Quantitative measurements of recombinant HIV surface glycoprotein 120 binding to several glycosphingolipids expressed in planar supported lipid bilayers

Article abstract of DOI:10.1021/ja011225s

The interaction of recombinant HIV-1 surface glycoprotein gp120 (rgp120) with natural isolates of lactosylceramide (LacCer), glucosylceramide (GcCer), and galactosylceramide (GaCer) has been quantitatively measured under equilibrium conditions using total

Full text of DOI:10.1021/ja011225s

Published on Web 01/18/2002
Quantitative Measurements of Recombinant HIV Surface
Glycoprotein 120 Binding to Several Glycosphingolipids
Expressed in Planar Supported Lipid Bilayers
John C. Conboy,† Katherine D. McReynolds,‡ Jacquelyn Gervay-Hague,*,§ and
S. Scott Saavedra*
Contribution from the Department of Chemistry, UniVersity of Arizona,
Tucson, Arizona 85721-0041
Received May 17, 2001
Abstract: The interaction of recombinant HIV-1 surface glycoprotein gp120 (rgp120) with natural isolates
of lactosylceramide (LacCer), glucosylceramide (GlcCer), and galactosylceramide (GalCer) has been
quantitatively measured under equilibrium conditions using total internal reflection fluorescence (TIRF)
spectroscopy. The binding affinity (Ka) of rgp120 to these glycosphingolipids (GSLs), reconstituted at 5
mol % in supported planar lipid bilayers composed of 95 mol % POPC, is ca. 106 M-1 for dissolved rgp120
concentrations greater than 25 nM. In contrast, at concentrations of rgp120 between 0.2 and 15 nM, rgp120
does not bind significantly to LacCer and GlcCer, but has a high affinity for GalCer with a measured Ka
value of 1.6 × 109 M-1. However, protein surface coverage measurements show that this strong binding
process accounts for very little of the total protein adsorbed over the entire concentration range studied. At
a protein concentration of ca. 20 nM, the surface coverage is only 3% of that achieved at apparent saturation
(i.e., when the protein concentration is ca. 220 nM). Thus the “high affinity” binding sites comprise only a
small fraction of the total number of binding sites. Several other variables were investigated. Rgp120 binding
behavior at membranes doped with R-hydroxygalactosylceramide (R-GalCer) was very similar to that
observed with GalCer, showing that the presence/absence of an R-hydroxy moiety does not significantly
affect galactosylceramide recognition. Phase segregation of GalCer, which occurs when the mole fraction
of this GSL in a POPC bilayer exceeds ca. 0.1, was also investigated and showed no effect on binding
affinity at low rgp120 concentrations. To investigate the influence of fatty acid chain length, GSLs with
monodisperse C18 and C24 chain lengths, both with and without an R-hydroxy moiety, were synthesized,
and their binding affinity to rgp120 was examined. Relative to the natural isolates (which contain a mixture
of chain lengths), minimal differences were observed; thus among the compounds tested, fatty acid chain
length does not affect GSL recognition. The results of this work should aid efforts to design anti-HIV-1
agents based on membrane-tethered, carbohydrate-based receptors for rgp120.
and CCR5.3 T-lymphocytes and macrophages are the primary
targets of HIV infection, and the process is initiated by
Introduction
The HIV virus gains entry into host cells through a cascade
of events mediated by the viral glycoproteins gp120 and gp41.1
Gp120 is a surface protein, noncovalently bound to the
transmembrane protein gp41. Together these two glycoproteins
are presented in trimeric form on the viral surface.2 HIV is
known to infect host cells expressing both a critical domain
protein called CD4 and chemokine co-receptors such as CXCR4
interactions between gp120, CD4, and the chemokine co-
receptor. These interactions lead to conformational changes in
gp120, which subsequently unmask a hydrophobic region of
gp41 that is capable of inserting into the host cell membrane.
Trimeric gp41 is a six-helix bundle comprised of three carboxy
terminal peptides linked to three amino-terminal peptides via
loop regions.4 In the native state of gp41, the C and N peptides
are believed to be in an extended form. Once gp41 is unmasked
and the fusion domain has been inserted into the host membrane,
the C-peptides associate with the N-peptides forming a hairpin
structure.2 This event brings the host-cell and viral membranes
into close proximity allowing viral entry into host cells.
Current chemotherapeutic interventions of HIV infection are
directed toward inhibiting enzymes that are required for viral
* To whom correspondence should be addressed. S.S.S.: voice, 520-
621-9761; fax, 520-621-8407; e-mail, saavedra@u.arizona.edu.
† Current address: Department of Chemistry, University of Utah, 315
S. 1400 E., Salt Lake City, UT 84112.
‡ Current address: Department of Chemistry, California State University,
Sacramento, 6000 J. Street, Sacramento, CA 95819.
§ Current address: Department of Chemistry, University of California,
Davis, One Shields Avenue, Davis, CA 95616.
(1) (a) Allan, J. S.; Coligan, J. E.; Barin, F.; McLane, M. F.; Sodroski, J. G.;
Rosen, C. A.; Haseltine, W. A.; Lee, T. H.; Essex, M. Science 1985, 228,
1091-1094. (b) Barin, F.; McLane, M. F.; Allan, J. S.; Lee, T. H.;
Groopman, J. E.; Essex, M. Science 1985, 228, 1094-1096. (c) McKeating,
J. A.; Willey, R. L. AIDS 1989, 3, S35-S41.
(2) (a) Chan, D. C.; Kim, P. S. Cell 1998, 93, 681-84. (b) Eckert, D. M.;
Kim, P. S. Annu. ReV. Biochem. 2001, 70, 777-810.
(3) Littman, D. R. Cell 1998, 93, 677-80 and references therein.
(4) (a) Chan, D. C.; Fass, D.; Berger, J. M.; Kim, P. S. Cell 1997, 89, 263-
273. (b) Weissenhorn, W.; Dessen, A.; Harrison, S. C.; Skehel, J. J.; Wiley,
D. C. Nature 1997, 387, 426-430.
9
968 VOL. 124, NO. 6, 2002 J. AM. CHEM. SOC.
10.1021/ja011225s CCC: $22.00 © 2002 American Chemical Society

Recombinant HIV Surface Glycoprotein 120
A R T I C L E S
replication, once a host cell has been invaded. Typically a
combination of HIV reverse transcriptase and protease inhibitors
is prescribed in what is known as highly active anti-retroviral
therapy (HAART).5 However, the emergence of viral strains
that are resistant to these drugs has fueled investigations into
alternative intervention strategies. One active area of research
has focused on preventing HIV entry into host cells. Recent
developments in the production of fusion inhibitors are par-
ticularly promising. Peptide-derived compounds have been
shown to block association of the C- and N-termini of gp41
preventing formation of the hairpin structure.6
We have concentrated our efforts on an earlier step in the
entry process in an effort to block gp120 interactions with
cellular receptors, which would effectively cap gp41 and prohibit
normal fusion events. Galactosyl ceramide (GalCer (1)) is a
glycosphingolipid (GSL) expressed on mucosal membrane cells
that do not express CD4. Several studies have shown that an
alternative pathway of HIV infection occurs in these CD4 cells
through gp120 interactions with GalCer.7-11 The GalCer binding
site is distinct from the region on gp120 that recognizes CD4,
which presents the opportunity for targeted drug delivery.12 We
have proposed that synthetic compounds, capable of binding to
the GalCer site, could be used to deliver a second ligand that
blocks CD4 and chemokine co-receptor interactions. Under-
standing the molecular events that are responsible for HIV
recognition of GalCer is a critical step in developing this dual
ligand approach to preventing HIV entry into host cells.
Literature on the affinity of gp120 for GalCer and related
GSLs is surprisingly inconsistent. Harouse et al.8,9 and Bhat et
al.13 used high performance thin-layer chromatography (HPTLC)
and enzyme-linked immunosorbent assays (ELISA) to determine
the binding of several GSLs to recombinant gp120 (rgp120).
GalCer (1), glucosylceramide (GlcCer (2)), lactosylceramide
(LacCer (3)), psychosine (4)), and ceramide (Cer (5)) were
compared, and only GalCer and psychosine were bound by
rgp120 (see structures in Figure 1). Latov and co-workers used
immunospot assays on nitrocellulose and thin-layer chroma-
tography plates as well as ELISA to test the binding affinity of
Figure 1. Structures of galactosyl ceramide (GalCer, 1), glucosyl ceramide
(GlcCer, 2), lactosyl ceramide (LacCer, 3), psychosine (4), ceramide (Cer,
5), and R-hydroxy galactosyl ceramide (R-GalCer, 6).
rgp120 for GalCer.10 Only the immunospot assay on TLC
indicated that gp120 recognized GalCer. Long et al.14 studied
interactions of rgp120 and liposomes doped with mixtures of
phosphatidylethanolamine, phosphatidylcholine, cholesterol, and
various glycolipids (GalCer, GlcCer, LacCer, psychosine). They
reported that rgp120 strongly bound liposomes containing
GalCer, while GlcCer and LacCer were bound less efficiently,
and psychosine was inactive. More recently, McReynolds et al.15
reported results from a biotin-NeutrAvidin adhesion assay,
which showed that rgp120 has similar binding affinities for
GalCer, GlcCer, and LacCer.
The assays used to study gp120/GSL interactions have
inherent limitations that may explain the aforementioned
anomalies. Conventional, heterogeneous solid-phase assays (e.g.,
an ELISA or TLC method) pose problems since proper
geometric presentation of membrane-bound receptors in a
bioactive conformation is not always possible. The presentation
problem can be overcome using a liposome-flotation14 or a cell-
based immunofluorescence assay,10 but it is difficult to obtain
a complete binding curve (i.e., from very low surface coverage
to saturation) using these methods. Furthermore, none of these
methods can provide a quantitative measurement of the binding
(5) Stephenson, J. J. Am. Med. Assoc. 1999, 282, 1317-18.
(6) (a) Jiang, S.; Lin, K.; Strick, N.; Neurath, A. R. Nature 1993, 365, 113.
(b) Wild, C. T.; Shugars, D. C.; Greenwall, T. K.; McDanal, C. B.;
Mathews, T. J. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 9770-9774. (c)
Furuta, R. A.; Wild, C. T.; Weng, T.; Weiss, C. D. Nat. Struct. Biol. 1998,
5, 276-279. (d) Rimsky, L. T.; Shugars, D. C.; Mathews, T. J. J. Virol.
1998, 72, 986-993. (e) Judice, J. K.; Tom, J. Y.; Huang, W.; Wrin, T.;
Vennari, J.; Petropoulos, C. J.; McDowell, R. S. Proc. Natl. Acad. Sci.
U.S.A. 1997, 94, 13426-30.
(7) Yahi, N.; Baghdiguian, S.; Moreau, H.; Fantini, J. J. Virol. 1992, 66, 4848-
4854.
(8) Harouse, J. M.; Kunsch, C.; Hartle, H. T.; Laughlin, M. A.; Hoxie, J. A.;
Wigdahl, B.; Gonzalezscarano, F. J. Virol. 1989, 63, 2527-2533.
(9) Harouse, J. M.; Bhat, S.; Spitalnik, S. L.; Laughlin, M.; Stefano, K.;
Silberberg, D. H.; Gonzalezscarano, F. Science 1991, 253, 320-323.
(10) McAlarney, T.; Apostolski, S.; Lederman, S.; Latov, N. J. Neurosci. Res.
1994, 37, 453-460.
(11) (a) Brogi, A.; Presentini, R.; Solazzo, D.; Piomboni, P.; Constantinocec-
carini, E. AIDS Res. Hum. RetroViruses 1996, 12, 483-489. (b) Brogi, A.;
Presentini, R.; Piomboni, P.; Collodel, G.; Strazza, M.; Solazzo, D.;
Costantinoceccarini, E. J. Submicrosc. Cytol. Pathol. 1995, 27, 565-571.
(12) (a) Bhat, S.; Mettus, R. V.; Reddy, E. P.; Ugen, K. E.; Srikanthan, V.;
Williams, W. V.; Weiner, D. B. AIDS Res. Hum. RetroViruses 1993, 9,
175. (b) Fantini, J.; Hammache, D.; Dele´zay, Yahi, N.; Andre´-Barres, C.;
Rico-Lattes, O.; Lattes, A. J. Biol. Chem. 1997, 272, 7245. (c) Speck, R.
F.; Wehrly, K.; Platt, E. J.; Atchison, R. E.; Charo, I. F.; Kabat, D.;
Chesebro, B.; Goldsmith, M. A. J. Virol. 1997, 71, 7136-7139. (d)
Olshevsky, U.; Helseth, E.; Furman, C.; Li, J.; Haseltine, W.; Sodroski, J.
J. Virol. 1990, 64, 5701-5707.
(14) Long, D.; Berson, J. F.; Cook, D. G.; Doms, R. W. J. Virol. 1994, 68,
5890-5898.
(13) Bhat, S.; Spitalnik, S. L.; Gonzalezscarano, F.; Silberberg, D. H. Proc.
Natl. Acad. Sci. U.S.A. 1991, 88, 7131-7134.
(15) McReynolds, K. D.; Hadd, M. J.; Gervay-Hague, J. Bioconjugate Chem.
1999, 10, 1021.
9
J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002 969

A R T I C L E S
Conboy et al.
determined by absorbance and fluorescence spectroscopy to be 3.8 (
1.2 FITC:rgp120. FITC conjugated streptavidin (Pierce) with a labeling
ratio of 2.9:1 was used for nonspecific protein adsorption studies.
affinity at equilibrium, which makes comparisons of binding
data problematic.
To overcome these limitations, we have implemented total
internal reflection fluorescence (TIRF) spectroscopy to study
rgp120 binding to a series of GSLs. TIRF spectroscopy is a
well-established technique for the quantitative study of protein
interfacial behavior.16-22 By labeling the protein with a fluo-
rophore, protein adsorption and desorption kinetics, lateral and
rotational diffusion processes, and membrane-induced structural
changes have been investigated. Furthermore, exciting the
sample in a TIR geometry confines the incident light to the
near-surface region of the sample which significantly reduces
the contribution of bulk (dissolved) protein to the measured
signal.20 Since it is not necessary to remove the aqueous protein
solution in contact with the membrane prior to making a
measurement, receptor-ligand binding processes can be studied
under equilibrium conditions. TIRF, in combination with
methods to deposit supported phospholipid bilayers23,24 on planar
optical substrates, allows a diverse array of behaviors at model
membrane surfaces to be examined. In addition, a planar
supported lipid bilayer more closely approximates a cellular
surface as compared to a unilamellar vesicle, which has a high
radius of curvature. Consequently, TIRF microscopy at the
surface of a planar supported bilayer overcomes some of the
limitations of the previous methods used to study gp120-GSL
binding.
We recently reported the use of TIRF to measure equilibrium
binding affinities of rgp120 to natural isolates of GalCer, GlcCer,
and LacCer.25 In this paper, we extend those studies to include
R-hydroxy GalCer (6) (Figure 1). In addition to determining
protein-GSL affinity constants, a modified form of a TIRF
quantitation method has been used to determine the relative
surface coverages of rgp120 bound to GSL-bearing membranes.
Several physical parameters that may affect rgp120 binding
behavior have also been investigated, including (i) the effect
of concentration-dependent phase segregation of GSLs when
incorporated into a fluid lipid bilayer,14 and (ii) the effect of
the length of the fatty acid chain in a GSL25 on rgp120
recognition.
1-Palmitoyl-2-oleoylphosphatidylcholine (POPC) was purchased
from Avanti Polar lipids. Fluorescein isothiocyanate (FITC) labeled
phosphoethanolamine (FITC-PE) and FITC conjugated dextran (10 000
MW, 2.2:1 FITC:dextran) were purchased from Molecular Probes. The
naturally occurring GSLs (structures shown in Figure 1) lactosylcera-
mide (LacCer) (bovine), glucosylceramide (GlcCer) (Gaucher’s spleen),
and ceramide (Cer) were purchased from Matreya. Galactosylcermaide
(GalCer) (bovine, Type I) and R-hydroxy galactosylceramide (R-
GalCer) (bovine, Type II) were purchased from Sigma. The com-
mercially available GSLs were natural isolates and therefore contained
saturated and unsaturated fatty acid chains ranging from C16-C24 and
were not hydroxylated except in the case of R-GalCer.27 All com-
mercially obtained lipids were used without further purification
Preparation of Synthetic GSLs. To investigate the influence of
fatty acid chain length and R-hydroxylation on rgp120/GSL recognition,
several GSLs having monodisperse fatty acid chains were synthesized.
Two analogues of R-hydroxy GalCer, â-D-galactopyranosyl-(1 f 1)-
2-N-(2-hydroxy-stearamide)-sphingenine (R-GalCer-C18) (7a) and â-D-
galactopyranosyl-(1 f 1)-2-N-(2-hydroxy-tetracosamide)-sphingenine
(R-GalCer-C24) (7b), were prepared according to Scheme 1. Briefly,
the R-hydroxy fatty acid (8 or 9) was acetylated prior to activation
with tetrafluorophenyl trifluoroacetate. The activated acid was con-
densed with psychosine and subsequently deacetylated with sodium
methoxide to afford the desired products 7a/b. The corresponding
analogues of GalCer, â-D-galactopyranosyl-(1 f 1)-2-N-stearamide-
sphingenine (GalCer-C18) (20) and â-D-galactopyranosyl-(1 f 1)-2-
N-tetracosamide-sphingenine (GalCer-C24) (21), were prepared accord-
ing to Scheme 2. Briefly, the fatty acid (16 or 17) was first activated
with tetrafluorophenyl trifluoroacetate. The intermediate was then
condensed with psychosine to afford the desired products 20 and 21.
A detailed description of the synthesis of all four products (20, 21,
7a/b) is presented in the Supporting Information.
Preparation of Planar Membrane Substrates. Planar supported
lipid bilayers (PSLBs) used in TIRF adsorption assays were prepared
on fused silica slides (4 × 2.5 cm, Dynasil). The fused silica substrates
were first sonicated in 50% isopropyl alcohol/50% water (v/v), rinsed
in Nano-Pure water with a measured resistivity of 18 MΩ cm, treated
with 30% H2O2/70% concentrated H2SO4 for 30 min, and then rinsed
repeatedly in Nano-Pure water. PSLBs doped with GSLs used in binding
assays were deposited using the Langmuir-Blodgett-Schaefer tech-
nique.28 The bilayers were asymmetric in that GSLs were present only
in the outer lipid monolayer (i.e., the monolayer in contact with the
overlying aqueous solution). The inner monolayer of the PSLB (i.e.,
the monolayer in contact with the fused silica substrate) was composed
of pure POPC. GSLs were incorporated into the outer monolayer of
the PSLB at molar ratios of 20:1 and 1:1 POPC:GSL. Monolayers were
transferred to the substrate at a surface pressure of 35 mN/m,
corresponding to 58 Å2/molecule. Nano-Pure water was used as the
subphase. All depositions were carried out at 25 °C. After formation,
PSLBs were maintained in an aqueous environment at all times.
Experimental Section
Materials. The FITC-rgp120 (IIIB) purchased from Intracel was the
full-length, glycosylated protein obtained from the baculovirus expres-
sion system. According to the commercial supplier,26 (i) the protein
purity was >90% as estimated by Coomassie blue stained gel, (ii) the
protein was recognized by anti-gp120 MAB in ELISA and Western
Blot assays, and (iii) the protein reacted with anti-gp120 antibodies
from human serum in Dot Blot. The fluorescein labeling density was
(16) Burmeister, J. S.; Olivier, L. A.; Reichert, W. M.; Truskey, G. A.
Biomaterials 1998, 19, 307-325.
Protein Adsorption Studies. A micro-TIRF flow cell (volume of
ca. 50 µL) was constructed to minimize the amount of rgp120 needed
to perform binding studies. A schematic of the cell is shown in Figure
2. The cell consists of a Viton O-ring held in place above a quartz
slide by a Teflon block. Two 0.5 mm diameter holes on either side of
the block allow for insertion of a syringe needle to introduce solutions
into the volume above the fused silica substrate. The light coupling
arrangement of the cell consists of two right-angle quartz prisms. One
(17) Kalb, E.; Engel, J.; Tamm, L. K. Biochemistry 1990, 29, 1607-1613.
(18) Thompson, N. L.; Poglitsch, C. L.; Timbs, M. M.; Pisarchick, M. L. Acc.
Chem. Res. 1993, 26, 568-573.
(19) Pisarchick, M. L.; Thompson, N. L. Biophys. J. 1990, 58, 1235-1249.
(20) Hlady, V.; Reinecke, D. R.; Andrade, J. D. J. Colloid Interface Sci. 1986,
111, 555-569.
(21) Pisarchick, M. L.; Gesty, D.; Thompson, N. L. Biophys. J. 1992, 63, 215-
223.
(22) Timbs, M. M.; Poglitsch, C. L.; Pisarchick, M. L.; Sumner, M. T.;
Thompson, N. L. Biochim. Biophys. Acta 1991, 1064, 219-228.
(23) Tamm, L. K.; McConnell, H. M. Biophys. J. 1985, 47, 105-113.
(24) Thompson, N. L.; Palmer, A. G., III. Mol. Cell. Biophys. 1988, 5, 39-56.
(25) Conboy, J. C.; McReynolds, K. D.; Gervay-Hague, J.; Saavedra, S. S.
Angew. Chem., Int. Ed. 2000, 39, 2882-2884.
(27) Technical Bulletin #131303, Avanti Polar Lipids, Inc.
(28) McConnell, H. M.; Watts, T. H.; Weis, R. M. Biochim. Biophys. Acta 1984,
864, 95-106.
(26) (a) Technical Bulletin for Catalog #12003, Intracel Corp., 8/14/1996. (b)
Intracel Retroviral Products Catalog, 1998; p 11.
9
970 J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002

Recombinant HIV Surface Glycoprotein 120
A R T I C L E S
Scheme 1
Scheme 2
prism is used to couple the excitation light (488 nm from an argon ion
laser) into the fused silica slide. After propagating down the length of
the slide by total internal reflection, the second prism is used to
outcouple the beam, thereby reducing scattered light in the cell volume.
Index matching fluid (1.458 nd, Cargille) was used for efficient in-
and outcoupling of the laser light.
Increasing concentrations of FITC-rgp120 in 150 mM phosphate
buffered saline (50 mM phosphate) were injected into the flow cell
and allowed to equilibrate with the surface for 30 min prior to each
measurement, which was determined experimentally to be a sufficient
time for a steady-state response to be observed. The flow cell was
mounted on a Nikon Diaphot inverted microscope. Fluorescence
emission was back-collected through the quartz slide with a 4×
objective, optically filtered with a 514 nm long-pass filter, and detected
with a photomultiplier tube. The incident excitation light at 488 nm
was modulated at a frequency of 2.5 kHz. Phase sensitive detection
was used to retrieve the fluorescence intensity. The experiment was
interfaced to a PC using Lab Windows for data collection. All
experiments were performed at 25 °C.
Determination of Binding Affinities. Two different models were
used to extract the equilibrium binding constants from the fluorescence
data: (i) a conventional Langmuir model which assumes no cooper-
ativity (i.e., the affinity of each binding site is identical); (ii) a Langmuir-
like model that includes a cooperative process, in which the binding
of FITC-rgp120 to the ligand-bearing membrane promotes subsequent
binding of additional rgp120 molecules.29 The Langmuir model used
had the following form:
F
[rgp120]Ka )
(1)
Fmax - F
9
J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002 971

A R T I C L E S
Conboy et al.
λ
dp )
(4)
1/2
n2
n1
2πn1 sin2θ -
(
)
[
]
The fluorescence measured from adsorbed FITC-rgp120 at equilibrium
is the sum of the fluorescence from the adsorbed protein and the
fluorescence from the dissolved protein in the near surface region. The
total fluorescence intensity from the surface (Fs) can be expressed as
Γrgp120
∆
∞
Fs ) φsdeꢀIo
e-2z/d dz + φbdeꢀIocrgp120 e-2z/d dz (5)
p
p
∫
∫
0
∆
Figure 2. Schematic of the micro-TIRF cell and associated instrumental
components: (a) fused silica slide, (b) quartz prism, (c) Teflon block and
Viton O-ring, (d) 4× microscope objective, (e) 510 nm long pass filter, (f)
PMT, (g) lock-in amplifier, (h) frequency generator, (i) data acquisition
computer, and (j) reference photodiode.
∆
where ∆ is the thickness of the protein layer (a thickness of 93 Å was
used for the adsorbed rgp120 layer - see below), Γrgp120 is the surface
concentration of rgp120, and φs and φb are the quantum efficiencies of
FITC at the surface and in the bulk solution, respectively. At low
concentrations (less than 10-7 M), the solution contribution to the total
fluorescence (right side of eq 5) is minimal and can be ignored. Equation
5 can be used to determine the surface density of adsorbed protein if
the values of φs, de, and Io are known. In the absence of direct
knowledge of these values, the quantity of adsorbed protein can be
determined empirically by calibration of the measured fluorescence
intensity.
where Fmax is the maximum fluorescence intensity at surface saturation,
F is the fluorescence intensity measured as a function of the bulk rgp120
concentration, [rgp120], and Ka is the equilibrium binding constant.
The cooperative Langmuir model used was of the following form:
F
)
ω(F/F [rgp120]Ka )
(2)
max
Fmax - F
Calibration was performed by measuring the fluorescence from
several standard solutions of FITC-labeled dextran introduced above
the substrate surface. Prior to the injection of the standards, the surface
was first “washed” with a 1% solution of Triton X-100 to remove the
supported lipid bilayer and adsorbed FITC-rgp120. This procedure
provided a “clean” background to measure the evanescently excited
FITC-dextran fluorescence and precluded nonspecific interactions of
the dextran with the protein-coated surface. The evanescently excited
solution fluorescence (Fb) is expressed by
where ω is an additional term to account for surface bound protein-
protein interactions. A derivation of eq 2 is given in ref 29. For values
of ω < 1 the interaction is noncooperative, and for ω > 1 the interaction
is cooperative. When ω ) 1, eq 2 reverts to eq 1 (no protein-protein
interaction). Depending on the observed binding behavior, the fluo-
rescence data were fit to eq 1 using the fitting parameters of Fmax and
Ka, or to eq 2 using the fitting parameters Fmax, Ka, and ω. An iterative
solution was obtained by nonlinear regression using SigmaPlot. In both
cases, it was assumed that the activity coefficient of rgp120 is unity.
However, this assumption may not be valid if rgp120 self-association
occurs over the concentration range examined (note that gp120 self-
associates in vivo2). With respect to GSLs, the Langmuir model assumes
unity activity, but this assumption may not be valid when the GSLs
comprise a significant mole fraction of the lipids in the upper leaflet
of a PSLB. In contrast, the cooperative model allows for a distribution
of binding site activities. Such a distribution could arise, for example,
from GSL self-association (i.e., lipid rafting30) in a PSLB. Some of
these issues are discussed in more detail below.
Fluorescence Calibration. The relative surface coverage of rgp120
adsorbed to PSLBs was determined by calibration of the measured
fluorescence intensity. Since the measured fluorescence intensity is
directly proportional to the adsorption cross section and the fluorescence
quantum yield of the FITC label on rgp120, it is possible to determine
the interfacial concentration of adsorbed protein species.20,31 The
fluorescence intensity measured in TIRF geometry can be expressed
as
∞
Fb ) φbdeꢀIocdextran e-2z/d dz
(6)
p
∫
0
where the limits of integration describe the extent of the evanescent
field as it decays in the rarer aqueous medium. By taking the ratio of
the measured fluorescence from the adsorbed FITC-rgp120 (eq 5) and
the FITC-dextran solution (eq 6), the following expression is obtained:
∆
φsfdeꢀIoΓrgp120 e-2z/d dz
p
∫
Fs
0
)
(7)
∞
Fb
φblkdeꢀIo∆cdextran e-2z/d dz
p
∫
0
The assumption is made that the quantum yield of FITC conjugated to
rgp120 is identical to that of FITC conjugated to dissolved dextran,
that is, φs/φb ≈ 1. Equation 7 then reduces to the following:
∆
Γrgp120 e-2z/d dz
p
∞
I ) φd ꢀI c e-2z/d dz
(3)
∫
p
Fs
0
∫
F
e
o
0
)
(8)
∞
Fb
∆cdextran e-2z/d dz
p
∫
0
where φ is the fluorescence quantum yield, de is the detector efficiency,
ꢀ is the extinction coefficient, c is the concentration, Io is the incident
intensity, and the integral represents the decay of the evanescent field
in the rarer medium32 with a characteristic penetration depth of dp, which
is given by the following equation:
A plot of measured fluorescence intensity versus the bulk concentration
of the FITC-dextran standards yields the proportionality constant
necessary to calculate the surface coverage of rgp120.
Results and Discussion
(29) Zhao, S. L.; Walker, D. S.; Reichert, W. M. Langmuir 1993, 9, 3166-
3173.
Binding Affinity of Rgp120 to Natural GSLs. Representa-
tive adsorption isotherms of FITC-rgp120 binding to LacCer,
GlcCer, GalCer, and Cer doped at 5 mol % into the outer leaflet
of a PSLB are plotted in Figure 3 over the entire concentration
range (0.2-220 nM) of dissolved rgp120 that was studied. Also
displayed is the binding curve for FITC-rgp120 incubated with
(30) (a) Simons, K.; Ikonen, E. Functional Rafts in Cell Membranes. Nature
1997, 387, 569-572. (b) Dietrich, C.; Bagatolli, L. A.; Volovyk, Z. N.;
Thompson, N. L.; Levi, M.; Jacobson, K.; Gratton, E. Lipid Rafts
Reconstituted in Model Membranes. Biophys. J. 2001, 80, 1417-1428.
(31) (a) Rebar, V. A.; Santore, M. M. Macromolecules 1996, 29, 6262-6272.
(b) Shibata, C. T.; Lenhoff, A. M. J. Colloid Interface Sci. 1992, 148, 469.
(32) Harrick, N. J. Internal Reflection Spectroscopy; Interscience: New York,
1967.
9
972 J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002

Recombinant HIV Surface Glycoprotein 120
A R T I C L E S
Figure 3. TIRF adsorption isotherms of FITC-rgp120 adsorption to POPC
membranes doped with 5 mol % GalCer (b), GlcCer (9), LacCer (2), Cer
([), and to a pure POPC membrane (]). The data plotted represent the
average of three independent measurements for each sample. The solid lines
through the data of GlcCer and LacCer represent fits to the data using a
cooperative Langmuir isotherm. For GalCer, the sum of two independent
Langmuir isotherms was used to fit the data. The solid lines through the
data for Cer and POPC are shown as a guide to the eye only.
Figure 4. TIRF adsorption isotherms of FITC-rgp120 adsorption at low
protein concentrations (0-15 nM) to POPC membranes doped with 5 mol
% GalCer (b), GlcCer (9), LacCer (2), Cer ([), and to a pure POPC
membrane (]). The data plotted represent the average of three independent
measurements for each sample. The solid lines through the data of GlcCer
and LacCer represent fits to the data using a cooperative Langmuir isotherm.
A single Langmuir isotherm was used to fit the GalCer data. The solid
lines through the data for Cer and POPC are shown only as a guide to the
eye.
a pure POPC bilayer. The fluorescence intensities plotted in
Figure 3 are normalized relative to the intensity measured for
a standard solution of FITC-dextran, which was introduced into
the cell after measurements for each binding curve were
completed. The fluorescence intensities plotted in Figure 3 can
therefore be compared directly.
Table 1. Summary of Binding Constants and Surface Coverages
for Rgp120 Binding to Planar POPC Membranes Doped with
Various GSLs
low rgp120 concentration
high rgp120 concentration
(0.2−15 nM)
(25−220 nM)
2
2
K (M-1) ×
Γ (mol/cm ) ×
K (M-1) × Γ (mol/cm ) ×
a
a
A comparison of the binding curves shows that there is only
minimal nonspecific interaction of rgp120 with a pure POPC
bilayer. Rgp120 adsorption to a Cer-doped bilayer, which lacks
the terminal sugar group, is comparable to that observed for
rgp120 to POPC. In contrast, considerable protein adsorption
occurs at bilayers doped with LacCer, GlcCer, and GalCer. Thus
glycosylation of the lipid appears to be crucial to rgp120
recognition and binding, as has been observed by a number of
other researchers.13,14,33 An additional, negative control was
performed using FITC-streptavidin. Adsorption of this protein
to a membrane doped with 5 mol % GalCer was minor (i.e.,
comparable to streptavidin adsorption to pure POPC membrane;
data not shown). A protein lacking a carbohydrate recognition
site therefore exhibits minimal nonspecific adsorption to a GSL-
bearing PSLB.
The TIRF isotherms for binding to PSLBs bearing LacCer,
GlcCer, and GalCer are very similar in the protein concentration
range of 25-220 nM (Figure 3). However, a closer examination
of the data at concentrations of dissolved rgp120 less than 25
nM reveals differences in binding behavior among these three
GSLs. In Figure 4, the data presented in Figure 3 are replotted
over the protein concentration range of 0.2-15 nM. Rgp120
exhibits a high affinity for a GalCer-doped bilayer in this
concentration regime. In contrast, PSLBs doped with LacCer
and GlcCer show only a small, linear increase in protein
adsorption when dissolved rgp120 is less than 15 nM. Similar
to the behavior observed in the high concentration regime, there
is no significant rgp120 interaction with either pure POPC or
Cer-doped bilayers.
GSL
109
10-14 a
106
10-14 a
GalCer (1)
R-GalCer (6)
GlcCer (2)
1.6 ( 0.9
0.15 ( 0.05
7.5 ( 3.2 5.7 ( 0.9
4.6 ( 1.3 5.0 ( 1.1
3.8 ( 1.1b 3.1 ( 0.7
3.6 ( 2.1b 3.1 ( 0.8
0.40 ( 0.05
0.31 ( 0.12 0.30 ( 0.08
LacCer (3)
ceramide (5)
POPC (no GSL)
GalCer-C18 (20)
GalCer-C24 (21)
R-GalCer-C18 (7a) 0.37 ( 0.6
R-GalCer-C24 (7b) 0.28 ( 0.1
0.014 ( 0.003
0.001 ( 0.0006
0.20 ( 0.06
0.14 ( 0.02
0.10 ( 0.01
1.2 ( 0.5
1.1 ( 0.3
a Surface coverages were calculated at the apparent saturation level (Fmax
in eqs 1 and 2) in each concentration range. b For GlcCer- and LacCer-
doped membranes, Ka was recovered by fitting the isotherm data to a
cooperative model (eq 2) over the entire rgp120 concentration range studied
(0.2-220 nM).
The equilibrium constants for rgp120 binding to the various
GSLs in PSLBs were extracted by fitting the TIRF isotherms
to either a Langmuir (eq 1) or a cooperative binding (eq 2)
model. The recovered binding constants are summarized in
Table 1. A satisfactory fit to the adsorption data for LacCer-
and GlcCer-doped membranes over the entire rgp120 concentra-
tion range studied (0.2-220 nM) was achieved only when a
cooperative model was used. In these cases, we observed binding
behavior that was qualitatively very similar to that reported by
Zhao and Reichert29 for streptavidin binding to biotinylated
lipids in a PSLB. F-tests were performed to statistically compare
fitting the data to both models.34 The null hypothesis was that
the Langmuir model adequately described the data. The resulting
p-values of 1.6 × 10-6 and 2.7 × 10-4 for GlcCer and LacCer
(33) Delezay, O.; Yahi, N.; Fantini, J. Perspect. Drug DiscoVery Des. 1996, 5,
192-202.
(34) Abramowitz, M.; Stegun, L. A. Handbook of Mathematical Functions;
Dover: New York, 1965; p 299.
9
J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002 973

A R T I C L E S
Conboy et al.
provide strong evidence that the observed binding behavior is
cooperative rather than Langmuir-like.
cyanovirin-N and mannose-containing oligomers.37 Other ex-
amples, discussed in two recent reviews,38 illustrate how both
preorganized protein-protein complexes and clustered carbo-
hydrate ligands can produce protein-carbohydrate binding
affinities in the micro- and nanomolar regimes. However, at
this time the mechanism responsible for the strong rgp120-
GSL binding observed herein is not known. On the basis of the
literature in this area,14,35-38 multivalent binding is the most
probable explanation. Aggregation of GSLs into microdomains
(i.e., lipid rafts30) could promote multivalent interactions
(however, see below). Protein-protein binding is an additional
consideration, given that gp120 is presented as a trimer on the
surface of HIV.2 Initial binding of rgp120 molecules to GSLs
at the surface of a PSLB could promote additional, thermody-
namically enhanced protein adsorption due to rgp120 self-
recognition; this mechanism would explain the cooperative
binding behavior described above.
Rgp120 Surface Coverage on Planar Membranes Bearing
Natural GSLs. A modification of the TIRF calibration method
first described by Hlady et al.20 was used to determine the
surface coverage (Γ) of rgp120 adsorbed to PSLBs doped with
GSLs. Surface coverages were calculated at the apparent
saturation level in each of the two dissolved rgp120 concentra-
tion regimes studied: the low regime (0.2-15 nM) and the high
regime (25-220 nM). When expressed in fluorescence intensity
units, the apparent saturation level refers to the parameter Fmax
in eqs 1 and 2. Implementing the TIRF calibration method
allowed the recovered Fmax values to be converted to Γ values,
which are listed in Table 1.
The binding constants (Ka) recovered using the cooperative
model are 3.6 × 106 and 3.8 × 106 M-1 for LacCer and GlcCer,
respectively. The corresponding interaction constants (ω) are
2.0 and 1.8, respectively. For ω > 1 the binding process is
considered cooperative in nature, which means that the surface
free energy increases with the addition of rgp120 to the
membrane surface, thereby promoting the adsorption of ad-
ditional protein molecules. This explains the minimal protein
adsorption observed at low dissolved concentrations of rgp120
and subsequent sharp increase at higher protein concentrations.
For rgp120 adsorption to GalCer-doped membranes, the
adsorption process was fit using a sum of two independent
Langmuir isotherms (eq 1). In other words, the binding data
obtained at low dissolved rgp120 concentrations (0.2-15 nM)
and at high dissolved concentrations (25-220 nM) were treated
independently. This assumption is reasonable since (i) there is
a difference of 3 orders of magnitude between the Ka values
for two binding processes (Table 1), and (ii) the binding process
in the low concentration regime (<20 nM) undergoes apparent
saturation (Figure 4) and therefore does not contribute signifi-
cantly to the protein adsorption process at concentrations greater
than 25 nM.
In the region of high dissolved rgp120 concentration (25-
220 nM), the recovered Ka value is 7.5 × 106 M-1. Although
this value is slightly greater than that observed for LacCer and
GlcCer in the same concentration range, it does not represent a
statistically significant difference. At low concentrations of
dissolved rgp120 (<20 nM), GalCer is clearly the preferred
receptor, relative to LacCer and GlcCer, with a recovered Ka
of 1.6 × 109 M-1. Although the GalCer adsorption data are
modeled here as two discrete binding events, the possibility that
cooperative rgp120 binding occurs, as observed for LacCer and
GlcCer, cannot be eliminated. The initial, high affinity binding
of rgp120 to GalCer may serve to stimulate further protein
adsorption at higher concentrations. This explanation is con-
sistent with the similar binding behavior observed at high
concentrations of dissolved rgp120 for all three GSLs. The
strong binding event at low rgp120 concentrations is what
differentiates the activity of GalCer from LacCer and GlcCer.
The rgp120-GSL binding affinities measured here, most
notably for GalCer, are unusually high for proteins and their
carbohydrate ligands; comparable binding constants have been
reported for the interaction between gp120 and its natural protein
target, the CD4 receptor.39 However, the observation of a strong
binding event between GalCer and rgp120 is consistent with
data obtained by Long et al. using a liposome flotation assay.14
An additional comparison can be obtained by quantitative
analysis of the surface tension data reported by Hammache et
al.35 for the adsorption of rgp120 to a monolayer of R-GalCer
at the air-water interface. Using a differential form of the Gibbs
adsorption equation and assuming Langmuir binding, a binding
constant of 2.1 × 109 is obtained,36 which is comparable to the
results reported herein. Unusually strong binding between
proteins and carbohydrates has also been reported for other
ligand-receptor pairs. Bewley and Otero-Quintero measured
nanomolar binding constants for polyvalent interactions between
Before discussing these data, it is important to note that use
of the TIRF calibration method to determine the absolute surface
coverage of an adsorbed protein film can be problematic. As
discussed previously by a number of authors,20,31 there are
several contributing factors that may yield an inaccurate Γ
value: (1) Self-quenching of FITC fluorescence, due to energy
transfer, may occur when FITC-labeled protein molecules are
(36) Analysis of the surface pressure data published by Hammache et al.35 was
done using the differential form of the Gibbs adsorption equation,
1
2RT
∂π
∂ ln(argp120
Γgp120
)
(
)
)
T
where Γrgp120 is the surface excess of rgp120, π is the surface pressure,
and a is the activity of rgp120. For the dilute solutions used in the work of
Hammache, the activity has been replaced with the subphase concentration.
Assuming a Langmuir adsorption process, an integral form of the surface
pressure as a function bulk concentration can be derived,
π ) Γmax2RT ln(1 + Kacgp120
)
where Γmax is the maximum surface excess, and Ka is the affinity constant.
A satisfactory fit to the surface tension data with a calculated association
constant of 9.8 × 109 M-1 and a surface coverage of approximately
0.22 × 10-12 mol/cm2 at saturation was extracted from the data. Note:
Errors in the measured surface pressure, the choice of an appropriate
isotherm model, and the relatively few number of experimental data points
can result in errors of greater than 40% for the calculated surface densities
and binding affinities.
(37) Bewley, C. A.; Otero-Quintero, S. J. Am. Chem. Soc. 2001, 123, 3892-
3902.
(38) (a) Lee, R. T.; Lee, Y. C. Glycoconjugate J. 2001, 17, 543-551. (b)
Mammen, M.; Choi, S.-K.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998,
37, 2754-2794.
(39) An affinity constant of Ka ) 7.7 × 108 M-1 for rgp120 binding to intact,
solubilized radioiodinated CD4 was measured by saturation binding analysis
(Lundin, K.; Nygren, A.; Arthur, L. O.; Robey, W. G.; Morein, B.;
Ramstedt, U.; Gidlund, M.; Wigzell, H. J. Immunol. Methods 1987, 97,
93-100). In the case of rgp120 binding to CD4 on whole cells, a Ka
)
2.5 × 108 M-1 was measured (Smith, D. H.; Byrn, R. A.; Marsters, S. A.;
Gregory, T.; Groopman, J. E.; Capon, D. J. Science 1987, 238, 1704-
1707). Thus the strength of rgp120-GalCer association at a POPC
membrane is clearly comparable to that of the initial binding event in the
primary HIV infection pathway.
(35) Hammache, D.; Pieroni, G.; Yahi, N.; Delezay, O.; Koch, N.; Lafont, H.;
Tamalet, C.; Fantini, J. J. Biol. Chem. 1998, 273, 7967-7971.
9
974 J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002

Recombinant HIV Surface Glycoprotein 120
A R T I C L E S
closely packed in an adsorbed film. In this case, the ratio φs/φb
(in eq 7) will be significantly less than 1, and systematically
low Γ values will result. (2) In the TIRF geometry, excitation
of a fluorophore can occur via either evanescent or scattered
radiation.20,31 The ratio of evanescent- to scatter-excited emission
varies depending on the distance of the fluorophore from the
interface where TIR occurs. Since FITC-dextran calibration
solution occupies the entire evanescent volume, whereas the
FITC-rgp120 film occupies only the near-surface region, the
ratio of evanescent- to scatter-excited emission generated from
the standards and the film will be quite different. This problem
will also produce systematically low Γ values. (3) A film
thickness (∆ in eq 8) of 93 Å has been assumed for the adsorbed
protein layer. This value is based on a calculated hydrodynamic
radius of 46 Å for rgp120, which was estimated empirically
from dynamic light scattering data for globular proteins of
various size.40 However, if the shape of rgp120 is not spherical,
or if rgp120 molecules undergo conformational changes upon
adsorption, the protein layer thickness may be significantly
different from the assumed value, which may yield an inaccurate
determination of Γ.
Because of these limitations, surface coverage measurements
determined using the TIRF calibration are predicted to be
systematically underestimated.41 Thus we cannot, for example,
determine the stoichiometry of rgp120-GSL binding. However,
the results from measurements on different films can still be
quantitatively compared, since they are highly accurate on a
relatiVe basis. Thus we turn to a comparison of relative Γ values
(listed in Table 1) determined for rgp120 adsorbed to 5 mol %
GalCer, LacCer, GlcCer, Cer, and POPC PSLBs. The highest
surface coverage, 5.7 × 10-14 mol/cm2, is found for GalCer in
the high rgp120 concentration regime. On Cer-doped and pure
POPC membranes, the respective Γ values of 0.40 × 10-14 and
0.10 × 10-14 mol/cm2 are less than 10% of that measured for
GalCer. Thus the level of nonspecific protein adsorption to GSL-
bearing membranes appears to be insignificant. At LacCer- and
GlcCer-bearing membranes, equal coverages of 3.1 × 10-14
mol/cm2 were found. This Γ value is about 60% of the coverage
observed for GalCer, which is consistent with the binding
affinity measurements at high rgp120 concentrations, showing
that GalCer is the preferred receptor.
A more significant result is that the GalCer surface coverage
is 0.15 × 10-14 mol/cm2, which accounts for only a small
fraction of the surface coverage achieved at saturation in the
high concentration regime. Thus, despite the fact that the binding
of rgp120 to GalCer is very strong (Ka ) 1.6 × 109 M-1) at
protein concentrations <20 nM, this event produces a low
protein surface coverage. If we assume that two classes of
rgp120 binding sites exist on a GalCer-doped membrane, the
high affinity sites account for only ca. 3% of the total sites,
with the balance of ca. 97% being “low” affinity” sites.
Consequently, at a dissolved rgp120 concentration of ca. 220
nM, almost all of the adsorbed protein molecules are bound by
a relatively weak (and likely different) mechanism than that
which is responsible for adsorption of the strongly bound
fraction.
The physical basis for the existence of two classes of binding
sites is not clear at this time. As discussed above, one possibility
is that multivalent, high affinity binding occurs at GSL
microdomains (i.e., lipid rafts30), and lower affinity binding
occurs at regions on the PSLB where the GSLs are more
spatially dispersed. However, the experimental evidence ac-
quired to date (see below) does not definitively support (or
refute) this hypothesis. Regardless, it is clear that the two-site
behavior is what distinguishes rgp120 binding at a GalCer-
bearing membrane from rgp120 binding at membranes doped
with LacCer and GlcCer. Finally, the structural basis for
preferential recognition of GalCer is also unclear at this time,
and it will probably remain unclear until the structure of the
GSL binding site on gp120 is determined.
Effect of r-Hydroxylation of GalCer on Rgp120 Recogni-
tion. Several previous investigations have suggested that the
structure of the fatty acid chain of GalCer can modulate the
binding of this molecule to gp120. In particular, Hammache et
al.35 suggested that R-hydroxylation of the fatty acid chain is
required for gp120 recognition, since only minimal binding was
observed when the fatty acid was not R-hydroxylated. This
suggests that the hydroxy moiety is involved in gp120 recogni-
tion of GalCer. However, the TIRF adsorption data presented
in Figures 3 and 4 and in Table 1 for GalCer, which lacks the
R-hydroxy group, indicate otherwise. To further address this
issue, rgp120 binding to R-hydroxy galactosylceramide (R-
GalCer) was examined. The results are presented in Table 1.
As with GalCer, a strong binding event is observed at low
protein concentrations with an association constant of Ka )
3.1 × 108 M-1, although this value is 5-fold less than the Ka of
1.6 × 109 M-1 measured for GalCer (see Table 1). At rgp120
concentrations greater than 25 nM, the binding behavior of
GalCer and R-GalCer is quite similar. The recovered Ka and Γ
values of 4.6 × 106 M-1 and 5.0 × 10-14 mol/cm2 for R-GalCer
are only slightly lower than the corresponding values of 7.5 ×
106 M-1 and 5.7 × 10-14 mol/cm2 for GalCer, respectively
(Table 1). In summary, these data clearly show that the affinity
of rgp120 for galactosylceramides is not significantly affected
by R-hydroxylation of the fatty acid chain.
In the low rgp120 concentration regime, nonspecific adsorp-
tion also accounts for less than 10% of the protein adsorption
to a GalCer-doped membrane, based on a comparison of the Γ
values determined for GalCer-doped (0.15 × 10-14 mol/cm2),
Cer-doped (0.014 × 10-14 mol/cm2), and pure POPC (0.001 ×
10-14 mol/cm2) membranes. (Note: Surface coverages were not
determined for LacCer- and GlcCer-doped membranes, due to
the minimal binding observed at low rgp120 concentrations.)
(40) Claes, P.; Dunford, M.; Kenney, A.; Vardy, P. In Laser Light Scattering
in Biochemistry; Harding, S. E., Satelle, D. B., Bloomfield, V. A., Eds.;
Royal Society of Chemistry: Cambridge, U.K., 1992; pp 66-67.
(41) TIRF measurements on near monolayer films of albumin and streptavidin
in our laboratory consistently yield surface coverages that are 5-10% of
a monolayer. An inaccurate estimate for protein film thickness could cause
the surface coverage to be either underestimated or overestimated. However,
even a highly inaccurate estimate of film thickness (e.g., different from
the actual film thickness by 100%) would account for only a minor portion
of this error. Energy transfer quenching would yield an underestimated
surface coverage, although given typical values for hydrodynamic protein
radii and dye:protein labeling ratios, it should not be a predominate source
of error. Consequently, the contribution of scattered excited fluorescence
to the total measured signal is likely to be the predominate source of the
systematic error.
Effect of Phase Segregation on Rgp120 Binding. In addition
to the chemical structure of the carbohydrate, the binding of
gp120 to GalCer may be influenced by the extent of phase
segregation of the lipid bilayer (i.e., the formation of domains
enriched in the GSL). This hypothesis has been suggested by
Long et al.,14 based on differences observed in rgp120 binding
9
J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002 975

A R T I C L E S
Conboy et al.
it is important to note that these data and prior studies do not
eliminate the possibility that much smaller GSL domains (i.e.,
smaller than the optical diffraction limit) are present at GSL
mole fractions <0.1. If present, such microdomains could
faciliate multivalent, high affinity rgp120 binding.
Binding of Rgp120 to Synthetic GSLs. The various GSLs
used above were natural isolates and therefore contained a
mixture of various fatty acid chain lengths. By synthesizing
galactosylceramides with defined chain lengths, we hoped to
address the rgp120-GalCer binding event in more detail. We
hypothesized that a longer fatty acid chain may make the
galactosyl group on GalCer more accessible to rgp120 by
projecting the carbohydrate further from the bilayer surface. In
the natural mixtures of GalCer used in the TIRF assay, the
predominant fatty acid chain lengths were C18 and C24,27 whereas
the fatty acid chains in POPC are C16 and C18. Based on an
extended all-trans hydrocarbon chain, there is an ca. 7.5 Å
difference in the projection of the carbohydrate above the lipid
bilayer surface for the C24 versus that of the C18 fatty acid chain
length of GalCer.
Figure 5. Epifluorescence micrographs of (a) 5 mol % GalCer in POPC
monolayer and (b) 50 mol % GalCer in POPC monolayer. Films were
deposited by the Langmuir-Blodgett method on fused silica substrates.
DOPE labeled with FITC was added (at 0.1 mol %) to visualize phase
segregation in the lipid monolayer. The dark areas in (b) represent gel-
phase domains of GalCer which have phase-segregated from the liquid-
crystalline POPC matrix, in which the FITC-DOPE is preferentially solvated.
as a function of the mole fraction of GalCer in POPC liposomes.
Phase segregation of GalCer in a 1-stearoyl-2-oleoyl-sn-glycero-
3-phosphocholine (SOPC) matrix has been investigated in detail
using NMR.42-45 Above a mole fraction of 0.1, a GalCer-rich
gel phase and a liquid-crystalline SOPC-rich phase coexist,
which is primarily due to the disparity in the transition
temperatures (Tm) for GalCer and SOPC (∼80 and 6 °C,
respectively). There is a similar disparity in Tm between GalCer
and the matrix lipid POPC (Tm ≈ -2 °C) used herein and by
Long et al.14 We therefore investigated the effect of GalCer
phase segregation on rgp120 binding.
Two synthetic analogues of GalCer, GalCer-C18 (20), and
GalCer-C24 (21) were prepared according to Scheme 2. The
corresponding analogues of R-hydroxy GalCer, R-GalCer-C18
(7a), and R-GalCer-C24 (7b) were prepared according to Scheme
1. The chain lengths of these analogues, steroyl (C18) and
tetracosanoyl (C24), were chosen because they are the two
predominant chain lengths in the natural GalCer mixture.27
All four compounds were incorporated into PSLBs and
screened for their binding affinity to rgp120 in the low
concentration regime (<20 nM). The results are presented in
Table 1. The GalCer-C18 and GalCer-C24 derivatives have
binding affinities for rgp120 of 1.2 × 109 and 1.1 × 109 M-1
respectively, which are equivalent to the Ka measured for the
natural GalCer isolate (1). In addition, the saturation surface
coverages measured for these two derivatives were comparable
to those measured for the natural mixture: 0.20 × 10-14 mol/
cm2 for GalCer-C18 and 0.14 × 10-14 mol/cm2 for GalCer-C24.
The Ka values for the R-hydroxylated analogues, R-GalCer-C18
To verify the existence of phase-segregated domains for
GalCer/POPC binary mixtures, epifluorescence microscopy was
used to image Langmuir-Blodgett monolayers deposited on
fused silica slides. To visualize domain formation, a fluorescent
probe, FITC-labeled phosphatidylethanolamine (FITC-DOPE),
was introduced at 0.1 mol %. At 5 mol % GalCer/95 mol %
POPC, the film exhibits uniform fluorescence emission, indicat-
ing that the probe is uniformly distributed at optical resolution
(Figure 5a). However, at 50 mol % GalCer/50 mol % POPC,
the existence of discrete domains is clearly evident (Figure 5b).
The FITC-PE is soluble in the liquid-crystalline phase of the
lipid monolayer. The dark regions, which are approximately 5
µm in diameter and roughly circular in shape, are solidlike
domains of high GalCer concentration, while the lighter areas
correspond to fluidlike regions composed predominately of
POPC.
and R-GalCer-C24, were 0.37 × 109 and 0.28 × 109 M-1
,
respectively. These values are equivalent to the Ka measured
for the natural R-GalCer isolate (6).
Thus for the compounds examined here, the length of the
fatty acid chain of GalCer appears to have little influence on
the binding of rgp120 when the carbohydrate is presented at
the surface of a POPC PSLB. This finding is supported by NMR
studies of GalCers of various chain lengths in liposomes in
which no direct effect on the presentation or orientation of the
terminal carbohydrate group was demonstrated.46 Furthermore,
the effect of R-hydroxylation is identical to that observed when
the natural isolates GalCer and R-GalCer were compared (see
Table 1). Since the results for the synthetic analogues were
equivalent to those obtained with the natural isolates in the low
rgp120 concentration regime, binding studies at higher protein
concentrations were not pursued.
Isotherms for rgp120 binding to 5 and 50 mol % GalCer lipid
bilayers were measured in the low concentration regime (<20
nM). The binding constants recovered using a Langmuir model
were statistically equivalent (Ka ) 1.6 ( 0.9 × 109 M-1 for 5
mol % and 2.1 ( 1.3 × 109 M-1 for 50 mol %). These results
clearly show that the binding affinity of rgp120 to GalCer is
not dependent on the formation of large GalCer domains (i.e.,
optically resolvable) within the PSLB. Given the statistically
equivalent Ka values, further studies of the effect of GalCer
concentration on rgp120 binding were not pursued. However,
Conclusions
(42) Morrow, M. R.; Singh, D.; Grant, C. W. M. Biochim. Biophys. Acta 1995,
1235, 239-248.
The results presented here represent the first quantitative
measurements of the complex binding process between rgp120
(43) Morrow, M. R.; Singh, D.; Lu, D. L.; Grant, C. W. M. Biophys. J. 1993,
64, 654-664.
(44) Morrow, M. R.; Singh, D.; Lu, D.; Grant, C. W. M. Biochim. Biophys.
Acta 1992, 1106, 85-93.
(45) Lu, D. L.; Singh, D.; Morrow, M. R.; Grant, C. W. M. Biochemistry 1993,
32, 290-297.
(46) Morrow, M. R.; Singh, D. M.; Grant, C. W. M. Biophys. J. 1995, 69, 955-
964.
9
976 J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002

Recombinant HIV Surface Glycoprotein 120
A R T I C L E S
and GSL receptors at a model lipid membrane surface. The TIRF
results clearly demonstrate that a carbohydrate on ceramide is
a requirement for rgp120 recognition.
in terms of both hydroxylation and fatty acid chain length.
Furthermore, formation of gel-phase GalCer domains in fluid
lipid bilayer does not affect rgp120 binding affinity at low
protein concentrations. In closing, we emphasize that the
quantitative results presented here cannot be obtained from
ELISA and other nonequilibrium-based techniques. However,
our TIRF measurements do complement and, in some cases,
substantiate many of the results obtained in previous studies.
Finally, the results of this work should aid efforts to design
anti-HIV-1 agents based on membrane-tethered, carbohydrate-
based receptors for rgp120.
In the low rgp120 concentration regime (<20 nM), GalCer
is the preferred receptor, with an affinity constant of ca. 109
M-1, which is at the upper range of measured protein-receptor
interactions at membrane surfaces. The strength of this binding
interaction is comparable to that found for gp120 binding to
the surface cell receptor CD4.39 At higher protein concentrations
(25-220 nM), the chemical structure of the carbohydrate plays
less of a role. In this regime, rgp120 binds to PSLBs doped
with 5 mol % GalCer, LacCer, or GlcCer with an affinity
constant of ca. 106 M-1. Protein surface coverage measurements
are consistent with the binding affinity measurements and show
that the high affinity GalCer recognition event observed at low
rgp120 concentrations involves only a small fraction of the total
number of binding sites. For all three GSLs, the isotherm data
suggest that the rgp120 binding process is cooperative, possibly
due to GSL aggregation and/or protein-protein interactions at
the membrane surface.
Acknowledgment. Financial support for this research was
received from NIH (AI40359-02), NSF (CHE-9726132 and
CHE-9623583), Eli Lilly (J.G.H.), and the Alfred P. Sloan
Foundation (J.G.H.). J.C.C. was supported as a NRSA fellow
under NIH grant F32 GM19914. K.D.M. gratefully acknowl-
edges receipt of the University of Arizona Dean’s Fellowship
and the Department of Chemistry Carl S. Marvel Fellowship.
Supporting Information Available: Detailed synthetic schemes
and characterization data for 7a, 7b, 20, and 21 (PDF). This
material is available free of charge via the Internet at
http://pubs.acs.org.
The predominate selectivity mechanism for rgp120 to the
class of GSL examined here appears to be the chemical structure
of the sugar group on ceramide. There is little dependence on
the chemical composition of the fatty acid chain of ceramide,
JA011225S
9
J. AM. CHEM. SOC. VOL. 124, NO. 6, 2002 977