A high-throughput screening to identify small molecules that suppress huntingtin promoter activity or activate huntingtin-antisense promoter activity

Huntington’s disease (HD) is a neurodegenerative disorder caused by a CAG repeat expansion in exon 1 of huntingtin (HTT). While there are currently no disease-modifying treatments for HD, recent efforts have focused on the development of nucleotide-based therapeutics to lower HTT expression. As an alternative to siRNA or oligonucleotide methods, we hypothesized that suppression of HTT expression might be accomplished by small molecules that either (1) directly decrease HTT expression by suppressing HTT promoter activity or (2) indirectly decrease HTT expression by increasing the promoter activity of HTT-AS, the gene antisense to HTT that appears to inhibit expression of HTT. We developed and employed a high-throughput screen for modifiers of HTT and HTT-AS promoter activity using luminescent reporter HEK293 cells; of the 52,041 compounds tested, we identified 898 replicable hits. We used a rigorous stepwise approach to assess compound toxicity and the capacity of the compounds to specifically lower huntingtin protein in 5 different cell lines, including HEK293 cells, HD lymphoblastoid cells, mouse primary neurons, HD iPSCs differentiated into cortical-like neurons, and HD hESCs. We found no compounds which were able to lower huntingtin without lowering cell viability in all assays, though the potential efficacy of a few compounds at non-toxic doses could not be excluded. Our results suggest that more specific targets may facilitate a small molecule approach to HTT suppression.

A high-throughput screening to identify small molecules that suppress huntingtin promoter activity or activate huntingtin-antisense promoter activity

Huntington's disease (HD) is a neurodegenerative disorder caused by a CAG repeat expansion in exon 1 of huntingtin (HTT). While there are currently no disease-modifying treatments for HD, recent efforts have focused on the development of nucleotide-based therapeutics to lower HTT expression.
As an alternative to siRNA or oligonucleotide methods, we hypothesized that suppression of HTT expression might be accomplished by small molecules that either (1) directly decrease HTT expression by suppressing HTT promoter activity or (2) indirectly decrease HTT expression by increasing the promoter activity of HTT-AS, the gene antisense to HTT that appears to inhibit expression of HTT. We developed and employed a high-throughput screen for modifiers of HTT and HTT-AS promoter activity using luminescent reporter HEK293 cells; of the 52,041 compounds tested, we identified 898 replicable hits. We used a rigorous stepwise approach to assess compound toxicity and the capacity of the compounds to specifically lower huntingtin protein in 5 different cell lines, including HEK293 cells, HD lymphoblastoid cells, mouse primary neurons, HD iPSCs differentiated into cortical-like neurons, and HD hESCs. We found no compounds which were able to lower huntingtin without lowering cell viability in all assays, though the potential efficacy of a few compounds at non-toxic doses could not be excluded. Our results suggest that more specific targets may facilitate a small molecule approach to HTT suppression.
Huntington's disease (HD) is an autosomal dominant neurodegenerative disorder characterized by abnormalities of movement, cognition, and emotion, with relentless progression to death 1 . HD is caused by expansion of a CAG repeat in exon 1 of the ubiquitously expressed Huntingtin (HTT) gene; the repeat is in-frame to encode polyglutamine 2 . In individuals with HD, the mutant huntingtin protein (mHTT) containing a long polyglutamine expansion is found throughout the CNS and also in non-CNS tissues 3 . HD pathogenesis is thought to derive primarily from a toxic gain-of-function conferred on mHTT by the polyglutamine expansion 4 , although some evidence exists for other forms of toxicity, including loss-of-function 5 Stem cell differentiation and compound treatment. Neural progenitor cells derived from induced pluripotent stem cells (iPSCs) 37 were revived from day in vitro (DIV) 18 onto 6-well plates and differentiated into human cortical neurons (hCNs) until DIV25, at which point cells were seeded onto coated 24-well plates and penicillin/streptomycin was added to media. Media was topped up at DIV26 and cells underwent a 50% media change at DIV29. Cells were incubated with compound from DIV32 to DIV39, at which point lysates from each plate were split for use in two cell-based assays. For assays involving the GENEA020 human embryonic stem cells (hESCs) 38 , cells were plated onto collagen-coated plates and allowed to attach overnight prior to treatment with compound for 48 h. Lysate from one plate was split for use in three cell-based assays. Cell viability assays. For all ATP-based cell viability assays, the Cell Titer Glo kit (Promega) was used according to manufacturer's instructions. For total nuclear stain in hESCs, cells were fixed and permeabilized followed by Hoechst 33342 staining. Cells were imaged using the InCell2200 widefield automated microscope (GE Healthcare) and data analyzed using the InCell Developer software. The STHdh Q111 1536-well viability assay was performed as previously described 39 . Briefly, cells were plated in black-wall, clear-bottom 1536-well cyclic olefin polymer-type imaging plates (Edition Eight; Whitefish, MT) at 1200 cells per well in 5 μL volume using a Multidrop Combi Reagent Dispenser (ThermoFisher). Cells were incubated for 16 h in an incubator at 33 °C and 5% CO 2 . 46 nL of compounds were transferred using a pintransfer tool and plates were returned to 33 °C for 2 h. Cells were then shifted to 37 °C and 5% CO 2 for 24 h prior to staining and imaging. Hoechst 33342 and propidium iodide (PI) were prepared in phosphate buffer solution (PBS) and 1 μL was added to each well; final concentration was 4 μg/mL Hoechst 33342 and 5 μg/mL PI. Plates were incubated at room temperature for 30 min prior to imaging on an InCell2200 microscope using a 10 × 0.45 NA air objective and standard DAPI (390/18 ×, 432/48 m) and Cy3 (542/27 ×, 587/45 m) filter sets. One field of Table 3. Antibodies used in HTRF assays.

MW1-Alexa488
Acceptor 520 Total-AKT-Eu 3+ -Cryptate Donor 620 Total-AKT-d2 Acceptor 665 Total-Tau-Eu 3+ -Cryptate Donor 620 Total-Tau-d2 Acceptor 665 www.nature.com/scientificreports/ view per well, encompassing the entire well, was chosen for imaging and percent PI positive cells were determined using Columbus Image Data Analysis System (PerkinElmer) to identify Hoechst nuclei and PI positive nuclei on a cell-by-cell basis. A mean PI intensity of more than 3 standard deviations above the background mean was classified as PI positive. Data were normalized to controls on a per-plate basis (Q7 + vehicle = 100% protection, and Q111 + vehicle = 0% protection).

Data analysis.
Concentration-response curves (CRCs) were generated using NCATS software (https :// tripo d.nih.gov/curve fit/) and analyzed using previously described methods 40 . Curves are classified based on quality of fit to the data, the magnitude of response, and the number of asymptotes to the calculated curve. A positive CRC would indicate a positive correlation of activity with dose concentration, while a negative CRC would indicate a negative correlation. Structural clustering of active compounds was performed using Leadscope Hosted Client (Leadscope Inc., Columbus, OH). The EC50 values of compounds in the confirmation and follow-up experiments were calculated from the dose-response curves by nonlinear regression analysis using Prism software (GraphPad Software, San Diego, CA).

Results
Development of stable reporter cell lines to measure HTT and HTT-AS promoter activity. In order to identify small molecules that alter HTT or HTT-AS expression, we developed a reporter-based screen to assess the effects of small molecules on their respective promoters. We transfected Flp-In T-Rex HEK293 cells with a dual-reporter gene plasmid linked with the HTT (1.7 kb) or HTT-AS (1.5 kb) promoter. The HTT promoter sequence included 1736 bp upstream of the transcription initiation site (chr4:3073089-3074829 human genome 38; hg38) 41 . The HTT-AS promoter sequence (1503 bp upstream of the transcription initiation site;   16 . Additional control plasmids were also constructed, including no promoter ("Null" control), or a CMV tet-on inducible promoter (Fig. 1A). We performed stable monoclonal selection and analyzed a series of monoclones. HTT-2 and HTT-AS-8 clones were selected based on results showing uniform baseline firefly and renilla luciferase activities ( Fig. 1B-E), normal cell morphology, lack of effect of DMSO (1%) on growth and reporter expression, high proliferation rates, and no loss of transgene through at least P10.
Primary screen: screening for compounds which activate the HTT-AS promoter or inhibit the HTT promoter using a luciferase reporter system. 52,041 compounds from five compound libraries (Table 1)  Confirmation screen: identification of reproducible hits from dual-reporter screen. 898 compounds identified in the initial promoter screen were reevaluated at 7 concentration (57 μM top concentration, 1:3 titration) to confirm their dose-response activity based on CRCs (1/2/3 for HTT-AS or − 1/− 2/− 3 for HTT) (Fig. 2). 739/809 HTT lowering compounds (91%) and 52/92 (57%) HTT-AS activating compounds confirmed the validation screen, for a composite confirmation rate of 88% (791/898). Compounds that were active in the confirmation screen were further triaged by structural clustering, elimination of compounds with promiscuous and undesirable functionalities, and removal of compounds which showed an increase in HTT promoter activity as such compounds were unlikely to prove of clinical interest. 123 compounds met these criteria and were selected for further study. Of this group, 112 decreased HTT promoter activity and 15 increased HTT-AS activity, with four compounds both decreasing HTT and increasing HTT-AS promoter activity.
Secondary screen: effect on huntingtin protein in HEK293 and HD patient lymphoblast cell. In order to ascertain huntingtin protein-lowering capacity, a FRET-based assay was used to measure the dose-response effect of the 123 selected compounds on huntingtin protein levels. Compounds were screened at 10 concentration dose response in both HEK293 cells (38 μM with 1:3 dilutions) and a lymphoblast cell line from an HD patient (long allele = CAG 82 ; 17 μM with 1:3 dilutions). HD patient-derived lymphoblasts were selected for use in this study as they have been successfully used to identify novel aspects of HD pathophysiology 42-44 , as Table 4. Summary of compounds that showed huntingtin protein-lowering activity in the HEK293 and HD lymphoblast assays. N/T = Not tested. † AC 50 indicates luminescent suppression (HTT) or induction (HTT-AS). † † AC 50 indicates cytoprotection or cytotoxicity.

Compound name ID
Promoter screens (AC 50 ; μM) † Protein-lowering screens (AC 50 ; μM) Neuron and stem cell-based screens (AC 50 ; μM)  Fig. 3); 10 in HEK293 cells only, 6 in HD lymphoblasts only, and 2 compounds in both cell lines. Of the two compounds that lowered huntingtin protein in both cell types, one (NCGC00274038) was a hit derived from the HTT-AS screen, and one (NCGC00104681) from the HTT screen.

HTT HTT-AS HEK293 (HTT) HD Lymph (HTT) HD hESC (mHTT) HD hCN (mHTT)
Tertiary screen: huntingtin protein lowering in stem cell and neuronal HD lines. 12 out of the 14 compounds active compounds in the HTRF assay above were further tested in human embryonic stem cells (hESCs) from an HD patient (Charles River, long allele = CAG 48 ) and cortical neurons differentiated from an HD induced pluripotent stem cell (iPSC) line (Evotec, long allele = CAG 72 ). Two compounds were excluded due to hazardous material transportation restrictions or limited supplies.
Cortical neurons were treated with compound for seven days at 10 doses (30 μM with 1:3 dilutions). A FRET-based assay was used to assess the capacity of each compound to lower mutant huntingtin, with an ATPbased cell viability assay run in parallel. One compound (NCGC00274038) exhibited some lowering of mHTT at concentrations that did not alter ATP levels (Fig. 4A,B). However, it also lowered Tau as assayed by FRET, indicating a lack of specificity. HD hESCs were treated with compound at 10 doses (30 μM with 1:3 dilutions). Both mutant and total huntingtin were measured, as well as protein kinase B (Akt) to assess specificity of action.  www.nature.com/scientificreports/ Cell viability was measured both by an ATP-based assay and by total cell count based on Hoechst staining. One compound (NCGC00274038) showed measurable huntingtin-lowering activity relative to ATP-based cell viability ( Fig. 4C-F). However, this compound had equipotent effects on huntingtin-lowering cell viability with toxicity measured by cell count. Consistent with this effect and with results in cortical neurons, Akt suppression also paralleled HTT-lowering, indicating non-specific action by the compound. 8 of the 14 hits derived from HEK293 and HD lymphoblast studies were also tested for their cytoprotective capacity in STHdh Q111 mouse striatal cells. Striatal neurons were treated with compound for 24 h at 11 doses (30 μM with 1:3 dilutions) while under stress conditions (low temperature, low serum). Cells were stained with Hoechst to label all nuclei and propidium iodide (PI) to detect dead cells. Percentage of PI positive neurons in compound-treated cells relative to untreated Q7 neurons was used to measure cytoprotection. Two compounds (NCGC00100369, NCGC00113437) showed strong cytoprotective effects (Fig. 4G), three compounds showed mild cytoprotective effects (NCGC00122597, NCGC00136813, NCGC00140752), two compounds were cytotoxic (NCGC00099051, NCGC00274038), and one compound was inactive (NCGC00131485).

Discussion
We developed a high-throughput luciferase reporter assay to screen for compounds that act on the HTT or HTT-AS promoter, using HEK293 cells expressing the HTT or HTT-AS promoter linked to a luciferase reporter. This assay is 1536-well compatible, enabling us to screen a diverse collection of ~ 50,000 compounds. We identified 739 compounds that reproducibly decreased HTT promoter-driven luciferase activity, and 52 that increased HTT-AS promoter-driven luciferase activity. Validation, based on the capacity of hits to lower huntingtin protein at nontoxic concentrations, was performed in two tiers, first, using HEK293 cells and lymphoblasts from HD patients; and second, using HD patient ESCs and iPSCs, and primary neurons from an HD mouse model. None of the small molecules identified in the initial screen were efficacious at non-toxic doses in all tests of validity (Fig. 5).
The discrepancy of results between HEK cells, lymphoblasts, and neuronal and stem cell lines may in part derive from differences in the mechanisms that regulate HTT expression. While huntingtin is ubiquitously expressed, expression levels are presumably under at least some cell-type specific regulation, as indirectly demonstrated by varying levels of HTT expression in different tissues (e.g., the expression of normal HTT is 5 × higher in transformed lymphocytes than in caudate). Other small molecules detected by moderate-or high-throughput screenings as potential HD therapeutic agents have also acted differently on specific cell types, and different assays may yield different results in the same cell-type 47,48 . The mechanisms of HTT transcriptional regulation, and how this regulation differs by cell type, is poorly understood 49 . Our strategy of testing efficacy, specificity, and toxicity in non-neuronal cells first was intended to reduce false positives while increasing the efficiency and decreasing the cost of the screen. The trade-off is a likely increase in false negatives. Other high-throughput screenings for small molecule therapeutics for HD have faced similar obstacles 50,51 .
Our approach emphasized cultured neurons as a final step in validation and test of toxicity. However, cultured neurons are more vulnerable to toxicity than either neurons in vivo or in co-culture with astrocytes 52-54 , increasing the likelihood that potential compounds of interest will be screened-out based on toxicity 55 . Use of co-culture systems, previously developed for investigation of HD pathogenesis 56 , or 3D cultures 57 , will likely decrease this source of false negatives. In addition, our results reflect the methodology used for evaluation of toxicity. ATP-based toxicity screening was used to assess toxicity in HEK293 cells, HD lymphoblasts, and induced cortical neurons, while both ATP and cell count-based methods were used to assess toxicity in HD hESCs. While cell counting methods avoid the issues of cell growth that can confound assays based on ATP production, the advantage comes with the cost of scalability.
One compound, NCGC00274038, a TGBβ-activated kinase (TAK1) inhibitor, lowered HTT and was not toxic by ATP measures in induced neurons, though it was toxic when assayed by cell count in hESCs and in a mouse striatal-derived cell line. We excluded this compound from further consideration based on our a priori analytic  Table 4. www.nature.com/scientificreports/ plan and evidence that direct cell counting is superior to ATP-based assays, recognizing that this decision is quite conservative. However, this does not necessarily exclude this compound or related compounds from further investigation. Previous studies have shown TAK1 inhibitors to have neuroprotective properties 58,59 . TAK1 is a positive regulator of MAPK signaling, and dysregulation of this pathway has been implicated in HD 60 and other neurodegenerative diseases 61 . Two recent studies have identified MAPK-related kinases as a positive regulators of mHTT protein levels, potentially via increased stability of HTT mRNA 62,63 . NCGC00274038 may therefore reduce HTT expression via suppression of MAPK signaling; its detection in our screen for compounds acting on the HTT promoter may have been artifactual, from an unrelated or indirect activity, or due to a stabilizing effect on luciferase mRNA. Other TAK1 inhibitors without the toxicity we detected in NGC00274038 may therefore merit investigation in HD, though likely independent of an effect on the HTT promoter.
This study included a screen for small molecules that upregulated HTT-AS, guided by the rationale that, as found in sense-antisense pairs at other loci, HTT-AS appears to suppress HTT. This screen had the advantage of minimizing detection of compounds on the basis of nonspecific toxicity. We focused on the HTT-AS promoter that drives transcription of HTT-AS-v1, which includes exon 1 with the CUG repeat. One challenge is that over-induction of this transcript may be toxic via the toxicity of expanded CUG repeats 64,65 . More generally, the regulation of expression at bidirectional loci is complex, and it is possible that small molecule manipulation of the antisense promoter is insufficient to elicit the response detected in model systems 16 . We have recently detected a second promoter that appears to drive expression of HTT-AS-v2; this transcript does not incorporate the HTT-AS exon 1, which contains the repeat, but does incorporate HTT-AS exon 2, which is precisely located antisense to the HTT promoter region. While the extent to which activating this second promoter can lead to suppression of HTT remains unclear, it may prove a more amenable target for manipulation than the promoter targeted in this study.
Our assay was designed to detect any potential activator of the HTT or HTT-AS promoter, in part based on limited information on more specific regulators of HTT or HTT-AS promoter function. If detected, such factors might provide a more focused therapeutic target. Work outside of HD suggests that other targets with the Figure 5. Summary of high throughput screening for HTT and HTT-AS promoter modifiers. 52,041 compounds were tested in the primary screen, with a total of 123 hits (112 hits in the HTT promoter assay and 15 in the HTT-AS promoter assay, with 4 compounds active in both assays. Of these 123 compounds, 14 were identified as huntingtin-lowering and non-toxic in the secondary screen. 10 were found to lower huntingtin protein in HEK293 cells and 6 in HD-patient derived lymphoblasts, with 2 active in both cell types. 12 of these compounds were tested in HD human embryonic stem cells (hESCs) and HD cortical neurons (hCNs), and 8 in immortalized HD mouse striatal neurons (mStNs). Compound IDs (Table 4) in chart indicate which compounds moved forward to each tertiary screen. 2 compounds (NCGC00100369; B) and (NCGC00113437; D) were identified as active in the mStNs, but not in hESCs or hCNs. www.nature.com/scientificreports/ potential of lowering HTT expression might include promoter specific G-Quadruplexes 66 , the point of interaction between a transcription factor and Pol II 67 , promoter-regulating lncRNAs 68 , or translational machinery 69 . We conclude that reducing mHTT expression remains an important goal in the field. Antisense and RNAi strategies look promising, but a small molecule approach may avoid some of the obstacles facing these strategies or could lead to an adjuvant therapy. Our results suggest that identifying more specific targets may facilitate this approach.