Phage-display selection of a human single-chain Fv antibody highly specific for melanoma and breast cancer cells using a chemoenzymatically synthesized GM3-carbohydrate antigen

Article abstract of DOI:10.1021/ja020737j

Overexpression of the cell-surface glycosphingolipid G_M3 is associated with a number of different cancers, including those of the skin, colon, breast, and lung. Antibodies against the G_M3 epitope have potential application as therape

Full text of DOI:10.1021/ja020737j

Published on Web 09/28/2002
Phage-Display Selection of a Human Single-Chain Fv
Antibody Highly Specific for Melanoma and Breast Cancer
Cells Using a Chemoenzymatically Synthesized
GM3-Carbohydrate Antigen
Kyung Joo Lee,† Shenlan Mao,† Chengzao Sun,† Changshou Gao,† Ola Blixt,‡
Sandra Arrues,§ Louis G. Hom,† Gunnar F. Kaufmann,† Timothy Z. Hoffman,†
Avery R. Coyle,† James Paulson,‡ Brunhilde Felding-Habermann,§ and
Kim D. Janda*,†
Contribution from the Department of Chemistry, The Scripps Research Institute and the Skaggs
Institute for Chemical Biology, 10550 N. Torrey Pines Road, La Jolla, California 92037, and
Department of Molecular Biology and Department of Molecular and Experimental Medicine,
The Scripps Research Institute, 10550 N. Torrey Pines Road, La Jolla, California 92037
Received May 24, 2002
Abstract: Overexpression of the cell-surface glycosphingolipid GM3 is associated with a number of different
cancers, including those of the skin, colon, breast, and lung. Antibodies against the GM3 epitope have
potential application as therapeutic agents in the treatment of these cancers. We describe the chemo-
enzymatic synthesis of two GM3-derived reagents and their use in the panning of a phage-displayed human
single-chain Fv (scFv) antibody library derived from the blood of cancer patients. Three scFv-phage clones,
GM3A6, GM3A8, and GM3A15, were selected for recombinant expression and were characterized using
BIAcore and flow cytometry. BIAcore measurements using the purified, soluble scFvs yielded dissociation
constants (Kd) ranging from 4.2 × 10-7 to 2.1 × 10-5 M. Flow cytometry was used to evaluate the ability
of each scFv to discriminate between normal human cells (human dermal fibroblast, HDFa), melanoma
cells (HMV-1, M21, and C-8161), and breast cancer cells (BCM-1, BCM-2, and BMS). GM3A6 displayed
cross-reactivity with normal cells, as well as tumor cells, and GM3A15 possessed little or no binding activity
toward any of the cell lines tested. However, GM3A8 bound to five of the six tumor cell lines and showed
no measurable reactivity against the HDFa cells. Hence, we have demonstrated that a synthetic GM3 panning
reagent can be used to isolate a fully human scFv that is highly specific for native GM3 on the surface of
tumor cells. The result is a significant step toward effective immunotherapies for the treatment of cancer.
The changes in glycosphingolipid (GSL) metabolism that
occur in cancerous cells were first noted more than 30 years
ago.1 Since that time, substantial evidence has accumulated that
now concretely associates major changes in GSL and glyco-
protein expression and composition with oncogenic transforma-
tion.2 Significantly, abnormal glycosylation and overexpression
of GSLs and glycoproteins have been found in a wide range of
cancerous tissues and have been shown to correlate with tumor
progression, metastasis, and patient survival rates.3 Although
these GSLs are found on the surface of both normal cells and
cancer cells, they are strongly immunogenic only when present
on tumors owing to the formation of GSL “microdomains” when
a threshold density level is exceeded. These clusters of GSL
molecules are important not only from the standpoint of
immunogenicity but also for the function of GSLs as adhesion
molecules and as modulators of membrane signaling pathways.4
Given the differences in GSL and glycoprotein markers between
the normal and cancerous state, the development of monoclonal
antibodies (mAbs) against carbohydrate epitopes for the treat-
ment of cancer has generated considerable interest.
The most significant problem associated with immunotherapy
is that mAbs are typically obtained from hybridomas of murine
origin. Repeated administration of these mAbs invariably results
in a human antimurine antibody (HAMA) response leading to
undesirable side effects as well as accelerated clearance of the
antibodies from the circulation.5 Unfortunately, human hybri-
domas are difficult to prepare, often unstable, and generally
secrete IgM antibody at low levels.6 Antibody-engineering
* Corresponding author. E-mail: kdjanda@scripps.edu.
† Department of Chemistry, The Scripps Research Institute and the
Skaggs Institute for Chemical Biology.
‡ Department of Molecular Biology, The Scripps Research Institute.
§ Department of Molecular and Experimental Medicine, The Scripps
Research Institute.
(1) (a) Mora, P. T.; Brady, R. O.; Bradley, R. M.; McFarland, V. W. Proc.
Natl. Acad. Sci. 1969, 63, 1290-1296. (b) Hakomori, S.; Murakami, W.
T. Proc. Natl. Acad. Sci. 1968, 59, 254-261.
(2) (a) Hounsell, E. F.; Young, M.; Davies, M. J. Clin. Sci. 1997, 93, 287-
293. (b) Kim, Y. S.; Gum, J., Jr.; Brockhausen, I. Glycoconjugate J. 1996,
13, 693-707. (c) Bhavanandan, V. P. Glycobiology 1991, 1, 493-503.
(d) Hakomori, S. AdV. Cancer Res. 1989, 52, 257-331. (e) Hakomori, S.
Sci. Am. 1986, 254, 44-53.
(3) (a) Hakomori, S. Cancer Res. 1996, 56, 5309-5318. (b) Muramatsu, T.
Glycobiology 1993, 3, 291-296.
(4) Hakomori, S.; Zhang, Y. Chem. Biol. 1997, 4, 97-104.
9
10.1021/ja020737j CCC: $22.00 © 2002 American Chemical Society
J. AM. CHEM. SOC. 2002, 124, 12439-12446
12439

A R T I C L E S
Lee et al.
Chart 1. Structures of Compounds under Discussion
strategies aimed at humanization of murine mAbs, such as
complementarity-determining region (CDR) grafting,7 often do
not lead to satisfactory clinical results. An alternative solution
is to generate completely human antibodies using phage-display
technology.8 It is possible, for example, to create large antibody
libraries from the peripheral blood lymphocytes of cancer
patients in order to screen for antibodies specific for particular
cancer antigens. We previously reported using this approach to
identify single-chain Fv (scFv) antibody fragments specific for
the carbohydrate antigens Lewisx and sialyl Lewisx.9
Overexpression of the ganglioside GM3 is associated with
melanoma and a number of other cancer types, such as colon,
breast, and lung.10 It was previously demonstrated that the
murine mAb DH2 recognized GM3 and inhibited the growth of
melanoma cells in vitro and in vivo.11 In addition, human IgM
mAbs having various specificities for GD2, GD3, GM2, and GM3
were isolated.12 However, for the reasons outlined above, the
use of these antibodies in a clinical setting is extremely limited.
Herein, we describe the chemoenzymatic synthesis of GM3
-
carbohydrate haptens that contain a linker moiety for attachment
to a carrier protein and the use of these synthetic panning
reagents to select human scFvs specific for GM3 from a phage-
display library. BIAcore was used to measure the antibody
affinities for the GM3 hapten itself, and flow cytometry was
performed to evaluate scFv recognition of the intact antigen on
the surface of melanoma and breast cancer cells. Significantly,
our strategy allowed the isolation and characterization of a
human scFv highly specific for three separate melanoma cell
lines, HMV-1, M21, and C-8161, the latter of which is highly
metastatic, and two breast cancer cell lines, BCM-1 and BCM-
2, derived from stage IV breast cancer patients. Finally, the
results suggested that GSL clustering may not be necessary to
obtain anti-GSL antibodies that bind specifically to tumor cells.
protein for aspects of recognition, although the effect is not yet
well understood.13 Thus, 2 incorporates the natural O-linked
carbohydrate fragment that occurs in GM3 1, while 3 utilizes an
acetylated N-glycosyl linkage that is less congruent to GM3, but
somewhat more stable. The synthesis of a similar N-linked GM3
hapten was recently reported.14 Both linkers provide a tether
that contains hydrogen-bonding functionalities, are much more
hydrophilic than the fatty ceramide portion of GM3, are ap-
proximately 10-12 Å in length, which is generally optimum
for hapten recognition, and are terminated with a maleimide
functionality. Since the haptens possess the distal sialic acid
residue, using a linker that terminates in a carboxylic acid would
not allow fully site-specific coupling to amine-containing
partners. Hence, a maleimide moiety was chosen that readily
reacts with either cysteine residues on proteins for the construc-
tion of panning reagents or sulfur functional groups on BIAcore
sensor chips for the measurement of kinetic parameters.
The synthesis of GM3-A was accomplished as shown (Scheme
1). Lactosyl bromide 6 was prepared from D-lactose as described
in the literature.15 Displacement of bromide from 6 with
2-azidoethanol in the presence of silver trifluoromethane-
sulfonate followed by deprotection of all acetyl groups gave
the key intermediate 7. In the enzymatic sialylation of 7, CMP-
sialic acid 9 was used as a sialyl donor and the reaction
proceeded smoothly to completion at room temperature in the
presence of excess R-2,3-sialyltransferase (R-2,3-SiaT) to
provide 8.16 The final compound was prepared in two consecu-
Results and Discussion
Two GM3 haptens, 2 and 3, which differ from each other only
in the appended linker group, were designed (Chart 1). We
examined different linkers on the basis of our previous work,
particularly in the elicitation of immune responses, that showed
the importance of the linker structure between hapten and carrier
(5) (a) Sakahara, H.; Saga, T.; Onodera, H.; Yao, Z.; Nakamoto, Y.; Zhang,
M.; Sato, N.; Nakada, H.; Yamashina, I.; Endo, K. Jpn. J. Cancer Res.
(GANN) 1997, 88, 895-899. (b) Dillman, R. O.; Beauregard, J. C.; Halpern,
S. E.; Clutter, M. J. Biol. Response Mod. 1986, 5, 73-84. (c) Schroff, R.
W.; Foon, K. A.; Beatty, S. M.; Olkham, R. K.; Morgan, A. C., Jr. Cancer
Res. 1985, 45, 879-885. (d) Shawler, D. L.; Bartholomew, R. M.; Smith,
L. M.; Dillman, R. O. J. Immunol. 1985, 135, 1530-1535.
(6) James, K. In Handbook of Experimental Pharmacology: The Pharmacology
of Monoclonal Antibodies; Rosenberg, M., Moore, G. P., Eds.; Springer-
Verlag: New York, 1994; Vol. 113, pp 3-19.
(7) (a) Daugherty, B. L.; DeMartino, J. A.; Law, M. F.; Kawka, D. W.; Singer,
I. I.; Mark, G. E. Nucleic Acids Res. 1991, 19, 2471-2476. (b) Riechmann,
L.; Clark, M.; Waldmann, H.; Winter, G. Nature 1988, 332, 323-327.
(8) (a) Graus, Y. F.; Verschurren, J. J.; Degenhardt, A.; van Breda Vriesman,
P. J.; De Baets, M. H.; Posner, J. B.; Burton, D. R.; Dalmau, J. J.
Neuroimmunol. 1998, 82, 200-209. (b) Henderikx, P.; Kandilogiannaki,
M.; Petrarca, C.; von Mensdorff-Pouilly, S.; Hilgers, J. H.; Krambovitis,
E.; Arends, J. W.; Hoogenboom, H. R. Cancer Res. 1998, 58, 4324-4332.
(9) Mao, S.; Gao, C.; Lo, L. C.-H.; Wirsching, P.; Wong, C.-H.; Janda, K. D.
Proc. Natl. Acad. Sci. USA 1999, 96, 6953-6958.
(10) Hoon, D. S.; Wang, Y.; Sze, L.; Kanda, H.; Watanabe, T.; Morrison, S.
L.; Morton, D. L.; Irie, R. F. Cancer Res. 1993, 53, 5244-5250.
(11) Dohi, T.; Nores, G.; Hakomori, S. Cancer Res. 1988, 48, 5680-5685.
(12) (a) Nishinaka, Y.; Ravindranath, M. H.; Irie, R. F. Cancer Res. 1996, 56,
5666-5671. (b) Lloyd, K. O. Ann. N. Y. Acad. Sci. 1993, 690, 50-58. (c)
Yamaguchi, H.; Furukawa, K.; Fortunato, S. R.; Livingston, P. O.; Lloyd,
K. O.; Oettgen, H. F.; Old, L. J. Proc. Natl. Acad. Sci. USA 1990, 87,
3333-3337. (d) Furukawa, K.; Yamaguchi, H.; Oettgen, H. F.; Old, L. J.;
Lloyd, K. O. Cancer Res. 1989, 49, 191-196. (e) Tai, T.; Paulson, J. C.;
Cahan, L. D.; Irie, R. F. Proc. Natl. Acad. Sci. USA 1983, 80, 5392-5396.
(13) (a) Isomura, S.; Ashley, J. A.; Wirsching, P.; Janda, K. D. Bioorg. Med.
Chem. Lett. 2002, 12, 861-864. (b) Matsushita, M.; Hoffman, T. Z.; Ashley,
J. A.; Zhou, B.; Wirsching, P.; Janda, K. D. Bioorg. Med. Chem. Lett.
2001, 11, 87-90.
(14) Xue, J.; Pan, Y.; Guo, Z. Tetrahedron Lett. 2002, 43, 1599-1602.
(15) Dahmen, J.; Frejd, T.; Gronberg, G.; Lave, T.; Magnusson, G.; Noori, G.
Carbohydr. Res. 1983, 116, 303-307.
9
12440 J. AM. CHEM. SOC. VOL. 124, NO. 42, 2002

Phage-Display Selection of a Human Fv Antibody
A R T I C L E S
Scheme 1. Synthesis of the GM3-A Haptena
Reagents and conditions: (a) (i) 2-azidoethanol, AgOTf, CH2Cl2, -78 °C to room temperature, 24 h, (ii) NaOMe, MeOH (70% yield for two steps); (b)
9, R-2,3-SiaT, cacodylate buffer, pH 6.5, rt, 1 d (77% yield); (c) (i) H2, 10% Pd/C, MeOH/H2O (50/50), rt, 2 h, (ii) 10, 0.1 M NaHCO3/MeOH (50/50), rt,
1 h (42% yield for two steps).
tive steps. First, the terminal azido functionality was reduced
by catalytic hydrogenation in methanol-water, then the resulting
amine was coupled directly with the activated ester of male-
imidobutyric acid in slightly basic solution, affording 2.
The enzyme R-2,3-SiaT has been used previously for oligo-
saccharide synthesis17 and generally transfers N-Ac-neuraminic
acid (sialic acid) to the 3-position of terminal galactose or N-Ac-
galactosamine residues. Several recent advances, such as the
cloning of glycosyltransferases, development of a cofactor
regeneration system, and the large-scale syntheses of nucleotide
sugars have made the enzymatic approach especially attractive
for the chemical glycosylation of certain problematic linkages.
In fact, R-2,3-SiaT has been used in the synthesis of natural
accharide. After deprotection of all acetyl groups, the intermedi-
ate 13 was obtained in good overall yield. Since N-lactosyl
derivatives have not often been used in enzymatic sialylation,20
we were pleased to observe that the sialyl group was successfully
transferred from 9 to 13, yielding compound 14 nearly
quantitatively. The terminal benzyloxycarbonyl group was
removed by catalytic hydrogenation and the resulting amine was
coupled with activated ester 16 to provide hapten 3.
We chose bovine serum albumin (BSA) as the protein for
conjugation to form the panning reagents. To this end, we first
treated BSA with Traut’s reagent, which modifies lysine residues
through an imidate ester bond to afford terminal thiols. The
modified BSA was then directly coupled with 2 or 3 in
phosphate buffer, pH 7.4, at room temperature and dialyzed to
give 4 and 5, respectively (Chart 1). The average numbers of
carbohydrate epitopes attached to the modified BSA (17.0 for
2 and 10.2 for 3) were determined by MALDI-TOF MS analysis.
The GM3-A-BSA and GM3-B-BSA conjugates were im-
mobilized on a solid support and used to pan a previously
constructed scFv-phage library derived from the blood samples
of 20 human subjects diagnosed with a variety of cancers.9 Over
the course of four rounds of panning, phage titers increased
from 1.8 × 105 to 1.4 × 109 (7800-fold enrichment) using 4
and from 5.9 × 105 to 1.6 × 108 (270-fold enrichment) using
5. Eleven clones from the panning with 4 and 10 clones from
using 5 were selected for sequence analysis (Tables 1 and 2).
Several distinct groups of related antibodies emerged, and
representative clones from each of these groups were chosen
for BIAcore analysis. From 4, one group comprised clones
GM3A2, GM3A3, GM3A6, and GM3A11. A distinguishing
feature arising from sequence analysis of these clones was that
CDR L2 contained nine amino acids, rather than the more typical
GM3.
18 The advantages of using enzymatic glycosylation are that
elaborate protection-deprotection steps are not required and the
carbohydrates are obtained with only the correct, naturally
occurring stereochemistry.
The synthesis of GM3-B also proceeded smoothly (Scheme
2). Compound 12 was prepared from R-D-lactose 11 in neat
allylamine at room temperature, a manner similar to the known
procedure.19 Ozonolysis of the terminal alkene at low temper-
ature followed by reductive workup gave the aldehyde, which
was directly coupled with secondary amine 15 in the presence
of sodium cyanoborohydride, affording the peracetylated dis-
(16) Blixt, O.; Brown, J.; Schur, M. J.; Wakarchuk, W.; Paulson, J. C. J. Org.
Chem. 2001, 66, 2442-2448.
(17) (a) Wittmann, V.; Takayama, S.; Gong, K. W.; Weitz-Schmodt, G.; Wong,
C.-H. J. Org. Chem. 1998, 63, 5137-5143. (b) Seitz, O.; Wong, C.-H. J.
Am. Chem. Soc. 1997, 119, 8766-8776. (c) Wong, C.-H.; Halcomb, R.
L.; Ichikawa, Y.; Kajimoto, T. Angew. Chem. Int. Ed. 1995, 34, 521-546.
(d) Ichikawa, Y.; Lin, Y.-C.; Dumas, D. P.; Shen, G.-J.; Wong, C.-H. et
al. J. Am. Chem. Soc. 1992, 114, 9283-9298. (e) Ichikawa, Y.; Shen, G.-
J.; Wong, C.-H. J. Am. Chem. Soc. 1991, 113, 4698-4700.
(18) (a) Liu, K.-C.; Danishefsky, S. J. Chem. Eur. J. 1996, 2, 1359-1362. (b)
Liu, K. K.-C.; Danishefsky, S. J. J. Am. Chem. Soc. 1993, 115, 4933-
4934. (c) Ito, Y.; Paulson, J. C. J. Am. Chem. Soc. 1993, 115, 1603-1605.
(19) (a) Ramon, D.; Rollin, P.; Klaffke, W. Angew. Chem. Int. Ed. 2000, 39,
396-398. (b) Spevak, W.; Dasgupta, F.; Hobbs, C. J.; Nagy, J. O. J. Org.
Chem. 1996, 61, 3417-3422.
(20) Wittmann, V.; Datta, A. K.; Koeller, K. M.; Wong, C.-H. Chem. Eur. J.
2000, 6, 162-171.
9
J. AM. CHEM. SOC. VOL. 124, NO. 42, 2002 12441

A R T I C L E S
Lee et al.
Scheme 2. Synthesis of the GM3-B Haptena
Reagents and conditions: (a) (i) allylamine, rt, 70 h, (ii) Ac2O, pyridine, 12 h (53% yield for two steps); (b) (i) O3, MeOH, -78 °C to room temperature,
then Me2S, (ii) 15, NaCNBH3, AcOH (0.5 equiv), MeOH, rt, 24 h, (iii) NaOMe, MeOH (35% yield for three steps); (c) 9, R-2,3-SiaT, cacodylate buffer,
pH 6.5, rt, 1 d (65% yield); (d) (i) H2, 10% Pd/C, MeOH/H2O (50/50), rt, 2 h, (ii) 16, 0.1 M NaHCO3/MeOH (50/50), rt, 1 h (45% yield for two steps).
Table 1. Aligned Heavy-Chain (H) Amino Acid Sequences of Selected ScFv Clones
clone
H family
CDR H1
CDR H2
CDR H3
GM3A2
GM3A3
GM3A6
GM3A11
GM3A1
GM3A4
GM3A9
GM3A15
GM3A8
GM3A13
GM3A5
GM3B
IGHV1
IGHV1
IGHV1
IGHV1
IGHV4
IGHV4
IGHV4
IGHV4
IGHV4
IGHV1
IGHV3
IGHV3
GYTFRNHAIS
GYTFRNHAIS
GYTFRNHAIS
GYTFRNHAIS
GASITSYYWN
GMSITSSYWN
GMSITSSYWN
GMSITSSYWN
GGSISSSYWN
GYTFTSYDIN
GFSFSSYAMT
GFSFSSYAMT
WISGYNGNTNYAQRFQG
WISGYNGNTNYAQRFQG
WISGYNGNTNYAQRFQG
WISGYNGNTNYAQRFQG
YIHHRGASNYNPSLKS
YVHYNGNTLSNPSLQS
YVHYNGNTLSNPSLQS
YVHYNGNTLSNPSLQS
YVHYTGSTHYNPSLQS
WMNPNSGNTGYAQKFQG
TITARGDRTYNADSVKG
TITARGDRTYNADSVKG
DLMNWGRFPLDY
DLMNWGRFPLDY
DLMNWGRFPLDY
DLMNWGRFPLDY
TGTSSDIGRYNYVS
WNGQNNAFDT
WNGQNNAFDT
WNGQNNAFDT
WNGVNNAFDT
ARYSGSYRPLDYYMDV
DSSYVDWPPYGLHV
DSSYVDWPPYGLHV
Table 2. Aligned Light-Chain (L) Amino Acid Sequences of Selected ScFv Clones
clone
L family
CDR L1
CDR L2
CDR L3
GM3A2
GM3A3
GM3A6
GM3A11
GM3A1
GM3A4
GM3A9
GM3A15
GM3A8
GM3A13
GM3A5
GM3B
IGLV5
IGLV5
IGLV5
IGLV5
IGLV2
IGLV2
IGLV2
IGLV2
IGKV1
IGKV1
IGLV2
IGLV2
TLRSGIDVGTYRI
TLRSGIDVGTYRI
TLRSGIDVGTYRI
TLRSGIDVGTYRI
TGTSSDIGRYNYVS
SVISSDVGANKRVS
SVISSDVGANKRVS
SVISSDVGANKRVS
RASQSISSYLN
KSDSDQKQG
KSDSDQKQG
KSDSDQKQG
KSDSDQKQG
EVDKRPS
EVNKRRS
EVNKRRS
EVNKRRS
AASRLRS
MIWHSSAWV
MIWHSSAWV
MIWHSSAWV
MIWHSSAWV
SSFGAGKV
SSYTRTELL
SSYTRTELL
SSYTRTELL
QQSYSTPVT
QQSYSTPLT
SSYTGTLLL
SSYTGTLLL
RASQRIDSFLN
TGPSSDVGNYDYVS
TGPSSDVGNYDYVS
AASSLQS
DVSNRPS
DVSNRPS
seven. This structure has been observed previously but is
considered rare.21
for GM3 (Kd ) 21 µM). Another clone, GM3A8, contained a
heavy-chain sequence very similar to clones GM3A4, GM3A9,
and GM3A15 but a light chain that more closely resembled
clone GM3A13. On the basis of several factors such as
expression level, enzyme-linked immunosorbent assay (ELISA)
using GM3-A-BSA (data not shown), and BIAcore (Kd ) 1.2
µM), GM3A8 was considered, at this point, the most desirable
scFv.
The soluble scFv GM3A6 was expressed in Escherichia coli,
purified, and found to possess a dissociation constant (Kd) of
0.42 µM determined using BIAcore (Table 3). The second major
sequence group included the identical clones GM3A4, GM3A9,
and GM3A15 and the related clone GM3A1 [85% homology
at the amino acid level, excluding the (Gly4Ser)3 linker].
BIAcore analysis of scFv GM3A15 revealed a moderate affinity
Interestingly, all 10 clones derived from conjugate 5 consisted
of the same DNA sequence, and this sequence was also found
in clone GM3A5. This suggested that, although linker differ-
(21) Kasaian, M. T.; Ikematsu, H.; Balow, J. E.; Casali, P. J. Immunol. 1994,
152, 3137-3151.
9
12442 J. AM. CHEM. SOC. VOL. 124, NO. 42, 2002

Phage-Display Selection of a Human Fv Antibody
A R T I C L E S
Table 3. Kinetic and Thermodynamic Parameters for Binding of
a
scFvs to GM3
scFv
k
on (103 M-1 s-1
)
koff (104 s-1
)
K (106 M)
d
GM3A6
GM3A8
GM3A15
2.1
0.44
4.9
8.9
5.3
1000
0.42
1.2
21
a The kinetic constants were measured by BIAcore using GM3-A
covalently attached to the BIAcore biosensor. The Kd was calculated (Kd
) koff/kon) from the experimentally determined kon (association rate constant)
and koff (dissociation rate constant) values.
ences between GM3-A and GM3-B afforded antibody populations
that were on the whole quite distinct, some crossover in hapten
recognition did occur in the panning process. Due to the
comparatively low level of enrichment, suggestive of low
affinity, and a lack of specificity by ELISA, no clones obtained
using 5 were chosen for recombinant expression or BIAcore
analysis. Notably, the results reflected many of our previous
findings that linker structure can be critical during recognition
processes, whether in vivo during an immune response or in
vitro for panning procedures, suggesting that variations in the
linkage between hapten and carrier should be evaluated for such
experiments.13
To determine whether the scFvs could recognize the native
GM3 epitope on the surface of tumor cells, the purified scFvs
were used to stain three melanoma and three breast cancer cell
lines, as well as normal dermal fibroblast (HDFa) cells as a
nonmalignant cell control. The melanoma cell line HMV-1 has
been characterized as poorly metastatic,22 M21 is moderately
metastatic,23 and C-8161 is regarded as a highly metastatic cell
line.24 The three breast cancer cell lines, BCM-1, BCM-2, and
BMS, were derived from circulating metastatic cells in the
peripheral blood of stage IV breast cancer patients. Binding of
anti-GM3 scFvs GM3A6, GM3A8, and GM3A15 to these cell
lines was assessed using flow cytometry (Figure 1). It was found
that clone GM3A15 displayed little or no binding activity toward
any cells relative to antibody-free buffer controls. GM3A6
displayed low to moderate affinity toward four of the six tumor
cell lines, but also appeared to bind to HDFa cells more strongly
than the other two scFvs. Antibody cross-reactivity with normal
cells is an unfavorable property for clinical applications of
cancer immunotherapy. In contrast, scFv GM3A8 showed
excellent specificity for tumor cells versus normal cells. Five
of the six tumor cell lines were strongly stained by this scFv
and, significantly, no measurable binding activity toward HDFa
cells was detected. Hence, scFv GM3A8 is a candidate for future
studies directed at killing tumor cells using a scFv-drug
conjugate or by elaboration into a whole IgG format for
complement-mediated and antibody-dependent cytotoxic path-
ways.
Figure 1. Analysis of the binding of anti-GM3 scFvs to various cell lines
using flow cytometry: (A) normal HDFa cells; (B) melanoma, HMV-1;
(C) melanoma, M21; (D) melanoma, C-8161; (E) breast cancer, BMS; (F)
breast cancer, BCM-1; (G) breast cancer, BCM-2. Control ) PBS.
plasma membrane and not exposed for antibody binding,
ensuring that the carbohydrate portion is presented on the cell
surface. A hapten linker was designed that recapitulated this
effect by appropriately attaching and presenting the GM3
trisaccharide on the surface of the carrier protein. It is likely
that the hydrophilic nature of the hapten-linker construct of 4
precludes sequestration within the relatively hydrophobic protein
matrix and therefore allows extensive access to the haptenic
structure. Clearly, there was a difference between 4 and 5,
suggesting perhaps that the N-acetyl group and/or positively
charged N-methyl group in the linker of 5 associate with the
protein to cause some shielding of carbohydrate determinants.
This might have resulted in the inability to isolate clones of
adequate affinity using 5 as noted above.
Recent studies of carbohydrate anticancer vaccine designs
indicated that multivalent display of the carbohydrate antigen
Lewisy in synthetic haptens was required for eliciting high
antibody titers.25 The antibodies produced in response to these
vaccines were predominantly of the IgM class due to the
chemical nature of the antigen and its reliance upon T cell-
independent stimulation of B-cell activity. Elicitation of in vivo
antibody responses against GM3 has also been particularly
problematic. A biantennary hapten was recently used to obtain
reasonable titers consisting of IgM and IgG antibodies.26
The results also demonstrated that a BSA carrier protein
displaying the carbohydrate portion of a tumor-associated GSL
could be used to isolate antibodies that recognized the intact,
native GSL antigen on the cell surface. In the hapten design, it
was not necessary to model the long, hydrophobic ceramide
tails of GM3, since these are generally concealed within the
(22) Katagiri, Y.; Hayashi, Y.; Baba, I.; Suzuki, H.; Tanoue, K.; Yamazaki, H.
Cancer Res. 1991, 51, 1286-1293.
(23) Schulz, G.; Staffileno, L. K.; Reisfeld, R. A.; Dennert, G. J. Exp. Med.
1985, 161, 1315-1325.
(25) Kudryashov, V.; Glunz, P. W.; Williams, L. J.; Hintermann, S.; Danishefsky,
S. J.; Lloyd, K. O. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 3264-3269.
(26) Zou, W.; Abraham, M.; Gilbert, M.; Wakarchuk, W. W.; Jennings, H. J.
Glycoconjugate J. 1999, 16, 507-515.
(24) Welch, D. R.; Bisi, J. E.; Miller, B. E.; Conaway, D.; Seftor, E. A.; Yohem,
K. H.; Gilmore, L. B.; Seftor, R. E.; Nakajima, M.; Hendrix, M. J. Int. J.
Cancer 1991, 47, 227-237.
9
J. AM. CHEM. SOC. VOL. 124, NO. 42, 2002 12443

A R T I C L E S
Lee et al.
dichloromethane was stirred for 30 min at room temperature under Ar.
Silver trifluoromethanesulfonate (1.38 g, 5.4 mmol) was then added.
The mixture was stirred for another 30 min in the dark and then cooled
to -50 °C. Compound 6 (2.53 g, 3.6 mmol), dissolved in dried
dichloromethane, was added slowly to the above mixture and allowed
to warm to room temperature for 20 h in the dark. The yellowish-
white precipitate was removed by filtration through a pad of Celite,
and the colorless filtrate was concentrated and then purified by column
chromatography (CHCl3/MeOH, 30/1) to provide the peracetylated
derivative (1.77 g, 70%). This was dissolved in anhydrous methanol
and a catalytic amount of sodium methoxide (25wt % solution in
methanol) was added (pH ∼ 10). The reaction was stirred at ambient
temperature for 7 h. Amberlite IR-120 (plus) ion-exchange resin was
added to neutralize the base, and then the solution was filtered and
In contrast, our approach is based on a human scFv library and
the isolation in vitro of preexisting antibodies that can bind
carbohydrate epitopes. As in our previous work,9 the coupling
of a monovalent hapten to a carrier protein afforded a panning
reagent that was sufficient to provide selection of the relevant
antibody clones. Yet, it is also conceivable that some of the
GM3 ligands coupled to lysine residues in close proximity on
BSA mimic a GSL-like cell-surface “microdomain” or similar
multivalent cluster. Regardless, when given a large library of
possible clones, the panning reagent strategy provides what is
functionally required to allow isolation of at least a few scFvs
that have the desired characteristics of specificity and affinity
for the native antigen. Consequently, in light of the simpler
design, haptens and panning reagents based on other members
of the GSL family and of carbohydrates present on glycoprotein
and mucin structures will be much more synthetically accessible.
In this way, a variety of scFvs should be available against
virtually any tumor-associated carbohydrate.
The scFv GM3A8 with its favorable binding properties for
melanoma and breast tumor cells provides a solid foundation
for further development. The in vitro affinity of the scFv is
estimated to be comparable to a previously reported GM3-binding
murine IgG and anti-GM3 polyconal antibodies,11,26 but with the
significant advantages of possessing a human antibody sequence
and having the ability to specifically recognize highly metastatic
cancer cells. Moreover, the scFv format greatly facilitates genetic
manipulations, such as mutagenesis for affinity enhancement,
introduction of a cysteine residue for precise attachment of
drugs, and grafting onto the appropriate immunoglobulin
scaffold to enlist relevant effector functions. In future work,
we will utilize GM3A8 variants bearing such modifications for
preliminary analyses in cell culture. Some of these antibodies
may then be worthy of subsequent evaluation against tumors
in animal models and ultimately in clinical studies for treatment
of human cancers.
1
evaporated to dryness to give 7 as a fluffy solid (1.02 g, 99%). H
NMR (600 MHz, CD3OD) δ: 4.37-4.35 (m, 2H), 4.03-4.00 (m, 1H),
3.93-3.91 (m, 1H), 3.86-3.69 (m, 5H), 3.61-3.42 (m, 8H), 3.28 (dd,
J ) 8.8, 7.9 Hz, 1H). 13C NMR (150 MHz, CD3OD) δ: 105.04, 104.28,
80.59, 77.02, 76.48, 76.34, 74.76, 74.66, 72.51, 70.25, 69.37, 62.46,
61.91, 52.00. HRMS (M + Na+): calcd for C14H25N3O11Na 434.1387,
found 434.1380.
Compound 8. Azide 7 (0.23 g, 0.56 mmol) and CMP-sialic acid 9
(0.41 g, 0.65 mmol) were dissolved in sodium cacodylate (100 mL,
200 mM, pH 6.5) containing MnCl2 (20 mM). The R-2,3-sialyltrans-
ferase-CMPNeuAc synthetase (12 U) was added and the reaction was
slowly stirred at room temperature for 18 h. When all of the starting
material was converted to product (TLC; EtOAc/HOAc/MeOH/H2O,
45/20/20/15), the mixture was ultrafiltered through a 10 000 MWCO
membrane. The filtrate was passed through Dowex anion resin (1-X8,
200-400 mesh, phosphate-form 4 × 10 cm) and lyophilized. The
residue was loaded onto a column of Sephadex G15 (5 × 140 cm),
equilibrated, and eluted with 5% n-BuOH in H2O. Appropriate fractions
were collected and lyophilized to give 8 (0.31 g, 0.43 mmol, 77%). 1H
NMR (600 MHz, D2O) δ: 4.52 (m, 2H), 4.11-3.54 (m, 22H), 3.33 (t,
J ) 8.5 Hz, 1H), 2.75 (dd, J ) 12.3, 4.4 Hz, 1H, Heq of sialic acid),
2.02 (s, 3H, NAc), 1.79 (t, J ) 11.8 Hz, 1H, Hax of sialic acid). 13C
NMR (150 MHz, D2O, trace acetone added as internal reference at δ
29.92) δ: 174.70, 173.61, 102.32, 101.87, 99.48, 77.84, 75.16, 74.85,
74.50, 73.99, 72.56, 72.44, 71.45, 69.05, 68.22, 68.04, 67.76, 67.14,
62.25, 60.71, 59.70, 51.36, 50.20, 39.31, 21.72. ESI (M + H+) calcd
for C25H42N4NaO19 725, found 725.
Experimental Section
General Synthetic Methods. Unless otherwise noted, materials were
obtained from commercial suppliers and used without further purifica-
tion. Dichloromethane (CH2Cl2) was distilled from calcium hydride
under N2, and methanol (MeOH) was distilled from magnesium turnings
and iodine under N2. All reactions were run under an inert atmosphere
with dry reagents and solvents and flame-dried glassware. Flash column
chromatography was carried out using Merck 60 230-400 mesh silica
gel. Thin-layer chromatography was performed using 0.25 mm silica
gel coated Kieselgel 60 F254 plates and visualized using aqueous cerium
Compound 2 (GM3-A). A mixture of compound 8 (20 mg, 27.6
µmol) and 10% Pd/C (10 mg) in MeOH/H2O (50/50) was hydrogenated
at atmospheric pressure for 2 h at room temperature. The reaction
mixture was filtered through a Celite pad and the filtrate was evaporated
and dissolved in 0.1 M NaHCO3 solution (0.5 mL). Compound 10 (15
mg, 55 µmol, suspended in 0.5 mL of MeOH) was added and the
mixture was stirred at room temperature for 1 h, during which time
the suspended 10 slowly dissolved, affording a clear solution. After
concentration of the reaction mixture, the residue was purified by gel
permeation chromatography (Sephadex G-15, 5% n-BuOH in H2O) to
give hapten 2 (10 mg, 42%) as a fluffy white solid after lyophilization.
1H NMR (600 MHz, CD3OD) δ: 6.82 (s, 2H), 4.43 (d, J ) 7.5 Hz,
1H), 4.33 (d, J ) 7.9 Hz, 1H), 4.05 (dd, J ) 9.7, 3.1 Hz, 1H), 3.93-
3.43 (m, 23H), 3.27 (t, J ) 8.5 Hz, 1H), 2.87 (dd, J ) 12.1, 3.8 Hz,
1H, Heq of sialic acid), 2.21 (t, J ) 7.5 Hz, 2H), 2.01 (s, 3H), 1.88 (q,
J ) 7.0 Hz, 2H), 1.73 (t, J ) 11.8 Hz, 1H, Hax of sialic acid). 13C
NMR (150 MHz, CD3OD) δ: 175.51, 175.09, 174.93, 172.62, 135.43,
105.05, 104.28, 101.08, 80.79, 77.63, 77.06, 76.46, 76.18, 74.92, 74.74,
72.97, 70.81, 70.06, 69.68, 69.33, 68.96, 64.55, 62.72, 61.85, 53.94,
42.08, 40.64, 38.08, 34.10, 25.66, 22.61. HRMS (M + H+): calcd for
C33H51N3O22Na 864.2862, found 864.2847.
1
molybdate or 5% sulfuric acid in ethanol with heating. H NMR and
13C NMR spectra were recorded on a Bruker AMX-500 spectrometer
at 500 and 125 MHz or a Bruker AMX-600 at 600 and 150 MHz,
respectively, using the indicated solvents at 25 °C. Chemical shifts are
reported in ppm on the δ scale. Compounds 10 and 16 were purchased
from Sigma. The R-2,3-sialyltransferase-CMPNeuAc synthetase fusion
protein,27 CMP-sialic acid 9,16 and compound 1528 were prepared
according to literature procedures.
Compound 7. A mixture of 2-azidoethanol (2.5 g, 28.8 mmol), 2,4,6-
collidine (0.57 mL, 4.3 mmol), and 4 Å molecular sieves in dried
(27) Gilbert, M.; Bayer, R.; Cunningham, A. M.; DeFrees, S.; Gao, Y.; Watson,
D. C.; Young, N. M.; Wakarchuk, W. W. Nat. Biotechnol. 1998, 16, 769-
772.
(28) Li, Q.; Chu, D. T. W.; Claiborne, A.; Cooper, C. S.; Lee, C. M.; Raye, K.;
Berst, K. B.; Donner, P.; Wang, W.; Hasvold, L.; Fung, A.; Ma, Z.; Tufano,
M.; Flamm, R.; Shen, L. L.; Baranowski, J.; Nillius, A.; Alder, J.;
Muelbroek, J.; Marsh, K.; Crowell, D.; Hui, Y.; Seif, L.; Melcher, L. M.;
Henry, R.; Spanton, S.; Faghih, R.; Klein, L. L.; Tanaka, S. K.; Plattner,
J. J. J. Med. Chem. 1996, 39, 3070-3088.
Compound 12. This compound was prepared using a literature
procedure.19 Neat allylamine (15 mL) was added to the R-D-lactose
monohydrate (2.0 g, 5.5 mmol), which gradually became a clear
solution. This mixture was stirred at room temperature for 70 h, then
9
12444 J. AM. CHEM. SOC. VOL. 124, NO. 42, 2002

Phage-Display Selection of a Human Fv Antibody
A R T I C L E S
the solvent was removed by evaporation and the crude white solid was
treated with toluene and evaporated three times. The flask was chilled
in an ice bath, then pyridine (30 mL) and acetic anhydride (22 mL)
were added, and the mixture was allowed to warm to room temperature
overnight. The solvent was removed and the residue was taken up in
CHCl3, washed with 1 M HCl, 1 M NaHCO3, and brine, and then dried
over Na2SO4 and concentrated. The crude product was purified by
column chromatography (CHCl3/MeOH, 20/1) to give 12 as a white
Preparation of 4. To a solution of BSA (5 mg/mL) in 50 mM Tris
buffer (pH 8.0) was added 2-iminothiolane-HCl (Traut’s reagent; Pierce)
(18 µL of a 0.25 M solution in H2O). The mixture was shaken for 1 h
at room temperature and then dialyzed twice against phosphate buffer
(10 mM, pH 7.4). The solution was transferred to a microcentrifuge
tube and hapten 2 (2.3 mg, 2.7 µmol), dissolved in phosphate buffer
(100 µL), was added. After shaking for 4 h at room temperature, the
mixture was dialyzed three times against phosphate buffer. The average
number of hapten molecules 2 attached per molecule BSA was
determined to be 17.0 by MALDI-TOF analysis.
1
solid (2.1 g, 53%). H NMR spectrum matched with that reported in
the literature.19 HRMS (M + Na+): calcd for C31H43NO18 740.2372,
found 740.2342. Mp: 85-90 °C (scinters).
Preparation of 5. This conjugate was prepared using hapten 3 and
following the same procedure as for 4. The average number of hapten
molecules 3 attached to BSA was 10.2.
Compound 13. Ozonized oxygen was bubbled into a solution of
12 (1.5 g, 2.1 mmol) in anhydrous MeOH at -78 °C until the blue
color persisted (∼30 min). The excess ozonized oxygen was purged
with the passage of Ar, then dimethyl sulfide (1 mL) was added at
-78 °C, and the reaction mixture was warmed to room temperature
by removal of the cooling bath. After 3 h, the solvent was removed by
evaporation to give the aldehyde as a white solid, which was used for
the subsequent reductive amination without purification. 1H NMR (500
MHz, DMSO-d6) δ: 9.07 (s, CHO). HRMS (M + Na+): calcd for
C30H41NO19Na 742.2170; found 742.2136.
Preparation of scFv-phage. Construction of the cancer patient
scFv-phage library has been described previously.9 Approximately 2
× 109 cells from the library glycerol stock were used to inoculate 200
mL of Super Broth (SB) medium containing 2% glucose, 100 µg/mL
carbenicillin, and 10 µg/mL tetracycline. The culture was shaken at 37
°C until OD600 ∼ 0.5, then ∼1012 colony-forming units (cfu) of
VCSM13 helper phage were added. After a 30 min incubation at room
temperature, the culture was diluted in 400 mL of SB (containing
carbenicillin and tetracycline) and grown at 30 °C. After 2 h, 70 µg/
mL kanamycin and 0.5 mM isopropyl â-D-thiogalactopyranoside (IPTG)
were added, and the culture was allowed to grow overnight. Phage
particles were recovered from the culture medium by precipitating with
3% (w/v) NaCl and 4% (w/v) poly(ethyleneglycol) 8000.
Selection of scFv-phage. The library was subjected to four rounds
of panning. Immunotubes (Maxisorb, Nunc) were coated overnight at
4 °C with 0.25 mL of 10 µg/mL GM3-A-BSA or GM3-B-BSA in PBS
(10 mM phosphate, 150 mM NaCl, pH 7.4). After blocking with
BLOTTO (4% nonfat dry milk in PBS) for 1 h at 37 °C, ∼1012 cfu of
scFv-phage in 2 mL PBS containing 1% milk and 3% BSA were added
and incubated for 2 h with rocking at 37 °C. The tube was washed
with PBS/ 0.1% Tween and PBS (twice for round 1, five times for
round 2, and 20 times for subsequent rounds). Bound phage were eluted
from the tube with 0.25 mL of 0.1 M glycine (pH 2.2) and neutralized
with 15 µL of 2 M Tris base. Eluted phage were amplified by infecting
fresh E. coli XL-1 Blue cells (Stratagene) and rescued as outlined in
the preceding section. An initial panel of scFv-phage clones was selected
using ELISA.
To compound 15 (0.43 g, 2.1 mmol), sodium cyanoborohydride (0.11
g, 1.7 mmol), and acetic acid (43 µL, 0.7 mmol) in anhydrous MeOH
was added the above aldehyde (1.0 g, 1.4 mmol) dissolved in MeOH.
After 3 h at room temperature, the solvent was removed by evaporation
and the crude compound was directly purified by column chromatog-
raphy (CHCl3/MeOH, 10/1) to give the peracetylated intermediate as
a white solid (0.4 g). HRMS (M + H+) calcd for C41H58N3O20 912.3613,
found 912.3590. All acetyl groups were then removed using a catalytic
amount of sodium methoxide (25wt
% solution in meth-
anol) in anhydrous methanol (pH ∼ 10) for 7 h to give 13 as a
white solid (0.43 g, 35% from 12). 1H NMR (600 MHz, CD3OD)
δ: 7.36-7.29 (m, 5H), 5.07 (s, 2H), 4.81 (d, J ) 8.8 Hz, 1H), 4.38
(d, J ) 7.9 Hz, 1H), 3.92-3.27 (m, 14H), 3.25 (m, 2H), 2.87 (m, 1H),
2.62-2.50 (m, 3H), 2.36 (s, 3H, NCH3), 2.19 (s, 3H, NAc). 13C
NMR (150 MHz, CD3OD) δ: 175.64, 158.72, 138.33, 129.44, 128.95,
128.81, 105.07, 88.93, 80.37, 78.84, 77.03, 76.56, 74.75, 72.44,
71.28, 70.19, 67.42, 62.41, 61.89, 58.35, 57.09, 42.96, 40.38, 39.02,
22.14. HRMS (M + H+): calcd for C27H44N3O13 618.2874, found
618.2893.
Compound 14. This compound was prepared from 13 (20 mg, 32
µmol), 9 (26.5 mg, 49 µmol), and R-2,3-sialyltransferase-CMPNeuAc
synthetase (12 U) following the same procedures as described for 8,
affording a fluffy white solid (20 mg, 65%). 1H NMR (600 MHz, CD3-
OD) δ: 7.37-7.31 (m, 5H), 5.11 (s, 2H), 4.44 (d, J ) 7.9 Hz, 1H),
4.06 (dd, J ) 9.7, 3.1 Hz, 1H), 3.92-3.37 (m, 22H), 3.14-3.05 (m,
4H), 2.87 (dd, J ) 12.3, 4.4 Hz, 1H, Heq of sialic acid), 2.76 (s, 3H,
NCH3), 2.22 (s, 3H, NAc), 2.02 (s, 3H), 1.74 (t, J ) 11.8 Hz, 1H, Hax
of sialic acid). 13C NMR (150 MHz, CD3OD) δ: 174.58 (2), 173.91,
158.10, 137.14, 128.57, 128.16, 127.95, 104.11, 100.09, 87.49, 79.21,
77.89, 76.71, 76.12, 75.66, 73.96, 72.03, 70.34, 69.80, 69.12, 68.26,
68.00, 66.82, 66.74, 63.71, 61.67, 60.44, 57.34, 56.07, 52.99, 42.80,
41.50, 41.08, 21.63, 21.25; HRMS (M + H+): calcd for C38H60N4-
O21Na 931.3648, found 931.3657.
ELISA Methods. Relative affinities and specificity of scFv-phage
and soluble scFvs were evaluated by ELISA. GM3-A-BSA and GM3
-
B-BSA were applied to a microtiter plate at a concentration of 5 µg/
mL in PBS at 4 °C overnight. Unmodified BSA (5 µg/mL) was used
as a negative control. After being washed three times with water, the
wells were blocked with BLOTTO for 1 h at 37 °C. Following
additional washing, for phage ELISA, 25 µL of scFv-phage suspension
was added to each well, and the plate was incubated 1 h at 37 °C.
After washing, 25 µL horseradish peroxidase (HRP)-conjugated anti-
M13 mAb (Amersham Pharmacia) was then added for 30 min at 37
°C. For ELISA using soluble scFvs, 25 µL of purified scFv was added
to each well, and the plate was incubated 1 h at 37 °C. Then, after
washing, 25 µL of HRP-conjugated anti-Flag M2 mAb (Sigma) was
added for 30 min at 37 °C. For detection, 50 µL/well of tetramethyl-
benzidine (TMB) substrate (Pierce) was added and the plate was
incubated at room temperature. After adding 50 µL of 2 M H2SO4 to
each well, the absorbance was read at 450 nm. Sequencing of clones
was performed using an Applied Biosystems 377 automated DNA
sequencer, and sequences were analyzed using the MacVector program
and DNA Plot (http:// www.dnaplot.de).
Compound 3 (GM3-B). This compound was prepared from 14 (18
mg, 19.3 µmol) and 16 (9.6 mg, 38 µmol) using a procedure analogous
to the preparation of hapten 2, affording a fluffy white solid (8 mg,
45%). 1H NMR (600 MHz, CD3OD) δ: 6.82 (s, 2H), 4.46 (d, J ) 7.9
Hz, 1H), 4.06 (dd, J ) 9.7, 3.1 Hz, 1H), 3.96-3.48 (m, 24H), 3.26
(m, 4H), 2.87-2.85 (m, 4H), 2.23 (s, 3H), 2.01 (s, 3H), 1.73 (t, J )
11.8 Hz, 1H, Hax of sialic acid). 13C NMR (150 MHz, CD3OD) δ:
175.51, 175.35, 174.91, 172.78, 172.45, 135.47, 105.05, 101.08, 88.54,
80.33, 78.95, 77.62, 77.13, 76.73, 74.93, 72.98, 71.41, 70.73, 70.06,
69.34, 68.97, 64.56, 62.70, 61.86, 57.48, 55.88, 53.94, 43.87, 42.10,
41.70, 41.54, 35.69, 22.60, 22.11. HRMS (M + H+): calcd for
C36H57N5O22Na 934.3393, found 934.3396.
Expression and Purification of scFvs. Three representative scFv
M3-A clones (GM3A6, GM3A8, and GM3A15) were ligated into the
expression vector pETFlag (derived from pET-15b, Novagen) and
transformed into E. coli BL21 cells (Stratagene). Expression of soluble
scFv was induced by growth of each culture overnight at 30 °C in SB
containing 0.5 mM IPTG. The Flag-tagged scFvs were purified from
G
9
J. AM. CHEM. SOC. VOL. 124, NO. 42, 2002 12445

A R T I C L E S
Lee et al.
the growth media using anti-Flag M2 mAb affinity agarose (Sigma),
with 0.1 M glycine (pH 2.7) as eluant. The yield of purified scFv
typically varied from 0.3 to 3 mg/L, depending on the clone.
BIAcore Analysis. Binding kinetics was determined by surface
plasmon resonance using a BIAcore 2000 instrument (Pharmacia).29
Hapten 2 was immobilized on a CM5 sensor chip using the BIAcore
surface thiol method, per the manufacturer’s instructions. All measure-
ments were conducted in HEPES-buffered saline with a flow rate of
10 µL/min. After each measurement, the chip surface was regenerated
with 10 µL of 1 mM HCl.
Flow Cytometry. We used flow cytometry to analyze antibody
reactivity toward three human melanoma cell lines: HMV-1, a poorly
metastatic cell line;22 M21, which is moderately metastatic;23 and
C-8161, a highly metastatic cell line.24 We also tested three human
breast cancer cell lines that were established from circulating metastatic
cells isolated from peripheral blood samples of stage IV breast cancer
patients. Blood samples were incubated with supramagnetic beads
(Dynal) coated with an antibody against a human epithelial antigen
(mAb Ber-Ep4). Tumor cells were recovered and expanded in culture.
The cell lines established from these cultures, BCM-1, BCM-2, and
BMS, exhibit adhesive, invasive, and migratory properties of highly
metastatic tumor cells and will be described elsewhere. Normal human
dermal fibroblast (HDFa) cells (Cascade Biologics, Inc.) were used as
a nonmalignant cell control. The cells were harvested in PBS/EDTA
and incubated with an anti-GM3 scFv (10 µg/mL in 0.5% BSA/PBS),
or only 0.5% BSA/PBS without scFv as a negative control, for 45 min
on ice. After washing twice in PBS/BSA and blocking with normal
goat serum (10 min at 22 °C), the cells were incubated with anti-Flag
M2 mAb (10 µg/mL) (Sigma) in PBS/BSA on ice for an additional 30
min. Finally, the cells were washed twice and incubated with FITC-
labeled anti-mouse goat F(ab′)2 (Pierce) for 30 min on ice. After
washing and counter-staining with propidium iodide (2 µg/mL) to
identify and exclude dead cells, the samples were analyzed in a
FACScan flow cytometer (Becton Dickinson).
Acknowledgment. We thank Dr. Peter Wirsching for invalu-
able assistance in the preparation of the manuscript. This work
was supported by funding from the Skaggs Institute for
Chemical Biology, The National Institutes of Health [Grants
NIAID-AI47127 (K.D.J.) and CA95458-01 (B.F.-H.)], Califor-
nia Cancer Research Program [00-00757V-20012 (K.D.J.)], and
US Army [Grant DAMD 17-99-1-9368 (B.F.-H.)].
(29) (a) Medaglia, M. V.; Fisher, R. J. Protein-Protein Interact. 2002, 255-
272. (b) Roder, U. W.; Markey, F. Methods Mol. Med. 1998, 13, 531-
554. (c) Malmborg, A.-C.; Michae¨lsson, A.; Ohlin, M.; Jansson, B.;
Borrebaeck, C. A. K. Scand. J. Immunol. 1992, 35, 643-650. (d) Jo¨nsson,
U.; Fa¨gerstam, L.; Ivarsson, B.; Johnsson, B.; Karlsson, R.; Lundh, K.;
Lo¨fås, S.; Persson, B.; Roos, H.; Ro¨nnberg, I.; Sjo¨lander, S.; Stenberg, B.;
Ståhlberg, R.; Urbaniczky, C.; O¨ stlin, H.; Malmqvist, M. BioTechniques
1991, 11, 620-627.
Supporting Information Available: Copies of the 1H NMR
and 13C NMR spectra for all prepared compounds. This material
is available free of charge via the Internet at http://pubs.acs.org.
JA020737J
9
12446 J. AM. CHEM. SOC. VOL. 124, NO. 42, 2002