Towards the engineering of an orthogonal protein kinase/nucleotide triphosphate pair

Article abstract of DOI:10.1016/S0040-4020(00)00834-6

Remolding protein/ligand interfaces has led to the development of new tools for the study of biological systems. Such methods allow one to engineer proteins with specificity for designed biological probes such as inhibitors, substrates or small molecule dimerizers. Previous work in our laboratory has resulted in engineered protein kinases with specificity for ATP analogs which are otherwise orthogonal ligands for wild type kinases. Kinase reactions of analog-sensitive mutant kinases in cell lysates using γ³²P labeled ATP analogs allow identification of the direct substrates of the sensitized kinase. As an extension of this methodology, we have designed and evaluated an ATP analog, N⁴ (benzyl) ribavirin triphosphate, which may be a suitable phosphodonor for a kinase that does not utilize ATP. Such a modification is likely to be necessary for in vivo kinase substrate labeling experiments where competitive phosphodonors are present in high concentrations. (C) 2000 Elsevier Science Ltd.

Full text of DOI:10.1016/S0040-4020(00)00834-6

TETRAHEDRON
Pergamon
Tetrahedron 56 (2000) 9495±9502
Towards the Engineering of an Orthogonal Protein
Kinase/Nucleotide Triphosphate Pair
Scott M. Ulrich,a Oleksandr Buzko,a Kavita Shahb and Kevan M. Shokata,
*
aDepartment of Cellular and Molecular Pharmacology, Box 0450, University of California at San Francisco, San Francisco, CA 94143, USA
bNovartis Institute for Functional Genomics, La Jolla, CA 92121, USA
Received 8 June 2000; accepted 3 August 2000
AbstractÐRemolding protein/ligand interfaces has led to the development of new tools for the study of biological systems. Such methods
allow one to engineer proteins with speci®city for designed biological probes such as inhibitors, substrates or small molecule dimerizers.
Previous work in our laboratory has resulted in engineered protein kinases with speci®city for ATP analogs which are otherwise orthogonal
ligands for wild type kinases. Kinase reactions of analog-sensitive mutant kinases in cell lysates using g32P labeled ATP analogs allow
identi®cation of the direct substrates of the sensitized kinase. As an extension of this methodology, we have designed and evaluated an ATP
analog, N4 (benzyl) ribavirin triphosphate, which may be a suitable phosphodonor for a kinase that does not utilize ATP. Such a modi®cation
is likely to be necessary for in vivo kinase substrate labeling experiments where competitive phosphodonors are present in high concentra-
tions. q 2000 Elsevier Science Ltd. All rights reserved.
Introduction
kinases invariably contain one or more binding domains
that have regulatory and/or targeting functions. The earliest
studied and most prevalent is the Src Homology 2 (SH2)
domain, which binds speci®c phosphotyrosine sequences.
Receptor kinase phosphorylations create SH2 binding
sites, which modulate cytosolic protein kinases and other
signaling enzymes. Src Homology 3 (SH3) domains are
another module recognizing intra- or inter-molecular poly-
proline motifs for targeting and regulation.5 It is the regula-
tion and localization provided by these and other binding
domains that direct kinase activity to the correct cellular
location and ensure appropriate activation and correct
signaling.6
Protein kinases are a family of enzymes that catalyze the
transfer of g-phosphate from ATP to tyrosine, serine or
threonine amino acid residues of substrate proteins. Phos-
phorylation alters the enzymatic activity, binding capability
or cellular localization of the substrate protein as a means to
relay cellular signals from the environment such as the
extracellular matrix, antigens and hormones such as insulin
and growth factors.1 Since the discovery of protein phos-
phorylation as a mechanism of signal transduction, the
discovery of the v-Src and v-Abl oncogenes,2,3 and the
realization that protein kinases are an immense superfamily
of proteins (2.1% of C. Elegans genes are protein kinases);
protein kinases have moved to center stage in the ®eld of
signal transduction. Owing to their centrality in cell signal-
ing, protein kinases have also become attractive therapeutic
targets for such diverse diseases as diabetes and cancer.23
Despite many years of intense study, the identity of bona
®de substrates of most kinases remain unknown. Determin-
ing physiological kinase substrates is dif®cult for several
reasons. First, kinases often have very little preference for
a particular sequence ¯anking the phosphorylated residue.7
While the targeting/binding domains serve to present
meaningful physiological substrates to a particular kinase
domain upon activation, the kinase domain itself is rela-
tively promiscuous.1,8±10 It is also well known that many
important kinases are genetically redundant due to the
presence of related family members. Thus, determining
the targets of one kinase in a complex signaling network
involving multiple kinases has proven intractable.9
Protein kinases occur as cytosolic domains of receptor
proteins, stimulated by extracellular ligand binding, or as
non-receptor cytosolic enzymes. The molecular details of
how protein kinases are regulated and activated have been
elucidated by structural studies. Receptor protein kinases
typically relay signals across the membrane by dimeriza-
tion/oligimerization followed by trans autophosphorylation
of the intracellular kinase domains, activating them to phos-
phorylate cellular protein substrates.4 Cytosolic protein
To address this problem, we have previously reported a
`chemical genetic' approach to trace the substrates of a
single kinase in the presence of all cellular kinases. This
method involves engineering a kinase such that it is
Keywords: protein/ligand interfaces; dimerizers; phosphodonors.
* Corresponding author. Tel.: 11-415-514-0472, fax: 11-415-514-0822;
e-mail: shokat@cmp.ucsf.edu
0040±4020/00/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved.
PII: S0040-4020(00)00834-6

9496
S. M. Ulrich et al. / Tetrahedron 56 (2000) 9495±9502
Figure 1. Schematic outlining the method of orthogonal ApTP analogs and the analog-sensitive allele v-Src-as1 to tag the direct v-Src substrates in the
presence of multiple kinases.
uniquely able to recognize an unnatural ATP analog. The
unnatural ApTP molecules are chemically modi®ed with a
benzyl `bump' on N6 of the adenine ring of ATP, rendering
the nucleotide orthogonal to (a dead substrate for) all wild-
type kinases. A single mutation (I338G in v-Src) sensitizes
the mutant kinase to accept N6 (benzyl) ATP (Fig. 1). The
analog-sensitive allele is termed V-SRC-AS1. Addition of
[g32P] N6 (benzyl) ATP to cell lysates speci®cally radio-
labels the direct substrates of v-Src-as1.10,11 The mutation
that renders v-Src-as1 sensitive to ATP analogs is readily
generalizable to other Src family kinases and many other
unrelated kinases (K. M. S., unpublished observations).
1. Synthesize modi®ed ApTP molecules orthogonal to wild
type kinases.
2. Identify mutations that allow kinases to accept the ortho-
gonal ApTP as an alternative phosphodonor.
3. Further modify ApTP!AppTP such that it can be
accepted by a modi®ed kinase that does not accept ATP.
4. Identify a mutant kinase that can accept this orthogonal
AppTP while itself being orthogonal to ATP.
Design goals #1 and #2 are satis®ed by the use of N6
(benzyl) ATP as an alternate phosphodonor for v-Src-as1
previously reported by our laboratory.11 Here we describe
the design, synthesis, and characterization of an AppTP that
may satisfy the third design goal. In order to design such an
AppTP analog, we had to ®rst design a kinase active site
which does not bind ATP. In creating such an orthogonal
kinase, we sought to use the same strategy that rendered N6
(benzyl) ATP orthogonal to wild type kinases. Adding a
steric bump to the N6 position of adenine precludes N6
(benzyl) ATP binding to the wild type active site. In an
analogous fashion, an amino acid mutation from a small
! large side chain in the active site of a kinase should
block binding and phosphotransfer with ATP (Fig. 2). The
conclusion from this exercise is that a kinase that does not
recognize ATP for steric reasons would likely contain a
smaller active site than wild type kinases.
Here we describe our efforts to improve the sensitivity of the
above method of tagging direct protein kinase substrates.
Speci®cally, we aim to create a kinase which only utilizes
the synthetic ApTP derivative. Engineering the kinase to be
orthogonal to all other nucleotide triphosphates in the cell
(ATP, GTP, CTP, TTP, etc.) will dramatically increase the
ef®ciency of substrate labeling. Eliminating or drastically
minimizing the ability of the engineered kinase to use ATP
is likely to be necessary for in vivo protein kinase substrate
labeling experiments, as the concentration of cellular ATP is
quite high (1±5 mM).24
Design
Starting with the N6 (benzyl) ATP scaffold, we sought to
create a smaller nucleotide base that could be recognized by
an active site that will exclude ATP as per design goal #4.
The design goals (Fig. 2) for an ideal engineered kinase/
synthetic nucleotide are:

S. M. Ulrich et al. / Tetrahedron 56 (2000) 9495±9502
9497
Figure 2. Second generation of AppTP and orthogonal kinase.
The C2 and N3 positions of ATP (blue in Fig. 3) do not make
any speci®c hydrogen bonds with the kinase, so removing
these atoms to create a smaller nucleotide would not disrupt
any speci®c contacts present in the kinase±substrate
complex. The resulting nucleotide (shown in Fig. 3b) is an
amidine imidazole and has similar sterics and functionality
to the triazole carboxamide (Fig. 3c) which is more stable
and synthetically accessible. This nucleoside derivative
(Fig. 3c) is N4 (benzyl) ribavirin, the unsubstituted parent
of which is an antiviral used clinically to treat Hepatitis C.12
The amide moiety of ribavirin is capable of presenting a
benzyl group to the I338G pocket similar to N6 (benzyl)
ATP. The ribavirin scaffold also satis®es hydrogen bond
donor/acceptor interactions with the kinase seen in all
ADP and inhibitor bound kinase crystal structures.27
Previous work on carboxamide substituted 5-membered
ring nucleotides has shown their capacity to retain
adenine-like hydrogen bonding in DNA duplexes.29
AutoDock.17 To con®rm that our docking algorithm
could accurately predict small molecule-protein inter-
actions, we tested it with the known ApTP analog, N6
(benzyl) ATP and the c-Src-as1 kinase. The X-ray crystal
structure of c-Src bound to ADP has been determined.20 The
structure of c-Src is likely to be identical to that of v-Src
since the two proteins have almost identical active site
residues, hence, we used it as a model in our docking
studies. We used the molecular graphics package
insightII19 to modify the structure of the wild type enzyme
and introduce the T338G change with the assumption that
such a mutation would be unlikely to affect the position of
the protein backbone.25 N6 (benzyl) ATP was docked to both
c-Src and c-Src-as1. The top scoring orientations were
compared to the structure of the c-Src/ADP complex.
This analysis showed that N6 (benzyl) ATP binds to c-Src-
as1 in a fashion similar to that of ADP bound to wild type
c-Src (Fig. 4b). The binding orientation difference has a root
This report describes the synthesis of N4 (benzyl) ribavirin
triphosphate (N4 (benzyl) RTP) and its structure±activity
relationship with both v-Src and v-Src-as1.
mean square deviation (RMSD) of 1.7^0.4 A (calculated
Ê
for a set of the ®ve top scoring orientations). The adenine
ring system of N6 (benzyl) ATP occupies the same region in
space within the binding site of the enzyme and forms essen-
tially the same interactions with surrounding residues as
ADP. These include the two essential hydrogen bonds
between the N6 amino group and NH group of E339 and
N1 nitrogen of ADP and carbonyl of M341. The ribose ring
is positioned in the same location with a minor shift, which
does not signi®cantly affect orientation of the phosphate
Results and Discussion
We ®rst used computer aided modeling to determine
whether N4 (benzyl) RTP, (AppTP) could bind v-Src-as1
using a modi®ed version of the molecular docking suite
Figure 3. N4 (benzyl) RTP as an AppTP. N6 (benzyl) ATP (a) was modi®ed to remove N2 and C3 in order to accommodate a smaller kinase active site which is
orthogonal to ATP. (b) The structure was further modi®ed for greater stability to give (c), N4 (benzyl) ribavirin.

9498
S. M. Ulrich et al. / Tetrahedron 56 (2000) 9495±9502
Figure 4. Graphic representation of docking results. N6 (benzyl) ATP (a, b) and N4 (benzyl) RTP (c, d) docked to wild type c-Src and v-Src-as1 overlaid with
ADP from the experimental structure solved by X-ray crystallography.
moiety. The benzyl `bump' of N6 (benzyl) ATP ®ts into the
engineered pocket as predicted by our general binding
model.26
Next, N4 (benzyl) RTP was docked to c-Src-as1 to test if the
new scaffold could recapitulate the appropriate interactions
that make N6 (benzyl) ATP a substrate for v-Src-as1. N4
(benzyl) RTP appears to bind in a fashion comparable but
not identical to the one observed for ADP (Fig. 4d). Root
Furthermore, when N6 (benzyl) ATP was docked to c-Src,
which lacks the engineered pocket to accommodate the
benzyl `bump', no consensus binding orientation was
found (one orientation is shown in Fig. 4a). This con®rms
the experimental observation that N6 (benzyl) ATP cannot
be accommodated by the wild type active site. These
calculations are in agreement with the experimental obser-
vation that N6 (benzyl) ATP is orthogonal to wild type v-Src
and is an ef®cient substrate of v-Src-as1.11 The ability of
computer modeling to con®rm N6 (benzyl) ATP binding to
c-Src-as1 validates the use of docking to guide our AppTP
design.
Ê
mean square deviation of common atoms is 2.2^0.5 A
relative to ADP. The benzyl group of N4 (benzyl) RTP is
positioned similarly to N6 (benzyl) ATP within the
engineered pocket. However, the triazole and ribose rings
do not overlay well with the imidazole and ribose rings
of ADP. In addition, the N4 of the analog is positioned
in such a way that one of the hydrogen bonds (to E339)
is not satis®ed. These observations re¯ect the greater
¯exibility of N4 (benzyl) RTP relative to the purine
scaffold.
Overall, our docking results predict that N4 (benzyl) RTP
should behave like N6 (benzyl) ATP, being orthogonal to
wild type v-Src and accepted by v-Src-as1.
The AppTP analog N4 (benzyl) RTP was ®rst docked to wild
type c-Src to test orthogonality. Docking of N4 (benzyl)
RTP to wild type c-Src showed no distinct consensus
binding orientation (one orientation is shown in Fig. 4c)
suggesting it is poorly recognized by wild type active
sites. The benzyl substituent, which confers orthogonality
to N6 (benzyl) ATP, is able to rotate out of the pocket into
the solvent in the case of N4 (benzyl) RTP because of the
¯exible linker (Fig. 4c). It appears that it could bind the
kinase by virtue of extensive hydrophobic contacts, which
compensate for the lack of speci®c hydrogen bonding.
To experimentally test our modeling predictions, N4
(benzyl) RTP was synthesized according to Scheme 1.
Amino triazole carboxylic acid 3 was converted to the C3
diazo compound and deaminated in situ by heating in
methanol.16 The resulting acid 5 was coupled to benzyl-
amine with PyBOP.13 Glycosylation of 6 was performed
using the procedure of Vorbruggen14 and the resulting
product 7 deprotected by treatment with ammonia in

S. M. Ulrich et al. / Tetrahedron 56 (2000) 9495±9502
9499
Scheme 1.
methanol to afford 8. Addition of the triphosphate moiety to
yield N4 (benzyl) RTP 9 was carried out as described in
Ludwig et al.28
The data on the new AppTP analog, N4 (benzyl) RTP is
surprising based on our modeling predictions. First, N4
(benzyl) RTP is not orthogonal to wild type v-Src as indi-
cated by the amount of phospho-GFP product (Fig. 5, lane
4). The structural modi®cations to N6 (benzyl) ATP to give
N4 (benzyl) RTP make the nucleotide lose its orthogonality,
presumably due to the ¯exible linker attaching the benzyl
`bump'. This ¯exible linker must give suf®cient conforma-
tional freedom to allow the benzyl moiety of N4 (benzyl)
RTP to swing out of the wild type active site. Such a confor-
mation can been seen in the docked structure of N4 (benzyl)
RTP to wild type c-Src (Fig. 4c). This alternate binding
mode is not available to N6 (benzyl) ATP (Fig. 4a).
The ability of the wild type v-Src and v-Src-as1 to utilize N4
(benzyl) RTP as a phosphodonor was tested in a kinase-
mediated reaction. Truncated forms of v-Src and v-Src-
as1 D(77±225) were expressed as GST fusions from E.
coli and puri®ed by glutathione af®nity chromatography.
A green ¯uorescent protein modi®ed with an optimized
Src phosphorylation site (IYGEF addition at position 235,
GFP235IYGEF) served as an exogenous protein substrate.15
Phosphorylated GFP235IYGEF was detected with a mono-
clonal antiphosphotyrosine speci®c antibody, 4G10.
An even more striking deviation from our design hypothesis
is that adding the I338G `hole' (v-Src-as1) abrogated the
ability of N4 (benzyl) RTP to be a phosphodonor (Fig. 5,
lane 8), exactly the opposite of design goal #3. The
additional conformational ¯exibility of N4 (benzyl) RTP
may again explain why v-Src-as1 is not catalytically compe-
tent with N4 (benzyl) RTP but uses N6 (benzyl) ATP with
good ef®ciency. The hydrophobic benzyl substituent of N4
(benzyl) RTP may be able to bind deeper and tighter to the
active site, inhibiting catalytic turnover, as suggested by our
modeling studies. The modeled structure of N4 (benzyl)
RTP docked to v-Src-as1 (Fig. 4d) shows that a key hydro-
gen bond from the amide N±H to the protein backbone
carbonyl of glutamate 339 is not satis®ed. It may also be
that N4 (benzyl) RTP binds the active site in a catalytically
unproductive orientation. In any case, N4 (benzyl) RTP does
not serve as an alternate substrate for v-Src-as1 as we
expected.
The results from this assay show that GFP235IYGEF is an
ef®cient kinase substrate (Fig. 5, lane 1 and 2). As a control,
the I338G mutation sensitizes v-Src-as1 to the orthogonal
analog, N6 (benzyl) ATP (Fig. 5, lanes 3 and 7).
Figure 5. GFP235IYGEF western blot assay of wild type v-Src (lanes 1±4)
and v-Src-as1 (lanes 5±8) with various nucleotides. Lane 1, 5, no GFP
control; lane 2, 6, 100 mM ATP; lane 3, 7, 100 mM N6 (benzyl) ATP;
lanes 4, 8, 100 mM N4 (benzyl) RTP.
These results suggest that N4 (benzyl) RTP is not a good

9500
S. M. Ulrich et al. / Tetrahedron 56 (2000) 9495±9502
AppTP for an orthogonal protein kinase. However, the assay
in Fig. 5 was performed in the absence of ATP. Our model-
ing predicts weak binding of N4 (benzyl) RTP by wild type
v-Src; a weak binder should compete poorly with the natural
substrate (ATP) for wild type kinase active sites.
nature of the chemical `bump' or engineered pocket to
achieve catalysis.
Conclusion
Since the impetus of the new design goals is to be able to
trace kinase substrates in cells, we asked if N4 (benzyl) RTP
is conditionally orthogonal to wild type v-Src in the
presence of the competitive phosphodonor, ATP. This was
carried out by measuring N4 (benzyl) RTP inhibition of a
wild type v-Src/[g32P] ATP reaction. A trace amount (nM)
of [g32P] ATP was added to wild type v-Src and the
GFP235IYGEF protein substrate in kinase buffer. N4
(benzyl) RTP was added at varying concentrations. No inhi-
bition was observed with up to 100 mM N4 (benzyl) RTP
(Fig. 6a). This result demonstrates that in the presence of
even nanomolar amounts of ATP, N4 (benzyl) RTP is a very
poor alternate substrate for wild type v-Src (IC50.1 mM).
Therefore, N4 (benzyl) RTP satis®es design goal #3 under
cellular conditions where the ATP concentration is .1 mM
and thus is conditionally orthogonal to wild type kinases,
and is potentially useful as a substrate for an engineered
kinase in a substrate labeling experiment.
Reengineering small molecule/protein interfaces to achieve
high speci®city has been a powerful tool in the design of
kinase inhibitors and small molecule dimerizers.21 Rational
remolding of binding surfaces requires sophisticated model-
ing of noncovalent interactions and chemical intuition. A
signi®cantly more dif®cult proposition, the engineering of a
new catalytic binding site also requires engineering new
binding interactions, but this alone does not guarantee
catalysis. Enzyme catalysis requires subtleties of reactive
group positioning and a ®ne balance of substrate binding,
transition state stabilization and product release. Our initial
efforts to design a second generation of unnatural nucleotide
triphosphate (AppTP)/engineered kinase underscore the
dif®culty of capturing these elements in a rationally
designed system.
Although the N4 (benzyl) RTP molecule is not strictly
orthogonal to wild type v-Src as we hoped, the af®nity of
the active site for N4 (benzyl) RTP is .100 fold lower than
for ATP. In a substrate labeling experiment, cellular ATP
will out compete N4 (benzyl) RTP with wild type kinases.
The nucleotide will behave as a dead substrate for wild type
kinases under the high ATP concentrations found in cells.
The structure activity relationship of N4 (benzyl) RTP and
v-Src-as1 was also explored using the same competition
assay. Again, v-Src-as1 and GFP235IYGEF were incubated
with a trace amount of [g32P] ATP and N4 (benzyl) RTP was
added at varying concentrations. The antiphosphotyrosine
blot shows that N4 (benzyl) RTP has remarkably high
af®nity for the v-Src-as1 active site (IC50,1 mM),
(Fig. 6b). This result, although disappointing, suggests
that the structural modi®cations converting N6 (benzyl)
ATP to N4 (benzyl) RTP do not abolish interaction between
the nucleotide and kinase. However, we designed N4
(benzyl) RTP to be phosphodonor, not an inhibitor. If N4
(benzyl) ATP is not catalytically competent due to its high
af®nity, its orientation may be modulated by changing the
Another dif®culty encountered for this new AppTP analog is
a result of removing the N2 and C3 atoms from the purine
ring to generate a smaller base. The added ¯exibility yielded
an AppTP that has remarkable af®nity for the v-Src-as1
kinase containing the engineered pocket yet is not a compe-
tent substrate. Different I338G-like mutations in v-Src-as1
may modulate the af®nity and orientation of N4 (benzyl)
RTP and allow catalysis. Similarly, other chemical
`bumps' may modulate the orientation and af®nity of inter-
action, and such experiments on newer generations of
AppTP analogs are underway. Potentially, alternatives to
rational design such as error prone PCR, phage display or
gene shuf¯ing could be used to identify mutant kinases that
are ef®cient catalysts with N4 (benzyl) RTP.
Materials and Methods
Docking
We used a modi®ed molecular docking suite, AutoDock,
originally developed at the Scripps Research Institute.17
The suite is based on the simulated annealing algorithm
applying adiabatic cooling as the docking process
progresses. It includes binding energy grid calculation
program, evaluating interaction energies in every point of
the scoring grid. The molecule of the ligand is processed to
include electrostatic charges, as well as information about
rotatable bonds. The docking routine introduces stepwise
translational or rotational changes to the small molecule
evaluating its binding parameters via the scoring grid,
thus, moving it in the direction of descending energy
gradient. Binding energy is evaluated as a sum of
Figure 6. Inhibition of wild type v-Src (a) and v-Src-as1 (b) by N4 (benzyl)
RTP.
GFP235IYGEF protein substrate and varying amounts of N4 (benzyl) RTP
in kinase buffer.
A
trace amount of [g32P] ATP was incubated with kinase,

S. M. Ulrich et al. / Tetrahedron 56 (2000) 9495±9502
9501
non-bonded and electrostatic interactions, as well as hydro-
gen bonding. The ligand is ¯exible, whereas the protein
target is kept rigid.
(5.0 g, 11.3 mmol) was added and followed immediately
by benzylamine (1.5 mL, 13 mmol). The reaction was stir-
red at rt under argon for 2 h. The insoluble tan product was
collected by ®ltration and washed with saturated aqueous
1
In order to add reliability to the performance of the
program, we have introduced elements modeling hydro-
phobic interactions within the binding site. Among other
improvements are introduction of additional atom types,
including `polar hydrogen' type and hydrogen bonding
potential energy function with a higher attraction com-
ponent to optimize modeling of hydrogen bonds.22 The
most apparent improvement has been observed for protein
targets with relatively open binding sites lacking multiple
Van der Waals contacts and/or hydrogen bonding. In such
cases, as it could be seen from analysis of the binding site,
hydrophobic effects played major role in positioning the
small molecule.
sodium bicarbonate and methanol to yield 6 (55%). H
NMR (DMSO-d6, 400 MHz) d 9.12 (s, 1H), 8.43 (s, 1H),
7.21 (m, 3H), 7.18 (m, 3H), 4.41 (d, Jˆ4.5 Hz, 2H). HRMS
(1EI) 202.0855; calcd 202.2057 for C10H10N4O.
N4 (Benzyl) ribavirin 2,3,5 tri-O-benzoate (8). To b-d-
ribofuranose 1-acetate 2,3,5-tri-O-benzoate (0.74 g, 1.48
mmol) in dry acetonitrile, compound 6 (0.3 g, 1.48 mmol)
was added and the suspension stirred at rt under argon.
Hexamethyldisilazane (0.38 mL, 1.6 mmol) was added
and the reaction stirred for 0.5 h. To this, chlorotrimethyl-
silane (0.29 mL, 1.6 mmol) was added, followed immedi-
ately by tri¯oromethanesulfonic acid (0.20 mL, 2.1 mmol),
upon which the solution turned clear. The solution was stir-
red at rt under argon for 5 h. The reaction was quenched by
the addition of saturated sodium bicarbonate (200 mL), and
the product was extracted with EtOAc (3£200 mL). The
organic layers were combined and dried over Na2SO4 and
concentrated in vacuo. The residue was chromatographed
on silica gel (2:1 CH2Cl2:EtOAc, Rfˆ0.5) to yield 8 (0.81 g,
85%). 1H NMR (CDCl3, 400 MHz) d 8.40 (s, 1H), 8.03 (m,
2H), 7.91 (m, 4H), 7.48 (m, 5H), 7.32 (m, 10H), 6.32 (d,
Jˆ3.2 Hz, 1H), 6.17 (m, 1H), 6.07 (t, Jˆ5.6 Hz, 1H), 4.75
(m, 2H), 4.62 (m, 3H).
Each experiment included 25 independent docking runs to
ensure statistical signi®cance of the results. Every run
started with the small molecule placed outside of the bind-
Ê
ing site approximately 15 to 20 A from the target. Resulting
orientations have been stored as separate coordinate ®les
and clustered according to their scores. We used spartan18
to minimize the structures of the ligands and to derive elec-
trostatic charges using MNDO basis set. The protein atoms
have been assigned charges with CVFF91 force ®eld as a
part of molecular modeling suite insightII.19 All docking
experiments were performed on a workstation Silicon
Graphics O2.
N4 (Benzyl) ribavirin (9). Compound 8, (0.8 g, 1.25 mmol)
was stirred in 2 M ammonia in methanol (25 mL) for 2 days
at rt. The solvent was removed in vacuo and the residue
chromatographed on silica gel (7:1 EtOAc/MeOH) to
yield 9 (0.41 g, 96%, Rfˆ0.4). 1H NMR (DMSO- d6,
400 MHz) d 9.04 (t, Jˆ4.8 Hz, 1H), 8.85 (s, 1H), 7.27 (m,
5H), 5.79 (d, Jˆ3 Hz, 1H), 5.54 (d, Jˆ4.2 Hz, 1H), 5.16 (d,
Jˆ4.2 Hz, 1H), 4.87 (t, Jˆ3.9 Hz, 1H), 4.40 (d, Jˆ4.5 Hz,
2H), 4.31 (q, Jˆ3.6 Hz, 1H), 4.10 (q, Jˆ3.9 Hz, 1H), 3.91
(q, Jˆ3.3 Hz, 1H), 3.57 (m, 1H), 3.46 (m, 1H). HRMS
(1EI) 334.1263; calcd 334.3322 for C15H18O5N4.
v-Src mutagenesis, expression and puri®cation was
carried out as previously described.10
Kinase reactions were carried out in 20 mM Tris pH 8, with
100 mM KCl and 10 mM MgCl2. 3 mg GFP-IYGEF and
1±10 ng kinase was added15 and the reactions carried out
at room temperature for 20 min. The reactions were
separated by SDS-PAGE electrophoresis and transferred
to nitrocellulose membranes (Bio-Rad). Phosphotyrosine
was detected by 4G10, an antiphosphotyrosine speci®c
monoclonal antibody followed by GaM HRP (Upstate
Biotechnology) chemiluminescence.
N4 (Benzyl) ribavirin triphosphate (10). Benzyl ribavirin
9 (0.06 g, 0.18 mmol) was dissolved in trimethyl phosphate
(0.5 mL) and stirred under argon on ice. Phosphorous
oxychloride (60 mL, 0.65 mmol) was added and the reaction
stirred at 08 for 1.5 h. Bis (tributylammonium) pyro-
phosphate [prepared by dissolving 80 % pyrophosphoric
acid (0.18 g, 0.81 mmol) in 1:1 water/ethanol (2 mL)
followed by addition of tributylamine (0.5 mL, 2.1 mmol)
and evaporated in vacuo to dryness] in dry DMF (1 mL) was
added quickly and stirred for 1 min. The reaction was
quenched by addition of 1 M TEA-B pH 7 (5 mL) and the
solvent was removed in vacuo. The residue was dissolved in
water (2 mL) and puri®ed by HPLC (Rainin) over a Poros
HQ/M column, 0±1 M gradient of TEA-B buffer pH7. MS.
(-ESI), m/z[M21]ˆ573.
IC50 determinations for N4 (benzyl) RTP was carried
out by incubating wild type and I338G v-Src kinase
with 10 mCi [g32P] ATP and varying amounts of cold N4
(benzyl) RTP in 20 mM Tris pH 8, 100 mM KCl, and
10 mM MgCl2. These reactions were separated by SDS-
PAGE electrophoresis, dried and exposed to X-ray ®lm
overnight.
Chemical Synthesis
1,2,4-Triazole-3-carboxylate (5) was prepared by the
method of Grinshtein in Ref. 16. The carboxylate is
thermally labile and used without further puri®cation.
Acknowledgements
1,2,4-Triazole-3-benzamide (6). Compound 5, (1.3 g,
11.5 mmol) was suspended in dry acetonitrile (60 mL). To
this, triethylamine (2.5 mL, 34 mmol) was added, upon
which the solution turned clear. To this solution, PyBOP
Thanks to the Balmain lab, P. J. Alaimo and other members
of the Shokat lab for helpful comments on the text. This

9502
S. M. Ulrich et al. / Tetrahedron 56 (2000) 9495±9502
15. Yang, F.; Liu, Y.; Bixby, S.; Friedman, J.; Shah, K.; Shokat,
K. M. Anal. Biochem. 1999, 266, 167±173.
work was supported by a grant from the National Institutes
of Health (2R01CA70331) and the Sandler Foundation.
16. Chiipen, G. I.; Grinshtein, V. Y. Chem. Heterocycl. Comp.
1965, 1, 420±421.
References
17. Morris, G. M.; Goodsell, D. S.; Huey, R.; Olson, A. J.
J. Comput.-Aided Mol. Des. 1996, 10, 293±304.
18. spartan 5.1, Wavefunction, Inc. 18401 Von Karman Ave.,
Ste. 370, Irvine, CA 92612, USA.
1. Hunter, T. Cell 2000, 100, 113±127.
2. Hunter, T.; Cooper, J. A. Ann. Rev. Biochem. 1985, 54, 897±
930.
3. Bishop, J. Cell 1985, 42, 23±28.
19. insightII 98.0 Molecular Simulations, Inc., 9685 Scranton
Road, San Diego, CA 92121-3752 USA.
4. Schlessinger, J. Trends Biochem. Sci. 1988, 13, 443±447.
5. Pawson, T. Nature 1995, 373, 573±580.
20. Xu, W.; Harrison, S.; Eck, M. J. Nature 1997, 385, 595±602.
21. Clackson, T. Curr. Opin. Struct. Biol. 1998, 8, 451±458.
22. Buzko, O.; Bishop, A.; Liu, Y.; Shokat, K. M., in preparation.
23. Bishop, A. C.; Shokat, K. M. Pharmacol.Ther. 1999, 82, 337±
346.
6. Pawson, T.; Scott, J. D. Science 1997, 278, 2075±2080.
7. Songyang, Z.; Carraway, K. L. I.; Eck, M. J.; Harrison, S. C.;
Feldman, R. A.; Mohammadi, M.; Schlessinger, J.; Hubbard, S. R.;
Smith, D. P.; Eng, C.; Lorenzo, M. J.; Ponder, B. A. J.; Mayer, B. J.;
Cantley, L. C. Nature 1995, 373, 536±539.
24. Voet, D.; Voet, J. G. Biochemistry; Wiley: New York, 1995,
p 430.
8. Kamps, M. P.; Sefton, B. M. Oncogene Res. 1988, 3, 105±115.
9. Brown, M. T.; Cooper, J. A. Biochim. Biophys. Acta 1996,
1287, 121±149.
25. Position of the backbone of the glycine mutant of p38 MAP
kinase is identical to that of the wild type protein, as derived from
crystallographic studies (unpublished results).
26. Liu, Y.; Shah, K.; Yang, F.; Witucki, L.; Shokat, K. M. Bioorg.
Med. Chem. 1998, 8, 1219±1226.
10. Shah, K.; Liu, Y.; Deimengian, C.; Shokat, K. M. Proc. Nat.
Acad. Sci. USA 1997, 94, 3565±3570.
11. Liu, Y.; Shah, K.; Yang, F.; Witucki, L.; Shokat, K. M. Chem.
Biol. 1998, 5, 91±101.
27. Sicheri, F.; Moar®, I.; Kuriyan, J. Nature 1997, 385, 602±609.
28. Ludwig, J. Acta. Biochim. Biophys. Acad. Sci. Hung. 1981, 16,
131±133.
12. Witkowski, J. T.; Robins, R. K.; Sidwell, R. W.; Simon, L. N.
J. Med. Chem. 1972, 15, 1150±1154.
29. Johnson, W. T.; Zhang, P.; Bergstrom, D. E. Nucleic Acids
Res. 1997, 25, 559±567.
13. Coste, J.; Le-Nguyen, D.; Castro, B. Tetrahedron Lett. 1990,
31, 205±208.
14. Vorbruggen, H.; Krolikiewicz, K.; Bennua, B. Chem. Ber.
1981, 114, 1234±1255.