Regulation of gene expression by synthetic dimerizers with novel specificity

Article abstract of DOI:10.1016/S0960-894X(03)00707-8

New synthetic chemical inducers of dimerization, comprising homodimeric FKBP ligands with engineered specificity for the designed point mutant F36V, have been evaluated for inducing targeted gene expression in mammalian cells. Structure-activity studies indicated that high-affinity dimerizers such as AP1903 are ineffective, perhaps due to kinetic trapping of non-productive dimers, whereas lower-affinity molecules, exemplified by AP1889 and AP1966, potently activate transcription.

Full text of DOI:10.1016/S0960-894X(03)00707-8

Bioorganic & Medicinal Chemistry Letters 13 (2003) 3181–3184
Regulation of Gene Expression by Synthetic Dimerizers
with Novel Specificity
Wu Yang,* Terence P. Keenan, Leonard W. Rozamus, Xiurong Wang,
Victor M. Rivera, Carl T. Rollins, Tim Clackson* and Dennis A. Holt
ARIAD Gene Therapeutics, Inc., 26 Landsdowne Street, Cambridge, MA 02139-4234, USA
Received 13 May 2003; accepted 1 July 2003
Abstract—New synthetic chemical inducers of dimerization, comprising homodimeric FKBP ligands with engineered specificity for
the designed point mutant F36V, have been evaluated for inducing targeted gene expression in mammalian cells. Structure–activity
studies indicated that high-affinity dimerizers such as AP1903 are ineffective, perhaps due to kinetic trapping of non-productive
dimers, whereas lower-affinity molecules, exemplified by AP1889 and AP1966, potently activate transcription.
# 2003 Elsevier Ltd. All rights reserved.
The use of chemical inducers of dimerization (CIDs), or
dimerizers, to control protein–protein interactions is
now well established. Proteins of interest are expressed
in cells as fusions to ligand-binding domains; addition
of a bivalent ligand drives association of the proteins.
Since the initial report of FK1012, a semi-synthetic
dimer of FK506, and its use to control the association
of FKBP fusion proteins,1 chemical dimerization has
been used to create inducible alleles of numerous sig-
naling proteins that are naturally activated by oligo-
merization.2 Induced dimerization can also be used to
control gene expression, by fusing drug binding domains
to the separate DNA-binding and activation domains of
a transcription factor: addition of dimerizer reconstitutes
an active transcription factor, initiating expression of a
gene engineered with an appropriate promoter.3ꢀ5
Dimerizer-induced gene expression is a useful research
tool for analyzing gene function, and also has potential
therapeutic utility for the pharmacologic control of gene
expression in the context of gene therapy.6
Fas-mediated apoptosis of engineered cells, and in acti-
vating the expression of target genes.7
Despite the advantages of AP1510 and its congeners,
their use in vivo is potentially complicated by their
interaction with endogenous FKBP molecules, which are
ubiquitously expressed in eukaryotic cells, often at high
levels. Interaction with endogenous FKBPs could poten-
tially interfere with natural FKBP functions and could
also blunt dimerizer potency through the sequestration of
dimerizer in FKBP-rich cells, such as erythrocytes. To
address this issue, we further modified the compounds to
enhance their specificity for engineered fusion proteins
using a ‘bump and hole’ strategy.9,10 A bulky substituent
(‘bump’) was added to each monomeric unit to sterically
disrupt FKBP binding, and a complementary ‘hole’ was
then generated in the binding pocket of the fusion FKBP
construct by structure-guided site-directed mutagenesis.
We have reported the successful use of this strategy to
generate the dimerizer AP1903 (1b) (Fig. 1), which binds
with high affinity and >1000-fold selectivity to the
Phe36Val (F36V) mutant of FKBP compared to the
wild-type protein.9 AP1903 and its congeners potently
activate signaling in cells expressing FKBP-F36V fusion
proteins, both in cell culture and in a variety of in vivo
settings.9,11 However, the evaluation of these engineered
dimerizers for use in regulated gene expression has not
been reported. We describe here the synthesis and
structure–activity studies of a series of AP1903 congeners
optimized for the regulated transcription of target genes.
Previously, we reported the synthesis and structure–
activity relationship of a family of bivalent FKBP ligands
typified by the homodimeric compound AP1510 (1a, Fig.
1).7,8 These molecules are much simpler than FK1012, are
of lower complexity and molecular weight than FK1012,
and are fully synthetic. In addition, AP1510 was found to
be significantly more potent than FK1012 in inducing
*Corresponding authors. Tel.:+1-617-494-0400; fax:+1-617-225-
2589; e-mail: clackson@ariad.com
0960-894X/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0960-894X(03)00707-8

3182
W. Yang et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3181–3184
identified as the most potent in this series. To investigate
the relationship between binding affinity and potency, we
also determined the relative F36V-FKBP binding affi-
nities of the monomeric FKBP ligands used to assemble
dimerizers using a fluorescence polarization (FP) assay.9
Consistent with previous assays using the transcrip-
tional system with wild-type FKBP fusion proteins,7
AP1510 (1a) induced much higher levels of transcription
than FK1012 in transient assays using the F36V-FKBP
‘holed’ system (Table 1), despite its weaker affinity (IC50
of monomer: 44 nM vs 12 nM). Furthermore in stably
transfected cells FK1012 proved completely inactive
while AP1510 (1a) continued to function as a dimerizer,
albeit at somewhat higher concentrations (EC50: 177
nM vs 294 nM). Unexpectedly, we found that the
divergence of behavior of FK1012 in transiently versus
stably transfected cells was recapitulated by AP1903
(1b). Despite exhibiting excellent activity in transiently
transfected cells (EC50: 3 nM), AP1903 (1b) displayed
minimal activity in the stably transfected setting.
Figure 1. Structures of AP1510 and AP1903.
Chemistry
The general methods used for the preparation of syn-
thetic monomers and dimerizers 1a–1k have been repor-
ted.7,8,10 The synthesis of the urea dimer 1k is outlined in
Scheme 1. Pipecolyl ester 210 was treated with an acti-
vated carbamate obtained from reaction of 3-amino-
pentane with para-nitrophenyl chloroformate. The
resulting urea was then treated with TFA to remove the
tert-butyl ester and the resulting carboxylic acid coupled
via BOP with a half of an equivalent of ethylenediamine
to afford 1k.
One plausible hypothesis for this behavior of AP1903
(1b) is that it is related to the exceptionally high affi-
nity of the compound for F36V-FKBP. The monomer
subunit of AP1903 binds to the mutant protein with
an IC50 of 1.8 nM, but a direct Kd measurement for a
fluorescently derivative was determined as ꢁ100 pM.9
Similarly, FK1012 binds with high affinity to both
wild-type and F36V-FKBPs. We wondered whether
non-productive dimers, consisting of DNA binding
domains (BD–BD) or activation domains (AD–AD),
might fail to dissociate readily if induced by high-affi-
nity dimerizers. This would reduce the effective con-
centration of fusion proteins available to form
functional heterodimeric BD–AD complexes. In addi-
tion, active BD–AD complexes induced by high-affi-
nity dimerization of multivalent fusion proteins might
not dissociate on a timescale compatible with multi-
ple, rapid transcriptional initiation events at a pro-
moter. These problems would both be exacerbated in
stably transfected cell lines, in which expression levels
of the transcription factor fusion constructs are much
lower than in transiently transfected cells. To test this
hypothesis, we prepared a number of bumped dimer-
izers with reduced affinity for F36V-FKBP.
Results
The assay system for determining the ability of AP1903-
type bumped dimerizers to activate gene expression12 is
similar to that described previously for the unbumped
dimerizer AP1510.7,8 In short, HT1080 human fibro-
sarcoma cells are transiently or stably transfected with
constructs encoding separate DNA-binding domain and
activation domain proteins, each fused to three tandem
copies of F36V-FKBP. The cells additionally harbor a
SEAP (secreted alkaline phosphatase) reporter gene
responsive to the regulated transcription factor, so that
productive dimerization of the fusion proteins leads to
SEAP production. We tested both transiently and stably
transfected cells since these represent two frequent uses
of the regulated gene expression system, and differ in the
expression levels of the fusion proteins (which are much
higher in the transient system). Table 1 presents the
transcriptional activity of each dimerizer tested, reported
both as an EC50 and a peak expression level as a per-
centage of that of AP1889 (1f), the compound ultimately
One approach to attenuating the binding affinity of
dimerizers for F36V-FKBP is modification of the 3,4-
dimethoxyphenyl moiety of AP1903 (1b). However, simi-
lar to previous observations in the AP1510 series,8 we
found that this modification was far more detrimental
to transcriptional activity than to FKBP binding.
Scheme 1. Preparation of 1k.

W. Yang et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3181–3184
3183
Table 1. Affinities of F36V-FKBP12 monomeric ligands and in vitro transcriptional activation activities of synthetic dimerizers
Compd
R
R0
Monomer
IC50 (nM)a
Dimer induced transcription
EC50 (nM)b /% expressionc
Transient
Stable
FK1012
1a (AP1510)
12
44
189 (39%)
177 (149%)
No induction
294 (17%)
1b (AP1903)
1.8
5.5
15
3 (79%)
57 (97%)
82 (29%)
No induction
No induction
No induction
1c
1d
1e
8.2
40
4 (89%)
4 (100%)
6 (105%)
No induction
8 (100%)
1f (AP1889)
1g
1h
1i
26
9 (53%)
1530
8.2 and 40
3 (115%)
2 (68%)
1j
8.2 and 1530
20 (150%)
11 (138%)
10 (7%)
1k (AP1966)
64
19 (106%)
aThe IC50 is the concentration of monomeric ligand that reduces probe binding by 50% in a fluorescence polarization assay.9
bThe EC50 is the concentration of dimerizer that induces 50% peak SEAP reporter protein expression levels.
cWhere peak expression levels obtained with 1f (AP1889)=100%.
Compared to AP1903 (1b), replacement with a phenyl (1c)
or 3-pyridyl (1d) group caused 3- and 8-fold reductions in
monomer binding affinity to F36V-FKBP respectively,
but 20- and 30-fold reductions in EC50 (with lower levels
of SEAP production) in transiently transfected cells, and
no activity in stable cell lines (Table 1). An alternate
strategy for reducing affinity for F36V-FKBP is replace-
ment of the ethyl bump of AP1903 (1b) with the smaller
methyl group. This reduced monomeric binding affinity 5-
fold, but the resulting dimerizer 1e still exhibited tran-
scriptional activities equivalent to those of AP1903 (1b).
In an effort to more radically reduce binding affinity, the
entire 3,4,5-trimethoxyphenyl ‘bottom’ portion of
AP1903 (1b) was replaced by either a phenyl or a 3,4-
methylenedioxyphenyl group. Strikingly, this resulted in
significant improvements in transcriptional activity.
Dimerizer AP1889 (1f), whose monomer exhibits a 22-
fold reduced affinity for F36V-FKBP, not only proved
as effective as AP1903 (1b) in transiently transfected
cells (with a similar EC50, yet a higher level of protein
produced), but also exhibited excellent potency in stably
transfected cells, with an EC50 of 8 nM.12 The inter-

3184
W. Yang et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3181–3184
mediate affinity 3,4-methylenedioxyphenyl analogue 1g,
which exhibits a 14-fold reduced affinity for F36V-
FKBP, similarly proved as effective as AP1889 (1f),
although induced expression levels in stably transfected
cells were lower. All the dimerizers remained selectively
active on cells expressing F36V-FKBP fusion proteins
compared to wild-type FKBP fusions, demonstrating
that the modifications did not affect the binding specifi-
city of the engineered ‘bump’ (data not shown).
F36V, potently activates Fas-mediated apoptosis and
other signalling processes in appropriately engineered
cells.9,11 However the molecule is surprisingly inefficient
at activating transcription, only functioning in tran-
siently transfected cells where F36V-transcription factor
fusion proteins are present at high levels. The analysis
presented here indicates that while dimerizer potency in
signalling appears largely driven by binding affinity, the
structure activity relationship for transcription is more
subtle, with kinetic and/or geometric factors also con-
tributing. Indeed, we found that reducing the binding
affinity of AP1903 dramatically enhanced activity, sug-
gesting that tight binding prevents the efficient and
dynamic assembly of functional dimerized complexes.
These insights led to the design of novel, simpler
dimerizers such as AP1889 (1f) and AP1966 (1k) that
potently activate targeted gene expression in both tran-
siently and stably transfected cells expressing F36V-
FKBP fusion proteins. These homodimeric ligands,
engineered to bind minimally to endogenous FKBPs,
offer an alternative approach to the rapamycin-based
heterodimerizer system3,6 for inducing gene expression
in stable cell lines and transgenic animals.
With the availability of a dimerizer able to induce tran-
scription in stably transfected cells, we investigated the
activity of a heterodimeric compound 1i incorporating
one monomer unit of AP1889 (1f) and one of AP1903
(1b). We speculated that if slow dissociation of dimer-
ized protein complexes accounts for AP1903’s poor
activity, replacement of even one monomer with a
weaker binding moiety should still improve potency,
since non-productive homodimeric species could be
converted to productive ones by dissociation of only
one fusion protein followed by rebinding of another.
Indeed, 1i proved to be more potent than either AP1903
(1b) or AP1889 (1f) in both transiently and stably
transfected cells. Its EC50 of 2 nM in stably transfected
cells is the most potent to date of any dimerizer for the
activation of gene transcription. To assess how weak a
protein-ligand interaction is compatible with efficient
transcription, the AP1889 (1f) half of 1i was replaced
with the drastically weaker monomer of 1h (monomer
IC50: 1530 nM) to afford dimerizer 1j. Surprisingly, 1j
exhibited good, though 7-fold weaker, activity in tran-
siently transfected cells versus AP1903 (1b), and
retained some activity in stable cell lines. 1j also dis-
played a unique dose–response curve. Whereas increas-
ing doses of other dimerizers induced a peak level of
expression and then leveled off, 1j quickly dropped to
baseline after achieving its peak. This is most probably
due to an excess of the AP1903 subunit of 1j serving as a
self-antagonist through the accumulation of monomeric
(rather than dimeric) complexes.
References and Notes
1. Spencer, D. M.; Wandless, T. J.; Schreiber, S. L.; Crabtree,
G. R. Science 1993, 262, 1019.
2. (a) Klemm, J. D.; Schreiber, S. L.; Crabtree, G. R. Ann.
Rev. Immunol. 1998, 16, 569. (b) For a compilation of pub-
lications describing chemically-induced dimerization, see
http://www.ariad.com/regulationkits/reg_ref.html
3. Rivera, V. M.; Clackson, T.; Natesan, S.; Pollock, R.;
Amara, J. F.; Keenan, T.; Magari, S. R.; Phillips, T.; Courage,
N. L.; Cerasoli, F.; Holt, D. A.; Gilman, M. Nat. Med. 1996,
2, 1028.
4. Ho, S. N.; Biggar, S. R.; Spencer, D. M.; Schreiber, S. L.;
Crabtree, G. R. Nature 1996, 382, 822.
5. Pollock, R.; Clackson, T. Curr. Opin. Biotechnol. 2002, 13, 459.
6. Ye, X.; Rivera, V. M.; Zoltick, P.; Cerasoli, F., Jr.; Schnell,
M. A.; Gao, G.; Hughes, J. V.; Gilman, M.; Ailson, J. M.
Science 1999, 238, 88.
To further probe the role of binding affinity, and to
reduce the number of chiral centers and overall molec-
ular weight, we replaced the 2-arylbutyric acid bottom
portion of AP1903 with radically different structures. Of
particular interest was dimerizer AP1966 (1k), in which
the trimethoxyphenylbutyric acid bottom was replaced
with a 3-pentyl urea. While urea-based ligands of wild-
type FKBP had been reported in the literature,13 due to
the branched nature of the 3-pentyl group, the mono-
mer of 1k exhibited far greater selectivity for F36V-
FKBP than wt-FKBP (64 nM vs 14 mM). The resulting
dimerizer exhibited an EC50 only 2–3-fold reduced from
that of AP1889 (1f) in both transient and stable assays,
despite its radically simplified structure, and peak
expression levels were also greater than those for
AP1889 in both settings.
7. Amara, J. F.; Clackson, T.; Rivera, V. M.; Guo, T.;
Keenan, T.; Natesan, S.; Pollock, R.; Yang, W.; Courage,
N. L.; Holt, D. A.; Gilman, M. Proc. Natl. Acad. Sci. U.S.A.
1997, 94, 10618.
8. Keenan, T.; Yaeger, D. R.; Courage, N. L.; Rollins, C. T.;
Pavone, M.-E.; Rivera, V. M.; Yang, W.; Guo, T.; Amara,
J. F.; Clackson, T.; Gilman, M.; Holt, D. A. Bioorg. Med.
Chem. 1998, 6, 1309.
9. Clackson, T.; Yang, W.; Rozamus, L. W.; Amara, J. F.;
Hatada, M.; Rollins, C. T.; Stevenson, L. F.; Magari, S. R.;
Wood, S. A.; Courage, N. L.; Lu, X.; Cerasoli, F. J.; Gilman,
M.; Holt, D. A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 10437.
10. Yang, W.; Rozamus, L. W.; Narula, S.; Rollins, C. T.;
Yuan, R.; Andrede, L. J.; Ram, M. K.; Phillips, T. B.; Schra-
vendijk, M. R.; Dalgarno, D.; Clackson, T.; Holt, D. A. J.
Med. Chem. 2000, 43, 1135.
11. Jin, L.; Zeng, H.; Chien, S.; Otto, K. G.; Richard, R. E.;
Emery, D. W.; Blau, C. A. Nat. Genet. 2000, 26, 64.
12. Pollock, R.; Rivera, V. M. Meth. Enzymol. 1999, 306, 263.
13. Dragovich, P. S.; Barker, J. E.; French, J.; Imbacuan, M.;
Kalish, V. J.; Kissinger, C. R.; Knighton, D. R.; Lewis, C. T.;
Moomaw, E. W.; Parge, H. E.; Pelletier, L. A. K.; Prins, T. J.;
Showalter, R. E.; Tatlock, J. H.; Tucker, K. D.; Villafranca,
J. E. J. Med. Chem. 1996, 39, 1872.
Conclusion
AP1903 (1b), a dimerizer with engineered specificity and
tight binding affinity for the designed FKBP mutant