The authors have declared that no competing interests exist.
Conceived and designed the experiments: AG US YZ MS AE JT MF KP. Performed the experiments: AG US PB JS YZ MS NP AE ML. Analyzed the data: AG US PB JS YZ MS NP AE ML JT MF KP. Contributed reagents/materials/analysis tools: US AE JT MF KP. Wrote the paper: AG US JS JT MF KP.
Drug repurposing or repositioning is an important part of drug discovery that has been growing in the last few years for the development of therapeutic options in oncology. We applied this paradigm in a screening of a library of about 3,800 compounds (including FDA-approved drugs and pharmacologically active compounds) employing a model of metastatic pheochromocytoma, the most common tumor of the adrenal medulla in children and adults. The collection of approved drugs was screened in quantitative mode, testing the compounds in compound-titration series (dose-response curves). Analysis of the dose-response screening data facilitated the selection of 50 molecules with potential bioactivity in pheochromocytoma cells. These drugs were classified based on molecular/cellular targets and signaling pathways affected, and selected drugs were further validated in a proliferation assay and by flow cytometric cell death analysis. Using meta-analysis information from molecular targets of the top drugs identified by our screening with gene expression data from human and murine microarrays, we identified potential drugs to be used as single drugs or in combination. An example of a combination with a synergistic effect is presented. Our study exemplifies a promising model to identify potential drugs from a group of clinically approved compounds that can more rapidly be implemented into clinical trials in patients with metastatic pheochromocytoma or paraganglioma.
Pheochromocytoma (PHEO) is a rare neuroendocrine tumor that develops in the adrenal medulla and represents the most common tumor in this location in children and adults
From an industry perspective, drug development programs for rare (“orphan”) diseases such as PHEO/PGL (approximately 1,000 new cases are diagnosed in the US each year) are less appealing because of the low return on investment. Thus, alternative approaches must be sought to discover novel therapeutic options for these tumors. One potential strategy is to “recycle” drugs that have been approved for use in the treatment of other diseases, a strategy known as drug repurposing or repositioning
In the present study we identified and validated new therapeutic options for PHEO/PGL by screening the NIH Chemical Genomic Center (NCGC) Pharmaceutical Collection (NPC), a large library of clinically approved drugs
We used the following cell lines, which represent the only available permanent PHEO cell lines available to the scientific community and include a range of models of PHEO. The rat PHEO cell line PC12, developed in 1976, has a MAX gene deletion that has been recently discovered in a human PHEO kindred
Cell viability after compound treatment was measured using a luciferase-coupled ATP quantitation assay (CellTiter-Glo, Promega) in MTT cells. The change of intracellular ATP content indicates the number of metabolically competent cells after compound treatment. MTT cells were harvested from T225 flasks and resuspended in DMEM medium with 5% FBS and 1% horse serum. Then 5 µl of a suspension of 200,000 cells/ml was dispensed into each well of white, solid bottom, 1536-well tissue culture–treated plates using a Multidrop Combi dispenser. After overnight culture at 37°C with 5% CO2, a total of 23 nl of compounds at 8 selected concentrations from the NPC or positive control (10 mM stock of doxorubicin hydrochloride) in DMSO was transferred to each well of the assay plate using a pintool (Kalypsys, San Diego, CA), and the plates were further incubated at 37°C with 5% CO2 for 24 or 48 hrs. After that, 4 µl of CellTilter-Glo™ luminescent substrate mix (Promega) was added to each well. The plate was incubated at room temperature for 15 minutes. The plates were measured on a ViewLux plate reader (PerkinElmer) with a clear filter. The final concentration of the compounds in the 5 µl assay volume ranged from 0.5 nM to 46 µM.
The NPC consists of 3,826 small molecule compounds, with 52% of the drugs approved for human or animal use by the FDA
The compounds from the NPC library were prepared as 15 interplate titrations, which were serially diluted 1∶2.236 in dimethyl sulfoxide (DMSO) (Thermo Fisher Scientific, Waltham, MA) in 384-well plates. The stock concentrations of the test compounds ranged from 10 mM to 0.13 µM. The transfer of the diluted compounds from 384-well plates to 1536-well plates was performed using an Evolution P3 system (PerkinElmer Life and Analytical Sciences, Waltham, MA). Each treatment plate included concurrent DMSO and positive control wells and concentration-response titrations of controls, all occupying columns 1 to 4. During screening, the compound plates were sealed and kept at room temperature, whereas other copies were maintained at −80°C for storage.
To determine compound activity in the qHTS assay, the titration-response data for each sample were plotted and modeled by a four parameter logistic fit yielding IC50 and efficacy (maximal response) values. Raw plate reads for each titration point were first normalized relative to positive control (doxorubicin hydrochloride, 100% inhibition) and DMSO-only wells (basal, 0%). Curve-fits were then classified by the criteria described
There were a total of 30 plates in the primary qHTS screen, which included 24 plates corresponding to the NPC library set and 6 DMSO plates. Compounds from the primary qHTS screen were classified into three categories according to the quality of curve fit and efficacy. Actives were compounds in curve class 1.1, 1.2, 2.1 and 2.2 curves with efficacy higher than 60%; inactives were compounds with class 4 curves; and inconclusive included all other compounds including those shallow curves and curves with single point activity.
Cell proliferation was determined by the MTT assay (also referred as 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay). MTT cells (15×103) were incubated in 96-well plates for 24 hours in complete medium before the addition of the indicated compound. A solution of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (1 mg/ml; Sigma-Aldrich) was added and plates were incubated at 37°C for 3 hours before measuring absorbance at 562 nm using a Wallac Victor 3 1420 Multilabel plate reader (Perkin Elmer).
For PARP cleavage, MTT cells were treated with the indicated dose and concentration of drug for 20 hrs at 37°C and 5% CO2. Cleaved and full-length PARP (rabbit anti-PARP antibody from Cell Signaling and HRP-conjugated anti-rabbit antibody from Jackson ImmunoResearch) and actin (mouse anti-actin antibody from Millipore and HRP-conjugated anti-mouse antibody from Jackson ImmunoResearch), for loading control, were measured by immunoblotting.
PC12 cells were plated overnight and incubated with three different concentrations of the various drugs for 24 hrs. To determine apoptotic and viable cells, cells were washed with PBS buffer (2 mM EDTA, 0.05% BSA in PBS) and stained with 5 µl of 7-AAD (BioLegend) for 10 min. Cells were acquired by flow cytometry (MACSQuant Analyzer, Miltenyi Biotec) and 7-AAD-negative viable cells (%) were analyzed by FlowJo software (Tree Star).
Three separate MTT cell samples were prepared according to Affymetrix protocols (Affymetrix, Inc.). RNA quality and quantity was ensured using a Bioanalyzer (Agilent, Inc.) and NanoDrop (Thermo Scientific, Inc.) respectively. Per RNA labeling, 200 ng of total RNA was used in conjunction with the Affymetrix recommended protocol for GeneChip 1.0 ST chips.
The hybridization cocktail containing fragmented and labeled cDNAs was hybridized to the Affymetrix Mouse Genome ST 1.0 GeneChip. The chips were washed and stained by the Affymetrix Fluidics Station using the standard format and protocols as described by Affymetrix. The probe arrays were stained with streptavidin phycoerythrin solution (Molecular Probes, Carlsbad, CA) and enhanced using an antibody solution containing 0.5 mg/ml of biotinylated anti-streptavidin (Vector Laboratories, Burlingame, CA). An Affymetrix Gene Chip Scanner 3000 was used to scan the probe arrays. Gene expression intensities were calculated using Affymetrix AGCC software.
Partek Genomic Suite was used to RMA normalize (Robust Multichip Analysis), summarize, logtransform the data, and run ANOVA analysis and hierarchical clustering. The Series entry for the murine microarray in the NCBI Gene Expression Omnibus (GEO) database has been approaved with the following number: GSE51832 (
RNA extraction from MTT cells was performed as previously described
A subset of patient data that was previously published
Genes corresponding to <2-fold change between the MTT cell line and SDHB PHEO datasets were considered for further analysis. This selection criterion resulted in a list of 1440 genes with an overall correlation of 0.86. To evaluate the similarity of drug targets between MTT vs. SDHB samples, a Pearson correlation coefficient was computed between a set of target genes from each drug with an overall average correlation of 0.8.
Two metabolic models corresponding to different levels of information were constructed. One is a general model that contains the 22 tested drugs and their curated target genes. The second network derives from the first one, which is based on the top 20 hub nodes. Network assembling, visualization and determination of statistical parameters were performed using Cytoscape v2.8.3
We have included a cytoscape network file (for
A) Drug name and number of high correlated genes mapped from the human PHEO and the murine MTT microarray data set. B) Connected node network representation in which all the targets of each drug are connected to the targets of all other drugs in the network.
For the synergism study we used the CalcuSyn Windows software for dose-effect analysis and synergism/antagonism quantification, following the manufacturer's instructions. Drug synergism was determined from median effect analysis equations developed by Chou-Talalay
We screened the NIH Chemical Genomic Center Pharmaceutical Collection of clinically approved drugs, which contains 1,760 US FDA approved drugs, 785 drugs approved by other countries, 1,225 compounds in clinical trials and 56 bioactive molecules, employing a model of metastatic PHEO (by using MTT mouse PHEO cells), the most common tumor of the adrenal medulla.
The assay, in a 1536 HTS plate format, measured cell viability by determining metabolically active cells (viable cells) in culture using a luciferase, ATP-dependent readout (as described in Material and Methods), at two different time points of compound incubation (24 hrs and 48 hrs). A number of compounds showed significant cell killing, with potency below 10 µM (at 48 hrs, when the compounds showed to be more potent and efficacious). Compounds were tested at 8 doses using a quantitative HTS (qHTS) approach
The qHTS identified 76 high-confidence active compounds (
Sample Name | Curve Class | IC50 (µM) | Efficacy (%) |
Colchicine |
−1.1 | 0.47 | −85 |
Dipyrithione (Crimanex) | −1.1 | 1.50 | −81 |
Zinc pyrithione | −1.1 | 2.11 | −91 |
1-Hydroxypyridine-2-thione zinc salt | −1.1 | 2.11 | −96 |
Mersalyl sodium | −1.1 | 2.66 | −92 |
Auranofin (Ridaura) | −1.1 | 2.66 | −87 |
Thimerosal |
−1.1 | 2.99 | −87 |
Deslorelin acetate (Suprelorin) | −1.1 | 2.99 | −89 |
Paclitaxel |
−1.2 | 0.04 | −51 |
5-Aza-2′-deoxycytidine, Decitabine |
−1.2 | 0.07 | −52 |
Homoharringtonine | −1.2 | 0.24 | −74 |
Trimetrexate glucuronate |
−1.2 | 0.38 | −51 |
Rubitecan |
−1.2 | 0.42 | −55 |
Nocodazole |
−1.2 | 0.53 | −60 |
Fenbendazole |
−1.2 | 1.01 | −61 |
Artemisinimum | −1.2 | 1.33 | −60 |
Carmofur (Mifurol) | −1.2 | 1.88 | −57 |
Suberoylanilide hydroxamic acid |
−1.2 | 2.37 | −59 |
Tenovin-1 | −1.2 | 2.66 | −56 |
Carubicinum |
−1.2 | 2.99 | −59 |
Captan |
−1.2 | 2.99 | −66 |
Lissamine green B | −1.2 | 6.68 | −54 |
Mycophenolic acid (CellCept, Myfortic) | −2.1 | 1.14 | −97 |
Tyrothricin |
−2.1 | 1.50 | −84 |
Mycophenolate mofetil | −2.1 | 1.68 | −92 |
Brilliant Green |
−2.1 | 1.68 | −108 |
Rotenone |
−2.1 | 2.66 | −124 |
Lestaurtinib | −2.1 | 2.66 | −83 |
Ciclopirox ethanolamine | −2.1 | 4.53 | −85 |
RTA 402 | −2.1 | 6.68 | −108 |
Sanguinarine | −2.1 | 8.41 | −88 |
Proflavine hemisulfate |
−2.1 | 8.41 | −92 |
Parthenolide | −2.1 | 8.41 | −104 |
Bortezomib |
−2.2 | 0.60 | −75 |
Albendazole |
−2.2 | 0.72 | −58 |
Sobuzoxane |
−2.2 | 0.84 | −56 |
Azacitidine |
−2.2 | 1.33 | −71 |
Tiquizium bromide | −2.2 | 1.88 | −52 |
Flavopiridol hydrochloride hydrate | −2.2 | 1.88 | −71 |
Ancitabina | −2.2 | 2.37 | −68 |
Ethaverine hydrochloride | −2.2 | 3.60 | −60 |
2,2′,4′-Trichloroacetophenone | −2.2 | 3.76 | −71 |
Berberine chloride | −2.2 | 4.22 | −54 |
17-Allylamino-geldanamycin (17-AAG) | −2.2 | 4.22 | −81 |
Proguanil hydrochloride |
−2.2 | 4.53 | −60 |
Topotecan hydrochloride |
−2.2 | 4.73 | −90 |
Phenelzine sulfate | −2.2 | 5.08 | −53 |
Oxapium iodide | −2.2 | 6.68 | −60 |
Methyl violet |
−2.2 | 7.50 | −70 |
Malachite Green Oxalate | −2.2 | 8.41 | −61 |
The table illustrates the drug name, efficacy, IC50 (in µM) and curve class.
Anti-tubulin agents;
Drugs targeting DNA and nucleotide analogues;
proteasome inhibitors;
antimicrobial agents;
antimetabolite.
The top 50 drugs considered active compounds based on the primary screening were grouped and mostly classified in 5 main functional categories: 1) antitubulin agents; 2) drugs targeting DNA and nucleotide analogues; 3) proteasome inhibitors; 4) antimicrobial agents; 5) antimetabolite. We found that antitubulin agents, which have a prominent class effect on protein polymerization and mitotic spindle organization, were the drug category with the most entries in the top compounds in the high-confidence list (five entries, which included colchicine, paclitaxel, and several –azoles drugs). Another class effect was represented by topoisomerase inhibitors (including carubicinum, rubitecan, sobuzoxane and topotecan), which have a prominent effect on DNA replication and telomerase maintenance.
A representative selection of compounds from the categories described above were selected for further investigation and validated in additional assays to confirm the activity of the hits from the HTS screen. For the secondary assays, we took advantage of a traditional proliferation assay, namely the colorimetric MTT assay, and a flow cytometric analysis for viable cells after drug treatment. We used three different PHEO cell lines, namely PC12, MPC and MTT cells (see Material and Methods for a complete description). All the compounds tested produced dose-response curves in all three cell lines, and were potent compounds having IC50 values in the low nanomolar range (
To further validate the activity of the selected compounds, we performed a flow cytometry cell viability assay using 7-AAD after overnight treatment with the selected compounds (
A) To quantify viable cells, the membrane impermeable dye 7-amino actinomycin D (7-ADD) was added to a cell suspension of PC12 cells, after overnight treatment with the respective compound. The bar graph represents relative live cell percentage. B) MTT cells were cultured with the indicated drug and dose for 20 hr and total cell lysates were subjected to Western blot for full-length PARP, cleaved PARP and actin.
Because the primary and secondary screening was performed in murine cell lines (due to the unavailability of human PHEO cell lines), a strategy using microarray data from murine and human samples was used to further select a clinically relevant set of drugs from our data that could be potentially developed in human clinical trials. We first compiled a list of genes directly targeted by the top drugs discovered by our primary screening using the NPC library databank. Compounds for which information on specific targets was not available were excluded from the analysis. Based on the list of genes compiled, we selected a subset of microarray data from a new murine MTT cell microarray and a published human SDHB microarray. In particular, we focused our attention on the human microarray of patients with SDHB disease, as these patients are the ones that can benefit the most from the discovery of novel therapies. To extrapolate the murine screening data for clinical relevance, for each drug we measured the Pearson correlation coefficient between the target genes and the expression profiles from these two microarray data sets (
Two metabolic models were constructed with compounds and targets that have been compiled from our data set. The first model was constructed as a complete hypothetical network for visualization of the network assembly of 22 compounds tested on the MTT cell line (
IC50 folds | EPI/SAHA [µM] | Fraction Affected | CI | Effect | DRI EPI/SAHA |
20/4000 | 0.808566317 | 0.719 | Synergistic | 8.227/1.674 | |
10/2000 | 0.740589432 | 0.537 | Synergistic | 8.593/2.378 | |
5/1000 | 0.588437471 | 0.568 | Synergistic | 5.459/2.6 | |
2.5/500 | 0.446667935 | 0.553 | Synergistic | 4.23/3.156 | |
1.25/250 | 0.328056979 | 0.518 | Synergistic | 3.675/4.069 |
CI = Combination index. DRI = Dose reduction index.
The compiled list of drugs derived from our screening represented a useful selection of compounds with the potential to work effectively in the treatment of metastatic PHEO/PGL, either as single drugs or in combination. Moreover, our combined analysis of the drug screening and the microarray data sets suggested possible drug combinations that are supported by sporadic, but useful, reports in the literature on similar drugs or drug categories in the treatment of metastatic PHEO/PGL patients.
The subnetwork represented in
A and B) Dose response curves for SAHA and Epirubicin; C) Median-effect plot; D) Algebraic estimate of the combinational index (CI) for the combination of epirubicin with SAHA relative to the fraction of affected cells.
In the present study, we identified drugs by a repurposing approach from the screening of an extensive library containing FDA-approved compounds that are cytotoxic for PHEO cell lines. We further validated active compounds from the screen and used data from human and murine microarrays to extrapolate potential clinically useful information to help select the hits with highest clinical potential. We used our approach to suggest an example of a synergistic drug combination using the HDAC inhibitor SAHA and the topoisomerase inhibitor epirubicin, which could represent an example of possible successful treatment for metastatic PHEO/PGL.
HTS of chemical libraries is a powerful generator of potentially clinically useful compounds to treat metastatic PHEO/PGL. This approach is novel for the field of PHEO/PGL, which is now experiencing a renaissance of interest in drug discovery, in particular in the evaluation of novel targeted therapies, especially following the discovery of several novel gene mutations (and consequently several intracellular signaling pathways with potential cellular targets) predisposing to the disease
Compared with other cancer types, one of the major challenges in the development of novel therapeutics for metastatic PHEO/PGL is the absence of
Using meta-analysis information from molecular targets of the top drugs identified by our screening with gene expression data from human and murine microarrays, we identified potential drugs to be used as single drugs or in combination.
There are multiple advantages of screening small molecule libraries of clinically approved compounds by repurposing strategies, including the possibility of introducing these drugs more rapidly into clinical trials because those drugs have already been used in humans for the treatment of other diseases, including other types of cancer. Drugs that received regulatory approval have already proven to be safe and effective for a particular disease, and we have historical information regarding the long-term side effects of the drug in a large population, which makes repurposing them for other diseases less time-consuming. Moreover this approach overcomes several of the economic and technical bottlenecks that are inherent to the drug discovery process
Our primary qHTS screening has identified several compounds with potential activity on PHEO/PGL. The repertoire of compounds identified is a potential source available to the PHEO/PGL community for further testing, either as single drugs or in combination, either with other drugs in our list or with other drugs that have shown activity on PHEO. Of particular interest, the antitubulin drugs seem to emerge from our screening as a group of drug with a prominent “class effect” on PHEO, with potent activity in the low nanomolar range. This suggests that these compounds may have activity as single agents in the treatment of metastatic PHEO and possibly PGL.
The secondary screening confirmed activity of several compounds, including the antitubulin agents, the HDAC inhibitor SAHA and the topoisomerase inhibitor rubitecan. Subsequent analysis combining the results from the screening with data from expression microarray assays resulted in the selection of two drugs for combination testing.
The drug combination we chose as an example of using data from the human microarray crossed with the cellular network of gene targets of the drug identified by the screening find validation also in the literature. Interestingly, SAHA has been described as a drug that induces SDHB protein stabilization and the entry of this protein (although mutated but still functional) into the mitochondria
Of note, our screening was able to identify compounds that we have already explored for the potential treatment of PHEO, and so they acted as internal controls of the validity of our approach. For example, one of the drugs in the top 50 active compound list was the Hsp90 inhibitor 17-AAG, which we investigated in further detail in another study
Another prominent category from our screening was represented by drugs targeting DNA. Recent evidence points to a role of DNA methylation in the pathogenesis of SDH mutant PHEOs
In conclusion, we have presented here the adoption of an integrated approach to discover potentially clinically useful compounds for the treatment of metastatic PHEO/PGL, and we hope that this strategy will help to move forward the field of drug development for other orphan diseases.
Supplemental data.
(CYS)
Enrichment analysis. Enrichment analysis (by therapeutic category) of active compounds from the primary screening of the NPC drug library. Gray bars represent the total number of drugs in the specific therapeutic category, white bars represents the number of active compounds and black bars represents the results of the enrichment analysis as described in the Material and Methods section. The green dash line marks an enrichment ratio >20% in the active drugs.
(TIF)
Subnetwork and hubnodes. A) Subnetwork with all interactions associated with drugs that include topoisomerase (DNA replication) and histone deacetylase (telomerase maintenance). B) Top 20 hubnodes network derived from the global network, colored by up- and down-regulated genes common to human PHEO SDHB and murine MTT cells.
(TIF)
NPC library screening, top compounds. Seventy-six high-confidence active compounds identified in the quantitative high-throughput screening with curve class 1.1, 1.2, 2.1 and 2.2, and maximal inhibition over 60%.
(XLSX)
NPC library screening, complete list.Complete list of compounds from the NPC library including active, inconclusive and inactive compounds. The table includes the NCG compound identification number (Sample ID), chemical structure composition, type of curve class based on the screening analysis, IC50 (in µM), efficacy and the identification of the compound as active, inconclusive or inactive based on the curve classification illustrated in
(XLS)
Gene targets-drugs correlation. List of drugs and their targets compiled from databases as described in
(DOCX)
Centrality scores. Top 20 genes based on the centrality score from the network common to SDHB (human) and MTT cells (murine), to look at the similarity of the dataset (MTT with SDHB) based on the microarray data.
(DOCX)
Hubnodes scores. Top 20 hubnodes (nodes with the most association in the network) showing the node importance score (derived from the eccentricity method). The hubnode can be either a gene or a drug in the interrelating network. The highest is the number of associations of a hubnode, the highest is the score.
(DOCX)