BG505-
sort = antibodies isolated by bait-sorting, Both = antibody found both by microculture and bait-sorting.
(D) Neutralization of isolated antibody family members (IC
50
) against global and f61 virus panels. Legend as in (B). Neutralization testing performed in duplicates,
average shown.
(E) Neutralization coverage and potency of SF5 and SF12 on a 119-virus cross clade panel. Neutralization testing performed in duplicates, average sh
own.
(F) Neutralization fingerprinting of SF5 and SF12 in comparison to other known anti-HIV-1 bNAbs.
See also
Figure S1
and
Tables S1
and
S2
.
Immunity
50
, 1513–1529, June 18, 2019
1515
polyreactivity for SF5 and SF12 (
Figures S1
A and S1B). In addi-
tion, the pharmacokinetics of SF12 in mice were similar to those
of 3BNC117, a bNAb that exhibits a typical IgG1 half-life in ma-
caques (
Gautam et al., 2016
) and humans (
Caskey et al., 2015
)
(
Figure S1
C). We conclude that the SF antibody family achieves
substantial anti-HIV-1 neutralization with an intermediate degree
of somatic hypermutation and no evidence for autoreactivity.
Antibodies SF5 and SF12 Bind a Distinct Epitope on the
gp120 Portion of Env
To map the epitope recognized by SF5 and SF12, we performed
ELISAs using HIV-1 Env proteins. Both antibodies bound to
monomeric YU2 gp120, indicating that a portion of the epitope
is contained within the gp120 subunit of Env (
Figure 2
A). We sub-
sequently evaluated ELISA binding to site-directed mutants in
A
BC
D
E
SF5 N448 (n=98)
SF12 N448 (n=98)
SF12 - 448 (n=21)
SF5 -448 (n=21)
F
119-virus panel
10
-2
10
-1
10
0
10
1
10
2
0
20
40
60
80
100
SF5 YU2 WT
SF12 YU2 WT
SF5 YU2 N448K
SF12 YU2 N448
K
10
-2
10
-1
10
0
10
1
10
2
0
20
40
60
80
100
%Neu
t
raliza
tion
WT
N160K
T162N
N276D
T278K
N279H
N279K
N280Y
A281T
N301D
N332K
S334N
G366E
E429K
G458D
G459D
G471R
N280Y N160K N332K
CD4bs
V3-glycan
V2-apex
N295S
Triple KO
YU2 site mutant panel
YU2 pseudovirus
log10 antibody (
μ
g/ml)
log10 antibody (
μ
g/ml)
Competition BG505.SOSIP.664
YU2 gp140 foldon
YU2 gp120
log10 antibody (
μ
g/ml)
log10 antibody (
μ
g/ml)
BG505.SOSIP.664
10
-3
10
-2
10
-1
10
0
10
1
10
2
0
1
2
3
OD
SF5
10
-3
10
-2
10
-1
10
0
10
1
10
2
0
1
2
3
OD
10
-3
10
-2
10
-1
10
0
10
1
10
2
0
1
2
3
OD415
SF12
10
-3
10
-2
10
-1
10
0
10
1
10
2
0
1
2
3
SF12
N160K
WT
D368R
N332A
A281T D368K
WT
Triple mutant
3BNC117
8ANC131
10-1074
PGDM1400
8ANC195
PGT151
35O22
SF5
SF12
10
-3
10
-2
10
-1
10
0
10
1
10
2
0
1
2
3
O
D
SF5
10
-2
10
-1
10
0
10
1
10
2
0
1
2
3
OD41
5
SF12
10
-2
10
-1
10
0
10
1
10
2
0
1
2
3
OD
SF5
3BNC117
8ANC131
10-1074
PGDM1400
8ANC195
PGT151
35O22
SF5
SF12
log10 antibody (
μ
g/ml)
log10 antibody (
μ
g/ml)
10
-3
10
-2
10
-1
10
0
10
1
10
2
0
20
40
60
80
100
Neutralization potency
IC
50
(
μ
g/ml)
Neutraliz
a
t
ion b
r
eadt
h (%)
% Neu
tr
ali
z
ation
Figure 2. Antibodies SF5 and SF12 Bind a Distinct Epitope on the gp120 Portion of Env
(A) ELISA of SF5 and SF12 against a gp120 monomer and a gp140 foldon trimer derived from HIV-1 strain YU2. Wild-type proteins and various site mutants of
the
proteins in common bNAb epitopes (CD4-binding site, V3-glycan, Apex) were tested. Triple mutant = N160K, A281T + D368K, N332K. Data representative o
f3
repeat assays.
(B) ELISA of SF5 and SF12 as well as reference bNAbs targeting 6 known epitopes against the BG505.SOSIP.664 trimer. Data representative of 3 repeat ass
ays.
(C) Competition ELISA with reference bNAbs targeting 6 known epitopes to evaluate interference with SF5 and SF12 binding to the BG505.SOSIP.664 trim
er.
Competing antibodies were added in a dilution series starting at 32
m
g/mL. SF5 and SF12 were added at a constant concentration of 0.5
m
g/mL. Data repre-
sentative of 3 repeat assays.
(D) Neutralization testing of SF12 against a panel of YU2 site mutants covering major epitopes on the HIV-1 spike. Neutralization testing performed i
n duplicates,
average curves shown.
(E) Computational analysis of 119-virus cross clade panel neutralization.
(F) Neutralization testing of SF5 and SF12 against an HIV-1 pseudovirus based on strain YU2 carrying a mutation at the PNGS N448
gp120
. Testing done in
duplicates, average shown.
1516
Immunity
50
, 1513–1529, June 18, 2019
monomeric YU2 gp120 and an uncleaved YU2 gp140 foldon
trimer (
Yang et al., 2000
) that define common epitopes. Mutation
of the CD4-binding site (D368R/D368K
gp120
, A281T
gp120
)(
Dose-
novic et al., 2015; Horwitz et al., 2013; Olshevsky et al., 1990
), the
V3 glycan patch (N332A/N332K
gp120
)(
Horwitz et al., 2013; Mou-
quet et al., 2012
), and the V2 apex epitope (N160K
gp120
)(
Walker
et al., 2009
), alone or in combination, did not abrogate SF5 or
SF12 binding (
Figure 2
A). SF5 and SF12 also bound to a cleaved
soluble native-like BG505.SOSIP.664 trimer (
Sanders et al.,
2013
)(
Figure 2
B). Taken together, these results suggested that
SF5 and SF12 bind an epitope that is present on gp120 mono-
mers and both cleaved and uncleaved Env trimers.
We performed competition ELISAs to assess binding to the
BG505.SOSIP.664 trimer using antibodies targeting the CD4-
binding site (3BNC117, 8ANC131), the V3 glycan patch (10-
1074), the V2-apex (PGDM1400) and the gp120-gp41 interface
(8ANC195, PGT151 and 35O22). Both SF5 and SF12 competed
strongly with themselves and each other (
Figure 2
C). 3BNC117,
a CD4-binding site antibody that bridges adjacent protomers
within a trimer and has a broad contact surface with gp120
(
Lee et al., 2017
), showed competition with both SF5 and
SF12. Incomplete competition was also observed for the CD4-
binding site antibody 8ANC131 and for the gp120-gp41 interface
antibodies 8ANC195 and PGT151. Moreover, the V3-glycan tar-
geting antibody 10-1074 competed strongly with SF5 but not
with SF12 (
Figure 2
C). We also assessed the neutralizing activity
of SF12 on a YU2 pseudovirus mutant panel comprising a num-
ber of mutations that impair the activity of CD4-binding site, V3-
glycan and V2-apex antibodies using a TZM.bl-based
in vitro
neutralization assay. SF12 neutralizing activity was insensitive
to the mutations, including a triple mutant carrying mutations in
all three epitopes (N280Y
gp120
, N160K
gp120
, N332K
gp120
)(
Fig-
ure 2
D). These data indicate that SF5 and SF12 bind a distinct
epitope near the epitopes for CD4-binding site bNAbs and
gp120-gp41 interface bNAbs 8ANC195 and PGT151.
Computational analysis (
West et al., 2013
) of available neutral-
ization data suggested that SF5/SF12 depend on the presence of
a glycan at N448
gp120
(
Figure 2
E). To verify that the neutralizing
activity of SF5 and SF12 depended on this potential N-linked
glycosylation site (PNGS), we showed that these antibodies failed
to neutralize a mutant HIV
YU2
pseudovirus lacking the N448
gp120
glycan (
Figure 2
F). The PNGS at position 448
gp120
is at the center
of one of the most highly glycosylated parts of the HIV-1 trimer,
also known as the silent face (
Wyatt et al., 1998
). Although
comparisons of synonymous versus non-synonymous mutations
suggested that the silent face is under immunologic pressure
(
Stewart et al., 2001
), antibodies that bind to the center of this
region have been difficult to isolate. Indeed, VRC-PG05 repre-
sented an, until now, unique example of a host-derived bNAb
that specifically targets the center of the silent face with a focus
on the glycan site at N448
gp120
(
Zhou et al., 2018
). The discovery
and characterization of SF12 and related silent face bNAbs
shows that this epitope can be targeted by antibodies with
greater breadth and potency than VRC-PG05.
Structure of the Natively Glycosylated SF12-Env
Complex
We determined a 3.1 A
̊
crystal structure of the SF12 Fab and a
3.3 A
̊
cryo-EM structure of a natively glycosylated clade B B41
SOSIP.664 trimer in complex with the SF12 Fab and a Fab
from the V3/glycan patch bNAb 10-1074 (
Figures 3
A and 3B).
Although 10-1074 Fab normally binds with a 3:1 Fab:Env trimer
stoichiometry (
Gristick et al., 2016
), EM class averages showed
either three or two SF12 Fabs bound to the Env trimer and only
one 10-1074 Fab (
Figures S2
and
S3
). Like VRC-PG05, for which
a crystal structure was solved in complex with a monomeric
gp120 core (
Zhou et al., 2018
), the SF12-trimer complex reveals
recognition of an epitope focused on the N262
gp120
, N295
gp120
,
and N448
gp120
glycans on the silent face of Env, rationalizing our
binding and
in vitro
neutralization results (
Figures 2
A–2F). Super-
imposition of the free and Env-bound SF12 Fab structures
showed only minor conformational changes resulting from Env
glycan interactions with the SF12 Fab in the Env-bound struc-
ture, as evidenced by the 1.1 A
̊
root-mean-square deviation
(RMSD) relating 245 C
a
atoms in the V
H
and V
L
domains of the
free and bound Fabs (
Figure 3
C).
We found three distinct differences between the structures of
a VRC-PG05 Fab-monomeric CNE55 gp120 core and the SF12-
Env trimer complexes (
Zhou et al., 2018
). First, the longer CDRH3
of SF12 was extended in a different conformation from that of the
shorter CDRH3 in VRC-PG05, resulting in a RMSD of 3.1 A
̊
across 130 C
a
atoms when superposing the V
H
domains from
the Fab-bound structures (
Figure 3
C). Second, the SF12
CDRL1 and CDRL3 loops adopted conformations different
from those of their VRC-PG05 counterparts (
Figure S4
). In the
SF12 Fab, the CDRH3 and CDRL loops form a groove at the
Fab-antigen interface that accommodates the N448
gp120
glycan,
which contrasts with the wedge between the VRC-PG05 CDRH3
and CDRL1 loops that penetrates through Env glycans (
Figures
3
D and 3E). Third, the orientation of the SF12 Fab differed from
that of VRC-PG05 Fab, with the SF12 Fab exhibiting an almost
perpendicular binding angle to the silent face epitope compared
with the VRC-PG05 orientation (
Figures 3
F and
S4
). To evaluate
this difference, we calculated the rotation and translation of the
V
H
-V
L
domains of the Fab portions of the SF12-Env trimer and
VRC-PG05-gp120 complex structures, finding that the orienta-
tions of the SF12 and VRC-PG05 V
H
-V
L
domains differed, with
the axis of the SF12 Fab at a steeper angle (by
71
) to the silent
face epitope than the axis of VRC-PG05. Despite differences
in approach angles to the silent face epitope, SF12 and
VRC-PG05 shared a common mode of interaction with the
N448
gp120
glycan, mediated in each case by their CDRH3 loops
(
Figure 3
G). We conclude that SF12 binds a VRC-PG05-related
epitope with a different angle of approach and an altered mode
of recognition from VRC-PG05.
SF12 Recognizes a Mostly Glycan-Focused Epitope on
HIV-1 Env
In contrast to the non-natively glycosylated monomeric CNE55
gp120 core that was complexed with VRC-PG05 Fab (
Zhou
et al., 2018
), the relatively high resolution cryo-EM structure of
the natively glycosylated SF12-Env trimer complex allowed
modeling of N-linked glycans in the B41 Env trimer (
Figure 3
).
Given the asymmetric Fab binding in our complex, we character-
ized the SF12 epitope and paratope using a gp140 protomer in
which the SF12, but not the 10-1074, Fab was bound (
Figures
4
A–4C). Consistent with differing binding angles and CDR loop
conformations, SF12’s footprint on Env differed from that of
Immunity
50
, 1513–1529, June 18, 2019
1517
AB
CDE
FG
Figure 3. Structural Overview of the SF12-B41-10-1074 complex
(A and B) Side-view (A) and top-view (B) of the final 3.3 A
̊
single-particle cryo-EM reconstruction of the SF12-B41-10-1074 complex colored by components (dark
gray, gp41; light gray, gp120; magenta, SF12 V
H
; pink, SF12 V
L
; blue, 10-1074 V
H
; light blue, 10-1074 V
L
; cyan, N-glycans).
(C) Superposition of V
H
-V
L
domains (235 C
a
atoms) of unliganded SF12 (orange), Env-bound SF12 (magenta), and core gp120-bound VRC-PG05 (green) Fabs,
showing differences in CDR conformations between SF12 and VRC-PG05.
(D) Surface representation of SF12 (magenta/pink) and VRC-PG05 (green/pale green) Fabs illustrating differences in CDRL1 and CDRH3 loop conformat
ions.
(E) Surface representation of Env-bound SF12 Fab showing interactions with the N262
gp120
(pale blue), N295
gp120
(pale green) and N448
gp120
(red) glycans at the
SF12-Env interface. Cryo-EM density for individual glycans is shown contoured at 6
s
.
(F) Comparison of V
H
-V
L
domain orientations of SF12 (magenta/pink; cartoon) and VRC-PG05 (green/pale green; surface). The V
H
-V
L
domain orientation of SF12
on Env trimer is related by a 71
rotation and 0.5 A
̊
translation to the VRC-PG05 variable domains after alignment against gp120 (gray; surface).
(G) Overlay of CDRH3 loops of SF12 (magenta) and VRC-PG05 (green) after alignment of bound gp120s illustrates CDRH3-mediated recognition of the N448
gp120
glycan (red; sticks) by both antibodies.
See also
Figures S2
,
S3
, and
S4
and
Tables S3
and
S4
.
1518
Immunity
50
, 1513–1529, June 18, 2019
VRC-PG05, such that interactions with both N-linked glycan and
peptide components mapped almost exclusively to the SF12
heavy chain (
Figures 4
A–4C). For example,
7% of the buried
epitope surface resulted from interactions with SF12’s light chain
(
1,843 A
̊
2
buried surface area (BSA) of epitope against SF12
heavy chain versus
135 A
̊
2
against the SF12 light chain;
Table
S5
), compared with
36% of buried epitope surface at the VRC-
PG05 light chain interface. This difference is likely due to the
longer CDRL1 and L3 loops on VRC-PG05, which penetrate
the glycan-rich epitope (
Figures 3
C and 3D).
In the Env protomer used for epitope analysis, we interpreted
densities for an ordered GlcNAc
2
Man
7
at N262 gp
120
, a GlcNAc
2
Man
6
at N295
gp120
, and a GlcNAc
2
Man
5
at N448
gp120
(
Figures
4
D–4F). Similar to VRC-PG05, the N262
gp120
, N295
gp120
, and
N448
gp120
glycans constituted
75% of the epitope surface
(
Table S5
), although comparisons must be interpreted cautiously
A
B
DEF
C
Figure 4. Details of SF12 Epitope and Glycan Recognition
(A) Sequence of SF12 variable domains with antibody regions annotated using IMGT sequence analysis (CDR loops are bracketed). SF12 residues that con
tact N-
linked glycans are in blue (N262
gp120
), green (N295
gp120
), and red (N448
gp120
), while gp120-contacting residues are boxed. Contacting residues in the SF12
paratope and epitope were defined as two residues containing any atom within 4 A
̊
of each other.
(B) Structure of a SF12-B41 gp120 protomer from the trimer complex, showing paratope residues as spheres (inset). Color scheme is the same as in (A).
(C) Surface representation of B41 trimer, with SF12 epitope highlighted in magenta.
(D–F) Stick representation of residue level contacts for N262
gp120
(D), N295
gp120
(E), and N448
gp120
(F) glycans. Potential hydrogen bonds are shown as black
dashes. Cryo-EM density maps contoured at 6
s
are shown for individual glycans.
See also
Figure S4
and
Table S5
.
Immunity
50
, 1513–1529, June 18, 2019
1519
due to (1) the use of a high mannose-only monomeric gp120 core
for the VRC-PG05 complex structure (
Zhou et al., 2018
) versus a
natively glycosylated native-like Env trimer for the SF12 complex
structure, and (2) the higher resolution (2.4 A
̊
) of the VRC-PG05-
gp120 crystal structure than that of the SF12-Env trimer cryo-EM
structure (3.3 A
̊
). Mapping key residues involved in SF12-Env
interactions identified determinants of glycan recognition medi-
ated by specific regions in the SF12 paratope. For example,
SF12’s CDRH1 and framework region 3 (FWR3) interacted exclu-
sively with the N295
gp120
glycan, while CDRH2 solely engaged
the N262
gp120
glycan (
Figures 4
D and 4E). SF12 interactions
with the N448
gp120
glycan were mediated mainly by CDRH3,
with additional contacts observed with the light chain CDRL1
and CDRL3 loops (
Figure 4
F).
Because the frequency of SHM in SF12 is lower than typical
for many HIV-1 bNAbs (
Figure 1
E;
Table S1
), we analyzed the
contributions of mutated amino acid residues in the SF12 para-
tope to epitope recognition. Of the 17 V gene segment-en-
coded residues that contact the epitope, 9 arose through
SHM, including an insertion in CDRH2 (
Figure S5
A). Consistent
with glycans comprising most of the SF12 epitope, SHMs were
mostly observed for residues at the antibody-glycan interface.
However, unlike VRC-PG05, where SHMs mainly resulted in
the removal of bulky tyrosine residues to accommodate gly-
cans (
Zhou et al., 2018
), SF12 utilized tyrosines, as well as
bulky hydrophilic residues, to facilitate interactions with the
A
B
D
C
Figure 5. SF12 Engages Two Distinct
Regions of gp120 Peptide Epitope
(A) Stick representation of SF12 CDRH3 (magenta)
and gp120 (gray) contacts at the SF12-Env inter-
face. Trp100D
HC
inserts into a hydrophobic
pocket (inset) stabilized by potential hydrogen
bond interactions (black dashes) with neighboring
residues.
(B) Stick representation of SF12 CDRH1 and H2
residues (magenta) contacting gp120 residues
(gray). Potential hydrogen bonds are shown as
black dashes. Density maps for SF12 and gp120
residues are shown as magenta and gray meshes,
respectively, contoured at 8
s
.
(C) Comparison of SF12 and VRC-PG05 neutrali-
zation breadth for different viral characteristics.
The red dashed line indicates neutralization
breadth for SF12 (62%) and VRC-PG05 (27%)
against a cross-clade panel.
(D) Modeling of the N442
gp120
glycan from clade C
426c SOSIP trimer (teal; PDB: 6MYY) was ach-
ieved by aligning gp120 coordinates from the two
structures. Potential clashes with SF12 heavy
chain (magenta) regions are highlighted.
See also
Figure S5
.
glycopeptide epitope (
Figure S5
). This
demonstrates that SHMs adding bulky
residues to the paratopes of antibodies
against the glycan-rich silent face of
HIV-1 Env are not necessarily an imped-
iment to broad and potent neutraliza-
tion by these antibodies (
Figures 4
D–4F
and
S5
A).
The protein component of the SF12 epitope (
25% of the
epitope surface) mapped to two regions of gp120 (
Figures 5
and
S4
). The first region involved residues from the gp120
b
4
and
b
7 strands and the N terminus of gp120 that were engaged
by regions of CDRH3 that penetrated the glycan shield (
Fig-
ure 5
A). In this interaction, SF12 CDRH3 residues R100C
HC
and D100B
HC
formed potential hydrogen bonds with Env resi-
dues K59
gp120
and R252
gp120
, respectively (
Figure 5
A). These
interactions contributed to the formation of a hydrophobic
pocket on gp120 into which SF12 residue Trp100D
HC
inserted,
shielding this exposed hydrophobic residue at the tip of the
CDRH3 loop (
Figure 5
A, inset). The second protein component
of the SF12 epitope resembled part of the VRC-PG05 gp120
protein epitope, involving residues from gp120
b
12 and
b
22
that interacted with the SF12 CDRH1 and H2 loops (
Figure 5
B).
In this region, SF12 utilized aspartates at positions 30
HC
and
54
HC
to mediate contacts with gp120 residues N295
gp120
,
R444
gp120
, and S446
gp120
(
Figure 5
B). Unlike VRC-PG05,
SF12 directly engaged the protein component of the epitope,
forming extensive hydrogen bonds with surrounding residues.
The increased epitope surface area contributed by gp120 pep-
tide components (
25% for SF12 versus
12% for VRC-PG05)
likely contributes to the observed differences in neutralization
potency and breadth for the two antibodies (
Figure 1
E). Overall,
our structure of SF12 bound to a natively glycosylated Env
trimer allows detailed insights into SF12 Env-glycan interactions
1520
Immunity
50
, 1513–1529, June 18, 2019