Molecular docking
The Center for Structure-based Drug Design and Development (CSD3) chemical library, consisting of 11,500 drug-like chemicals, was prepared in electronic format as two-dimensional (2D) SDF files. Using Pipeline Pilot [19], the protonation state of all compounds was adjusted to reflect the most prevalent form at a pH of 7.4. CORINA [20] was used to convert these files to three-dimensional (3D) PDB coordinate files, which resulted in energy-minimized 3D structures. The files were then processed with the python script prepare_ligand4.py, which comes with the Autodock Tools Suite [21]. This script generates a pdbqt file and adds partial charges to the ligand, sets all torsions in the ligand to active (to permit rotation), and merges all non-polar hydrogen atoms.
The DUSP5 PD structure (PDB:2G6Z) [16] was prepared for docking using the Autodock Tools Suite [21]. Grid maps were used in the energy calculations performed by Autodock. Partial charges were added and all non-polar hydrogen atoms were merged, resulting in a pdbqt file. The 13 different grid maps, one for each of the different atoms present in the chemical library of compounds (ex. C, H, F, Cl, etc.), were generated using Autogrid4 [21]. A grid box, the site used to dock the ligands, was positioned to cover the entire protein in a blind docking experiment to ensure unbiased identification of binding location and orientation.
The docking parameter file (dpf), which contains the parameters that Autodock4 uses to dock ligands into the protein, was prepared using the python script prepare_dpf4.py, and default docking parameters were used, except that 50 separate docking calculations were performed with each calculation consisting of 1,750,000 energy evaluations, and a root mean square deviation (rmsd) tolerance set to 2.0 angstroms (to define entry of structure into a given cluster). The dpf files were then automatically docked using the MUGrid Cluster (Marquette University) with HTCondor [22, 23] and AutoDock4 [21, 24] using the Lamarckian genetic algorithm local search method to perform the optimization of docking poses. The docking poses were then clustered on the basis of the rmsd between the coordinates of the atoms in a given ligand, and were ranked on the basis of calculated free energy of binding. The docking log files were then analyzed using the python script summarize_results4.py contained in the shell script sumresults_4.py [21], which rank orders all the dockings by binding energy. The results were then analyzed to find the best-clustered compounds with lowest free energy of binding as determined by Autodock4.2. Additional docking of all experimentally tested chemicals was performed as described above, but with 100 dockings trials.
Ligand-based searching
As previously described, the CSD3 chemical library was electronically prepared and protonation state adjusted using Pipeline Pilot [19]. Using OpenEye Scientific Software’s Omega2 [25, 26], three dimensional coordinates were calculated and stored in OpenEye Scientific Software’s preferred file format, .oeb.gz, for subsequent molecular overlay evaluation. OpenEye Scientific Software’s Rapid Overlay of Chemical Structures (ROCS) [25] software was used to search for molecules with similar shape and electronic properties to a lead molecule. Lead molecules identified from DUSP5 docking and inhibition studies were used as chemical queries to search a database of Food and Drug Administration (FDA) approved drugs, to identify FDA approved drugs that might also be DUSP5 inhibitors.
The ZINC library [27] of 13 million commercially available chemicals was obtained as 2D SDF files and prepared similarly to the CSD3 chemical library for use with OpenEye Scientific Software’s ROCS. This further expanded the availability of chemical analogs available for experimental screening. ROCS calculations were also performed against the DrugBank [28, 29] database of FDA approved drugs.
Synthesis of RR505 and RR506. Synthesis of Carbazole-1,3,6-trisulfonic acid, trisodium salt (RR505)

Solid carbazole (3.0 g, 17.9 mmol) was placed in a 50-mL round-bottom flask and 67 % H2SO4 (12 mL) was added drop-wise at 22 °C and a slurry thus obtained was stirred and heated at 115 ± 5 °C for 6 h. The resulting dark solution was cooled to room temperature and poured into a saturated NaCl solution (100 mL) containing NaOH (2.4 g, 60 mmol) to afford an ash-colored precipitate, which was filtered, washed with saturated NaCl solution (50 mL) and dried at 90 °C for 10 h to get 7.5 g of the crude product.
The crude solid was dissolved in distilled/deionized water (150 mL), treated with activated charcoal Norit (1.1 g) and the resulting mixture was refluxed for 15 min. The solution was filtered hot through a pad of Celite®, and evaporated slowly to afford a white powder of RR505 (5.5 g, 65 % yield). 1H-NMR (400 MHz, D2O): 7.60 (1H, d, J = 8.8 Hz), 7.79 (1H, d, J = 8.8 Hz), 8.08 (1H, s), 8.53 (1H, s), 8.64 (1H, s), 1H-NMR (400 MHz, DMSO-d
6
): 7.58-7.66 (2H, m), 7.96 (1H, s), 8.24 (1H, s), 8.26 (1H, s), 10.81 (1H, s). 13C-NMR (100 MHz, D2O): 112.3, 118.8, 121.6, 121.64, 121.7, 124.5, 124.7, 125.5, 133.7, 134.7, 136.9, 142.1.
Synthesis of Carbazole-1,3,6,8-tetrasulfonic acid, tetrasodium salt (RR506)

Solid carbazole (3.0 g, 17.9 mmol) was placed in a 50-mL round-bottom flask and chlorosulfonic acid (41.7 g, 358 mmol) was added in small portions with vigorous shaking at 22 °C, after which the mixture was stirred and heated at 100 ± 5 °C for 1 h. The resulting dark solution was cooled to room temperature and then poured slowly onto crushed ice (~100 g). The resulting precipitate was filtered by gravity filtration and was dried by placing between paper towels. The resulting semi-dried solid was dissolved in ethyl acetate (150 mL), treated with Norit (1.5 g), and refluxed for 15 min and filtered hot through a pad of silica gel (~1x1.8 inch). The filtrate was concentrated in vacuo and recrystallized from a 1:9 mixture of ethyl acetate and hexanes to afford a yellow solid, which was filtered and dried in vacuo.
The dried solid was dissolved in a mixture of dioxane (20 mL) and distilled/deionized water (20 mL) and heated under reflux for 12 h. The resulting solution was cooled to room temperature and was extracted with diethyl ether (2 x 50 mL) to remove nonpolar impurities. The aqueous layer was neutralized by a dropwise addition of NaOH solution (1 M) with continuous monitoring of pH using pH paper. The resulting solution was concentrated to ~10 mL and acetone was added to afford a white powder of RR506 (3.5 g, 22 % yield, average yield from 3 runs). 1H-NMR (400 MHz, D2O): 8.12 (2H, s), 8.68 (2H, s), 1H-NMR (400 MHz, DMSO-d
6
): 7.97 (2H, s), 8.26 (2H, s), 10.61 (1H, s); 13C-NMR (100 MHz, D2O): 121.9, 122.1, 123.8, 125.7, 134.7, 136.7.
Alternative synthesis of RR506
Solid carbazole (1.0 g, 6 mmol) and nitrobenzene (20 mL) were placed in a 50-mL round-bottom flask and chlorosulfonic acid (14 g, 120 mmol) was added in small portions at 22 °C, after which the mixture was stirred at 22 °C for 72 h. The resulting solution was poured into aqueous saturated NaCl solution (100 mL) containing NaOH (0.96 g, 24 mmol) which resulted in a fluffy precipitate. The precipitate thus formed was filtered and dried. The solid was dissolved in distilled/deionized water (100 mL) and refluxed with 1.5 g of Norit for 15 min and filtered hot through a pad of celite. The filtrate was concentrated to ~25 mL and RR506 was precipitated by addition of acetone. The precipitate was filtered and dried to afford RR506 as a white solid (2.9 g, 84 % yield). 1H-NMR (400 MHz, D2O): 8.12 (2H, s), 8.68 (2H, s), 1H-NMR (400 MHz, DMSO-d
6
): 7.97 (2H, s), 8.26 (2H, s), 10.60 (1H, s).
Protein production
The DUSP5 PD gene was synthesized by Blue Heron (Bothell, WA) in both an active wild type form (DUSP5 PD(WT)) and an inactive form, where the catalytic cysteine was mutated to a serine (DUSP5 PD(C263S)). The genes were inserted into Origene pEX plasmids with ampicillin resistance and an N-terminal hexa-histidine tag to facilitate protein purification. Plasmids were transformed into BL21(DE3) cells (Invitrogen) for expression.
For unlabeled DUSP5 PD(WT) preparation, an overnight culture was used to inoculate 2 L of LB (Luria-Bertani) media, containing 50 μg/mL of ampicillin. Cells were grown at 37 °C to an OD600 of 0.7 and then induced with 0.6 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 4 h at 37 °C, then for 14 h at 16 °C. Cells were harvested using centrifugation and frozen prior to purification. Thawed cells were lysed in a buffer containing 50 mM Tris, 300 mM NaCl, 5 mM imidazole, and 10 % glycerol at pH 7.8. Lysate was centrifuged at 15,000 rpm for 1 h. The supernatant was loaded on to Ni-Sepharose Fast Flow resin (GE Healthcare) and washed three times successively with five column volumes of lysis buffer containing 25 mM imidazole. Protein was eluted with lysis buffer containing 305 mM imidazole. Protein was then dialyzed in a buffer containing 50 mM potassium phosphate and 2 mM dithiothreitol (DTT) at pH 6.8.
For 15 N-labeled DUSP5 PD(C263S) preparation (for NMR titrations), an overnight culture was used to inoculate 2 L of LB media supplemented with 50 μg/mL of ampicillin. Cells were grown to an OD600 of 0.7 at 37 °C, then harvested and washed with M9 minimal media (pH 7.0) [26]. Cells were resuspended in 500 mL M9 minimal media containing 0.5 g 15NH4Cl, 2 g D-glucose, 5 mL Basal Medium Eagle with Earle’s salts and sodium bicarbonate (Sigma Aldrich), 0.146 g L-glutamine (Sigma Aldrich) 1.0 mL 1 M MgSO4, and 0.5 mL 1 M CaCl2 (pH 7.2) [30]. Additionally, a metal mix containing Zn, Mn, Cu, Co, B, and Mo salts was added to supply cells with necessary micronutrients [30]. Cells were allowed to acclimate for 30 min at 37 °C, then induced with 1 mM IPTG for an additional 4 h at 37 °C. Cells were harvested and 15 N-labeled protein was purified as described before with the addition of 2 mM DTT during all purifications steps.
p-nitrophenol phosphate (pNPP) activity assay
To measure enzymatic activity of the DUSP5 PD and the inhibitory capacity of selected molecules, an in vitro phosphatase assay was developed based on previous studies [31]. In this assay, DUSP5 PD will dephosphorylate the substrate p-nitrophenol phosphate (pNPP, Sigma Aldrich), yielding p-nitrophenolate, which absorbs at 405 nm with an extinction coefficient of 18,000 M−1 cm−1.

Thus, an increase in absorbance at 405 nm corresponds to the turnover of pNPP to p-nitrophenolate. The assay was initially optimized in 1 mL quartz cuvettes, then was subsequently optimized for and validated in a 96-well plate format. All IC50 values were obtained using the 96-well plate assay format (see below). The assay buffer contained 100 mM Tris, 100 mM sodium chloride, 5 mM magnesium chloride, and 1 mM DTT at pH 7.5. The pNPP substrate was prepared as a 50 mM stock by dissolving the solid substrate in assay buffer. The DUSP5 PD and pNPP were assayed initially in a cuvette (1 mL total volume) and initial velocities were fitted to the Michaelis-Menten equation:
$$ v=\frac{V_{\max}\left[S\right]}{K_m+\left[S\right]} $$
(1)
where v is the initial velocity, V
max is the maximum velocity, K
m is the Michaelis constant, and [S] is the concentration of pNPP. Data were fitted using a nonlinear least squares fit to eq. 1, with GraphPad Prism 6 software.
Validation of pNPP assay for high throughput screening (HTS)
For the 96-well plate validation assay, sodium orthovanadate (Sigma Aldrich) was utilized as a positive control for inhibition [32] at a final concentration of 10 μM, to completely block DUSP5(WT) enzymatic activity. All plate assays were performed in standard 96-well clear bottom plates (Thermo Scientific Nunc) with a total assay volume of 200 μL, using a SpectraMax M5 Microplate Reader (Molecular Devices). The plate validation assay was performed with replicate columns of positive control wells, negative control wells and blank wells. Blank columns contained only buffer and pNPP. Negative control (uninhibited) contained buffer, pNPP, and DUSP5 PD(WT); and, positive control contained the same components, but also contained 10 μM sodium orthovanadate. The plate was then shaken and allowed to equilibrate in the spectrophotometer at 25 °C for 30 min. After incubation, 4 μL of a 50 μM enzyme stock was dispensed into appropriate wells utilizing a single-channel pipette. This produced a final enzyme concentration of 1 μM. Before a read was taken, the plate was shaken for five seconds. The initial rate for the DUSP5 PD(WT) reaction was linear for approximately 90 min; and, the plate was kept in the spectrophotometer at 25 °C for an additional 80 min after the kinetic read. The endpoint reading was subsequently taken at 90 min after initiation of reaction.
Slopes from the kinetic read, as well as single-point absorbance values at the 90-minute endpoint read, were then averaged. For blank wells and positive control wells, both slope values (continuous assay) and single point absorbance values (fixed time assay) were approximately zero, as expected (Table 2). Standard deviations were calculated and a Z’ value [33] subsequently determined using the following equation:
$$ Z\hbox{'}=1-\frac{3\left({\sigma}_p+{\sigma}_n\right)}{\left|{\mu}_p-{\mu}_n\right|} $$
(2)
where σp is the standard deviation for the positive control, σn is the standard deviation for the negative control, μp is the mean for the positive control, and μn is the mean for the negative control. The Z’ value is a coefficient denoting the quality of a high throughput screening assay, reflecting both the variation in data and dynamic range for the assay. A good assay exhibits a high signal to background ratio. A Z’-factor of 1.0 reflects an ideal assay; and, for an assay to be considered reliable, must exceed 0.5 [33].
IC50 measurements
IC50 values were obtained using the assay described above, in 96 well plates. The maximum inhibitor concentration screened in any plate was 300 mM and the minimum screened concentration was 1 μM. The IC50 plate was designed so that the first column of wells served as blanks, with wells containing only buffer and substrate. The second column of wells functioned as the plate negative control, with each well containing buffer, substrate and enzyme. The remaining wells in the plate contained buffer, substrate, enzyme, and varying amounts of inhibitor, with inhibitor concentration increasing from left to right across the plate. Data points were collected at a minimum in triplicate, and inhibitor concentrations were chosen to provide data equally spread on a logarithmic scale. The composition of buffer and the concentrations of substrate and enzyme utilized were identical to those in the plate validation assay. After initiation and shaking, a ten-minute kinetic read was taken.
For each plate assayed, the slope values for all negative control wells were averaged and the measured value considered representative of full enzymatic activity. Fractional activity was then calculated by dividing the slope of each inhibitor well by this value, determining the relative amount of enzyme activity observed at each concentration of inhibitor. Values were then plotted as percent activity versus the log of the concentration of inhibitor, and fitted to the following equation:
$$ y= Bottom+\frac{\left( Top- Bottom\right)}{1+{10}^{x- \log I{C}_{50}}} $$
(3)
where Top and Bottom are plateaus for the values of initial velocity when uninhibited and fully inhibited, respectively.
Nephelometry
Nephelometry is a technique for measuring the relative aggregation of particles in solution, based on the light-scattering properties of molecular aggregates [34]. We performed nephelometry to explore the ability of the chemicals studied herein to form aggregates, which can lead to artifactual inhibition effects. Compounds were tested for aggregation in 96-well plates using a buffer containing 100 mM Tris base, 100 mM sodium chloride, and 5 mM magnesium chloride at pH 7.5. Each compound analyzed in these experiments contained concentrations of compound ranging from 10-100 μM, recorded in quadruplet. Each plate was analyzed at two separate gain values of 52 and 72. Data were collected using a BMG NEPHELOstar Plus, equipped with a 635 nm laser.
NMR binding assay
NMR samples of DUSP5 PD(C263S) were prepared for 2D 1H-15N HSQC (heteronuclear single quantum coherence) spectral titration studies. The 15 N-labeled DUSP5 PD(C263S) protein was concentrated using an Amicon Ultra-4 centrifugal device (Millipore) to 600 μM. NMR samples were prepared with the following conditions for RR505: 250 μM RR505, 250 μM DUSP5 PD(C263S), 10 % D2O, 50 mM potassium phosphate, 100 mM KCl, and 2 mM DTT at pH 6.8 and for CSD
3
-2320: 0 or 500 μM CSD
3
-2320, 500 μM DUSP5 PD(C263S), 10 % D2O, 50 mM potassium phosphate, 100 mM KCl, and 2 mM DTT at pH 6.8. NMR experiments were performed on a 500 MHz Varian NMR System using a triple resonance probe with z-axis gradients at 25 °C.
ERK dephosphorylation assay
For this assay, 10 ng of GST-tagged recombinant phosphorylated ERK2 (R&D Systems, 1230-KS) was incubated with and without the indicated DUSP5 proteins (0.5 nM final concentration) for 15 min at room temperature, with or without the indicated drugs. The reactions were halted with 2x Laemmli sample buffer and subjected to SDS-PAGE. The proteins were transferred to polyvinylidene difluoride (PVDF) and immunoblotted using antibodies to pERK (Cell Signaling Tech., #9106) and total ERK, which includes both phosphorylated and unphosphorylated ERK1 and ERK2 (Cell Signaling Tech., #9102). Bound antibodies were visualized using horseradish peroxidase-linked anti-mouse IgG (Cell Signaling Tech, #7076S) and anti-rabbit IgG (Cell Signaling Tech, #7074S), respectively, and ECL reagents (Pierce, #34708) according to the manufacturer’s protocol. For calculating IC50 values, gel bands were imaged by chemiluminescence with either film or digital image capture by a FluorChem HD2 imager (Alpha Innotech). Density of each band was quantified with ImageJ software by using the gel analysis tool. Relative values of phosphorylated ERK present for each drug concentration treatment compared to pERK only controls were calculated. These relative values were then used to obtain IC50 values with GraphPad Prism 6 software. Each experiment was repeated at least three independent times, and IC50 values provided as a range.