A nonself sugar mimic of the HIV glycan shield shows enhanced antigenicity

Article abstract of DOI:10.1073/pnas.1002717107

Antibody 2G12 uniquely neutralizes a broad range of HIV-1 isolates by binding the high-mannose glycans on the HIV-1 surface glycoprotein, gp120. Antigens that resemble these natural epitopes of 2G12 would be highly desirable components for an HIV-1 vaccine. However, host-produced (self)-carbohydrate motifs have been unsuccessful so far at eliciting 2G12-like antibodies that cross-react with gp120. Based on the surprising observation that 2G12 binds nonproteinaceous monosaccharide D-fructose with higher affinity than D-mannose, we show here that a designed set of nonself, synthetic monosaccharides are potent antigens. When introduced to the terminus of the D1 arm of protein glycans recognized by 2G12, their antigenicity is significantly enhanced. Logical variation of these unnatural sugars pinpointed key modifications, and the molecular basis of this increased antigenicity was elucidated using high-resolution crystallographic analyses. Virus-like particle protein conjugates containing such nonself glycans are bound more tightly by 2G12. As immunogens they elicit higher titers of antibodies than those immunogenic conjugates containing the self D1 glycan motif. These antibodies generated from nonself immunogens also cross-react with this self motif, which is found in the glycan shield, when it is presented in a range of different conjugates and glycans. However, these antibodies did not bind this glycan motif when present on gp120.

Full text of DOI:10.1073/pnas.1002717107

A nonself sugar mimic of the HIV glycan
shield shows enhanced antigenicity
Katie J. Dooresa,b,f,h,1, Zara Fultonc,1, Vu Hongd,e, Mitul K. Patela, Christopher N. Scanlang, Mark R. Wormaldg,
M. G. Finnd,e, Dennis R. Burtonb,f,h, Ian A. Wilsonc,e,2, and Benjamin G. Davisa,2
aDepartment of Chemistry, University of Oxford, Chemistry Research Laboratory, Mansfield Road, Oxford, United Kingdom, OX1 3TA; bDepartments of
f
Immunology and Microbial Science, cMolecular Biology, and dChemistry, eSkaggs Institute for Chemical Biology, International AIDS Vaccine Initiative
Neutralising Antibody Center, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037; gGlycobiology Institute, Department of
Biochemistry, University of Oxford, South Parks Road, Oxford, United Kingdom, OX1 3QU; and hRagon Institute of Massachusetts General Hospital,
Massachusetts Institute of Technology and Harvard, Boston, MA, 02114
Edited by Carolyn R. Bertozzi, University of California, Berkeley, CA, and approved July 28, 2010 (received for review March 8, 2010)
Antibody 2G12 uniquely neutralizes a broad range of HIV-1 isolates
by binding the high-mannose glycans on the HIV-1 surface glyco-
protein, gp120. Antigens that resemble these natural epitopes of
2G12 would be highly desirable components for an HIV-1 vaccine.
However, host-produced (self)-carbohydrate motifs have been un-
successful so far at eliciting 2G12-like antibodies that cross-react
with gp120. Based on the surprising observation that 2G12 binds
nonproteinaceous monosaccharide D-fructose with higher affinity
than D-mannose, we show here that a designed set of nonself,
synthetic monosaccharides are potent antigens. When introduced
to the terminus of the D1 arm of protein glycans recognized by
2G12, their antigenicity is significantly enhanced. Logical variation
of these unnatural sugars pinpointed key modifications, and the
molecular basis of this increased antigenicity was elucidated using
high-resolution crystallographic analyses. Virus-like particle protein
conjugates containing such nonself glycans are bound more tightly
by 2G12. As immunogens they elicit higher titers of antibodies than
those immunogenic conjugates containing the self D1 glycan motif.
These antibodies generated from nonself immunogens also cross-
react with this self motif, which is found in the glycan shield, when
it is presented in a range of different conjugates and glycans.
However, these antibodies did not bind this glycan motif when
present on gp120.
“silent” carbohydrate face of gp120 (part of HIV’s glycan shield)
and escaping immune tolerance (11, 17, 18). Carbohydrates are
typically considered poor immunogens. One major limitation to
the recognition and, hence, immunogenicity of carbohydrate
structures is their exhibition of microheterogeneity. A single pro-
tein may display many variations of carbohydrate structure
(glycoforms) (19) that can result in a polyclonal and reduced
antigenic response (20). Furthermore, viruses typically rely on
host glycosylation machinery; their glycosylation patterns are,
therefore, inevitably similar to that of the host. As a result, host
immune mechanisms should normally recognize such sugars as
self and display tolerance, thus not eliciting antibodies to host-
derived sugars. Carbohydrate–protein interactions tend also to
be weaker than protein–protein interactions (μM rather that nM)
(21, 22) further reducing affinities of antibodies to glycans. Large
glycan structures can also shield underlying peptide epitopes,
thereby further reducing the effectiveness of the immunogenic
response (4).
It is therefore remarkable that, in spite of the expected poor
immunogenicity of carbohydrate epitopes, 2G12 recognizes a
conserved cluster of oligomannose residues on the silent face
of gp120 with high affinity (Fig. 1A and B) (11). The crystal struc-
ture of 2G12 (11) revealed a unique domain-exchanged dimeric
structure; the variable domains of the heavy chains (VH) swap
over to form an extended binding surface that includes a novel
2G12 ∣ gp120 ∣ oligosaccharide ∣ glycoconjugate ∣ glycomimetic
0
VH∕VH interface in addition to two conventional VH∕VL com-
IV-1 is thought to have infected up to 60 million people since
bining sites (11), and that greatly enhances the carbohydrate-
antibody affinity to the nanomolar range (11). Thus, it is the clus-
ter of high-mannose sugars on gp120 (Fig. 1A) that creates novel
epitopes that are recognized as foreign by the immune system and
are the ultimate target of the domain-exchanged 2G12. Structural
studies (11, 23), binding assays (11, 17, 18, 23, 24), enzymatic
hydrolysis (17), and mutational analysis (17) have highlighted
terminal mannosides and, in particular, the D1 arm (Fig. 1B)
as key binding motifs for 2G12. This unique binding solution
reveals a potential vaccine strategy that may be accessible by
H
its discovery over 20 years ago (1). Of those infected, more
than 20 million have died, with the vast majority of individuals
affected being from developing countries (1). An effective vaccine
is, therefore, paramount to combat the epidemic. The HIV-1
envelope spike, critical for viral infectivity, consists of a compact,
unstable trimer of the glycoproteins gp120 and gp41 (Fig. 1A) and
is a key target for design of an antibody-based HIV-1 vaccine. The
envelope spike undergoes rapid evolution in each individual pa-
tient, resulting in enormous sequence heterogeneity among indi-
vidual isolates of HIV-1 (2). Moreover, neutralization-sensitive
epitopes on gp120 and gp41 are either difficult to access or
shielded from recognition by the immune system by an extensive
display of host-derived N-glycans (3, 4). Nevertheless, a small
group of rare, broadly neutralizing antibodies (b12, 2G12, 2F5,
4E10, Z13) against gp120 and gp41 have been previously isolated
from HIV-1-infected patients that provide protection against viral
challenge in animal models (5–9) as well as the more recent
discovery of new highly potent human antibodies (PG9, PG16)
(10). Structural analyses have revealed how they broadly neutra-
lize HIV-1 and the mechanism by which the virus normally evades
detection by the immune system (8, 11–16). Identification of
antigens that could generate similar types of broadly neutralizing
antibodies is, therefore, an important step in the development of
an HIV-1 vaccine.
Author contributions: K.J.D., Z.F., C.N.S., I.A.W., and B.G.D. designed research; K.J.D., Z.F.,
V.H., M.K.P., and M.R.W. performed research; K.J.D., Z.F., M.R.W., M.F., D.R.B., I.A.W., and
B.G.D. analyzed data; M.K.P. contributed new reagents/analytic tools; and K.J.D., Z.F.,
I.A.W., and B.G.D. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
Data deposition: The coordinates and structure factors for 2G12/D-fructose, 2G12/C-6
methyl monosaccharide 10, and 2G12/C-6′″ methyl tetrasaccharide 5 crystal structures have
been deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 3OAY, 3OAZ, and
3OB0, respectively).
1K.J.D. and Z.F. contributed equally to this work.
2To whom correspondence may be addressed. E-mail: wilson@scripps.edu or ben.davis@
chem.ox.ac.uk.
From this small group, broadly neutralizing antibody 2G12 is
uniquely capable of recognizing sugars on the immunologically
This article contains supporting information online at www.pnas.org/lookup/suppl/
doi:10.1073/pnas.1002717107/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1002717107
PNAS ∣ October 5, 2010 ∣ vol. 107 ∣ no. 40 ∣ 17107–17112

Fig. 1. Carbohydrate epitopes on gp120 and man-
nose mimics. (A) Model of the gp120 component of
the trimeric HIV-1 envelope spike based on the struc-
ture of core gp120, (51) with three gp120 monomers
shown in purple, green, and pink. Carbohydrate
chains are shown in yellow. The high-mannose clus-
ter, corresponding to glycans attached to Asn295,
Asn332, Asn339, Asn386, and Asn392, which is recog-
nized by 2G12, is represented in red. Figure was
made using PyMOL (52). (B) 2G12 high-mannose car-
bohydrate epitope. (C) D-fructose as a 1-deoxy-5-hy-
droxy analogue of D-mannose. (D) Proposed nonself,
D1-arm ligands (R ¼ alkyl).
targeting these conserved oligomannose clusters on gp120 as
antigens, and that may be probed by using 2G12 as a recognition
template. The potential therapeutic value of the 2G12 epitope as
an attractive HIV-1 vaccine target has recently been highlighted
by the finding that 2G12 can protect macaques against HIV-1
challenge at remarkably lower serum neutralizing titers than most
other broadly neutralizing antibodies (25).
nonself glycans with high affinity to 2G12. We identify a unique
nonself D1 arm mimic that displays better inhibition of 2G12/
gp120 binding than the natural D1 arm, both as an isolated sugar
and in protein–sugar conjugates, and reveal the mechanism of
inhibition of this most potent monovalent 2G12 ligand known.
We also uncover the molecular mechanism by which D-fructose
achieves better inhibition of 2G12/gp120 binding than D-man-
nose. We further show that this nonself glycan can elicit higher
titers of antibodies that can react with the self glycan D1 arm
motif, a motif that is found in the glycan shield of HIV. However,
these antibodies do not neutralize HIV.
To date, the use of glycans containing the natural sugar D-man-
nose or variants in a number of elegantly designed oligosacchar-
ide constructs have failed to elicit 2G12-like antibodies (26–33).
The antibodies generated fail to cross-react with gp120 (26, 30)
and may reflect an inability to break tolerance for these host-gen-
erated self sugars. Moreover, syntheses of fragments (or modified
fragments) of the high-mannose glycans found on gp120 have
failed to reveal a monovalent ligand that binds more tightly to
2G12 than the D1 arm (Man4) or Man8–9GlcNAc2 (Fig. 1B);
indeed, modifications or truncations have created weaker binders
(26–33). Elegant studies using a three-component conjugate have
shown that reactivity toward cancer antigens containing self-
carbohydrates can be enhanced by attaching them to scaffolds
containing Toll-like receptor and related ligands that increase
antigen-presenting cell uptake and modulate cytokine responses
(34). However, the generation of antigens in which the carbohy-
drate epitope itself is successfully engineered to be both nonself
and more antigenic has not heretofore been described.
We sought to address these challenges by adopting a design
and synthesis strategy inspired by the prior unexpected observa-
tion that, from a panel of monosaccharides, D-fructose (1) binds
most strongly, some 9 times greater than D-mannose (2) itself,
to 2G12 (11). The keto-sugar fructose is absent from protein gly-
cans, yet when displayed in its pyranose form and rotated by 180°
it strongly resembles D-mannose, differing only in anomeric and
C-5 substituents (Fig. 1C). This enhanced binding affinity for
an unusual sugar not found in protein epitopes suggested that
a hybrid nonself sugar could be designed, based on the structure
of D-fructose, that might engage additional interactions (binding
to additional residues perhaps by virtue of the additional substi-
tution found in D-fructose in proximity to C-5 compared to
D-mannose) in the 2G12 binding site.
Results
Crystal Structure of Fab 2G12 in Complex with D-Fructose. To under-
stand the molecular basis for the higher affinity of 2G12 for
D-fructose over D-mannose (and Manα1-2Man) (11) and aid
our design strategy, we determined the crystal structure of Fab
2G12 in complex with the nonself monosaccharide to 1.95 Å
resolution (see SI Appendix, SI Materials and Methods and
Table S1). The electron density for D-fructose at the two primary
combining sites of the domain-exchanged Fab dimer is excellent
and readily interpretable and reveals D-fructose adopts a pyra-
nose form that does, indeed, resemble D-mannopyranose in
the 2G12 binding sites (Fig. 2A). The structure validates our pre-
vious postulate that the contacts formed by D-fructose with 2G12
would be similar to those made by the terminal mannoside in
Manα1-2Man (11) (Fig. 2A). The C-5 hydroxyl in D-fructose
forms an additional direct H bond with SerH100A O and four
additional water-mediated H bonds with AlaH31 O, LeuH100 O,
SerH100A O, and TyrL94 Oη. These additional water-mediated
interactions engaging the C-5 substituent appear to explain
why D-fructose is a better inhibitor of 2G12/gp120 binding than
D-mannose. Moreover, this enhanced network of H bonds is also
likely to be responsible for the higher affinity of Fab 2G12 for
D-fructose over Manα1-2Man because it also mimics the direct
H bond between O3 in the next mannose of the disaccharide and
AlaH31 O. A total of 173 Å2 of molecular surface on Fab 2G12
and 155 Å2 of molecular surface on D-fructose are buried in the
complex, with 10 direct and 10 water-mediated H bonds and 42
van der Waals interactions in each antigen binding site. This crys-
tal structure confirms that additional substituents around a term-
We report here the synthesis and insertion of unnatural sugars
into oligomannose constructs to break tolerance and create
17108
∣
www.pnas.org/cgi/doi/10.1073/pnas.1002717107
Doores et al.

oligosaccharides 3-7 containing substitutions and modifications
at the C-3, C-5, and C-6 positions of the terminal mannoside was
designed for synthesis (Fig. 1D).
As a first step, a larger panel of eight nonself, monosaccharide
sugars was synthesized and evaluated prior to selection of the
most potent for introduction onto the terminus of the D1 arm.
Monosaccharides 8, 9a-e, 10, and 11 (Scheme 1) were prepared
from parent sugar D-mannose via an oxidation and addition pro-
cedure that allowed late-stage diversification for rapid substituent
variation. Thus, carbonyl compounds 12, 13, and 14 were each
regioselectively protected to allow selective access to OH-3,
OH-6, and OH-5, respectively; this, in turn, allowed alteration
and substituent introduction at C-3, C-6, and C-5. Oxidation un-
der Swern conditions allowed mild preparation of these target
carbonyl intermediates. Reaction of 12, 13, and 14 with appropri-
ate organometallic nucleophiles followed by global deprotection
gave 9a-e, 10, and 11. These synthetic routes were high yielding
and allowed ready introduction of varied substituents with high
stereoselectivity (Scheme 1). Absolute configuration of stereo-
chemistry in 10 was confirmed using nOe NMR analysis of a
4,6 benzylidene derivative; nOe was also used to confirm the con-
figuration of C-3 derivative 11 (see SI Appendix).
A competitive ELISA assay was used to assess the ability of
these nonself monosaccharides to inhibit binding of 2G12 to
HIV-1 glycoprotein gp120 (Fig. 3). Of the fructose-like mono-
mers with substitution at the C-5 position (9a-e), those with large
substituent groups (vinyl 9c and ethynyl 9d) that are rotationally
restricted by hybridization were found to have much higher IC50
values than D-mannose. These results suggested that larger R
groups are not tolerated in the 2G12 binding site, an observation
consistent with the modeling experiments (see SI Appendix).
However, smaller and less-restricted substitutions at C-5 (com-
pounds methyl 9a, ethyl 9b, and hydroxylmethyl 9e) as well as
C-3 methyl-modified 11 and 6-deoxy 8 structures were all found
to have IC50 values similar to the natural sugar D-mannose.
These results indicated that such designed substitutions were
indeed tolerated in the 2G12 binding site. Excitingly, from this
panel, C-6 methyl-substituted compound 10 emerged with an
IC50 value not only more potent than D-mannose but even more
Fig. 2. Crystal structures of Fab 2G12 in complex with D-fructose, C-6 methyl
monosaccharide 10 and C-6′″ methyl tetrasaccharide 5. (A) Fo–Fc simulated
annealing omit map of D-fructose bound to Fab 2G12 contoured at 3σ. The
light and heavy chains of Fab 2G12 are shown in green and purple, respec-
tively, and the CDR loops are labeled. (B) Overlay of D-fructose (shown in yel-
low) with Manα1-2Man (shown in blue) from a previous Fab 2G12 structure
(11). Additional direct and water-mediated H bonds (H bonds are shown as
black dashed lines; waters are shown as pink spheres) in the Fab 2G12/D-fruc-
tose structure are shown. (C) Fo–Fc simulated annealing omit map of C-6
methyl monosaccharide 10 (shown in stick format) bound to Fab 2G12 con-
toured at 3σ. (D) Overlay of the modified monosaccharide (shown in yellow)
with Manα1-2Man (shown in blue) (11). Additional water-mediated H bonds
in the Fab 2G12/C-6 methyl monosaccharide 10 structure are shown. (E) Fo–Fc
simulated annealing omit map of C-6′″ methyl tetrasaccharide 5 bound to
Fab 2G12 contoured at 2.5σ. (F) Overlay of the modified tetrasaccharide
(shown in yellow) with the D1 arm of Man7 (shown in blue) (23). The man-
nose residues are labeled. Figure was made using PyMOL (52).
inal mannoside mimic (here D-fructose) could mediate beneficial
interactions.
Modeling, Synthesis, and Affinity for 2G12 of Nonself Monosacchar-
ides. Based on this understanding of the mechanism of enhanced
D-fructose binding by 2G12, we sought to design 2G12 antigens
that incorporated D-fructose-like, nonself monosaccharides. The
D1-arm tetrasaccharide (3) was chosen as a scaffold on which to
mount these nonself prototypes because, in its natural form, it
shows near maximal interaction with 2G12 compared to other
oligomannose fragments (23, 24).
To investigate to what extent such and other nonself substitu-
tions on the D-mannose structure could be tolerated in the 2G12
binding site, we carried out docking studies of substitutions to
D-mannose in complex with Fab 2G12. This modeling revealed
that nonself substitutions not only at C-5 (as in D-fructopyra-
nose), but also at C-3 and C-6 (see SI Appendix, Fig. S1), could
be potentially tolerated by 2G12 and might enhance binding (11).
Based on these results, a representative target panel of nonself
Scheme 1. Synthesis of nonself monosaccharides. Reagents and conditions:
(i) 2,2-dimethoxypropane, pTsOH, DMF, 73%, then TBDPSCl, imidazole, DCM,
95%, then c.HCl, MeOH, 82%, then TBDPSCl, imidazole, DMF, 79%, then
DMSO, C2O2Cl2, DCM, −78 °C → RT, 97%; (ii) RMgBr, THF or Bu3SnCH2OPMB,
BuLi, THF, 76%, then TBAF, THF, then TFA, H2O; (iii) BnOH, AcOH, 60%, then
TrtCl, DMAP, Pyr, 80 °C, 77%, then BnBr, NaH, DMF, 99%, then AcOH, ethanol,
80 °C, 90%, then DMSO, C2O2Cl2, DCM, −78 °C → RT, 97%; (iv) MeMgBr, THF,
then H2, Pd∕C, MeOH, 95%; (v) BnOH, AcOH, 60%, then dimethoxybenzal-
dehyde, pTsOH, DMF, 75 °C, 61%, then DIBAL-H, toluene, −40 °C, 76%, DMSO,
Ac2O; (vi) MeMgBr, THF, then H2, Pd∕C, MeOH; (vii) MeOH, AcCl, then I2, PPh3,
imidazole, THF 65 °C, then Ac2O, Pyr; (viii) H2, Pd∕C, NEt3, MeOH, then AcOH,
H2SO4, then NaOMe, MeOH.
Doores et al.
PNAS
∣
October 5, 2010
∣
vol. 107
∣
no. 40
∣
17109

Fig. 3. (A) IC50 values of synthetic monosaccharides and tetrasaccharides
compared to the IC50 of D-mannose monosaccharide in the binding of
2G12 to gp120. See SI Appendix. (B) ELISA assay of binding of 2G12 to Qβ
glycoconjugates Qβ-3 and Qβ-5. (C) Average 50% maximum binding titer
of serum against oligomannose glycoconjugates elicited by immunization
with Qβ-3 and Qβ-5 (measured by ELISA). For each group, n ¼ 4.
ꢀ ¼ p < 0.05 in t-test.
potent than D-fructose. This compound is the most potent mono-
saccharide 2G12/gp120 inhibitor reported to date.
Synthesis and Affinity for 2G12 of Tetrasaccharide D1-Arm Mimics.
Based on these encouraging results, key representative, nonself,
monosaccharides were next incorporated at the terminus of the
oligosaccharide D1 arm. A 1 þ 2 þ 1 oligosaccharide assembly
strategy was adopted that was intended not only to be high
yielding and allow easy deprotection but to also allow late-stage
introduction of diversity at both reducing and nonreducing
termini (Scheme 2). The central Manα1-2Man disaccharide
block 16 was synthesised prior to formation of the α1-3 linkage
to give trisaccharide 18. The terminal residues were added last
(Scheme 2 A and B) using perbenzylated thiophenyl glycosyl do-
nors 19, 20a, 20b, and 21 and DMTSTas an activating agent (35).
These donors (19, 20a, 20b, and 21) were prepared in four high
yielding steps from the corresponding unprotected monomers 8,
9a, 9b, and 10 (Scheme 2A). Following assembly, tetrasaccharides
22, 23a, 23b, 24A, and 25A were readily deprotected in high yields
under standard conditions to yield target D1-arm sugars 3, 4a, 4b,
5, and 7.
All but one D1-arm sugar variant inhibited binding of gp120 to
2G12 (Fig. 3A). The least potent tetrasaccharide 4a (modified at
C-5′″ with a methyl substituent, where each substituent is denoted
according to sugar residue from reducing to nonreducing termi-
nus X, X′, X″, X′″) displayed an IC50 10 times higher than the D1
arm 3 (4 times lower than that of monomeric D-mannose). Mod-
eling data suggested that substitution at the C-5′″ position causes
changes in the conformation of mannose ring C (SI Appendix,
Fig. S2) thereby reducing interaction of this mannose ring with
the 2G12 binding site (see SI Appendix). This was further sup-
ported by the result that tetrasaccharide 4b modified at C-5′″ with
a bulkier ethyl substituent showed no inhibition at the concentra-
tions tested. Encouragingly, C-6′″ deoxygenated tetrasaccharide
7 had only a 4 times higher IC50 than the D1 arm (6 times lower
than that of D-mannose) and suggested alterations at C-6′″ of the
terminal mannose residue, such as removal of a hydroxyl group,
were well tolerated by 2G12. This result was strikingly confirmed
by C-6′″-methyl tetrasaccharide 5, which not only binds better
than the other nonself variants but has a more potent binding
IC50 value than the D1 -arm 3 itself to 2G12 (over 4 times lower
than the D1 arm). This is the only nonself D1-arm derivative to
show better inhibition of 2G12/gp120 binding than the natural
D1-arm and the most potent monovalent 2G12 ligand known.
Scheme 2. (A) Synthesis of nonself glycosyl donor building blocks. Reagents
and conditions: (i) Ac2O, Pyr then PhSH, BF3•OEt2, DCM then NaOMe, MeOH
then NaH, BnBr, DMF. (B) Synthesis of nonself tetrasaccharides with (B) and
without (A) linker group glycoconjugate synthesis. Reagents and conditions:
(ii) DMTST, TTBP, DCM, 4 Å mol sieves, −78 °C → RT; (iii) NaOMe, MeOH:
(iv) DMTST, TTBP, DCM, 4 Å mol sieves, −78 °C → RT; (v) AcOH, H2O, 50 °C;
(vi) Pd∕C, H2, MeOH,; (vii) TfN3, NH4HCO3, CuCl2, DCM, H2O. (C) Glycoconju-
gate synthesis. Reagents and conditions: (viii) alkynyl N-hydroxysuccinimide
ester (HCCðCH2Þ2CO2-NHS), 15% DMSO, TRIS buffer pH 7.0; (ix) aminoguani-
dine, CuSO4, tris(3-hydroxypropyltrazolylmethyl)amine, sodium ascorbate,
PBS (0.1 M, pH 7.0).
Crystal Structures of Fab 2G12 in Complex with C-6 Methyl-Substituted
Monosaccharide 10 and C-6′″-Methyl Tetrasaccharide 5. To elucidate
the structural basis of the higher affinity of 2G12 for C-6′
″-methyl-substituted tetrasaccharide 5 over the natural D1-arm
3, we determined crystal structures of Fab 2G12 in complex with
both C-6 methyl-substituted compound 10 and C-6′″-methyl
tetrasaccharide 5 to 1.75 Å and 2.85 Å resolution, respectively
(see SI Appendix, SI Materials and Methods and Table S1). In
the former high-resolution structure, compound 10 is bound at
the two primary combining sites of the Fab dimer with extremely
well-defined electron density (Fig. 2C). As in the Fab 2G12/
D-fructose structure, the contacts formed by the modified mono-
saccharide with 2G12 are similar to those made by the terminal
mannose in Manα1-2Man (11) (Fig. 2D). However, the C-6
methyl group forms additional van der Waals interactions with
the aromatic side chain of TyrL94 and AspH100B O, which results
in partial burial of the hydrophobic methyl group and appears to
account for the enhanced affinity of 2G12 for compound 10 over
D-mannose. Moreover, similar to the Fab 2G12/D-fructose struc-
ture, a water-mediated H-bond relay bridges the anomeric oxygen
in compound 10 with AlaH31 O and SerH100A O and mimics
the direct H bond between O3 in the reducing terminal mannose
in Manα1-2Man disaccharide (11) and AlaH31 O and further
17110
∣
www.pnas.org/cgi/doi/10.1073/pnas.1002717107
Doores et al.

explains the stronger affinity of 2G12 for the modified monosac-
charide over Manα1-2Man. A total of 197 Å2 of molecular sur-
face on Fab 2G12 and 182 Å2 of molecular surface on C-6 methyl
monosaccharide 10 are buried in the complex, with 9 direct and 9
water-mediated H-bonds and 61 van der Waals interactions in
each antigen binding site.
Discussion
Our design and synthetic strategy has led to identification of non-
self modifications to the D1 arm that show enhanced antigenicity
to 2G12. Although nonself modifications to the terminus of the
D1 arm are tolerated in the 2G12 binding site, it is clear that sev-
eral mechanisms influence the antigenicity of these glycans for
the 2G12 antibody. Small alterations at the C-3 and C-5 positions
of D-mannose, as well as removal of the hydroxyl group at the C-6
position (Scheme 1, 8, 9a, 9b, and 11), can be tolerated in the
2G12 binding site to create binding affinity similar to that of
D-mannose. Specific tolerance was also observed for the tetrasac-
charides: 4a (C-5 methyl) and 7 (C-6 deoxy) both show binding
levels similar to that of D1-arm 3. However, when the size of the
alteration is increased and/or its conformations restricted by car-
bon hybridization (9c and 9d), these modifications become much
less tolerated in the 2G12 binding site leading to IC50 values
much higher than D-mannose. For the tetrasaccharide 4b, with
the bulkier ethyl group at the C-5′″ position, there is no inhibition
of 2G12/gp120 binding. Our modeling studies (see SI Appendix)
suggest that large substituents at the C-5 position lead to altera-
tions in the conformation of the mannose ring C (Fig. 1B) result-
ing in reduced interaction of this sugar with the 2G12 binding site.
Modifications that lead to enhancement of the inhibition of
2G12/gp120 binding arose from additional interactions of the
nonself modification with the 2G12 binding site. This hypothesis
is supported by the crystal structures of D-fructose and C-6
methyl monosaccharide 10. In the case of D-fructose, the addi-
tional direct and water-mediated interactions between the C-5
hydroxyl and AlaH31 O, LeuH100 O and SerH100A O appear to ex-
plain why this sugar in its pyranose form is an inhibitor of 2G12/
D-mannose binding. The additional van der Waals contacts be-
tween the designed C-6-methyl modification and the aromatic
side chain of TyrL94 appear to correlate with the higher affinity
of 2G12 for the modified monosaccharide 10 over D-mannose
and even D-fructose. The finding that 2G12 uses the same mode
of recognition for the nonself C-6′″-modified tetrasaccharide 5 as
for the natural D1 arm, apart from the beneficial additional in-
teractions formed by the designed C-6′″-methyl group modifica-
tion, led us to evaluate the immunogenicity of this antigen in vivo.
Conjugates that contain this nonself glycan elicited antibodies
that react with the self glycan, suggesting that the additional
methyl group, although enhancing antigenicity, does not dominate
the specificity of the resulting antibody binding sites. It should be
noted that previous examples of glycoconjugates designed as vac-
cines [containing, e.g., S-linked gangliosides (38) or N-propiony-
lated polysialic acids (39)] have led to antibodies being generated
that were highly complementary to the modification in preference
to the unmodified sugar. Although here HIV-reactive antibodies
were not elicited, higher titers of self-D1 arm (3) reactive antibo-
dies were elicited by immunization with the nonself glycoconjugate
compared to the self glycoconjugate; it also cannot be discounted
that other differences such as loading distribution may play a role.
One possible central factor is the need to also elicit domain-
exchanged antibodies. Immunization is, therefore, potentially
limited by the presentation mode of the glycan on conjugates (such
as in Qβ particles), and alternative clustered presentations of this
nonself glycan to better mimic the glycan shield of HIV are now
under investigation. It is also important to note that given the un-
iqueness of 2G12 and its domain-exchanged structure, the as-yet
unknown conditions required for such exchange might be difficult
to achieve. Such a structure or similar might therefore only be
formed very rarely in response to vaccine challenge.
Although the Fab 2G12/C-6′″ methyl tetrasaccharide 5 cocrys-
tals were highly anisotropic and diffracted to modest resolution,
the electron density for the entire modified tetrasaccharide is also
well-defined at both primary combining sites (Fig. 2E). The tet-
rasaccharide is bound with an overall conformation similar to that
of the D1 arm in Man7, Man8, and Man9GlcNAc2 in complexes
with Fab 2G12 (11, 23) (Fig. 2F). The buried surface area is
approximately 300 Å2 for Fab 2G12 and 295 Å2 for the nonself,
D1-arm mimic 5. Together, the Fab 2G12/C-6 methyl mono-
saccharide 10 and Fab 2G12/C-6′″ methyl tetrasaccharide 5 struc-
tures uncover the molecular basis for the higher affinity of 2G12
for C-6′″-methyl tetrasaccharide 5 over Man4. That C-6′″-methyl
tetrasaccharide 5 adopts the same overall conformation at the
antigen binding sites of 2G12 as the D1 arm of Man9GlcNAc2 sug-
gests that the interactions with the C-6′″ methyl group are the
only difference between the mechanism of 2G12 binding to the
modified tetrasaccharide and the D1 arm.
Synthesis of Glycoconjugates for Immunogenicity Studies. Having
identified a nonself modification that showed enhanced 2G12
antigenicity, we next investigated whether this correlated with
enhanced immunogenicity. Both C-6′″-methyl-tetrasaccharide 5
and D1-arm tetrasaccharide 3 were equipped with a five-carbon
linker terminating with a reactive azide group (Scheme 2B).
Reaction with alkyne-modified lysine residues on the surface
of the virus-like particle Qβ using copper-catalyzed azide-alkyne
chemistry (Scheme 2C) (36, 37) created glycan shield mimetics
Qβ-3 and Qβ-5 that bound 2G12 with nanomolar affinity
(Fig. 3B). Comparable glycan loadings (average ∼300 per parti-
cle) was confirmed by MALDI mass spectrometry (SI Appendix,
Fig. S7A), which also showed slight differences in distribution of
the sugars across the capsid subunits. The enhanced IC50 ob-
served for isolated C-6′″-methyl-tetrasaccharide 5 over D1 arm
also correlated with an enhancement in affinity of 2G12 for
the glycoconjugate Qβ-5 that contained 5 as compared to that
bearing only D1 arm (Qβ-3).
Immunogenicity of Nonself Glycan 5. New Zealand white rabbits
were immunized with a prime and three boost immunizations
of either self conjugate Qβ-3 or nonself Qβ-5, and the 50% max-
imum binding titers were measured in an ELISA format. In all
cases, higher titers of mannose-reactive antibodies were gener-
ated from immunization with the nonself glycoconjugate Qβ-5
compared to the self conjugate Qβ-3 (Fig. 3C). The increased
titers were, importantly, most significant for reactivity toward
D1 arm (3) on two unrelated protein scaffolds, cowpea mosaic
virus (CPMV-3) and bovine serum albumin (BSA-3), but were
also observed for Man8 and Man9 glycans as well as 5 (Fig. 3C).
Both immunogens showed a similar time course trend in D1-arm-
specific antibody production (see SI Appendix). Only moderate
reactivity was observed against CPMV particles displaying the
linker and one mannose sugar (see SI Appendix) confirming spe-
cificity for the sugar moiety. Indeed, depletion of this linker re-
activity did not result in significant decreases in D1-arm-specific
antibodies. The ability of the nonself glycoconjugate Qβ-5 to elicit
HIV-reactive antibodies was assessed by measuring both binding
to recombinant gp120 and the ability to neutralize HIV pseudo-
virus. This assay revealed that there was no binding by the anti-
bodies to gp120 and no neutralization (Figs. S8 and S9).
In conclusion, based on the previous observation that D-fruc-
tose is a more potent inhibitor of 2G12 binding to gp120 than
D-mannose (11), a series of nonself modifications have been in-
troduced into the terminus of the D1 arm of high-mannose sugars
that represents an unusual self-carbohydrate epitope in HIV-1
gp120. These modified sugars are not only well tolerated in
Doores et al.
PNAS
∣
October 5, 2010
∣
vol. 107
∣
no. 40
∣
17111

the 2G12 binding site but also enhance antigenicity. In particular,
introduction of a methyl group at the C-6′″ position (in tetra-
saccharide 5) of the terminal mannose residue of the D1 arm
enhances binding to 2G12 compared to the natural D1 arm found
on gp120. Crystal structures of this modified sugar with 2G12 in
comparison to structures of 2G12 with unmodified sugars (11, 23)
and D-fructose enabled us to elucidate the mechanism of this en-
hanced affinity as well as the increased affinity of D-fructose for
2G12 compared to D-mannose. HIV-reactive antibodies were not
elicited. The generation of HIV neutralizing activity or even gp120
binding activity from antibodies using synthetic conjugates as
immunogens has, thus far, been elusive for the field in general.
However, higher titers of self-D1 arm (3) reactive antibodies were
elicited from the nonself glycan than theself. The exact mechanism
of this enhanced immunogenicity is uncertain, yet the observed
cross-reactivity between natural and unnatural glycans may have
broader significance and potential use in other vaccine programs
directed toward other (e.g., microbial or fungal) pathogens. We
are currently investigating alternative presentations of these non-
self D1-arm mimics to more effectively mimic the clustered oligo-
mannose glycans of HIV in an attempt to elicit domain-exchanged
2G12-like antibodies or other neutralizing antibodies. When
displayed in the “correct” format this nonself glycan could be an
important component of a carbohydrate anti-HIV vaccine.
Materials and Methods
Chemical Synthesis. Details of the synthesis of all compounds can be found
in the SI Appendix. Briefly, glycosylations typically used acceptor (1 equiv),
thioglycoside donor (1.2 equiv), and dimethylthiosulfonium triflate (4 equiv)
at −78 °C in dichloromethane.
Biological Assays. Details of all binding assays, immunization protocols, and
titer determinations can be found in the SI Appendix.
Crystal Structure Determinations and Analyses. Full details can be found in
the SI Appendix. Briefly, Fab 2G12 fragments were prepared as previously
described (11) and concentrated to 20 mg∕mL. For each complex, the solid
sugar ligand was added to the Fab solution to saturation. Cocrystals were
cryoprotected with glycerol or sodium malonate. Data collection and refine-
ment statistics are summarized in SI Appendix, Table S1.
ACKNOWLEDGMENTS. We thank the staffs of the Advanced Light Source and
Advanced Photon Source, and Sander van Kasteren for technical assistance.
This work was supported by National Institutes of Health Grants GM46192
and AI84817 (to I.A.W.) and funds from the Phillip Leverhulme Prize in
Molecular Biology (B.G.D.), the Neutralizing Antibody Consortium of the In-
ternational AIDS Vaccine Initiative, and the Skaggs Institute for Chemical
Biology. B.G.D. is a Royal Society–Wolfson Research Merit Award recipient
and supported by an Engineering and Physical Sciences Research Council Life
Sciences Interface Platform Grant.
1. AIDS Epidemic Update. Joint United Nations Programme on HIV/AIDS (United Nations,
Geneva).
22. Wilson IA, Stanfield RL (1995) A Trojan horse with a sweet tooth. Nat Struct Biol
2:433–436.
2. Johnson WE, Desrosiers RC (2002) Viral persistance: HIV’s strategies of immune system
evasion. Annu Rev Med 53:499–518.
3. Allan JS, et al. (1985) Major glycoprotein antigens that induce antibodies in AIDS
patients are encoded by HTLV-III/LAV. Science 228(4703):1091–1094.
4. Wyatt R, et al. (1998) The antigenic structure of the HIV gp120 envelope glycoprotein.
Nature 393(6686):705–711.
5. Saphire EO, et al. (2003) Crystal structure of an intact human IgG: Antibody asymmetry,
flexibility, and a guide for HIV-1 vaccine design. Adv Exp Med Biol 535:55–66.
6. Haynes BF, et al. (2005) Cardiolipin polyspecific autoreactivity in two broadly neutra-
lizing HIV-1 antibodies. Science 308:1906–1908.
7. Zwick MB, et al. (2004) The long third complementarity-determining region of the
heavy chain is important in the activity of the broadly neutralizing anti-human
immunodeficiency virus type 1 antibody 2F5. J Virol 78:3155–3161.
8. Cardoso RMF, et al. (2005) Broadly neutralizing anti-HIV antibody 4E10 recognizes a
helical conformation of a highly conserved fusion-associated motif in gp41. Immunity
22:163–173.
23. Calarese DA, et al. (2005) Dissection of the carbohydrate specificity of the broadly
neutralizing-anti-HIV-1 antibody 2G12. Proc Natl Acad Sci USA 102:13372–13377.
24. Lee HK, et al. (2004) Reactivity-based one-pot synthesis of oligomannoses: Defining
antigens recognized by 2G12, a broadly neutralizing anti-HIV-1 antibody. Angew
Chem Int Edit 43:1000–1003.
25. Hessell AJ, et al. (2009) Broadly neutralizing human anti-HIV antibody 2G12 is effective
in protection against mucosal SHIV challenge even at low serum neutralizing titers.
PLoS Pathog 5:e1000433.
26. Astronomo RD, et al. (2008) A glycoconjugate antigen based on the recognition
motif of a broadly neutralizing human immunodeficiency virus antibody, 2G12, is
immunogenic but elicits antibodies unable to bind to the self glycans of gp120. J Virol
82:6359–6368.
27. Dudkin VY, et al. (2004) Toward fully synthetic carbohydrate-based HIV antigen
design: On the critical role of bivalency. J Am Chem Soc 126:9560–9562.
28. Krauss IJ, et al. (2007) Fully synthetic carbohydrate HIV antigens designed on the logic
of the 2G12 antibody. J Am Chem Soc 129:11042–11044.
9. Trkola A, et al. (1996) Human monoclonal antibody 2G12 defines a distinctive neutra-
lization epitope on the gp120 glycoprotein of human immunodeficiency virus type 1.
J Virol 70:1100–1108.
29. Li H, Wang LX (2004) Design and synthesis of a template-assembled oligomannose
cluster as an epitope mimic for human HIV-neutralizing antibody 2G12. Org Biomol
Chem 2:483–488.
10. Walker L, et al. (2009) Broad and potent neutralizing antibodies from an African
donor reveal a new HIV-1 vaccine target. Science 326:285–289.
11. Calarese DA, et al. (2003) Antibody domain exchange is an immunological solution to
carbohydrate cluster recognition. Science 300:2065–2071.
12. Saphire EO, et al. (2001) Crystal structure of a neutralizing human IgG against HIV-1: A
template for vaccine design. Science 293:1155–1159.
30. Ni J, Song H, Wang Y, Stamatos NM, Wang LX (2006) Toward a carbohydrate-based
HIV-1 vaccine: Synthesis and immunological studies of oligomannose-containing
glycoconjugates. Bioconjugate Chem 17:493–500.
31. Wang LX, Ni J, Singh S, Li H (2004) Binding of high-mannose-type oligosaccharides
and synthetic oligomannose clusters to human antibody 2G12: implications for
HIV-1 vaccine design. Chem Biol 11:127–134.
13. Kwong PD, Wilson IA (2009) HIV-1 and influenza antibodies: seeing antigens in new
ways. Nat Immunol 10:573–578.
14. Ofek G, et al. (2004) Structure and mechanistic analysis of the anti-human immuno-
32. Wang SK, et al. (2008) Targeting the carbohydrates on HIV-1: Interaction of oligoman-
nose dendrons with human monoclonal antibody 2G12 and DC-SIGN. Proc Natl Acad
Sci USA 105:3690–3695.
deficiency virus type
78:10724–10737.
1
antibody 2F5 in complex with its gp41 epitope.
J
Virol
33. Wang J, Li H, Zou G, Wang L-X (2007) Novel template-assembled oligosaccharide
clusters as epitope mimics for HIV-neutralizing antibody 2G12. Design, synthesis,
and antibody binding study. Org Biomol Chem 5:1529–1540.
34. Ingale S, Wolfert MA, Gaekwad J, Buskas T, Boons G-J (2007) Robust immune responses
elicited by a fully synthetic three-component vaccine. Nat Chem Biol 3:663–667.
35. Fugedi P, Garegg PJ (1986) A novel promoter for the efficient construction of 1,2-trans
linkages in glycoside synthesis, using thioglycosides as glycosyl donors. Carbohydr Res
149:C9–C12.
15. Zhou T, et al. (2007) Structural definition of a conserved neutralization epitope on
HIV-1 gp120. Nature 445:732–737.
16. Pejchal R, et al. (2009) A conformational switch in human immunodeficiency virus
gp41 revealed by the structures of overlapping epitopes recognized by neutralizing
antibodies. J Virol 83:8451–8462.
17. Scanlan CN, et al. (2002) The broadly neutralizing anti-human immunodeficiency virus
type 1 antibody 2G12 recognizes a cluster of alpha1—>2 mannose residues on the
outer face of gp120. J Virol 76(14):7306–7321.
36. Astronomo RD, et al. (2010) Defining criteria for oligomannose immunogens for HIV
using icosahedral virus capsid scaffolds. Chem Biol 17:357–370.
18. Sanders RW, et al. (2002) The mannose-dependent epitope for neutralizing antibody
37. Hong V, Presolski SI, Ma C, Finn MG (2009) Analysis and optimization of copper-
catalyzed azide-alkyne cycloaddition for bioconjugation. Angew Chem Int Edit
48:9879–9883.
2G12 on human immunodeficiency virus type
1 glycoprotein gp120. J Virol
76:7293–7305.
19. Rademacher TW, Parekh RB, Dwek RA (1988) Glycobiology. Annu Rev Biochem
57:785–838.
20. Rudd PM, Dwek RA (1997) Glycosylation: Heterogeneity and the 3D structure of
proteins. Crit Rev Biochem Mol 32:1–100.
21. Weis WI (1997) Cell-surface carbohydrate recognition by animal and viral lectins. Curr
Opin Struct Biol 7:624–630.
38. Rich JR, Wakarchuk WW, Bundle DR (2006) Chemical and chemoenzymatic synthesis of
S-Linked ganglioside analogues and their protein conjugates for use as immunogens.
Chem Eur J 12(3):845–858.
39. Krug LM, et al. (2004) Vaccination of small cell lung cancer patients with polysialic
acid or N-propionylated polysialic acid conjugated to keyhole limpet hemocyanin.
Clin Cancer Res 10(3):916–923.
17112
∣
www.pnas.org/cgi/doi/10.1073/pnas.1002717107
Doores et al.