Supplementary Information
Table of Contents
Compound Library Screened S4
Parasite and cell-based assay methods S6
Enzymatic and protein assays S10
Data processing and screening results S21
Multiple-lab cross-validation study S24
Chemical structure network graph S26
Additional mechanistic studies S28
Detailed analysis of three early lead compounds S31
Other high-value compounds (<200 nM) S35
Data availability S37
Supplementary references S38
A summary of the St. Jude screening workflow is reported in Supplementary Figure 3. Relevant screening parameters are defined in Table S4. A chemical library of 309474 compounds was screened in a phenotypic assay of P. falciparum. 1152 compounds with a single-point growth inhibition ≥80% in both SYBR Green and YOYO-3 assays were selected and further validated by establishing dose response relationships against 3D7 and K1. These 1152 compounds plus the true positives that resulted from the ROC analysis generated a total of 1300 compounds that had a saturable dose response curves. A set of 561 compounds was selected as high priority hits (EC50 < 2µM against 3D7 or K1). From this set, 264 were commercially available and were resourced and checked for identity and purity by our analytical chemistry group. 228 compounds passed QC (average purity greater than 85% by UV and ELSD) and were subsequently tested in further experiments (multiple lab cross- validation, phylochemogenetics, mechanism of action, and synergy).
Supplementary Figure 1. Screening flowchart.
Compound Library Screened
The current SJCRH CBT chemical library consists of roughly 515,000 unique molecules purchased from a variety of commercial sources. The library can be subdivided into 4 categories: approved drugs (~ 1,100 compounds); other known bioactives (~ 2,500 compounds); focused sets directed at defined targets including GPCR’s, kinases, proteases, and phosphatases (~ 45,000 compounds), and the diversity collection. For this project the bioactives and a portion of the diversity library were screened. The composition of the screening set, along with the primary screening data, and measured potencies of the compounds reported in this paper are available to download from http://www.stjuderesearch.org/guy/data/malaria/.
The bioactives collection consists of compounds from a wide variety of sources that include older drugs, veterinary drugs, enzyme inhibitors, tool compounds, and natural products and synthetic materials with public reports of activities on specified pathways or targets and has been assembled from diverse sources including Prestwick, Microsource, and Sigma-LOPAC
All samples in the CBT library are carefully chosen to provide a balanced, functionally diverse collection suitable for discovery of chemical matter active against a wide variety of targets and for phenotypic screening. In particular, the diversity subset has been designed using a maximally diverse cluster philosophy so that the population is made up of multiple clusters, each containing a series of related compounds, where the clusters are diverse with respect to one another. The general design paradigm for this collection is as follows:
First, commercially available compounds are filtered using a combination of physiochemical metrics, to improve probability of bioavailability, and functional group metrics, to reduce the probability of non-specific or artifact effects. The former is guided by our published work on the correlation of physiochemical parameters with bioactivity, as opposed to oral availability.1 The latter is guided by our implementation of the Vertex “Rapid Elimination of Swill” model,2 which utilizes a numeric scoring method with each functional group being assigned a score from -5 (always excluded) to 0 (never excluded) and allowing an aggregate score of -2 before elimination. In these filters, certain classes of molecule (cysteine protease inhibitors, for example) are handled specially.
Next, the filtered compound list is used to generate maximally diverse clusters. In order to do this, the compounds are reduced to core fragments (or “scaffolds”) using the method of Bemis and Murcko.3 We have found that Murcko fragmentation affords an intuitively useful level of abstraction, and have applied the technique in an analysis of several popular screening collections.4 We then prioritize compound clusters for purchase based on the balance of the diversity of the clusters from our existing library as assessed by Tanimoto similarity and the presence of a reasonable number of analogs within a cluster. We generally require 5 to 20 compounds per cluster and prefer clusters with more than 20 compounds, from which we would purchase a maximum of only 20 representative structures.
The method described above provides compounds that are most likely to work in cellular models for activity and that are least likely to carry with them mechanistically derived issues that make further development difficult. From that subset of molecules, we select chemotypes with maximal diversity, but populate each chemotype with a series of related compounds. This provides a library with a balance between the maximal likelihood of discovery of a novel activity and the maximal interpretability of a hit because of an enhanced likelihood of observing structure-activity relationships within the screening collection. We further enhance the ability to follow up hits by biasing towards chemotypes where a substantial number of analogs are available for purchase during hit-to-lead phases, thus allowing preliminary evaluation of the utility of a chemotype for a given system before committing to chemistry.
Parasite and cell-based assay methods
Data processing and visualization
All data processing and visualization, and chemical similarity and substructure analysis was performed using custom programs written in the Pipeline Pilot platform (Accelrys, v.7.0.1) and the R program.5,6 The R drc package was used to fit sigmoidal curves.6 ROC statistics were computed using the R rocr package.7
P. falciparum phenotypic screen assay (St. Jude, Pf3D7a)
Parasites were grown as previously described.8 For EC50 determinations, 20 µl of RPMI 1640 (L-glutamine and HEPES, Invitrogen) with final 5 µg/ml gentamicin were dispensed per well with a liquid dispenser (Matrix Wellmate, Thermo Scientific) in an assay plate (384-well black polystyrene microplate, clear bottom, tissue culture treated, Corning). DMSO inhibitor stock solutions were pin-transferred (V&P Scientific) into the assay plate and background fluorescence was measured in both relevant channels (485 nm / 535 nm, 612 nm / 665nm) in an Envision plate reader (Perkin Elmer). Then 20 µl of a synchronized culture suspension (1% rings, 10% hematocrit) were added per well thus making a final hematocrit and parasitemia of 5% and 1% respectively. Assay plates were incubated for 72 h and parasitemia was determined based on a modified protocol.9 Briefly, 5 μl of mixed DNA dye solution in RPMI (5X SYBR Green I, 1 μM YOYO-3, 1% v/v Triton X-100, 1 mg/ml saponin) were added per well, assay plates were shaken for 30 s at 2000 rpm, incubated in the dark for 4 h, then read as previously described.
Mammalian cell drug susceptibility assay
Raji, HEK293, Hep G2 and BJ cell lines were purchased from the American Type Culture Collection and were cultured according to recommendations. Exponentially growing cells respectively 1200, 400, 400, 1000 were plated per well (25 μl) in white polystyrene flat bottom sterile 384-well tissue culture treated plates (Corning), and incubated overnight at 37 °C in a humidified 5% CO2 incubator. DMSO inhibitor stock solutions were pin-transferred (V&P Scientific) the following day. Plates were placed back in the incubator for 72 h incubation and equilibrated at room temperature for 20 min before addition of 25 μl Cell Titer Glo (Promega) to each well. Plates were shaken on an orbital shaker for 2 min at 500 rpm. Luminescence was read after 15 min on an Envision plate reader (Perkin Elmer).
Trypanosoma brucei drug susceptibility assay
Culture adapted T. brucei brucei were grown at 37º C, 5% CO2 in HMI-9 medium (HyClone) supplemented with penicillin/streptomycin (50 μg/ml, Invitrogen), 10% heat inactivated FBS (Omega Scientific) and 10% Serum Plus (JHR Biosciences) to a density of 1x106 cells/mL, then diluted to 1 x 104 cells/mL 100 μL of the diluted culture was added to each well of a white polycarbonate flat bottom sterile 96-well tissue culture treated plate (Greiner), and 1 μl of inhibitor in DMSO was added by pin-transfer using 200 nL slotted hydrophobic coated pins (V&P Scientific). Plates were incubated for 48 h at 37ºC, 5% CO2, then equilibrated at room temperature for 30 min before addition of 50 μl Cell Titer Glo (Promega) to each well. Plates were shaken on an orbital shaker for 2 min at 500 rpm. Luminescence was read after 8 min on an Envision plate reader (Perkin Elmer).
Toxoplasma gondii drug susceptibility assay
10,000 U-2OS human host cells/well were grown for 24 h in 384-well plates, followed by introduction of test compound and 5000 freshly harvested RH strain T. gondii parasites expressing luciferase in a final volume of 40 µl.10 44 h post-infection, DMNPE-caged luciferin (Sigma-Aldrich, St. Louis, MO) was added to each well to a final concentration of 10 µM, and 4 h later luminescence was read using an Analyst HT (Molecular Devices, Sunnyvale, CA).
Leishmania major promastigote drug susceptibility assay
Leishmania major promastigotes were harvested in exponential growth phase (~2.0-3.0 X 107 parasites/ml) and were seeded (5,000 parasites/22 µl) into each well of a 384 well microtiter plate using a MAPC2 bulk dispenser (Titertek, Huntsville, AL). Test and control chemicals (3 µl) were initially assayed at a single concentration (10 µM) and were added to individual microtiter plate wells using a Velocity 11 V-prep (Menlo Park, CA) liquid handling system, equipped with a 384-well dispensing head, followed by centrifugation at 50 g for 1 min. Negative (vehicle) controls contained 1% DMSO and positive controls contained 10% DMSO (final well concentrations). Assay plates were incubated for 44 h at 28oC in the presence of 5% CO2. 5 µl of Alamar Blue reagent were added to each assay plate well and incubated for 4 h at 28oC with 5% CO2. Data were captured (excitation A560; emission A590) on a Molecular Devices SpectraMax M5 (Sunnyvale, CA). Individual assay plate Z-primes (for primary screening activities) were calculated from the vehicle and positive controls and data from plates were used only if calculated Z-prime was > 0.5. Primary actives were defined as chemicals displaying ≥50% inhibition of signal readout at 10 µM. For subsequent EC50 determinations, L. major promastigotes were treated with a 20 point concentration range of compound from 0 to 25 µM for 44 h at 28oC in the presence of 5% CO2. 5 µl Alamar blue was added to each well, plates were incubated for 4 h at 28oC with 5% CO2, and data were captured as described above.
P. falciparum drug susceptibility assay (Avery lab, Pf3D7b)
Compounds were transferred into 384 well poly-D-lysine coated wells (PerkinElmer) using a 384 well mini track dispenser. 25 μl of assay media was added per well followed by 20 μl of 2 or 3% synchronized ring stage parasite in 0.75% hematocrit (3D7 or Dd2). The final assay parameters were 50 μl total volume, 0.3% final hematocrit and DMSO concentration of 0.4%. The plates were incubated for 72 h in a humidified atmosphere at 37oC, 5% O2 and 5% CO2.
After incubation the plates were brought to room temperature and 30 μl of PBS added to all wells using a Biomek FX (Beckman Instruments). 60ml of supernatant was then aspirated from the wells and 30 μl of DAPI staining buffer (4’, 6-diamidino-2-phenylindole, (Invitrogen) in PBS containing 0.001% saponin and 0.01% Triton X-100) added. Plates were then incubated at RT in the dark for at least 5 h before imaging on the OPERA™ HTS confocal imaging system (Perkin Elmer).
The digital images obtained were then analyzed using the PerkinElmer Acapella spot detection software where spots that fulfill the criteria established for a stained parasite are determined and counted. The % inhibition of parasite replication was calculated using DMSO and 2 μM Artemisinin control data.
P. falciparum drug susceptibility assay (DeRisi lab, Pf3D7c)
Compounds were transferred using a Biomek FX robotic liquid handling device (Beckman Instruments) and incubated in presence of 2% hematocrit and 0.8% parasite in a total assay volume of 200 μl for 72 h in a humidified atmosphere at 37oC, 5% O2 and 5% CO2, in 96-well round bottom plates (Fisher). After incubation, 170 μl supernatant was discarded and cells were washed with 150 ml 1X PBS. 15 μl re-suspended cells were transferred to 384-well flat bottom non-sterile plates (Corning) already containing 15 ml of the SYBR Green lysis buffer (2X SYBR Green, 20mM Tris base pH 7.5, 20 mM EDTA, 0.008% w/v saponin, 0.08% w/v Triton X-100). Plates were incubated for 15 min and then read on a LJL Biosystems Analyst AD 96-384 spectrophotometer.
Enzymatic and protein assays
Heme crystallization assay
Drugs were assayed in 100 μl of sodium acetate pH 4.8 containing 50 μM mono oleoyl-glycerol (Sigma) and 50 μM bovine hemin chloride (Sigma). Plates were next shaken for 5 min on a bench top plate rocker and incubated at 37º C for 4-6 h. Plates were read in HTS 7000 Plus 96-well plate reader (PerkinElmer) after successive additions of 100 μl of a 2% sodium dodecyl sulfate (SDS), 200 mM sodium bicarbonate solutionand then 20 μl of 1 M NaOH. The OD405 value of each measurement was converted to nmoles of soluble heme using the formula: nmoles of heme crystal = OD405 x 20.338 -1.6369 / 5. The nmoles of heme crystal formed in the presence of drug were then divided by the number of nmoles heme crystal formed in the 3 averaged vehicle control wells to determine a relative % inhibition for each drug. Each drug was performed in triplicate with experiments repeated at least twice.
PfDHOD inhibition assay
Compounds were initially assayed in 384 well plate format using a Biomek FX robotic liquid handling device (Beckman Instruments). The cross-validated set was screened for P. falciparum dihydroorotate dehydrogenase (PfDHODH) inhibitory activity at 0.1, 1.0 and 10 µM compound concentrations. Assay buffer solution (0.048 ml) (100 mM HEPES pH 8.0, 150 mM NaCl, 10% Glycerol, 0.05%Triton X-100, 20 μM CoQ0, 200 μM L-dihydroorotate, 120 μM 2,6- dichloroindophenol) was transferred to each well, followed by addition of the stock solution (0.5 μl of 100X DMSO stock) for each compound. The assay was started by the addition of a 25X (2 μl) stock solution of enzyme (50 nM final in assay plate). The assay was allowed to progress at room temperature for 20 min then 10% sodium dodecyl sulfate (5 μl) was added to stop the reaction. The plate was spun at 3000 rpm at room temperature for 1 min and the absorbance in each well was measured at 595 nm using an Envision plate reader (Perkin Elmer). Hits were re-assayed to determine the EC50 using a previously described protocol.11 This assay was performed in a Beckman DU650 spectrophotometer using the same buffers and reagents (final volume 0.5 ml) described for the plate assay except that the enzyme concentration was 10 nM, the triton concentration was increased to 0.1%, and a rate assay was used in place of the endpoint assay.