Large-scale screening using familial dysautonomia induced pluripotent stem cells identifies compounds that rescue IKBKAP expression.

Patient-specific induced pluripotent stem cells (iPSCs) represent a novel system for modeling human genetic disease and could provide a source of cells for large-scale drug-discovery screens. Here we demonstrate the feasibility of performing a primary screen in neural crest precursors derived from iPSCs that were generated from individuals with familial dysautonomia (FD), a rare, fatal genetic disorder affecting neural crest lineages. We tested 6,912 small-molecule compounds and characterized eight that rescued expression of IKBKAP, the gene responsible for FD. One of the hits, SKF-86466, was found to induce IKBKAP transcription through modulation of intracellular cAMP levels and PKA-dependent CREB phosphorylation. SKF-86466 also rescued IKAP protein expression and the disease-specific loss of autonomic neuronal marker expression. Our data implicate alpha-2 adrenergic receptor activity in regulating IKBKAP expression and demonstrate that small-molecule discovery using an iPSC-based disease model can identify candidate drugs for potential therapeutic intervention.

Identification of Compounds that Rescue IKBKAP Expression in Familial Dysautonomia-iPS Cells

Familial dysautonomia (FD) is an autosomal recessive disease caused by a single point mutation in the I-κ-B kinase complex-associated protein (IKBKAP) gene leading to IKBKAP mis-splicing and a marked reduction in IKAP protein[1]. We previously reported the isolation of iPSC lines from FD-fibroblasts and identified multiple disease-specific phenotypes[2]. Among those phenotypes, we observed low levels of wild type (WT) IKBKAP expression, particularly in FD-iPSC derived neural crest (FD-NC), which may partly explain why the disease specifically affects the peripheral nervous system. FD is a progressive neurodegenerative disorder without effective treatment. The identification of compounds that increase IKBKAP expression in affected cell types could yield novel therapies for FD.

Developing screening conditions for FD-iPSC-derived NC

Several recent disease modeling studies have demonstrated the use of iPSC derived cell types for validating small sets of drug candidates[2-9]. However, to move from validating a few compounds to screening larger chemical libraries, it is critical to define disease-relevant conditions suitable for use in HTS (Fig. 1a). The first step in this process is the large-scale production of the pertinent cell type. We have previously reported the prospective isolation of FD-NC precursors by flow cytometry (Fig. 1b). Here we proliferated purified FD-NC precursors in the presence of FGF2 and EGF for 2 weeks (Fig. 1c) and cryo-preserved large batches (108–109 cells) for subsequent screening applications. NC precursors showed stable growth properties (population doubling time: 44.2 hrs; Ki-67+ cells: 51.7 ± 2.4%), high purity, maintenance of neurogenic differentiation potential, and excellent post-thawing recovery rates (91.6 ± 5.7 %; Fig. 1c, Supplementary Fig. 1). Optimized cell plating conditions for 384-well microtiter plates were obtained using a laminin/fibronectin-based coating method ("in house" coating; Supplementary Material & Methods, Supplementary Fig. 2a) that achieved reproducible cell attachment with high viability (Fig. 1d). Cellular growth was monitored using Alamar Blue[10] and Hoechst nuclear staining[11] which defined suitable plating densities at 2,500 – 7,500 cells/well (Fig. 1d, Supplementary Fig. 2b–c). DMSO had no major impact on growth of FD-NC precursors up to 1% DMSO (v/v). The final step in developing an FD-NC based HTS assay was the selection of a sensitive and disease-relevant readout. We hypothesized that promising compounds should increase levels of WT-IKBKAP in patient specific cells and thereby increase the levels of IKAP protein. Therefore, we developed a qRT-PCR assay for measuring levels of WT-IKBKAP against the 18S internal control based on published primer sets[12] (Supplementary Fig. 2d). Levels of mutant (MU)-IKBKAP were also determined to address whether compounds increase both WT and MU-IKBKAP or act via IKBKAP splicing[12]. Transcript levels were measured following cell lysis, RNA extraction, qRT-PCR reaction and data quantification (Supplementary Table 1). RT-PCR technology is not commonly used in HTS and is highly dependent on the quality and quantity of the isolated RNA. We found that a plating density of 7,500 cells/well yielded excellent reproducibility. To further validate our RT-PCR assay, we performed a control study mimicking the screening work flow (Suppl. Fig. 3) on three 384-well plates and demonstrating high consistency in Ct values (coefficient of variation (CV) ranging from 1 to 2%; average Z' values of 0.78 (Fig. 1e)). Differences in Ct values for WT-IKBKAP were minimal among replicate wells of the same plate. However, more marked differences were observed between wells of independent plates prompting us to perform the final screen in triplicates.

Figure 1

HTS Assay Development

a, Schematic representation of HTS assay: FD patient specific fibroblasts were reprogrammed into FD-iPSC and FD-iPSCs were differentiated into neural crest cells. FD-NCs were purified and expanded for 2 weeks. After optimization, FD-NCs were plated in 384 well plate and further treated with compound library (24 hours after plating). Each plate contained 32 control wells (DMSO only, yellow columns). Treated FD-NCs were analyzed by quantitative RT-PCR (48 hours after treatment). b, FACS-purified HNK1+ FD-NCs. c, Number of different batch of purified FD-NCs during expansion. d, Representative images of 384-well-plated FD-NC and CalceinAM staining. e, Robustness of primary HTS read-out: Control run statistics with average Ct values (AVG), their standard deviation (STDEV) and coefficient of variation (CV) and the Z’ values are summarized for plates 1, 2, and 3.

Screening of 6,912 compounds

We screened a library of 6,912 small molecules (20,736 treatment conditions in total). The library contained a broad set of biologically active and structurally diverse compounds from multiple commercial sources covering FDA approved drugs, known bioactives and experimental substances[13]. Compounds were screened at 10 µM in 1% DMSO (v/v) with each plate containing 1% DMSO (v/v) control wells. Z' values for the control plates in all three sets of the screen indicated excellent reproducibility (Suppl. Fig. 4). Using fold difference analysis of WT-IKBKAP to S18 ratios and a binning method taking into account robustness of replicate data, we nominated 28 compounds (0.4% hit rate) as potential rescuers of the WT-IKBKAP levels. In addition, we used an alternative rank-based selection algorithm that nominated an additional 15 candidate hits (Fig. 2a). For the rank-based method, we calculated the average WT-IKBKAP rank across the three replicate wells (ranks from 1–352), following elimination of compounds with low S18 values and correction for respective MU-IKBKAP ranks (see Material and Methods). This method was used as complementary strategy to the “fold difference” method to reduce the impact of variability in lowly expressed genes and to find potential compounds that selectively induce WT-IKBKAP (increase ratio of WT- to MU-IKBKAP). The list of 43 nominated compounds including structure information is provided in Supplementary Table 2). Out of 43 hits, only 35 were commercially available for further studies. After additional validations including dose-response and cytotoxicity assays (12 different concentration points), use of alternative housekeeping gene (GAPDH), and finally re-validation of WT-IKBKAP expression under standard 6-well culture conditions (Fig. 2a,b), we pursued 8 hits for further studies: Glucosaminic acid (); SKF-86466 Hydrochloride (); Phenindione (); Dihydrojasmonic acid (); Cyclosporine (); Cyproterone (); Atropine (); 5-Thia-1-azabicyclo(4.2.0)oct-2-ene-2-carboxylic acid, 7-(((2Z)-(2-amino-4-thiazolyl)(methoxyimino)acetyl)amino)-3-(((5-(carboxymethyl)-4-methyl-2-thiazolyl)thio)methyl)-8-oxo-, disodium salt, (6R,7R) (). Interestingly, the 8 compounds significantly increasing WT-IKBKAP levels in FD-NC precursors showed differential potency for IKBKAP induction in other cell types such as undifferentiated FD-iPSC, FD-fibroblasts and FD-lymphoblasts (Supplementary Fig. 5a–c). The tissue-specific bias in regulating WT-IKBKAP levels supports the argument for performing primary screens in symptom-relevant cells such as FD-NCs. Surprisingly, compounds did not trigger significant increases in WT-IKBKAP in control-NCs (derived from H9 hESCs; Supplementary Fig. 5d) suggesting a possible feedback mechanism preventing supra-physiological WT-IKBKAP levels in control cells.

Figure 2

Results of 6912 compounds screen in FD-NCs

a, Out of 6,912 compounds 28 and 15 compounds were nominated as hits respectively using two distinct selection algorithms (see text for details). A total of 8 compounds were pursued for detailed follow up studies under non-HTS conditions. b, Validating impact of treatment with 8 compounds on WT-IKBKAP expression in FD-NC maintained under 6-well plate culture conditions. c, Reproducibility of HTS screening results in FD-NC derived from independent FD-iPSC clones from the same (FD4) and from a different (FD101) patient. d, IKAP protein levels after treatment with individual compounds. e,f, ASCL1 (e) and SCG10 (f) expression in FD-NC following long-term treatment with 8 compounds from FD-iPSC to FD-NC stage. g, The wound-healing (scratch) assay in FD-NCs following long-term treatment. n = 3–5; * P < 0.05; ** P < 0.01; *** P < 0.001. All values are mean and s.d. F.C., Fold Change based on DMSO. N.S., Not Significant.

We next tested whether regulation of WT-IKBKAP expression by the 8 compounds is reproduced in FD-NCs from additional iPSC clones either from the same (clone 4) or a different (clone 101) patient (Fig. 2c). All compounds with the exception of THIA showed comparable potency across the various FD-NC precursor lines suggesting that the effects are generally not clone or patient dependent. THIA was shown to increase WT-IKBKAP in most clones when testing FD-iPSC NCs from yet additional patients (Supplementary Fig. 5e).

Hit validation and rescue studies

Only WT-IKBKAP but not MU-IKBKAP is translated into IKAP protein. We observed that several of the 8 compounds increased IKAP protein expression as measured by semi-quantitative Western blotting (Fig. 2d). Induction of IKAP protein levels was comparable to that observed following treatment with kinetin (K), the compound previously shown to partially rescue disease phenotypes in our FD-iPSC model[2], and comparable to the levels in NC precursors derived from hESCs or control iPSC lines (H9, C14; Fig. 2d). There was a significant (p = 0.0343) correlation between WT-IKBKAP transcript and IKAP protein levels, while no such correlation was seen between protein expression and MU IKBKAP (Supplementary Fig. 5f–h). None of the 8 compounds significantly affected cell proliferation or cell death, parameters that could indirectly affect IKBKAP levels (Supplementary Fig. 6).

While short-term (48 hours) compound treatment was sufficient to induce a robust increase in expression of WT-IKBKAP (Fig. 2b), it did not affect expression of autonomic markers such as ASCL1 (Mash1) and SCG10 (Supplementary Fig. 7a,b). Therefore we next tested the potency of each compound in long-term assays by treating cells for 28 days (from FD-iPSC through FD-NC stage). None of the compounds affected the percentage of HNK1+ FD-NC-precursors at day 28 (Supplementary Fig. 7c) suggesting that treatment does not interfere with cell differentiation. The levels of WT-IKBKAP in purified FD-NCs were increased for most compounds following long-term treatment compared with DMSO group (Supplementary Fig. 7d). Importantly, we observed significant increases in ASCL1 (Mash1) and SCG10 expression, particularly following treatment with SKF (Fig. 2e,f). Remarkably, SKF rescued ASCL1 (Mash1) and SCG10 expression to the levels found in NCs from wild-type lines. Despite a nearly complete rescue of autonomic marker expression, none of the 8 compounds rescued the reduced migratory propensity observed in FD-NC precursors (Fig. 2g), a result that mimicked our previous findings with kinetin[2]. It also remains to be determined whether long-term SKF treatment is required to capture the developmental stage most sensitive for rescuing autonomic neurogenesis or whether long-term treatment starting at late time points can be equally effective.

SKF implicates α2AR signaling in the regulation of IKBKAP

In a first attempt to address potential mechanisms by which compounds can increase IKBKAP expression, we focused on SKF (dose-response in Supplementary Fig. 7e). SKF is a known α2-adrenergic receptor (α2AR) antagonist without known inverse agonist activity[14]. SKF showed particularly promising results in rescuing expression of autonomic markers raising the question whether α2AR activity could be involved in regulating IKBKAP expression. The family of α2ARs acts as G protein-coupled receptors (GPCRs) in association with the Gi heterotrimeric G-protein. There are three different subtypes of α2ARs: α2A, α2B, α2C. The alpha subunit of the inhibitory G protein (dissociated Gi) associates with adenylcyclase and inhibits its activity resulting in decreased intracellular cAMP levels. We first confirmed expression for each of the three subunits (α2A, α2B and α2C) of α2AR in undifferentiated FD-iPSCs, differentiated FD-iPSC (Day 12 of neural induction) and purified/expanded FD-NCs by RT-PCR[15] (Fig. 3a, Supplementary Fig. 8a,b and Supplementary Table 1). Next we tested functional analogs of SKF that are either generic α2AR antagonists (Yohimbine, YOH) or specific to each subunit of α2AR (BRL-44408 (BRL) for subunit A, Imiloxan (IMI) for B, Spiroxatrine (Spi) and JP-1302 (JP) for C and RX821002 (RX) for D). YOH, BRL, IMI, Spi and JP (at 1 µM) significantly increased IKBKAP level immediately following exposure (Fig. 3b,c; 30 min. exposure), whereas no such effects were observed upon treatment with RX. RX exhibits selectivity for α2D-adrenoceptors[16], a receptor subtype expressed in murine but not in human tissues. IKAP protein levels were significantly increased compared to control (Fig. 3d) 48 hours after treatment with YOH or SKF. Despite the dramatic spike of IKBKAP mRNA induction 30 minutes after YOH treatment, the levels of mRNA and protein induction at 48 hours were comparable between YOH and SKF.

Figure 3

Functional studies on the mechanism of SKF action

a, Expression of α2 adrenergic receptor subunits (A, B and C) in FD-NC. b,c, WT-IKBKAP expression in FD-NC following treatment with other α2-adrenergic receptor antagonists (RX, RX821002; IMI, Imiloxan; Yoh, Yohimbine; Spi, Spiroxatrine and JP, JP-1302). d, IKAP protein level after treatment of SKF and YOH (48 hrs). e, Time course of WT-IKBKAP expression following treatment with Xylazine (Xyl), p-Iodoclonidine hydrochloride (pCI), Kinetin (K), SKF (S) and Yohimbine (Y) or DMSO control (30 min). f,g, WT-IKBKAP expression in FD-NC after SKF or YOH (Y) treatment and co-treatment with Xyl (30 min, f) or various pre-/post-treatments (g). h, cAMP levels following SKF or Yoh treatment i, cAMP levels after treatment of inhibitor and stimulators in α2 adrenergic receptor pathway. j, WT-IKBKAP expression in FD-NC after treatment with SKF, Yoh, cAMP and Forskolin (FK) in the presence or absence of the specific PKA inhibitor compounds H89 and KT5720 (10 µM). k,l, Representative images (k) and quantification (l) of phosphorylated CREB levels in treated FD-NCs. m, Luciferase activity of IKBKAP promoter following each treatment. n = 3–5; * P < 0.05; ** P < 0.01; *** P < 0.001. All values are mean and s.d. F.C., Fold Change based on DMSO. N.S., Not Significant. Scale bars, 20 µm.

Functional analogs of SKF, such as YOH, were included in our HTS collection but were not identified as ‘hit’ in our initial analysis. This is likely due to the fact that the compound concentrations for the primary screen (10 µM) were suboptimal for YOH (Fig. 3b), and that a different time period was selected between compound exposure and IKBKAP measurement (30 minutes versus 48 hours). Exposure of FD-iPSC derived NCs to agonists of α2AR (Xylazine or p-Iodoclonidine hydrochloride) resulted in time-dependent decreases in IKBKAP levels (Fig. 3e). However, the combination treatment of Xylazine and SKF or YOH did not significantly decrease IKBKAP levels (Fig. 3f) suggesting complex interactions of agonists and antagonists at α2AR. To further address the specificity of α2AR modulation on IKBKAP expression we tested for the impact of SKF and Xylazine following a sequential, time-dependent treatment paradigm. FD-NCs were exposed to DMSO, Xylazine or SKF (20uM, 10uM or 2uM) for 5 minutes and then switched to a medium containing DMSO, Xylazine or SKF (20uM, 10uM or 2uM) for an additional 30 minutes in various permutations (Fig. 3g). We observed that the effects of SKF pre-treatment on FD-NC were abolished by Xylazine. Conversely, the effects of Xylazine pre-treatment on FD-NC were overcome by SKF treatment in dose-dependent manner further arguing that SKF acts primarily through modulating A2AR.

Regulation of CREB phosphorylation is a well-known downstream mediator of α2AR signaling mediated cAMP-induced regulation of pKA. Accordingly α2-adrenoceptor antagonists, treatment with either SKF or YOH led to an increase in intracellular cAMP levels that could be blocked by exposure to Xylazine (Fig. 3h, i). We observed that treatment with cell permeable dibutyryl-cAMP (cAMP) or Forskolin (FK) was able to mimic SKF effect (Fig. 3j). Furthermore, exposure to SKF, IMI and YOH increased the levels of phosphorylated CREB (p-CREB) in FD-NCs while exposure to H89 or KT5720 (selective Protein Kinase A inhibitors) blocked the effect of both SKF and YOH on IKBKAP expression (Fig. 3k,l). These data confirm that pKA activity is required to mediate SKF and YOH function. IKBKAP promoter analysis using the Transcription Element Search System (TESS, http://www.cbil.upenn.edu/cgi-bin/tess/tess) shows the presence of conserved CREB binding domains[17]. We developed a luciferase assay based on the corresponding human IKBKAP promoter region which demonstrated strong promoter activation following SKF treatment that was repressed by Xylazine or H89 (Fig. 3m). Our pharmacological studies strongly suggest that SKF induces IKBKAP expression, at least in part, via classical α2AR signaling and cAMP-mediated activation of pKA and pCREB. The regulation of cAMP levels and its impact on IKBKAP expression levels appear to follow a complex temporal pattern and likely intersect with other signaling pathways. Interestingly, CSP, GA and AT also showed evidence of increased cAMP or p-CREB levels suggesting that this pathway could contribute to the action of other molecules from our screen (Supplementary Fig. 8c,d). Kinetin, a compound known to act on IKBKAP splicing, did not activate cAMP in FD-iPSC NC precursors supporting the specificity of this effect. The different mechanisms of action for SKF and Kinetin point to the possibility of combining modalities for future therapeutic applications.

DISCUSSION

Currently there are no animal models of FD that faithfully recapitulate the clinical symptoms of the disease beyond measuring tissue-specific IKBKAP splicing[18],[19]. Other unresolved issues include our lack of understanding as to why SKF or other compounds including Kinetin do not rescue the migratory defect of FD-NCs and whether any of the compounds will affect degenerative aspects of the disease such as the survival of postmitotic sensory or autonomic neurons. In either case, an important challenge for the future will be to make rational decisions in selecting the most promising compounds to pursue for clinical translation. While there are obvious criteria such as data on safe in vivo levels of a compound in relation to its IKBKAP activity or the ability to the cross blood-brain and blood-ganglia barriers, additional in vivo assays will be required for pre-clinical testing of our candidate compounds in a disease background. On the other hand, it is tempting to speculate whether an "in vitro clinical trial" that systematically tests compounds in iPSC-derived cells from FD-patients to be treated may contribute to a rationale clinical trial design. Since the first reports on establishing human iPSCs the use of disease-specific cells for drug validation and drug discovery has been an important goal. Our data demonstrate that it is feasible to perform primary screens in iPSC-derived cells for thousands of clinically relevant compounds. The development of similar assays for other diseases could herald a new drug discovery paradigm moving us closer towards the era of personalized medicine.