Address correspondence and reprint requests to Elena Cattaneo, Department of BioSciences and Centre for Stem Cell Research, Università degli Studi di Milano, Via Viotti 3/5, 20133 Milano, Italy. E-mail: firstname.lastname@example.org (or) Alessandro Contini, Department of Pharmaceutical Science, Division of General and Organic Chemistry “A. Marchesini”, Università degli Studi di Milano, Via Venezian 21, 20133 Milano, Italy. E-mail: email@example.com
Transcriptional dysregulation is a hallmark of Huntington's disease (HD) and one cause of this dysregulation is enhanced activity of the REST-mSIN3a-mSIN3b-CoREST-HDAC repressor complex, which silences transcription through REST binding to the RE1/NRSE silencer. Normally, huntingtin (HTT) prevents this binding, allowing expressing of REST target genes. Here, we aimed to identify HTT mimetics that disrupt REST complex formation in HD. From a structure-based virtual screening of 7 million molecules, we selected 94 compounds predicted to interfere with REST complex formation by targeting the PAH1 domain of mSIN3b. Primary screening using DiaNRSELuc8 cells revealed two classes of compounds causing a greater than two-fold increase in luciferase. In particular, quinolone-like compound 91 (C91) at a non-toxic nanomolar concentration reduced mSIN3b nuclear entry and occupancy at the RE1/NRSE within the Bdnf locus, and restored brain-derived neurotrophic factor (BDNF) protein levels in HD cells. The mRNA levels of other RE1/NRSE-regulated genes were similarly increased while non-REST-regulated genes were unaffected. C91 stimulated REST-regulated gene expression in HTT-knockdown Zebrafish and increased BDNF mRNA in the presence of mutant HTT. Thus, a combination of virtual screening and biological approaches can lead to compounds reducing REST complex formation, which may be useful in HD and in other pathological conditions.
Dysregulation of REST and its target genes have been implicated in Huntington's disease. We have coupled structured-based virtual screening approaches to biological assays and selected molecules that interfere with the repressor complex REST-mSIN3b. In particular, at the non-toxic dose, compound C91 is able to increase neuronal gene transcription and to reverse low Bdnf mRNA levels in HD models.
repressor element 1 silencing transcription factor/neuron-restrictive silencer factor
Huntington's disease (HD) is an adult-onset neurodegenerative disorder caused by expansion of the CAG repeat in the HD gene (Huntington's Disease Collaborative Research Group's 1993). The CAG repeat is translated into a polyglutamine (polyQ) tract near the N-terminus of huntingtin (HTT), a protein with beneficial functions in neurons (Zuccato et al. 2010). The pathological mechanisms underlying HD are multi-faceted, and nearly all aspects of cell physiology appear to be affected (Zuccato et al. 2010), including transcriptional dysregulation, an early event considered to contribute considerably to HD pathogenesis.
One of the verified transcriptional abnormalities in HD is the increased repression of neuronal genes by repressor element 1 silencing transcription factor/neuron-restrictive silencer factor (REST/NRSF) (Zuccato et al. 2003; Zuccato and Cattaneo 2007; Bithell et al. 2009; Buckley et al. 2010). REST is part of a repressor complex that regulates the expression of several neuronal genes by binding to repressor element 1/neuron-restrictive silencer element (RE1/NRSE) (Bruce et al. 2004; Johnson et al. 2007; Ooi and Wood 2007). Reduced REST activity and reduced target gene expression have been described in cerebral ischemia (Calderone et al. 2003; Formisano et al. 2007), cancer (Lawinger et al. 2000; Coulson 2005; Fuller et al. 2005; Westbrook et al. 2005; Lv et al. 2010), Down syndrome (Bahn et al. 2002; Canzonetta et al. 2008; Lepagnol-Bestel et al. 2009), cardiac hypertrophy (Kuwahara et al. 2003; Bingham et al. 2007), epilepsy (Garriga-Canut et al. 2006; Cai et al. 2011), and HD (Zuccato et al. 2003; Zuccato and Cattaneo 2007; Marullo et al. 2010). In the brain, normal HTT has been shown to act by retaining REST in the cytoplasm, reducing its repressor function and promoting REST-regulated neuronal gene transcription (Zuccato et al. 2003). In HD, REST translocates to the nucleus, causing reduced transcription of REST target genes (Zuccato et al. 2003, 2007; Zuccato and Cattaneo 2007; Shimojo 2008). Among these targets, BDNF is a neurotrophin important for the survival and activity of striatal neurons that die in HD. A 50 and 80% reduction in the level of BDNF has been reported in different HD models (Zuccato et al. 2007; Zuccato and Cattaneo 2009) and human brains post-mortem, respectively (Zuccato et al. 2008). Subsequent studies have shown that delivery of a dominant-negative form of REST (DN:REST) to HD cells and mice restores the transcription of many RE1/NRSE-controlled neuronal genes (Zuccato et al. 2007; Conforti et al. 2012).
REST recruits distinct corepressor platforms including CoREST, mSIN3a, and mSIN3b [mammalian SIN3 transcription factor (mSIN3)] (Naruse et al. 1999; Grimes et al. 2000; Roopra et al. 2000; Sun et al. 2005; Ooi and Wood 2007) in a tissue and locus-specific manner. mSIN3a interacts with REST in vitro, and its presence is necessary for the repressive function of REST (Grimes et al. 2000; Roopra et al. 2000). Immunohistochemical studies of mSIN3a distribution have shown that mSIN3a is excluded from the nucleus of surviving neurons in post-mortem tissue from HD patients, which could be a secondary result of the damage (Boutell et al. 1999).
The structure of REST in complex with mSIN3a has not been resolved, but the N-terminal repressor domain of REST is known to interact with the four-α-helix amphipathic Alpha-Helix 1 (PAH1) domain of mSIN3b and the NMR structure of this domain, in association with the REST minimal repressor domain, has been made available (Nomura et al. 2005). We reasoned that compounds able to compete with the N-terminal repressor domain of REST binding to this PAH1 domain may prevent formation of the REST-mSIN3 complex.
To identify small organic molecules that are able to competitively bind the PAH1 hydrophobic cleft of mSIN3 molecules, we adopted a virtual screening approach to target the mSIN3-PAH1 domain obtained from the mSIN3b-PAH1 complex with the REST fragment by homology modeling. Among the 7 million molecules screened, 94 compounds were selected with a binding energy < −6.5 kcal/mol, an arbitrarily chosen threshold as no reference compounds are available for this target. The compounds were tested in a primary assay of the DiaNRSELuc8 cell line, a clonally derived rodent brain cell line stably engineered to express a luciferase reporter under a RE1/NRSE promoter (Rigamonti et al. 2007; Leone et al. 2008). Compound 91 (C91) was selected and evaluated in neural stem cell lines (NS) carrying the mutant HTT gene. At the non-toxic dose of 250 nM, C91 was able to block the silencing activity of RE1/NRSE in luciferase reporter assays using a RE1/NRSEBDNF-LUC construct. The same experiment revealed increased mSIN3b cytoplasmic localization and reduced mSIN3b occupancy at the RE1/NRSE Bdnf locus, concomitant with increased Bdnf mRNA and protein levels. Rescue of BDNF mRNA levels was also observed in an Htt-knockdown Zebrafish model treated with the same dose of C91 and in Zebrafish embryos transiently expressing the human N-terminal mutant HTT.
ST14A neural progenitor cells with RE1/NRSEBDNF-LUC (DiaNRSELuc8) (Rigamonti et al. 2007) were grown in high glucose Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum at 33°C in a 5% CO2 atmosphere. Heterozygous Hdh Knock-in NS cells were routinely cultured in NS proliferation medium composed of Euromed-N medium (Euroclone, Milano, Italy) supplemented with N2 supplement (Invitrogen, Carlsbad, CA, USA), 10 ng/mL EGF (Peprotech Inc., distributed by DBA, Milano, Italy), and 10 ng/mL FGF-2 (Peprotech Inc.) (Conforti et al. 2013).
AB strain wild-type Zebrafish were maintained at 28°C on a 14-h light/10-h dark cycle at the Department of BioSciences, Università degli Studi di Milano, according to Westerfield M. (Westerfield 1993). Embryos were collected by natural spawning, staged, and raised at 28°C in fish water (Instant Ocean, 0.1% methylene blue) in Petri dishes.
Homology models of human mSIN3-PAH1 (Q96ST3, amino acids 119–189) were generated with MOE software (Chemical Computing Group Inc., Montreal, Canada, 2010), protonated using the H++ server assuming a pH of 7.4 and salinity of 0.15 mol/L (Gordon et al. 2005), and refined by molecular dynamic simulations using the AMBER9 suite (Case et al. 2005) with the ff03 force field (Duan et al. 2003). All dockings were performed with two different docking software programs, MOE (Chemical Computing Group Inc.) and AD4 (Morris et al. 2009). MOE docking was performed using ‘triangle matcher’ (TM) placement and the ‘affinity’ scoring function with the default parameters; as in previous studies this protocol was fast and accurate enough for primary selection (Ferri et al. 2009, 2013). The binding site was selected according to the MOE ‘site finder’ procedure using the previously equilibrated models as the receptor. All unique compounds with a predicted docking energy < −6.5 kcal/mol were considered for further refinement using the MMFF94x force field. All compounds with a predicted docking energy < −6.5 kcal/mol were selected and processed in AD4 (Morris et al. 2009). Molecular graphic images were produced by either MOE or the UCSF Chimera package (Pettersen et al. 2004). For more details see supplementary informations.
Microinjection of human mutant and wild-type htt mRNA
Injections were carried out on one- to two-cell stage Zebrafish embryos. To perform htt knock-down, a specific translation-blocking htt morpholino (htt ATG MO, 5’-GCCATTTTAACAGAAGCTGTGATGA-3’, Gene Tools, LLC) was injected at a concentration of 1 pmol per embryo in 1X Danieau buffer (pH 7.6). A standard control morpholino (SC-MO) oligonucleotide specific for human β-thalassemia was used as a negative control. To over-express human N-terminal HTT with 128Q, synthetic capped mRNAs prepared using ‘mMESSAGE mMACHINE™ SP6 Kit’ (Ambion, distributed by Life Technologies, Monza, Italy) were injected at a concentration of 75 pg per embryo.
C91 administration to Zebrafish embryos
C91 stock solutions were diluted in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM MgSO4) supplemented with 0.1% methylene blue. At 24 hpf, wild-type, htt ATG MO, and SC-MO-injected embryos were dechorionated to allow access to the compound and incubated in 5 mL compound dilutions in six-well plates for 48 h. Embryos treated with DMSO vehicle were used as controls. For each group, 50 embryos were treated and analyzed. Images of 72 hpf embryos were acquired after treatment using a Leica DM6000 B microscope equipped with a DFC 480 digital camera and LAS Leica imaging software (Leica, Milano, Italy).
Compounds 87 and 91 were obtained from ChemDiv Inc. (San Diego, CA, USA). Compounds 81, 82, and 113 were obtained from InterBioScreen Ltd. (Institutsky Prospect, Chemogolovka, Russia) and guaranteed by the resellers to be at least 95% pure. The lyophilized powder was dissolved in 100% DMSO (Sigma, Milano, Italy) to a final concentration of 10 mM. Working solutions were prepared following the protocol reported in Leone et al. 2008.
Biological assays were performed both in reporter cell line DiaNRSELuc8 (Rigamonti et al. 2007) and in HD NS cell lines (Conforti et al. 2013). The protocols used to evaluate luciferase activity (Promega, Milano, Italy) and the cell viability (Sigma) have been already described (Rigamonti et al. 2007). DiaNRSELuc8 was treated for 72 h with different doses of each compound tested (100 nM, 500 nM, 1 μM, 5 μM, and 10 μM). NS-HdhQ111/7 cells transfected with RE1/NRSEBDNF-LUC construct were treated for 72 h with different doses of C91 (1 nM, 50 nM, 250 nM, and 1.5 μM). Luciferase activity was measured and reported RLUs.
Cell proliferation assays
NS cells were plated in proliferation medium in 96-well microplates at 5000 cells/well. The CyQuant® NF Cell Proliferation Assay Kit (Molecular Probes, Invitrogen) was used to assay cell proliferation according to the manufacturer's instructions. Cellular DNA content was used to provide an accurate indication of relative cell number.
RNA isolation and reverse transcription
We isolated total RNA from cells and tissues using Trizol Reagent (Invitrogen). Genomic DNA was digested with DNAse (Applied Biosystems, Foster City, CA, USA) at 37°C for 5 min. Total RNA (1 μg) was reverse-transcribed to single-stranded cDNA using Superscript III RNaseH-reverse transcriptase (Invitrogen) and random primers in a 20 μL reaction according to the manufacturer's instructions.
Transfection with siRNA for REST, mSin3a, mSin3b, and CoREST
For siRNA knock-down studies, HD NS cells were transfected with REST, mSin3a, mSin3b, scramble (Santa Cruz, distributed by DBA), or CoREST (Dharmacon, distributed by Dasit Sciences, Milano, Italy) siRNA according to the optimized protocol for mouse neural stem cells (Lonza, distributed by Euroclone, Milano, Italy). Briefly, 4 × 106 cells were re-suspended in 100 μL of Mouse NSC Nucleofector® Solution containing 100 nM of siRNA and nucleofected using Nucleofector® Program A-033 (Amaxa, distributed by Euroclone). RNA was collected at 24 and 48 h after transfection and processed for QPCR analysis.
Cells (2 × 107) were treated with 1% formaldehyde in phosphate buffered saline (PBS) by rotation for 8–10 min at 4°C. Fixation was stopped by the addition of glycine as a final concentration of 125 mM. Cells were washed two times with PBS, the pellets suspended in chromatin immunoprecipitation (ChIP) lysis buffer. All steps of ChIP were performed and included protease inhibitor mixture MINI (Roche, Basel, Switzerland) (Zuccato and Cattaneo 2007; Zuccato et al. 2007; Conforti et al. 2012). Sonication was performed in glass tube to minimize foaming of the solution. Under these conditions, DNA fragments with an average size of 200–700 bp were obtained. Sonicated extracts were centrifuged and chromatin yield was evaluated by UV spectrometry. Chromatin was precleared with blocked protein G-sepharose and incubated overnight with rotation using 1 μg of primary anti-REST (cat. #07-579; Upstate, Millipore, Milano, Italy), anti-mSIN3a (cat. #SC-994; Santa Cruz), and anti-mSIN3b (cat. #SC-996; Santa Cruz) antibodies. Purified DNA has been analyzed using Quantitative real-time PCR (QPCR) (iCycler Thermal Cycler with Multicolor Real-time PCR Detection System; Bio-Rad, Hercules, CA, USA) using SYBR Green to assess REST (Upstate, Millipore), mSIN3a (Santa Cruz), and mSIN3b (Santa Cruz) occupancy. Three independent PCR experiments were performed for each RE1/NRSE. Primers flanking the RE1/NRSE of Bdnf have been reported previously (Zuccato et al. 2007; Conforti et al. 2012).
Three independent reactions were set up for each RNA stock using 500 ng of total RNA and gene expression assessed by QPCR. The analyses were performed in triplicate for each of the genes to obtain replicates for statistical analysis. QPCR reactions were performed using an iCycler Thermal Cycler with a multicolor real-time PCR detection system (BioRad). All of the reactions were performed in a total volume of 25 μL containing 25 ng of cDNA, 50 mM KCl, 20 mM Tris-HCl (pH 8.4), 0.2 mM dNTPs, 25 units/mL iTaq DNA polymerase, 3 mM MgCl2, 10 nM SYBR Green I fluorescein, stabilizers (iQ SYBR Green Supermix BioRad), and 0.2 μM of forward and reverse primers. The amplification cycles for total BDNF consisted of an initial 10 min denaturing cycle at 95°C followed by 40 cycles of 10 s at 95°C, 10 s at 56°C, and 20 s at 72°C. The amounts of target mRNA were normalized to a reference gene (β-actin) amplified following the same protocol. The primer sequences were as follows: mBDNF S, 5′-TCGTTCCTTTCGAGTTAGCC-3′; mBDNF AS, ′5-TTGGTAAACGGCACAAAAC-3′; mSyn1 S, 5′-GAGCAGATTGCCATGTCTGA-3′; mSyn1 AS, 5′-CACTGCGCAGATGTCAAGTC-3′; mChrm4 S, 5′-TCCTCACCTGGACACCCTAC-3′; mChrm4 AS, 5′-ACGTAGCAGAGCCAGTAGCC-3′; mSNAP25 S, 5′-GAACAACTGGAACGCATTGA-3′; mSNAP25 AS, 5′-AAGCCCGCAGAATTTTCCTA-3′; mβ-actin S, 5′-AGTGTGACGTTGACATCCGTA-3′; mβ-actin AS, 5′-GCCAGAGCAGTAATCTCCTTCT3′; mREST S, 5′-CGAACTCACACAGGAGAACG-3′; and mREST AS, 5′-GAGGCCACATAATTGCACTG-3′.
Cell lysates were prepared in lysis buffer consisting of 10% glycerol, 25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, and 1 mM EGTA supplemented with 1 : 100 of Protease Inhibitor Mixture (Sigma). The samples were homogenized, sonicated, and centrifuged (Biofuge, distributed by AHSI, Monza, Italy) for 15 min at 4°C at maximum speed, and the supernatants were collected and stored at 80°C. The samples were assayed for BDNF using an Immuno Assay System (Promega) according to the manufacturer's instructions. The proteins were quantified by a bicinchoninic acid protein assay (Pierce, distributed by Euroclone) before loading.
Cells were fixed in 4% paraformaldehyde for 15 min at 25°C and washed three times with PBS. The cells were permeabilized with 0.5% Triton (Sigma) and blocked with 5% fetal bovine serum (Euroclone) for 1 h at 25°C. Cells were incubated overnight at 4°C with the appropriate primary antibody: anti-REST (cat. #SC-15118; Santa Cruz), anti-SIN3a (cat. #SC-994; Santa Cruz), anti-SIN3b (cat. #SC-996; Santa Cruz), anti-CoREST (cat. #612146; BD Transduction Laboratories, Milano, Italy). After incubation with the primary antibody, the cells were washed three times with PBS. The appropriate secondary antibodies conjugated to Alexa fluorophores 488 or 568 (Molecular Probes, Invitrogen) were added and incubated for 1 h at 25°C, the cells washed three times with PBS, and the nuclei stained with Hoechst 33258 (5 μg/mL Molecular Probes, Invitrogen). Images were acquired using a Leica DMI 6000B microscope and LAS-AF imaging software (Rasband, W.S., ImageJ, U.S. National Institutes of Health, Bethesda MD, USA).
Statistical tests were used on the basis of the number of the cell lines (wild-type and HD) and the treatment (Compound 91). One-way anova followed by Bonferroni post-test was used to analyze gene transcription changes in Bdnf and in REST-regulated genes after transfection with siRNA for REST, mSin3a, mSin3b, and CoREST. One-way anova followed by Bonferroni post-test was used to analyze data from luciferase activity and cell viability in test of dose-response curves both in DiaNRSELuc8 and HD NS cells. The same test was used to evaluate changes in mRNA levels of Bdnf and REST-regulated genes after treatment with CP91 at different doses in HD NS cells. ChIP experiments on HD NS cells were analyzed with one-way anova followed by Bonferroni post-test. Finally, the results of gene expression in Zebrafish embryos treated with compound were analyzed by one-way anova followed by Bonferroni post-test.
Knock-down of selective REST complex components positively influences neuronal gene transcription in HD cells
Prior to virtual screening, we performed individual siRNA experiments to knock-down each specific component of the REST complex in heterozygous NS carrying full-length HTT with 140 glutamines expressed from the CAG knock-in allele (NS-HdhQ140/7). These NS cells were generated previously in our laboratory from the correspondent ES cells (Conforti et al. 2013). As a consequence of the mutation, NS-HdhQ140/7 cells exhibited a 30% reduction in the Bdnf mRNA level with respect to NS cells carrying the normal Hdh gene and were characterized by a significant reduction in cell survival (Conforti et al. 2013). As expected, transfection of REST siRNA in proliferating NS-HdhQ140/7 cells triggered a 25 ± 5% decrease in the correspondent mRNA level and a 60 ± 10% increase in Bdnf mRNA, whereas a higher three and five-fold increase in Syn1 and Snap25 mRNA levels was measured with respect to control scramble-transfected cells (Fig. 1a). Similarly, specific knock-down of mSin3a or mSin3b resulted in a 60 and 70% reduction in the corresponding mRNA levels and a significant increase in Bdnf, Syn1, and Snap25 mRNA levels (Fig. 1b–c). In contrast, CoREST knock-down was ineffective (Fig. 1d). These data support previous results showing that REST recruits specific cofactors in individual cells (Greenway et al. 2007) and highlights mSIN3 binding to REST as a potential target for screening compounds able to disrupt the repressor complex and restore neuronal gene transcription in HD cells.
Virtual screening of mSIN3 modulators
Next, we designed a virtual screening for compounds potentially able to interfere with the interaction between mSIN3 molecules and REST. The mSIN3-PAH1 domain was obtained by homology modeling, starting from the NMR coordinates of the mSIN3b-PAH1 complex with the N-terminus of REST (Nomura et al. 2005). Virtual screenings using an ensemble docking approach can provide better quality results than those expected from single experimental X-ray structures (Novoa et al. 2010). Accordingly, two different models based on two different coordinate sets (model 1 and model 2) of the original NMR structure were employed.
The virtual screening procedure is summarized in Fig. 2a. Starting from the ZINC database (Irwin and Shoichet 2005) we obtained a 3D conformational database (DB) of chemically diverse drug-like compounds (Ferri et al. 2009). Using a consensus approach, two docking software programs, MOE and Autodock4 (AD4) (Morris et al. 2009), were used to screen the DB (step1 of Fig. 2a). From the first screening of model 1, 1343 compounds were found to be capable of interacting with the selected binding site with a predicted docking energy of < −6.5 kcal/mol (step2 of Fig. 2a). This energy threshold was chosen on the basis of our experience in the docking of protein–protein interaction inhibitors (Ferri et al. 2009, 2013), as no compounds are currently available as a reference. The selected compounds were subjected to a force field refinement procedure (step3 of Fig. 2a) and a total of 1000 compounds were selected and re-docked to mSIN3a-PAH1 using the AD4 package (step4 of Fig. 2a). Only compounds with a docking energy of < −6.5 kcal/mol predicted by both docking software programs were selected for further refinement (221 compounds). Finally, to restrict the selection to compounds capable of binding different receptor conformations, the selected compounds were docked on the model 2 receptor using both MOE and AD4 (step5 of Fig. 2a). A total of 94 compounds were predicted to have a binding energy < −6.5 kcal/mol by both software programs for both receptor models.
Primary screening in brain-derived DiaNRSELuc8 cells
The 94 compounds selected were initially tested in DiaNRSELuc8 reporter cells (Rigamonti et al. 2007; Leone et al. 2008). In this primary screening each compound was tested for cell viability and luciferase activity at 5 μM for 72 h. As a control, the same cell line was treated with DMSO (vehicle). From this biological screening five compounds were selected because they led to a 2- and 3.5-fold increase in luciferase activity, as measured by a dual luciferase reporter assay and 40–60% of cell viability according to a CyQUANT assay (Table 1). The other 89 compounds were screened out because they did not produce any significant change in luciferase activity (Table S1). Hit compounds belonged to two different chemical subclasses. C81, C87, and C113 exhibited a coumarin-like structure, whereas C82 and C91 were quinolone-like compounds. These five compounds were subjected to a dose-response test.
Table 1. Active compounds in primary screening. Each compound was administered to DiaNRSELuc8 cells at 5 μM dose for 72 h. Cell viability was assessed by CyQUANT assay and ability to inhibit the RE1/NRSE silencer was evaluated by measuring Luciferase activity
Primary screening (5 μM for 72 h)
Cell viability (%)
Hit compounds were retested in DiaNRSELuc8 cells using a range of concentrations, from 100 nM to 10 μM, for 72 h. Treatment with C81 (Fig. 2b) or C87 (Fig. 2c) resulted in a 1.5-fold increase of luciferase activity at 5 and 10 μM, but at these concentrations cell survival was reduced to 50% compared with untreated cultures (Fig. 2b and c, respectively). C113 resulted in a 1.5- to 2-fold increase in luciferase activity at 1 and 5 μM with a 30 and 50% reduction in cell viability, respectively (Fig. 2d). C82 and C91 generated a 1.35- and 1.25-fold increase, respectively, in luciferase activity at 1 μM and no detrimental effect on cell survival (Fig. 2e and f respectively). C91 was thus selected for further testing.
C91 restores neuronal gene transcription in a HD cell model
Next, we evaluated the activity of C91 on RE1/NRSE-controlled neuronal gene transcription in a RE1/NRSEBDNF-LUC transfected NS control cell line (NS-HdhQ7/7) and two heterozygous HD knock-in NS cell lines carrying 111 (NS-HdhQ111/7) or 140 (NS-HdhQ140/7) glutamines (Q) (Conforti et al. 2013). C91 was administered at different doses (from 1 nM to 1.5 μM) for 72 h. Figure 3a and b show that even at the highest dose (1.5 μM), C91 did not affect cell viability or morphology.
Analysis of luciferase activity showed that, in RE1/NRSEBDNF-LUC transfected NS-HdhQ111/7 cells, C91 led to a three and two-fold increase in luciferase activity at 250 nM and 1.5 μM, respectively, compared with DMSO-treated cells (Fig. 3c). Exposure of NS-HdhQ111/7 cells transfected with a sterol regulatory element fused upstream of the luciferase gene (SRELUC construct) to the same concentrations of C91 did not produce any change in luciferase activity (Figure S1). We concluded that C91 interferes with REST complex formation without affecting the transcriptional pathway that regulates the expression of cholesterol genes.
To verify that C91 has the capacity to restore endogenous transcription of RE1/NRSE-controlled neuronal genes, we measured mRNA levels of Bdnf, Chrm4, Snap25 and Syn1 by QPCR in control and HD heterozygous knock-in cells. Bdnf, Syn1, ChrM4, and Snap25 mRNAs were down-regulated in HD NS cells compare with NS-HdhQ7/7 control cells (Fig. 3d). Treatment with C91 led to an increase in transcription of RE1/NRSE–regulated genes both in wild-type and HD cells. Figure S2 shows that C91 is able to increase transcription of the BDNF gene in control cells, with the highest effect observed at 50 nM. NS-HdhQ111/7 cells also showed a six to eight-fold increase of Bdnf mRNA when treated with C91 at 50 and 250 nM, compared with untreated cultures (Fig. 3d). NS-HdhQ140/7 cells also showed a significant nine-fold increase in the mRNA level after treatment with 250 nM of C91 (Fig. 3d). Positive results were also obtained when analyzing the levels of Chrm4, Snap25, and Syn1 mRNA. Treatment with 250 nM C91 resulted in a 5.3-fold increase in Chrm4 mRNA levels in NS-HdhQ111/7 cells and 6.5-fold increase in NS-HdhQ140/7 cells compared with controls (Fig. 3d), and Snap25 and Syn1 mRNA levels were increased only in NS-HdhQ140/7 cells, with an eight and six-fold increase, respectively (Fig. 3d). Importantly, C91 did not seem to act as a broad transcriptional activator, as the mRNA levels of Lamp, Ctsf, Tpp1, and HmgCR, which do not contain RE1/NRSE sites, were not affected (Figure S3).
BDNF protein was significantly reduced in NS-HdhQ111/7 and NS-HdhQ140/7 cells compared with wild-type NS cells (198 ± 22 and 217 ± 19 pg BDNF/mg total protein, respectively, versus 278 ± 11 pg BDNF/mg total protein; Fig. 3e). Therefore, we tested whether C91, at its most active and non-toxic dose of 250 nM, was able to restore normal BDNF protein levels in HD cells. Figure 3e shows that C91 was able to increase BDNF protein levels in NS-HdhQ111/7 and NS-HdhQ140/7 cells up to 263 ± 48 and 287 ± 39 pg BDNF/mg total protein, respectively, which represent an increase of 30 and 35%, respectively, compared to the basal BDNF protein content in these cells.
We concluded that C91 affects the activity of the REST complex by restoring BDNF protein production and the transcription of a number of REST-controlled neuronal genes.
C91 reduces genomic binding of mSIN3b to the RE1/NRSE site within the Bdnf locus and increases its cytoplasmic localization in HD cells
To address the mechanism by which C91 affects the activity of the REST complex in HD cells, we performed ChIP analysis to measure the occupancy of the REST, mSIN3a, and mSIN3b at different RE1/NRSE loci after treatment. Chromatin from proliferating NS-HdhQ140/7 cells treated with 250 nM C91 for 72 h was immunoprecipitated with REST, mSIN3b, and mSIN3a antibodies, respectively, and the occupancy of REST, mSIN3b, and mSIN3a at RE1/NRSE in the Bdnf locus was evaluated by QPCR. mSIN3b occupancy at the Bdnf locus was significantly reduced in the absence of changes in the mSIN3b mRNA and protein levels (Fig. 4a–c). In contrast, no changes were observed in the occupancy of REST or mSIN3a under the same conditions (Fig. 4a).
Immunocytochemical analysis of mSIN3b in NS-HdhQ140/7 cells exposed to C91 at 250 and 500 nM demonstrated increased cytoplasmic localization of the protein, but no changes were detected when cells were exposed to DMSO or 50 nM C91 (Fig. 4d). Additional analyses reported in Figure S4 revealed specific accumulation of mSIN3b in the nucleus of NS-HdhQ140/7 cells. In contrast, wild-type NS cells exhibited a widespread mSIN3b signal in both the nucleus and cytoplasm (Figure S4). Immunocytochemistry for REST and mSIN3a did not reveal any change in their localization in the presence of treatment (Fig. 4d). This evidence correlates with ChIP data indicating reduced mSIN3b occupancy at the Bdnf locus (Fig. 4a).
Taken together, these results indicate that C91 counteracts REST-mediated repressor activity in two HD NS cell lines. In addition, we found that C91 increases mSIN3b cytoplasmic localization and restores Bdnf transcription by reducing mSIN3b DNA binding at the Bdnf locus.
C91 mimics HTT and restores BDNF levels in Zebrafish carrying the HD gene or knock-down for the Hdh gene
To test the effects of C91 in vivo, we employed Zebrafish embryos as a model system. To identify the temporal window for treatment, we used QPCR to measure BDNF mRNA levels at different embryonic stages, ranging from 1–2 cells to 5 dpf (days post-fertilization). Starting at 5–6 somites, Bdnf mRNA transcription increased in parallel with the development of the nervous system (Figure S5).
On the basis of these findings, we decided to add C91 to the fish water between 24 and 72 hpf (hours post-fertilization) at concentrations ranging from 10 nM to 1 μM. A total of five independent experiments were performed and the survival, gross morphology, and BDNF transcript levels analyzed in C91 and DMSO-treated embryos. After 48 h of exposure to C91, we observed no changes in morphology compared with controls (Figure S6), nor any significant changes in the mortality rate, although variability amongst the five different experiments was high (Fig. 5a). However, when total RNA samples were collected from 72 hpf C91 and DMSO-treated embryos to measure Bdnf transcript levels, C91 administration resulted in a 2.5-fold increase in Bdnf mRNA compared with controls, at the dose of 250 nM of C91 (Fig. 5b). Using the same methodological approach, we analyzed the effect of C91 on the expression of other REST-controlled genes, including proenkephalin (pENK) (Figure S7). pENK mRNA levels increased four-fold compared with untreated embryos. In contrast, a transcript that did not contain RE1/NRSE regulatory sites, such as neurotrophin-7 (NT7), was not modulated in Zebrafish treated with C91 (Figure S7).
Next, we tested the effects of C91 after htt-knockdown via the injection of a specific translation blocking morpholino (httMO, morphants) (Lo Sardo et al. 2012). This treatment is known to reduce BDNF mRNA levels, with a phenotype similar to Bdnf-knockdown in fishes (Diekmann et al. 2009). We measured a 55% reduction in Bdnf mRNA at 3 dpf in htt-knockdown Zebrafish (Fig. 5c). Importantly, though the treatment was not able to rescue the overall fish morphology because of htt-knockdown, the Bdnf transcript level in morphant embryos treated with C91 was similar to that of DMSO-treated control embryos (Fig. 5c). These data indicate that C91 stimulates BDNF gene transcription in vivo.
To evaluate the activity of C91 in an HD-like in vivo context, we over-expressed human N-terminal HTT with 128 Q through microinjection of synthetic capped mRNAs in 1- to 2-cell stage Zebrafish embryos. At 24 hpf, we observed that expression of human N-terminal form in Zebrafish embryos caused a severe phenotype characterized by a reduction of anterior neural tube, necrosis of the head, and altered body plan (data not shown). We also found that over-expression of the mutant gene with 128Q did not affect Bdnf mRNA production. However, treatment of the mutant HTT Zebrafish embryos with 250 nM C91 led to a threefold increase in Bdnf mRNA compared with untreated Zebrafish embryos (Fig. 5d). This finding suggests that C91 is able to increase Bdnf mRNA levels in a mutant HTT background in vivo.
HD is an orphan therapy disease as the majority of therapeutics currently used in HD ameliorate its primary symptomatology but do not address the inexorable disease progression (Venuto et al. 2012). Therefore, research efforts are directed at studying key cellular pathogenic mechanisms to identify new targets of therapeutics (Zuccato et al. 2010).
One of the transcriptional abnormalities described in HD involves neuronal gene repression by the transcription factor REST (Zuccato et al. 2003, 2007; Zuccato and Cattaneo 2007). Mutated HTT does not have the ability to sequester REST within the cytoplasm and consequently, the repressor pathologically accumulates in the nucleus of HD cells and suppresses production of BDNF, a neurotrophin important for striatal neurons, which are the primary site of degeneration in HD (Zuccato et al. 2003, 2007; Zuccato and Cattaneo 2007). Several studies have addressed the direct interaction of REST with RE1/NRSEs in target genes in cells and mice expressing mutant HTT or in models of HTT depletion (Zuccato et al. 2003, 2007; Zuccato and Cattaneo 2007; Soldati et al. 2011). In addition, a ChIP-on-ChIP approach demonstrated an increased occupancy of multiple REST target genes in post-mortem brains from HD patients and parallel down-regulation of several REST-regulated genes (Zuccato and Cattaneo 2007; Zuccato et al. 2007; Johnson and Buckley 2009). Increased REST occupancy has also been reported in peripheral cells from HD patients (Marullo et al. 2008). Furthermore, in vivo interference of the activity of the REST repressor complex by genetic manipulation has restores BDNF levels in HD mouse models via reduced endogenous REST occupancy at the Bdnf locus (Conforti et al. 2012). Taken together, these observations identified REST as a suitable target for high-throughput screening to develop compounds useful in HD (Rigamonti et al. 2009).
Here, by coupling virtual screening approaches to in vitro and in vivo experiments, we validated that the REST repressor complex is a molecular target in HD. Our multidisciplinary study allowed us to select a group of molecules on the basis of their chemical structure and interactions with the different REST cofactors. Several independent studies have shown that the transcriptional repression mediated by REST involves the recruitment of mSIN3a/b. In addition, in this study we demonstrated for the first time that mSIN3b has a nuclear localization in HD cells. Our virtual screening approach specifically identified small organic molecules that are able to bind the PAH1 hydrophobic cleft of mSIN3 competitively, inhibiting the PAH1-REST interaction. Among these molecules, C91 was the most effective compound in vitro. Figure 6 shows that C91 is predicted to be buried in the PAH1 cleft. Hydrophobic interactions were observed between the 2-fluorobenzyl moiety and the Phe36 side chain, whereas the main quinolone scaffold lies in the hydrophobic pocket formed by Ala8, Leu9, Leu12, Phe29, and Met33. The chemical structure of C91 is ascribable to quinolones, a family of synthetic broad-spectrum, chemotherapeutic, bactericidal drugs, suggesting that the molecule already possesses good drug-like properties. This finding may be useful for future studies in animals. C91 was not toxic in any of the HD NS cells tested or in vivo in the Zebrafish model. C91 was able to increase Bdnf mRNA levels both in vitro and in vivo and to increase the transcription of other RE1/NRSE-regulated genes. The transcription of selected genes that are not regulated by REST was unaffected, supporting the idea that C91 may preferentially affect the REST transcriptional pathway. Future studies with animal models of HD are needed to test the efficacy of C91 as a potential therapeutic for HD.
In conclusion, we demonstrated that compounds that directly interfere with the RE1/NRSE transcriptional machinery can be obtained through a virtual screening procedure designed to identify modulators of protein–protein interactions that affect the formation of the REST-mSIN3 complex. We also demonstrated that the RE1/NRSE transcriptional machinery can be targeted by chemical compounds both in vitro and in vivo. Although C91 does not provide a full rescue from mutant HTT toxicity, we showed that C91 increases neuronal gene transcription and counteracts low Bdnf mRNA levels in the presence of the HD mutation. Given the multiple pathogenic mechanisms involved in HD, a compound targeting one pathological mechanism may not be effective alone. Combinations of therapeutics that target different pathogenic mechanisms should have greater efficacy.
This study was supported by STEM-HD (FP6, European Union, LSHB-CT-2006-037349) and partially by the Huntington's Disease Society of America Coalition for the Cure (NY, U.S.A.). We also wish to thank one anonymous donor. A. Contini is grateful to the ‘Consorzio Interuniversitario Lombardo per l'Elaborazione Automatica’ (CILEA) for computational facilities.
Conflict of interest statement
The authors have no conflicts of interest to declare.