| Virology |
Full-Length Text
Short CDRL1 in intermediate VRC01-like mAbs is not
sufficient
to
overcome key glycan barriers on HIV-1 Env
Parul Agrawal,
1
Maria L. Knudsen,
1
Anna MacCamy,
1
Nicholas K. Hurlburt,
1
Arineh Khechaduri,
1
Kelsey R. Salladay,
1
Gargi M. Kher,
1
Latha Kallur Siddaramaiah,
1
Andrew B. Stuart,
1
Ilja Bontjer,
2,3
Xiaoying Shen,
4
David
Montefiori,
4
Harry B. Gristick,
5
Pamela J.
Bjorkman,
5
Rogier W. Sanders,
2,3,6
Marie Pancera,
1,7
Leonidas Stamatatos
1,8
AUTHOR AFFILIATIONS
See
affiliation
list on p.
17
.
ABSTRACT
VRC01-class broadly neutralizing antibodies (bnAbs) have been isolated
from people with HIV-1, but they have not yet been elicited by vaccination. They are
extensively somatically mutated and sometimes accumulate CDRL1 deletions. Such
indels may allow VRC01-class antibodies to accommodate the glycans expressed on a
conserved N276 N-linked glycosylation site in loop D of the gp120 subunit. These glycans
constitute a major obstacle in the development of VRC01-class antibodies, as unmutated
antibody forms are unable to accommodate them. Although immunizations of knock-in
mice expressing human VRC01-class B-cell receptors (BCRs) with
specifically
designed
Env-derived immunogens lead to the accumulation of somatic mutations in VRC01-class
BCRs, CDRL1 deletions are rarely observed, and the elicited antibodies display narrow
neutralizing activities. The lack of broad neutralizing potential could be due to the
absence of deletions, the lack of appropriate somatic mutations, or both. To address
this point, we
modified
our previously determined prime-boost immunization with a
germline-targeting immunogen nanoparticle (426c.Mod.Core), followed by a heterolo
gous core nanoparticle (HxB2.WT.Core), by adding a
final
boost with a cocktail of various
stabilized soluble Env trimers. We isolated VRC01-like antibodies with extensive somatic
mutations and, in one case, a seven-amino acid CDRL1 deletion. We generated chimeric
antibodies that combine the vaccine-elicited somatic mutations with CDRL1 deletions
present in human mature VRC01 bnAbs. We observed that CDRL1 indels did not improve
the neutralizing antibody activities. Our study indicates that CDRL1 length by itself is not
sufficient
for the broadly neutralizing phenotype of this class of antibodies.
IMPORTANCE
HIV-1 broadly neutralizing antibodies will be a key component of an
effective
HIV-1 vaccine, as they prevent viral acquisition. Over the past decade, numer
ous broadly neutralizing antibodies (bnAbs) have been isolated from people with HIV.
Despite an in-depth knowledge of their structures, epitopes, ontogenies, and, in a few
rare cases, their maturation pathways during infection, bnAbs have, so far, not been
elicited by vaccination. This necessitates the
identification
of key obstacles that prevent
their elicitation by immunization and overcoming them. Here we examined whether
CDRL1 shortening is a prerequisite for the broadly neutralizing potential of VRC01-class
bnAbs, which bind within the CD4 receptor binding site of Env. Our
findings
indicate that
CDRL1 shortening by itself is important but not
sufficient
for the acquisition of neutrali
zation breadth, and suggest that particular combinations of amino acid mutations, not
elicited so far by vaccination, are most likely required for the development of such a
feature.
KEYWORDS
HIV-1, VRC01-class antibodies, CDRL1, neutralization, BCR sequencing
Month XXXX Volume 0
Issue 0
10.1128/jvi.00744-24
1
Editor
Viviana Simon, Icahn School of Medicine at
Mount Sinai, New York, New York, USA
Address correspondence to Leonidas Stamatatos,
lstamata@fredhutch.org.
The authors declare no
conflict
of interest.
See the funding table on p.
18
.
Received
26 April 2024
Accepted
12 July 2024
Published
6 September 2024
Copyright © 2024 Agrawal et al. This is an open-
access article distributed under the terms of the
Creative Commons Attribution 4.0 International
license
.
Downloaded from https://journals.asm.org/journal/jvi on 16 September 2024 by 131.215.250.65.
V
RC01-class broadly neutralizing HIV-1 antibodies have been isolated from several
people living with HIV-1 infected by
different
HIV-1 viruses and bind a
well-defined
epitope within the CD4-binding site (CD4-BS) (
1
–6
). Antibodies in this class are all
derived from the pairing of VH1-2*02 heavy chains (HCs) with a limited number of
light chains (LCs) expressing rare
five-amino
acid (5-aa) CDRL3s (
7
). VRC01-class bnAbs
isolated during infection (referred to as “mature,” m, antibodies) are extensively mutated
in both their HCs and LCs and can be up to 50% divergent in their amino acid sequen
ces. Nevertheless, they share similar structures and engage their epitope in a similar
manner (
4
–6
). Contrary to most other anti-CD4-BS antibodies, they recognize the CD4-BS
primarily through their gene-encoded CDRH2 domains (
5,
6,
8,
9
), although their CDRH3s
also contribute to this interaction (
10,
11
). VRC01-class bnAbs have been shown to
prevent infection of humanized mice by HIV-1 and of non-human primates by S(H)IV
(
12
–14
). Importantly, the
first
VRC01-class antibody isolated (VRC01) prevented HIV-1
acquisition from susceptible viruses in two phase 3 clinical trials (HVTN 703/704) (
15
).
Although the VRC01-class bnAbs isolated from HIV-1-infected individuals bind to
diverse HIV-1 Env proteins and neutralize diverse HIV-1 isolates, their unmutated
precursors (referred to as “germline,” gl) do neither (
16
–20
). In fact, Envs derived from
circulating viruses capable of binding glVRC01-class antibodies are presently not known.
As a result, during immunizations with such Envs, naïve B cells expressing glVRC01-class
B-cell receptors (BCRs) do not become activated (
21
), and hence, VRC01-class bnAbs have
not yet been elicited through vaccination.
The key obstacles preventing glVRC01-class antibodies from binding Envs are as
follows: (i) glycans expressed on the conserved N-linked glycosylation site (NLGS) at
position 276 (in loop D), (ii) glycans present on the V5 region of Env, and (iii) the length
and glycosylation of the V1–V3 Env regions (
16,
19,
20,
22
). Consequently,
specifically
designed Env-derived proteins capable of engaging glVRC01-class antibodies (“gl-target
ing” Envs) and their corresponding BCRs have been generated (
16,
19,
20,
22,
23
).
As wild-type (WT) animal species do not express orthologs of the human VH1-2*02
allele (
7
), transgenic mice engineered to express
different
forms of glVRC01-class BCRs
have been developed and are routinely employed to test immunogens and immuniza
tion schemes that would elicit VRC01-class bnAbs (
19,
24
–32
). One such knock-in (KI)
mouse is heterozygous for the gl HC of VRC01 mAb (VRC01
glHC
) but expresses diverse
endogenous mouse LCs (
28
). In total, approximately 0.08% of naïve B cells in these mice
express glVRC01-like BCRs (
28,
33,
34
). We have reported that a prime immunization
with the gl-targeting clade C 426c.Mod.Core immunogen (expressed as self-assembling
nanoparticles) followed by a booster immunization with self-assembling nanoparticles
expressing the heterologous clade B HxB2.WT.Core immunogen (which does not activate
naïve glVRC01 B cells) results in the elicitation of partially mutated VRC01-like B cells
and antibody responses in these KI mice (
30
–32
). The vast majority of these antibodies
express the mouse
κ8–
30*01, which has a CDRL1 of 17 aa in length. In contrast, mVRC01-
class bnAbs generated during HIV-1 infection typically have indels in their CDRL1 (
6
).
A shorter CDRL1 (7- to 11-aa long) is believed to be the reason why mVRC01-class
bnAbs can bypass the glycans present at the conserved position N276 of the gp120
subunit and neutralize diverse HIV-1 strains. Indeed, CDRL1
modifications
(along with
modifications
in LFWR3 and LCDR3, associated with adaptations to N276 glycans and
Loop D contacting) appear toward the end of the maturation process of VRC01-class
antibodies during HIV-1 infection (
35
).
Here, we employed a prime-boost immunization schema in the VRC01
glHC
mouse
model that led to the isolation of a VRC01-like antibody whose κ8–30*01 LC had acquired
a CDRL1 indel. The neutralizing potential of this antibody, however, was not broad and
did not lead to neutralization of tier 2 HIV-1 viruses. In parallel, by replacing the long
CDRL1 of partially mutated VRC01-like antibodies elicited during this immunization with
the much shorter CDRL1 present on the human mVRC01-class bnAb VRC01 (mVRC01),
we
confirmed
that CDRL1 indels did not improve their neutralizing potentials. Collec
Full-Length Text
Journal of Virology
Month XXXX Volume 0
Issue 0
10.1128/jvi.00744-24
2
Downloaded from https://journals.asm.org/journal/jvi on 16 September 2024 by 131.215.250.65.
tively, therefore, our data indicate that a shorter CDRL1 is important but not
sufficient
for
improving the neutralizing properties of these elicited VRC01-like antibodies.
RESULTS
Serological analysis during immunization
A prime immunization of VRC01
glHC
KI mice with the clade C-derived 426c.Mod.Core
self-assembling nanoparticles (see Materials and Methods and [
19
] for details) followed
by a boost immunization with the clade B-derived HxB2.WT.Core self-assembling
nanoparticles results in the development of partially mutated VRC01-like antibodies that
can neutralize the autologous 426c virus whose Env
artificially
lacks the N276 NLGS (426c
single mutant, 426c.SM) but not the 426c virus expressing the WT fully glycosylated
Env (426c.WT) (
31
). To improve the maturation of these antibodies, here, we examined
whether adding a second boost immunization (following the HxB2.WT.Core immuniza
tion), with a cocktail of 11 heterologous soluble, stabilized Env trimers (SOSIP) (Table S1),
will lead to the elicitation of VRC01-like antibodies that more
efficiently
accommodate
the N276-associated glycans on functional Envs (
Fig. 1A
). One reason for employing
SOSIP immunogens as
final
boost was that SOSIPs may select VRC01-like BCRs with
mutations that allow them to engage the VRC01 epitope as present on functional Envs.
Another reason for using SOSIP booster immunogens after the two immunizations with
core gp120s was to avoid the activation of B cells targeting epitopes present only on the
cores, which are occluded in native Envs.
In agreement with our previous observations, immunization with 426c.Mod.Core
nanoparticles elicited robust autologous plasma antibody responses (
Fig. 1B
, blue line;
Fig. S1A
), the majority of which targeted the CD4-BS on that Env (
Fig. S1B
, blue line) as
demonstrated by the reduced binding to the CD4-BS knock-out (KO) version (D368R/
E370A/D279A) of this protein (
Fig. 1B
). These antibodies also recognized the heterolo
gous glVRC01-targeting protein eOD-GT8 ((
16
);
Fig. 1B
, green line;
Fig. S1A
) in a CD4-BS-
dependent manner (
Fig. S1B
, green line). All animals generated durable anti-
HxB2.WT.Core plasma antibody responses (
Fig. 1B
, red line;
Fig. S1A
), a fraction of which
targeted the CD4-BS on that Env (
Fig. S1B
, red line). The boost immunization with
HxB2.WT.Core nanoparticles increased the plasma antibody responses against all
proteins tested (
Fig. 1B
, post week 10), in a CD4-BS dependent manner (
Fig. 1C
, open
bars). Our second heterologous boost with the SOSIP cocktail, however, did not result in
an apparent increase in the plasma antibody responses to these proteins (
Fig. 1B
, post
week 18), including the CD4-BS fraction (
Fig. 1C
,
filled
bars).
Characterization of Env+ B-cell receptors isolated after the
final
immuniza
tion
Two weeks post
final
immunization, class-switched memory IgG+ Env + B cells were
isolated from the spleens of the immunized animals and individually sorted. To isolate
Abs capable of recognizing the VRC01 epitope within the trimeric HIV-1 Env spike,
tetramers of 426c.TM.SOSIP (which lacks the Loop D N276 (S278A), and V5 N460 (T462A)
and N463 (T465A) NLGSs (see Materials and Methods for additional information and
[
20
]), along with eOD-GT8 and eOD-GT8.KO (D276N/W277F/R278T/D279A/D368R), to
exclude non-CD4-BS B cells, were employed as baits during B-cell sorting. Of note,
426c.TM.SOSIP was not part of the immunogens in the SOSIP cocktail employed during
the
final
boost. B cells that were 426c.TM.SOSIP+/eOD-GT8+/eOD-GT8.KO
− were isolated
and their HC/LC genes sequenced.
A total of 162 Env+ class-switched memory B cells were sorted, and 52 HCs were
successfully sequenced, of which 35 (67%) expressed the human inferred glVRC01 HC
(
Fig. 2A
). Of the 67 LCs that were successfully sequenced, 19 (28%) contained the 5-aa
long CDRL3s (
Fig. 2B
), consistent with our and other’s (
24,
28,
31,
32
) previous observa
tions. The majority of LCs (95%, 18 of 19 5-aa-long CDRL3 containing sequences) were
derived from the mouse
κ8–
30*01 LC V gene (
Fig. 2C
, shown in pink). The non-
κ8–30*01
Full-Length Text
Journal of Virology
Month XXXX Volume 0
Issue 0
10.1128/jvi.00744-24
3
Downloaded from https://journals.asm.org/journal/jvi on 16 September 2024 by 131.215.250.65.
LC with a 5-aa-long CDRL3 was derived from the 12–46*01 V-gene (
Fig. 2C
, shown in
blue). Interestingly, this LC was 9.7% somatically mutated and paired with VH1-2*02 HC.
In total, 61% (11 of 18) of the VH1-2*02 HC sequences that were paired with LCs
expressing 5-aa-long CDRL3s contained an asparagine at position 35 instead of the
gene-encoded histidine (H35N, in CDRH1,
Fig. 2D
). This H35N mutation introduces an
additional hydrogen bond with N100a in CDRH3 and increases the stability of the
interaction between CDRH1 and CDRH3 during the maturation of VRC01-class antibodies
(
28
). In total, 53% (nine of 17) of the paired
κ8–
30*01-derived 5-aa-long CDRL3s con
tained a glutamic acid at position 96 (Glu
96,
Fig. 2E
), which is a key feature of mVRC01-
class antibodies and forms a hydrogen bond with Gly459 in gp120 at the amino terminus
of the V5 region (
6,
7,
36
).
The mean number of HC aa mutations in VH1-2*02 sequences that were paired with
LCs expressing 5-aa-long CDRL3 was approximately 8.6, and the mean number of the
corresponding LC aa mutations was approximately 7.3 (
Fig. 2F
). We previously reported
that after the HxB2.WT.Core booster immunization, the mean number of aa mutations
is approximately 6.6 and approximately 5.4 in the HCs and LCs, respectively (
31
). Thus,
the immunization with the cocktail of SOSIPs increased the number of somatic mutations
in VRC01-like BCRs. Importantly, despite the accumulation of somatic mutations, the
three key amino acids, Trp
50HC
, Asn
58HC
, and Arg
71HC
, that make critical contacts with
FIG
1
Antibody responses 2 weeks following the
final
boost immunization. (
A
) Mice (
n
= 5) were primed with 426c.Mod.Core ferritin nanoparticles with GLA-LSQ
at week 0, followed by adjuvanted HxB2.WT.Core ferritin nanoparticle boost at week 10, and adjuvanted SOSIP cocktail as the
final
boost at week 18. Mice were
bled at the indicated time points (red circles) and sacked for blood, spleen, and lymph node tissues, at week 20 (green circle). (
B
) Plasma was assayed by ELISA
for binding against 426.Mod.Core (blue solid line), eOD-GT8 (green solid line), HxB2.WT.Core (red solid line), as well as their corresponding antigens with CD4-BS
or VRC01 epitope knock-out (KO) (dotted lines); mean endpoint titers against the indicated proteins with SEM values are shown over time. Black dotted lines
indicate the time of the booster immunizations. (
C
)
CD4-BS-specific
EC50 values against the indicated proteins are shown post each boost where open bars
represent response post Boost 1 and
filled
bars, post Boost 2. See also Table S1 and
Fig. S1
.
Full-Length Text
Journal of Virology
Month XXXX Volume 0
Issue 0
10.1128/jvi.00744-24
4
Downloaded from https://journals.asm.org/journal/jvi on 16 September 2024 by 131.215.250.65.
the CD4-BS of Env (
7
), remained unaltered, similar to what occurs during the
affinity
maturation process of VRC01-class bnAbs during HIV-1 infection (
4,
5
). Also, the fact that
a range of mutations accumulated in the HCs and LCs of these BCRs (
Table S2
) indicates
that B cells that express variations of the VRC01-like BCRs co-evolved in these animals
following the SOSIP cocktail booster immunization.
Env-binding properties of monoclonal VRC01-like antibodies isolated after
the
final
immunization
To determine whether
differences
in aa mutations result in functional antibody
differences,
we generated IgGs from paired HC/LC sequences and investigated their
binding and neutralizing properties. In total, we characterized 17 VRC01-like
antibodies
.
FIG
2
Heavy-chain/light-chain sequence analysis after the
final
boost immunization at week 20. Pie charts indicate HC (
A, D
) and LC (
B, C, E
) characteristics
from individually sorted B cells from pooled mouse samples collected 2 weeks post
final
immunization. The number of HC and LC sequences analyzed is shown
in the middle of each pie chart. (
A
) VH-gene usage, (
B
) aa length of the CDRL3 domains in the LC, and (
C
) LC-gene usage, where shades of gray/black slices
represent non 5-aa-long CDRL3s. (
D–F
) Analysis of paired sequences: HCs with the H35N mutation are shown in (
D
), presence or absence of Glu
96
LC within the LC
sequences with 5-aa-long CDRL3 domains are shown in (
E
), and (
F
) number of amino acid changes in the HC and LC of paired sequences at week 20, where each
circle represents a paired sequence (see also
Table S2
).
Full-Length Text
Journal of Virology
Month XXXX Volume 0
Issue 0
10.1128/jvi.00744-24
5
Downloaded from https://journals.asm.org/journal/jvi on 16 September 2024 by 131.215.250.65.
Their VH and VL sequences are shown in
Table S2
, and their ontogenies are shown in
Table S3. Sixteen antibodies expressed
κ8–
30*01-derived LCs with 5-aa-long CDRL3s,
and one antibody (8-7) expressed the
κ12–
46*01 derived LC with 5-aa-long CDRL3s, as
discussed above. Interestingly, mAb 8–27 with
κ8–
30*01-derived LC had a 7-aa-long
deletion in its CDRL1 domain (
Table S2
). The Env-binding properties of these 17
VRC01-like antibodies were assessed using biolayer interferometry (BLI) (
Fig. 3
;
Table
S4
).
We
first
confirmed
that the antibodies recognize the CD4-BS by determining their
binding to the 426c.Mod.Core Env and its CD4-BS KO version (
Fig. 3A
, top panel). The
human mVRC01 and glVRC01 mAbs served as internal controls (
Fig. 3
, red, solid, and
dotted lines). With the exception of 8–9 and 8–11 (data not shown), all mAbs bound
426c.Mod.Core, with 8–7 (in black) showing the fastest
off-rate
(
Fig. 3A
, top left panel),
and as expected, none displayed any reactivity to 426c.Mod.Core.KO (
Fig. 3A
, top right
panel). We next examined binding to eOD-GT8 (that only contains the outer domain of
gp120 and lacks the N276 NLGS [
16,
28
]) and its CD4-BS KO version. All mAbs (including
8–9 and 8–11; data not shown) bound eOD-GT8 and not eOD-GT8.KO (
Fig. 3A
, bottom
left and right panels, respectively). These results
confirmed
that the mAbs recognize the
CD4-BS in the absence of glycans in Loop D (N276) and V5, but sequence variations
affect
their binding properties.
We next examined the binding of these VRC01-like mAbs to the autologous and
various heterologous fully glycosylated gp120 cores (
Fig. 3B
). These proteins, described
as WT.Cores, lack the V1, V2, and V3 domains (as well as the N and C termini of gp120),
but are otherwise fully glycosylated, including at position N276 in Loop D and at the
NXT/S sequon(s) in the V5 loop. A summary of the antibody recognition properties is
shown in
Table S4
. All 15 mAbs bound the 426c.WT.Core with 8–7 and 8–26 showing the
fastest
off-rate,
and 12 mAbs bound the heterologous HxB2.WT.Core, with mAbs 8–18, 8–
19, and 8–26 not binding whereas mAbs 8–7, 8–20, and 8–23 showing very weak
binding. Six of the 15 mAbs bound QH0692.WT.Core (clade B) with 8–24 displaying the
slowest
off-rate.
Nine of the 15 mAbs bound 4501dH1.WT.Core (clade B) with 8–24
displaying the slowest
off-rate.
Four of the 15 mAbs bound Q168a2.WT.Core (clade A), of
which only three bound the 93TH057.WT.Core (clade A/E), with 8–24 showing the
strongest binding to all Envs (
Fig. 3B
). We conclude, therefore, that these VRC01-like
antibodies can bypass the N276- and V5-associated glycans on autologous and heterolo
gous gp120s lacking the variable domains 1–3, something that, as discussed above, the
human glVRC01-class antibodies are unable to do (
Fig. 3B
, red dotted line). These results
indicate that a heterogenous population of VRC01-like B cells co-exists in these immu
nized animals, with
different
CD4-BS recognition properties.
We next examined whether these antibodies could bind the VRC01 epitope on SOSIP
proteins (i.e., in the presence of appropriately positioned V1–V3 domains) with and
without NLGS at position N276 in Loop D and/or in V5 (
Fig. 3C
). All antibodies bound
426c.TM.SOSIP (N276-/N460-/N463-) (
Fig. 3C
, top left panel), indicating that these
antibodies can bind in the presence of well-ordered V1–V3 loops when the Loop D and
V5 NLGS are unoccupied. All antibodies also bound 426c.SM.SOSIP (N276-) (
Fig. 3C
, top
right panel), indicating that they can bind in the presence of V1–V3 even when the V5
NLGS are present. While many mAbs were not able to bind to Env trimers that had the
N276 glycan, mAbs 8–10, 8–21, 8–24, and 8–27 showed binding to a version of
426c.SOSIP that lacks the V5 NLGS (N460 and N463) but expresses the N276 NLGS in
Loop D (426c.DM.SOSIP) (
Fig. 3C
, bottom left panel), with 8–10, 8–24, and 8–27 being the
better binders. These data
confirm
that N276 poses the main obstacle for the maturing
VRC01-like antibodies and also show that a subset of mAbs are able to partially over
come that obstacle. One antibody, 8–27, displayed binding (albeit very weak) to the fully
glycosylated 426c.WT.SOSIP in this assay (
Fig. 3C
, bottom right panel). Collectively, these
results indicate that most VRC01-like antibodies isolated at this stage of immunization
can bypass the V1–V3 and N276 NLGS-associated glycans, as long as the V5 NLGS are
unoccupied.
Full-Length Text
Journal of Virology
Month XXXX Volume 0
Issue 0
10.1128/jvi.00744-24
6
Downloaded from https://journals.asm.org/journal/jvi on 16 September 2024 by 131.215.250.65.
FIG
3
Binding properties of VRC01-like mAbs generated after the
final
boost immunization. (
A
) Fifteen VRC01-like mAbs were evaluated against the indicated
soluble monomeric Envs. 8-7 mAb with fastest
off
rate for 426c.Mod.Core and eOD-GT8 is shown in black. (
B
) Binding of 15 mAbs against the indicated
autologous and heterologous WT.Cores, where the mAbs that displayed cross-reactivity are shown in
different
colors. (
C
) Binding of 15 mAbs against the
indicated variants of 426c.SOSIP are shown. mVRC01 (solid red line) and glVRC01 (dotted red line) were included as internal controls in all assays. Black dotted
lines indicate end of association and dissociation phases (see also Tables S3 and S4).
Full-Length Text
Journal of Virology
Month XXXX Volume 0
Issue 0
10.1128/jvi.00744-24
7
Downloaded from https://journals.asm.org/journal/jvi on 16 September 2024 by 131.215.250.65.