Quantitative microarray analysis of intact glycolipid-CD1d interaction and correlation with cell-based cytokine production

Article abstract of DOI:10.1021/ja8012787

The protein CD1d binds self and foreign glycolipids for presentation to CD1-restricted T cells by means of TCR recognition and activates T_H1 and T_H2 chemokine release. In this study, a variety of glycolipid ligands were attached to a microarray surface and their binding with dimeric CD1d was investigated. An α-galactosyl ceramide (α-GalCer) bearing a carbamate group at the 6′-OH position was tethered to the surface, and the dissociation constant on surface with CD1d was determined to reflect the multivalent interaction. Competition assays were then used to determine the dissociation constants (K_i) of new and intact glycolipids in solution. The 4-fluorophenyloctanoyl-modified α-GalCer (18) was found to bind most strongly with CD1d (K_i 0.21 μM), 2 orders of magnitude stronger than α-GalCer and more than three times more selective than α-GalCer for IFN-γ release from NKT cells. Various α-GalCer analogues were analyzed, and the results showed that the binding affinity of glycolipids to CD1d correlates well with IFN-γ production but poorly with IL-4 secretion by NKT cells, suggesting that tighter binding ligands could bias cytokine release through the T_H1 pathway.

Full text of DOI:10.1021/ja8012787

Published on Web 08/20/2008
Quantitative Microarray Analysis of Intact Glycolipid-CD1d
Interaction and Correlation with Cell-Based Cytokine
Production
Pi-Hui Liang,†,‡ Masakazu Imamura,§ Xiangming Li,| Douglass Wu,§
Masakazu Fujio,§ Richard T. Guy,§ Bing-Ching Wu,† Moriya Tsuji,*,| and
Chi-Huey Wong*,§,†
The Genomics Research Center, Academia Sinica, 128 Academia Road, Section 2, Nankang
Dist., Taipei, Taiwan 11529, Department of Chemistry, The Scripps Research Institute,
10550 North Torrey Pines Road, La Jolla, California 92037, School of Pharmacy, National
Taiwan UniVersity, Taipei, Taiwan 100, and Aaron Diamond AIDS Research Center,
455 First AVenue, The Rockefeller UniVersity, New York, New York 10016
Received February 20, 2008; E-mail: wong@scripps.edu; chwong@gate.sinica.edu.tw; mtsuji@adarc.org
Abstract: The protein CD1d binds self and foreign glycolipids for presentation to CD1-restricted T cells by
means of TCR recognition and activates TH1 and TH2 chemokine release. In this study, a variety of glycolipid
ligands were attached to a microarray surface and their binding with dimeric CD1d was investigated. An
R-galactosyl ceramide (R-GalCer) bearing a carbamate group at the 6′-OH position was tethered to the
surface, and the dissociation constant on surface with CD1d was determined to reflect the multivalent
interaction. Competition assays were then used to determine the dissociation constants (Ki) of new and
intact glycolipids in solution. The 4-fluorophenyloctanoyl-modified R-GalCer (18) was found to bind most
strongly with CD1d (Ki 0.21 µM), 2 orders of magnitude stronger than R-GalCer and more than three times
more selective than R-GalCer for IFN-γ release from NKT cells. Various R-GalCer analogues were analyzed,
and the results showed that the binding affinity of glycolipids to CD1d correlates well with IFN-γ production
but poorly with IL-4 secretion by NKT cells, suggesting that tighter binding ligands could bias cytokine
release through the TH1 pathway.
Introduction
CD1molecules are heterodimers, composed of a heavy
polypeptide chain noncovalently associated with a 2-microglo-
bulin and have substantial structural similarity to major histo-
compatibility complex (MHC) class I proteins.1 There are four
members of the CD1 family, denoted group I (CD1a, b, c, and
e) and group II (CD1d) on the basis of sequence identity and
chromosomal location.2,3 Each CD1 mediates T-cell responses
through the presentation of self and foreign lipids, glycolipids,
lipopeptides, or amphipathic small molecules to T-cell receptors
(TCR).4-6
R-GalCer (Figure 1), a glycolipid found in the marine sponge
Agelas mauritianus, is the most extensively studied ligand for
† Academia Sinica.
‡ National Taiwan University.
§ The Scripps Research Institute.
| Aaron Diamond AIDS Research Center.
(1) Zeng, Z.; Castano, A. R.; Segelke, B. W.; Stura, E. A.; Peterson, P. A.;
Wilson, I. A. Science 1997, 277, 339–345.
(2) Calabi, F.; Jarvis, J. M.; Martin, L.; Milstein, C. Eur. J. Immunol.
1989, 19, 285–292.
Figure 1. R-GC and its derivatives 1-4.
CD1d.7 When bound to CD1d, R-GalCer stimulates rapid TH1
(such as IFN-γ) and TH2 (IL-4) cytokine production by VR14i
natural killer T (VR14i NKT) cells in mice and the human
(3) Brigl, M.; Brenner, M. B. Annu. ReV. Immunol. 2004, 22, 817–890.
(4) Porcelli, S. A.; Modlin, R. L. Annu. ReV. Immunol. 1999, 17, 297–
329.
(5) Jayawardena-Wolf, J.; Bendelac, A. Curr. Opin. Immunol. 2001, 13,
109–113.
(6) Porcelli, S. A. AdV. Immunol. 1995, 59, 1–98.
(7) Tsuji, M. Cell. Mol. Life Sci. 2006, 63, 1889–1898.
9
12348 J. AM. CHEM. SOC. 2008, 130, 12348–12354
10.1021/ja8012787 CCC: $40.75
2008 American Chemical Society

Analysis of Intact Glycolipid-CD1d Interaction
A R T I C L E S
homologue VR24i NKT cells by means of TCR recognition.8-12
The production of TH1 cytokine is thought to be responsible
for the antitumor, antiviral/bacterial/parasitic, and adjuvant
effects of R-GalCer, while TH2 cytokine production is thought
to correlate with the amelioration of certain autoimmune diseases
(e.g., type 1 diabetes and multiple sclerosis).13-15 R-GalCer has
been the subject of several clinical trials for its anticancer
potential but was found to be ineffective, possibly because the
therapeutic effects of IFN-γ were hindered by IL-4.16 Therefore,
compounds which increase the selectivity of either the TH1 or
TH2 cytokine response may be more therapeutically useful.17,18
Among the factors that could cause a cytokine profile shift,
the stability of the CD1d/glycolipid complex may play a
significant role. A less stable association between the glycolipid
and CD1d, for example, could result in a shorter half-life for
NKT cell stimulation. For IFN-γ production to occur, a longer
TCR stimulation is required. IL-4 production occurs after only
2 h of stimulation, while IFN-γ production by NKT cells
requires an additional stimulation period, at least 1-2 h.19,20
Thus, improving the stability of the R-GalCer/CD1d complex
could potentially enhance the TH1 response by prolonging
stimulation of NKT cells. The first R-GalCer analogue known
to enhance the TH1 response is a stable C-glycoside analogue
(R-C-GalCer).21 Recently, we have synthesized a series of
glycolipids bearing aromatic groups on the acyl side chain and
found these molecules to skew the cytokine release profile
toward a TH1 response.22
association and dissociation of binding, may make the study
difficult. Because lipid antigens presented by CD1d molecules
can trigger and regulate a wide variety of immune responses, a
sensitive, accurate, and reproducible high-throughput assay to
probe the antigen binding properties to CD1d would be very
useful. Several previous studies have examined the lipid binding
properties of CD1d, using surface plasmon resonance (SPR),23
isoelectric focusing (IEF),24 and isothermal calorimetry (ITC).24
The SPR method suffered from a low signal-to-noise ratio and
the ITC assay required a large amount of protein for each assay.
The utility of fluorescent lipid probes in the study of ligand
binding by recombinant soluble single chain CD1 proteins has
also been evaluated,25 but this method is only sufficiently
sensitive for the study of group I CD1 proteins. One possible
reason is the binding and kinetics of association of lipid probes
are too slow for detection, because all spectra were obtained
immediately after combining the probes with the CD1 proteins.25
The fluorescent modification of the probe at the lipid tail may
also affect the interaction with CD1d.
Carbohydrate microarrays allow the rapid screening of
interactions between glycans and other molecules.26-30 Recently,
we have developed a quantitative glycan microarray method to
determine the dissociation constants of lectins/antibodies and
carbohydrate interactions on the surface at the atto-mol level.31
Here, we report a new method for the quantitative analysis of
glycolipid-receptor interactions. In this method, R-GalCer
derivatives are covalently bound to a glass slide and incubated
with CD1d, and their binding properties were examined (e.g.,
dissociation constant on the surface). Competition experiments,
in which an intact glycolipid antigen and CD1d were mixed in
solution and allowed to interact with surface R-GalCer, were
used to determine the dissociation constants of new glycolipids
in solution. As part of our ongoing search for potent CD1d
agonists, this microarray platform was used to quickly determine
the dissociation constants of intact R-GalCer derivatives bearing
different alkylphenyl chains at either the acyl or phytosphin-
gosine positions.
While the designed glycolipids were evaluated by functional
assay, examination of the binding affinity between these
glycolipids and CD1d was less easily addressed. Progress toward
an understanding of the binding properties between glycolipids
and CD1d has been slow. A major problem is the lack of a
method for measuring the binding constant of the intact
glycolipid to CD1d. In addition, the physical properties of lipids,
e.g., critical micelle concentration (CMC), solubility, or the slow
(8) Singh, N.; Hong, S.; Scherer, D. C.; Serizawa, I.; Burdin, N.;
Kronenberg, M.; Koezuka, Y.; Van Kaer, L. J. Immunol. 1999, 163,
2373–2377.
Result and Discussion
(9) Hong, S.; Wilson, M. T.; Serizawa, I.; Wu, L.; Singh, N.; Naidenko,
O. V.; Miura, T.; Haba, T.; Scherer, D. C.; Wei, J.; Kronenberg, M.;
Koezuka, Y.; Van Kaer, L. Nature Med. 2001, 7, 1052–1056.
(10) Burdin, N.; Brossay, L.; Kronenberg, M. Eur. J. Immunol. 1999, 29,
2014–2025.
Preparation of Ligands on Array Surface. In order to minimize
the amount of R-GalCer derivatives required for the microarray
while maintaining a high signal/noise ratio, a ligand with higher
affinity for CD1d was sought. X-ray crystallography studies32-35
of CD1d and glycolipid showed that the 1′-anomeric O and 2′-
(11) Kronenberg, M.; Gapin, L. Nat. ReV. Immunol. 2002, 2, 557–568.
(12) Brossay, L.; Chioda, M.; Burdin, N.; Koezuka, Y.; Casorati, G.;
Dellabona, P.; Kronenberg, M. J. Exp. Med. 1998, 188, 1521–1528.
(13) Taniguchi, M.; Harada, M.; Kojo, S.; Nakayama, T.; Wakao, H. Annu.
ReV. Immunol. 2003, 21, 483–513.
(23) Naidenko, O. V.; Maher, J. K.; Ernst, W. A.; Sakai, T.; Modlin, R. L.;
Kronenberg, M. J. Exp. Med. 1999, 190, 1069–1080.
(24) Cantu III, C.; Benlagha, K.; Savage, P. B.; Bendelac, A.; Teyton, L.
J. Immunol. 2003, 170, 4673–4682.
(14) Gonzalez-Aseguinolaza, G.; Van Kaer, L.; Bergmann, C. C.; Wilson,
J. M.; Schmieg, J.; Kronenberg, M.; Nakayama, T.; Taniguchi, M.;
Koezuka, Y.; Tsuji, M. J. Exp. Med. 2002, 195, 617–624.
(15) Chang, Y.-J.; Huang, J.-R.; Tsai, Y.-C.; Hung, J.-T.; Wu, D.; Fujio,
M.; Wong, C.-H.; Yu, A. L. Proc. Natl. Acad. Sci. U.S.A. 2007, 104,
10299–10304.
(25) Im, J. S.; Yu, K. O. A.; Illarionov, P. A.; LeClair, K. P.; Storey, J. R.;
Kennedy, M. W.; Besra, G. S.; Porcelli, S. A. J. Biol. Chem. 2004,
279, 299–310.
(16) Giaccone, G.; Punt, C. J.; Ando, Y.; Ruijter, R.; Nishi, N.; Peters,
M.; von Blomberg, B. M.; Scheper, R. J.; van der Vliet, H. J.; van
den Eertwegh, A. J.; Roelvink, M.; Beijnen, J.; Zwierzina, H.; Pinedo,
H. M. Clin. Cancer Res. 2002, 8, 3702–3709.
(26) Willats, W. G.; Rasmussen, S. E.; Kristensen, T.; Mikkelsen, J. D.;
Knox, J. P. Proteomics 2002, 2, 1666–1671.
(27) Wang, D.; Liu, S.; Trummer, B. J.; Deng, C.; Wang, A. Nat.
Biotechnol. 2002, 20, 275–281.
(17) Smyth, M. J.; Godfrey, D. I. Nat. Immunol. 2000, 1, 459–460.
(18) Berkers, C. R.; Ovaa, H. Trends Pharmacol. Sci. 2005, 26, 252–257.
(19) Schmieg, J.; Yang, G.; Franck, R. W.; Tsuji, M. J. Exp. Med. 2003,
198, 1631–1641.
(28) Liang, P.-H.; Wu, C.-Y.; Greenberg, W.; Wong, C.-H. Curr. Opin.
Chem. Biol. 2008, 12, 86–92.
(29) Feizi, T.; Fazio, F.; Chai, W.; Wong, C.-H. Curr. Opin. Struct. Biol.
2003, 13, 637–645.
(20) Yu, K. O.; Im, J. S.; Molano, A.; Dutronc, Y.; Illarionov, P. A.;
Forestier, C.; Fujiwara, N.; Arias, I.; Miyake, S.; Yamamura, T.;
Chang, Y. T.; Besra, G. S.; Porcelli, S. A. Proc. Natl. Acad. Sci. U.S.A.
2005, 102, 3383–3388.
(30) Hirabayashi, J. Trends Biotechnol. 2003, 21, 141–143.
(31) Liang, P.-H.; Wang, S.-K.; Wong, C.-H. J. Am. Chem. Soc. 2007,
129, 11177–11184.
(32) Zajonc, D. M.; Maricic, I.; Wu, D.; Halder, R.; Roy, K.; Wong, C.-
H.; Kumar, V.; Wilson, I. A. J. Exp. Med. 2005, 202, 1517–1526.
(33) Koch, M.; Stronge, V. S.; Shepherd, D.; Gadola, S. D.; Mathew, B.;
Ritter, G.; Fersht, A. R.; Besra, G. S.; Schmidt, R. R.; Jones, E. Y.;
Cerundolo, V. Nat. Immunol. 2005, 6, 819–826.
(21) Yang, G.; Schmieg, J.; Tsuji, M.; Franck, R. W. Angew. Chem., Int.
Ed. 2004, 43, 3818–3822.
(22) Fujio, M.; Wu, D.; Garcia-Navarro, R.; Ho, D. D.; Tsuji, M.; Wong,
C.-H. J. Am. Chem. Soc. 2006, 128, 9022–9023.
9
J. AM. CHEM. SOC. VOL. 130, NO. 37, 2008 12349

A R T I C L E S
Liang et al.
Scheme 1. Synthesis of Compound 1a
a Reagents and conditions: (a) TIPSCl, imidazol, THF. (b) BnBr, NaH, DMF, 0 °C to rt, 55%. (c) 7, BSP, Tf2O, DCM, -78 °C. (d) TBAF, THF, 15%
(2 steps). (e) 9, DPPA, Et3N, Tol, reflux, 77%. (f) 10% Pd/C, H2, TFA, EtOH, 95%.
equatorial OH of the galactose moiety and a ceramide moiety
bearing two lipid tails are essential for binding. Modification
at the 6′-OH position of the sugar moiety of R-GalCer
(compound 1, Figure 1) for the attachment to the glass slide
was therefore pursued. All derivatives synthesized incorporated
a linker bearing a primary amine for attachment of the derivative
to the glass slides. In the comparative studies, compounds 2
(comprising only a lipid part), 3 (modified at the acyl tail), and
4 (comprising only a galactose part), all truncated structures of
R-GalCer, were used to study the binding with CD1d. Amine 1
was synthesized from thiocresol galactose (5) as shown in
Scheme 1. First, the hydroxyl group of galactose 5 was protected
using TIPSCl and benzyl bromide to give 6, which was
glycosylated with lipid acceptor 7 in the presence of a catalytic
amount of 1-benzenesulfinylpiperidine (BSP) and Tf2O to afford
compound 8. Silyl deprotection was achieved using TBAF/THF,
and the resulting primary alcohol was coupled with an isocy-
anate intermediate (prepared from linker 9 and diphenylphos-
phoryl azide (DPPA)) to afford the carbamate 10. Global
deprotection of compound 10 was achieved by hydrogenolysis
to give compound 1. Compounds 2 and 3 were synthesized using
a similar strategy (see the Supporting Information).
Figure 2. (a) Array fabrication and detection. The glycolipid was printed
onto the slide, and its presence was detected using recombinant dimeric
CD1d protein. The spot image was generated from the precomplex of biotin-
anti-mCD1d antibody and Cy3 labeled streptavidin. (b) Images from the
array scanner, each derivative (1-4) was printed on the slide at concentra-
tions of 1000 (right-most column), 800, 600, 400, 200, 100, 80, 60, 40, 20,
10, 8, 6, 4, 2, 1, 0.8, 0.6, 0.4, 0.2 µM (left-most column) and incubated
with CD1d. The arrow indicates the limit of detection for compound 1 (1
µM). The bar indicates 1 mm in length.
Derivatives 1-4 were printed at different concentrations
(between 1 mM to 0.2 µM) on the slide coated with N-
hydroxysuccinimide (NHS)-activated ester hydrogel. The bind-
ing was visualized in a two-step procedure (Figure 2a): first,
commercially available recombinant dimeric mouse CD1d
(mCD1d) was bound to the surface molecules and then detected
using a biotin-labeled rat anti-mCD1d antibody, precomplexed
with Cy3 labeled streptavidin. Compound 1 showed the best
binding with CD1d, with a detection limit of 1 µM printing
concentration (the lowest concentration of 1 used in the printing
process); this corresponds to approximately 1 fmol of 1 per spot
(Figure 2b). As expected, modification at the 6′-OH position of
the sugar had a smaller effect on binding with CD1d than
modification at other positions.24 Compounds 2 and 3 also
interacted with CD1d with low affinity. This result shows that
the galactose moiety also participates in binding and that the
two hydrophobic tails are necessary to fit into the two
hydrophobic pockets of CD1d, a conclusion consistent with the
crystal studies.32-35 Compound 4, which lacked the lipid moiety,
(34) Zajonc, D. M.; Cantu, C, III; Mattner, J.; Zhou, D.; Savage, P. B.;
Bendelac, A.; Wilson, I. A.; Teyton, L. Nat. Immunol. 2005, 6, 810–
818.
(35) Borg, N. A.; Wun, K. S.; Kjer-Nielsen, L.; Wilce, M. C.; Pellicci,
D. G.; Koh, R.; Besra, G. S.; Bharadwaj, M.; Godfrey, D. I.;
McCluskey, J.; Rossjohn, J. Nature 2007, 488, 44–49.
9
12350 J. AM. CHEM. SOC. VOL. 130, NO. 37, 2008

Analysis of Intact Glycolipid-CD1d Interaction
A R T I C L E S
did not bind CD1d. This glycolipid-bearing, CD1d-bound array
(Figure 2a) is unique in structure, as most of the previously
reported microarrays (e.g., ganglosides microarray) tether the
lipid moiety to the solid support and expose the sugar moiety
for binding.36 The new glycolipid array was more sensitive than
the conventional ELISA for detecting the CD1d-glycolipid
interaction.36
Printing and Incubation Conditions. The optimal printing
condition was next sought. Due to the low solubility of R-GalCer
in aqueous solution, 100% DMSO was used as the solvent for
printing and found to give the best morphology of spots. After
being assayed with Cy3-streptavidin, the fluorescent microspots
were found to have an average diameter of 220 ( 30 µm (n )
50, printing concentration ) 500 µM), about 80% larger than
carbohydrates fabricated with a CMP2B pin (120 ( 20 µm, n
) 100, printing concentration ) 500 µM), suggesting that there
was a significant spreading of lipid molecules within the
microspots during the array fabrication and postfabrication
processes (Figure 2b). The spot size became smaller at lower
printing concentration. The fluorescence pattern also indicated
that the glycolipid is randomly and almost evenly distributed
within each microspot.
In the incubation step, parameters thought to affect binding
were studied: pH, detergent concentrations, DMSO concentra-
tions, immobilization time, and lipid transporter proteins.
Previous studies of the pH dependence on ligand association
with CD1d proteins have yielded conflicting results. Although
a number of studies showed that lipid binding to CD1d may
occur preferentially in acidic endocytic compartments,37,38 in
vitro studies using the surface plasmon resonance method
indicated that recombinant CD1d proteins could interact with
R-GalCer at neutral pH.23 Using compound 1, the current study
confirmed the intrinsic ability of recombinant CD1d to bind
ligands at neutral pH.
In addition, the effect of Tween 20, the most commonly used
detergent for the binding and cell-based assays, was evaluated.
Binding of CD1d with surface R-GalCer was found to be
inhibited by Tween 20 at concentrations above 0.01%. There-
fore, the assay was performed in 0.005% PBS buffer. Due to
the low water solubility of glycolipids (less than 100 µM), the
stock solutions were prepared as 1 mM (for R-GalCer) and 10
mM (other glycolipids) in DMSO solution. In the assay, DMSO
was found to inhibit the binding at concentrations greater than
2.0%. In order to mitigate the influence of DMSO when diluting
the glycolipids for assay, 1.0% DMSO was used in the assay
systems.
Figure 3. (a) Compound 1 was printed with concentrations of 1000 (left-
most column), 900, 800, 700, 600, 500, 400, 300, 200, 100, 80, 60, 40, 20,
10, 5 µM (right-most column). The images were obtained from the slides
after incubation with different concentrations of CD1d (from 800 nM to
6.25 nM as indicated above each square) followed by detection with a
precomplex of biotin-labeled anti-CD1d and Cy3-labeled streptavidin. (b)
Binding curves for compound 1 printed at different concentrations are
shown. (c) KD,surf values vs printing concentrations were obtained.
of CD1d at their preference pH (pH 5.0 for saposin A, C and
pH 6.0 for saposin B, D). No significant alteration was observed.
Determination of KD,surf. To determine the dissociation
constant on the surface, protein concentration was plotted against
fluorescence intensity at different concentrations of compound
1 (ranging from 1 mM to 1 µM) printed on the slide. Figure 3a
depicts photographs of glass slides printed at 16 different
concentrations with a 16 × 16 pattern of compound 1 from 1
mM (left-most column) to 5 µM (right-most column). The arrays
were incubated with eight concentrations of CD1d, ranging from
800 nM to 6.25 nM. CD1d concentrations were plotted against
median fluorescence intensities of replicate spots to give a set
of curves for each printed concentration (Figure 3b). Under
equilibrium conditions, the curves were analyzed as Langmuir
isotherms,
The incubation time was also varied from 1 to 6 h; it was
found that the interaction of surface R-GalCer and CD1d reached
an equilibrium after 1 h.
In ViVo, glycolipids are bound to the endosomal system with
lysosomal localization by lipid transfer proteins prior to delivery
to CD1 proteins.39 Recently, saposin B was discovered to be
the dominant saposin.40 To investigate the effect of saposin
binding on the interactions described in this study, glycolipid 1
was incubated with various saposins (A, B, C, D) in the presence
Fmax[P]
F )
(1)
[P] + KD,surf
(36) Kanter, J. L.; Narayana, S.; Ho, P. P.; Catz, I.; Warren, K. G.; Sobel,
R. A.; Steinman, L.; Robinson, W. H. Nature Med. 2006, 12, 138–
143.
Fmax is the maximum fluorescence intensity; [P] is the total
CD1d concentration; and KD,surf is the equilibrium dissociation
constant for the surface compound and CD1d. KD,surf is the
-1
(37) Spada, F. M.; Koezuka, Y.; Porcelli, S. A. J. Exp. Med. 1998, 188,
1529–1534.
(38) Gumperz, J. E.; Roy, C.; Makowska, A.; Lum, D.; Sugita, M.;
Podrebarac, T.; Koezuka, Y.; Porcelli, S. A.; Cardell, S.; Brenner,
M. B.; Behar, S. M. Immunity 2000, 12, 211–221.
(39) Kang, S. J.; Cresswell, P. Nat. Immunol 2004, 5, 175–181.
(40) Yuan, W.; Qi, X.; Tsang, P.; Kang, S. J.; Illarionov, P. A.; Besra,
G. S.; Gumperz, J.; Cresswell, P. Proc. Natl. Acad. Sci. U.S A. 2007,
104, 5551–5556.
9
J. AM. CHEM. SOC. VOL. 130, NO. 37, 2008 12351

A R T I C L E S
Liang et al.
protein concentration at which one-half of the surface binding
sites are occupied with protein. The KD, surf values are plotted
in Figure 3c as a function of printing concentration. The data
show that KD,surf () ∼100 nM at 1 mM printing concentration
and ) 500 nM at 5 µM printing concentration) decreased by a
factor of approximately 5 as the ligand density increased,
reflecting tighter binding at the higher printing concentration.
In a previous glycan microarray study,31 we found that the
fluorescence intensities became saturated when the ligand was
printed at concentrations over 100 µM. At printing concentra-
tions greater than 100 µM, the surface density is calculated to
be ∼1014 molecules/cm2, this value being close to the maximum
molecular surface density of the slide. When the surface density
is saturated, values of KD, surf and Fmax would become constant.
In this assay, the graph of KD, surf against printing concentrations
(Figure 3c) shows an inflection point at the 100 µM printing
concentration, indicating the surface saturation. However, the
KD,surf values are gradually decreased when the printing
concentration is over 100 µM, and a printing concentration of
1 mM resulted in the highest fluorescence intensity. This may
be due to the noncovalent interaction of compound 1, due to
the lipid nature of the glycolipid. The “real” surface density
cannot be directly ascertained using current technology. One
proposed method is to print a fluorescence labeled ligand directly
onto the surface and use fluorescence intensity to analyze surface
density. Such experiments are currently underway in our
laboratory.
binding sites act independently and have identical binding
properties. (3) The initial concentration of ligand ([I]) is much
greater than the initial concentration of protein so that the
concentration of unbound ligand is approximately equal to the
total concentration of ligand (i.e., [Io] ) [I]). (4) There is no
monomeric CD1d interaction for surface compound 1 because
sufficient ligand is attached to the slide. The glass slide was
incubated with solutions containing varying concentrations of
the glycolipid (covering ∼2 orders of magnitude) and the
protein-CD1d. Because of the slow association and slow
dissociation of CD1d with R-GalCer (demonstrated by IEF
assay24), the compounds were incubated with CD1d in a
microtiter plate for 5-8 h prior to being applied onto the glass
surface.
Two types of glycolipids, the acyl modified, phenyl-alkanoyl
chain analogues 11-25, and the phytosphingosine truncated
analogue 26 were examined (Table 1). R-GalCer has a Ki value
of 16 µM, similar to the values obtained from IEF (5.9 µM)
and ITC (9.7 µM).24 The value is 30-140-fold less than the
KD,surf value from the interaction of CD1d with surface R-Gal-
Cer, a reflection of the multivalent effects which take place when
1 is attached to a surface.42 This effect can be understood using
the simple model proposed by Gargano et al., which correlates
the strength of multivalent association with that of monovalent
association43 and allows the bivalent association (KA,surf) to be
expressed as
KA,surf ) F(S/10-2 n-1(KAmono n
)
)
Determination of Ki. In the previous section, the surface
dissociation constant (KD,surf), a measure of functional affinity,
was obtained by attaching compound 1 to a glass slide and
allowing molecules of CD1d to bind to it. The solution
equilibrium dissociation constant (Ki), a measure of the intrinsic
affinity of glycolipid for CD1d, can also be determined using
microarrays, this time in a competitive assay. Here, glass slides
bearing compound 1 are exposed to a solution containing both
CD1d and one of compounds 11-26. The equation that
describes the binding of the two ligands (from solution and
surface) to the same site on the protein is identical to that for
the competitive inhibition of an enzyme-catalyzed reaction.41
The final forms of the relationship can be expressed as
where n is the valency number, F is a statistical factor defined
by the system, and s is equal to 30/[inter-receptor distance (Å)].
In this system, n is 2 and the distance between two binding site
is expected to be more than 120 Å (for mouse IgG, the distance
of two Fab domain is ∼120 Å); KA, surf is calculated to be 9.7
× 106 M-1, a value similar to the experimental value of the
binding association constant, KA,surf () KD, surf -1, 1 × 107 M-1
to 2 × 106 M-1). That the calculated value of KA,surf is close to
that predicted by the equation constitutes confirmation that the
interaction between surface compound 1 and dimeric CD1d is
indeed a bivalent one (i.e., n does equal 2). This theoretical
equation for calculating KA,surf indicates that the strength of the
multivalent association increases exponentially as a function of
valency. In such a bivalent protein case, the intrinsic affinity
(KAmono) is especially important for the contribution of binding
enhancement for multivalent association.
Phenylalkanoyl analogues 12, 13, 14, and 16 were found to
bind ∼1 order of magnitude more strongly with CD1d than
R-GalCer. Compounds 18-25 were designed to have different
electronic properties by varying the substitutions at the para
positions of the phenyloctanoyl (18-21) and phenyldecanoyl
(22-25) groups. Substitution with fluorine (18) was found to
enhance the binding by approximately 1 order of magnitude,
compared with the parent compound 13 (Ki ) 1.3 µM). The
lower binding activity of the para-phenyl group (compounds
21 and 25) could be due to the lack of space in the binding
pocket. The phytosphingosine analogue (26) was found to be a
weaker agonist of mCD1d.
Fmax[Po]
F )
(2)
[Io]
[Po] + KD,surf 1 +
(
)
Ki
where [Io] is the concentration of inhibitor (the newly added
glycolipids), [Po] is the concentration of protein (CD1d) used
in the system; and Ki is the concentration of ligand at which
one-half of protein binding sites are occupied with ligand in
solution. In this assay, the concentration of the protein is kept
constant at or near the KD, surf (300-500 nM of CD1d), and the
printing concentration of compound 1 is at 100 µM where the
surface density is close to saturation and this can avoid
nonspecific binding interference of compound 1 with surface.
The following four assumptions were made: (1) Only one ligand
(compound 1) on the surface can bind to one binding site of
the protein at a time; no aggregation is assumed to take place.
This is thought to be reasonable because steric hindrance should
preclude any further interaction between a ligand and a binding
site to which another ligand is already bound. (2) All CD1d
Cytokine Releasing Profile for New Glycolipids and Their
Correlation with Array Results. The ability of these glycolipids
to activate human VRi24NKT cells was then evaluated by
(42) Mammen, M.; Choi, S.-K.; Whitesides, G. M. Angew. Chem., Int. Ed.
1998, 37, 2754–2794.
(41) Copeland, R. A. Enzymes: a practical introduction to structure,
mechanism, and data analysis. Wiley-VCH, Inc.: New York, 1996.
(43) Gargano, J. M.; Ngo, T.; Kim, J. Y.; Acheson, D. W.; Lees, W. J.
J. Am. Chem. Soc. 2001, 123, 12909–12910.
9
12352 J. AM. CHEM. SOC. VOL. 130, NO. 37, 2008

Analysis of Intact Glycolipid-CD1d Interaction
A R T I C L E S
Table 1. Ki (µM) Values of Glycolipids Obtained Using Microarray and NKT Cell-Based Cytokine Secretions by Glycolipid Stimulationa
microarray
NKT cell cytokine secretion
IL-4 secretionb
compd
structure
Ki (µM)a
IFN-γ secretionb
IFN-γ/IL-4 ratioc
R-GalCer
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
R ) ((CH2)24CH3
R ) (CH2)5Ph
16 ( 2.1
2.1 ( 0.8
1.7 ( 1.1
1.3 ( 0.4
3.1 ( 0.3
4.3 ( 1.2
1.6 ( 0.9
16 ( 3.2
0.21 ( 0.05
0.45 ( 0.15
1.2 ( 0.6
13 ( 4.8
0.27 ( 0.08
0.66 ( 0.13
4.4 ( 1.3
>100
1.00
1.00
1.00
1.12 ( 0.08
1.28 ( 0.08
1.80 ( 0.15
1.56 ( 0.11
0.86 ( 0.06
1.22 ( 0.08
0.17 ( 0.01
2.39 ( 0.13
2.25 ( 0.06
1.43 ( 0.07
0.17 ( 0.003
2.13 ( 0.003
1.91 ( 0.19
1.65 ( 0.20
0.33 ( 0.001
0.12 ( 0.006
0.73 ( 0.16
0.74 ( 0.16
0.87 ( 0.27
0.72 ( 0.19
0.59 ( 0.15
0.87 ( 0.28
0.71 ( 0.10
0.80 ( 0.09
0.79 ( 0.06
0.76 ( 0.08
0.56 ( 0.04
0.81 ( 0.13
0.87 ( 0.07
0.87 ( 0.21
0.61 ( 0.06
0.59 ( 0.15
1.53 ( 0.15
1.73 ( 0.17
2.07 ( 0.26
2.17 ( 0.24
1.46 ( 0.15
1.40 ( 0.15
0.24 ( 0.03
3.00 ( 0.16
2.85 ( 0.10
1.88 ( 0.08
0.30 ( 0.01
2.62 ( 0.18
2.19 ( 0.12
1.89 ( 0.22
0.55 ( 0.04
0.20 ( 0.02
R ) (CH2)6Ph
R ) (CH2)7Ph
R ) (CH2)8Ph
R ) (CH2)9Ph
R ) (CH2)10Ph
R ) (CH2)14Ph
R ) (CH2)7Ph(p-F)
R ) (CH2)7Ph(p-OMe)
R ) (CH2)7Ph(p-CF3)
R ) (CH2)7Ph(p-Ph)
R ) (CH2)10Ph(p-F)
R ) (CH2)10Ph(p-OMe)
R ) (CH2)10Ph(p-CF3)
R ) (CH2)10Ph(p-Ph)
-
58 ( 6.3
a All the experiments were performed in triplicate, and the data were presented as mean ( SD. b Values were obtained from the cytokine (IFN-γ and
IL-4) productions of glycolipids normalized to the cytokine production of R-GalCer. c Ratios were obtained from the IFN-γ ratio/IL-4 ratio.
comparing their proinflammatory (IFN-γ) and immunomodu-
latory (IL-4) responses relative to R-GalCer (Table 1). Gly-
colipids (11-26) induce a similar or lower level of IL-4 cytokine
secretion compared with R-GalCer. In the IFN-γ secretions,
modification of the acyl chain resulted in a higher production
compared with R-GalCer and phytospingosine modification (26).
Truncation of the phytosphingosine chain has been previously
reported to diminish IFN-γ production.44,45 Compound 13, a
phenyloctanoyl modification, had a peak value of IFN-γ
secretion among the spacer chain elongation in acyl modification
(11-17). para-Fluoro and methoxy substitutions (18, 19)
resulted in the highest IFN-γ cytokine production, more than
double that of the R-GalCer induced IFN-γ secretion. The values
of IFN-γ secretion, IL-4 secretion, and IFN-γ/IL-4 ratios were
plotted against the Ki values, and the relationships were obtained
by fitting the lines to the data points (Figure 4). The resulting
trend is interesting: compounds having a lower Ki induced higher
Figure 4. Ki values against IFN-γ secretion (blue), IL-4 secretion (pink),
IFN-γ production. Those with weaker affinities to CD1d such
and IFN-γ/IL-4 ratios (red) were plotted. The relationships were obtained
as compounds 17, 21, and 25 induced less IFN-γ production,
and this resulted in a low selectivity between IFN-γ and IL-4.
IFN-γ secretion and IFN-γ/IL-4 ratios correlated well with the
Ki values with R2 ) 0.83. This observation supports the earlier
assumption that compounds which had tighter binding with
CD1d could bias the cytokine releasing through the TH1
pathway.22 However, almost all glycolipids, regardless of their
affinity to CD1d, induce similar levels of IL-4 secretion by NKT
cells; the correlation of IL-4 secretion and the binding affinity
is poor (R2 ) 0.22). It was found that NKT cells are preactivated
and ready to secrete IL-4 through Notch-regulated conserved
by fitting the lines between them.
noncoding sequence-2 (CNS-2) enhancer activation.46 Thus,
even a weak signal through their invariant TCRs can induce a
high level of IL-4 secretion by NKT cells.
Conclusion
The interaction between a glycolipid and the protein CD1d
has been analyzed using a microarray-based method. The 6′-
OH derivative of R-GalCer was covalently bound to a glass
surface and shown to have a KD,surf of 100 to 500 nM with
recombinant dimeric mCD1d at a variety of different printing
concentrations. The solution dissociation constants (Ki) of intact
(44) Goff, R. D.; Gao, Y.; Mattner, J.; Zhou, D.; Yin, N.; Cantu, C., III;
Teyton, L.; Bendelac, A.; Savage, P. B. J. Am. Chem. Soc. 2004, 126,
13602–13603.
(45) Oki, S.; Chiba, A.; Yamamura, T.; Miyake, S. J. Clin. InVest. 2004,
113, 1631–1640.
(46) Tanaka, S.; Tsukada, J.; Suzuki, W.; Hayashi, K.; Tanigaki, K.; Tsuji,
M.; Inoue, H.; Honjo, T.; Kubo, M. Immunity 2006, 24, 689–701.
9
J. AM. CHEM. SOC. VOL. 130, NO. 37, 2008 12353

A R T I C L E S
Liang et al.
placed in a grid (9 × 9 pattern) and in total 44 subarrays replicates
(4 × 11 pattern, each subarray is approximately 5 mm (width) ×
6 mm (length) area) in one slide. After 24 h of reaction, the slides
were washed with PBST buffer (0.05% Tween 20) for 30 min and
then blocked with blocking solution (superblock blocking buffer
in PBS, Pierce) for another 1 h. The slides were dried under a flow
of argon gas and stored at room temperature in a desiccator. The
slides were washed with PBS buffer (pH 7.4) before use.
R-GalCer and its derivatives were obtained by performing
competitive binding experiments. The para-F derivative 18 was
found to bind most strongly to CD1d with a Ki of 0.21 µM, 2
orders of magnitude higher than R-GalCer, and induce the
highest IFN-γ secretion. Using various R-GalCer analogues, the
binding affinity of glycolipids to CD1d was found to correlate
well with IFN-γ production by NKT cells, but less so with IL-4
secretion, consistent with our assumption that tighter binding
compounds could bias cytokine release through the TH1
pathway. This microarray method should facilitate the analysis
of lipid binding and constitutes a relatively simple in Vitro
system for probing the physiological function of proteins and
elucidating their role in the immune response.
Direct Binding Assay. mCD1d:Ig fusion protein was diluted in
PBS buffer and then directly applied to the subarray of glass slides.
Humidifying incubation was performed under foil for 1 h at room
temperature. The slide was followed by a wash procedure. The wash
procedure is (1) washing three times with incubation buffer, (2)
washing three times with distilled water, and (3) drying under a
flow of argon gas. A solution of precomplexed biotinylated rat anti-
mCD1d antibody (20 µg/mL) and Cy3-labeled streptavidin (10 µg/
mL) was incubated on the slide for 1 h. The slide was again washed
following the above-mentioned wash procedure, and the fluores-
cence was visualized at a resolution of 5 Å m with a 595 nm laser
using an ArrayWorx microarray reader (Applied Precision).
Competitive Binding Assay. Solutions of the competitors at
different concentrations (400 to 0.1 µM) were prepared, and 5 µL
aliquots were incubated with protein (5 µL, 500 nM). The mixture
was then placed in a 384 well microtiter plate, covered, and left
for 5 to 8 h. An aliquot of this solution (8 µL) was loaded onto the
slides and incubated for 1 h under a humidifying container at room
temperature. Then, the slide was treated with the wash procedure,
and secondary antibodies were applied. To obtain the data, each
inhibitor was assayed under this system at least three times and
the average Ki values are reported.
Experimental Section
Materials. NHS-coated glass slides (Nexterion H slide, SCHOTT
North America), mCD1d:Ig fusion protein (BD Bioscience), biotin-
labeled rat anti-mCD1d IgG2b (BD Bioscience), Cy3 labeled
streptavidin (Jackson ImmunoResearch), and other standard chemi-
cals were purchased from commercial suppliers and used as
received.
Generation of Vr24+ Human NKT Cell Lines. Human NKT
cell lines, expressing the VR24+ T cell receptor (TCR), were
generated as follows: VR24+ T cells and CD14+ cells were isolated
from leukopaks using magnetic beads (Miltenyi biotec, Auburn,
CA) coupled to an anti-VR24+ TCR monoclonal antibody and an
anti-CD14 monoclonal antibody, respectively. Immature dendritic
cells were generated from the CD14+ cells after a 3-day incubation
in the presence of 300 U/mL GM-CSF (R&D systems, Minneapolis,
MN) and 100 U/mL IL-4 (R&D systems, Minneapolis, MN).
Following irradiation with 3000 rads, the immature dendritic cells
were cocultured with syngeneic VR24+ T cells in the presence of
100 ng/mL of R-galactosylceramide and 10 IU/mL of IL-2 (R&D
systems, Minneapolis, MN) for 10 days. After stimulating the
VR24+ T cells a second time with alpha-galactosylceramide-pulsed,
irradiated immature dendritic cells, NKT cell lines were shown to
express VR24+ TCR (99% purity) by a flow cytometric assay.
Cytokine Secretion Assay Using Human NKT Cell Lines.
20 000 cells of a human NKT cell line and 2 × 104 Hela cells
transfected with human CD1d that had been irradiated for 10 000
rads were cocultured in the presence of 10 ng/mL of each glycolipid
in a 96-well plate. After culture for 24 h, the concentration of IFN-γ
or IL-4 in the culture supernatants was determined by ELISA (BD
Pharmingen, San Diego, CA).
Microarray Fabrication. Microarrays were printed (Genomic
Solutions, Gene Machine) by robotic pin (SMP2B, TeleChem
International Inc.). Approximately 0.7 nL of various concentrations
of amine-containing compounds 1-4 in DMSO were deposited onto
slides from a 384 well plate. (1) The slide for the scope of printing
concentration studies: NHS-coated glass slides were printed with
compounds 1-4 at 20 different concentrations (1000, 800, 600,
400, 200, 100, 80, 60, 40, 20, 10, 8, 6, 4, 2, 1, 0.8, 0.6, 0.4, 0.2
µM) from right to left with 10 replicates vertically placed in each
subarray. Six identical subarrays were fabricated in a 1 × 6 pattern,
and each subarray consisted of a 20 × 10 pattern of spots, with a
0.3 mm pitch. After 2 h, the surface of the slide was divided by
permanent marker to avoid contamination for later protein incuba-
tion; (2) the slide for Figure 2 was printed with compound 1 with
concentrations of 1000 (left most column), 900, 800, 700, 600, 500,
400, 300, 200, 100, 80, 60, 40, 20, 10, 5 µM (right-most column,
16 different concns) with 16 replicates vertically placed in each
grid, and in total 16 replicate (8 × 2 pattern) subarrays in one slide;
(3) the slides for Ki determinations: the slides were printed of
compound 1 with a concentration of 100 µM and 81 replicates
Data Analysis
Extraction of spot intensity data was performed using
ArrayVision 8.0 (Applied Precision). The background, calculated
as the median of pixel intensities from the local area around
each spot, was subtracted from the mean pixel intensity within
each spot. The mean pixel intensity of each spot was extracted
from the circle of each spot fixed to 0.1 mm and automatically
made changes to the spot outline from 75% to 125%. To
graphically represent the data, the values of the background-
subtracted signal intensities were plotted against the known
concentration of the protein spotted in the array, using the
commercial nonlinear regression program GraphPad PRISM 4
(GraphPad, San Diego). The values obtained from the competi-
tion assay were plotted against the known concentrations of the
competitors applied on the slide. The error bars indicated in
the figures show the average percentage error for all data points
reported in the figures. Sensitivity of detection for each spot
was defined as a signal-to-noise ratio (S/N) of 2-fold above
background. S/N was calculated as S/N ) (background-
subtracted median signal intensity)/(standard deviation of
background signal intensity).
Acknowledgment. This work was supported by TSRI, Aca-
demia Sinica, and the Gates Foundation. We thank Dr. Takayuki
Shiratsuchi for providing saposin proteins. P.-H.L. is a recipient
of the NHRI postdoctoral award.
Supporting Information Available: Synthesis and spectral
information for compounds 1-3, 6-25. This information is
available free of charge via the Internet at http://pubs.acs.org/.
JA8012787
9
12354 J. AM. CHEM. SOC. VOL. 130, NO. 37, 2008