High-Throughput Fluorescence Polarization Assay for Chemical Library Screening against Anti-Apoptotic Bcl-2 Family Member Bfl-1

Protein Purification

His₆-“truncated” Bid (t-Bid) was prepared as previously reported. ²⁴ Glutathione S-transferase (GST)–tagged proteins containing Bcl-X_L, Bcl-2, Bcl-W, Bcl-B, Bfl-1, and Mcl-1 lacking their C-terminal transmembrane domains (~ last 20 amino acids) (“ΔTM”) were expressed from the pGEX 4T-1 plasmid in XL-1 Blue cells (Stratagene, La Jolla, CA) as previously reported. ²⁵ Briefly, cells were grown in 2 L of LB with 50 µg/mL ampicillin at 37 °C to an optical density (OD) 600 nm of 1.0, then isopropyl-β-D-thiogalactopyranoside (IPTG; 0.5 mM) was added, and the cultures were incubated at 25 °C for 6 h. Cells were then recovered in 20 mM phosphate buffer (pH 7.4), 150 mM NaCl, 1 mM dithiothreitol (DTT), 1 mM EDTA, and 1 mM phenylmethylsulfonyl fluoride (PMSF), followed by sonication. Cellular debris was sedimented by centrifugation at 25 000 g for 20 min, and the resulting supernatants were incubated with 10 mL of glutathionine-Sepharose (Amersham Biosciences, Piscataway, NJ) at 4 °C for 2 h. The resin was washed three times with 20 mM phosphate buffer (pH 7.4), 150 mM NaCl, and 1 mM DTT, and then 10 mM of reduced glutathione dissolved in 50 mM Tris-HCl (pH 8.0) was used to elute the GST fusion proteins.

Fluorescence Polarization Assays

FPAs were performed as described previously using various Bcl-2 family proteins and fluorescein isothiocyanate (FITC)–conjugated Bid BH3 peptide. ²⁵ Briefly, Bfl-1 proteins were incubated in 384-well black round-bottom plates with 5 or 2.2 nM FITC–Bid BH3 peptide (FITC-Ahx-EDIIRNIARHLAQVGDSMDR) in a total volume of 20 µL in the dark. Fluorescence polarization was measured using a LJL Analyst HT plate reader (Molecular Devices Corp., Sunnyvale, CA) in phosphate-buffered saline (PBS) or 25 mM Bis-Tris buffer containing 1 mM of TECP, 0.005% Tween-20 (pH 7.0). IC₅₀ determinations were performed using a sigmoidal dose–response nonlinear regression model (Y = Bottom + (Top – Bottom)/1 + 10^{(LogEC50 – X)*Hill slope}) with GraphPad Prism software (GraphPad, Inc., San Diego, CA).

High-Throughput Screening

Approximately 66 000 compounds (10 mM stock solution in DMSO) from the National Institutes of Health (NIH) Small Molecule library were plated into black, 384-well, round-bottom plates (Greiner Bio-One, Frickenhausen, Germany) using a BioMek FX Laboratory Automation Workstation (Beckman Coulter, Fullerton, CA). The ThermoScientific Matrix WellMate bulk dispenser (Thermo Fisher Scientific, Hudson, NH) was then used to add GST–Bfl-1, followed by FITC–Bid BH3 peptide, both diluted in 25 mM Bis-Tris, 1 mM TCEP, 0.005% Tween-20 (pH 7.0). The first column of the screening plate was positive controls (16 wells), and the second column of the plate was negative controls (16 wells). The remaining wells contained ~20 µM compound, 3 nM GST–Bfl-1, and 2.2 nM FITC–Bid BH3 peptide in a total volume of 20 µL. After a 4-h incubation at room temperature in the dark, fluorescence polarization was measured using an Analyst HT Multi-Mode Plate Reader (Molecular Devices), with an excitation filter at 485 nM and an emission filter at 530 nM. Data analysis was performed using CBIS software (ChemInnovation Software Inc, San Diego, CA).

Competitive Peptide Displacement Assays for Assessing Compound Selectivity

Methods for competitive peptide displacement assays were similar to previous publications. ²⁵ Briefly, empirically optimized amounts of various GST fusion proteins (3 nM for Bfl-1; 20 nM for Bcl-2, Bcl-X_L, Bcl-W, and Bcl-B; and 10 nM for Mcl-1) were incubated with the compounds at various concentrations for 5 min at room temperature in 25 mM Bis-Tris, 1 mM TCEP, 0.005% Tween-20 (pH 7.0). Then, 2.2 nM FITC-Bid (for Bfl-1), 10 nM FITC-Bim (for Bcl-2, Bcl-X_L, Bcl-W, and Bcl-B), or 5 nM FITC-Bak (for Mcl-1) BH3 peptide was added, resulting in a total volume of 20 µL per well. DMSO concentration (resulting from compound addition) did not exceed 0.2% (vol:vol). Fluorescence polarization was measured after 20 min. IC₅₀ determinations were generated by fitting the experimental data using a sigmoidal dose–response nonlinear regression model with GraphPad Prism software (GraphPad, Inc.).

Time-Resolved Fluorescence Resonance Energy Transfer Assays

For TR-FRET assays, GST–Bfl-1 and anti-GST-terbium (Invitrogen, Carlsbad, CA) were mixed together with the FITC–Bid BH3 peptide in 20 mM Bis-Tris containing 0.005% Tween-20 in 384-well plates in a total volume of 20 µL per well. After incubation at room temperature for 10 min, 2 µL of compound-containing solutions was added to the reaction mixtures containing 3 nM Bfl-1, 2.2 nM FITC–Bid BH3 peptide, and 2 nM anti-GST-terbium for 30 min at room temperature. TR-FRET signals were measured with a SpectraMax M5 plate reader (Molecular Devices) using the following settings: excitation at 330 nm, emission for FITC signal at 490 nm, and emission for terbium signal at 520 nm. IC₅₀ determinations were generated by fitting the experimental data using a sigmoidal dose–response nonlinear regression model with GraphPad Prism software (GraphPad, Inc.).

Mitochondria Purification and Protein Release Assays

HeLa cells were pelleted by centrifugation and then washed once in HM buffer (10 mM HEPES [pH 7.4], 250 mM mannitol, 10 mM KCl, 5 mM MgCl₂, 1 mM EGTA), containing 1 mM PMSF and a mixture of protease inhibitors (Roche Molecular Biochemicals, Indianapolis, IN). The cell pellet was then homogenized in HM buffer by 60 strokes of a dounce homogenizer, using a B-type pestle. The homogenate was centrifuged twice at 600 g for 5 min to remove nuclei and debris. The resulting supernatant was centrifuged at 10 000 g for 10 min, and the resulting mitochondria-containing pellet was washed twice with HM buffer.

For mitochondrial protein release assays, 10 µL of mitochondria (50 µg) was added to a final volume of 50 µL HM buffer containing compounds, with or without t-Bid or Bcl-2 family proteins, at 30 °C for 15 min. The reactions were further incubated at 30 °C for 40 to 60 min, and then mitochondria were pelleted by centrifugation and the supernatants were collected, boiled in Laemmli sample buffer, and analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE)/immunoblotting using anti-SMAC antibody. ²⁶

Compound Synthesis

Development of Bfl-1 Fluorescence Polarization Assays

We previously reported a nonoptimized FPA for Bfl-1 that proved suitable for characterizing known inhibitors of Bcl-2 family proteins. ²⁵ This FPA that measures binding of GST–Bfl-1 protein to FITC-conjugated Bid BH3 peptide in PBS without detergents was further optimized for HTS applications. To this end, various buffer, salt concentrations, and detergent conditions were explored. Using a modified Bis-Tris buffer, we achieved a 30% increase in the assay window (signal-to-noise ratio), 3-fold relative to the K_d, allowing a 33-fold reduction of protein concentration and an improvement in the signal stability from 15 min to 24 h compared to PBS buffer without detergent ( Fig. 1A , 1B , data not shown). In the optimized FPAs, the concentration of Bfl-1 is approximately equivalent to the dissociation constant (K_d = 9.2 ± 0.2 nM) of its complex with the fluorescent BH3 peptide ligand. Similarly, a TR-FRET assay was optimized using the same modified Bis-Tris buffer ( Fig. 1C ; K_d = 2.9 ± 0.3 nM). In addition to decreasing the assay protein requirements, the optimized conditions also improved assay sensitivity. We note that our experience has been that each batch of recombinant Bfl-1 protein should be tested and protein concentration adjusted as shown in Figure 1A .

OPEN IN VIEWER

Figure 1.

Bfl-1 binds to labeled Bid BH3 peptide. (A) Various concentrations of recombinant glutathione S-transferase (GST)–Bfl-1 (black circles) or GST control (white circles) proteins were mixed with 5 nM of FITC–Bid BH3 peptide in 20 mM Bis-Tris (pH 7.0), 1 mM TCEP, 0.005% Tween-20. Fluorescence polarization was measured after 10 min. Data represent mean of three determinations (standard deviations were <10% for all values). (B) Fluorescence polarization assay (FPA) was conducted as above using 5 nM FITC–Bid BH3 peptide and various concentrations of GST–Bfl-1 protein for various times (ranging from 10 min to 6 h). Data represent mean of three determinations (standard deviations were <10% for all values). (C) Various concentrations of GST–Bfl-1 proteins were mixed with 2.2 nM FITC–Bid BH3 peptide in 20 mM Bis-Tris (pH 7.0), 1 mM TCEP, 0.005% Tween-20, and then 2 nM of anti-GST-terbium was added. Time-resolved fluorescence resonance energy transfer (TR-FRET) signals were measured after a 30-min incubation at room temperature. Data represent mean of three determinations (standard deviations were <10% for all values).

Another issue observed during assay optimization was a time-dependent Bfl-1 binding to the fluorescent probe. To compensate for this phenomenon, we reformatted the assay to include a 1-h incubation of Bfl-1 in the presence of compounds and a 4-h incubation after the subsequent addition of the fluorescent probe (FITC–BH3 peptide).

Previously, we observed that the binding characteristics of the GST fusion Bfl-1 protein is similar to untagged Bfl-1 in BH3 peptide binding assays. ²⁶ Thus, for convenience, we choose GST fusion Bfl-1 protein for conducting a proof-of-concept HTS campaign.

High-Throughput Screening

The performance of the optimized assay was evaluated by performing multiple replicate tests in a 384-well format using either FITC–Bid BH3 peptide alone (assay minimum) or with Bfl-1 protein (assay maximum). The assay demonstrated robust performance, with a Z′ factor of 0.83 ( Fig. 2A ). Moreover, the Z′ factor remained >0.7 after overnight incubation (data not shown).

OPEN IN VIEWER

Figure 2.

Compound library screening. (A) Glutathione S-transferase (GST)–Bfl-1 (10 nM) was incubated with 5 nM FITC–Bid BH3 peptide without (black squares) or with 50 nM of unlabeled Bid BH3 peptide (black circles). Alternatively, FITC–Bid peptide was added to wells without Bfl-1 protein (first column, 16 wells, white circles). Fluorescence polarization was measured after 10 min. Data represent fluorescence polarization (mP) measured in individual wells, constituting 176 replicates for each condition in a 384-well format. The Z′ score was calculated. (B) For chemical library screening, GST–Bfl-1 (3 nM) was incubated with FITC–Bid BH3 peptide (2.2 nM) with DMSO alone or with 20 µM of the National Institutes of Health (NIH) library compounds (blue). Fluorescence polarization was measured as previously described. Blue bars represent fluorescence polarization assay (FPA) results for compound library screening, indicating the number of compounds (y-axis) that displayed a specific percentage inhibition (x-axis). Red bars indicate control wells included in all plates, which comprised either FITC–BH3 Bid peptide alone (right) or FITC–Bid BH3 peptide plus Bfl-1 protein (left). The box shows the result for “hits,” where ≥50% inhibition of FPA signal is considered positive.

The assay was screened against the entire NIH small-molecule library, which at the time consisted of ~66,000 compounds, searching for chemical inhibitors that displaced FITC–Bid BH3 peptide from GST–Bfl-1 protein, resulting in 44 primary positive results, with an average Z′ factor of ~0.75 ± 0.04 ( Fig. 2B ). “Hits” were defined as compounds that induced a ≥50% decrease in fluorescence polarization without increasing fluorescence intensity by ≥25% (thus eliminating compounds that interfered with the assay due to their autofluorescence). After correcting for fluorescence interferences, the primary screen then yielded 20 hits. These hits were cherry picked from assay plates and retested, resulting in 14 confirmed hits, representing a hit rate of 0.021%.

A TR-FRET secondary assay used the same molecular interaction between the FITC–Bid BH3 peptide and the GST–Bfl-1 protein, except terbium-conjugated anti-GST served as the fluorescent donor for exciting the FITC BH3 peptide. Of 20 compounds active in the primary FPA-based screen, 14 demonstrated inhibitory activity in the TR-FRET assay. These were the same 14 compounds that showed reproducible activity in the FPA (above).

Compound dose–response curves were generated using both fluorescence polarization (FP) and TR-FRET assays for the 14 validated hits. Compounds that did not display a sigmoidal dose–response curve or that increased fluorescence intensity (some compounds increased fluorescence intensity at concentrations higher than the screening concentration) were eliminated. Finally, seven compounds appeared to be promising based on dose–response results from both the FP and TR-FRET assays (IC₅₀ values summarized in Table 1 ). The potency of these screening hits varied, with IC₅₀ values as low as <1 µM to as high as approximately 30 µM, depending on whether measured by FPA or TR-FRET.

Characterization of Compound Selectivity for Bfl-1

The seven validated hits from the screening could be divided into three groups based on structural features of the compounds: #1 (MLS-0053105 = CID-1180676) is a chloromaleimide, #2 to #5 (#2, MLS-0047123 = CID-2980973; #3, MLS-0009480 = CID-1151859; #4, MLS-0051609 = CID-3244140; #5, MLS-0051509 = CID-3237441) comprise a sulfonylpyrimidine series; and #6 and #7 MLS-0019296 (CID-5389865) and MLS-0025736 (CID-8423532) are quinolinyl-oxadiazoles. The reactivity of these compounds with other targets previously screened and deposited into the PubChem database at the time of our discovery revealed that the chloromaleimide showed reactivity in only 1 of 30 assays tested, the sulfonylpyrimidines in 3 of 60 assays, and the quinolinyl-oxadiazoles in none of 29 assays tested. Thus, these three chemical series do not show promiscuous activity with other classes of targets.

To test compound selectivity against all six human anti-apoptotic Bcl-2 family protein members, we established FPAs that monitor the binding of various fluorochrome-conjugated BH3 peptides to the recombinant GST fusion proteins. BH3 peptides were selected based on our prior comparison of binding affinities, ²⁷ resulting in pairings for the binding assays as follows: Bfl-1/FITC–Bid BH3 peptide, Bcl-W/FITC–Bim BH3 peptide, Bcl-2/FITC–Bim BH3 peptide, Bcl-X_L/FITC–Bim BH3 peptide, and Mcl-1/FITC–Bak BH3 peptide. Table 2 summarizes the IC₅₀ values of seven validated compounds for each of the six anti-apoptotic Bcl-2 family members. The chloromaleimide compounds showed submicromolar inhibition of Bfl-1, with over 10 times less potent inhibition of Bcl-W, Bcl-2, and Bcl-X_L and no activity against Bcl-B and Mcl-1, indicating good selectivity for Bfl-1. The sulfonylpyrimidine compounds showed superb selectivity against Bfl-1, with no inhibition of other Bcl-2 family members at concentrations as high as 100 µM. One of the quinolinyl-oxadiazole series compounds showed cross-reactivity with Bcl-W, but otherwise this series was also Bfl-1 selective ( Table 2 ).

OPEN IN VIEWER

Table 2.

Selectivity of Bfl-1 Compounds against Six Human Anti-Apoptotic Bcl-2 Family Proteins

Compound #	Bfl-1/F-Bid IC50 (µM)	Bcl-W/F-Bim IC50 (µM)	Bcl-2/F-Bim IC50 (µM)	Bcl-XL/F-Bim IC50 (µM)	Bcl-B/F-Bim IC50 (µM)	Mcl-1/F-Bak IC50 (µM)
1 = MLS-0053105	0.4	12.5	12.5	50	>100	>100
2 = MLS-0047123	1.5	>100	>100	>100	>100	>100
3 = MLS-0009480	2.0	>100	>100	>100	>100	>100
4 = MLS-0051609	16.7	>100	>100	>100	>100	>100
5 = MLS-0051509	4.0	>100	>100	>100	>100	>100
6 = MLS-0019296	23.9	12.5	>100	>100	>100	>100
7 = MLS-0025736	4.7	>100	>100	>100	>100	>100

Various concentrations of compounds were mixed with anti-apoptotic Bcl-2 family proteins (Bfl-1, Bcl-W, Bcl-2, Bcl-XL, Bcl-B, Mcl-1). FITC-BH3 peptides (Bid, Bim, or Bak as indicated) were then added and fluorescence polarization was measured after 30 min. IC₅₀ determinations were generated by fitting the experimental data using a sigmoidal dose–response. Data are the mean of four determinations (standard derivations ≤15%).

Bioactivity of Compounds Tested Using Mitochondrial Assay

Isolated mitochondria were used to test the bioactivity of compounds under circumstances where recombinant Bfl-1 protein nullifies mitochondrial release of proteins induced by t-Bid protein, similar to our previous report where we analyzed BH3 peptide activity. ²⁶ Interestingly, sulfonylpyrimidine CID-1151859 caused mitochondria release of SMAC protein in a concentration-dependent manner ( Fig. 3A ), perhaps reflecting activity against endogenous proteins associated with the mitochondria. Recombinant Bfl-1 protein, but not Bcl-2 or Bcl-X_L, blocked the release of SMAC protein induced by this compound, consistent with our data indicating selective binding to Bfl-1 ( Fig. 3B , 3C ). In contrast, chloromaleimide compound CID-1180676 did not cause SMAC release from mitochondria ( Fig. 3D ). We then employed recombinant Bid (t-Bid) protein as an inducer of SMAC release from mitochondria and tested the effects of compounds with respect to neutralization by recombinant anti-apoptotic Bcl-2 proteins Bfl-1 and Bcl-X_L. The chloromaleimide compound CID-1180676 negated Bfl-1 protection against t-Bid-induced release of SMAC from isolated mitochondria, whereas Bcl-X_L did not, suggesting selectivity for Bfl-1 ( Fig. 3E ).

OPEN IN VIEWER

Figure 3.

Analysis of Bfl-1-binding compounds using a mitochondria-based functional assay. Various concentrations of compound CID-1151859 or CID-1180676 (µM) were preincubated with specific concentrations of purified recombinant t-Bid (20 nM in A–D) (0 vs 20 µM in E), without (A, D) or with (B, C, E) various concentrations of anti-apoptotic Bcl-2 family proteins, including Bfl-1 at 0-20 µM (B), Bcl-X_L at 0-20 µM (C), Bcl-2 at 0-20 µM (C), or 0-1 µM Bfl-1 or 0-1 µM Bcl-X_L (E) in HM buffer, before adding 50 µg of isolated HeLa mitochondria for 1 h at 30 °C. Samples were centrifuged to generate supernatants that were analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis/immunoblotting using the anti-SMAC antibody. The “Pellet” (P) represents intact, untreated mitochondria, run in gels as a control. Data are representative of at least three experiments.

Structure–Activity Relation Analysis of Chemical Inhibitors of Bfl-1

During the course of this project, more than 180 analogs of the chloromaleimide and sulfonylpyrimidine class compounds were purchased or synthesized, comparing their activity against Bfl-1 using both the FPA and a TR-FRET assay. The structure–activity relation (SAR) analysis of the chloromaleimide series was published recently. ²³ Here we describe a brief SAR analysis of the sulfonylpyrimidines (n = 11 compounds tested). Table 3 summarizes the IC₅₀ values of the compounds. Altogether, the activity of the compounds tested against Bfl-1 varied by as much as approximately one log, ranging from 2.6 µM to 25.1 µM in IC₅₀ values.

The synthetic chemistry strategy employed to follow up on hit compound CID-1151859 is shown in scheme 1 ( Fig. 4 ). Specifically, 1,3-diketones were reacted with S-substituted carbamimidothioates to afford the corresponding S-alkylpyrimidines. Oxidation of these products at 25 °C with mCPBA furnishes the desired sulfonylpyrimidine analogs. We synthesized and tested a series of compounds, and in vitro data were generated to provide SAR data on the hit series. Initial studies focused on analogs in which the phenyl ring and/or the sulfone substituent were modified, and some of these compounds are shown in Table 3 . For example, a 3-methoxy substituent is tolerated (MLS-0066988 = CID-751044), whereas a 3,4-dimethoxy derivative (MLS-0051609 = CID-3244140) is less potent than the parent molecule. It also appears that replacement of the phenyl ring with a heteroaryl moiety such as furan (MLS-0050257 = CID-3237865) or thiophene (MLS-0041503 = CID-655335 and MLS-0051774 = CID-3245258) is viable. Truncating the sulfonyl side chain to a methylsulfone moiety as in MLS-0066988 (CID-751044), MLS-0050257 (CID-3237865), MLS-0051609 (CID-3244140), and MLS-0051774 (CID-3245258) appears to lead to a modest reduction in potency compared with the lead compound. Homologation of the carbethoxymethylene side chain to the carbethoxyethylene variant (MLS-0046458 = CID-2219262) leads to a slight reduction in potency. More pronounced increases in side chain length, as in compounds MLS-0041503 (CID-655335), MLS-0039077 (CID-2037235), MLS-0050749 (CID-3238559), and MLS-0008909 (CID-654688) leads to further reduction in potency. Ultimately, we selected the 4-fluorophenyl derivative CID-2980973, which is twice as potent as the best hit compound CID-1151859 as the Bfl-1 sulfonylpyrimidine series chemical probe, which has received the designation (http://mli.nih.gov/mli/mlp-probes/?dl_id=1289, ML042 = CID-2980973) from the NIH Molecular Libraries Initiative.

OPEN IN VIEWER

Figure 4.

Synthetic scheme for production of hit compound CID-1151859. (a) 4,4,4-trifluoro-1-phenylbutane-1,3-dione and ethyl 2-(carbamimidoylthio)acetate (1.2 eq.) were heated in ethanol at 100 °C for 5 h to yield ethyl 2-((4-phenyl-6-(trifluoromethyl)pyrimidin-2-yl)thio)acetate in 68% yield. (b) To this product, 2 eq of mCPBA in CH₂Cl₂ was added, and the mixture was stirred at room temperature for 20 h. After evaporation of solvent and reverse-phase preparative high-performance liquid chromatography, CID-1151859 was isolated in 82% yield.