Rational design and evaluation of a branched-chain-containing glycolipid antigen that binds to cd1d

Article abstract of DOI:10.1002/asia.201000120

Circumnavigating or clamping: CD1 molecules recognize glycolipid antigens with a straight alkyl chain moiety. Owing to the presence of the central pole in the A pocket of the CD1 binding groove, the straight alkyl chain of antigens circles around the pole

Full text of DOI:10.1002/asia.201000120

COMMUNICATION
DOI: 10.1002/asia.201000120
Rational Design and Evaluation of a Branched-Chain-Containing Glycolipid
Antigen That Binds to CD1d
Dong Jae Baek,[a] Yoon-Sook Lee,[a] Chaemin Lim,[a] Doohyun Lee,[a] Taeho Lee,[a]
Jae-Young Lee,[a] Kyoo-A Lee,[a] Won-Jea Cho,[b] Chang-Yuil Kang,[a] and
Sanghee Kim*[a]
CD1 molecules are cell surface glycoproteins that are
mainly expressed by cells of the hematopoietic lineage. The
main function of CD1 is to recognize glycolipid antigens
and present them to T cells for activation of the immune
system.[1] Since the discovery of the role of CD1, it has been
of considerable interest to identify the foreign and self-lipid
antigens that can be presented by CD1 molecules. The
known CD1-presented lipid antigens have diverse molecular
architectures, but they share a long, straight alkyl chain
moiety in their lipid structures as a common structural fea-
ture.[2] The discovery of structurally distinct CD1 ligands
would enhance our understanding of the biology of lipid rec-
ognition by T cells and would provide opportunities for the
development of immune modulating agents. Thus, it would
be of interest to design and synthesize novel ligands that
structurally differ from the hitherto known stereotypical
CD1-binding lipid antigens.
Currently, there are five known isoforms of CD1 (CD1a–
CD1e) in mammalian species. The crystal structures of
mouse CD1d and human CD1a, CD1b, and CD1d have
been solved.[2] The binding grooves of CD1 molecules are
hydrophobic and accommodate the hydrophobic lipid chains
of glycolipid antigens. The hydrophilic head group of the
glycolipid protrudes from the CD1 protein surface to be rec-
ognized by T cells. The structure and position of the hydro-
philic portion of the glycolipid have proven to be key factors
for the activation of T cells. More recently, however, it has
been shown that subtle differences in the hydrophobic lipid
structure also significantly impact T cell activation.[2,3]
Although the number of pockets in the CD1 binding
groove varies among the isoforms, there are two major hy-
drophobic pockets (A’ and F’). An interesting and common
feature of all CD1 isoforms is that a central pole, formed by
residues Phe70 and Val12/Cys12, transects the A’ pocket.[2]
Thus, the A’ pocket forms a narrow curved tunnel around
the pole. CD1a has a curved A’ pocket with a defined termi-
nus, whereas CD1b and CD1d have a curved A’ pocket that
forms a tunnel with two openings (Figure 1a).
Owing to the presence of the central pole in the A’
pocket, the straight alkyl chain of a glycolipid must circle
around the pole like a hook as shown in Figure 1a.[2] When
considering the architecture of the A’ pocket, it is plausible
that it could be filled with something other than a straight
alkyl chain. We envisioned that a branched alkyl chain
might also bind in the A’ pocket of CD1b and CD1d as
[a] D. J. Baek, Y.-S. Lee, C. Lim, D. Lee, Dr. T. Lee, J.-Y. Lee,
K.-A. Lee, Prof. C.-Y. Kang, Prof. S. Kim
College of Pharmacy
Seoul National University
San 56-1, Shilim, Kwanak, Seoul 151-742 (Korea)
Fax : (+82)2-762-8322
E-mail: pennkim@snu.ac.kr
Figure 1. Design of non-stereotypical glycolipid antigens. a) Crystal struc-
tures of ligand-bound hCD1a (Protein Data Bank (PDB) code 1ONQ),
hCD1b (PDB code 1GZQ), and hCD1d (PDB code 2PO6). The bound
ligands are sulfatide, phosphatidylinositol, and a-GalCer, respectively.
The A’ pocket region is indicated by yellow shading, and the central pole
is marked with an arrow. b) Schematic representation of the envisioned
binding mode of the branched analogues inside the A’ pocket of CD1.
[b] Prof. W.-J. Cho
College of Pharmacy
Chonnam National University
Yongbong, Buk, Kwangju 500-757 (Korea)
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/asia.201000120.
1560
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Asian J. 2010, 5, 1560 – 1564

schematically represented in Figure 1b. Two terminal alkyl
groups of the branched chain were envisaged to be inde-
pendently delivered into the curved Aꢁ pocket through each
of the two openings and to be anchored in place, not by cir-
cumnavigating, but by clamping onto the central pole. Thus,
we decided to investigate this type of ligand to identify non-
stereotypical CD1-binding antigens.
acyl chain-containing a-GalCer analogues 3–5, 7–9, and 11–
13 (Scheme 1) were designed in a combinatorial manner. In
addition, a series of analogues 6, 10, and 14–22 were also de-
signed as negative controls, in which the branched acyl
chains were too long for the A’ pocket binding.
As a proof of the concept, we chose a-GalCer (1,
Figure 2) as a template for designing the branched-chain-
containing ligands. a-GalCer is a well-characterized CD1d
Figure 2. Structures of the glycolipids 1 and 2.
ligand and the first-defined potent agonist of natural killer T
(NKT) cells.[4] When the a-GalCer/CD1d complex is recog-
nized by the conserved T cell receptor (TCR) of NKT cells,
the NKT cells are activated and promptly secrete large
amounts of T helper 1 (Th1) and Th2 cytokines, such as in-
terferon-g (IFN-g) and interleukin-4 (IL-4), respectively.[5]
The released cytokines play critical roles in inducing a series
of cellular events; Th1 cytokine production is believed to
correlate with antitumor, antiviral/antibacterial, and adju-
vant activities, whereas Th2 cytokine production is thought
to ameliorate autoimmune diseases.[6] Although a-GalCer
has been the most extensively studied valuable compound in
exploration of NKT cell biology,[6b,7] there are several unfav-
orable properties that hinder its clinical application.[8] One
of these is its poor solubility in both aqueous and organic
media.[9] The other major drawback is that a-GalCer induces
the secretion of Th1 and Th2 cytokines without preferen-
ce.[8b,c] Therefore, it is highly desirable to identify com-
pounds with better solubility and cytokine-inducing selectiv-
ity.
Scheme 1. Synthetic scheme for glycolipids 3–22. The glycolipids designed
as negative controls are shown in red. Reagents and conditions:
a) HOOCACHTNURTGENNG(U CH2)nAHCUTGNTERNN(UNG CmH2m+1)2, 1-ethyl-3-(3-dimethylaminopropyl)carbodi-
imide, 4-(dimethylamino)pyridine, CH2Cl2, RT, 75–84%. b) H2, Pd(OH)2,
EtOH/CH2Cl2 (3:1), RT, 73–82%. PMB=p-methoxybenzyl.
For the synthesis of the designed compounds, we em-
ployed the known a-galactosyl phytosphingosine 23[12]
(Scheme 1) as the starting material. Acylation of the amino
group of 23 with branched acids followed by global depro-
tection of the benzyl-type protecting groups in 24 by hydro-
genolysis afforded the desired branched a-GalCer analogues
3–22. It is worth noting that the branched glycolipids synthe-
sized in this study had considerably higher solubilities in
both aqueous and organic solvents than a-GalCer. For ex-
ample, the solubility of the glycolipid 7 at room temperature
is approximately 20 mgmLÀ1 in DMSO, 12 mgmLÀ1 in etha-
nol, and 18 mgmLÀ1 in water.
Using X-ray crystal structure information of the CD1d/a-
GalCer complex[10,11] (Figure 1a), we designed branched-
chain-containing a-GalCer derivatives as depicted in general
structure 2 (Figure 2). In this study, a nitrogen atom was
used as the branch point. The optimal number of carbon
atoms between the carbonyl carbon atom and the nitrogen
atom was six to eight, which caused the nitrogen atom to be
located close to the entrance of the A’ pocket. The optimal
length of the two alkyl chains attached to the nitrogen atom
to effectively fill the A’ pocket was approximately eight
carbon atoms. Larger alkyl substituents were likely to
exceed the size of the pocket. Thus, a series of branched
As a preliminary evaluation of the branched acyl chain
analogues, mouse CD1d-specific NKT hybridoma cells
(DN32.D3)[13] were used; these cells are immortalized NKT
cells and tend to produce IL-2 upon stimulation. As shown
in Figure 3a, the rationally designed analogues 3–5, 7–9, and
11–13, where n=6 to 8 and m=7 to 9, demonstrated greater
IL-2 secretion than a-GalCer (1). The negatively designed,
longer-alkyl-chained analogues, such as compounds with n=
10 (19–22) and m=12 (6, 10, 14, and 18), were virtually in-
active. By contrast, the analogues with n=9 and m=7 to 9
(15–17) were effective at eliciting the production of IL-2,
but were less efficient than the corresponding analogues
Chem. Asian J. 2010, 5, 1560 – 1564
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemasianj.org
1561

S. Kim et al.
COMMUNICATION
ing causes was the stability of the CD1d/glycolipid com-
plex.[16] Since Thl cytokine production requires longer TCR
stimulation than Th2, it has been believed that the more
stable CD1d/glycolipid complex induces a biased Th1 re-
sponse.[8c,17] Thus, we assessed the stability of the branched
analogue/CD1d complex to examine the origin of the en-
hanced Th2 selectivity. NKT hybridoma cells were stained
with mouse CD1d dimers that were loaded with the selected
analogues (7–9) that were more efficient than a-GalCer
with respect to IL-2 production. Each analogue-loaded
CD1d dimer stained NKT cells more strongly than the
dimer loaded with a-GalCer (Figure 4a). These results are
Figure 3. Biological evaluation of glycolipids 3–22. a) IL-2 secretion by
DN32.D3 NKT hybridoma cells. IL-2 production was measured in cocul-
tured supernatants of DN32.D3 cells and mouse CD1d-transfected RBL
cells loaded with glycolipids after 16 h of culture. Representative data of
two individual experiments are expressed as the mean ÆSD of dupli-
cates. b) IFN-g and IL-4 secretion by mouse splenocytes when stimulated
with glycolipids. IFN-g and IL-4 production was measured after 72 h of
culture. Results are expressed as relative activity. Representative data of
two individual experiments are expressed as the mean ÆSD of tripli-
cates.
Figure 4. Binding properties of the branched analogues. a) CD1d dimer
binding of glycolipids 7–9 to NKT hybridoma cells. Mean fluorescence in-
tensity (MFI) of dimer binding is plotted against glycolipid concentration.
Representative data of two individual experiments are shown. b) Stability
of the CD1d/glycolipid 7 complexes. CD1d/glycolipid complexes were in-
cubated for 24 h and washed at set times before the addition of DN32.D3
cells. IL-2 production was measured after 24 h of culture (in duplicate
cultures). Results are expressed as the relative quantification value.
with n=6 to 8. Since IL-2 secretion is dependent on anti-
gen-loaded CD1d,[14] these results imply that branched ana-
logues of the proper length could efficiently bind to CD1d
and trigger TCR signaling.
Because the balance of Th1 and Th2 cytokines is crucial
to an immune response,[5c,6a,8] the cytokine release profiles
of the branched analogues were assessed by measuring the
levels of IFN-g and IL-4 in in vitro culture supernatants of
mouse splenocytes stimulated with the compounds. Fig-
ure 3b shows the relative IFN-g and IL-4 production levels
resulting from treatment with the branched analogues as
compared with those resulting from a-GalCer. The patterns
of cytokine release in response to the tested compounds
were well correlated with IL-2 secretion. Interestingly, all of
the positive compounds induced NKT cell cytokine respons-
es with a Th2 bias. A Th2-biased cytokine profile was also
observed in an in vivo evaluation in mice (Figure S1 in the
Supporting Information).
indicative of efficient binding of 7–9 to CD1d and higher af-
finity of the branched analogue/CD1d complex to the TCR.
Additionally, we assessed the stability of the complex of 7
with CD1d by performing an ꢂoff rateꢁ stimulation assay.[10b]
We found that the branched analogue/CD1d complex, simi-
lar to the a-GalCer/CD1d complex, has a very long half-life
and slow dissociation rate (Figure 4b). Although more evi-
dence is needed, these results suggest that stability of the
CD1d/glycolipid complex might not be a major contributing
cause in shifting the cytokine profile of activated NKT
cells.[5c]
The above biological results strongly suggest that the
branched acyl chains of the analogues are favorably accom-
modated within the CD1d binding groove. However, they
do not unequivocally reveal the binding mode. Thus, we per-
Although the shift of Thl and Th2 cytokine release profile
could result from numerous factors,[5c,15] one of the convinc-
1562
www.chemasianj.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Asian J. 2010, 5, 1560 – 1564

Branched-Chain-Containing Glycolipid Antigen That Binds to CD1d
formed molecular modeling studies on the interaction of
CD1d with the selected analogues. Surflex–Dock was uti-
lized to search for the best possible geometry of the ana-
logue within the CD1d binding groove. As shown in
Figure 5 (see also Figure S2 in the Supporting Information),
sight into NKT cell biology and lipid–CD1d interactions as
well as for ameliorating the battle against autoimmune dis-
eases.
Acknowledgements
This work was supported by grants from the National Research Labora-
tory program (M10500000055-06J0000-05510) and the Korea Biotech
R&D group of the next-generation growth engine project
(F104AC010004-08A0301-00420). Y.-S.L. acknowledges support from the
Seoul Science Fellowship.
Keywords: antigens · cytokines · drug design · glycolipids ·
medicinal chemistry
[1] a) S. Porcelli, C. T. Morita, M. B. Brenner, Nature 1992, 360, 593–
597; b) M. S. Vincent, J. E. Gumperz, M. B. Brenner, Nat. Immunol.
2003, 4, 517–523; c) G. Bricard, S. A. Porcelli, Cell. Mol. Life Sci.
2007, 64, 1824–1840.
[2] a) D. B. Moody, D. M. Zajonc, I. A. Wilson, Nat. Rev. Immunol.
2005, 5, 387–399; b) D. M. Zajonc, M. Kronenberg, Curr. Opin.
Struct. Biol. 2007, 17, 521–529; c) J. D. Silk, M. Salio, J. Brown,
E. Y. Jones, V. Cerundolo, Annu. Rev. Cell Dev. Biol. 2008, 24, 369–
395.
[3] a) K. O. Yu, J. S. Im, A. Molano, Y. Dutronc, P. A. Illarionov, C. For-
estier, N. Fujiwara, I. Arias, S. Miyake, T. Yamamura, Y.-T. Chang,
G. S. Besra, S. A. Porcelli, Proc. Natl. Acad. Sci. USA 2005, 102,
3383–3388; b) C. McCarthy, D. Shpherd, S. Fleire, V. S. Stronge, M.
Koch, P. A. Illarionov, G. Bossi, M. Salio, G. Denkberg, F. Redding-
ton, A. Tarlton, B. G. Reddy, R. R. Schmidt, Y. Reiter, G. M. Grif-
fiths, P. A. Merwe, G. S. Besra, E. Y. Jones, F. D. Batista, V. Cerun-
dolo, J. Exp. Med. 2007, 204, 1131–1144.
[4] T. Kawano, J. Cui, Y. Koezuka, I. Toura, Y. Kaneko, K. Motoki, H.
Ueno, R. Nakagawa, H. Sato, E. Kondo, H. Koseki, M. Taniguchi,
Science 1997, 278, 1626–1629.
[5] a) A. Bendelac, P. B. Savage, L. Teyton, Annu. Rev. Immunol. 2007,
25, 297–336; b) W. C. Florence, R. K. Bhat, S. Joyce, Expert Rev.
Mol. Med. 2008, 10, e20; c) M. Kronenberg, Annu. Rev. Immunol.
2005, 23, 877–900.
Figure 5. Docking model of compound 7 (green backbone) within the
hCD1d (PDB code 2PO6) binding groove. H bonds are shown as red
dotted lines. The amino acid residues that form the central pole are
shown in pink.
the branched acyl chain in analogue 7 effectively filled the
A’ pocket as we hypothesized. The nitrogen atom of the
branched chain was located close to the entrance of the A’
pocket. The two alkyl chains attached to the nitrogen atom
were introduced into each of the two openings of the curved
tunnel and fitted into the A’ pocket. Docking of compounds
with a much longer alkyl chain showed that the branched
acyl chain was oversized and did not fit into the A’ pocket
properly.
[6] a) A. Balato, D. Unutmaz, A. A. Gaspari, J. Invest. Dermatol. 2009,
129, 1628–1642; b) M. Taniguchi, M. Harada, S. Kojo, T. Nakayama,
H. Wakao, Annu. Rev. Immunol. 2003, 21, 483–513.
In this work, we demonstrated the feasibility of immune
system activation by non-stereotypical glycolipid antigens.
This study not only adds to our understanding of the struc-
tural and functional aspects of lipid presentation by CD1
molecules, but also provides a rational basis for the design
of new CD1-mediated immune modulating agents. Replace-
ment of the long straight acyl chain moiety of a-GalCer gly-
colipid with a branched acyl chain did not impair CD1d
binding and NKT cell-activating abilities. The binding mode
of the branched analogues in the Aꢁ pocket of CD1d would
differ from that of a-GalCer; the two alkyl chains were in-
dependently delivered into the curved Aꢁ pocket through
each of the two openings. The branched analogue has inter-
esting and distinct properties, such as high solubility, high
antigenic potency, and Th2-biased NKT cell cytokine re-
sponse, whereas a-GalCer has a low solubility and unbiased
cytokine release profile, which have limited its utility in
many aspects. We believe thus that our newly developed
branched glycolipids would be useful in gaining further in-
[7] a) D. Wu, M. Fujioa, C.-H. Wong, Bioorg. Med. Chem. 2008, 16,
1073–1083; b) P. B. Savage, L. Teytonc, A. Bendelac, Chem. Soc.
Rev. 2006, 35, 771–779.
[8] a) G. Giaccone, C. J. Punt, Y. Ando, R. Ruijter, N. Nishi, M. Peters,
B. M. E. Blomberg, R. J. Scheper, H. J. J. Vilet, A. J. M. Eertwegh,
M. Roelvink, J. Beijnen, H. Zwierzina, H. M. Pinedo, Clin. Cancer
Res. 2002, 8, 3702–3709; b) M. J. Smyth, D. I. Godfrey, Nat. Immu-
nol. 2000, 1, 459–460; c) C. R. Berkers, H. Ovaa, Trends Pharmacol.
Sci. 2005, 26, 252–257.
[9] Y. Liu, R. D. Goff, D. Zhou, J. Mattner, B. A. Sullivan, A. Khurana,
C. 3rd Cantu, E. V. Ravkov, C. C. Ibegbu, J. D. Altman, L. Teyton,
A. Bendelac, P. B. Savage, J. Immunol. Methods 2006, 312, 34–39.
[10] a) M. Koch, V. S Stronge, D. Shepherd, S. D Gadola, B. Mathew, G.
Ritter, A. R Fersht, G. S Besra, R. R Schmidt, E Y. Jones, V. Cerun-
dolo, Nat. Immunol. 2005, 6, 819–826; b) D. M. Zajonc, C. 3rd
Cantu, J. Mattner, D. Zhou, P. B. Savage, A. Bendelac, I. A. Wilson,
L. Teyton, Nat. Immunol. 2005, 6, 810–818.
[11] a) N. Borg, K. S. Wun, L. Kjer-Nielsen, M. C. J. Wilce, D. G. Pellicci,
R. Koh, G. S. Besra, M. Bharadwaj, D. I. Godfrey, J. McCluskey, J.
Rossjohn, Nature 2007, 448, 44–49; b) D. G. Pellicci, O. Patel, L.
Kjer-Nielsen, S. S. Pang, L. C. Sullivan, K. Kyparissoudis, A. G.
Brooks, H. H. Reid, S. Gras, I. S. Lucet, R. Koh, M. J. Smyth, T.
Chem. Asian J. 2010, 5, 1560 – 1564
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemasianj.org
1563

S. Kim et al.
COMMUNICATION
Mallevaey, J. L. Matsuda, L. Gapin, J. McCluskey, D. I. Godfrey, J.
Rossjihn, Immunity 2009, 31, 47–59.
16469; b) L. Leung, C. Tomassi, K. V. Beneden, T. Decruy, M. Trap-
peniers, D. Elewaut, Y. Gao, T. Elliott, A. Al-Shamkhani, C. Otten-
smeier, J. M. Werner, A. Williams, S. V. Calenbergh, B. Linclau,
ChemMedChem 2009, 4, 329–334.
[12] W. Du, J. Gervay-Hague, Org. Lett. 2005, 7, 2063–2065.
[13] D. Zhou, J. Mattner, C. 3rd Cantu, N. Schrantz, N. Yi, Y. Gao, Y.
Sagiv, K. Hudspeth, Y.-P. Wu, T. Yamashita, S. Teneberg, D. Wang,
R. L. Proia, S. B Levery, P. B. Savage, L. Teyton, A. Bendelac, Sci-
ence 2004, 306, 1786–1789.
[14] a) S. Jiang, D. S. Game, D. Davies, G. Lombardi, R. I. Lechler, Eur.
J. Immunol. 2005, 35, 1193–1200; b) T. I. Prigozy, O. Naidenko, P.
Qasba, D. Elewaut, L. Brossay, A. Khurana, T. Natori, Y. Koezuka,
A. Kulkarni, M. Kronenberg, Science 2001, 291, 664–667.
[15] a) M. Trappeniers, K. V. Beneden, T. Decruy, U. Hillaert, B. Linclau,
D. Elewaut, S. V. Calenbergh, J. Am. Chem. Soc. 2008, 130, 16468–
[16] a) M. Fujio, D. Wu, R. Garcia-Navarro, D. D. Ho, M. Tsuji, C.-H.
Wong, J. Am. Chem. Soc. 2006, 128, 9022–9023; b) R. D. Goff, Y.
Gao, J. Mattner, D. Zhou, N. Yin, C. 3rd Cantu, L. Teyton, A. Bend-
elac, P. B. Savage, J. Am. Chem. Soc. 2004, 126, 13602–13603; c) K.
Miyamoto, S. Miyake, T. Yamamura, Nature 2001, 413, 531–534.
[17] S. Oki, A. Chiba, T. Yamamura, S. Miyake, J. Clin. Invest. 2004, 113,
1631–1640.
Received: February 10, 2010
Published online: April 21, 2010
1564
www.chemasianj.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Asian J. 2010, 5, 1560 – 1564