A Membrane Transporter Determines the Spectrum of Activity of a Potent DNA-Targeted Hybrid Anticancer Agent

Cytotoxic drugs that are
mechanistically distinct from current chemotherapies are attractive components
of personalized combination regimens for combating aggressive forms of cancer.
To gain insight into the cellular mechanism of a highly potent platinum–acridine hybrid agent, we
performed a correlation analysis of NCI-60 compound screening results and gene
expression profiles. We discovered a plasma membrane transporter, human
multidrug and toxin extrusion protein 1 (hMATE1, SLC47A1), as the dominant
pan-cancer predictor for cancer cell chemosensitivity to the hybrid agent. We
have validated the role of hMATE1 using transporter inhibition, gene knockdown,
and chemical sensitization assays. The results suggest that hMATE1 may have
applications as a molecular marker to identify and target tumors that are likely
to respond to platinum–acridines. Furthermore, enhancement of hMATE1 expression
by epigenetic drugs emerges as a potential co-treatment strategy to sensitize
tumor tissue to platinum–acridines and other anticancer drugs transported by
hMATE1.

cancer cell chemosensitivity to the hybrid agent. We have validated the role of hMATE1 using transporter inhibition, gene knockdown, and chemical sensitization assays. The results suggest that hMATE1 may have applications as a molecular marker to identify and target tumors that are likely to respond to platinum-acridines. Furthermore, enhancement of hMATE1 expression by epigenetic drugs emerges as a potential co-treatment strategy to sensitize tumor tissue to platinum-acridines and other anticancer drugs transported by hMATE1. ABSTRACT: Cytotoxic drugs that are mechanistically distinct from current chemotherapies are attractive components of personalized combination regimens for combatting aggressive forms of cancer.
To gain insight into the cellular mechanism of a potent platinum-acridine anticancer agent (compound 1), a correlation analysis of NCI-60 compound screening results and gene expression profiles was performed. EED226. The results suggest that hMATE1 may have applications as a pan-cancer molecular marker to identify and target tumors that are likely to respond to platinum-acridines. Furthermore, enhancement of hMATE1 expression by epigenetic drugs may be a potential co-treatment strategy to efficiently deliver platinum-acridines and other clinical anticancer drugs transported by hMATE1 to tumor tissue.

INTRODUCTION
Since the FDA approval of cisplatin (Figure 1a), chemically unique approaches have been pursued to improve the efficacy and safety of platinum-based chemotherapy. 1 The design of several of the newergeneration nonclassical metallodrugs is based on the premise that tumor resistance can be overcome at the DNA level as a consequence of the agents' unique DNA binding modes and DNA damage response (DDR) patterns. 2 This reasoning has redefined the landscape of platinum anticancer drug discovery and resulted in promising new clinical and preclinical candidates. 3 One type of compound in preclinical development are platinum-acridine agents, represented by compound 1 (Figure 1b), the most potent derivative 4 identified in this class of cytotoxics. 5,6 Platinum-acridines bind to DNA by a mechanism that involves intercalation and platination of nucleobase nitrogen, causing a more severe form of DNA damage than the cross-links observed for cisplatin. 5 On a per-adduct basis, the hybrid agents are more potent inhibitors of DNA synthesis than cisplatin, which induce replication fork arrest and a high level of DNA double-strand breaks requiring specialized DNA repair modules, 7 and are more efficient transcription inhibitors. 8 These mechanisms most likely contribute to the high cytotoxicity of platinumacridines, particularly in non-small-cell lung cancer (NSCLC), where the hybrid agents show up to 1000fold higher activity than cisplatin. 6 Collectively, the results from mechanistic studies in cell-free systems, human cancer cells, and chemical genomic fitness profiling in S. Cerevisiae 9 are consistent with nuclear DNA as the principal target of these agents.
Platinum-acridines show a dramatically higher activity than cisplatin in NSCLC, even though the hybrid adducts are repaired more rapidly than the classical cross-links in these notoriously DNA repairproficient cells. 7 These findings call into question whether the damage at the genome level and cellular response platinum-acridines cause alone overcome chemoresistance in NSCLC. In this article, we report the results of a study that combined activity screening and gene expression correlation analysis, as well as functional target validation performed on compound 1. We not only demonstrate a complete lack of similarity of the compound's antitumor profile with that of the classical platinum drugs, but also discovered a membrane transporter, human multidrug and toxin extrusion protein, hMATE1 (SLC47A1), as the single most predictive marker of chemosensitivity to platinum-acridines and demonstrate its potential utility as a target for personalized cancer treatment.

EXPERIMENTAL SECTION
Compound Screening. Compound 1 was tested by the NCI Developmental Therapeutics Program in a panel of 59 cancer cell lines in a one-dose screen at 10 M test compound and in five-dose screens over a concentration range of 10 -4 to 10 -8 M. Five-dose screens were performed in duplicate.
Reported GI50 values and the chemosensitivity profiles (mean graph) are means of the two experiments.
All correlation analyses were based on GI50 assay endpoints. 10 Correlation and Gene Set Overlap Analysis. Comparative analysis of NCI-60 activity profiles based on GI50 end points was performed with the COMPARE analysis tools 11  gene expression patterns based on transcript levels (z-scores) from 5 different microarray platforms were analyzed in a similar manner for a total of 58 cell lines with a minimum correlation of R = ± 0.30 (for N = 58, R = ± 0.259 is statistically significant at p < 0.05). The CellMiner tool 12,13 was used to compare the gene expression, DNA copy number alteration, and DNA methylation status for SLC47A1 across NCI-60 cell lines (database version 2.2, https://discover.nci.nih.gov.cellminer; human genome version HG19, number of genes: 25683). Correlation analysis of SLC47A1 transcript levels (average log2 intensities) and DNA methylation (scores 0-1 for completely unmethylated to completely methylated gene promoters) was done with CellMinerCDB (version 1.1; discover.nci.nih.gov/cellminercdb), which implements the GDSC (Sanger Institute) cell line set and databases. 12 Correlations between ad-hoc processing. Panels were assembled and annotated without any additional enhancements of images, unless explicitly stated, in Adobe Photoshop CC, version 2017. 1 For RNAi knockdown of hMATE1 (SLC47A1) in imaging assays, A549 cells were harvested from T-75 cell culture flasks and seeded on a 6-well plate at a density of 150,000 cells per well. Silencer Select siRNA1, Silencer Select siRNA2, and scrambled siRNA (Silencer Select Negative Control #1) were thoroughly mixed with RNAiMAX in Opti-Mem media according to manufacturer's protocol and added to each well. Cells transfected at a final siRNA concentration of 20 nM for 48 hours at 37 °C were detached with trypsin and subcultured at a 1:2 ratio into 35 mm MatTek plates and allowed to attach overnight. Cell culture medium was replaced and supplemented with 10 M compound 1, and dishes were incubated for 4 hours at 37 °C. Cells were fixed with 4% formaldehyde, permeabilized with 0.5% Triton X-100, washed with PBS, incubated with 7.5% BSA) for 30 minutes at room temperature, and incubated with appropriately diluted primary antibody in 1% BSA for 1 hour at room temperature. After the cells were washed with PBS, they were incubated with the secondary antibody (Goat-anti-Rabbit IgG Alexa Fluor-635), diluted in 1% BSA (1:400 anti-rabbit) for 1 hour at 37 °C. After three PBS washes, samples were imaged immediately or stored in PBS at 4 °C for further testing.
For sensitization assays, HCT-116 cells were harvested from T-75 flasks, seeded on a 24-well plate with glass-like polymer bottom (P24-1.5P, Cellvis, Sunnyvale, CA) with 25,000 cells per well, and allowed to attached overnight. Single drugs or drug combinations were tested in this assay at final concentrations of 2.5 M EED226, 2.5 M EPZ-6438, 500 M valproic acid, and 10 M decitabine (see assay layout 1, AL1, in the SI). Cells were incubated at 37 °C for 72 hours. Each well was replaced with fresh medium supplemented with 10 M compound 1, and incubation was continued for 4 additional hours. Each well was washed with 3 times with warm PBS, before cells were fixed with 0.5 mL of 4% formaldehyde at room temperature for 15 minutes. After 3 PBS washes, plates were immediately imaged or stored at 4 °C until analyzed. Subsequent incubations of HCT-116 cells at escalating doses of EED226 and EPZ-6438 were performed analogously (see AL2 in the SI). Representative conditions that were screened for expression levels of hMATE1 were determined by immunofluorescence with anti-hMATE1 antibody as described in RNAi knockdown experiments.

Uptake of Compound 1 Studied by ICP-MS.
Protocols for the quantification of intracellular platinum-acridines by ICP-MS have been described previously. 16 Briefly, cells collected from the transporter inhibition and hMATE1 knockdown assays (see below) were pelleted and homogenized by microwave-assisted digestion (ETHOS UP Milestone, Sorisole, Italy) in a mixture of dilute, trace-metal grade HCl and HNO3. Standard curves appropriate for quantification of platinum in specified uptake assays were generated using concentrations of 0 ppt, 20 ppt, 50 ppt, 100 ppt, 200 ppt, and 500 ppt of a diluted Pt standard (High-Purity Standards, Charleston, SC, USA). An 8800 Triple Quadrupole ICP-MS spectrometer (Agilent, Tokyo, Japan) equipped with a SPS 4 automatic sampler, a Scott-type double pass spray chamber operated at 2°C, and a Micromist concentric nebulizer was used for analysis. Helium gas (99.999% purity, Airgas, Colfax, NC, USA) was used in the collision/reaction cell to minimize potential spectral interferences while monitoring the isotope 195 Pt.
For transporter inhibition assays, 700,000 A549 cells in 2.5 mL of F12K media (ATCC , supplemented with 10% FBS, and 10% penstrep, and 10% L-glutamine (Thermo Fisher, 25030-081), were seeded into T-25 flasks and allowed to attach overnight. Cells pre-treated with 100 nM pyrimethamine for 25 minutes and untreated cells were then dosed with 100 nM compound 1 for 3 hours.
After treatment, medium was aspirated, and cells were washed 3 times with fresh media. Trypsin was added to detach cells, and 3 mL of fresh media were added to each flask to collect the cell suspensions, which were pelleted by centrifugation at 250  g for 3 minutes. After the supernatant was aspirated, pellets were washed with 3 mL of PBS solution twice and centrifuged again at 250  g for 3 minutes.
Pellets were stored at -80 °C until analyzed by ICP-MS. The assay was performed in triplicate for each treatment group.
For uptake studies after hMATE1 (SLC47A1) knockdown, A549 cells were reverse-transfected with Silencer Select siRNA1 or Silencer Select Negative Control #1 scrambled RNA for 48 hours using the RNAiMAX system in Opti-Mem media. Media was replaced with fresh antibiotics-free DMEM/F12K medium and incubation was continued for an additional 24 hours. Cells were then incubated with 100 nM compound 1 at 37 °C for 4 hours, and cell pellets were prepared as described above. The assay was performed in triplicate for each treatment group. Microwave digestions and ICP-MS analysis for Pt were performed as described above.  In pyrimethamine competition assays, A549 cells were seeded at a density of 5000 cells per well and allowed to attach for 24 hours. Cells were then pre-treated with 10 or 100 nM pyrimethamine for 20 minutes and subsequently incubated with 100 nM compound 1 or DMF-containing media (control) for 72 hours. No-treatment controls were also included. Assays were run in duplicate with 6 replicates per plate. Cell viability was assessed as described above.

Cell Proliferation
Cell viability in RNAi knockdown assays was assessed by transfecting A549 cells on 96-well plates using a reverse transfection protocol. Briefly, Silencer siRNA, scrambled RNA (Silencer Negative Control #3 siRNA), and lipofectamine (RNAiMAX) were diluted with Opti-Mem prior to mixing in each well to generate a final siRNA concentration of 10 nM. Mixtures were incubated for 20 minutes at room temperature. Cells were then seeded into new wells in DMEM/F12 medium without antibiotics at a density of 5000 cells/well, incubated in the presence of transfection reagent for 24 hours at 37 °C in 5% CO2, and finally treated with compound 1 at fixed concentrations of 100 nM or 1 M (or DMF-containing media in control groups) for an additional 24 or 48 hours. Cell viability after 48 and 72 hours was assessed as described above.
In HCT-116 sensitization experiments, cells were seeded at a density of 1100 cells/well in 100 L of media and allowed to attach overnight. Medium in each well was replaced with fresh medium containing a combination of EED226 and EPZ-6438 to generate final concentrations of 2.5 and 5 M of each drug. Medium supplemented with epigenetic drugs was replaced every 24 hours and finally removed after 72 hours to begin treatment with compound 1 at concentrations of 1 M and 10 M for an additional 72 hours. Cell viability after 72 hours was assessed as described above.
Immunoblotting. Cells were lysed in RIPA buffer (25 mM   NaCl and 0.05% Tween 20, adjusted to pH 7.6, 5% non-fat milk) at room temperature for 1 hour, (ii) incubated with primary anti-MATE1 antibody or GAPDH antibody in TBST buffer (2% non-fat milk) at 4 °C overnight, (iii) washed 4 times for 5 minutes with TBST buffer and incubated with goat-anti-rabbit IgG-HRP secondary antibody in TBST buffer (2% non-fat milk) at room temperature for 1 hour, (iv) washed with TBST 4 times for 5 minutes, and (v) finally incubated with SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo, 34580) at room temperature for 5 minutes. The protein bands were visualized alongside pre-stained protein ladder (PageRuler, Thermo Fisher) using an Amersham Imager 600 (GE Healthcare). Band intensities were integrated using Image J (version 1.52a, National Institutes of Health, Bethesda, MD).
To generate sufficient quantities of cell-free extract for Western blot analysis accompanying knockdown experiments, A549 cells were seeded at a density of 150,000 cells per well on 6-well plates, and transfections were performed with an optimized siRNA concentration of 2.5 nM. Likewise, to quantify hMATE1 (SLC47A1) in epigenetic sensitization assays, 100,000 HCT-116 cells were seeded into 60-mm dishes and treated with a 2.5 M or 5 M mixture of EPZ-6438 and EED226. Cell lysates were generated in both cases as described above.  (Table S1). In six of these cell lines (incl. 3 NSCLC), compound 1 resulted in 50% growth inhibition at single-digit nanomolar concentrations (logGI50 < -8) (Table S1). Compound 1 showed approximately two orders of magnitude higher activity across the entire spectrum of cell lines than cisplatin, which results in an average growth inhibition similar to that achieved by doxorubicin and topotecan, two oncology drugs also acting through DNA damage-mediated mechanisms (Figure 1b).

Compound 1 Shows High Potency and a Unique Activity
While the two topoisomerase poisons kill cancer cells at similar inhibitory concentrations as compound 1, they do not show the cell line-specific cytotoxic enhancement of our hybrid agent, which is most notable in NSCLC. Of the four agents in comparison, compound 1 shows the widest range of activity from lownanomolar to micromolar GI50 values with a more than 2000-fold difference between the most sensitive and the most resistant cell lines (logGI50 > 3.3, Figure S1).
We then used the COMPARE algorithm in conjunction with Pearson correlation analysis 17 to search the NCI database for test compounds that resulted in NCI-60 activity patterns similar to that of compound 1. The results demonstrate that the mechanism of compound 1 is unique among DNA-targeted cytotoxic drugs and other classes of cancer chemotherapeutics (R < 0.5) ( Table 1). Of the approved oncology drugs tested in NCI-60, transcription inhibitors and topoisomerase poisons revealed the highest similarity with compound 1. Importantly, cisplatin and oxaliplatin were among the drugs that showed the lowest level of correlation. These results suggest that our hybrid molecule and the traditional platinumbased drugs may not share any relevant mechanistic features except their proven ability to form adducts with nuclear DNA. This raises the question as to whether the unique activity profile of compound 1 might be associated with specific molecular targets or gene expression patterns in cancer cells.  Table S3.  Mean centered -log 10 (GI 50 )

Transcript intensity (Z Scores)
Chemosensitivity (NCI-60) hMATE1 (SLC47A1) Expression COMPARE analysis yielded 806 unique genes correlated positively, and 849 genes correlated negatively (p < 0.05) with the growth inhibition of compound 1 (GI50 endpoint) across the entire range of cell lines (Table S2). The by far strongest positive correlation (R = 0.69, p < 10 -5 ) was observed with the gene SLC47A1, which encodes a member of the solute carrier (SLC) family of proteins: human multidrug and toxin extrusion protein 1, hMATE1. hMATE1, a 13-helix transmembrane protein, 20 shows high expression levels in normal liver and renal tissue ( Figure S2), where it serves as a proton-coupled antiporter. 21 Its primary function is the ATP-independent efflux of organic cations across apical membranes into the bile and urine, which renders hMATE1 an essential modulator of drug response, drug toxicity, and drug-drug interactions. 22 Aberrantly high expression of hMATE1 is also observed in cancerous tissues ( Figure S2).
The above analysis is consistent with a mechanism by which MATE promotes the uptake of compound 1 into cancer cells rather than acting as an efflux pump, which would cause a more resistant phenotype and would have resulted in a negative correlation. A comparison of the NCI-60 screening results for compound 1 with the SLC47A1 expression profile (Figure 1c) supports the findings of the COMPARE analysis and illustrates the extent to which the transport protein dominates chemosensitivity.
With a few exceptions, cell lines showing high levels of SLC47A1 transcript are generally exquisitely sensitive to compound 1, while the opposite is true for cell lines expressing low levels ( Figure 1c).
Compound 1 performs poorly relative to other DNA-targeted drugs (e. g., doxorubicin and topotecan, Figure S1) across all leukemia cell lines, which invariably show low SLC47A1 expression. In cell lines representing solid tumors, considerable cell line-dependent variability exists. For instance, in the two prostate cancer cell lines tested, PC-3 (GI50  5 M, low SLC47A1 expression) and DU-145 (GI50 < 10 nM, high SLC47A1 expression), compound 1 shows a more than 500-fold difference in growth inhibition, which is not observed for any other oncology drug in NCI-60. Likewise, the renal carcinoma cell line, SN12C, which shows the highest level of SLC47A1 expression of all NCI-60 cell lines, most likely due to a gene copy number amplification 23 ( Figure S3), was also the most sensitive to compound 1.
SLC47A1 is not the only solute carrier gene whose expression showed a positive correlation with growth inhibition in NCI-60, but only SLC47A1 correlated at such a high level (p < 10 -5 vs. p < 0.01 for all other SLC genes; see Figure 1d and Table S3), suggesting a specific and dominant role of this transporter in the mechanism of compound 1. When calculating overlaps between the > 800 genes that were positively correlated with the activity of compound 1 and gene ontology (GO) gene sets deposited in the Molecular Signatures Database (MSigDB 14 ), GO terms such as plasma membrane function and components, and intracellular transport ranked highest (Table S4). This is in stark contrast to doxorubicin and topotecan, which showed the greatest overlap with GO sets annotated chromatin, DNA damage recognition and repair, and chromosome organization (data not shown), as would be expected for a genotoxic agent. 9 These observations underpin the notion that, contrary to our expectation, the chemosensitivity of cancer cells to compound 1 is not controlled at the genome level, but by the transportome.

Pyrimethamine, a Selective hMATE1 Inhibitor, Effectively Blocks Cellular Accumulation of
Compound 1 and Quenches Its Cytotoxicity in A549 Cells. To validate hMATE1 protein as a mediator of chemosensitivity, we first performed a transporter inhibition assay in A549 human lung adenocarcinoma cells. A549 expresses high levels of hMATE1 (SLC47A1) (The Human Genome Database; see Figure S2), which we confirmed by Western blot analysis ( Figure S4). Unsurprisingly, the cell line proved to be highly sensitive to compound 1 in the NCI-60 screen (GI50 < 10 nM) and in previous colorimetric cell proliferation assays (IC50 = 3.9 nM). 4 In this assay, prior to treatment with compound 1, cultured A549 cells were pre-treated with the antimalarial drug pyrimethamine (PM, Figure   2a), a potent and selective inhibitor of hMATE1 (reported Ki values: 77-93 nM 24 ). Since the assay required co-incubation of compound 1 and PM, we first confirmed that no undesired reactivity exists between the two agents (Supporting Information). When A549 cells were pre-treated with PM, followed by a 4-hour exposure to compound 1, confocal microscopy images showed a reduction of intracellular acridine fluorescence by 60% relative to cells not treated with PM (Figure 2b,c). These results suggest that hMATE1-mediated transport across the plasma membrane is directly involved in the cellular uptake of compound 1. Because the microscopy experiments were performed at relatively high concentrations of platinum-acridine and PM (10 M), contributions from non-specific transport by other membrane proteins cannot be ruled out under these conditions. 25 To overcome this drawback, we took advantage of the parts-per-trillion-level limit of detection of inductively coupled plasma mass spectrometry (ICP-MS) and also quantified uptake of compound 1 from cellular platinum levels under therapeutically more relevant conditions. When cells were pre-incubated with 100 nM PM to avoid non-specific inhibition of other organic cation transporters and subsequently treated with 100 nM compound 1, corresponding to the compound's IC90 value in A549, a decrease of uptake by 85% was observed ( Figure 2d). Together, these findings corroborate that compound 1 is selectively transported across the plasma membrane by hMATE1.
To determine if blocking hMATE1 by PM had an effect on the cytotoxicity of compound 1 in A549 cells, we performed a colorimetric cell proliferation assay ( Figure 2e). Exposure to 100 nM compound 1 for 72 hours causes severe cell death with less than 10% of the cells surviving treatment.
When A549 cells were pre-treated with PM at concentrations that did not compromise cell viability, a pronounced cytoprotective effect was observed. PM at a concentration of 10 nM was able to significantly  . Such an assay is complicated by the non-trivial task of combining transient gene silencing with a long-term cell proliferation assay. Using transfection of appropriate siRNAs, we were able to generate a A549 model in which hMATE1 was transiently reduced by 40-50 % relative to scrambled control, which is consistent with reported knockdown efficiencies achieved for the SLC47A1 gene in this cells line using RNAi. 26 Knockdown was confirmed by Western blot analysis and immunofluorescence intensity evaluation of transfected cells (Figure 3a,b). The cellular uptake of compound 1 was studied under the same conditions as in the transporter inhibition assay using PM. In hMATE1 knockdown cells, accumulation of platinum was significantly (p = 0.0091) reduced by 50% relative to control cells transfected with a scrambled RNA sequence (Figure 3c). We then designed a 96-well plate assay that allowed us to assess the performance of compound 1 in A549 cells after hMATE1 knockdown. After 24 hours of continuous treatment, the dose-and time-dependent cytotoxicity of compound 1 was reduced in A549 cells at concentrations of 100 nM and 1 M by 12% and 35%, respectively. At the higher concentration, the level of protection persists after 48 hours of treatment, which resulted in a 36% higher survival of hMATE1-silenced cells compared to mock-treated cells.
These results unequivocally confirm that hMATE1 protein plays a direct role in the mechanism of compound 1 by mediating its cellular uptake, which ultimately controls the chemosensitivity of the lung cancer cell line. The assay was performed several times under slightly varied conditions with similar results (see Figure   S5); p < 0.01, **). d) Effect of hMATE1 knockdown on the cytotoxicity of compound 1 in A549 cells assessed by a cell proliferation assay (MTS). Data are the mean ± S.E.M of two independent experiments performed in triplicate (n = 6; the results were significant at p < 0.05 (*) and p < 0.001 (***), respectively; two-tailed t-test). For additional data and replicates, see Figure S5.

Transcriptomics and Gene Set Overlap Analysis Suggest that hMATE1 Expression is Epigenetically Regulated in Many Types of Cancer. Significant correlations exist between SLC47A1
transcript levels and DNA methylation status (CpG islands, CGI) of the gene (p < 0.001), as well as correlations involving epigenetic repressors of gene expression, such as DNA methyltransferase I (DNMT1) and the histone methyltransferase, enhancer of zeste homolog 2 (EZH2) (Table S5). Thus, in addition to DNA copy number amplifications (Table S3) Table S6). A recent study demonstrates that hMATE1 expression in normal liver tissue is attenuated epigenetically by promoter hypermethylation, 28 which supports the above observations.
We also discovered a link between genes whose methylation status is negatively correlated with SLC47A1 transcript levels in NCI-60 (CellMiner), including SLC47A1 itself, and specific gene sets in the Molecular Signatures Database (MSigDB, gsea-msigdb.org) (Table S7). Hypergeometric distribution analysis 14 of our list of genes (Table S8) showed the highest correlation with genes epigenetically silenced in embryonic stem cells. 29 The process involves EZH2-mediated histone protein H3 trimethylation at lysine 27 (H3K27me3) by the polycomb repressive complex 2 (PRC2) and downstream promoter CGI hypermethylation. These data provide additional clues about hMATE1 regulation at the epigenome level and a potential link between hMATE1 expression, cancer stemness, and drug resistance. Taken together, these observations led us to hypothesize that epigenetic drugs reversing the repression of hMATE1 might increase the uptake of compound 1 and sensitize resistant cancer cells to this agent. HCT-116 cells show low hMATE1 expression caused by repressive modifications in its SLC47A1

Treatment of HCT-116 Colon Cancer Cells with Epigenetic Drugs Activates hMATE1 Expression
promoter region (see Figure S6) and proved to be relatively resistant to compound 1 in NCI-60 ( Figure   1c, Figure S1).
We first pre-screened several epigenetic drugs in cultured HCT-116 cells in a multi-well plate format for their ability to increase the uptake of compound 1 using fluorescence microscopy (

DISCUSSION
At physiological pH, compound 1 and its derivatives exist as 2+ charged, hydrophilic cations comprising a positively charged platinum(II) moiety and a protonated 9-aminoacridine chromophore (pKa = 9-10). 16 In earlier work, we have demonstrated that the most potent platinum-acridines accumulate in NSCLC cells at a 60-100-fold faster rate than cisplatin, 38 which is consistent with the efficient, SLC transportermediated uptake mechanism established in this study. Compound 1 is the first chemotherapeutic agent for which bioinformatics and high-throughput screening tools have identified an overexpressed transport protein as a target that confers a high level of chemosensitivity to cancer cells.
Compound 1 has emerged from a pipeline of platinum-acridine agents that were designed based on the guiding principle that rapid formation of unique DNA adducts would overcome tumor resistance to DNA-targeted drugs, including platinum-based pharmaceuticals. While DNA damage indisputably is the ultimate cause of cancer cell death produced by the hybrid agent, its low-nanomolar activity critically depends on a transport protein, which is an unprecedented feature among anticancer drugs in the NCI-60 database. hMATE1 controls the pattern of activity with a high level of predictability. Cancer cells overexpressing the membrane transporter are highly sensitive to compound 1 regardless of genetic background and phenotypic abnormalities. 39,40 Efficient transmembrane transport that leads to high intracellular drug concentrations has the potential to overcome common resistance mechanisms such as DNA repair 41 or multidrug resistance-mediated drug efflux. 42 hMATE1 expression is high in most NSCLC cell lines (Fig. 1c), which explains why the advantage of platinum-acridines over cisplatin and other cytotoxic agents was first noted in this aggressive type of cancer. 5 Membrane transporters that help drugs accumulate in diseased tissue may ultimately result in a more favorable therapeutic window for systemic treatment. 43 Compound 1 has already demonstrated efficacy in xenograft models of A549 in mice when administered intravenously, both directly and as liposomal formulation. 44 Using a non-optimized dosing schedule, the agent was able to reduce tumor growth by 65% with less than 20% weight loss in test animals, which was reversible, without causing other signs of systemic toxicity. It is possible that hMATE1-enhanced uptake into tumors contributes to the efficacy of compound 1 in vivo.
A few cases have been reported of membrane transporters typically involved in drug elimination that may also enhance drug uptake into tumor tissue. Organic cation transporters (hOCT, SLC22A) are an example of such a dual pharmacokinetic role. 45 hOCTs have been shown to enhance the cytotoxicity and efficacy of platinum-containing drugs. 43,46 For instance, in colorectal cancer tissue, high levels of hOCT assist in the cellular uptake of oxaliplatin, which has provided a rationale for the drug's therapeutic use in this form of cancer. 47 hMATE1 protein, which mediates efflux of substrate from polarized epithelial cells in excretory organs, may play a similar role by transporting substrates across the plasma membrane into cells. 21 This has recently been demonstrated for the clinical kinase inhibitor imatinib (Gleevec) in chronic myeloid leukemia (CML) cells, which enhances the drug's potency in this hematological cancer. 48 Importantly, in the same study hMATE1 expression levels have been validated as a predictor of interindividual differences in imatinib response and clinical outcome in CML patients. 48 These findings corroborate the critical role solute carrier (SLC) transporters may play in mediating delivery of pharmacologically relevant levels of drug to diseased tissue. 49 Finally, we provide proof-of-concept data to demonstrate that colorectal cancer cells treated with epigenetic drugs can be sensitized to compound 1 and that the enhanced cytotoxicity is caused by hMATE1-mediated drug accumulation. A growing body of clinical evidence supports the utility of coadministering cytotoxic drugs with epigenetic drugs (see also clinicaltrials.gov). Liu et al. 50 recently demonstrated that renal cell carcinoma (RCC) cells can be sensitized to oxaliplatin by pre-treatment with the hypomethylating agent decitabine, which promotes hOCT2 expression and oxaliplatin accumulation.
Another compelling case of epigenetic sensitization has been reported by Gardner et al. 51 for the Schlafen-11 protein (SLFN11), a putative RNA/DNA helicase that acts as a sensor of replicative stress and tumor suppressor. 52 In patient-derived small-cell lung cancer (SCLC) tissue, Schlafen-11, which sensitizes cancer cells to topoisomerase I poisons, was epigenetically silenced. 51 Treatment with epigenetic drugs restores Schlafen-11 levels, which reverses resistance in SCLC and re-sensitizes cells to the drug topotecan. 51 There also appears to be an epigenetic component to hMATE1 (SLC47A1) expression in SCLC 53 (sclccelllines.cancer.gov). Since topotecan is a substrate of hMATE1, 54 the reported level of sensitization to the topoisomerase I poison in SCLC cell lines after treatment with EPZ-6438 51 may also reflect higher drug accumulation due to increased levels of hMATE1. Using compound 1 as a cytotoxic component in similar combination regimens to treat SCLC and other cancers not responding optimally to our hybrid agent (e.g., leukemias, colorectal cancer, ovarian cancer, see Figure   1), would be an attractive opportunity.

CONCLUSION
In summary, the current study provides the mechanistic basis for the unique spectrum of anticancer activity of a platinum-acridine hybrid agent, compound 1. The data demonstrates that the fate of a cancer  Table of Contents  page   Table S1. Ten NCI-60 cell lines most sensitive to compound 1 S3 Figure S1. Comparison of NCI-60 chemosensitivity profiles S4 Table S2. Results of NCI COMPARE analysis S5 Table S3. NCI-60/COMPARE analysis S27 Figure S2. Expression of hMATE1 (SLC47A1) in normal human tissue and in cancer cells S28 Figure S3. SLC47A1 gene copy numbers and transcript levels in NCI-60 S29 Table S4. Summary of top 10 overlaps  Table S5. Pattern comparisons for SLC47A1 expression in NCI CellMiner S34 Figure S6. Correlation between CPI methylation status and expression levels of SLC47A1 S35 Table S6. Summary of significant (p < 0.05) correlations identified between CPI methylation status and expression levels of the SLC47A1 gene S36 Table S7. Summary of correlations for chemosensitivity and omics data for compound 1 S37 Table S8.   Figure S1. Comparison of NCI-60 chemosensitivity profiles (averages of at least 2 assays) for cisplatin (CDDP, NSC 119875), doxorubicin (DOX, NSC 123127), topotecan (TOP, NSC 609699), and compound 1 (NSC # not disclosed). -0.328 organic anion transporter * a N = 58. b * P < 0.05; ** P < 0.01; *** P < 0.001; ***** P < 0.00001.      Table S7. Summary of correlations observed for chemosensitivity and omics data for compound 1.