A High-Throughput,Homogeneous,Fluorescence Polarization Assay for Inhibitors of Hedgehog Protein Autoprocessing

Abstract

Hedgehog (Hh) signaling plays an important role in embryonic patterning and adult stem cell renewal but has recently been found also to be involved in certain stem cell cancers. One of the first steps in Hh signaling is the autoprocessing of Hh protein, in which the C-terminal domain (Hh-C) catalyzes a cholesterol-dependent autocleavage reaction that leads to the production of the cholesterol ester of the N-terminal Hh domain (Hh-N), thereby yielding a signaling molecule that activates the Hh pathway by binding to the Patched receptor. This article describes an in vitro, homogeneous assay system that measures changes in fluorescence polarization that accompany the cholesterol-dependent autocleavage of Hh protein. The assay system makes use of a modified Hh protein in which Hh-N, which is not essential for autocleavage, is replaced by a 25-residue peptide containing a tetracysteine motif, complexed with a bisarsenical fluorophore. The assay is quite robust and easily adapted to high-throughput screening in 384-well plates with Z′ factors above 0.8. It has been used to screen the National Institutes of Health Clinical Collection, which has led to the identification of 2 compounds that inhibit the cholesterol-dependent autocleavage of Hh protein at micromolar concentrations.

Introduction

Hedgehog (hh) signaling plays an important role in embryonic patterning and adult stem cell renewal but has recently been found also to be involved in certain stem cell cancers, including pancreatic, small cell lung, intestinal, and prostate, all of which are notoriously difficult to treat. ¹ One of the first steps in Hh signaling is the autoprocessing of Hh protein, in which the C-terminal domain (Hh-C) catalyzes a cholesterol-dependent autocleavage reaction that leads to the production of the cholesterol ester of the N-terminal Hh domain (Hh-N), thereby yielding a signaling molecule. ² The secreted Hh ligand binds to the Patched receptor (Ptch1), leading to activation of Smoothened (Smo) and a signal transduction cascade that culminates in the activation of Gli transcription factors and the transcription of Gli and other genes. ³

Most research on the role of Hh signaling is focused on the series of events that lie between the activation of Smo and the activation of Gli, with the aim of understanding this complex signaling pathway and in the hope of developing potential cancer therapeutics. In contrast, very little attention has been given to the role of Hh autoprocessing and modification, despite the fact that it constitutes the initial step in the Hh signaling pathway. Owing to the absence of pharmacologic tools for its study, the role of Hh autocleavage and esterification with cholesterol in Hh signaling, both in the developing embryo and in the growth of Hh ligand-dependent cancers, is still not entirely clear.

All earlier studies on the autoprocessing of Hh protein used denaturing sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) to follow the autocleavage reaction, ^2,4-6 which does not lend itself easily to a rapid, quantitative assay. This article describes a high-throughput screen (HTS) for identifying compounds that attenuate Hh autoprocessing and can be used as research tools for dissecting the role of autoprocessing and the attendant cholesterol modification in Hh function, both in the many different contexts of embryonic development and in Hh ligand-dependent stem cell cancers. The HTS makes use of an in vitro, homogeneous assay system that measures changes in fluorescence polarization that accompany the cholesterol-dependent autocleavage of Hh protein. It involves replacing the signaling domain of Hh protein, which is not essential for autocleavage, with a 25-residue peptide containing a FlAsH-tag, which is complexed with a bisarsenical fluorophore based either on fluorescein (FlAsH) or resorufin (ReAsH). ⁷

The data presented in this article show that this assay is quite robust and easily adapted to HTS in 384-well plates with Z′ factors ⁸ above 0.8. The assay has been used to screen a small compound library and has led to the identification of 2 compounds that inhibit the cholesterol-dependent autocleavage of Hh protein at micromolar concentrations.

Materials and Methods

Chemicals and reagents

4′,5′-bis(1,3,2-dithioarsolan-2-yl)-fluorescein (FlAsH) and 4′,5′-bis(1,3,2-dithioarsolan-2-yl)-resorufin (ReAsH) as the 1,2-ethanedithiol (EDT) complexes were synthesized as described ^7,9 or purchased from Invitrogen (Carlsbad, CA). The National Institutes of Health (NIH) Clinical Collection was obtained from BioFocus (South San Francisco, CA)

Plasmid construction

The tetracysteine motif, CCPGCC or FlAsH tag, was inserted into the inducible coding region of plasmid pET45b (Novagen, Madison, WI) adjacent to the His-tag, using the Kpn I and NotI restriction sites and appropriate synthetic oligonucleotides (5′-GTGTTGCCCGGGTTGCTGTGC-3′ and 5′-GGCCGCACAGCAACCCGGGCAACACGTAC-3′), custom synthesized by Invitrogen, to yield plasmid pET45H. A plasmid encoding the hedgehog protein of Drosophila melanogaster was generously provided by Dr. Kent Nybakken of this institute. The entire hedgehog coding sequence was amplified by the polymerase chain reaction using Failsafe polymerase (Epicentre, Madison, WI) and the primers 5′-GGCGGCCGCTTCCCACGTGCACGGCTG-3′ and 5′-GGCGGCCGCTCGTTGGGATCAGGATCGCTC-3′ (Invitrogen). The amplification product was cloned into PCR-XL-TOPO (Invitrogen), the recombinant plasmid was digested with NotI restriction endonuclease, and the 0.71-kb fragment containing the hedgehog coding sequence was inserted into the NotI site of plasmid pET45H to yield plasmid pET45Dmel. The predicted amino acid sequence of the 26.3-kDa hedgehog protein with its N-terminal His-tags and FlAsH-tags is shown in Figure 1 .

Fig. 1.

Deduced amino acid sequence of the recombinant Hh protein expressed by plasmid pET45Dmel. The His-tag and FlAsH-tag sequences are underlined, the site of autocleavage is indicated by the vertical arrow, and start of the autoprocessing domain, with the N-terminal sequence, CFT, is indicated.

Bacterial growth and protein expression

Escherichia coli Rosetta 2 (DE3) (Novagen), transformed with plasmid pET45Dmel, was cultured in Luria-Bertani (LB) medium supplemented with chloramphenicol (34 µg/mL) at 37 °C. In mid-exponential phase (A₆₀₀ = 0.5), isopropyl β-D-thiogalactopyranoside was added to a final concentration of 0.4 mM, and the cells were harvested by centrifugation 3 h later, washed once with Buffer A (20 mM Na phosphate [pH 7.5], 0.5 M NaCl, 1 mM p-toluenesulfonyl fluoride), and stored at −70 °C.

The bacterial cells were thawed, suspended in Buffer A (3 mL per 50 mL of culture), and disrupted in a French pressure cell at 4 °C. The extract was centrifuged at 20,000 g for 30 min, and the pellet was washed twice with Buffer A and extracted twice with 3 mL of Buffer B (20 mM Na phosphate [pH 7.5], 0.5 M NaCl, 8 M urea, and 1 mM p-toluenesulfonyl fluoride) to solubilize inclusion bodies. SDS-PAGE analysis of the solubilized inclusion bodies showed the major protein component to be the 26-kDa hedgehog protein. The inclusion bodies were stored at 4 °C but could also be frozen at −80 °C without significant loss of autoprocessing activity.

Fluorescence polarization assay for Hh autoprocessing

The assay was typically carried out in black 384-well plates (Nunc 264558) with 70 µL per well and was initiated by mixing equal volumes of a solution containing 1.4 µM (40 µg/mL) of solubilized inclusion bodies diluted in 1 mM Tris(2-carboxyethyl)phosphine (TCEP), 0.4 mM 2-(dimethylamino)ethanethiol, and 0.7 µM FlAsH-EDT₂, in Buffer D (20 mM Na phosphate [pH 6.9], 0.5 M NaCl, and 0.5 M L-arginine hydrochloride) with a solution of 68 µM cholesterol in 0.04% Triton X-100. In larger assays, liquid dispensing was done with a Thermo Fisher Multidrop 384 dispenser (Thermo Fisher, Waltham, MA). The mixtures were incubated at 25 °C, and fluorescence polarization was measured at 10-min intervals in a Tecan Safire 2 microplate reader (Tecan, Männedorf, Switzerland) with excitation at 390 nm and emission at 550 nm. A time delay of 300 ms was imposed between plate movement and flash to eliminate a fluorescence polarization (FP) artifact caused by vibration of the meniscus, which can give rise to systematic errors in FP values. To measure only the first step in the Hh autoprocessing reaction involving formation of the thioester intermediate ( Fig. 2 ), cholesterol in the second solution was replaced with neutral hydroxylamine (2.2 M final concentration) as a nucleophile to bring about the uncatalyzed cleavage of the thioester.

Fig. 2.

Mechanism of hedgehog (Hh) autoprocessing. The first step in autoprocessing is a self-catalyzed rearrangement of the peptide bond at the N-terminus of the autoprocessing domain to a thioester involving the side chain of its conserved N-terminal Cys residue. This is followed by transesterification to cholesterol, bound to the C-terminal autoprocessing domain, to yield the cholesterol ester of the signaling domain, which functions in intracellular signaling. Alternatively, the thioester intermediate can be cleaved by a strong nucleophile such as hydroxylamine.

Results

Assay development

Hh autoprocessing is a 2-step reaction, the first step being the N/S acyl rearrangement of the scissile bond between the signaling and the autoprocessing domains and the second the transesterification of the resulting ester intermediate with cholesterol ( Fig. 2 ). ¹⁰ The first step can be measured independently of transesterification by cleaving the thioester intermediate with high concentrations of strong nucleophiles such as hydroxylamine or thiols. ⁵

Our assay takes advantage of the fact that Hh autoprocessing requires only the C-terminal autoprocessing domain and that the N-terminal signaling domain can be replaced with any peptide, ² for example, a 25-residue peptide that contains a tetracysteine motif (FlAsH-tag), as shown in Figure 1 . The FlAsH-tag reacts almost irreversibly with a bisarsenical fluorophore based either on fluorescein or resorufin (FlAsH or ReAsH, respectively). ⁷ Autocleavage of the modified Hh protein leads to the conversion of a fluorescent 27-kDa molecule to 3 kDa, with a significant decrease in fluorescence anisotropy that can be measured by FP. Upon incubation at 25 °C, the FlAsH complex of the fusion protein undergoes autocleavage when incubated with the hydroxylamine or with cholesterol in the presence of a detergent such as Triton X-100, as indicated by molecular weight reduction ( Fig. 3A ) or a FP decrease ( Fig. 3B ). The time course of the FP measurements indicates a reaction half-time for the self-catalyzed reaction with cholesterol of about 20 min, considerably faster than the uncatalyzed reaction of the thioester intermediate with hydroxylamine ( Fig. 3B ). It should be noted that the change in FP is significantly greater in the hydroxylamine-induced cleavage reaction, suggesting that the anisotropy of the cholesterol adduct of the N-terminal cleavage fragment is greater than that of the free peptide produced by hydroxylamine cleavage. Similar results were obtained with the resorufin-based bisarsenical fluorophore, ReAsH ( Fig. 4 ), except that the change in FP upon autocleavage was somewhat less than with the FlAsH reagent. The FlAsH-based assay has excellent performance characteristics, with a Z′ factor of ≈0.8 in a 384-well format. Z′ factors were calculated from the relationship Z′ = 1 – [3 SD of sample – 3 SD of control]/[mean of sample – mean of control], ⁸ where SD refers to standard deviation. Typical values for SD in 384-well plates were 2.5 and for [mean of sample – mean of control] were 70 to 80.

Fig. 3.

Autoprocessing of the HhC fusion protein with an N-terminal FlAsH-tag. (A) Monitoring of the autoprocessing reaction by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and staining with Coomassie blue. The samples were incubated for 16 h at 25 °C under the conditions indicated. The position of the uncleaved precursor and the autoprocessing product is indicated at the right. (B) Monitoring of the autoprocessing reaction by fluorescence polarization (FP). FP was measured in duplicate wells of a 384-well plate upon excitation at 470 nm and emission at 550 nm, without nucleophile or with either 34 µM cholesterol or 2.2 M hydroxylamine, using 1 µM Hh protein and 1.7 µM FlAsH-EDT₂, which had been dialyzed after complex formation to remove unbound bisarsenical fluorophore.

Fig. 4.

Monitoring of the autoprocessing reaction by fluorescence polarization (FP) using ReAsH. FP was measured in triplicate wells of a 384-well plate upon excitation at 590 nm and emission at 632 nm, without nucleophile or with either 34 µM cholesterol or 2.2 M hydroxylamine, using 1 µM Hh protein and 4.4 µM ReAsH-EDT₂, which had been dialyzed after complex formation to remove unbound bisarsenical fluorophore.

Critical parameters and further assay optimization

The rate and extent of reaction of FlAsH-tagged peptides with FlAsH-EDT₂ depend on reagent concentration as well as on the presence of low concentrations of monothiols, ⁹ which makes the relative concentrations of these components critical factors in the FP polarization assay. TCEP (1 mM) was added to ensure full reduction of the Cys residues of the FlAsH-tag, and 2-(dimethylamino)ethanethiol (0.4 mM) was used to promote complex formation between the FlAsH-tag and the bisarsenical fluorophore. Under these conditions, FlAsH-bisarsenical complex formation went to completion in about an hour at 25 °C, requiring a preincubation period prior to running the FP assay. This made for a relatively reproducible procedure and was used in our pilot screen of the NIH Clinical Collection (see below) but may lead to some plate-to-plate variation in the context of a large HTS. Moreover, the presence of fluorescent impurities in some batches of FlAsH-EDT₂ or ReAsH-EDT₂ required careful optimization of the relative concentrations of the FlAsH-tagged protein and each lot of bisarsenical fluorophore.

We subsequently developed a more reproducible procedure for fluorescent labeling of the FlAsH-tagged Hh-C protein and the bisarsenical fluorophore, which used a molar excess of the fluorophore to induce complex formation, followed by dialysis against buffer containing 1 mM TCEP but no free thiol to remove unbound fluorophore. This dialyzed mixture gave reproducible assays over a period of 4 days and is therefore well suited for large-scale HTS efforts. The data shown in Figure 3B and Figure 4 were obtained by this optimized procedure.

Another important assay parameter is the manner of dispersion of the water-insoluble cholesterol substrate. We examined a number of detergents, of which octyl-β-D-glucopyranoside and Triton X-100 gave the most robust cholesterol-dependent activities. Triton X-100 was the preferred reagent because it gave stable cholesterol dispersions slightly above the critical micelle concentration (CMC) but could be diluted in the assay to 0.2%, which was just below the CMC and had no effect on autoprocessing activity but was still high enough to minimize “promiscuous” inhibition. ¹¹ Under these conditions, the apparent K_m for cholesterol was about 14 µM, compared to a K_m of 29 µM for the cholesterol analog, ergosterol ( Fig. 5 ).

Fig. 5.

Substrate kinetics of the autoprocessing of the HhC fusion protein with an N-terminal FlAsH-tag. The assay was conducted as described in the legend to Figure 3 , except that the cholesterol concentration was varied or replaced by ergosterol as indicated.

Pilot screen of the NIH Clinical Collection

To evaluate the suitability of our FP assay for HTS, we conducted a pilot screen of the NIH Clinical Collection. The 446 compounds of this library were assayed in duplicate at 25 µM, using the FlAsH-based assay in 384-well plates. Every assay plate contained at least 2 rows of negative controls without inhibitor and 2 rows of positive controls in which 100% inhibition was simulated by omitting the substrate cholesterol. The assay was conducted as described in Figure 3B , with a signal window of 23% and a Z′ factor of 0.81, with duplicate values in close agreement. Two compounds that showed about 65% inhibition came from wells NGP-001-02-B06 and NGP-001-04-C06 in the library and were identified as PubChem CID 5717 and 72303. Ten additional compounds, which inhibited between 25% and 40%, were not given further consideration.

The 2 hits with >50% inhibition in the pilot screen were subjected to secondary screening using the resorufin-based bisarsenical reagent, ReAsH, with excitation at 590 nm and emission at 632 nm, to rule out any interference of the compounds with the fluorescence measurements at 550 nm used in the primary screen. The secondary screen measured concentration dependence and also compared cholesterol- and hydroxylamine-dependent autocleavage. As shown in Figure 6 , both compounds (PubChem CID 5717 and 72303) inhibited cholesterol-dependent autocleavage and not the hydroxylamine-dependent reaction, suggesting that inhibition targeted the nucleophilic attack by cholesterol or cholesterol binding rather than the N-S acyl rearrangement. The measurement of inhibition as a function of time, which provides an indication of reversibility, showed that CID 72303 inhibited in a time-independent manner (i.e., irreversibly), whereas inhibition by CID 5717 was time dependent (i.e., reversible), as expected for a competitive inhibitor.

Fig. 6.

Secondary screening of inhibitors identified in the pilot screen. The 2 compounds from the National Institutes of Health Clinical Collection that gave more than 50% inhibition at 25 µM were assayed at the concentrations indicated either in the presence of 34 µM cholesterol in Triton X-100 dispersion or 2.2 M hydroxylamine as an artificial nucleophile, as indicated on the chart. Fluorescence polarization (FP) was measured in a 384-well plate upon excitation at 590 nm and emission at 632 nm, using 2 µM Hh protein and 0.4 µM ReAsH-EDT₂, either after 1 h (solid lines) or 4 h (dotted lines). (A) NGP-001-04-C06 or CID 72303; (B) NGP-001-02-B06 or CID 5717 (zafirlukast). (C) Structures of PubChem CID 72303 and CID 5717.

Discussion

The Hh signaling pathway is the subject of intense investigation, owing to its important role in embryonic development and certain types of cancer. ^1,3 However, most current research focuses on the steps that occur after the Hh ligand binds to the patched receptor (Ptch1). In contrast, the steps by which the Hh ligand is generated from its precursor form, which include 3 different processing events ¹⁰ and interaction with at least 2 additional proteins prior to binding to Ptch1, ^12,13 have not been studied in detail, not least owing to the lack of chemical probes for their dissection. Until the recent identification of a small molecule, robotnikinin, which binds to the sonic Hh signaling domain, ¹⁴ the only probes available for studying the early steps in the Hh pathway had been antibodies that bind to Hh-N and block its activity. ¹⁵ The assay described in this article represents the first attempt to identify inhibitors of the self-catalyzed autocleavage reaction, which is coupled to the conjugation of the Hh signaling domain with cholesterol. Such inhibitors would be valuable tools for investigating the role of Hh autocleavage and esterification with cholesterol in Hh signaling, about which there is still some uncertainty because catalytically impaired Hh mutants have significant residual function, ⁴ and artificial Hh-N variants without lipid modification can function in some signaling contexts (e.g., Fujita et al. ¹⁶ ).

Our FP assay relies on the reduction of anisotropy as the fluorophore-labeled N-terminal protein segment is reduced in size from 27 kDa to 3 kDa by hedgehog protein autoprocessing. When autoprocessing is induced by hydroxylamine as an artificial nucleophile, the change in FP is nearly 50% ( Figs. 3, 4 ), consistent with such a large change in Mr and a fluorescence lifetime of about 4 ns. ⁷ On the other hand, the FP change is much less when autoprocessing is coupled to nucleophilic attack by cholesterol on the scissile bond, being only 27% for the FlAsH fluorophore ( Fig. 3B ) and 10% for ReAsH ( Fig. 4 ), despite the fact that the conjugation of the autocleavage product with cholesterol adds less than 10% to its mass. A possible explanation for the unexpectedly high fluorescence anisotropy of the cholesterol esters of the cleavage products may be either the formation of aggregates of these lipopeptides or their association with Triton X-100, which is added to disperse the cholesterol substrate. Fortunately, despite the relatively small absolute change in FP, the fact that fluorescence polarization lends itself to homogeneous assays, which measure the ratio of fluorometric parameters rather than absolute fluorescence, allows for statistically robust assays even if the signal window is small. ¹⁷ In our case, the Z′ factors ⁸ were consistently >0.8 in our 384-well assays for both the FlAsH and ReAsH fluorophores.

When applied to the 446-compound NIH Clinical Collection using the FlAsH-based assay, our HTS behaved in a robust manner with only 2 positives, defined by more than 50% inhibition at 25 µM, which were confirmed by dose-dependent secondary assays, using the resorufin-based fluorophore, ReAsH, as an indicator to eliminate potential fluorescence artifacts in the primary screen ( Fig. 6 ). Both compounds inhibited cholesterol-dependent hedgehog autocleavage but not the hydroxylamine-dependent reaction, suggesting that their target might be cholesterol binding to the hedgehog autoprocessing domain. The inhibition by one of these compounds, CID 72303, was not time-dependent between 1 and 4 h, suggesting an irreversible interaction with the hedgehog protein, which was not further pursued. On the other hand, the other inhibitor identified in our screen, CID 5717, inhibited progressively less with increasing reaction time, suggesting a classical, reversible inhibition of hedgehog autoprocessing. CID 5717 is a well-characterized compound (zafirlukast), which has been shown to be a leukotriene receptor antagonist. ^18,19 As an antagonist of the binding of leukotrienes to their receptors, this compound could possibly also antagonize the binding of cholesterol to the sterol binding site of the hedgehog autoprocessing domain. Although the IC₅₀ value of 10 µM is too high to be useful as a pharmacological tool, CID 5717 can serve as a positive control in further high-throughput screens for hedgehog autoprocessing inhibitors that are now in progress.