A combinatorial method for solution-phase synthesis of labeled bivalent β-turn mimics

Article abstract of DOI:10.1021/ja074717z

Piperidine-functionalized, 1,4-disubstituted-1,2,3-triazoles of generic structure 1 were conceived as "minimalist" mimics of peptidic β-turn structures. Key features of these molecules include (i) the possibility of incorporating amino acid side chains co

Full text of DOI:10.1021/ja074717z

Published on Web 12/19/2007
A Combinatorial Method for Solution-Phase Synthesis of
Labeled Bivalent â-Turn Mimics
Yu Angell,† Dianjun Chen,† Fouad Brahimi,‡ H. Uri Saragovi,‡ and Kevin Burgess*,†
Department of Chemistry, Texas A & M UniVersity, Box 30012, College Station, Texas 77841,
and McGill UniVersity, Pharmacology and Therapeutics, Lady DaVis Institute-Jewish
General Hospital Montreal, Quebec, Canada
Received June 27, 2007; E-mail: burgess@tamu.edu
Abstract: Piperidine-functionalized, 1,4-disubstituted-1,2,3-triazoles of generic structure 1 were conceived
as “minimalist” mimics of peptidic â-turn structures. Key features of these molecules include (i) the possibility
of incorporating amino acid side chains corresponding to many of the protein amino acids; (ii) a close
correspondence of separations of these side chains to i + 1 to i + 2 residues in turns; (iii) facile adjustment
of the side-chain vectors on docking while only influencing two critical degrees of freedom; and (iv) some
electrostatic polarity. Fifteen monomers of this type were made via copper-mediated cycloaddition reactions.
Solution-phase methodologies were devised to assemble these monomers into bivalent compounds in
high purity states (typically >85%) so that they could be used in first-pass biological assays without further
purification. The skeleton for forming these bivalent compounds is triazine-based. There is a third site
which allowed for introduction of a fluorescent label (library of compounds 2) or an alkyne-functionalized
triethylene glycol chain (library of compounds 3) included to promote water-solubility and to allow
incorporation of probes via copper-mediated cycloaddition reactions. In the event, two 135-membered
libraries were prepared, one consisting of compounds 2 and the other of 3. No protecting groups or coupling
agents were required; these attributes of the method were important to allow most of the products to be
obtained in over 85% purities. The fluorescein-tagged library of compounds 2 was screened in a
fluorescence-activated cell sorting (FACS) assay using cells transfected to overexpress one of the following
neurotrophin receptors: TrkA, TrkC, and p75. Preliminary findings indicate four compounds 2gm, 2gn,
2gi, and 2gj bound the TrkA receptor selectively; all of these contain a threonine-lysine turn mimic. Thus,
a pharmacological probe for the TrkA receptor has been developed.
in the field of protein-protein interactions.13-15
Introduction
Combinatorial methods are ideally suited for the syntheses
Many proteins interact via two or more contacts that account
for most of the binding energy, that is, hot-spots.1 Molecules
that bring together two pharmacophores to mimic hot spots of
one protein component2 may therefore be referred to as
“bivalent”.3-6 The prevalence of such bivalent molecules in the
literature underscores their biological importance,7-12 especially
of bivalent products because of the numerical advantages of
combining libraries with themselves.16 If a library of n monova-
lent compounds were assembled into every different possible
bivalent compound, then n(n + 1)/2 products would result. Thus,
there is a “combinatorial advantage”17 to this type of strategy,
that may be defined as
† Texas A & M University.
‡ Lady Davis Institute-Jewish General Hospital Montreal.
(1) Fuh, G.; Li, B.; Crowley, C.; Cunningham, B.; Wells, J. A. J. Biol. Chem.
1998, 273, 11197-11204.
(2) Boger, D. L.; Desharnais, J.; Capps, K. Angew. Chem., Int. Ed. 2003, 42,
4138-4176.
The combinatorial advantage is small for small values of n but
increases rapidly as this number of monovalent compounds
increases (Figure 1a).
Three main obstacles must be overcome to prepare libraries
of discrete bivalent molecules from the corresponding monova-
(3) Zhang, A.; Liu, Z.; Kan, Y. Curr. Top. Med. Chem. 2007, 7, 343-345.
(4) Decker, M.; Lehmann, J. Curr. Top. Med. Chem. 2007, 7, 347-353.
(5) Peng, X.; Neumeyer, J. L. Curr. Top. Med. Chem. 2007, 7, 363-373.
(6) Haviv, H.; Wong, D. M.; Silman, I.; Sussman, J. L. Curr. Top. Med. Chem.
2007, 7, 375-387.
(7) Wong, D. M.; Greenblatt, H. M.; Dvir, H.; Carlier, P. R.; Han, Y.-F.; Pang,
Y.-P.; Silman, I.; Sussman, J. L. J. Am. Chem. Soc. 2003, 125, 363-373.
(8) Wittmann, V.; Datta, A. K.; Koeller, K. M.; Wong, C.-H. Chem.sEur. J.
2000, 6, 162-171.
(13) Klemm, J. D.; Schreiber, S. L.; Crabtree, G. R. Annu. ReV. Immunol. 1998,
16, 569-592.
(9) Tamiz, A. P.; Zhang, J.; Zhang, M.; Wang, C. Z.; Johnson, K. M.;
Kozikowski, A. P. J. Am. Chem. Soc. 2000, 122, 5393-5394.
(10) Profit, A. A.; Lee, T. R.; Lawrence, D. S. J. Am. Chem. Soc. 1999, 121,
280-283.
(14) Ho, S. N.; Biggar, S. R.; Spencer, D. M.; Schreiber, S. L.; Crabtree, G. R.
Nature 1996, 382, 822-826.
(15) Belshaw, P. J.; Ho, S. N.; Crabtree, G. R.; Schreiber, S. L. Proc. Natl.
Acad. Sci. 1996, 93, 4604-4607.
(11) Rao, J.; Whitesides, G. M. J. Am. Chem. Soc. 1997, 119, 10286-10290.
(12) Halazy, S.; Perez, M.; Fourrier, C.; Pallard, I.; Pauwels, P. J.; Palmier, C.;
John, G. W.; Valentin, J.-P.; Bonnafous, R.; Martinez, J. J. Med. Chem.
1996, 39, 4920-4927.
(16) 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, 1347-1378.
(17) Reyes, S. J.; Burgess, K. Chem. Soc. ReV. 2006, 35, 416-423.
9
556
J. AM. CHEM. SOC. 2008, 130, 556-565
10.1021/ja074717z CCC: $40.75 © 2008 American Chemical Society

Heterobivalent Selectivity and Bivalent-Turn Mimics
A R T I C L E S
Figure 2. (a) The divergent strategy featuring solid-supported peptidomi-
metics used in preliminary studies, and (b) the types of monovalent starting
materials used.
lent ones. First, a method must be available to combine two
different building blocks to form “heterobivalent” products
(Figure 1 parts b and c); this is hard to achieve selectively if
both building blocks are similar, especially if they also have
reactive, unprotected, side chains. Second, it is desirable to form
the bivalent products in high states of purity, preferably so that
they do not need to be chromatographed before a first-pass
biological assay. This means that any coupling agents that are
used should give only residues that are readily removed, and
that removal of protecting groups after the dimerization step is
unlikely to be satisfactory. Ideally, the library would be formed
by pipetting in aliquots of each monovalent component into each
well, without coupling agents and protecting group chemistry,
Figure 1. (a) The number of isolated bivalent molecules rises rapidly with
the number of monovalent starting materials; (b) monovalent starting
materials each having identical coupling groups; (c) formation of homo-
bivalent compounds is easy, but selectivity for heterobivalent ones is hard;
(d) perfect heterobivalent selectivity allows assembly of all permutations
of the monomers; and (e) given the same number of monomers, fewer
compounds can be made by combining two libraries with different coupling
groups.
9
J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008 557

A R T I C L E S
Angell et al.
and have them covalently assemble to give highly pure products.
Third, and particularly important for medicinal chemistry
applications, the coupling chemistry should not interfere with
pharmacophore groups on the monovalent compounds. If these
criteria can be satisfied then construction of a library according
to Figure 1d is extremely efficient.
labels could be added to derivatives for binding studies
performed after hits were identified.
It is relatively easy to make heterobivalent products from two
libraries wherein each compound contains a functional group
that reacts exclusively with the other (Figure 1e).18,19 However,
logistically this is a less effective strategy, and it tends to be
more work in practice. The strategy is logistically less effective
because members of one library cannot be reacted with each
other, so less bivalent molecules can be made from the same
number of monomers. Mathematically speaking, all combina-
tions are allowed but not all permutations. It is less practical
because two different libraries must be made (Figure 1e). It tends
to be more work than combining two similar libraries because
two different ones must be prepared.
One approach toward achieving selectivity for heterobivalent
products is to generate a library of a set of monovalent molecules
on a solid phase, divide each sample into two, then treat each
portion differently. One part might be cleaved into solution,
while the other would be chemically transformed into entities
that would react with the samples liberated into solution (Figure
2a). In preliminary work, this strategy was followed in our group
using peptidomimetics like the two shown in Figure 2b.20 Those
compounds were designed to mimic â-turns in proteins, both
conformationally, and because relevant amino acid side chains
(e.g., Arg, Lys, Glu) could be employed, not just the unreactive
ones (e.g., Glu, Ala, Val, Ile).20
A total of 78 bivalent compounds prepared via the method
shown in Figure 2a were obtained in less than 3 mg amounts.20
This library was restricted to 78 compounds because it was
difficult to obtain the monovalent compounds in sufficient
quantities to make more bivalent products. Further, resynthesis
of any bivalent molecule showing activity was time-consuming
and scale-up was difficult. Both of these problems can be
attributed to solid-phase syntheses; such approaches are expen-
sive and inconvenient for preparation of more than 10 mg of
material, especially if HPLC purification is required as the final
purification. Thus next steps in the project were clear. To obtain
larger libraries and to facilitate resynthesis it was necessary to
use (i) scalable, solution-phase syntheses of monomers and (ii)
solution-phase assembly into bivalent compounds.
This paper describes the design and syntheses of a series of
â-turn mimics 1. These were made via scalable, solution-phase
syntheses, typically in several gram amounts. The solid-phase
approach outlined in Figure 2a was adapted so that libraries of
bivalent molecules 2 and 3 could be assembled from these turn
mimics in solution. In one library, featuring the bivalent
compounds 2, all the constituent members are fluorescein labeled
to facilitate direct binding studies. In the other, each compound
3 was nonfluorescent so that the bivalent compounds could be
screened in assays designed for such substrates; however, each
constituent of that library contained a terminal alkyne so that
Results and Discussion
Design of the Monovalent-Turn Mimics. Our design of the
â-turn mimics 1 was based on two considerations regarding the
relative importance of main-chain amides and side-chain func-
tionalities. First, studies of protein complexes crystallographi-
cally have shown that main-chain carbonyl groups are involved
in only about 11% of protein-protein interface regions, whereas
side chains contribute about 80%.21 Thus we concluded it is
far more important to have amino acid side chains represented
in the mimics, than main-chain amides. This is consistent with
designs by, for instance, Hirschmann and Smith.22-24 Second,
syntheses of the mimics must allow for incorporation of the
side chains that occur most frequently at hot-spots, that is, those
of Trp, Arg, Tyr, Lys, Glu, Ser, Asn, Leu;21 syntheses of â-turn
mimics that do not allow this are completely unsatisfactory for
medicinal chemistry. Unfortunately, it is actually quite difficult
to devise syntheses of mimics that do satisfy this criterion
because of the diversity of amino acid side-chain functionalities
involved.
Conformational and structural issues also factored into our
design considerations for turn mimics 1. Using the arbitrary
(21) Conte, L. L.; Chothia, C.; Janin, J. J. Mol. Biol. 1999, 285, 2177-2198.
(22) Hirschmann, R.; Nicolaou, K. C.; Pietranico, S.; Salvino, J.; Leahy, E. M.;
Sprengeler, P. A.; Furst, G.; Smith, A. B., III. J. Am. Chem. Soc. 1992,
114, 9217-9218.
(23) Hirschmann, R.; Sprengeler, P. A.; Kawasaki, T.; Leahy, J. W.; Shakespeare,
W. C.; Amos, B.; Smith, I. J. Am. Chem. Soc. 1992, 114, 9699-9701.
(24) Hirschmann, R.; Nicolaou, K. C.; Pietranico, S.; Leahy, E. M.; Salvino, J.;
Arison, B.; Cichy, M. A.; Spoors, P. G.; Shakespeare, W. C.; Sprengeler,
P. A.; Hamley, P.; Smith, A. B., III; Reisine, T.; Raynor, K.; Maechler,
L.; Donaldson, C.; Vale, W.; Freidinger, R. M.; Cascieri, M. R.; Strader,
C. D. J. Am. Chem. Soc. 1993, 115, 12550-12568.
(18) Maly, D. J.; Choong, I. C.; Ellman, J. A. Proc. Natl. Acad. Sci. 2000, 97,
2419-2424.
(19) Su, S.; Acquilano, D. E.; Arumugasamy, J.; Beeler, A. B.; Eastwood, E.
L.; Giguere, J. R.; Lan, P.; Lei, X.; Min, G. K.; Schaus, S. E.; Porco, J. A.
Org. Lett. 2005, 7, 2751-2754.
(20) Pattarawarapan, M.; Reyes, S.; Xia, Z.; Zaccaro, M. C.; Saragovi, H. U.;
Burgess, K. J. Med. Chem. 2003, 46, 3565-3567.
9
558 J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008

Heterobivalent Selectivity and Bivalent-Turn Mimics
A R T I C L E S
Scheme 1. Two Methods for Preparing Monovalent Mimics 1a
a Generally the least expensive and most direct route used is route a
featuring unprotected amino acid starting materials (except for Lys which
was side-chain protected). Route b uses protected amino acids.
that are close to those of â-turns. Of the several different types
of â-turn conformations, type-1 is the most common (as in
Figure 3a).27 However, in solution, most â-turns probably are
somewhat flexible; good mimics will reflect that characteristic
without being so flexible that the entropic cost of binding at a
protein interface is prohibitively high. Of course, the turn mimic
1 does not always reside in the conformation indicated in Figure
1b, and this is not, in fact, the low-energy conformation.
However, the conformational state in Figure 1b is readily
accessible at room temperature since the enthalpic barrier to
rotation about C2-N3 dihedral is low as shown in Figure 3d.
The transition state (Figure 3d, iii) and minima (Figure 3d, i
and ii) were fully optimized at the B3LYP/6-311++G(d,p) level
of theory, and each stationary point was verified with a
frequency calculation (see Supporting Information). The lowest
energy conformer that was identified, i, has a C2-N3 dihedral
angle of 50° which is 140° away from the targeted turn
conformation, but the second lowest energy conformer, ii, was
only 0.85 kcal‚mol-1 higher in energy and had a C2-N3 dihedral
angle of -57° which matched the target turn conformation much
closer (deviation ) 33°). The calculations showed that the
Figure 3. (a) The key distance of Câ-separations of the i + 1 to i + 2
residues of a type 1 â-turn and of the monovalent turn mimics featured
here; (b) an overlay of the mimics onto a type 1 â-turn; (c) a comparison
of their electrostatic charge separations for these two entities; and (d) B3LYP
calculated pathway for rotation around the C2-N3 dihedral for 1 showing
the minimum energy conformation i, the next highest energy conformation
ii, and the transition state iii that separates them.
numbering shown in Figure 3a, the bond vectors 1-2, 2-3,
3-4, 4-5, and 5-6 populate relatively well-defined regions
of space, whereas rotation about the 1-2 and 5-6 bonds in
â-turns are relatively free.25,26 To draw an analogy, conforma-
tions of â-turns vaguely resemble “molecular aeroplanes” with
the R1 and R2 substituents spinning like propellers on a
reasonably rigid framework. Thus we reasoned that good mimics
should be able to access conformations with C1-C6 separations
(25) Garland, S. L.; Dean, P. M. J. Comput.-Aided Mol. Des. 1999, 13, 469-
483.
(26) Garland, S. L.; Dean, P. M. J. Comput.-Aided Mol. Des. 1999, 13, 485-
(27) Ball, J. B.; Hughes, R. A.; Alewood, P. F.; Andrews, P. R. Tetrahedron
1993, 49, 3467-3478.
498.
9
J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008 559

A R T I C L E S
Angell et al.
Scheme 2. Solution-Phase Method for the Preparation of the
Table 1. Peptidomimetics 5 Prepared via Method A
Library of Compounds 2
Syntheses of the Monovalent-Turn Mimics. Two methods
were used to prepare the monovalent building blocks 1. The
most economical and direct route (Scheme 1a) begins with
amino acids that were converted to the corresponding azides
via known azo-transfer reactions,28 coupled with Boc-piperazine,
and combined with appropriately substituted alkynes via copper-
mediated cycloadditon processes.29-31 Only the final products
5 were purified via flash chromatography; the intermediates were
isolated via aqueous workup (crude yields of 4 are shown) and
used in the next step without further purification. This route
was successful for all the amino acid/alkyne combinations tested,
that is, those shown in Table 1. Scheme 1a uses R1′ and R2′ to
denote protected side chains (and R1 and R2 to indicate
deprotected ones). In fact, the only side chain of the amino acid
starting materials (R1 shown in Table 1) that had to be protected
was the amino group of Lys (Boc-protected to mask it during
the azo-transfer step).
a From intermediate 4 after flash chromatography.
molecule only needs to surmount an enthalpic barrier of 0.99
kcal‚mol-1 to reach the target turn conformation; such barriers
are readily overcome at room temperature. In summary, the turn
conformation of peptidomimetics 1 is energetically accessible,
and it represents a low-energy conformation that is only slightly
less stable than the minima. The energy necessary to reach this
conformation is available at ambient temperatures, and the slight
increase of energy over the lowest minimum is likely to be easily
compensated for via induced fit binding to a protein surface.
All the considerations outlined above are satisfied for the
monovalent mimics 1. The separations between Câ atoms that
correspond to the i + 1 and i + 2 amino acid side chains (i.e.,
C1 and C6) in type-1 â-turns and in mimics 1 are close in an
energetically accessible conformation of 1. â-Turns have two
bonds that are essentially freely rotating (1-2 and 5-6). The
core of molecules 1 has three rotatable (1-2, 2-3, and 5-6),
and only one of these affects the C1-to-C6 distance (the 2-3
rotation). Consequently, compounds 1 can access appropriate
conformations to mimic the C1-to-C6 spacing in â-turns simply
via rotation around the 2-3 bond (Figure 3 panels a and b).
The polar triazole part of the mimics also introduces electrostatic
polarity that can also resemble turn conformations. Thus, we
were satisfied by the match of the mimics to the target secondary
structure.
Method B in Scheme 1 is a more conservative approach that
begins with Cbz-protected amino acids, and requires an ad-
ditional deprotection step. This was our original approach, but
as the work evolved, it was found that this route had no
advantages over method A for the amino acids shown in Table
1. However, method B is shown here just in case the work is
later expanded to other amino acids for which it might have
some merit.
(28) Alper, P. B.; Hung, S.-C.; Wong, C.-H. Tetrahedron Lett. 1996, 37, 6029-
6032.
(29) Tornoe, C. W.; Christensen, C.; Meldal, M. J. Org. Chem. 2002, 67, 3057-
3064.
(30) Rostovtsev, V. V.; Green, L. G.; Fokin, V. V.; Sharpless, K. B. Angew.
Chem. Int. Ed. 2002, 41, 2596-2599.
(31) Bock, V. D.; Hiemstra, H.; van Maarseveen, J. H. Eur. J. Org. Chem. 2006,
51-68.
9
560 J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008

Heterobivalent Selectivity and Bivalent-Turn Mimics
A R T I C L E S
Figure 4. Purities of the library of compounds 2 where the detection method was (a) UV set at 254 nm, and (b) SEDEX detection. The term “cap” is used
for morpholine.
Adaptation of the Dimerization Method to Solution Phase.
After considerable optimization studies, the solution-phase
method for assembly of the bivalent compounds shown in
Scheme 2 was developed. The protected compounds 5 were
unmasked to give mimics 1 and then reacted with a dichloro-
triazine derivative of aminofluorescein (i.e., DTAF) to give the
electrophiles 7. These were then reacted with second aliquots
of the mimics 1, in all permutations, to give the library of
bivalent mimics 2. Morpholine was also used in the second
couplings. This resulted in syntheses of the specific mimics
2a-o that represent controls for testing which contain only one
mimic, triazine with the fluorescein label, and the morpholine
part which represents the linker.
Some attributes of the method in Scheme 2 are as follows.
Our goal was to prepare compounds 2 in over 85% purity, so
the threshold of acceptable purity applied to the intermediates
7 had to be above this. The crude intermediates (chlorotriazines
7) were assayed by reverse phase HPLC (UV and SEDEX
detection) and found to be above 85% purity, so no chromato-
graphic isolation was necessary. In the first coupling, THF
facilitated the addition of just one monomer, giving monochloro-
intermediates 7 in satisfactory purities for use in the next stage
without further purification. Removal of THF after the synthesis
of the intermediates 7 is convenient because it is volatile relative
to some other solvents. It seems that DMSO facilitates faster
coupling of nucleophiles to triazines than THF. Indeed, we found
9
J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008 561

A R T I C L E S
Angell et al.
Scheme 3. Solution-Phase Method for the Preparation of the
Library of Compounds 3
Choice of the least polar solvent (THF) for the first coupling
reaction, and the most polar one (DMSO) for the second
coupling reaction is entirely logical.32-35 Nucleophilic aromatic
substitution reactions involve charged intermediates. It is
desirable to destabilize these intermediates in the first coupling
step to prevent two couplings occurring to the first peptidomi-
metic. However, in a polar solvent it is desirable to send the
second coupling to completion.
Application of the Solution-Phase Procedure To Make a
Library of Fluorescently Labeled Bivalent Mimics. The
method outlined in Scheme 2 was applied to the synthesis of a
library of 135 fluorescently labeled bivalent mimics. Thus, the
monovalent compounds 5 were deprotected with TFA; no
scavenger was required. The resulting unmasked intermediates
1 dissolved in THF were reacted with equimolar amounts of
DTAF and excess K2CO3 for 4 h to give the intermediates 7.
The THF was removed, then these intermediates were dissolved
in DMSO, split into small portions, then coupled with another
equivalent of the deprotected monomers 1 to give the bivalent
compounds 2. Alternatively, to generate a set of control
compounds, mophorline was used instead of the deprotected
monomers as a capping group to give compounds 2a′-o′.
The purities of intermediates 7 were monitored by analytical
HPLC, and they were judged to be high enough to use in the
next step without purification. After the second coupling, a major
practical issue was the removal of inorganic salts; two proce-
dures were developed for this. Most of the bivalent products
were not very water soluble so they could be precipitated from
the basic solutions by adding 5% HCl; presumably the acid
protonates the fluorescein labels transforming them into their
less soluble, ring-closed forms. However, some of the bivalent
molecules have side chains that promote solubilities in slightly
acidic aqueous media (e.g., those in the g, h, and k series).
Bivalent molecules in this second category were separated from
most of the inorganic salts by removing the solvent and
extracting them into methanol. After these procedures, more
than 80% of the bivalent compounds 2 were produced with a
minimum of 85% purity. It appears that the other 20% were
not as pure owing to pipetting errors because all gave purities
above 85% when they were reprepared. HPLC and MALDI
analyses were performed on all the compounds. Both UV and
SEDEX detectors were used to determine the purities shown in
Figure 4, which includes the ones that were reprepared. Purity
data before and after the repreparation procedures are given in
the Supporting Information. In summary, this new methodology
allows the monovalent compounds to combine with each other
cleanly in a one-compound-per-well format.
that utilizing DMSO as solvent for coupling the second
monovalent compounds was critical; other solvents tested gave
unsatisfactory results. The bivalent molecules formed in good
overall purities, even though the monovalent building blocks
used to assemble them were completely unprotected. The
reactions were typically performed on a small scale to form a
few milligrams of product. However, no problems were
encountered on scaling-up syntheses of some selected bivalent
compounds to give 100 mg of product.
Briefly, the exploratory work that led to the method shown
above illustrated several valuable conclusions. The solvents
tested for the coupling reactions were THF, MeCN, MeOH,
DMF, and DMSO; THF proved to be best for the first addition
and DMSO for the second (see above). When DMF or DMSO
was used for the first coupling reaction, the major products
tended to be homobivalent molecules. Methanol was unsuitable
for the first coupling because it tends to react with the DTAF.
Acetonitrile gave similar results to THF in the first coupling,
but it is more toxic and slightly harder to remove. For the second
addition, the reaction went very slow in THF, MeCN, and
MeOH at 25 °C and could not be completed in a reasonable
time period. It was harder to remove DMF than DMSO, but it
seems to be an equally good solvent for the second coupling in
all other respects. Use of (NH4)2CO3 as a base was investigated
but it gave incomplete coupling reactions. This is particularly
unfortunate because ammonium salt residues are volatile,
whereas potassium chloride residues are not and have to be
removed using differential solubilities.
Application of the Solution-Phase Procedure To Make a
Library of Alkyne-Tagged Bivalent Mimics. Libraries of
labeled compounds are ideal for direct binding assays, but for
others the fluorescent label may lead to unnecessary complica-
tions. For instance, it might be desirable to screen the com-
pounds for their positive or negative influence on apoptosis in
live cell lines; in this case a superfluous fluorescent label might
alter the bioactivity or obscure readings of the outcome.
However, fluorescent labeling might subsequently facilitate
(32) Miller, J.; Parker, A. J. J. Am. Chem. Soc. 1961, 83, 117-123.
(33) Isanbor, C.; Emokpae, T. A.; Crampton, M. R. J. Chem. Soc., Perkin Trans.
2 2002, 2019-2024.
(34) Bartoli, G.; Todesco, P. E. Tetrahedron Lett. 1968, 47, 4867-4870.
(35) Emokpae, T. A.; Atasie, N. V. Int. J. Chem. Kinet. 2005, 37, 744-750.
9
562 J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008

Heterobivalent Selectivity and Bivalent-Turn Mimics
A R T I C L E S
Figure 5. Purities of the library of compounds 3 where the detection method was (a) UV set at 254 nm, and (b) SEDEX detection. The term “cap” is used
for morpholine.
further exploration of hits identified in such assays. Conse-
quently, we searched for a nonfluorescent tagging-group that
would not make the compounds much more lipophilic, but which
could be easily used to add a fluorescent label. An oligoethylene
glycol-linked terminal alkyne was selected for this purpose. The
oligoethylene glycol fragment was intended to separate the
alkyne from the presumed pharmocophores in the peptidomi-
metics and to maintain some water solubility. Copper-mediated
azide-alkyne cycloadditions36,37 could then be used to add azide-
functionalized labels38 when the time was right.
labeled molecules 2. Substitution of a fluorescein label with
the alkyne tag 8 is conceptually trivial but has some important
practical consequences. First, both coupling reactions to the
triazine were slower than in the DTAF-containing compounds
so longer reaction time was used (8 rather than 4 h for the first
coupling and 7 d instead of 3 d for the second one). The
triethylene glycol-containing tag also changes the solubility
characteristics of the products; in all cases the DMSO was
(36) Kolb, H. C.; Finn, M. G.; Sharpless, K. B. Angew. Chem. Int. Ed. 2001,
40, 2004-2021.
(37) Angell, Y.; Burgess, K. Chem. Soc. ReV. 2007, 36, 1674-1689.
(38) Seo, T. S.; Li, Z.; Ruparel, H.; Ju, J. J. Org. Chem. 2003, 68, 609-612.
The reactions used to prepare compounds 3, Scheme 3, largely
parallel those shown in Scheme 2 for the library of fluorescein-
9
J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008 563

A R T I C L E S
Angell et al.
removed, the residue was acidified with 5% HCl, and then the
products were extracted into methanol with decantation away
from the precipitated salts. The methanol was removed, and
the product was extracted into methanol again, followed by
another decantation. Analyses via HPLC with SEDEX detections
indicated that more than 90% of the inorganic salts were
removed via this procedure.
Figure 5 shows the purities of the crude products obtained
after aqueous workup of the library of compounds 3. This library
was actually designed for the NIH repository for which the
threshold was >90% purity via either UV or SEDEX detection.
All the compounds in the library met this criterion.
Evaluation of the Libraries in a Model Assay for Trk
Selectivities. Many of the dipeptide sequences chosen in the
design of the monovalent molecules 5 designed for this library
were based on the turn regions of nerve growth factor (NGF).
We have a longstanding interest in evaluation of NGF mimics
as artificial ligands for the TrkA receptor.39,40 Fluorescently
labeled compounds, as in the library of compounds 2, are ideal
for a first-pass fluorescence-activated cell sorting (FACS) assay
that involves binding to transfectant cells expressing the TrkA
receptor. The FACS output gives a measure of direct binding
to the cells and provides an ideal way to access the selectivity
of the binding to the Trk receptors. Competition with labeled
forms of the natural ligand (NGF) is not required; this is highly
advantageous because NGF has such a high affinity for its
receptor (Kd ) ca. 10-11 M),41 early stage small molecule leads
are unlikely to able to compete with it, and because labeled
NGF is prohibitively expensive.
Direct FACS assays were performed as previously de-
scribed.42 Briefly, the FITC-peptidomimetics 2 were tested via
quantitative FACS for direct binding to NIH-3T3 cells trans-
fected to express a targeted receptor or a control receptor: NIH-
TrkA cells, NIH-TrkC cells, NIH-p75 cells. NIH-3T3 cells
overexpressing the transfected insulin-like growth factor-1
receptor (NIH-IGF-1R cells) were used as negative control,
because a neurotrophin mimetic should not bind to their surface.
As positive binding controls mAbs 5C3 (anti-TrkA), 2B7 (anti-
TrkC), MC192 (anti-p75), and RIR3 (anti-IGF-1R) were used,
and nonbinding mouse IgG-FITC was used as negative control.
All the FACS assays were repeated independently at least four
times.
Many of the peptidomimetics bound strongly to the Trk-
expressing cells; they gave even higher staining than monoclonal
antibodies that bind the Trk receptors with affinities in the 10-9
M range (see Supporting Information). However, the goal of
this study was to ascertain if selectiVity was possible, rather
than to determine absolute affinity. The reason for this is that it
is relatively easy to find molecules that stick to proteins,
especially if the molecules contain fluorescein that tends to be
a promiscuous binder.43 However, if a fluorescein-containing
Figure 6. Direct binding FACScan assays with FITC-peptidomimetics:
(a) structures of compounds 2gm, 2gn, 2gi, and 2gj; (b) plot of mean channel
fluorescence of direct-binding FACS assay (significant differences, (/) p
< 0.005 and (//) p < 0.05). The background MCFs of NIH-IGF-1R were
subtracted to analyze the specific MCF binding to test cells. Depending on
the peptidomimetic, the MCFs subtracted ranged from 50 to 200 fluorescent
units and was <30% of the calculated specific signal. As positive controls,
antireceptor mAbs were used for each cell line: anti-TrkA mAb 5C3 for
NIH-TrkA (raw MCF ≈ 550), anti-TrkC mAb 2B7 for NIH-TrkC (raw
MCF ≈ 300), anti-p75 mAb MC192 for NIH-p75 (raw MCF ≈ 700), and
aIR3 for NIH-IGF-1R (raw MCF ≈ 450) (data not shown).44,45 These
positive control mAbs did not bind to cells not expressing its cognate target,
and the MCF of negative control mouse IgG-FITC ranged from 8 to 20 for
all the cell lines.
(39) Pattawarawarapan, M.; Burgess, K. J. Med. Chem. 2003, 46, 5277-5291.
(40) Burgess, K. Acc. Chem. Res. 2001, 34, 826-835.
(41) Lee, F. S.; Kim, A. H.; Khursigara, G.; Chao, M. V. Curr. Opin. Neurobiol.
2001, 11, 281-286.
(42) Zaccaro, M. C.; Lee, H. B.; Pattarawarapan, M.; Xia, Z.; Caron, A.; Burgess,
K.; Saragovi, H. U. Chem. & Biol. 2005, 12, 1015-1028.
(43) McGovern, S. L.; Caselli, E.; Grigorieff, N.; Shoichet, B. K. J. Med. Chem.
2002, 45, 1712-1722.
(44) Zaccaro, M. C.; Lee, B. H.; Pattarawarapan, M.; Xia, Z.; Caron, A.;
L’Heureux, P.-J.; Bengio, Y.; Burgess, K.; Saragovi, H. U. Chem. Biol.
2005, 12, 1015-1028.
(45) Ivanisevic, L.; Zheng, W.; Woo, S. B.; Neet, K. E.; Saragovi, H. U. J.
Biol. Chem. 2007, 282, 16754-16763.
9
564 J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008

Heterobivalent Selectivity and Bivalent-Turn Mimics
A R T I C L E S
molecule binds to a Trk receptor selectivity, whereas other
fluorescein-containing molecules do not, then that constitutes
a lead that can then be optimized without fluorescein labels.
That process involves different assays, and extensive SAR that
again is beyond the scope of this paper.
preparing large libraries of bivalent materials require that
appreciable quantities of monovalent starting materials are
available. The methodologies outlined here facilitated construc-
tion of bivalent libraries without use of protecting groups or
coupling agents. A positive feature of the methodology is that
probes, sites for further functionalization, or groups to modulate
physiochemical properties can be added to the triazine backbone.
Further studies are required to determine if the mimics 1 can
be functional analogues of â-turns, but the FACS data described
here indicate they can contribute to binding. This has revealed
several possibilities for a selective TrkA probe.
In preliminary assays, four peptidomimetics, 2gj, 2gm, 2gn,
and 2gi exhibited preferential binding to TrkA-expressing cells
when compared versus IGF-1R-expressing cells. These four
compounds were resynthesized, purified (>95%) and retested
for binding to a panel of NIH3T3 cells expressing either TrkA,
TrkC, p75, or IGF-1R (Figure 6). Analyses on the data for each
compound show that they all bind significantly better to TrkA
than to TrkC. Both 2gm and 2gn bind significantly better to
TrkA than to p75, with 2gm being better than 2gn. Compounds
2gm (TK-GR) and 2gn (TK-YS) are the better ligands for TrkA;
2gj (TK-KI) and 2gi (TK-AR) bind TrkA relatively poorly. In
addition, the four peptidomimetics bind equally and relatively
poorly to p75. The differences in binding between 2gm and
2gi suggest that the alanine side chain may cause steric
hindrance. Functional group differences observed in these
compounds may reflect the fact that NGF has four different
turns (with different side chains) that have been implicated in
binding the TrkA receptor.
Acknowledgment. Antoine Caron and Carlis Rejon (Lady
Davis) performed preliminary work for the FACS assays.
Financial support for this project was provided by the National
Institutes of Health (Grants MH070040, GM076261) and the
Robert A. Welch Foundation. TAMU/LBMS-Applications
Laboratory headed by Dr Shane Tichy provided mass spectro-
metric support, and the Laboratory for Molecular Simulation’s
(Dr Lisa Thompson) supported our molecular simulations work.
Supporting Information Available: Procedures and charac-
terization data for preparation of the monovalent compounds
and the two bivalent libraries, methods for the FACS assays.
This material is available free of charge via the Internet at
http://pubs.acs.org.
Conclusion
Monovalent molecules 1 were shown to be readily accessible
on gram scales. This is important because the logistics of
JA074717Z
9
J. AM. CHEM. SOC. VOL. 130, NO. 2, 2008 565