Erythropoietin mimetics derived from solution phase combinatorial libraries

Article abstract of DOI:10.1021/ja0118789

The erythropoietin receptor (EPOr) is activated by ligand-induced homodimerization, which leads to the proliferation and differentiation of erythroid progenitors. Through the screening of combinatorial libraries of dimeric iminodiacetic acid diamides, nov

Full text of DOI:10.1021/ja0118789

Published on Web 12/29/2001
Erythropoietin Mimetics Derived from Solution Phase
Combinatorial Libraries
Joel Goldberg, Qing Jin, Yves Ambroise, Shigeki Satoh, Joel Desharnais,
Kevin Capps, and Dale L. Boger*
Contribution from the Department of Chemistry and The Skaggs Institute for Chemical Biology,
The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, California 92037
Received August 2, 2001
Abstract: The erythropoietin receptor (EPOr) is activated by ligand-induced homodimerization, which leads
to the proliferation and differentiation of erythroid progenitors. Through the screening of combinatorial libraries
of dimeric iminodiacetic acid diamides, novel small molecule binders of EPOr were identified in a protein
binding assay. Evaluation of a series of analogues led to optimization of binding subunits, and these were
utilized in the synthesis of higher order dimer, trimer, and tetramer libraries. Several of the most active
EPOr binders were found to be partial agonists and induced concentration-dependent proliferation of an
EPO-dependent cell line (UT-7/EPO) while having no effect on a cell line lacking the EPOr (FDC-P1). An
additional compound library, based on a symmetrical isoindoline-5,6-dicarboxylic acid template and including
the optimized binding subunits, was synthesized and screened leading to the identification of additional
EPO mimetics.
Introduction
fied through phage display or the random screening of com-
pound libraries. Though successful, the limited number of
Ligand-induced receptor dimerization or oligomerization has
emerged as a general mechanism for signal transduction,1 and
important members of several receptor superfamilies are acti-
vated by this process (Figure 1).2-5 Many of these receptors
appear to bind their ligands using only small clusters of residues
for a majority of the binding interaction,6-8 which has led to
the expectation that smaller molecules may be capable of
inducing their activation. Unlike the discovery of small molecule
antagonists of biological activity functioning through the
disruption of protein-protein interactions, such agonists must
function by promoting a productive protein-protein interaction.
Recently, the first peptide,7-9 as well as nonpeptide,10-12
agonists promoting receptor homodimerization have been identi-
examples, the size of the agonists,7-9,12 or questionable nature
of the receptor activation10,11 have not yet provided a generaliz-
able approach to the discovery or development of such small
molecule receptor agonists.
The erythropoietin receptor (EPOr)13 is a member of the
cytokine receptor superfamily, whose members share similar
structural motifs14 and mechanisms of signal transduction, in
which activation is believed to be achieved through an ap-
propriately oriented ligand-induced homodimerization.15-17
EPOr’s endogenous high affinity ligand, erythropoietin (EPO),
is a 34 kDa glycoprotein hormone that controls red blood cell
production by promoting the proliferation and differentiation
of erythroid progenitors.18 Recombinant EPO is one of the most
successful biotechnology products19 and is currently used to treat
anemias stemming from kidney failure, cancer chemotherapy,
and AIDS.20 Moreover, clinical observations on patients receiv-
(1) Hinterding, K.; Alonso-Diaz, D.; Waldmann, H. Angew. Chem., Int. Ed.
1998, 37, 688. Klemm, J. D.; Schreiber, S. L.; Crabtree, G. R. Annu. ReV.
Immunol. 1998, 16, 569. Heldin, C.-H. Cell 1995, 80, 213.
(2) Ullrich, A.; Schlessinger, J. Cell 1990, 61, 203.
(3) Moutoussamy, S.; Kelly, P. A.; Finidori, J. Eur. J. Biochem. 1998, 255, 1.
(4) Massague, J.; Attisano, L.; Wrana, J. L. Trends Cell Biol. 1994, 4, 172.
Lemmon, M. A.; Schlessinger, J. Trends Biochem. Sci. 1994, 19, 459.
(5) Smith, C. A.; Farrah, T.; Goodwin, R. G. Cell 1994, 76, 959.
(6) Wells, J. A. Science 1996, 273, 449. Reineke, U.; Schneider-Mergener, J.
Angew. Chem., Int. Ed. 1998, 37, 769.
(12) Qureshi, S. A.; Kim, R. M.; Konteatis, Z.; Biazzo, D. E.; Motamedi, H.;
Rodrigues, R.; Boice, J. A.; Calaycay, J. R.; Bednarek, M. A.; Griffin, P.;
Gao, Y.-D.; Chapman, K.; Mark, D. F. Proc. Natl. Acad. Sci. U.S.A. 1999,
96, 12156.
(7) Wrighton, N. C.; Farrell, F. X.; Chang, R.; Kashyap, A. K.; Barbone, F.
P.; Mulcahy, L. S.; Johnson, D. L.; Barrett, R. W.; Jolliffe, L. K.; Dower,
W. J. Science 1996, 273, 458.
(8) Livnah, O.; Stura, E. A.; Johnson, D. L.; Middleton, S. A.; Mulcahy, L.
S.; Wrighton, N. C.; Dower, W. J.; Jolliffe, L. K.; Wilson, I. A. Science
1996, 273, 464.
(13) Syed, R. S.; Reid, S. W.; Li, C.; Cheetham, J. C.; Aoki, K. H.; Liu, B.;
Zhan, H.; Osslund, T. D.; Chirino, A. J.; Zhang, J.; Finer-Moore, J.; Elliott,
S.; Sitney, K.; Katz, B. A.; Matthews, D. J.; Wendoloski, J. J.; Egrie, J.;
Stroud, R. M. Nature 1998, 395, 511.
(14) Bazan, F. J. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 6934.
(15) Boger, D. L.; Goldberg, J. Bioorg. Med. Chem. 2001, 9, 557.
(16) Touw, I. P.; De Koning, J. P.; Ward, A. C.; Hermans, M. H. A. Mol. Cell.
Endocrinol. 2000, 160, 1.
(9) Cwirla, S. E.; Balasubramanian, P.; Duffin, D. J.; Wagstrom, C. R.; Gates,
C. M.; Singer, S. C.; Davis, A. M.; Tansik, R. L.; Mattheakis, L. C.; Boytos,
C. M.; Schatz, P. J.; Baccanari, D. P.; Wrighton, N. C.; Barrett, R. W.;
Dower, W. J. Science 1997, 276, 1696. Kimura, T.; Kaburaki, H.;
Miyamoto, S.; Katayama, J.; Watanabe, Y. J. Biochem. 1997, 122, 1046.
(10) Tian, S.-S.; Lamb, P.; King, A. G.; Miller, S. G.; Kessler, L.; Luengo, J.
I.; Averill, L.; Johnson, R. K.; Gleason, J. G.; Pelus, L. M.; Dillon, S. B.;
Rosen, J. Science 1998, 281, 257.
(17) Wilson, I. A.; Jolliffe, L. K. Curr. Opin. Struct. Biol. 1999, 9, 696. Livnah,
O.; Stura, E. A.; Middleton, S. A.; Johnson, D. L.; Jolliffe, L. K.; Wilson,
I. A. Science 1999, 283, 987. Remy, I.; Wilson, I. A.; Michnick, S. W.
Science 1999, 283, 990. See also: Constantinescu, S. N.; Keren, T.;
Socolovsky, M.; Nam, H.; Henis, Y. I.; Lodish, H. F. Proc. Natl. Acad.
Sci. U.S.A. 2001, 98, 4379.
(11) Kimura, T.; Kaburaki, H.; Tsujino, T.; Ikeda, Y.; Kato, H.; Watanabe, Y.
FEBS Lett. 1998, 428, 250.
(18) Lacombe, C.; Mayeux, P. Haematologica 1998, 83, 724.
(19) Annual worldwide sales, $4 billion: Chem. Eng. News 2000, 77, 17.
9
544 VOL. 124, NO. 4, 2002 J. AM. CHEM. SOC.
10.1021/ja0118789 CCC: $22.00 © 2002 American Chemical Society

Erythropoietin Mimetics
A R T I C L E S
enlisted a technically nondemanding solution phase strategy
which dependably delivers pure individual compounds or small
to large combinatorial mixtures. Symmetrical cyclic anhydride
templates, such as N-BOC iminodiacetic acid anhydride (1),
which are readily substituted at up to three positions, can later
be combined through a variety of linking strategies to provide
symmetrical and unsymmetrical dimer and higher-order librar-
ies.23,24 In addition to dimerization through amide bond forming
reactions, recent efforts have expanded the scope of this
approach by introducing dimerization and trimerization through
the formation of carbon-carbon bonds via olefin metathesis,25
Stille,26 and biaryl27 coupling reactions. In all cases, reaction
and workup conditions were designed to permit the isolation
and purification of products through simple liquid-liquid or
liquid-solid extractions. Notably, solution phase, but not solid
phase, combinatorial chemistry uniquely permits the conver-
gent25 dimerization of such functionalized subunits to provide
candidate leads for promoting receptor activation, and this has
been the basis behind much of our prior work. We previously
described the synthesis of diverse libraries derived from 1 that
were suited for the study of such protein-protein interactions.24
Herein, we report the identification of novel erythropoietin
mimetics through the screening of these and related combina-
torial libraries based on iminodiacetic acid and isoindoline-5,6-
dicarboxylic acid templates that was conducted over a period
of several years.
Results and Discussion
Iminodiacetic Acid Libraries: EPOr Binding Activity. In
recent studies, the identification of cyclic peptide dimers with
the ability to mimic EPO,7,8 together with structural details of
the intricate ligand-receptor and receptor-receptor interactions,
have been described.8,28 In the X-ray structure of the complex
between EMP1 (erythropoietin mimetic peptide 1) and the
extracellular binding fragment of EPOr (EBP), the peptide ligand
and receptor were observed as a symmetrical 2:2 dimer.8
Although each EMP1-EBP interface was relatively large (420
Å2), key binding sites were identified that could potentially be
exploited by smaller molecules. Two main sites of interaction
were observed between the cyclic peptide ligand and the receptor
(Figure 1). A hydrophobic core comprised of two phenylalanine
side chains and a methionine residue of EBP binds the tyrosine,
phenylalanine, tryptophan, and cystine residues of EMP1. The
hydroxy group on the peptide tyrosine (Tyr4) also serves as a
hydrogen bond acceptor in this pocket. The second key
Figure 1. Top: General mechanism of cytokine activation of the JAK/
STAT signal transduction pathway. Bottom: Key residues of EMP1
involved in the agonist peptide-EPOr interaction, taken from the X-ray
structure of EMP1 bound to EBP.8 Two major sites of interaction are
evident: a hydrophobic pocket, outlined by Tyr4 (which also hydrogen bonds
to EPOr), Phe,8 and Trp13 of EMP1, and hydrogen bonds from three
contiguous residues (Gly,9 Pro,10 and Leu11) of the peptide backbone.
ing recombinant human EPO and subsequent in vivo testing
where it was found to induce complete myleoma tumor
regression in 30-60% of the treated mice have shown that it
may act as an antitumor therapeutic agent in its own right.21
Thus, the discovery of effective small molecule EPO mimetics
that might replace the use of the injectable recombinant protein
would be of significant scientific, medical, and commercial
importance.
In the absence of candidate lead structures, we turned to
combinatorial chemistry22 to identify potential EPO mimetics.
Moreover, we adapted an approach whereby candidate ligands
were initially screened for EPOr binding, then dimerized in
efforts to convert EPOr antagonists into EPOr agonists. We
(23) Cheng, S.; Comer, D. D.; Williams, J. P.; Myers, P. L.; Boger, D. L. J.
Am. Chem. Soc. 1996, 118, 2567. Boger, D. L.; Tarby, C. M.; Myers, P.
L.; Caporale, L. H. J. Am. Chem. Soc. 1996, 118, 2109. Cheng, S.; Tarby,
C. M.; Comer, D. D.; Williams, J. P.; Caporale, L. H.; Myers, P. L.; Boger,
D. L. Bioorg. Med. Chem. 1996, 4, 727. Boger, D. L.; Ducray, P.; Chai,
W.; Jiang, W.; Goldberg, J. Bioorg. Med. Chem. Lett. 1998, 8, 2339. Boger,
D. L.; Ozer, R. S.; Andersson, C.-M. Bioorg. Med. Chem. Lett. 1997, 7,
1903. Boger, D. L.; Goldberg, J. Multi-Step Solution Phase Combinatorial
Synthesis. In Combinatorial Chemistry: A Practical Approach; Finniri,
H., Ed.; Oxford University Press: New York, 2000; p 303.
(24) Boger, D. L.; Goldberg, J.; Jiang, W.; Chai, W.; Ducray, P.; Lee, J. K.;
Ozer, R. S.; Andersson, C.-M. Bioorg. Med. Chem. 1998, 6, 1347.
(25) Boger, D. L.; Chai, W.; Jing, Q. J. Am. Chem. Soc. 1998, 120, 7220. Boger,
D. L.; Chai, W. Tetrahedron 1998, 54, 3955. Boger, D. L.; Chai, W.; Ozer,
R. S.; Andersson, C.-M. Bioorg. Med. Chem. Lett. 1997, 7, 463.
(26) Boger, D. L.; Jiang, W.; Goldberg, J. J. Org. Chem. 1999, 64, 7094.
(27) Boger, D. L.; Goldberg, J.; Andersson, C.-M. J. Org. Chem. 1999, 64,
2422.
(20) Gabrilove, J. Semin. Hematol. 2000, 37(4 supp. 6), 1.
(21) Mittelman, M.; Neumann, D.; Peled, A.; Kanter, P.; Haran-Ghera, N. Proc.
Natl. Acad. Sci. U.S.A. 2001, 98, 5181.
(28) Johnson, D. L.; Farrell, F. X.; Barbone, F. P.; McMahon, F. J.; Tullai, J.;
Hoey, K.; Livnah, O.; Wrighton, N. C.; Middleton, S. A.; Loughney, D.
A.; Stura, E. A.; Dower, W. J.; Mulcahy, L. S.; Wilson, I. A.; Jolliffe, L.
K. Biochemistry 1998, 37, 3699.
(22) Floyd, C. D.; Leblanc, C.; Whittaker, M. Prog. Med. Chem. 1999, 36, 91.
9
J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002 545

A R T I C L E S
Goldberg et al.
Figure 3. Competitive binding activity (% inhibition of EPO binding) of
the individual compounds from the deconvolution of the most active
sublibraries of mixture library 2. 125I-EPO, in the presence of the library
members (50 µM), was incubated with immobilized EPOr, as described in
ref 27. Compounds were tested in triplicate, and the bars represent the
average % inhibition of 125I-EPO binding to the receptor. 2, A6B10C4 was
selected as a lead for further derivitizations.
competitive binding activity, the most active containing N-R-
CBZ-lysine methyl ester (B10) in the R2 position. Mixture
sublibraries were deconvoluted by synthesizing the individual
components of the mixtures that displayed the highest competi-
tive binding activity. Thus, the 50 compounds that comprised
the five most active mixtures were prepared to identify the
optimal linking dicarboxylic acid, R3.
Figure 2. The structure and building blocks of library 2. The library was
synthesized and screened as mixtures containing single R1 (A1-A6) and
R2 (B1-B10) substitutions, linked by an equimolar mixture of subunit R3
(C1-C10), 60 mixtures of 10 compounds. For the synthesis and charac-
terization of this library, see ref 24.
interaction entails three closely situated hydrogen bonds between
the peptide backbone of EMP1 and EBP.
These deconvoluted compounds were tested in the competi-
tive binding assay, and the results are shown in Figure 3.
Throughout the five series, the C4 linking dicarboxylic acid
(R3) exhibited the best binding activity. Thus, with R2 fixed as
B10 and R3 fixed as C4, a range of R1 derivatives exhibited
good competitive binding activity. Compound 2, A6B10C4
(Figure 3), which displayed 32% inhibition at 50 µM, was
selected as a lead for subsequent modifications. Its activity was
respectable when compared to that of EMP1 which has an IC50
of 5 µM in this binding assay.28,31 Compounds 2, A2B10C4
(42% inhibition at 50 µM) and 2, A5B10C4 (59% inhibition at
50 µM) were also identified as possessing significant activity.
Two series of follow-up libraries were prepared to optimize
the amine binding groups. At this stage, the N-R-CBZ-lysine
methyl ester (B10) group was identified as being critical for
the observed activity, which was modulated by the R1-NH2
(group A) substitution. Thus, a series of structurally diverse
primary and secondary amines (A7-A115, Figure 4) was
utilized, in conjunction with the N-R-CBZ-lysine methyl ester
(B10) and isophthalic acid (C4) linker, to prepare 109 new
derivatives. Screening results from the competitive binding assay
Solution phase combinatorial synthesis was employed to
generate an initial library of 600 C2-symmetrical compounds
(60 mixtures of 10 compounds, library 2, Figure 2) which were
designed to bind EPOr in the same manner as EMP1. The library
members contained functionality appropriately spaced to loosely
mimic the hydrophobic and hydrogen bonding interactions
revealed in the EMP1-EBP X-ray structure. The R1-NH2
(group A) amines were included to mimic Tyr4 of EMP1,
whereas the R2-NH2 (group B) amines included functionality
capable of emulating the hydrogen bonding peptide backbone
of EMP1. The approach constituted the dimerization linkage
of iminodiacetic acid diamides with a mixture of rigid dicar-
boxylic acids. The entire synthesis sequence to assemble the
library required three steps, and purification at each step was
achieved by the removal of excess reactants and reagents by
liquid-liquid extraction protocols.29
The library was screened (50 µM mixture concentration, 1%
DMSO) for binding in an assay that measured the ability of the
compounds to displace 125I-EPO from immobilized EPOr.30
Several of the compound mixtures displayed EPO-EPOr
(29) For a detailed description of the synthesis of this library, see ref 24.
(30) The protein binding assay was conducted by The R. W. Johnson
Pharmaceutical Research Institute, as described in ref 28.
(31) Connolly, P. J.; Wetter, S. K.; Murray, W. V.; Johnson, D. L.; McMahon,
F. J.; Farrell, F. X.; Tullai, J.; Jolliffe, L. K. Bioorg. Med. Chem. Lett.
2000, 10, 1995.
9
546 J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002

Erythropoietin Mimetics
A R T I C L E S
Figure 4. Structures of the 109 amine subunits (A7-A115) used in combination with the B10 and C4 subunits to explore the binding activity of lead
compound 2, A6B10C4. For the synthesis of compounds 2, A7B10C4-A115B10C4, see ref 24.
(Figure 5) revealed that most of the amine subunits did not
improve the activity and, in many cases, resulted in a loss in
affinity for the receptor, indicating the effectiveness of the
initially chosen R1-NH2 amines (A1-A6). In some cases,
interesting activity was observed, most notably with cyclic acetal
A8 (79% inhibition at 50 µM). The most active of these R1-
NH2 substitutions were included in the synthesis of subsequent
higher order libraries (vide infra).
The second series of analogues of 2, A6B10C4 contained
modifications to the N-R-CBZ-lysine methyl ester (B10) (Figure
6). These compounds maintained the 4-fluorophenethylamine
(A6) group and the isophthalic acid (C4) linker identified from
9
J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002 547

A R T I C L E S
Goldberg et al.
substitutions (B46-B65), although most larger substitutions
were less effective than the methyl ester, and none provided an
increase in potency. The carbamate functionality also surfaced
as necessary for activity, and none of the derivatives in which
it was removed (B23, B26-B35) were generally active. Many
substitutions on the aromatic ring are permitted, although none
of the new carbamates (B66-B84) led to activity that exceeded
that of the original benzyl carbamate, B10. Compound 2,
A6B36C4, which has the same general substitution as the N-R-
CBZ-lysine methyl ester but with the positions of the carbamate
oxygen and nitrogen atoms reversed, maintains some activity
(15% inhibition). The stereochemistry at the R-position is
important as the D-lysine derivative 2, A6B44C4 (5%) is
significantly weaker than the lead compound with the L-
stereochemistry. Replacement of the lysine R-amino by a
hydroxy group (included in B32-B36) also diminishes activity.
Recognizing the significance of the N-R-CBZ-lysine methyl
ester, this group was introduced in both positions of the
iminodiacetic acid diamide template (A7 ) B10) and was tested
in combination with all 10 linking dicarboxylic acids (com-
pounds 2, A7B10C1-10). Here, the acetylene dicarboxylic acid
linker (C1) provided the highest activity (45% inhibition at 50
µM), followed by the isophthalic acid (C4) linked agent (22%
inhibition at 50 µM).
Higher Order Libraries. Subsequent libraries were prepared
which expressed the iminodiacetic acid diamides as trimers and
tetramers, either through a tricarboxylic acid linkage or through
sequential dimerization couplings with dicarboxylic acids. The
structures in these libraries contained the key amine binding
groups identified in the initial EPOr binding screen. In the trimer
series, a total of 560 compounds were prepared (80 mixtures
of 7), and the five most active mixtures in the competitive
binding assay were deconvoluted by resynthesis of the 35
individual compounds.32 Several of these agents displayed
activity on the same level as the iminodiacetic acid dimers, but
none surpassed the initial series. As expected, compounds
containing the N-R-CBZ-lysine methyl ester (B10) exhibited
significant inhibition; however, the most potent agent in the
series was 3, A2B6C11, which at 50 µM resulted in a 38%
inhibition of EPO binding (Figure 8a). The tetramer compounds
were synthesized on a broader scale where a total of 1596
compounds were screened in mixtures of 8-10.32 Deconvolution
of the most active mixtures and screening 176 individual
compounds provided agents that were of similar activity to the
corresponding iminodiacetic acid diamide dimers discussed
above. The two most active tetramers in the binding assay
contained the A8 and A13 binding groups, combined with the
N-R-CBZ-lysine methyl ester (B10), which displayed 52% (4,
A8B10C12) and 47% (4, A13B10C12) inhibition of EPO
Figure 5. Competitive binding activity of compounds 2, A7B10C4-
A115B10C4. Each compound (50 µM) was incubated with 125I-EPO and
immobilized EPOr, as described in ref 27. Compounds were tested in
triplicate, and the bars represent the average % inhibition of 125I-EPO binding
to the receptor.
the initial deconvoluted library. These agents included com-
mercially available amines (B11-B26 and B85-B87) possess-
ing some of the structural features of N-R-CBZ-lysine methyl
ester, as well as synthesized amines which contained systematic
changes to the B10 subunit, especially the ester and carbamate
functionalities. Each of these compounds (2, A6B10C4-
A6B87C4) was screened in the EPO competitive binding assay
(50 µM concentration, Figure 7). Most of the analogues
displayed very little activity in comparison to lead compound
2, A6B10C4. The most active were 2, A6B17C4 and 2,
A6B18C4 which inhibited EPO binding by 32 and 36%,
respectively. The decreased activity of compounds containing
B37-B43 indicates that the ester functionality has importance
for binding, although lysinol derivative 2, A6B45C4 maintains
modest activity. A variety of sizes are tolerated among the ester
(32) In the trimer series (library 3), eight group A amines (R1-NH2) and 10
group B amines (R2-NH2) identified as the most active subunits from the
dimer library screening were utilized, each combination linked through a
mixture of seven tricarboxylic acid linkers, resulting in 560 homotrimers
as 80 mixtures, seven compounds per mixture. The general library structure
and the subunits of the most active compound, 3, A2B6C11, are shown in
Figure 8a. For the tetramer series (library 4), three of the most active group
A subunits (R1-NH2) and 15 of the most active group B subunits (R2-
NH2) were utilized, each combination linked through sequential dimerization
reactions with mixtures of 8-10 dicarboxylic acids used in the final
couplings, resulting in 1596 compounds as 168 mixtures of 8-10. The
general library structure and the subunits of the two most active agents
identified from the binding assay (4, A8B10C12 and 4, A13B10C12) are
shown in Figure 8b. Structures of the complete set of subunits used in
libraries 3 and 4 may be found in Supporting Information. For details of
their synthesis and characterization, see ref 24.
9
548 J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002

Erythropoietin Mimetics
A R T I C L E S
Figure 6. Structures of the 77 amine subunits (B11-B87) containing modifications to the N-R-CBZ-lysine methyl ester (B10), present in the most active
library compounds. Each of these amines was combined with the A6 and C4 subunits and screened in the competitive binding assay. For details of the
synthesis of compounds 2, A6B11C4-A65B87C4, see ref 24.
binding at 50 µM concentration (Figure 8b). The choice of the
linking dicarboxylic acid was less important for the observed
activity, but the highest inhibition was detected with octade-
canedioic acid (C12). Thus, the trimer and tetramer agents, in
which three or four copies of the same functionalized imino-
diacetic acid binding subunits are present, bind the target
receptor, and because of their higher valency have a greater
chance of interacting with a second receptor molecule. The
screening of such higher order libraries, in which multiple copies
of identical binding subunits are present, has been successfully
adopted by others for the discovery of EPOr agonists.12
A second series of tetramer agents was prepared in which 10
dicarboxylic acids were used to link each of the eight most active
compounds of the optimized iminodiacetic acid diamide dimers
9
J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002 549

A R T I C L E S
Goldberg et al.
Figure 7. Competitive binding activity of compounds 2, A6B11C4-
A65B87C4. Each compound (50 µM) was incubated with 125I-EPO and
immobilized EPOr, as described in ref 28. Compounds were tested in
triplicate, and the bars represent the average % inhibition of 125I-EPO binding
to the receptor.
(library 2). These agents (library 5, Figure 9)29 were prepared
for testing in a functional activity assay (vide infra) where
interaction with two receptor molecules is necessary.
Figure 8. Structures and binding activity (% inhibition of EPO binding)
for the most active trimer and tetramer iminodiacetic acid diamide libraries.
(a) A total of 560 trimers were prepared as 80 mixtures of seven compounds,
utilizing the key R1-NH2 and R2-NH2 amine binding subunits identified
from the dimer library screening. The most active mixtures were decon-
voluted, and compound 3, A2B6C11 displayed the highest activity,
inhibiting EPO binding by 38% at 50 µM. (b) A total of 1596 tetramers
were prepared as 168 mixtures of 8-10 compounds, utilizing the key R1-
NH2 and R2-NH2 amine binding subunits identified from the dimer library
screening. The most active mixtures were deconvoluted, and compounds
4, A8B10C12 and 4, A13B10C12 displayed the highest activity, inhibiting
EPO binding by 52 and 47%, respectively, at 50 µM. Structures of the full
set of subunits used in libraries 3 and 4 may be found in Supporting
Information.
Isoindoline Libraries. A second dimeric library was also
developed to provide an alternative structural framework for
the incorporation of the protein-binding subunits. Here, the
cyclic anhydride chemistry was extended to the preparation of
a prototypical 1000-member library of dimeric isoindoline-5,6-
diamides. The key amine binding groups identified from the
initial iminodiacetic acid libraries were incorporated into the
new library. Thus, an alternative rigid cyclic anhydride template
(6, Figure 10) was utilized, from which the pendant binding
groups are displayed. The use of such rigid templates may be
especially important for receptor dimerization agonist libraries
considering recent evidence that suggests activation is achievable
only by a precise (re)orientation of the target receptor chains.15,17
1
liquid-liquid extractions and characterized by H NMR, IR,
and HRMS. In each case, the desired product was isolated in
g95% purity. Each monoamide was then divided into 10
portions and coupled (EDCI) to amines B1′-B10′ providing
100 individual diamides 9, A1′B1′-A10′B10′ in 87% average
yield after acid/base extractive workup. Representative samples
The N-BOC-isoindoline template 6 was generated in situ by
treatment of dicarboxylic acid 7 with EDCI33 and was reacted
in parallel with amines A1′-A10′ (Figure 11), providing 10
individual monoamides 8, A1′-A10′ in good to excellent yields
(average ) 88%, Scheme 1). The products were purified by
1
were analyzed (matrix characterization) by H NMR, IR, and
HRMS, indicating the presence of the desired diamide products
in excellent purity regardless of the reaction efficiencies.
(33) Boger, D. L.; Lee, J. K.; Goldberg, J.; Qing, J. J. Org. Chem. 2000, 65,
1467.
9
550 J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002

Erythropoietin Mimetics
A R T I C L E S
Figure 11. The structures of the R1-NH2 (A1′-A10′) and R2-NH2 (B1′-
B10′) amines used to functionalize template 6.
Scheme 1
Figure 9. Structures of tetramer library 5. The most active dimeric
iminodiacetic acid diamide compounds (2), including the linking isophthalic
moiety, were dimerized through dicarboxylic acid couplings. This series
was prepared for testing in functional activity screens, where an increased
number of protein-binding subunits would be available for interaction with
two receptor molecules. For the synthesis and characterization of library 5,
see ref 24.
Figure 10. Comparison of cyclic anhydride templates 1 and 6.
The dimerization of the N-BOC-protected diamides 9,
A1′B1′-A10′B10′ was then investigated to provide the C2-
symmetrical ligands. Following protocols developed for the
iminodiacetic acid diamide libraries,24 samples of 9 were treated
with 4 N HCl-dioxane to remove the BOC protecting group
and provide the secondary amine. However, preliminary analysis
indicated an impure mixture of reaction products after the HCl
treatment. For example, treatment of 10 with 4 N HCl-dioxane
(25 °C, 1 h), followed by evaporation of the solvent, provided
a residue consisting of the desired deprotected diamide 11 (50%)
together with a 1:1 mixture of phthalimide 12 and 4-fluorophen-
ethylamine hydrochloride, indicating that acid-catalyzed cy-
clization of the isoindoline-5,6-dicarboxamide competes with
N-BOC deprotection (Scheme 2).34 The use of alternative
conditions (solvents and temperatures) for the HCl deprotection
reaction resulted in similar product mixtures.
Consequently, milder reagents were investigated for the
removal of the BOC protecting group. Treatment of 10 with
formic acid (25 °C, 2 h) followed by evaporation of the reaction
solvent was found to cleanly provide 11, without any detectable
competitive imide formation. Effective, yet technically nonde-
manding conditions, which are practical for a large library
(34) Diamide 10 was prepared from dicarboxylic acid 7 by successive couplings
(EDCI) with 4-fluorophenethylamine.
9
J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002 551

A R T I C L E S
Goldberg et al.
Scheme 2
Scheme 3
synthesis, were developed to free base the resulting formic acid
salts. A solution of the evaporated crude reaction product in
1:1 MeOH-CH2Cl2 was treated with the OH- form of Am-
berlite IRA-400 ion-exchange resin.35 Subsequent filtration and
evaporation were found to remove all traces of formic acid,
providing the free secondary amine (Scheme 3).
The dimerization of the isoindoline diamides was first
investigated using 9, A1′B1′ as a representative substrate. The
N-BOC protecting group was removed by treatment with formic
acid and subjected to dimerization coupling (EDCI) with each
of the 10 dicarboxylic acids C1′-C10′ (Figure 12), providing
the dimerized isoindoline products 13, A1′B1′C1′-C10′. The
reactions were worked-up by dilution of the reaction solvent
(DMF) with 1:1 MeOH-CH2Cl2 and treatment with a mixture
of cation (Dowex 50WX8-200) and anion (Amberlite IRA-400,
OH- form) exchange resins. This procedure was found to
effectively remove not only unreacted starting materials, but
also EDCI and its reaction byproducts. Thus, simple filtration
and evaporation of solvent provided the 10 individual products
13, A1′B1′C1′-A1′B1′C10′, free of any impurities. The yields
of the isolated products were good to excellent (79-94%), and
each was characterized (1H NMR, IR, HRMS) establishing the
identity and purity (g95%) of the products.
Figure 12. Dimerization of the functionalized isoindoline template through
dicarboxylic acid couplings.
Scheme 4
Mixture couplings were investigated by subjecting 9, A1′B1′
to the same deprotection and coupling conditions utilizing an
equimolar mixture of dicarboxylic acids C1′-C10′. A reaction
stoichiometry in which an excess of amine (2.5 equiv) relative
to dicarboxylic acid (0.1 equiv of each dicarboxylic acid, 1.0
equiv total) was utilized to drive the reactions to completion
and ensure that differences in the individual coupling rates would
not affect the integrity of the product mixture. After the ion-
exchange resin workup, a 79% yield (based on the average
product MW) of 13, A1′B1′C1′-C10′ was obtained. This
mixture was analyzed by MS, and the parent ion of each
expected product was observed.
provided library 13, A1′B1′C1′-A10′B10′C10′ as 100 mixtures
of 10 compounds in an average of 72% yield (Table 1).
EPO-Dependent Cell Proliferation. The most active imi-
nodiacetic acid-derived agents in the binding assay (2-5), as
well as the entire isoindoline library (13, as mixtures of 10
compounds), were screened for functional activity by measuring
the ability to stimulate the growth of an EPO-dependent cell
line (UT-7/EPO)36 through EPOr binding and dimerization.
Mitogenic activity was determined by measuring cellular
3
radioactivity (CPM) due to incorporation of H-thymidine in
By utilizing the optimized reaction and workup conditions
outlined above, the full set of 100 isoindoline-5,6-diamides 9,
A1′B1′-A10′B10′ was dimerized through dicarboxylic acid
linkers C1′-C10′ (Figures 11 and 12, Scheme 4). The reactions
the cell line (UT-7) developed to grow only in the presence of
EPO. Primary screening of 2-5 was conducted at a single
compound concentration (10 µM). Many of the compounds were
(36) Komatsu, N.; Yamamoto, M.; Fujita, H.; Miwa, A.; Hatake, K.; Endo, T.;
Okano, H.; Katsube, T.; Fukumaki, Y.; Sassa, S.; Miura, Y. Blood 1993,
82, 456.
(35) This was prepared by treating Amberlite IRA-400 resin with 1 N aqueous
NaOH and washing with H2O, MeOH, and CH2Cl2.
9
552 J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002

Erythropoietin Mimetics
A R T I C L E S
Table 1. % Yields (Based on the Average Product Molecular
Weight) for the Final Coupling Step for Mixture Library 13
Table 3. Mitogenic Activity of Dimeric Isoindoline Library, 13 (10
µM Mixture Concentration) in an EPO-Dependent Cell Line (UT-7)a
13
B1′ B2′ B3′ B4′ B5′ B6′ B7′ B8′ B9′ B10′
13
B1′
-4 -23 -20 -26 -16 -44 -34 -42 -58 -31
41
B2′
B3′
B4′
B5′
B6′
B7′
B8′
B9′
B10′
A1′
A2′
A3′
A4′
A5′
A6′
A7′
A8′
A9′
A10′
100 86 84 98 70 99 93 94 81 100
79 79 81 84 65 72 88 53 88 81
69 75 64 73 56 40 65 70 76 79
71 51 89 70 45 56 54 26 82 66
65 83 61 72 53 35 58 42 61 79
82 74 76 77 53 64 80 54 76 86
82 74 82 64 48 74 84 64 95 74
86 80 80 68 60 65 86 57 82 98
99 100 84 86 57 60 79 76 86 86
59 77 75 76 57 39 72 42 45 79
A1′
A2′
A3′
A4′
A5′
A6′
A7′
A8′
A9′
-1
-6 -16 -18 -23
16 -42 -14 -44
53 -11
44 -18
33 -14
18
16
-1 -18
-8 -14
70
11
12
2
10 -32 -32
14 -33 -52
10 -10 -30
-32 -32 -42
21 -14
24
4
-6
2
-6
9
-24
-50 -27 -18
16
15
15 -16
-3
-6
29
1
7
8
-4
18 -10
-5 -27
10 -14 -12 -32
3
4
-8
-2
-2
16 -17 -14
19
2
-26 -14
-1 -68
A10′ 184
16
5
-50 -30
51 -26
average yield ) 72.2%
a Data reported are the average % increases in proliferation, relative to
control, as measured by 3H-thymidine incorporation. Each A′B′ combination
is linked by a mixture of ten linking dicarboxylic acids. Ten of the most
active mixtures (italic) were deconvoluted by resynthesis of the individual
compounds.
Table 2. Mitogenic Activity of Iminodiacetic Acid Library
Compounds (10 µM) in an EPO-Dependent Cell Line (UT-7)a
% increase in
proliferation
% increase in
proliferation
compound no.
compound no.
Table 4. Mitogenic Activity of the Deconvoluted Mixtures (10 µM)
of Library 13a
2, A1B10C1
2, A2B10C1
2, A5B10C4
2, A5B10C8
2, A6B10C4
2, A6B17C4
2, A7B10C1
2, A7B10C4
2, A12B10C4
2, A13B10C4
2, A17B10C4
4, A8B10C12
4, A13B10C12
5, A1B10C2
5, A12B10C2
5, A12B10C4
5, A12B10C14
5, A12B10C16
5, A12B10C18
5, A2B10C2
5, A2B10C4
5, A2B10C18
5, A5B10C19
88
54
61
33
39
37
95
33
34
42
42
34
40
88
24
16
24
6
5, A5B10C4
5, A5B10C5
5, A5B10C14
5, A5B10C16
5, A5B10C18
5, A6B10C2
5, A6B10C19
5, A6B10C4
5, A6B10C16
5, A6B10C18
5, A6B18C2
5, A6B18C19
5, A6B18C4
5, A6B18C5
5, A6B18C13
5, A6B18C15
5, A6B18C17
5, A6B18C18
5, A7B10C4
5, A8B10C2
5, A8B10C19
5, A8B10C4
5, A8B10C16
69
65
64
74
44
35
35
28
23
38
46
26
41
-3
-5
29
19
6
C1′
C2′
C3′
12
14 -24 -26
C4′
C5′
C6′
12
12 -75 -64
C7′
C8′
C9′
C10′
A2′B1′
5
15
2
4
10
2
23 -10
16 -2
A3′B1′ -2
2
-24
21
A4′B1′
A5′B1′
0
12
2
16
0
-4 -21 -11 -18 -20 -18 -23
22
-15 -5 -54 -38 -41
-58 -48 -48
1
0
-2
22 -7
0
-4
-6
2
A10′B1′ -8 -16
A2′B7′ 26 -9
A3′B7′ -8 -51 -19 -96 -22 -66 -97 -86
10 -14
32 -1
A10′B7′ -8 -34
0
8
24
26 -25
7
A4′B7′
A5′B7′
6 -46 -18 -72 -68 -58 -36 -30
45 -12
19 -30
16 -51 -44 -6
-58 -56 -34
24 -17
7
6
0
a Data reported are the average % increases in proliferation, relative to
control, as measured by H-thymidine incorporation.
3
composition on functional activity. The mixture screening results
(Table 3) indicate that many of these C2-symmetrical compounds
were able to support proliferation of the EPO-dependent UT-7
cells. The most active mixtures of library 13 contained either
the 4-fluoro- or the 4-chlorophenethylamine (B1′ or B7′) binding
groups.
29
35
-14
18
40
33
29
11
30
33
a Data reported are the average % increases in proliferation, relative to
DMSO control, as measured by 3H-thymidine incorporation. Selected
compounds (italic) were retested over a concentration range; see Figure 13
and Table 5.
Deconvolution of the 10 most active mixtures containing the
B1′ and B7′ subunits was conducted by resynthesis of the
corresponding 100 individual compounds. These deconvoluted
samples were screened in the cell proliferation assay (10 µM,
1% DMSO), and the results are presented in Table 4. Many of
these agents were able to stimulate low levels of cell prolifera-
tion,37 and the highest activity was observed for compounds
containing the N-R-CBZ-lysine methyl ester (A2′) or the
homologous N-R-CBZ-ornithine methyl ester (A5′). The optimal
linking dicarboxylic acid (C1′-C10′) was dependent on the
particular combination of the R1-NH2 and R2-NH2 substitu-
tions, but, in general, the linear, more flexible linkers (i.e., C1′-
C4′, C9′) led to greater cellular proliferation. The high levels
of activity observed with the A2′B1′, A2′B7′, A5′B1′, and
A5′B7′ combinations are consistent with the most active
structures of the iminodiacetic acid diamide screening results,
illustrating the significance of the substituted lysine or ornithine
binding groups. Many of the same R1-NH2 and R2-NH2
binding groups (especially the N-R-CBZ-lysine methyl ester)
able to induce cell proliferation at this concentration, and
representative results are presented in Table 2. A standard curve
was also established for EPO, and a maximum response of 812%
was observed. The greatest increase in proliferation was
observed with compound 2, A7B10C1, which contains the key
N-R-CBZ-lysine methyl ester (B10) binding group in both the
R1-NH2 and R2-NH2 positions. Compounds from the tetramer
library 5, which contain four copies of the same functionalized
iminodiacetic acid diamides present in the dimer series (2),
induced a similar increase in proliferation, indicating that the
additional binding group copies did not further increase the
observed stimulated cell growth. Thus, the dimeric compounds,
which contain two sets of binding groups, were found to be
sufficient for functional activity.
Screening results from mixture library 13 (10 µM mixture
concentration, 1% DMSO) are presented in Table 3. Because
the R1-NH2 and R2-NH2 binding groups previously identified
from the screening of the iminodiacetic acid libraries were
incorporated into 13, the new library members were expected
to similarly bind EPOr. The extended, more rigid, nature of the
isoindoline template was included to probe the effects of linker
(37) The negative values observed for many of the compounds can be attributed
to the poor solubility of the dimeric isoindoline products or to possible
cytotoxicity of these agents at 10 µM.
9
J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002 553

A R T I C L E S
Goldberg et al.
Table 5. Comparison of the Maximal Mitogenic Response of
Selected Agents with EMP1, the Merck Agonist, and EPO
maximal response
(% increase in cell proliferation)
concentration
(µM)
compound
2, A1B10C1
2, A2B10C1
2, A5B10C4
2, A5B10C8
2, A6B10C4
2, A6B17C4
2, A7B10C1
2, A7B10C4
2, A12B10C4
2, A13B10C4
Merck agonist
EMP1
155
150
110
133
30
110
175
95
60
120
33
670
812
10
10a
10
25
10a
10a
25
5
10a
10a
0.1
0.5-10
0.01
EPO
a Highest concentration tested.
line that lacks the EPOr, FDC-P1.39 Across a full concentration
range (0.01-100 µM), no proliferation enhancement was
observed ((15% of control). The unique concentration depen-
dence of the cellular proliferation of the EPO-dependent UT-7
cell line and the lack of effect on the FDC-P1 cell line lacking
an EPO receptor, combined with the competitive binding activity
in the protein binding assay, suggest that the compounds are
able to productively interact with EPOr and function as weak
partial agonists. However, we cannot rule out the potential that
the agents act at alternative, downstream intracellular sites not
yet recognized. Nonetheless, this identification of small mol-
ecules that appear to act by promoting specific receptor
dimerization and activation represents a remarkable achieve-
ment.15 Although the discovery of antagonists of biological
activity functioning through the disruption of protein-protein
interactions has been achieved with increasing numbers and
levels of success, the discovery of agonists functioning by
promoting productive protein-protein interactions is extremely
rare. The extent of stimulated cell growth is moderate, roughly
15-20% of the maximum response observed with EPO, but
significant in comparison to the two previously reported EPOr
agonists (Table 5). Notably, they are substantially more effica-
cious than the Merck agonist (15-20% versus 4% of the activity
of EPO), and 2, A1B10C1 is equipotent with the Merck agonist
at its maximal response dose of 100 nM and lower. Compounds
2, A1B10C1; 2, A5B10C4; 2, A5B10C8; and 2, A7B10C1
(average MW ) 1246) are considerably smaller than EMP18
(MW ca. 2100), which itself functions only as a dimer (Figures
1 and 14, MW ca. 4200), and the recently reported Merck non-
peptide EPOr agonist12 (Figure 14, MW ca. 6400). They are
synthesized in four steps from commercially available starting
materials, allowing for easy derivatization and optimization. Key
binding elements have been identified, potentially allowing for
further reductions in the identified lead structures. The identi-
fication of these smaller, easily modified agents by a generaliz-
able strategy represents a major step toward the development
of a new generation of therapeutic agents.
Figure 13. Concentration dependence of the stimulated increase in cell
proliferation. UT-7 cells were incubated with varying concentrations (10-8
-
10-4 M) of 2, A1B10C1; 2, A5B10C4; 2, A5B10C8; 2, A7B10C1; or the
Merck agonist (10-9-10-5 M) in 1% DMSO. The % increase in prolifera-
tion, relative to DMSO control, was determined by measuring the uptake
3
of H-thymidine.
were found to lead to similar activity when combined through
linkers of different length and rigidity.
Dose-response curves were generated for the most active
agents over a concentration range of 10-8-10-4 M. Data for
compounds 2, A1B10C1; 2, A5B10C4; 2, A5B10C8; and 2,
A7B10C1 are shown in Figure 13. Each of these compounds
promotes a concentration-dependent increase in cell prolifera-
tion, beginning at submicromolar concentrations, that leads to
a maximal 100-180% increase in stimulated growth at con-
centrations between 5 and 25 µM (Table 5). At higher
concentrations, the observed proliferation begins to decrease
presumably due to the saturation of EPOr with ligand, resulting
in predominantly inactive 1:1 receptor-agent complexes, which
has been observed with EPO and is expected of EPOr agonists.38
To assess whether this effect was related to EPOr activation,
the compounds listed in Table 5 were examined against a cell
The generation and screening of diverse libraries based on
structurally unbiased or biased templates such as 1 or 6 using
solution phase methods are straightforward and efficient ap-
proaches to the discovery of receptor binders which can
subsequently be dimerized or oligomerized, allowing interactions
(38) Elliott, S.; Lorenzini, T.; Yanagihara, D.; Chang, D.; Elliott, G. J. Biol.
Chem. 1996, 271, 24691. Schneider, H.; Chaovapong, W.; Matthews, D.
J.; Karkaria, C.; Cass, R. T.; Zhan, H.; Boyle, M.; Lorenzini, T.; Elliott, S.
G.; Giebel, L. B. Blood 1997, 89, 473.
(39) Dexter, T. M.; Garland, J.; Scott, D.; Scolnick, E.; Metcalf, D. J. Exp.
Med. 1980, 152, 1036.
9
554 J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002

Erythropoietin Mimetics
A R T I C L E S
information, expediting the lead generation process. The same
strategy, however, has already been shown effective for the
discovery of selective protein-protein antagonists in the absence
of any specific structural information, utilizing randomized
subunits known to have general binding interactions with protein
surfaces.40
Conclusions
Through the screening of dimeric and higher order iminodi-
acetic acid diamide and isoindoline-5,6-dicarboxamide libraries,
compounds were identified that function as competitors of EPO
binding to its receptor. Many of the most active binders were
further found to promote concentration-dependent proliferation
of an EPO-dependent cell line, indicating their functional activity
as partial agonists of EPO. This functional activity was observed
with as few as two copies of the protein binding subunits, and
some variability in the choice of the linker moiety was tolerated.
They represent the third class of EPO partial agonists disclosed,
but the first examples of relatively small compounds that
function as EPO mimetics. Although their maximal functional
activity is modest (15-20% that of EPO) and less than that
reported for EMP1, they are substantially more efficacious than
the Merck agonist. Their relative small size, the generalizable
strategy for their discovery, and ease of synthesis makes them
superb leads amenable to further optimization and characteriza-
tion, and such studies are in progress.
Acknowledgment. This work was supported by the National
Institutes of Health (CA78045) and the award of a National
Defense Science and Engineering Graduate Fellowship (J.G.).
The authors would like to thank The R. W. Johnson Pharma-
ceutical Research Institute for supporting the initial stages of
this work, conducting the protein binding assay, and supplying
samples of EMP1, erythropoietin, and the UT-7 cell line.
Supporting Information Available: Structures of the full set
of subunits for libraries 3 and 4, full details of the synthesis of
library 13, full characterization of 8, A1′-A10′, matrix char-
acterization of 9, preparation and characterization of 13,
A1′B1′C1′-10′ as individual compounds and the full mixture,
and details of the cell proliferation assays (PDF). This material
is available free of charge via the Internet at http://pubs.acs.org.
Figure 14. Structures of the Merck non-peptide agonist12 and EMP1.8
with multiple protein binding sites leading to novel receptor
agonists. This methodology can be readily extended to other
receptor systems, including those activated by heterodimeriza-
tion versus homodimerization, providing an even more general
method for identifying novel small molecules which modulate
protein-protein interactions. In the present example, the library
subunits were chosen on the basis of the available structural
JA0118789
(40) Boger, D. L.; Goldberg, J.; Satoh, S.; Cohen, S. B.; Vogt, P. K. HelV. Chim.
Acta 2000, 83, 1825. Boger, D. L.; Goldberg, J.; Silletti, S.; Kessler, T.;
Cheresh, D. A. J. Am. Chem. Soc. 2001, 123, 1280. Silletti, S.; Kessler,
T.; Goldberg, J.; Boger, D. L.; Cheresh, D. A. Proc. Natl. Acad. Sci. U.S.A.
2001, 98, 119.
9
J. AM. CHEM. SOC. VOL. 124, NO. 4, 2002 555