Small Molecules that Target the Toxic RNA in Myotonic Dystrophy Type 2

Article abstract of DOI:10.1002/cmdc.201402095

Myotonic dystrophy type 2 (DM2) is caused by an expansion of CCTG repeats in the zinc-finger protein gene (ZNF9). Transcribed CCUG repeats sequester muscleblind-like protein 1 (MBNL1), an important alternative splicing regulator, preventing its normal function, leading to the disease phenotype. We describe a series of ligands that disrupt the MBNL1-r(CCUG)_n interaction as potential lead agents for developing DM2 therapeutics. A previously reported triaminopyrimidine-acridine conjugate was a moderate inhibitor in vitro, however it proved to be poorly water-soluble and not cell-permeable. To improve its therapeutic potential, the new set of ligands maintained the key triaminopyrimidine recognition unit but replaced the acridine intercalator with a bisamidinium groove binder. The optimized ligands exhibit low micromolar inhibition potency to MBNL1-r(CCUG)₈. Importantly, the ligands are the first to show the ability to disrupt the MBNL1-r(CCUG)_n foci in DM2 model cell culture and exhibit low cytotoxicity. Targeting toxic RNA: A series of optimized ligands with two triaminopyrimidine recognition units linked to a bisamidinium groove binder is described. They exhibit low-micromolar inhibition potency to the MBNL1-r(CCUG)₈ interaction and are the first to show the ability to disrupt the MBNL1-r(CCUG)_n foci in DM2 model cell culture and exhibit low cytotoxicity.

Full text of DOI:10.1002/cmdc.201402095

CHEMMEDCHEM
FULL PAPERS
DOI: 10.1002/cmdc.201402095
Small Molecules that Target the Toxic RNA in Myotonic
Dystrophy Type 2
Lien Nguyen, JuYeon Lee, Chun-Ho Wong, and Steven C. Zimmerman*[a]
Myotonic dystrophy type 2 (DM2) is caused by an expansion of
CCTG repeats in the zinc-finger protein gene (ZNF9). Tran-
scribed CCUG repeats sequester muscleblind-like protein 1
(MBNL1), an important alternative splicing regulator, prevent-
ing its normal function, leading to the disease phenotype. We
describe a series of ligands that disrupt the MBNL1–r(CCUG)n
interaction as potential lead agents for developing DM2 thera-
to be poorly water-soluble and not cell-permeable. To improve
its therapeutic potential, the new set of ligands maintained the
key triaminopyrimidine recognition unit but replaced the acri-
dine intercalator with a bisamidinium groove binder. The opti-
mized ligands exhibit low micromolar inhibition potency to
MBNL1–r(CCUG)8. Importantly, the ligands are the first to show
the ability to disrupt the MBNL1–r(CCUG)n foci in DM2 model
cell culture and exhibit low cytotoxicity.
peutics.
A previously reported triaminopyrimidine–acridine
conjugate was a moderate inhibitor in vitro, however it proved
Introduction
Myotonic dystrophy type 1 and type 2 (DM1 and DM2, respec-
tively) are the most common forms of muscular dystrophy, and
both are estimated to affect 1 in 8000.[1] DM1 is caused by an
expansion of CTG repeats (CTGexp) in the 3’-untranslated region
of the DMPK gene on chromosome 19, whereas DM2 is associ-
ated with an expansion of CCTG repeats (CCTGexp) in the
intron 1 of the ZNF9 gene on chromosome 3.[1] The minimum
lengths of the repeats indicating the presence of DM1 and
DM2 diseases are 50 CTG and 75 CCTG repeats, respectively. At
this length, the sequences become unstable, progressively ex-
panding and reaching thousands of CTG or CCTG repeats in
the most severe cases of the disease.
Transcribed CUG and CCUG repeats (abbreviated rCUGexp
and rCCUGexp, respectively) sequester muscleblind-like (MBNL)
proteins, including an important alternative splicing regulator
MBNL1. A decrease in cellular MBNL1 levels results in splicing
defects of more than 100 pre-mRNAs, leading to the disease
symptoms, which include myotonia, muscle weakness, and cat-
aracts.[2] In addition, the rCCUGexp transcript appears to interact
with other proteins that are involved in translation.[2–4] Thus, it
was reported that the CUG binding protein (CELF1) and a trans-
lation initiation factor (eIF2) form a complex, and both are se-
questered by rCCUGexp as are the proteasome, events that lead
to a decrease in the global rate of protein synthesis.[3–5] The
rCCUGexp transcript may affect the expression levels of the
ZNF9 protein, and this may contribute to the DM2 phenotype;
however, the reports on this point are somewhat contradic-
tory.[4–8]
An important potential approach to treating DM2 involves
molecular intervention, the discovery and development of
either oligonucleotides or small molecules that selectively dis-
rupt the interaction of rCCUGexp with MBNL1 and other pro-
teins. This same general approach has been applied to DM1
and is considerably more advanced.[9–19] To our knowledge,
there are only two reports of small molecules that selectively
target the toxic rCCUGexp transcript causing DM2.[20,21] In 2009,
Disney and co-workers described a series of multivalent ligands
displaying kanamycin A units with varying separations (see 1,
n=1–3, m=0–19), and the best compound identified was the
[a] L. Nguyen, J. Lee, C.-H. Wong, Prof. S. C. Zimmerman
Department of Chemistry
University of Illinois at Urbana-Champaign
600 S. Mathews Ave., Urbana, IL 61801 (USA)
E-mail: sczimmer@illinois.edu
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201402095.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
1
&
ÞÞ
These are not the final page numbers!

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
trimer, which inhibited the
r(CCUG)47–MBNL1
interaction
with nanomolar IC50 values in
vitro.[20]
Recently, we described ligand
2, a rationally designed “stacked
intercalator” featuring a triamino-
pyrimidine
recognition
unit
linked to an acridine intercalator,
and showed that this simple
ligand binds the rCCUG repeat
sequences with high selectivity
and micromolar KD values.[21] Al-
though 2 inhibited formation of
the MBNL1–r(CCUG)6 complex
with a moderate inhibition po-
tency (IC50 =52 mm), we found it
to be toxic in cellular assays.
Herein, we report a new series
of triaminopyrimidine–bisamidi-
nium conjugates as potential
DM2 ligands that not only main-
tain the low micromolar inhibi-
tion potency of 2 for MBNL1–
rCCUGexp complexes but also
have improved water solubility
and cell uptake, and decreased
cytotoxicity. Importantly, to the
best of our knowledge, the opti-
mized DM2 ligands are the first
documented to dissolve the foci
of MBNL1 and rCCUGexp in a cel-
lular model of DM2.
Results and Discussion
Design of bisamidinium-based
DM2 ligands
The current design combines
knowledge gained in several
recent efforts at targeting both
DM1 and DM2. As illustrated in
Figure 1,
rCCUGexp-targeting
ligand 2, was based on ligand 3,
a triaminotriazine–acridine con-
jugate designed to bind UU and
TT mismatches selectively by for-
mation of a Janus wedge base
triplet.[16] Thus, the higher pKa of
the pyrimidine unit was pro-
posed to result in its protona-
tion, leading to a Janus wedge
moiety that is complementary to
the CU mismatch.[21] Support for
this idea comes from the finding
that, despite the structural simi-
Figure 1. Center: The designed Janus-wedge recognition of UU mismatch by triaminotriazine (DM1) and CU mis-
match by triaminopyrimidine (DM2). Bottom: Acridine motif and acridine-based DM1 and DM2 ligands (3 and 2,
respectively). Top: Bisamidinium motif and bisamidinium based DM1 and DM2 ligands (5 and 6, respectively).
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
2
&
ÞÞ
These are not the final page numbers!

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
larities of the triaminotriazine and triaminopyrimidine moieties,
ligands 2 and 3 hold their selectivity toward rCUGexp and rCCU-
Gexp, respectively. The selective targeting of rCCUGexp by the tri-
aminopyrimidine recognition unit is key to the current DM2
ligand design.
Synthesis of bisamidinium-based DM2 ligands
Following the prior success[19] with ligand 5 and based on the
other considerations outlined above, several analogous bisami-
dinium ligands for DM2 were synthesized with triaminopyrimi-
dine Janus wedge units for targeting the CU mismatches in
rCCUGexp (Figure 1).[21] The substitution pattern of the bisamidi-
nium motif may affect both its cellular penetration and locali-
zation,[22] and its specific fit to rCCUGexp. Therefore, ligands
with both para- and meta-substituted bisamidinium moieties
were prepared, with alkyl linkers varying from two to four
methylene groups to determine the optimum length.
Both ligands 2 and 3 contain an acridine intercalator, which
was expected to stack on the Janus wedge unit and provide
a hydrophobic driving force for binding, while decreasing the
nonspecific intercalation. This “stacked intercalator” approach
was successful in so far as the off-target binding of duplex
DNA was significantly suppressed. However, ligand 3 exhibited
unacceptable toxicity, and similar studies with 2 in cell culture
experiments revealed poor water solubility, low cell permeabili-
ty, and high toxicity (see Supporting Information).
Concern about the acridine unit led to its replacement using
ligand 4, a groove binding bisa-
Ligands 6–11 were prepared from the appropriate meta- or
para-bisimidate and the corresponding aminoalkylpyrimidine
(Scheme 1; for details, see Supporting Information). The amino-
midinium ion. The selection of 4
was based on a number of re-
ports showing bisamidinium
compounds to possess excellent
water solubility, high cellular
uptake,[22] and low cytotoxicity.[23]
Indeed, with the appropriate
substitution pattern, the com-
pounds targeted the nucleus.[22]
Equally importantly, Butcher and
co-workers reported the NMR
structure of bisamidinium ligand
4
(DB213, R=NMe2, Figure 1)
bound within the groove of the
HIV-1 frameshift-inducing stem-
loop RNA.[24] The groove binding
site adopts a remarkably similar
A-form conformation to that in
Scheme 1. Synthesis of ligands 6–11. Reagents and conditions: a) Et3N, EtOH, 258C, 24 h, 39–52%.
r(CUG)12, and the distance between the ammonium ions in
bound 4 (DB213, R=NMe2) is the same as between every third
UU mismatch in r(CUG)12.
alkylpyrimidines were prepared using a modification of the
preparation reported by Mascal and co-workers.[27] Briefly, the
corresponding dicyanoalkyl phthalimide reacted with guani-
dine carbonate, followed by the deprotection using hydrazine
(see Scheme S1 in the Supporting Information).
Ligand 5 containing the bisamidinium groove binder 4 and
two UU recognition units was found to exhibit low toxicity in
cellular assays, be well tolerated by mice, bind rCUGexp, dis-
solve nuclear MBNL1–rCUGexp foci in vitro, and show phenotyp-
ic improvements in a DM1 Drosophila model.[19] In considering
whether a similar strategy might be successful for DM2 (e.g.,
ligand 6, Figure 1), it was important to know the rCCUGexp
structure, and particularly whether it similarly adopts an A-
form. It is predicted that rCCUGexp might adopt two possible
structures: one with two consecutive CU mismatches or one
with alternating CC and UU mismatches; the latter RNA is re-
ferred to as the slipped form.[21,25] The former structure is pre-
dicted to be more stable, and Disney and co-workers recently
reported the first high-resolution X-ray analysis of an r(CCUG)3
duplex within a larger RNA sequence. The rCCUG sequence
indeed adopts an A-form structure with two CU mismatches
separated by two GC base pairs.[26]
Inhibition of MBNL1–r(CCUG)8 complex by DM2-targeting
small molecules in vitro
An electrophoretic mobility shift assay (EMSA) was used to de-
termine the inhibition potencies of the ligands toward
MBNL1–r(CCUG)8. To avoid the slipped rCCUG repeat from
forming and restrict the r(CCUG)8 structure to that with two
consecutive CU mismatches, an RNA oligomer was used with
complementary GC-rich regions both at the 3’- and 5’-ends
(Figure 2A). The binding affinity of MBNL1 for r(CCUG)8 was
also measured by EMSA. Thus, 32P-labelled r(CCUG)8 was incu-
bated with MBNL1, serially diluted, and the fraction of RNA
that is free or complexed measured following their electropho-
retic separation. The percent RNA bound by MBNL1 was plot-
ted against the MBNL1 concentration, and curve fitting provid-
ing a KD value of 3.9ꢀ0.8 nm (Figure 2B,C).
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
3
&
ÞÞ
These are not the final page numbers!

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
Table 1. Half-maximal inhibitory concentration (IC50) values and inhibition
constants (Ki) for ligands 6, 8, 10 and 11 towards MBNL1–r(CCUG)8.[a]
Compd
KD [nm]
IC50 [mm]
Ki [mm]
(MBNL1-r(CCUG)8)
6
8
10
11
3.9ꢀ0.8
73.4ꢀ6.0
12.9ꢀ1.0
14.7ꢀ1.2
24.9ꢀ2.0
8.7ꢀ1.9
1.5ꢀ0.3
1.7ꢀ0.4
2.9ꢀ0.6
[a] Data represent the meanꢀSD of three independent experiments
performed in duplicate.
longer or shorter linkers failed to inhibit the MBNL1–r(CCUG)8
complex effectively (Figure S2 in the Supporting Information).
Given these qualitative results, IC50 values were measured
for ligands 6, 8, 10, and 11 using quantitative EMSA with vary-
ing ligand concentration (Figure 3B,C). The corresponding in-
hibition constants (Ki) were in the low micromolar range
(Table 1). Interestingly, there is no significant difference in in-
hibition potency between meta- and para-bisamidinium com-
pounds (i.e., 8 vs 10, 6 vs 11). This result was unexpected be-
cause the substitution pattern changes both the distance the
recognition units can span as well as the fit of the bisamidini-
um unit within the RNA groove. In the absence of additional
structural data, it is difficult to explain the insensitivity to the
benzamidinium substitution pattern.
Figure 2. A) Structure of r(CCUG)8; B) MBNL1-r(CCUG)8 binding gel; C) Plot of
RNA fraction bound against log[MBNL1]. Error bars represent the standard
deviation of three independent experiments.
The inhibitory potency of the ligands was screened qualita-
tively by incubating each at 100 mm with MBNL1–r(CCUG)8.
Free and protein-complexed r(CCUG)8 were separated on a gel,
and their relative ratios were measured. Incubation of MBNL1–
r(CCUG)8 with compounds 6, 8, 10, and 11 that each contain
linkers with three or four methylene groups lead to full disrup-
tion of the MBNL1–r(CCUG)8 complex (Figure 3A). In contrast,
a significant amount of complex remained after incubation
with the two methylene linker-containing ligands 7 and 9, as
well as for negative control 4 (DB213, R=NMe2) lacking the tri-
aminopyrimidine rings. Likewise, the DM1-targeting ligand 5,
its meta-substituted analogue, and analogues containing
Lead DM2 compounds show low cytotoxicity toward HeLa
cells
As noted above, development of the triaminopyrimidine–acri-
dine conjugate, ligand 2, was limited by its high cytotoxicity. In
particular, it was found that treatment of DM2 model cells (see
below) with 2 at 50 mm led to more than 90% cell death and
Figure 3. A) Screening gel of bisamidinium-based DM1 and DM2 ligands. In this gel, [r(CCUG)8]=0.22 nm, [MBNL1]=33 nm, [ligand]=100 mm; B) Inhibition
gel of ligand 10. Briefly, MBNL1–r(CCUG)8 complex was incubated with 10 at different concentrations (200 mm to 50 nm, twofold dilution); C) The inhibition
curve of ligand 10; the RNA fraction bound was normalized using the control. Error bars represent the standard deviation of at least three independent
experiments.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
4
&
ÞÞ
These are not the final page numbers!

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
a change in cell morphology with a presence of disease foci
after 48 h (Figure S6 in the Supporting Information). In con-
trast, a low cytotoxicity was observed when the bisamidinium-
based ligands 6, 8, 10, and 11 at 100 mm were incubated with
the DM2 model cells for 48 h.
tent with foci disruption, no direct observation of small mole-
cules disrupting foci in DM2 cell culture have been reported. It
is known that induced overexpression of MBNL1 prevents
muscle degeneration in a DM1 fly model and reverses the spli-
ceopathy and myotonia in a DM1 mouse model.[29–31] Impor-
tantly, the ligands targeting toxic CUG repeats and disrupting
the MBNL–CUGexp interaction in the nucleus reverse DM1 phe-
notypes.[9–19]
To determine more quantitatively whether these ligands ex-
hibited an improved cytotoxicity profile in comparison with
ligand 2, a sulforhodamine B (SRB) colorimetric assay[28] with
HeLa cells was performed. The percentage of cell death was
plotted as a function of ligand concentration (Figure 4; see
also Figure S5 in the Supporting Information). para-Bisamidini-
To examine whether the in vitro activity of lead ligands 6, 8,
10, and 11 translates into cellular activity, the ability to dissolve
foci in DM2 model cells was measured by fluorescence in situ
hybridization (FISH) experiments. Thus, HeLa cells were co-
transfected with GFP-MBNL1 and (CCTG)1200 plasmids to gener-
ate a DM2 cell model.[3,4,26] The locations of MBNL1 and the
r(CCUG)1200 transcripts were readily tracked by confocal micros-
copy, monitoring the fluorescent signal of GFP and a Cy3-
(CAGG)8 FISH probe, respectively (Figure 5). The rCCUG repeats
were found to localize in both nuclei and cytoplasm, which is
consistent with a report by Timchenko and co-workers.[4] There
was a clear co-localization of GFP-MBNL1 and r(CCUG)1200
within the disease foci, providing evidence of MBNL1 seques-
tration by rCCUG repeats (row 2 in Figure 5). Additionally, there
were some r(CCUG)1200 foci not co-localized with GFP-MBNL1,
which might suggest a limited co-transfection of the protein
or possibly the r(CCUG)1200 transcript bound by other protein
complexes, including the 20S catalytic core and the CELF1–
eIF2 complex.[3,4]
Whereas 4 (DB213, R=NMe2) showed almost no foci disper-
sion with the co-localization of Cy3 and GFP signals (row 3 in
Figure 5), all of the bisamidinium ligands that inhibited the
MBNL1–r(CCUG)8 complex in vitro (i.e., 6 and 11) dissolved the
disease foci (rows 4–7 in Figure 5; for more images with
a larger field, see the Supporting Information) with a significant
decrease in the number and size of the foci. In particular, treat-
ment of DM2 model cells with compounds 6, 8, 10, and 11 at
100 mm resulted in 50–70% foci disruption (Figure 6). There
was no correlation between the Ki value and the extent of foci
disruption, which is understandable given that many other
factors govern the cellular activity.
Figure 4. Example of cytotoxicity curves for ligands 2 and 10; see also Fig-
ure S5 in the Supporting Information. The IC50 values were estimated from
the plots. Error bars represent standard deviation of three independent
experiments.
Conclusions
um conjugates 6 and 8 were observed to have low cytotoxicity
with between 15 and 30% cell death observed. Interestingly,
the meta-bisamidinium ligands, compounds 10 and 11, exhibit-
ed essentially no cytotoxicity at concentrations as high as
100 mm.
A documented approach for the treatment of myotonic dystro-
phy is through small molecules that target CUG or CCUG re-
peats and disrupt their interaction with MBNL1 and other pro-
teins. Herein, we have reported a new set of ligands that inhib-
it the MBNL1–rCCUGexp interaction. The ligands all have low
molecular weights (MW ꢁ520) and contain two important
units: a triaminopyrimidine moiety for targeting the CU mis-
match and a bisamidinium moiety. The bisamidinium unit mini-
mally provides a general A-form RNA-targeting ability and con-
fers excellent water solubility and cellular uptake. The opti-
mized ligands show promising bioactivities, inhibiting MBNL1–
r(CCUG)8 with low micromolar Ki values, very low toxicity in
HeLa cells, and the ability to dissolve the disease foci in DM2
model cells. Indeed, the low cytotoxicity of ligands 6, 8, 10, 11
and the binding selectivity of ligand 2 toward a sequence con-
Foci disruption by small molecules in a DM2 cell model
The hallmark of DM2 disease is the sequestration of MBNL by
rCCUGexp. Immunofluorescence experiments showed a co-local-
ization of MBNL1 protein with rCCUGexp in DM2 patient cells in
clearly observable punctate foci.[29] With a similar pathogenesis
to DM1, it is believed that the DM2 phenotype might be re-
versed by ligands able to disrupt the interaction of MBNL1–
rCCUGexp in cells, thereby liberating MBNL1 to function normal-
ly. Although a report of downstream splicing rescue is consis-
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
5
&
ÞÞ
These are not the final page numbers!

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
Experimental Section
Biology
Materials: r(CCUG)8 (5’-GCCCUGC-
CUGCCUGCCUGCCUGCCUGCCUGC-
CUGGC-3’) was purchased from
Dharmacon RNAi Technologies (GE
Healthcare) with 2’-deprotection,
desalting, and HPLC purification.
The r(CCUG)8 was dissolved in
Ambionꢁ (Life Technologies) RNA
storage solution (pH 6.4) to give
1 mm stock solution.
Protein expression and purification:
An expression vector for a truncat-
ed MBNL1 contains amino acids 1–
272. This MBNL is comprised of the
four zinc-finger motifs of MBNL1
and a His6, which has a similar af-
finity as the full-length MBNL1. The
protein was expressed and purified
following the procedure described
previously.[32]
RNA radioactive labeling: r(CCUG)8
was labeled with [g-32P]-ATP using
T4 polynucleotide kinase (New
England Biolabs) and purified by
phenol extraction and ethanol pre-
cipitation. Labeled RNA was
heated at 958C for 5 min, placed
on ice for 10 min, and diluted in
a binding buffer (175 mm NaCl,
5 mm MgCl2, 20 mm Tris (pH 7.5),
1.25 mm
2-mercaptoethanol,
12.5% glycerol (v/v), 2 mgmLꢂ1
bovine serum albumin (BSA),
0.1 mgmLꢂ1 heparin, 0.05% Triton-
X (v/v)).
Binding experiments: r(CCUG)8 at
a final concentration of 0.22 nm
was incubated with MBNL1 at seri-
ally diluted concentrations (3.3 mm
to 2.1 pm). The reaction mixture
was incubated at RT for 15 min,
and loaded on 6% polyacrylamide
gel (80:1) at 48C. The gels were
run for 2 h at 200 V in 0.5m Tris-
borate buffer (pH 8.2). Gels were
visualized on a Molecular Dynam-
ics Storm PhosphorImager. The ap-
parent KD was obtained by fitting
fraction RNA bound versus protein
concentration using the equation:
Figure 5. Confocal microscopy images of foci dispersion. DM2 model cells were treated with compounds 4, 6, 8,
10, 11 at 100 mm, 48 h; r(CCUG)n was probed by Cy3-(CAGG)8; the nucleus was stained by To-Pro 3.
Fraction
RNA
bound=
1/(1+KD/[MBNL1]). The standard
deviation was obtained from three independent experiments.
taining CU mismatches[21] suggest a specificity of bisamidinium
ligands towards rCCUGexp. Our current efforts are focused on
improving these already promising properties with an eye
toward testing a second-generation series in Drosophila and
mouse models.
Screening EMSA experiments: Each ligand at 100 mm was incubated
with pre-incubated MBNL1–r(CCUG)8 complexes at RT for 15 min.
Final concentrations of MBNL1 and r(CCUG)8 were 33 nm and
0.22 nm, respectively. The reaction mixture was loaded on the gel
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
6
&
ÞÞ
These are not the final page numbers!

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
Chemistry
Triaminopyrimidine–bisamidinium conjugates were synthesized fol-
1
lowing Scheme S1 in the Supporting Information. H NMR spectra
were recorded on a Varian Unity Inova 500 spectrometer. All NMR
measurements were carried out in [D6]DMSO at RT unless other-
wise stated. Chemical shifts are reported as parts per million
(ppm). Mass spectra were obtained in the Mass Spectrometry Lab-
oratory, School of Chemical Sciences, University of Illinois, USA. The
analytical HPLC traces were obtained using a C18 column with
0.1% TFA in 1:1 (v/v) MeOH/H2O as eluent. All reactions were car-
ried out under a nitrogen atmosphere. All solvents and reagents
were purchased commercially with a reagent quality and used
without further purification. The progress of reaction was moni-
tored by thin-layer chromatography (TLC) using Merck pre-coated
Figure 6. Plotting percent normalized foci area per against different treat-
ment conditions. DM2 model cells were treated with compounds 4, 6, 8, 10,
11 at 100 mm. Error bars represent the standard error of the mean of at least
three independent experiments. *** indicates a statistically significant differ-
ence in comparison to untreated DM2 model cells, P<0.001.
silica gel 60F254
.
Ligand 2 was prepared using the procedure reported by Wong
and co-workers, the characterization data was identical to that re-
ported.[21] The synthesis of intermediates can be found in the Sup-
porting Information. HPLC traces of triaminopyrimidine–bisamidine
conjugates show a purity of ꢃ95% (see Supporting Information).
N1,N4-Bis(4-(2,4,6-triaminopyrimidin-5-yl)butyl)terephthalimida-
mide hydrochloride salt (6): Et3N (0.14 mL, 1.0 mmol) was added
to diethyl terephthalimidate dihydrochloride (150 mg, 0.5 mmol) in
EtOH (10 mL) cooled in an ice bath. A solution of 5-(4-aminobutyl)-
pyrimidine-2,4,6-triamine (215 mg, 1.1 mmol) in EtOH (3 mL) and
ethylene glycol (3 mL) was added dropwise. The mixture was
warmed to RT and stirred for 24 h. The reaction was monitored by
TLC using 7:3 H2O/HOAc as eluent. The solvent was removed in va-
cuo. The residue was dissolved in MeOH and loaded onto a silica
gel flash column and purified by chromatography (MeOH/CH2Cl2
1:4!4:1). The starting material and a side product were eluted,
and the desired product was eluted with 4:1 MeOH/CH2Cl2 contain-
ing 0.04–0.08 mL of a 4m aq HCl solution in 1,4-dioxane per liter
of solvent. The desired product fractions were combined, and the
solvent was removed in vacuo to afford 6 as a light yellow solid,
a hydrochloride salt (160 mg, 48%): mp: >2308C (decomp.);
1H NMR (500 MHz): d=10.36 (brs, NH2, 2H), 9.79 (brs, NH2, 2H),
9.45 (brs, NH2, 2H), 7.99 (s, ArH, 4H), 7.15 (brs, NH2, 12H), 3.49
(brs, CH2, 4H), 2.33 (brs, CH2, 4H), 1.75 (brs, CH2, 4H), 1.39 ppm
(brs, CH2, 4H); MS (ESI): m/z (%): 521.1 (100) [M+H]+.
as described above. Fraction RNA bound was compared to the
control samples (H2O or DMSO) to determine the hits under the
testing conditions.
Inhibition EMSA experiments: Inhibition experiments were done as
described above except MBNL1–r(CCUG)8 complexes were incubat-
ed with ligands to give a serially diluted concentrations (from
200 mm to 50 nm). The IC50 values were obtained by fitting fraction
RNA bound versus the log function of ligand concentrations using
the equation: Y=B+(AꢂB)/(1+10ðXꢂlogðIC ÞÞ), in which Y=fraction
50
RNA bound, B=minimum fraction RNA bound, A=maximum frac-
tion RNA bound, X=log [ligand]. The inhibition constants (Ki) were
obtained using the equation: Ki =IC50 ꢂKD(rCCUG8-MBNL1)/
[MBNL1]. Standard deviations were obtained from three independ-
ent experiments.
Cytotoxicity study: Toxicity of ligands in HeLa cells was studied fol-
lowing the protocol reported by Vichai and Kirtikara.[28] Briefly, in
a 96-well plate, HeLa cells were incubated at 378C for 72 h with
the treatment of the ligand. The ligand was added to each well to
have serially diluted concentrations from 100 mm to 5.1 nm. The
cells were fixed after 3 d, and stained with sulforhodamine B (SRB).
Bound SRB was dissolved by Tris base solution (pH 10.5) and the
OD was measured at 510 nm. IC50 values of toxicity were obtained
by plotting the percentage of cell death against concentrations of
ligands.
N1,N4-Bis(2-(2,4,6-triaminopyrimidin-5-yl)ethyl)terephthalimida-
mide hydrochloride salt (7): The desired compound was prepared
using a procedure analogous to that described for 6 using diethyl
terephthalimidate dihydrochloride (217 mg, 0.7 mmol) in EtOH
(5 mL), Et3N (0.2 mL, 1.4 mmol), and 5-(2-aminoethyl)pyrimidine-
2,4,6-triamine (309 mg, 1.8 mmol) to afford of 7 as a light yellow
solid (240 mg, 52%): 1H NMR (500 MHz): d=10.37 (brs, NH2, 2H),
9.75 (brs, NH2, 2H), 9.64 (brs, NH2, 2H), 8.04 (s, ArH, 4H), 6.85 (br s,
NH2, 8H), 6.54 ((br s, NH2, 4H), 3.49 (brt, J=2.5Hz, CH2, 4H),
2.71 ppm (brt, J=7.5 Hz, CH2, 4H); MS (ESI): m/z (%): 465.3 (100)
[M+H]+.
Fluorescence in situ hybridization (FISH): FISH experiments were per-
formed following the protocol described previously.[12] Briefly, HeLa
cells at 80% confluence were co-transfected with (CCTG)1200 and
GFP-MBNL1 plasmids. Transfected HeLa cells were treated with li-
gands 4, 6, 8, 10, and 11 at 100 mm, 48 h. rCCUG repeats were
probed with Cy3-(CAGG)8, and the nuclei were stained with To-
Pro 3. The cells were looked at under confocal microscope LSM
700 (Zeiss). GFP, Cy3 and To-Pro 3 were excited at 488, 564, and
639 nm, respectively. Four random areas were selected to deter-
mine the average foci area per cell, using Teton software (ver-
sion 4.7). The average foci area per cell of untreated cells was set
to 100%. Each experiment was repeated at least three times inde-
pendently.
N1,N4-Bis(3-(2,4,6-triaminopyrimidin-5-yl)propyl)terephthalimida-
mide hydrochloride salt (8): The desired compound was prepared
using a procedure analogous to that described for 6 using diethyl
terephthalimidate dihydrochloride (150 mg, 0.5 mmol) in EtOH
(10 mL), Et3N (0.14 mL, 1.0 mmol), and 5-(3-aminopropyl)pyrimi-
dine-2,4,6-triamine (205 mg, 1.1 mmol) to afford 8 as a light yellow
solid (150 mg, 47%): 1H NMR (500 MH): d=10.69 (brs, NH2, 2H),
10.30 (brs, NH2, 2H), 9.72 (brs, NH2, 2H), 7.93 (s, ArH, 4H), 7.25
(brs, NH2, 8H), 7.10 (brs, NH2, 4H), 3.13 (brs, CH2, 4H), 1.61 (brs,
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
7
&
ÞÞ
These are not the final page numbers!

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
[1] O. Raheem, S.-E. Olufemi, L. L. Bachinski, A. Vihola, M. Sirito, J. Holm-
lund-Hampf, H. Haapasalo, Y.-P. Li, B. Udd, R. Krahe, Am. J. Pathol. 2010,
177, 3025–3036.
[2] B. Schoser, L. Timchenko, Curr. Genomics 2010, 11, 77–90.
[3] E. Salisbury, B. Schoser, C. Schneider-Gold, G.-L. Wang, C. Huichalaf, B.
Jin, M. Sirito, P. Sarkar, R. Krahe, N. A. Timchenko, L. T. Timchenko, Am. J.
Pathol. 2009, 175, 748–762.
[4] C. Huichalaf, B. Schoser, C. Schneider-Gold, B. Jin, P. Sarkar, L. Timchen-
ko, J. Neurosci. 2009, 29, 9042–9049.
[5] R. Massa, M. B. Panico, S. Caldarola, F. R. Fusco, P. Sabatelli, C. Terraccia-
no, A. Botta, G. Novelli, G. Bernardi, F. Loreni, Neuropathol. Appl. Neuro-
biol. 2010, 36, 275–284.
[6] A. Botta, S. Caldarola, L. Vallo, E. Bonifazi, D. Fruci, F. Gullotta, R. Massa,
G. Novelli, F. Loreni, Biochim. Biophys. Acta Mol. Basis Dis. 2006, 1762,
329–334.
[7] J. M. Margolis, B. G. Schoser, M. L. Moseley, J. W. Day, L. P. W. Ranum,
Hum. Mol. Genet. 2006, 15, 1808–1815.
[8] W. Chen, Y. Wang, Y. Abe, L. Cheney, B. Udd, Y.-P. Li, J. Mol. Biol. 2007,
368, 8–17.
[9] P. C. Gareiss, K. Sobczak, B. R. McNaughton, P. B. Palde, C. A. Thornton,
B. L. Miller, J. Am. Chem. Soc. 2008, 130, 16254–16261.
[10] L. O. Ofori, J. Hoskins, M. Nakamori, C. A. Thornton, B. L. Miller, Nucleic
Acids Res. 2012, 40, 6380–6390.
[11] J. L. Childs-Disney, J. Hoskins, S. G. Rzuczek, C. A. Thornton, M. D. Disney,
ACS Chem. Biol. 2012, 7, 856–862.
[12] M. D. Disney, M. M. Lee, A. Pushechnikov, J. L. Childs-Disney, ChemBio-
Chem 2010, 11, 375–382.
[13] M. M. Lee, J. L. Childs-Disney, A. Pushechnikov, J. M. French, K. Sobczak,
C. A. Thornton, M. D. Disney, J. Am. Chem. Soc. 2009, 131, 17464–
17472.
[14] Z. Gao, T. A. Cooper, Hum. Gene Ther. 2013, 24, 499–507.
[15] M. B. Warf, M. Nakamori, C. M. Matthys, C. A. Thornton, J. A. Berglund,
Proc. Natl. Acad. Sci. USA 2009, 106, 18551–18556.
[16] J. F. Arambula, S. R. Ramisetty, A. M. Baranger, S. C. Zimmerman, Proc.
Natl. Acad. Sci. USA 2009, 106, 16068–16073.
[17] A. H. Jahromi, L. Nguyen, Y. Fu, K. A. Miller, A. M. Baranger, S. C. Zimmer-
man, ACS Chem. Biol. 2013, 8, 1037–1043.
[18] A. H. Jahromi, Y. Fu, K. A. Miller, L. Nguyen, L. M. Luu, A. M. Baranger,
S. C. Zimmerman, J. Med. Chem. 2013, 56, 9471–9481.
[19] C.-H. Wong, L. Nguyen, J. Peh, L. M. Luu, J. S. Sanchez, S. L. Richardson,
T. Tuccinardi, H. Tsoi, H. Y. E. Chan, W. Y. Chan, A. M. Baranger, P. J. Her-
genrother, S. C. Zimmerman, J. Am. Chem. Soc. 2014, 136, 6355–6361.
[20] M. M. Lee, A. Pushechnikov, M. D. Disney, ACS Chem. Biol. 2009, 4, 345–
355.
[21] C.-H. Wong, Y. Fu, S. R. Ramisetty, A. M. Baranger, S. C. Zimmerman, Nu-
cleic Acids Res. 2011, 39, 8881–8890.
[22] A. Lansiaux, F. Tanious, Z. Mishal, L. Dassonneville, A. Kumar, C. E. Ste-
phens, Q. Hu, W. D. Wilson, D. W. Boykin, C. Bailly, Cancer Res. 2002, 62,
7219–7229.
[23] M. Hung, P. Patel, S. Davis, S. R. Green, J. Virol. 1998, 72, 4819–4824.
[24] R. J. Marcheschi, M. Tonelli, A. Kumar, S. E. Butcher, ACS Chem. Biol.
2011, 6, 857–864.
[25] M. B. Warf, J. A. Berglund, RNA 2007, 13, 2238–2251.
[26] J. L. Childs-Disney, I. Yildirim, H. Park, J. R. Lohman, L. Guan, T. Tran, P.
Sarkar, G. C. Schatz, M. D. Disney, ACS Chem. Biol. 2014, 9, 538–550.
[27] M. Mascal, J. Hansen, P. S. Fallon, A. J. Blake, B. R. Heywood, M. H.
Moore, J. P. Turkenburg, Chem. Eur. J. 1999, 5, 381–384.
[28] V. Vichai, K. Kirtikara, Nat. Protoc. 2006, 1, 1112 – 1116.
[29] R. J. Osborne, C. A. Thornton, Hum. Mol. Genet. 2006, 15, R162–R169.
[30] M. de Haro, I. Al-Ramahi, B. De Gouyon, L. Ukani, A. Rosa, N. A. Faustino,
T. Ashizawa, T. A. Cooper, J. Botas, Hum. Mol. Genet. 2006, 15, 2138–
2145.
CH2, 4H), 1.17 ppm (brs, CH2, 4H). MS (ESI): m/z (%): 493.6 (100)
[M+H]+.
N1,N3-Bis(2-(2,4,6-triaminopyrimidin-5-yl)ethyl)isophthalimida-
mide hydrochloride salt (9): Et3N (0.25 mL, 1.8 mmol) was added
to diethyl isophthalimidate dihydrochloride (255 mg, 0.87 mmol) in
EtOH (10 mL) cooled in an ice bath. A solution of 5-(2-aminoethyl)-
pyrimidine-2,4,6-triamine (309 mg, 2.4 mmol) in EtOH (2 mL) and
ethylene glycol (2 mL) was added dropwise. The reaction mixture
was warmed to RT and stirred for 24 h. The reaction was monitored
by TLC using 7:3 H2O/HOAc as eluent. The solvent was removed in
vacuo. The residue was dissolved in MeOH and loaded onto a silica
gel flask column and purified by chromatography (MeOH/CH2Cl2
1:4!4:1). The starting material and a side product were eluted
and the desired product was eluted with 4:1 MeOH/CH2Cl2 contain-
ing 0.04–0.08 mL of a 4m aq HCl solution in 1,4-dioxane per liter
of solvent. The desired product fractions were combined, and the
solvent was removed in vacuo to afford 9 as a HCl salt as pale
yellow solid (265 mg, 50%): mp: >2308C (decomp.); 1H NMR
(500 MHz): d=10.36 (brs, ArCNH, 2H), 9.82 (brs, ArC=N+H2, 2H),
9.63 (brs, ArC=N+H2, 2H), 8.47 (s, ArH, 1H), 8.13 (dd, J=7.8, 1.8 Hz,
ArH, 2H), 7.81 (t, J=7.9 Hz, ArH, 1H), 7.01 (brs, CNH2C, 8H), 6.77
(brs, NCNH2C, 4H), 3.48 (brt, J=7.4 Hz, NH2CH2, 4H), 2.74 ppm
(brt, J=7.6 Hz, CCH2CH2, 4H); MS (ESI): m/z (%): 465.3 (100)
[M+H]+, 233.1 (20) [M+H]2+
.
N1,N3-Bis(3-(2,4,6-triaminopyrimidin-5-yl)propyl)isophthalimida-
mide hydrochloride salt (10): The desired compound was synthe-
sized following a procedure analogous to that described for 9
using diethyl isophthalimidate dihydrochloride (150 mg, 0.5 mmol)
in EtOH (10 mL), Et3N (0.14 mL, 1.0 mmol) and 5-(3-aminopropyl)-
pyrimidine-2,4,6-triamine (200 mg, 1.1 mmol) to afford 10 as a pale
1
yellow solid (120 mg, 39%): H NMR (500 MHz): d=10.25 (brs, NH2,
2H), 9.78 (brs, NH2, 2H), 9.49 (brs, NH2, 2H), 8.41 (s, ArH, 1H), 8.06
(dd, J=6.5, 1.5 Hz, ArH, 2H), 7.81 (dt, J=4, 1.5 Hz, ArH, 1H), 6.40
(brs, NH2, 8H), 6.01 (brs, NH2, 4H), 3.48 (brt, J=6.5 Hz, CH2, 4H),
2.38 (brt, J=8.5 Hz, CH2, 4H), 1.67 ppm (brq, J=3 Hz, CH2, 4H);
MS (ESI): m/z (%): 493.6 (100) [M+H]+.
N1,N3-Bis(4-(2,4,6-triaminopyrimidin-5-yl)butyl)isophthalimida-
mide hydrochloride salt (11): The desired compound was synthe-
sized following a procedure analogous that described for 9 using
diethyl isophthalimidate dihydrochloride (150 mg, 0.5 mmol) in
EtOH (10 mL), Et3N (0.14 mL, 1.0 mmol) and 5-(4-aminobutyl)pyrimi-
dine-2,4,6-triamine (220 mg, 1.1 mmol) to obtain 11 as a pale
1
yellow solid (130 mg, 39%): H NMR (500 MHz): d=10.35 (brs, NH2,
2H), 9.87 (brs, NH2, 2H), 9.47 (brs, NH2, 2H), 8.43 (s, ArH, 1H), 8.09
(dd, J=7, 1.5 Hz, ArH, 2H), 7.78 (t, J=7.5 Hz, ArH, 1H), 7.21 (brs,
NH2, 12H), 3.49 (brs, CH2, 4H), 2.32 (brs, CH2, 4H), 1.78 (brs, CH2,
4H), 1.42 ppm (brs, CH2, 4H); MS (ESI): m/z (%): 521.1 (100)
[M+H]+.
Acknowledgements
The authors thank Dr. Partha S. Sarkar (University of Texas Medi-
cal Branch at Galveston, USA) for providing the (CCTG)1200 plas-
mids. This work was supported by the US National Institutes of
Health (NIH) (RO1 AR058361).
[31] C. M. Chamberlain, L. P. W. Ranum, Hum. Mol. Genet. 2012, 21, 4645–
4654.
[32] Y. Yuan, S. A. Compton, K. Sobczak, M. G. Stenberg, C. A. Thornton, J. D.
Griffith, M. S. Swanson, Nucleic Acids Res. 2007, 35, 5474–5486.
Keywords: bisamidinium
dystrophy · triaminopyrimidine
· DM2 therapeutics · myotonic
Received: March 31, 2014
Published online on && &&, 0000
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
8
&
ÞÞ
These are not the final page numbers!

FULL PAPERS
Targeting toxic RNA: A series of opti-
mized ligands with two triaminopyrimi-
dine recognition units linked to a bisami-
dinium groove binder is described. They
exhibit low-micromolar inhibition po-
tency to the MBNL1–r(CCUG)8 interac-
tion and are the first to show the ability
to disrupt the MBNL1–r(CCUG)n foci in
DM2 model cell culture and exhibit low
cytotoxicity.
L. Nguyen, J. Lee, C.-H. Wong,
S. C. Zimmerman*
&& – &&
Small Molecules that Target the Toxic
RNA in Myotonic Dystrophy Type 2
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 9
&
9
&
ÞÞ
These are not the final page numbers!