|Home | About | Journals | Submit | Contact Us | Français|
In animal models, chemical disruption of the Hedgehog (Hh) signaling pathway during embryonic development causes severe birth defects including holoprosencephaly and cleft lip and palate. The exact etiological basis of correlate human birth defects remains uncertain but is likely multifactorial, involving the interaction of genetic and environmental or chemical influences. The Hh transduction mechanism relies upon endogenous small molecule regulation; conferring remarkable pathway sensitivity to inhibition by a structurally diverse set of exogenous small molecules. Here, we employed small molecule screening to identify human exposure-relevant Hh signaling inhibitors. From of a library of 4,240 compounds - including pharmaceuticals, natural products, and pesticides, three putative Hh pathway inhibitors were identified: tolnaftate, an antifungal agent; ipriflavone, a dietary supplement; and 17-β-estradiol, a human hormone and pharmaceutical agent. Each compound inhibited Hh signaling in both mouse and human cells. Dose-response assays determined the three compounds to be 8-to-30 fold less potent than the index Hh pathway inhibitor cyclopamine. Despite current limitations in chemical library availability, which narrowed the scope of this study to only a small fraction of all human exposure-relevant small molecules, three structurally diverse environmental Hh signaling inhibitors were identified, highlighting an inherent pathway vulnerability to teratogenic influences.
Hedgehog (Hh) signaling is a key mediator of growth, morphogenesis, and differentiation during embryonic development, playing a critical role in patterning the brain, face, and other tissues. In utero chemical antagonism of Hh signaling can induce a wide range of developmental anomalies including holoprosencephaly (HPE), cleft lip and palate (CLP), and limb defects (Coventry et al., 1998; Cordero et al., 2004; Lipinski et al., 2007; Lipinski et al., 2008a). While occurring infrequently in live births (1.3 in 10,000), HPE is estimated to occur in 1 in 250 human conceptuses (Matsunaga and Shiota, 1977; Leoncini et al., 2008). Characterized by median forebrain deficiency, HPE frequently occurs with facial abnormalities including hypotelorism, midface hypoplasia, and median and lateral forms of CLP. Isolated CLP represents the most common birth defect form of newborns occurring in approximately 1/700 births. The etiological basis for both HPE and CLP is poorly understood, with the preponderance of evidence suggesting an interaction of genetic predisposition with environmental and/or chemical influences (Murray, 2002; Graham and Shaw, 2005; Dubourg et al., 2007, Gritle-Linde, 2009; Schachter and Krauss, 2009). The Hh pathway is a tantalizing common target for such interactions. In addition to the finding that Hh antagonist exposure can induce HPE and CLP in animal models, heterozygous hypomorphic mutations in the Shh gene are associated with human HPE (Nani et al., 1999; Maity et al., 2005).
The Sonic Hedgehog (Shh) gene encodes a secreted, cholesterol-modified ligand, which binds to the membrane bound Hh ligand receptor, Patched (Ptc1). In the absence of Hh ligand, Ptc1 is postulated to secrete endogenous small molecules that act to inhibit activity of the signal mediating protein Smoothened (Smo). (Taipale et al., 2002; Bijlsma et al, 2006). Upon Hh ligand binding, Ptc1 undergoes a conformational change and no longer suppresses Smo, triggering a downstream signaling cascade culminating in the modulation of Hh target gene expression via the three vertebrate Gli transcription factors (reviewed in Ingham and McMahon, 2001). The plant alkaloid cyclopamine was identified as a teratogen in the 1960s, when it was discovered to be the causative agent of HPE-related malformations in lambs born of sheep grazing on the plant Veratrum californicum (Keeler, 1978). Following the characterization of the Hh signaling pathway and the HPE phenotype of Shh null mice (Chiang et al., 1996), cyclopamine was shown to inhibit Hh signaling transduction by binding to and preventing activation of Smo (Chen et al., 2002a). Cyclopamine exposure is not a thought to be a significant health risk for humans but the history of its teratogenic effects highlights the vulnerability of the Hh pathway to environmental disruption.
Recent studies have implicated inappropriate activation of the Hh pathway in a variety of cancers (reviewed in Briscoe and Therond, 2005) and high-throughput screens of synthetic chemical libraries have identified several potent, structurally diverse small molecule pathway antagonists as potential therapeutic agents (Chen et al, 2002b; Frank-Kamenetsky et al., 2002; Williams et al. 2003). The purpose of the study presented here was quite different; to assess the potential for human exposure to Hh signaling antagonists as a first step in examining the hypothesis that an interaction between environmental/chemical and genetic influences may contribute to the incidence of HPE and CLP in humans. As opposed to previous efforts to identify novel, synthetic antagonists, we describe here the identification of Hh signaling inhibitors from a library of compounds selected for the potential of human exposure.
A total of 4,240 compounds were screened, comprising several individual small molecule libraries including the Prestwick Chemical Library (Prestwick Chemical, Illkirch, France); the Library of Pharmacologically Active Compounds (Sigma, St. Louis, MO); and the Spectrum Chemical Collection and the Agro Plate Collection (Microsource Discovery Systems, Gaylordsville, CT). These libraries primarily consist of currently marketed drugs, failed drug candidates, natural products, hormones, and pesticides.
The seeding density, fluid volumes, and incubation schedules for the screening approach described herein were determined empirically through extensive pilot experiments to maximize signal/background ratio and minimize well to well variability. Immortalized Ptc1-/- MEFs, which demonstrate ligand-independent Hh signaling measurable by β-galactosidase activity, were plated in 96-well plates at a density of 2,500 cells/well in 100 μl DMEM with 2% FBS. Cells were allowed to attach for 6 hrs. Following, compounds in DMSO were added directly to culture media of individual wells yielding a final drug concentration of 10 μM (1.0% final DMSO concentration). After 64 hrs incubation in semi-humidified chambers, media was removed and 20 μl Reporter Lysis Buffer was added. Following 20 min incubation, 180 μl of ONPG substrate/Z buffer (Miller, 1972) was mixed into the cell lysate. Plates were then sealed in plastic bags and incubated at 37° C for 6 hrs in semi-humidified chambers. Finally, 420 nm wavelength absorbance was measured using a Tecan Safire II plate reader.
In each 96-well plate, two wells were treated with 1 μM cyclopamine as a positive control with eight wells treated with DMSO alone as a negative control. Experimental compounds were added to individual wells. To determine background absorbance, four wells were not seeded with cells but treated with vehicle. Following background subtraction, ratios were calculated for the absorbance of each compound treated well, divided by the absorbance measured in vehicle treated wells. Cyclopamine treatment resulted in a mean decrease in absorbance of 40±5%. Accordingly, experimental compounds yielding a 35% or greater reduction in absorbance (n=251) were selected for further evaluation.
Recognizing that reporter activity could also be decreased by cytotoxic effects, we next assessed the impact of these 251 compounds on the cellular monolayer (Figure 1), which was left intact following passive lysis and absorbance measurements in positive and negative control wells. Visualization at 40× magnification revealed that 194 compounds caused gross disruption of cellular monolayer, an endpoint indicative of cytotoxicity. A similar outcome was produced when each of these compounds was rescreened at a concentration of 1 μM. The majority of these compounds were flagged as promiscuous inhibitors in a database maintained by the small molecule screening facility, which tracks all assay results and flags compounds that identify as a hit in two or more screens. Compounds that did not cause monolayer disruption (n=57) were selected for further screening.
As the purpose of this screen was to identify Hh signaling inhibitors with human exposure potential, we sought to eliminate compounds without demonstrable routes for human exposure from further testing. Extensive literature and database searches failed to identify routes of exposure for 30 of these compounds, most of which were failed pharmaceutical drug candidates and semi-synthetic drug analogs.
Follow-up screening of the remaining 27 compounds was performed using Shh-LIGHTII cells as described previously. Cells were treated with compounds at 0.08, 0.40 and 2.0 μM concentrations. Those producing dose-dependent inhibition of reporter activity (n=6) were then tested for cytotoxic activity via MTT assay. Three compounds were excluded for causing cytotoxicity, leaving three lead compounds; 17-β-estradiol, tolnaftate and ipriflavone. A schematic illustrating the selection criteria used in this study is shown in Figure 1. The disposition of the 57 compounds that inhibited β-galactosidase activity without disrupting cell monolayer at each step of the subsequent selection process is listed in supplemental Table 1.
Assays were performed as described previously (Lipinski et al., 2007). NIH3T3 cells with stable expression of Gli-responsive firefly luciferase and TK-renilla luciferase (Shh LIGHTII) (Taipale et al., 2002) were plated in Multiwell Primaria™ 24 well plates (Falcon, Franklin Lakes, NJ) at 1.5 × 105 cells per well in 400 μl media and allowed to attach overnight. Following, media were replaced with DMEM containing 1% FCS ± 1 nM octylated Shh peptide (Curis, Cambridge, MA) and small molecules at given concentrations. Following a 48 hr incubation period, cell lysates were harvested and reporter activity was determined using a dual luciferase reporter assay system (Promega, Madison, WI).
Immortalized wildtype mouse embryonic fibroblasts (iMEFs) expressing hGli1-GFP or GFP alone (Lipinski et al., 2008b) were plated in DMEM containing 10% FBS at 2×105 cells/well in a Primaria 24 well plate and allowed to attach overnight. Following, media were replaced with DMEM containing 1% FBS ± octylated-Shh peptide at a concentration of 1 nM, ± small molecule compounds dissolved in DMSO to achieve given concentrations. Normal dermal human fibroblasts (primary foreskin fibroblasts) were generously provided by Dr. Lynn Allen-Hoffman. Cells were grown and plated in fibroblast growth media (Lonza, Rockland, ME) in Primaria 24 well plates and allowed to attach overnight. Following, media were replaced with DMEM/F12 containing 1% FBS ± octylated-Shh peptide at a concentration of 5 nM, ± lead compounds dissolved in DMSO to achieve given concentrations. A higher concentration of Shh peptide was employed because the human fibroblasts demonstrated a less robust transcriptional response to ligand stimulation than the murine cell lines. After a 48 hr incubation period, RNA was harvested, and gene expression was determined by Real-Time RT-PCR using species-specific primers as described in (Zhang et al., 2007).
The screening approach, selection process, and follow-up assays carried out in this study were designed to identify compounds with potential for human exposure that inhibit Hh signal transduction independent of cytotoxicity. A screen of 4,240 small molecules was performed using immortalized Ptc1-/- MEFs (Figure 1). Ptc1 is a conserved Hh target gene encoding a transmembrane receptor which represses signal transduction in the absence of Hh ligand. LacZ knocked into the Ptc1 locus results in constitutive pathway activation with β-galactosidase activity serving as a surrogate measure of pathway activity. At a 10 μM concentration, 251 compounds reduced β-galactosidase activity greater than 35% relative to vehicle. Of these, 194 compounds caused significant disruption of cell monolayer; a result indicative of cytotoxic activity. These compounds were re-screened at a concentration of 1 μM but even at this concentration, reporter activity reduction was associated with disruption of the cellular monolayer in each case. Of the remaining 57 compounds, extensive literature and database searches examining the plausibility of human exposure excluded an additional 30. The 27 remaining compounds with confirmed routes of human exposure were tested further.
To exclude compounds that reduced β-galactosidase activity independent of Hh signaling or acted through other non-specific mechanisms, inhibition activity was tested in Shh-LIGHTII cells incorporating a luciferase reporter. Only 6 of 27 compounds caused a dose-dependent inhibition of Shh ligand-induced reporter activation. Three of these compounds caused associated cytotoxicity as determined by MTT assays (data not shown) and were excluded from further testing. The names of the 57 compounds that reduced β-galactosidase activity without disrupting cell monolayer as well as their disposition at each step of the subsequent selection process is listed in supplemental Table 1.
In total, three compounds were identified that inhibited Hh signaling activity in two independent cell-based reporter assays without disrupting cellular monolayer or causing apparent cytotoxicity: tolnaftate, an antifungal agent; ipriflavone, a dietary supplement; and 17-β-estradiol, a human hormone and pharmaceutical agent.
Using Shh-LIGHTII cells (Taipale et al., 2002), dose response assays were performed to determine potency of the lead compounds relative to cyclopamine, the index Hh inhibitor, and tomatidine, a related dietary alkaloid and relatively weak Hh signaling inhibitor (Lipinski et al., 2007). We found that ipriflavone, 17-β-estradiol, and tolnaftate yielded fully efficacious dose-response curves with slopes similar to that of tomatidine and cyclopamine (Figure 2). Approximate EC50 and EC90 concentrations for each compound were generated from these dose-response curves.
Comparison of molecular structure among previously identified Hh inhibitors and the three compounds identified here reveals no compelling similarity. Steroids similar in structure to 17-β-estradiol, such as testosterone and progesterone, as well as genistein, the chemical precursor of synthesized ipriflavone, were represented in screened chemical libraries but were not identified in the primary screen (structures shown in Figure 3). When tested in this dose-response assay, progesterone demonstrated no inhibition activity while genistein caused significant inhibition only at the highest tested concentration (18.9 μM). These observations demonstrate that the Hh signaling inhibition capacity of ipriflavone and 17-β-estradiol is structure specific and is not mimicked by related compounds.
To validate the three identified compounds as bona fide Hh pathway antagonists, we tested the ability of each to inhibit endogenous target gene activation in normal cells. Applying the respective EC50 and EC90 concentrations from Figure 2, we found that each compound caused a dose-dependent inhibition of Shh ligand-mediated Gli1 induction in mouse embryonic fibroblasts (MEFs) (Figure 4A). Moreover, the EC90 concentration of each lead compound completely blocked Gli1 induction in Shh-responsive primary normal dermal human fibroblasts (Dennler et al., 2007) (Figure 4B).
The use of Ptc1-/- cells in initial screening assays was expected to identify inhibitors that act downstream of Ptc1. Most identified small molecule Hh signaling antagonists appear to act at the level of Smo but Lauth et al., (2007) identified two small molecule inhibitors of the Gli transcription factors. The effect of the three inhibitors identified here on Gli1 mediated transcription was tested using MEFs overexpressing hGli1. These cells demonstrate constitutive pathway activation as evidenced by increased mGli1 expression (Figure 5 inset). In the absence of Shh ligand, constitutive pathway activity is driven by the overexpression of hGli1 which directly activates target gene transcription independent of Gli2 and Gli3 (Lipinski et al., 2006). The elevated expression of mGli1 was not affected by treatment with any of the three compounds at their respective EC90 concentrations (Figure 5). These observations suggest that each of the three compounds identified here acts downstream of Ptc1 and upstream of Gli1. The similarity of slopes in the dose-response curves performed for these compounds to the slope of the dose-response curve for cyclopamine suggests further that these compounds act, like cyclopamine, at the level of Smo.
Small molecule screening identified three compounds with demonstrable human exposure as Hh signaling inhibitors, which appear to act downstream of Ptc1 and upstream of Gli1 with variably weaker potency than the index Hh inhibitor cyclopamine. The screening approach and execution appears to have been independently validated by a recent report examining the effect of exogenous 17-β-estradiol exposure on chondrocranial development in the zebrafish (Fushimi et al., 2009). The authors demonstrated that exposure above 5 μM caused craniofacial dysmorphia grossly mimicking that of both shh-mutant and cyclopamine-exposed zebrafish (Wada et al., 2005). Exposure to 20 μM 17-β-estradiol caused a reduction in ptc1 expression while not affecting the expression of shh. That these effects were not reversed by an estrogen receptor antagonist suggests an independence from nuclear estrogen receptor signaling and, along with the data presented herein, argues that 17-β-estradiol acts as a direct Hh pathway antagonist.
Each of the three compounds identified here has recognized routes for human exposure but none has been linked to human birth defects consistent with Hh signaling inhibition. One animal study investigating the teratogenic potential of tolnaftate, a synthetic antifungal agent, was negative (Noguchi et al., 1966). Treatment of pregnant mice and rats by subcutaneous injection of 1,000 mg/kg/d tolnaftate from E8.5 to 13.5 (where morning of vaginal plug = E0), did not cause a significant incidence of fetal abnormalities. No report of tolnaftate teratogenicity in humans has been published.
Ipriflavone is a synthetic isoflavone of worldwide use as a supplement for its touted anabolic and bone density building properties. Several studies have examined in vivo pharmacokinetics of ipriflavone in animal models as well as humans. In rats, the bioavailability of ipriflavone was approximately 24% (Kim et al., 2002). In healthy human volunteers, ipriflavone exhibited a half-life of 10-12 hrs (Ma et al., 1997). To date, no studies examining the teratogenic potential of ipriflavone in humans or animal models have been published.
17-β-estradiol, the most significant human estrogen, plays critical roles in sexual development, reproduction, and other biological effects. Serum levels fluctuate during the normal menstrual cycle, generally remaining below 25 nM. Circulating levels rise during pregnancy but generally remain under 100 nM, significantly lower than our observed in vitro EC50 concentration for Hh signaling inhibition of 2.75 μM. Exogenous 17-β-estradiol is also administered for several clinical purposes including hormonal supplementation, birth control, and prevention of osteoporosis; however, levels achieved by exogenous administration are generally within the physiologic range.
We find it unlikely, under normal circumstances, that humans are exposed to inhibitory concentrations of any one of these compounds in isolation. Indeed, epidemiological studies would likely have identified an environmental chemical with high potency and teratogenic potential. However, this does not exclude a role of these compounds or other yet to be identified environmental Hh inhibitors as contributors to etiologically-complex birth defects like CLP and HPE. The small molecule libraries sampled here represent only a small fraction of the total number of compounds with relevant human exposure and it is likely that other, more potent inhibitors exist in the environment. Moreover, Hh signaling antagonists can produce additive inhibitory effects (Lipinski et al., 2007).
A particularly noteworthy aspect of HPE etiopathogenesis in relation to this line of study is the presence of heterozygous mutations in several Hh pathway genes in both affected probands as well as clinically normal obligate carriers (Nanni et al., 1999; Roessler et al. 2003). This variable phenotypic penetrance suggests an interaction of multiple genetic lesions or single lesions and environmental or chemical insults. That three of 4,240 evaluated compounds exhibited specific Hh pathway inhibitory activity, demonstrates the potential for Hh pathway perturbation as a result of environmental exposure. These findings justify additional, more comprehensive searches for Hh inhibitors, as well as studies examining the interaction of such chemical insults in the context of identified genetic predispositions.
Supplemental Figure 1 – Subsequent testing of 27 compounds that lowered reporter activity without disrupting cellular monolayer in initial screening. Compounds were screened for demonstration of plausible routes of human exposure (green fill indicates a route was identified), dose-dependent inhibition in Shh-LIGHTII cells (inhibition activity is indicated by green fill) and cytotoxicity (green indicates negative). Note that tolnaftate and pentamidine isethionate were present in two different chemical libraries that were screened.
Small molecule screening was performed at the Keck-UWCCC Small Molecule Screening Center at the University of Wisconsin-Madison with the helpful guidance of Noël Peters. Shh peptide was kindly provided by Curis. This work was supported by grants from National Institutes of Health (DK065303); National Institute of Environmental Health Sciences (T32-ES00715); National Institutes of Health (DK056238).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.