(E)-alkene and ethylene isosteres substantially alter the hydrogen-bonding network in class II MHC Aq/glycopeptide complexes and affect T-cell recognition

Article abstract of DOI:10.1021/ja2038722

The structural basis for antigen presentation by class II major histocompatibility complex (MHC) proteins to CD4⁺ T-cells is important for understanding and possibly treating autoimmune diseases. In the work described in this paper, (E)-alkene and ethylene amide-bond isosteres were used to investigate the effect of removing hydrogen-bonding possibilities from the CII259-270 glycopeptide, which is bound by the arthritis-associated murine A^q class II MHC protein. The isostere-modified glycopeptides showed varying and unexpectedly large losses of A^q binding that could be linked to the dynamics of the system. Molecular dynamics (MD) simulations revealed that the backbone of CII259-270 and the A^q protein are able to form up to 11 hydrogen bonds, but fewer than this number are present at any one time. Most of the strong hydrogen-bond interactions were formed by the N-terminal part of the glycopeptide, i.e., in the region where the isosteric replacements were made. The structural dynamics also revealed that hydrogen bonds were strongly coupled to each other; the loss of one hydrogen-bond interaction had a profound effect on the entire hydrogen-bonding network. The A^q binding data revealed that an ethylene isostere glycopeptide unexpectedly bound more strongly to A^q than the corresponding (E)-alkene, which is in contrast to the trend observed for the other isosteres. Analysis of the MD trajectories revealed that the complex conformation of this ethylene isostere was structurally different and had an altered molecular interaction pattern compared to the other A^q/glycopeptide complexes. The introduced amide-bond isosteres also affected the interactions of the glycopeptide/A^q complexes with T-cell receptors. The dynamic variation of the patterns and strengths of the hydrogen-bond interactions in the class II MHC system is of critical importance for the class II MHC/peptide/TCR signaling system.

Full text of DOI:10.1021/ja2038722

ARTICLE
pubs.acs.org/JACS
(E)-Alkene and Ethylene Isosteres Substantially Alter the
Hydrogen-Bonding Network in Class II MHC Aq/Glycopeptide
Complexes and Affect T-Cell Recognition
Ida E. Andersson,† Tsvetelina Batsalova,‡ Sabrina Haag,‡ Balik Dzhambazov,‡ Rikard Holmdahl,‡
Jan Kihlberg,*,†,§ and Anna Linusson*,†
ꢀ
ꢀ
†Department of Chemistry, Umea University, SE-901 87 Umea, Sweden
‡Medical Inflammation Research, Department of Medical Biochemistry and Biophysics, Karolinska Institute, SE-171 77 Stockholm,
Sweden
§AstraZeneca Research and Development M€olndal, SE-431 83 M€olndal, Sweden
S Supporting Information
b
ABSTRACT: The structural basis for antigen presentation by class II
major histocompatibility complex (MHC) proteins to CD4+ T-cells
is important for understanding and possibly treating autoimmune
diseases. In the work described in this paper, (E)-alkene and ethylene
amide-bond isosteres were used to investigate the effect of removing
hydrogen-bonding possibilities from the CII259ꢁ270 glycopeptide,
which is bound by the arthritis-associated murine Aq class II MHC
protein. The isostere-modified glycopeptides showed varying and
unexpectedly large losses of Aq binding that could be linked to the
dynamics of the system. Molecular dynamics (MD) simulations
revealed that the backbone of CII259ꢁ270 and the Aq protein are able to form up to 11 hydrogen bonds, but fewer than this
number are present at any one time. Most of the strong hydrogen-bond interactions were formed by the N-terminal part of the
glycopeptide, i.e., in the region where the isosteric replacements were made. The structural dynamics also revealed that hydrogen
bonds were strongly coupled to each other; the loss of one hydrogen-bond interaction had a profound effect on the entire hydrogen-
bonding network. The Aq binding data revealed that an ethylene isostere glycopeptide unexpectedly bound more strongly to Aq than
the corresponding (E)-alkene, which is in contrast to the trend observed for the other isosteres. Analysis of the MD trajectories
revealed that the complex conformation of this ethylene isostere was structurally different and had an altered molecular interaction
pattern compared to the other Aq/glycopeptide complexes. The introduced amide-bond isosteres also affected the interactions of
the glycopeptide/Aq complexes with T-cell receptors. The dynamic variation of the patterns and strengths of the hydrogen-bond
interactions in the class II MHC system is of critical importance for the class II MHC/peptide/TCR signaling system.
’ INTRODUCTION
which is open at both ends, allowing long peptides to protrude
out of the groove.2,3 A class II MHC protein can bind a variety of
different peptides and present them to TCRs. This binding
promiscuity is enabled by an array of sequence-independent
hydrogen bonds formed between the peptide backbone and the
MHC protein, while the specificity displayed by different class II
MHC proteins is achieved through pockets in the MHC binding
groove that have preferences for certain peptide amino acid side
chains.2
The structure and the dynamics of class II MHC/peptide
complexes have been studied through molecular dynamics (MD)
simulations of the motions of the interacting atoms.4ꢁ12 This
information has been used to interpret biological processes
regulated by class II MHC/peptide complexes on a molecular
level. For example, structures extracted from MD simulations
Class II major histocompatibility complex (MHC) glycopro-
teins are expressed in antigen-presenting cells (APCs) where
they bind peptide antigens derived from processing proteins of
extracellular origin.1 The resulting class II MHC/peptide com-
plexes are transported to the cell surface of the APCs where they
are presented to T-cell receptors (TCRs) on CD4+ T-helper
cells. The interactions governing the ternary class II MHC/
peptide/TCR complexes play a crucial role in determining if an
immune response is initiated via activation of the T-cell. Knowl-
edge of the structural basis for antigen presentation by the class II
MHC protein2 and recognition by the TCR are important for
understanding and possibly treating autoimmune diseases, such
as type 1 diabetes, multiple sclerosis (MS), and rheumatoid
arthritis (RA), where the immune system recognizes and attacks
endogenous tissues.
Peptides consisting of up to 20 amino acid residues bind in an
extended conformation in the class II MHC binding groove,
Received: April 27, 2011
Published: July 18, 2011
r
2011 American Chemical Society
14368
dx.doi.org/10.1021/ja2038722 J. Am. Chem. Soc. 2011, 133, 14368–14378
|

Journal of the American Chemical Society
ARTICLE
were used in combination with binding affinity predictions and
analysis of the hydrogen-bonding network to explain the varia-
tion in class II MHC binding between related peptides.6
Although most MD simulations only include class II MHC/
peptide complexes and not TCRs because of the limited number
of TCR crystal structures, such data can still provide information
about the T-cell responses.5,7ꢁ9 For example, the exposed
peptide surface area derived from MD data for a series of class
II MHC-bound substituted hen egg lysozyme (HEL) peptides
was correlated with their T-cell responses.5 Furthermore, con-
formational changes in the MHC binding of different peptides
observed in MD simulations have been linked to T-cell respon-
siveness and experimental autoimmune encephalomyelitis (EAE)
induction in vivo.7 We have recently linked flexibility and
changes in the presented epitope of MHC anchor-modified
glycopeptides to different T-cell responses.11 Recently, parts of
the TCR were included in a simulation of a homology model of
an RA-associated TCR Vβ domain docked to a peptide/DR4
crystal structure, which indicated that certain residues of the
flexible complementarity determining region (CDR)3β loop
were critical for TCR recognition.12
To probe interactions in the ternary class II MHC/peptide/
TCR complexes and also overcome some of the limitations
associated with the therapeutic use of peptides, e.g., low bioavail-
ability and metabolic stability, different modifications have been
incorporated into peptides that interact with class II MHC
proteins.13ꢁ27 Examples include replacement of backbone amides
by different amide bond isosteres (e.g., retro-inverso16 and
(E)-alkene)15 or introduction of D-amino acids,14 aza-amino
acids,21 and other non-natural amino acids.20,24 There are also
examples where modified peptides have been shown to prevent
or treat disease in animal models for RA and MS.13,23
Figure 1. (a) Structure of glycopeptide 1 (CII259ꢁ273). The peptide
sequence that was modified by (E)-alkene and ethylene isosteres in
the present study is shown in bold. (b) Comparative model25,26 of the
complex between CII259ꢁ270 and the Aq protein. The Aq binding site
has a secondary structure consisting of two parallel R helices (R1 and β1)
on top of a β sheet. The glycopeptide 1 is bound in an extended backbone
conformation with the GalHyl264 side chain solvent exposed and the side
chains of Ile260 and Phe263 buried in the P1 and P4 pocket, respectively. The
side chains of GalHyl264, Ile260, and Phe263 are indicated by balls and sticks.
Vaccination of mice with glycopeptide 1 (Figure 1) alone28 or
in complex with the class II MHC Aq protein29 has been shown
to prevent development of collagen-induced arthritis (CIA), a
mouse model for RA. This glycopeptide, which is a fragment
from type II collagen (CII), was also recognized by T-cells
isolated from a cohort of RA patients.30 Comparative models25,26
of the Aq/glycopeptide complex have revealed an extensive
hydrogen-bonding network between the glycopeptide backbone
and the residues in Aq. Furthermore, the Ile260 and Phe263
residues in 1 are anchored in the P1 and P4 pockets of Aq,
isosteric analogues of glycopeptide 1 using Fmoc-based solid-
phase synthesis. The synthesis of Ile-Ala (E)-alkene and ethylene
derivatives 2 and 3 was based on a strategy developed by
Luthman and co-workers (Scheme 1).35 Phosphonium salt 11
was first prepared via a three-step process: N-trifluoroacetyl
protection of (S)-isoleucinol (8), conversion of the alcohol into
the corresponding bromide (10), and displacement with triphe-
nylphosphine. In the following Wittig reaction,35 11 was treated
with n-butyllithium and reacted with chiral aldehyde 12,36,37
which afforded (E)-alkene 13 as a single stereoisomer with a yield
of 87%. After removal of the silyl-protecting group, alcohol 14
was oxidized to carboxylic acid 15 in 75% yield using Jones
reagent. This reaction was accompanied by formation of ketone
16 in 15% yield via carbonꢁcarbon bond cleavage and thus
resembled oxidations of related systems reported previously.38,39
Final exchange of the N-protecting group gave the desired
(E)-alkene derivative 2 (89% from 15). Saturated analogue 3
(77%) was then obtained by palladium-catalyzed hydrogenation
of 2 during which some Fmoc cleavage was noted.
Application of the above route to Ala-Gly (E)-alkene deriva-
tive 4 (Figure 2) has been described previously.26 The corre-
sponding saturated analogue 5 was obtained by directly reducing
the E/Z-mixture of alkenes 17 obtained in the Wittig reaction
(Scheme 2), followed by hydroxyl group deprotection, oxidation,
and exchange of the N-protecting group (57% yield from 17).
Initial attempts to synthesize Gly-Phe (E)-alkene derivative 6
via the same protocol (Scheme 3), i.e., by reacting phosphonium
whereas the β-D-galactopyranosyl hydroxylysine (GalHyl264
)
side chain protrudes out of the binding site and is critical for
recognition by the TCR (Figure 1).31ꢁ33
In this study we investigated the effect of removing backbone
hydrogen-bonding possibilities from 1 on Aq binding and
recognition by T-cell hybridomas. Glycopeptide 1 was modified
by introducing (E)-alkene and ethylene isosteres instead of each
of the three amide bonds in between the two anchors, Ile260 and
Phe263 (Figure 1). These isostere moieties can mimic the geometry
of the amide bond but lack hydrogen-bonding capability.34 The
inclusion of the ethylene isostere also enabled the effect of
increasing the flexibility of the peptide backbone to be studied.
The structural dynamics of the glycopeptide/Aq complexes were
studied by MD simulations and compared to experimental
results for Aq binding and T-cell responses.
’ RESULTS AND DISCUSSION
Synthesis. This study required synthesis of the dipeptide
building blocks 2ꢁ7 (Figure 2), which were used to assemble
14369
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
salt 21 with chiral aldehyde 22, gave a 13:10 mixture of (E)- and
(Z)-isomers according to the 1H NMR spectrum of the
crude product. After purification by flash chromatography, the
(E)-alkene was obtained in a yield of only approximately 10%.
Most likely, the poor solubility of phosphonium salt 21 in THF
contributed to the low yield as a suspension remained even after
the addition of base. Attempts to find improved conditions
(Supporting Information) were unsuccessful, and an alternative
approach to the Gly-Phe isosteres was therefore employed that
instead relied on Evans alkylation of an oxazolidinone enolate to
establish the C-terminal stereocenter as described by Kelly and
co-workers (Scheme 4).40
peroxide, followed by acidic hydrolysis of the benzophenone
imine and protection of the amino group with an Fmoc group.
The crude product was purified by flash chromatography,
followed by preparative chiral HPLC, affording 6 from 29 with
a yield of 66% and >99% ee. Catalytic hydrogenation of 6 then
Scheme 2. Saturated Analogue 5 Synthesisa
Initial attempts to alkylate 27 with benzyl bromide gave only
trace amounts of product, while the chiral auxiliary was mainly
cleaved. It was hypothesized that the lithiated carbamate ob-
tained from 27 could interfere in the reaction, and therefore the
Boc-protecting group was exchanged to a benzophenone imine,
which has been successfully used in other alkylation reactions.41ꢁ44
Alkylation of 28 then proceeded to give 29 with a 70% yield as a
93:7 diastereomeric mixture. It was not possible to separate the
diastereomers using flash chromatography, and therefore the
remaining steps were performed on the diastereomeric mixture.
Thus, oxazolidinone 29 was treated with basic hydrogen
a Reagents and conditions: (a) H2, Pd/C, MeOH, rt; (b) TBAF, THF, rt
(71% from 17); (c) CrO3, H2SO4, acetone/H2O, rt (88%); (d) (i)
K2CO3, MeOH/H2O, rt, (ii) FmocOSu, Na2CO3, MeCN/H2O, rt
(92% from 20).
Figure 2. Target isostere derivatives protected for use in solid-phase
glycopeptide synthesis.
Scheme 1. Synthesis of Ile-Ala (E)-Alkene and Ethylene Derivatives 2 and 3a
a Reagents and conditions: (a) TFAA, Et3N, CH2Cl2, 0 °C (100%); (b) CBr4, PPh3, MeCN, 0 °C f rt (85%); (c) PPh3, toluene, reflux (97%); (d)
(i) n-BuLi, THF, ꢁ78 °C, (ii) aldehyde 12, THF, ꢁ78 °C f 0 °C (87%); (e) TBAF, THF, rt (94%); (f) CrO3, H2SO4, acetone/H2O, rt (75%);
(g) (i) K2CO3, MeOH/H2O, rt, (ii) FmocOSu, Na2CO3, MeCN/H2O, rt (89%); (h) H2, Pd/C, MeOH, rt (77%).
14370
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
also gave the corresponding saturated analogue 7 with a yield
of 84%.
effects on binding to Aq and T-cell recognition. Similar observa-
tions have also been reported for other murine class II MHC
systems.15,16,18,47ꢁ49 In addition, the dramatic effect of disrupt-
ing a hydrogen bond between a MHC class II protein and
peptides has been studied by selected amino acid mutations of
the Ad protein.50ꢁ52
The isostere building blocks 2, 3, and 5ꢁ7 were activated
using HATU and 2,4,6-collidine and then incorporated into the
CII259ꢁ273 sequence using Fmoc-based solid-phase glycopep-
tide synthesis.46 Cleavage of the isostere glycopeptides from the
solid support, deacetylation of the galactose moiety, and final
purification by reversed-phase HPLC afforded the trifluoroace-
tate salts of 31, 32, and 34ꢁ36 (Figure 3) with overall yields of
14ꢁ20% based on the resin capacities and in >98% purity. All
glycopeptides were homogeneous, according to analytical re-
versed-phase HPLC using two eluent systems, and their structures
were confirmed by 1H NMR spectroscopy and matrix-assisted laser
desorption ionization time-of-flight (MALDI-TOF) mass spectro-
metry. Synthesis of glycopeptide 33 from building block 4 using
these conditions has been reported previously.26
At two out of the three modified positions (i.e., the Ile-Ala and
Ala-Gly positions, Figure 1), stronger Aq binding was obtained
for the (E)-alkene isostere than the corresponding ethylene
isostere (cf., 31 with 32 and 33 with 34, Figure 4). This agrees
with the assumption that an increase in flexibility provided by the
ethylenes should lead to a higher entropy cost for binding to Aq
and thus result in lower binding affinity.53,54 Unexpectedly,
introduction of the more flexible ethylene isostere at the Gly-
Phe position (36) resulted in stronger binding to Aq compared to
the corresponding (E)-alkene (35). As both isosteres can mimic
the geometry of the native amide, one would expect that the
more flexible isostere would bind less strongly than the rigid
(E)-alkene. However, recent investigations involving thermody-
namic analysis, structural and dynamic characterization, and
solvation studies of series of ligands binding to proteins55ꢁ58
have shown that the relationship between the enthalpy and
entropy components, and their effect on binding affinity, follows
a more complex pattern than initially assumed, and as such it is
not yet fully understood.
Binding to the Aq Protein. Competitive binding experiments
were performed to investigate the effect of the introduced
(E)-alkene and ethylene isosteres in 31ꢁ36 on binding to the
Aq protein (Figure 4). It was found that all isostere glycopeptides
bound more weakly to Aq than the native 1. Four of the modified
glycopeptides exhibited moderate binding, i.e., the (E)-alkenes
31, 33, and 35 as well as the Gly-Phe ethylene 36. The remaining
Ile-Ala and Ala-Gly ethylene isosteres 32 and 34 bound weakly.
Thus, isosteric replacement of one out of 14 amide bonds in 1
had a profound effect on Aq binding. Selected amide bonds in 1
have previously been replaced by (E)-alkene,26 ketomethylene,26
methyleneamine,26 and methylene ether27 isosteres or oxazole25
moieties, and these modifications generally resulted in significant
Scheme 3. Attempt to Synthesize Gly-Phe (E)-Alkene
Derivative 6a
a Reagents and conditions: (a) (i) n-BuLi, THF or THF/DME 95:5,
ꢁ78 °C, (ii) aldehyde 22, THF, ꢁ78 °C f 0 °C.
Figure 3. Native CII259ꢁ273 glycopeptide 1 and analogues 31ꢁ36
modified with (E)-alkene and ethylene isosteres.
Scheme 4. Alternative Approach to the Gly-Phe Isosteresa
a Reagents and conditions: (a) and (b) see Allan et al.;45 (c) (i) Et3N, tBuCOCl, THF, ꢁ78 °C, (ii) (4R,5S)-4-methyl-5-phenyl-2-oxazolidinone, n-BuLi,
THF, ꢁ78 °C f rt (82%); (d) (i) TFA, CH2Cl2, rt, (ii) benzophenone imine, CH2Cl2, rt (72% from 27); (e) LDA, BnBr, THF, ꢁ78 °C f 0 °C (70% as
a mixture with dr 93:7); (f) LiOH, H2O2, THF/H2O, 0 °C; (g) (i) citric acid, THF/H2O, rt, (ii) FmocOSu, NaHCO3, acetone/H2O, rt, (iii) chiral
chromatography (66% from 28, 99% ee); (h) H2, Pd/C, MeOH, rt (84%).
14371
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
Table 1. Average and Standard Deviation for the Backbone
Dihedral Angle (Defined by CrꢁC0ꢁNꢁCr) for Each of the
Positions Ile260-Ala261, Ala261-Gly262, and Gly262-Phe263 for
the MD Trajectory between 6 and 18 ns
dihedral angle (°)
glycopeptide (position)
average
(SD
1
31
32
1
(Ile-Ala amide)
179.87
180.78
188.72
172.08
178.91
179.34
170.09
181.73
180.53
7.24
5.41
(Ile-Ala (E)-alkene)
(Ile-Ala ethylene)
(Ala-Gly amide)
13.69
8.35
33
34
1
(Ala-Gly (E)-alkene)
(Ala-Gly ethylene)
(Gly-Phe amide)
(Gly-Phe (E)-alkene)
(Gly-Phe ethylene)
5.25
13.07
7.40
35
36
5.41
11.55
of the Aq residues forming the P4 pocket were generally more
conserved than those forming the P1 pocket in the extracted
frames of the different complexes. This may be because the P4
pocket is defined by both the side chains and backbone of Aq,
whereas the P1 pocket is defined by several aromatic side chains.
These aromatic side chains seem to be more flexible in their
interactions with the glycopeptides.
The average values and standard deviations (SD) of the
backbone dihedral angles were comparable for the amides,
ethylenes, and (E)-alkenes at each of the three different positions
in the peptides (Table 1 and Supporting Information). Although
similar backbone conformations were adopted, the standard
deviations of the backbone dihedral angles revealed different
degrees of flexibility for the amide and the two isosteres. As
would be expected, the (E)-alkenes were slightly more rigid when
bound to Aq compared to the amides in the native 1, whereas the
ethylenes were more flexible.
Figure 4. Binding of native 1 and isostere glycopeptides 31ꢁ36 to Aq.
In the competitive binding experiments, increasing concentrations of the
glycopeptides 1 or 31ꢁ36 were incubated with recombinant Aq protein
and a fixed concentration of nonglycosylated and biotinylated CII259ꢁ
273. The amount of biotinylated peptide bound to Aq after incubation
was then detected in a time-resolved fluoroimmunoassay using euro-
pium-labeled streptavidin. The data are displayed as percent inhibition
compared to the biotinylated peptide in the absence of test glycopeptide.
The points represent the average of triplicates, and error bars are set to
(1 SD. Percent inhibition values are missing for 33 and 34 at 2.5 mM
due to abnormal data.
MD Simulations of Glycopeptide/Aq Complexes. MD
simulations were performed to study the dynamics of the
glycopeptide/Aq complexes over time. In particular, structural
variations and changes in the hydrogen-bonding patterns result-
ing from the introduced amide bond isosteres were monitored
and compared to Aq binding. The stronger Aq binding of the Gly-
Phe ethylene isostere 36 compared to the corresponding (E)-
alkene 35 was specifically addressed.
General Features of the Simulated Complexes. Seven 18 ns
MD simulations were performed of complexes between Aq and
the native 1 or one of the isostere glycopeptides 31ꢁ36. Residues
259ꢁ270 of each glycopeptide were included in the simulations,
which correspond to the glycopeptide sequence present in the
comparative model25,26 used as the starting structure. The root-
mean-square deviation (RMSD) of the backbone C and N atoms
from the corresponding initial structures was analyzed over the
simulation trajectories to verify the stability of the complexes
(Supporting Information). The RMSD profiles were similar for
all glycopeptide/Aq complexes and were generally found to
plateau within 6 ns, suggesting that a stable equilibrium had
been reached.
Visual inspection of frames extracted every 100 ps during the
last ns of the simulations showed that all glycopeptides had
similar Aq binding modes. As expected, the glycopeptides bound
with extended backbone conformations and with the side chains
of Ile260 and Phe263 anchored in the P1 and P4 pockets,
respectively. These anchor residues were identified in previous
experimental binding studies,33,59 and the extended peptide
backbone is generally observed in crystal structures of class II
MHC/peptide complexes.2 The positions and the conformations
Glycopeptide/Aq Hydrogen-Bond Interactions. Analysis of
the hydrogen-bonding network for the 259ꢁ270 sequence of
native 1 between 17 and 18 ns revealed a total of 11 possible
interactions with occupancy >5% between the glycopeptide
backbone and Aq (Table 2 (for definition of hydrogen bond)
and Figure 5). The results resembled the hydrogen-bond pattern
displayed in the comparative model26 of the Aq/CII260ꢁ267
complex where 13 possible hydrogen bonds were present; 7 of
which were identified in the MD simulations. Although the
glycopeptide and the Aq protein were able to form several
possible hydrogen bonds during the time of the simulation, not
all the bonds were present at a single occasion (snapshot). Six of
the hydrogen bonds occurred in more than 40% of the sampled
conformations. The average number of hydrogen bonds per
snapshot was 5.18 for the 259ꢁ270 sequence of 1, which is far
less than the 11 possible. Thus, although the static picture of the
peptide/MHC complex based on a comparative model revealed
several possible hydrogen bond interactions, the patterns and the
strengths of these interactions were found to vary greatly in the
dynamic system. The observations that less than half of the possible
hydrogen bonds are present as strong hydrogen bonds at any one
time and that the patterns varied between the different snapshots are
important findings and most likely of critical importance for
signaling in the class II MHC/peptide/TCR recognition system.
Most of the possible hydrogen-bond interactions, including 5
of the 6 hydrogen bonds with occupancy higher than 40%, were
14372
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
Table 2. Hydrogen-Bond Occupancya (in %) between Aq and the 259ꢁ270 Glycopeptide Backbones for Native 1 and the
Modified Glycopeptides 31ꢁ36 for the MD Trajectory between 17 and 18 ns
glycopeptide
residueb
31 Ile-Ala
32 Ile-Ala
33 Ala-Gly
34 Ala-Gly
35 Gly-Phe
36 Gly-Phe
Aq residuec
1 native
(E)-alkene
ethylene
(E)-alkene
ethylene
(E)-alkene
ethylene
d
Gly259 (a)
Gly259 (a)
Gly259 (d)
Ile260 (d)
Ala261 (d)
Ala261 (a)
Phe263 (d)
Phe263 (a)
Phe263 (a)
Lys264 (d)
Gly265 (d)
Gly265 (d)
Gly265 (a)
Glu266 (d)
Glu266 (a)
Gln267 (a)
Gly268 (a)
Gly268 (d)
Gly268 (d)
Pro269 (a)
βVal85 (m)
βHis81 (s)
RSer53 (s)
RSer53 (m)
βAsn82 (s)
βAsn82 (s)
RTyr9 (m)
RAsn62 (s)
RTyr22 (s)
βGlu74 (s)
RAsn62 (m)
RAsn62 (s)
RThr65 (s)
βTyr30 (s)
RAsn69 (s)
βTrp61 (s)
RHis68 (s)
RHis68 (s)
RAsn69 (s)
RHis68 (s)
ꢁ
ꢁ
52
ꢁ
ꢁ
10
12
ꢁ
ꢁ
ꢁ
14
ꢁ
ꢁ
28
ꢁ
6
53
ꢁ
35
41
72
19
63
ꢁ
61
ꢁ
ꢁ
ꢁ
8
ꢁ
ꢁ
ꢁ
48
ꢁ
5
ꢁ
ꢁ
ꢁ
57
modified
44
38
ꢁ
modified
54
7
53
modified
27
61
ꢁ
24
35
63
modified
57
ꢁ
ꢁ
40
ꢁ
modified
21
69
modified
ꢁ
17
ꢁ
58
57
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
59
ꢁ
30
ꢁ
ꢁ
ꢁ
ꢁ
44
ꢁ
ꢁ
ꢁ
ꢁ
36
13
46
ꢁ
17
24
61
ꢁ
6
55
ꢁ
44
57
17
74
ꢁ
16
ꢁ
ꢁ
ꢁ
ꢁ
62
41
60
ꢁ
70
ꢁ
68
70
ꢁ
ꢁ
25
22
ꢁ
ꢁ
32
ꢁ
ꢁ
ꢁ
ꢁ
18
ꢁ
40
ꢁ
35
30
ꢁ
ꢁ
39
ꢁ
a Calculated with the hydrogen-bond extension in VMD.60 A hydrogen bond was defined by a donorꢁacceptor distance of less than 3.3 Å and a donor-
H
acceptor angle of less than 20°. Only frequencies >5% are reported. b The residue involved in the hydrogen bond is defined as donor (d) or
3 3 3
acceptor (a). c The hydrogen bond is defined to involve either the side chain (s) or the main chain (m). d No hydrogen bond observed or the occupancy is
less than 5%.
reported for the related mouse class II MHC I-Ad 50ꢁ52 and
human class II MHC HLA-DR systems.20
The 3 amide bonds in the Ile260-Phe263 sequence of 1, which
were replaced by isosteres in this study, were each involved in one
hydrogen-bond interaction with Aq according to the MD data
(Table 2 and Figure 5). However, the effect of inserting either
(E)-alkene or ethylene moieties on Aq binding could not be
linked to the occupancy (i.e., the hydrogen-bond strength) of the
specific hydrogen bond that was lost. Because the MD simulation
of the complex of 1 and Aq revealed a dynamic hydrogen-bonding
network, we conducted a more detailed study of the corresponding
networks in the complexes of the modified glycopeptides and Aq,
focusing on the N-terminal part of the glycopeptides.
Although 7 possible hydrogen bonds were identified between
the 259ꢁ265 backbone sequence of 1 and residues in Aq, on
average only 3.47 interactions were present in each sampled
Figure 5. Hydrogen-bond network between the 259ꢁ270 backbone
sequence of 1 and the Aq protein during the last ns of the MD simulation.
complex conformation (Table 3). A correlation was found
The hydrogen-bond interactions (see Table 2 for definition) are
illustrated with dashed lines and are color coded according to their
occupancy, <40% (blue) and >40% (red).
between the average number of hydrogen bonds formed between
Aq and the “core” 259ꢁ265 sequence of the native and modified
glycopeptides and the experimentally determined Aq binding.
The group of glycopeptides that bound with moderate strength
to Aq, i.e., the (E)-alkenes 31, 33, and 35 as well as the Gly-Phe
found in the 259ꢁ265 sequence of 1 (Table 2 and Figure 5).
Thus, the hydrogen-bonding network between 1 and Aq was
ethylene 36, had on average 1.95ꢁ2.50 hydrogen bonds between
their 259ꢁ265 backbone sequences and residues in Aq (Table 3).
mainly formed by the N-terminal part of the glycopeptide and the
Aq protein. It should be noted that the 259ꢁ265 sequence is
The Ile-Ala (32) and Ala-Gly (34) ethylene glycopeptides, which
both bound weakly, had 1.25 and 1.84 interactions, respectively.
Thus, the loss of hydrogen-bonding interactions in the N-term-
almost identical to CII260ꢁ267, which was previously identified
as the minimal epitope of 1 required for Aq binding and T-cell
recognition.61 The importance of the hydrogen-bonding net-
inal 259ꢁ265 sequence of the modified glycopeptides probably
contributed to the observed loss of Aq binding. This result is
work in the N-terminal part of a peptide backbone has also been
14373
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
Table 3. Average Number of Hydrogen Bondsa with Standard
Deviation Present between Aq and the 259ꢁ265 Glycopep-
tide Backbones for the MD Trajectory between 17 and 18 ns
In summary, we found that the hydrogen-bonding network
between Aq and bound glycopeptides is highly dynamic over
time. Only half of the possible number of hydrogen bonds
between Aq and 1 were present in single snapshots, the majority
of which were found in the N-terminal 259ꢁ265 sequence of 1.
We also observed that site-specific introduction of an isostere
that necessitates the loss of one hydrogen-bond interaction had a
profound effect on the complete hydrogen-bonding network
between the N-terminal part of the glycopeptide backbone and
the Aq protein. The results from the MD study presented herein
show that these hydrogen bonds are strongly coupled to each
other. This finding is consistent with previous reports that
suggest the hydrogen-bond network formed between peptides
and class II MHC proteins follows a cooperative model in which
the strength and the presence of individual bonds are dependent
upon the integrity of neighboring interactions.2,50ꢁ52
number of hydrogen bonds
glycopeptide (position)
average
(SD
1
31
32
33
34
35
36
(native)
3.47
1.95
1.25
2.06
1.84
2.50
2.19
1.31
1.02
0.93
0.98
1.02
1.19
1.02
(Ile-Ala (E)-alkene)
(Ile-Ala ethylene)
(Ala-Gly (E)-alkene)
(Ala-Gly ethylene)
(Gly-Phe (E)-alkene)
(Gly-Phe ethylene)
a Calculated with the hydrogen-bond extension in VMD.60 A hydrogen
bond was defined by a donorꢁacceptor distance of less than 3.3 Å and a
Structural Changes in the Gly-Phe Ethylene Complex. The Aq
binding data revealed that the glycopeptide with an ethylene
isostere at the Gly-Phe position unexpectedly bound more
strongly to Aq than the corresponding (E)-alkene (cf. 36 and
35 in Figure 4), which is in contrast to the trend observed when
the isosteres were inserted at the two other positions. This
difference in binding is probably due to an improved Aq binding
of the ethylene isostere in the Gly-Phe position compared to the
other two positions rather than decreased binding strength of the
corresponding (E)-alkene. This is because similar interaction
patterns with Aq were observed for 35 as for alkenes 31 and 33,
and no conformational strain of the proposed bioactive con-
formation of 35 could be detected. The improved binding of 36
compared to 35 could be driven by enthalpy and/or entropy. In
the analysis of the trajectories no profound difference in move-
ments (flexibility) could be detected between the three ethylenes
or between the three alkenes when bound to Aq. Furthermore, in
this particular case it seems that the relative enthalpic and
entropic effects of water were small because the solvent acces-
sible surface of the complexes of the six isosteres and the native 1
at the three modified positions were similar throughout the MD
simulation. Based on these observations, together with the
assumption that the isostere glycopeptides have similar flexibility
and solvation in their unbound states, we suggest that the relative
difference in entropy between the different isosteres is too small
to account for the stronger binding of the Gly-Phe ethylene 36,
compared to the corresponding (E)-alkene 35. Instead, the
stronger binding of the Gly-Phe ethylene 36 could be mainly
enthalpy driven and directly dependent on molecular interac-
tions between 36 and Aq.
A main difference between the hydrogen-bonding network
involving the 259ꢁ265 sequences of Gly-Phe ethylene 36 and
(E)-alkene 35 was that the hydrogen bond between Phe263 and
RAsn62 of Aq was preserved in 36 to the same extent as observed
for 1, whereas this interaction was not observed in the complex
with the more rigid (E)-alkene 35. At the same time, the side
chain of RAsn62 formed a hydrogen bond to Gly265 in 35, which
was not observed for 36, whereas the main chain of RAsn62
interacted with Gly265 in 36, which was not observed for 35.
Furthermore, two hydrogen bonds (between Gly259 and βHis81
and Ile260 and RSer53, Table 2) at the N-terminal were weakened
for 36 compared to 35. Further analysis of the MD trajectories
revealed that the Aq/36 complex adopted an equilibrium con-
formation that was structurally different to the equilibrium con-
formations of the other Aq/glycopeptide complexes (Table 5 and
Supporting Information). This structural difference could be
donor-H acceptor angle of less than 20°.
3 3 3
Table 4. Average RMSDa for the R1 and β1 Helices of Aq in
Complex with Glycopeptides 1 and 31ꢁ36, Respectively, for
the MD Trajectory between 9 and 18 ns
RMSD R1 helix
RMSD β1 helix
glycopeptide
(native)
average
(SD
average
(SD
1
1.59
3.26
2.50
3.24
2.98
2.82
2.80
0.42
0.31
0.25
0.25
0.20
0.28
0.25
1.10
1.66
1.40
1.26
1.42
1.49
1.21
0.14
0.18
0.19
0.22
0.16
0.24
0.14
31 (Ile-Ala (E)-alkene)
32 (Ile-Ala ethylene)
33 (Ala-Gly (E)-alkene)
34 (Ala-Gly ethylene)
35 (Gly-Phe (E)-alkene)
36 (Gly-Phe ethylene)
a RMSD calculated with frame 1 as reference in the respective
simulation.
supported by earlier studies which showed that for peptides
binding to MHC class I molecules, the distribution of medium
and strong hydrogen bonds determined by short (200 ps)
MD simulations correlated with the binding potency of the
peptides.62
The analysis of the 259ꢁ265 fragments revealed that the
amide bond isosteres did not solely affect the hydrogen-bonding
interactions at the positions where they were introduced. In fact,
the different isosteres substantially altered the hydrogen-bonding
network along their glycopeptide backbones compared to the
native 1 (Table 2). These altered networks also affected the
dynamics and the structure of the helices that enclose the binding
site. The R1 helix of Aq in the complexes with the isosteres
adopted different conformations compared to the complex with
1 (cf. average RMSD values in Table 4). The β1 helix of Aq in the
complexes with the isosteres also adopted different conforma-
tions compared to the native complex, although the effect was
not as profound as for the R1 helix. In addition, the R1 helix of Aq
in the complex with 1 was more flexible than the β1 helix over the
simulation time (cf. SD for RMSD values in Table 4). Interest-
ingly, the helices of Aq in complex with isosteres showed a
different dynamic pattern, where the R1 helices were more rigid
and the β1 helices were slightly more flexible compared to the
native complex.
14374
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
Table 5. Average RMSD for the Sequence Ile260-Phe263 and χ1 Dihedral Angle for Phe263 Calculated for the MD Trajectory
between 6 and 18 ns
RMSD heavy atoms of Ile260-Phe263 (Å)a
dihedral angle χ1 (°) of the Phe side chain
glycopeptide
average
(SD
average
(SD
1
31
32
33
34
35
36
(native)
1.18
1.33
1.31
1.08
1.08
0.91
2.11
0.21
0.27
0.19
0.16
0.17
0.15
0.27
ꢁ80.48
ꢁ77.08
ꢁ75.16
ꢁ73.20
ꢁ76.83
ꢁ78.59
ꢁ101.79
9.22
12.17
9.55
(Ile-Ala (E)-alkene)
(Ile-Ala ethylene)
(Ala-Gly (E)-alkene)
(Ala-Gly ethylene)
(Gly-Phe (E)-alkene)
(Gly-Phe ethylene)
7.72
8.49
9.38
13.43
a RMSD calculated with frame 1 as reference in the respective simulation.
Figure 6. (a) Dihedral angle χ1 in the Phe263 residue versus simulation time (from 6 to 18 ns) for native 1, Gly-Phe (E)-alkene 35, and Gly-Phe ethylene
36. (b) Observed χ1 dihedral angle in the Phe263 residue illustrated for 1, defined by the bonds between the atoms NꢁCRꢁCβꢁCδ. Individual curves of
the dihedral angles for all glycopeptides are presented in the Supporting Information. (c) Example of superposed binding poses of glycopeptides
1 (blue), 35 (green), and 36 (red). The poses were selected at the end of the 18 ns simulation trajectory from complexes with an average RMSD for the
backbone heavy atoms.
traced to the Phe263 side chain in 36, whose dihedral angle χ1
differed significantly from the corresponding angles in the native 1
and the other isosteres 31ꢁ35 (Table 5, Figure 6, and Supporting
Information). As mentioned earlier, the side chain of the Phe263
residue is an important anchor that interacts with the P4 pocket
of Aq. Snapshots extracted from the simulation trajectories from
17 to 18 ns showed that the altered dihedral angle χ1 in 36
corresponded to a shift of the Phe side chain in the P4 pocket
(Figure 6). This allowed the Phe side chain to make new
interactions with residues in the P4 pocket, e.g., βTyr30, βPhe40,
and βPhe47, while at the same time some interactions were
weakened, e.g., with βVal78. It is possible that the flexibility
provided by the ethylene, which enabled the repositioning of the
Phe side chain in the P4 pocket, also allowed the Phe263-RAsn62
hydrogen bond to be preserved in the Aq complex with 36. One
may speculate that the increased binding strength of the complex
may be due to a cooperative effect of the van der Waals
interactions formed by the anchor residues of the glycopeptide
and the protein and the hydrogen-bonding network. Such
cooperative binding has recently been described for a series of
thrombin inhibitors.63 In summary, the altered interactions
between Phe263 of 36 and the Aq residues forming the P4 pocket,
together with the hydrogen-bonding network, may result in a
more favorable free energy of binding for 36 to Aq compared to 35.
T-Cell Recognition. The ability of the isostere-modified
glycopeptides to stimulate four Aq-restricted T-cell hybridomas,64
selected from groups with different specificities for the galactose
moiety of 1,32 was investigated in an assay of IL-2 secretion
(Figure 7). In general, T-cell stimulation is determined by the
strength of peptide binding to the class II MHC as well as by the
epitope displayed by the peptide/MHC complex to the TCR. In
this study, the native 1 induced similar strong response profiles
14375
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
Figure 7. Response of the Aq-restricted T-cell hybridomas HCQ.3, HCQ.10, 22a1-7E, and HM1R.2 to glycopeptides 1 and 31ꢁ36. The hybridomas
were evaluated by incubation with syngeneic spleen cells and increasing concentrations of 1 or 31ꢁ36. T-cell hybridomas that recognize glycopeptides
bound to Aq on the spleen cells secreted IL-2, which was quantified by a sandwich ELISA using the DELFIA system. The points represent the average of
duplicates. The fluorescence value for 36 at 9.38 μM for the HCQ.3 hybridoma is not reported due to a technical error when performing the assay.
for all four T-cell hybridomas. The (E)-alkene- and ethylene-
modified glycopeptides elicited weaker responses than 1 which,
at least partly, should be a result of their weaker binding to Aq. The
altered epitope that results from changes in the structure and the
dynamics of the R1 and β1 helices in Aq in the complexes with the
modified glycopeptides most likely provides another explanation
for the reduced responses. In addition, the four hybridomas
displayed very different response profiles for the modified glyco-
peptides, indicating that their TCRs have different specificities for
the presented Aq/glycopeptide epitopes. Hybridoma HCQ.3 re-
cognized all modified glycopeptides except Ile-Ala ethylene 32, and
generally the responses reflected how well the glycopeptides bound
to Aq. In contrast, HCQ.10 did not recognize any of the modified
glycopeptides. Hybridoma 22a1-7E recognized only the two Gly-
Phe modified glycopeptides 35 and 36 weakly, whereas HM1R.2
mainly recognized the two Ala-Gly modified glycopeptides 33 and
34, indicating that differences in the presented epitopes determine
which glycopeptides are recognized by these two hybridomas.
The Ile-Ala ethylene 32 was the only glycopeptide that was not
recognized by any of the T-cell hybridomas. The MD simulations
showed that the Ile side chain of this glycopeptide was not
positioned as deeply in the P1 pocket as in the complexes with
the other glycopeptides. This could be due to the high flexibility
of the ethylene isostere and the loss of hydrogen-bond interac-
tions, which probably alters the presentation of the epitope to
such an extent that T-cell recognition is completely abolished.
Changes in peptide structure that alter the epitope displayed to
the TCR, and in so doing dramatically influence T-cell recogni-
tion, have been described previously for a peptide bound by the
related murine I-Ek protein.65 Furthermore, we have previously
correlated differences in the presented epitope and the dynamics
of the glycopeptide/Aq complex to different T-cell responses for
glycopeptides with the Ile260 and Phe263 residues substituted
with non-natural amino acids.11 Previous reports have indicated
that the kinetics of the TCR’s interaction with the class II MHC/
peptide complex is a major determinant for T-cell activity.66ꢁ70 It
14376
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
is also likely that the dissociation rate of the TCR’s interaction
with the (E)-alkene- and ethylene-modified glycopeptides is an
important factor determining the T-cell responses observed in
this study. Furthermore, the different and dynamic hydrogen-
bond networks revealed by the MD studies may affect the
presentation of the glycopeptides by Aq and consequently
TCR recognition.
complexes plays a crucial role in determining if an immune
response is initiated via activation of the T-cell.
’ ASSOCIATED CONTENT
S
Supporting Information. Full experimental details for
b
synthesis of all compounds, the MHC-binding assay and the
T-cell recognition assay, details for the MD simulations, NMR
spectra for all new isolated compounds, Aq binding curves for
triplicates, complete ref 20, RMSD for Aq/glycopeptide com-
plexes vs simulation time, backbone dihedral angle for 1, 31, 33,
and 35 vs simulation time, backbone dihedral angle for 1, 32, 34,
and 36 vs simulation time, RMSD for Ile260-Phe263 sequences vs
simulation time, Phe dihedral angles vs simulation time, and
RMSD of the R1 and β1 helices of Aq vs simulation time. This
material is available free of charge via the Internet at http://pubs.
acs.org.
’ CONCLUSIONS
In this study the amides at positions Ile260-Ala261, Ala261
-
Gly262, and Gly262-Phe263 in the CII259ꢁ273 (1) glycopeptide
backbone were replaced by (E)-alkene and ethylene amide bond
isosteres to probe the effect of altering the hydrogen-bonding
capability on binding to the Aq class II MHC protein. Compe-
titive binding experiments revealed that modifying one out of 14
amide bonds in glycopeptide 1 reduced binding to Aq to a greater
extent than expected. Differences in the stability of the glycopep-
tide/Aq complexes were investigated by MD simulations, which
showed a dynamic network of hydrogen bonds between the
glycopeptide backbone and Aq.
’ AUTHOR INFORMATION
Corresponding Author
anna.linusson@chem.umu.se; jan.kihlberg@chem.umu.se
The Aq/1 complex displayed a lower number of hydrogen
bonds in each snapshot than both the total number of possible
hydrogen bonds observed during the simulation and the number
reported for a static model of the complex published recently.26
Thus, the introduced amide bond isosteres affected a smaller
hydrogen-bonding network than initially anticipated, providing
one possible explanation for the unexpectedly large reduction in
Aq binding. In addition, the MD simulations revealed that it was
primarily the N-terminal CII259ꢁ265 backbone sequence that
was involved in an extensive network of strong hydrogen-bond
interactions with Aq. This network was substantially altered by
the position and the nature of the introduced amide bond
isosteres, suggesting that a cooperative effect was responsible
for the reduction in binding. These findings are in close agree-
ment with previous reports of dense and cooperative hydrogen-
bonding networks in N-terminal parts of peptides bound by the
mouse class II MHC Ad 50ꢁ52 and human class II MHC HLA-DR
proteins.20
As expected, glycopeptides with the more rigid (E)-alkene
isostere were more strongly bound to Aq than glycopeptides with
the ethylene isosteres at the Ile-Ala and Ala-Gly positions.
Interestingly, the opposite relationship was found for the Gly-
Phe position, where the ethylene isostere bound more strongly to
Aq than the corresponding (E)-alkene. MD simulations revealed
that the increased flexibility of the ethylene isostere enabled
repositioning of the Phe side chain in the P4 pocket and allowed
the Phe263-RAsn62 hydrogen bond to be maintained, in contrast
to the corresponding (E)-alkene.
’ ACKNOWLEDGMENT
This work was funded by grants from the Swedish Research
Council (dnr 2008-3672), the Swedish Strategic Science Foun-
dation (SSF), the JC Kempe Foundation (SJCKMS), and the
EU project Masterswitch HEALTH-F2-2008-223404. The MD
simulations were conducted using the resources of the High
Performance Computing Center North (HPC2N).71 The
authors thank Anna Johansson at AstraZeneca R&D M€olndal
for chiral HPLC purification of compound 6.
’ REFERENCES
(1) Janeway, C. A.; Travers, P.; Walport, M.; Shlomchik, M. J.
Immunobiology; 6th ed.; Garland Science Publishing: Oxford, U.K.,
2005.
(2) McFarland, B. J.; Beeson, C. Med. Res. Rev. 2002, 22, 168–203.
(3) Stern, L. J.; Brown, J. H.; Jardetzky, T. S.; Gorga, J. C.; Urban,
R. G.; Strominger, J. L.; Wiley, D. C. Nature 1994, 368, 215–221.
(4) Yaneva, R.; Springer, S.; Zacharias, M. Biopolymers 2009,
91, 14–27.
(5) Weber, P.; Raynaud, I.; Ettouati, L.; Trescol-Biemont, M. C.;
Carrupt, P. A.; Paris, J.; Rabourdin-Combe, C.; Gerlier, D.; Testa, B. Int.
Immunol. 1998, 10, 1753–1764.
(6) Knapp, B.; Omasits, U.; Bohle, B.; Maillere, B.; Ebner, C.;
Schreiner, W.; Jahn-Schmid, B. Mol. Immunol. 2009, 46, 1839–1844.
(7) Knapp, B.; Omasits, U.; Schreiner, W.; Epstein, M. M. PLoS One
2010, 5, e11653.
(8) Toh, H.; Kamikawaji, N.; Tana, T.; Muta, S.; Sasazuki, T.;
Finally, the isostere-modified glycopeptides significantly af-
fected the interactions with four different T-cell hybridomas,
specific for CII259ꢁ273. A general reduction in T-cell response
was observed for all modified glycopeptides, which was assumed
to be due to the reduction in Aq binding and to structural and
dynamic changes in the R1 and β1 helices of Aq. The very
different response profiles that were displayed by the hybridomas
were most likely caused by differences in the hydrogen-bonding
networks and the secondary effects thereof, resulting in the
presentation of different glycopeptide/Aq epitopes to the TCRs.
The dynamic variation of the patterns and the strengths of the
hydrogen-bond interactions in the class II MHC/glycopeptide
Kuhara, S. Protein Eng. 2000, 13, 423–429.
(9) Toh, H.; Kamikawaji, N.; Tana, T.; Sasazuki, T.; Kuhara, S.
Protein Eng. 1998, 11, 1027–1032.
(10) Wan, S.; Flower, D. R.; Coveney, P. V. Mol. Immunol. 2008,
45, 1221–1230.
(11) Andersson, I. E.; Andersson, C. D.; Batsalova, T.; Dzhambazov,
B.; Holmdahl, R.; Kihlberg, J.; Linusson, L. PLoS One 2011, 6, e17881.
(12) De Rosa, M. C.; Giardina, B.; Bianchi, C.; Alinovi, C. C.; Pirolli,
D.; Ferraccioli, G.; De Santis, M.; Di Sante, G.; Ria, F. PLoS One 2010,
5, e11550.
(13) Smilek, D. E.; Wraith, D. C.; Hodgkinson, S.; Dwivedy, S.;
Steinman, L.; McDevitt, H. O. Proc. Natl. Acad. Sci. U.S.A. 1991,
88, 9633–9637.
14377
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378

Journal of the American Chemical Society
ARTICLE
(14) Gautam, A. M.; Lock, C. B.; Smilek, D. E.; Pearson, C. I.;
Steinman, L.; McDevitt, H. O. Proc. Natl. Acad. Sci. U.S.A. 1994,
91, 767–771.
(15) Ettouati, L.; Salvi, J. P.; Trescol-Biꢀemont, M. C.; Walchshofer,
N.; Gerlier, D.; Rabourdin-Combe, C.; Paris, J. Pept. Res. 1996,
9, 248–253.
(42) Ooi, T.; Kameda, M.; Taniguchi, M.; Maruoka, K. J. Am. Chem.
Soc. 2004, 126, 9685–9694.
(43) Maruoka, K.; Tayama, E.; Ooi, T. Proc. Natl. Acad. Sci. U.S.A.
2004, 101, 5824–5829.
(44) Chandra, A.; Viswanathan, R.; Johnston, J. N. Org. Lett. 2007,
9, 5027–5029.
(16) Mꢀeziꢁere, C.; Viguier, M.; Dumortier, H.; Lo-Man, R.; Leclerc,
C.; Guillet, J. G.; Briand, J. P.; Muller, S. J. Immunol. 1997, 159,
3230–3237.
(17) Cotton, J.; Hervꢀe, M.; Pouvelle, S.; Maillꢁere, B.; Mꢀenez, A. Int.
Immunol. 1998, 10, 159–166.
(18) Ettouati, L.; Casimir, J. R.; Raynaud, I.; Trescol-Biꢀemont, M. C.;
Carrupt, P. A.; Gerlier, D.; Rabourdin-Combe, C.; Testa, B.; Paris, J.
Protein Pept. Lett. 1998, 5, 221–229.
(45) Allan, R. D.; Dickenson, H. W.; Johnston, G. A. R.; Kazlauskas,
R.; Tran, H. W. Aust. J. Chem. 1985, 38, 1651–1656.
(46) Atherton, E.; Sheppard, R. C. Solid phase peptide synthesis. IRL
Press at Oxford University Press: Oxford, U.K., 1989.
(47) Herve, M.; Maillere, B.; Mourier, G.; Texier, C.; Leroy, S.;
Menez, A. Mol. Immunol. 1997, 34, 157–163.
(48) Cotton, J.; Herve, M.; Pouvelle, S.; Maillere, B.; Menez, A. Int.
Immunol. 1998, 10, 159–166.
(19) Smith, A. B., III; Benowitz, A. B.; Sprengeler, P. A.; Barbosa, J.;
Guzman, M. C.; Hirschmann, R.; Schweiger, E. J.; Bolin, D. R.; Nagy, Z.;
Campbell, R. M.; Cox, D. C.; Olson, G. L. J. Am. Chem. Soc. 1999,
121, 9286–9298.
(49) de Haan, E. C.; Wauben, M. H. M.; Grosfeld-Stulemeyer, M. C.;
Kruijtzer, J. A. W.; Liskamp, R. M. J.; Moret, E. E. Bioorg. Med. Chem.
2002, 10, 1939–1945.
(50) McFarland, B. J.; Katz, J. F.; Beeson, C.; Sant, A. J. Proc. Natl.
(20) Bolin, D. R.; et al. J. Med. Chem. 2000, 43, 2135–2148.
(21) Hart, M.; Beeson, C. J. Med. Chem. 2001, 44, 3700–3709.
(22) Koehler, N. K. U.; Yang, C. Y.; Varady, J.; Lu, Y.; Wu, X.; Liu,
M.; Yin, D.; Bartels, M.; Xu, B.; Roller, P. P.; Long, Y.; Li, P.; Kattah, M.;
Cohn, M. L.; Moran, K.; Tilley, E.; Richert, J. R.; Wang, S. J. Med. Chem.
2004, 47, 4989–4997.
(23) Rosloniec, E. F.; Brandstetter, T.; Leyer, S.; Schwaiger, F. W.;
Nagy, Z. A. J. Autoimmunity 2006, 27, 182–195.
(24) Boots, A. M. H.; Hubers, H.; Kouwijzer, M.; Zandbrink, L. D.;
Westrek-Esselink, B. M.; van Doorn, C.; Stenger, R.; Bos, E. S.; van
Lierop, M. J. C.; Verheijden, G. F.; Timmers, C. M.; van Staveren, C. J.
Arthritis Res. Ther. 2007, 9, R71.
(25) Andersson, I. E.; Batsalova, T.; Dzhambazov, B.; Edvinsson, L.;
Holmdahl, R.; Kihlberg, J.; Linusson, A. Org. Biomol. Chem. 2010,
8, 2931–2940.
(26) Andersson, I. E.; Dzhambazov, B.; Holmdahl, R.; Linusson, A.;
Kihlberg, J. J. Med. Chem. 2007, 50, 5627–5643.
Acad. Sci. U.S.A. 2001, 98, 9231–9236.
(51) McFarland, B. J.; Katz, J. F.; Sant, A. J.; Beeson, C. J. Mol. Biol.
2005, 350, 170–183.
(52) McFarland, B. J.; Beeson, C.; Sant, A. J. J. Immunol. 1999,
163, 3567–3571.
(53) Cram, D. J. Angew. Chem., Int. Ed. 1986, 25, 1039–1134.
(54) Houk, K. N.; Leach, A. G.; Kim, S. P.; Zhang, X. Y. Angew.
Chem., Int. Ed. 2003, 42, 4872–4897.
(55) Diehl, C.; Engstr€om, O.; Delaine, T.; Hakansson, M.; Genhe-
den, S.; Modig, K.; Leffler, H.; Ryde, U.; Nilsson, U. J.; Akke, M. J. Am.
Chem. Soc. 2010, 132, 14577–14589.
(56) DeLorbe, J. E.; Clements, J. H.; Teresk, M. G.; Benfield, A. P.;
Plake, H. R.; Millspaugh, L. E.; Martin, S. F. J. Am. Chem. Soc. 2009,
131, 16758–16770.
(57) Baum, B.; Muley, L.; Smolinski, M.; Heine, A.; Hangauer, D.;
Klebe, G. J. Mol. Biol. 2010, 397, 1042–1054.
ꢀ
(58) Ward, J. M.; Gorenstein, N. M.; Tian, J. H.; Martin, S. F.; Post,
(27) Holm, L.;Bockermann, R.;Wellner, E.;B€acklund, J.; Holmdahl, R.;
Kihlberg, J. Bioorg. Med. Chem. 2006, 14, 5921–5932.
(28) B€acklund, J.; Treschow, A.; Bockermann, R.; Holm, B.; Holm,
L.; Issazadeh-Navikas, S.; Kihlberg, J.; Holmdahl, R. Eur. J. Immunol.
2002, 32, 3776–3784.
C. B. J. Am. Chem. Soc. 2010, 132, 11058–11070.
(59) Rosloniec, E. F.; Whittington, K. B.; Brand, D. D.; Myers, L. K.;
Stuart, J. M. Cell. Immunol. 1996, 172, 21–28.
(60) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996,
14, 33–38.
(29) Dzhambazov, B.; Nandakumar, K. S.; Kihlberg, J.; Fugger, L.;
Holmdahl, R.; Vestberg, M. J. Immunol. 2006, 176, 1525–1533.
(30) B€acklund, J.; Carlsen, S.; H€oger, T.; Holm, B.; Fugger, L.;
Kihlberg, J.; Burkhardt, H.; Holmdahl, R. Proc. Natl. Acad. Sci. U.S.A.
2002, 99, 9960–9965.
(31) Broddefalk, J.; Forsgren, M.; Sethson, I.; Kihlberg, J. J. Org.
Chem. 1999, 64, 8948–8953.
(32) Holm, B.; B€acklund, J.; Recio, M. A. F.; Holmdahl, R.; Kihlberg,
J. ChemBioChem 2002, 3, 1209–1222.
(61) Holm, L.; Kjellꢀen, P.; Holmdahl, R.; Kihlberg, J. Bioorg. Med.
Chem. 2005, 13, 473–482.
(62) Rognan, D.; Krebs, S.; Kuonen, O.; Lamas, J. R.; deCastro,
J. A. L.; Folkers, G. J. Comput.-Aided Mol. Des. 1997, 11, 463–478.
(63) Muley, L.; Baum, B.; Smolinski, M.; Freindorf, M.; Heine, A.;
Klebe, G.; Hangauer, D. G. J. Med. Chem. 2010, 53, 2126–2135.
(64) Corthay, A.; B€acklund, J.; Broddefalk, J.; Micha€elsson, E.;
Goldschmidt, T. J.; Kihlberg, J.; Holmdahl, R. Eur. J. Immunol. 1998,
28, 2580–2590.
(33) Kjellꢀen, P.; Brunsberg, U.; Broddefalk, J.; Hansen, B.; Vestberg,
M.; Ivarsson, I.; Engstr€om, Å.; Svejgaard, A.; Kihlberg, J.; Fugger, L.;
Holmdahl, R. Eur. J. Immunol. 1998, 28, 755–767.
(65) Kersh, G. J.; Miley, M. J.; Nelson, C. A.; Grakoui, A.; Horvath,
S.; Donermeyer, D. L.; Kappler, J.; Allen, P. M.; Fremont, D. H.
J. Immunol. 2001, 166, 3345–3354.
(34) Venkatesan, N.; Kim, B. H. Curr. Med. Chem. 2002, 9, 2243–
2270.
(66) Kersh, G. J.; Kersh, E. N.; Fremont, D. H.; Allen, P. M.
Immunity 1998, 9, 817–826.
(35) Wiktelius, D.; Luthman, K. Org. Biomol. Chem. 2007, 5, 603–605.
(36) Ley, S. V.; Anthony, N. J.; Armstrong, A.; Brasca, M. G.; Clarke,
T.; Culshaw, D.; Greck, C.; Grice, P.; Jones, A. B.; Lygo, B.; Madin, A.;
Sheppard, R. N.; Slawin, A. M. Z.; Williams, D. J. Tetrahedron 1989,
45, 7161–7194.
(37) Guan, Y. C.; Wu, J. L.; Sun, L.; Dai, W. M. J. Org. Chem. 2007,
72, 4953–4960.
(38) He, J. F.; Wu, Y. L. Tetrahedron 1988, 44, 1933–1940.
(39) Platzek, J.; Snatzke, G. Tetrahedron 1987, 43, 4947–4968.
(40) Fu, Y.; Bieschke, J.; Kelly, J. W. J. Am. Chem. Soc. 2005, 127,
15366–15367.
(67) Lyons, D. S.; Lieberman, S. A.; Hampl, J.; Boniface, J. J.; Chien,
Y. H.; Berg, L. J.; Davis, M. M. Immunity 1996, 5, 53–61.
(68) Matsui, K.; Boniface, J. J.; Steffner, P.; Reay, P. A.; Davis, M. M.
Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 12862–12866.
(69) McKeithan, T. W. Proc. Natl. Acad. Sci. U.S.A. 1995,
92, 5042–5046.
(70) Rabinowitz, J. D.; Beeson, C.; Lyons, D. S.; Davis, M. M.;
McConnell, H. M. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 1401–1405.
(71) High Performance Computing Center North (HPC2N); HPC2N:
ꢀ
Umea, Sweden; www.hpc2n.umu.se/ (accessed Apr 27, 2011).
(41) Odonnell, M. J.; Bennett, W. D.; Bruder, W. A.; Jacobsen,
W. N.; Knuth, K.; Leclef, B.; Polt, R. L.; Bordwell, F. G.; Mrozack, S. R.;
Cripe, T. A. J. Am. Chem. Soc. 1988, 110, 8520–8525.
14378
dx.doi.org/10.1021/ja2038722 |J. Am. Chem. Soc. 2011, 133, 14368–14378