Published online 20 April 2014 Nucleic Acids Research, 2014, Vol. 42, No. 10 6591–6602 doi: 10.1093/nar/gku275 Lomofungin and dilomofungin: inhibitors of MBNL1-CUG RNA binding with distinct cellular effects Jason W. Hoskins1, Leslie O. Ofori2, Catherine Z. Chen3,AmitKumar4, Krzysztof Sobczak1, Masayuki Nakamori1, Noel Southall3, Samarjit Patnaik3, Juan J. Marugan3, Wei Zheng3, Christopher P. Austin3, Matthew D. Disney4, Benjamin L. Miller5 and Charles A. Thornton1,* 1Department of Neurology, University of Rochester, Rochester, NY 14642, USA, 2Department of Chemistry, University of Rochester, Rochester, NY 14642, USA, 3Division of Pre-clinical Innovation, National Center for Advancing Translational Sciences, National Institutes of Health, Bethesda, MD 20892, USA, 4Department of Chemistry, Scripps Florida, Jupiter, FL 33458, USA and 5Department of Dermatology, University of Rochester, Rochester, NY 14642, USA Received May 8, 2013; Revised March 21, 2014; Accepted March 24, 2014 ABSTRACT INTRODUCTION Myotonic dystrophy type 1 (DM1) is a dominantly Myotonic dystrophy type 1 (DM1) is the prototypical ge- inherited neuromuscular disorder resulting from ex- netic disease resulting from a toxic gain-of-function by mu- pression of RNA containing an expanded CUG repeat tant RNA (1). RNA toxicity has been implicated in several (CUGexp). The pathogenic RNA is retained in nuclear other disorders that share a distinctive set of genetic fea- foci. Poly-(CUG) binding proteins in the Muscleblind- tures: the causal mutations are large polymorphic expan- sions of simple tandem repeats that are genetically unsta- like (MBNL) family are sequestered in foci, caus- ble and located in non-coding regions of the genome (2–9). ing misregulated alternative splicing of specific pre- The underlying disease mechanisms are heterogeneous and exp mRNAs. Inhibitors of MBNL1-CUG binding have postulated to include activation of signalling pathways by been shown to restore splicing regulation and cor- mutant RNA, aberrant translation that initiates within the rect phenotypes in DM1 models. We therefore con- repeat tract without an AUG start codon and sequestration ducted a high-throughput screen to identify novel of proteins that bind to long tracts of CUG, CCUG, GGC- inhibitors of MBNL1-(CUG)12 binding. The most ac- CUG, GGGGCC, CGG, AUUCU or UGGAA repeats [re- tive compound was lomofungin, a natural antimi- cently reviewed by Krzyzosiak et al. (10)]. crobial agent. We found that lomofungin undergoes In DM1 the expanded sequence is a (CTG)n repeat in spontaneous dimerization in DMSO, producing dilo- the 3’ untranslated region (UTR) of DM protein kinase mofungin, whose inhibition of MBNL1–(CUG) bind- (DMPK), a gene expressed in muscle, heart and brain (11). 12 When pathologically expanded beyond 50 repeats, the CTG ing was 17-fold more potent than lomofungin itself. tract is highly unstable in germline and somatic cells, which However, while dilomofungin displayed the desired leads to DMPK alleles having several thousand repeats (12). binding characteristics in vitro, when applied to cells The DMPK mRNA containing an expanded CUG repeat it produced a large increase of CUGexp RNA in nu- (CUGexp) is retained in the nucleus in foci (13,14). Splic- clear foci, owing to reduced turnover of the CUGexp ing factors in the Muscleblind-like (MBNL) family, which transcript. By comparison, the monomer did not in- are the predominant r(CUG)n binding proteins in mam- duce CUGexp accumulation in cells and was more malian cells, are sequestered in the foci of CUGexp RNA effective at rescuing a CUGexp-induced splicing de- (15,16). The resulting loss of MBNL function causes mis- fect. These results support the feasibility of high- regulated alternative splicing and other changes of the mus- throughput screens to identify compounds target- cle transcriptome (17–19). For example, mis-splicing of in- ing toxic RNA, but also demonstrate that ligands for sulin receptor and chloride ion channel 1 lead to insulin resis- tance and muscle hyperexcitability (myotonia), respectively repetitive sequences may have unexpected effects (20,21). on RNA decay. As yet there are no disease-modifying treatments for DM1 or other RNA dominant disorders. However, sev- eral compounds or oligonucleotides have shown benefi- *To whom correspondence should be addressed. Tel: +1 585 275 2542; Fax: +1 585 273 1255; Email: Charles [email protected] C The Author(s) 2014. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. 6592 Nucleic Acids Research, 2014, Vol. 42, No. 10 cial effects in cell or animal models (22–31). For small anti-His6 antibody that bound to MBNL1-105-His6,and molecules the predominant approach has been to iden- the fluorescence acceptor was XL665 conjugated to strep- tify, from the repertoire of known nucleic acid binders, a tavidin that bound to biotin-(CUG)12. The confirmatory set of compounds that show preferential binding to CUG assay employed a different detection system, consisting of repeats (22,27–29,32–34). CUG-binding compounds have AlphaScreen donor beads (conjugated to streptavidin) that also been assembled using dynamic combinatorial libraries emitted a singlet oxygen to activate AlphaScreen acceptor (24,26). These studies have indicated that small molecules beads (coated with nickel, PerkinElmer). Both systems were can improve DM1-related splicing defects by inhibiting shown to detect MBNL1-105-His6 when in close proxim- exp MBNL-CUG binding, thus restoring MBNL function in ity to biotin-CUG12 in solution (37). The assay was used cells. to screen 279 433 compounds in the Molecular Libraries We performed a high-throughput screen to identify com- Small Molecule Repository (MLSMR, Supplementary Ta- pounds that inhibit r(CUG)n binding to MBNL1, the pre- ble S1). A description of the library and selection algorithm dominant MBNL protein of skeletal muscle. Out of 279 for the compounds can be found at http://mli.nih.gov/mli/ 433 compounds in the screen, the most potent inhibitor compound-repository/mlsmr-compounds/. Notably, the se- was lomofungin, a natural antimicrobial agent from Strep- lection algorithm was not designed to enrich for RNA bind- tomyces lomondensis (35,36). We found that lomofungin ing compounds. undergoes spontaneous dimerization in dimethyl sulfoxide (DMSO), producing dilomofungin, whose potency was 17- fold greater than lomofungin in the same screen. However, Identification and structural analysis of dilomofungin while dilomofungin displayed greater r(CUG)n affinity and Analysis of lomofungin (Enzo Life Sciences, BML-A245– stronger MBNL-(CUG)12 binding inhibition in vitro,itwas 0050) was performed on a Shimadzu LC2010 HPLC the monomer that showed greater correction of MBNL1- equipped with a C18 reverse-phase column. 1H NMR spec- dependent splicing in cells. This discrepancy between the tra were recorded at 25◦C on a Bruker Avance 400 (400 cell-free and intracellular activities appeared to result from MHz) or Bruker Avance 500 (500 MHz) instrument. Chem- a novel effect of dilomofungin on RNA turnover. Dilomo- ical shifts (␦) are reported in parts per million (ppm) down- fungin reduced the rate of CUGexp decay in cells, which led exp field from tetramethylsilane and referenced to the residual to a striking increase of CUG RNA in nuclear foci. These protium signal in the nuclear magnetic resonance (NMR) results are cautionary for development of high affinity RNA solvent (CDCl3, ␦ = 7.26). Data are reported as chemical ligands to mitigate RNA toxicity, which in some cases may shift, multiplicity (s = singlet, d = doublet, t = triplet, m = stabilize the target and promote the accumulation of mutant multiplet), integration, and coupling constant (J) in Hertz RNA in cells. (Hz). 13C NMR spectra were likewise recorded at 25◦Con a Bruker Avance 400 (100 MHz) or Bruker Avance 500 (125 MATERIALS AND METHODS MHz). Chemical shifts (␦) are reported in parts per million High-throughput screen for inhibitors of MBNL1-CUG (ppm) downfield from tetramethylsilane and referenced to RNA binding carbon resonances in the NMR solvent. High-resolution mass spectra were acquired at the University of Buffalo Our homogeneous time-resolved fluorescence energy trans- mass spectrometry facility, Buffalo, NY, USA. fer (HTRF) assay for inhibitors of MBNL1-(CUG)12 bind- ing was previously described (37). In brief, the assay employed a chemically synthesized (CUG)12 oligoribonu- Fluorescence titration cleotide and recombinant human MBNL1 protein. The (CUG)12 oligonucleotide was biotinylated at the 5 end and Fluorescence titrations were performed using a Varian Cary contained a 4 bp GC clamp at 5 and 3 ends to stabilize Eclipse spectrophotometer. A 50 ␮M stock of compound the repeat in a hairpin conformation, with G·C and C·G in 1X Hepes buffered saline with 0.5% DMSO and 0.005% base pairs separated by periodic U·U mismatch in the stem. tween-20 was titrated into 400 ␮l of Cy3- labeled (CUG)10 The recombinant MBNL1 had a 105 amino acid deletion RNA, also in the same buffer. After each addition the at the C-terminus, to reduce protein aggregation without mixture was allowed to equilibrate for at least 10 min, loss of r(CUG)n binding, a GST tag at the N-terminus and or until no change in the fluorescence spectrum was ob- a His6 tag at the C-terminus. The MBNL1-105-His6 was served. Changes in fluorescence emission at 565 nm (ex- expressed and purified from Escherichia coli as previously citation at 550 nm) were measured. Raw data were cor- described (37). The biotin-(CUG)12 RNA and MBNL1- rected for dilution-dependent changes, plotted against com- 105-His6 protein were mixed in equimolar concentrations pound concentration and fit to the one site binding equation (20 nM each) and dispensed in 1536 well plates at 2 ␮lper (Equation 1): well.