Proteomic profiling reveals that collismycin A is an iron chelator

Collismycin A (CMA), a microbial product, has anti-proliferative activity against cancer cells, but the mechanism of its action remains unknown. Here, we report the identification of the molecular target of CMA by ChemProteoBase, a proteome-based approach for drug target identification. ChemProteoBase profiling showed that CMA is closely clustered with di-2-pyridylketone 4,4-dimethyl-3-thiosemicarbazone, an iron chelator. CMA bound to both Fe(II) and Fe(III) ions and formed a 2:1 chelator-iron complex with a redox-inactive center. CMA-induced cell growth inhibition was completely canceled by Fe(II) and Fe(III) ions, but not by other metal ions such as Zn(II) or Cu(II). Proteomic and transcriptomic analyses showed that CMA affects the glycolytic pathway due to the accumulation of HIF-1α. These results suggest that CMA acts as a specific iron chelator, leading to the inhibition of cancer cell growth.

Our data indicate that CMA acts as a specific iron chelator in cells, as predicted by ChemProteoBase profiling. CMA binds to both Fe(II) and Fe(III) ions and forms 2:1 chelator-iron complex that inactivates the iron ion, resulting in the inhibition of cancer cell growth.

CMA inhibits cancer cell growth and causes G1 cell cycle arrest.
We first examined the growth inhibitory effects of CMA (Fig. 1a) against human cancer cell lines, and found that CMA inhibits their growth with IC 50 values ranging from 0.1 to 0.4 μ M for 72 h (Table 1). When HeLa cells or HL-60 cells were treated with CMA for 12 h, G1 phase population increased significantly (Fig. 1b). This effect was reversible as the cell cycle arrest was canceled by the depletion of CMA from culture media ( Supplementary Fig. S1). Western blot analysis after incubation with CMA demonstrated that the expression of cyclin D1 was markedly decreased in a time-dependent manner (Fig. 1c). The expression rates of cyclin D1 after treatment of CMA decreased to 44.4% (6 h, P < 0.001), 12.2% (12 h, P < 0.001), 7.7% (18 h, P < 0.001), and 7.1% (24 h, P < 0.001) ( Supplementary Fig. S2a). CDK4 was also decreased after 12 h treatment of CMA (Fig. 1c). The expression rates of CDK4 after treatment of CMA decreased to 67.3% (12 h, P < 0.001), 38.3% (18 h, P < 0.001), and 30.1% (24 h, P < 0.001) ( Supplementary Fig. S2b). On the other hand, the expression of p27 Kip1 was not changed significantly ( Fig. 1c and Supplementary Fig. S2c). Thus, CMA leads to cell cycle arrest at the G1 phase with a marked decrease of cyclin D1.
Target prediction of CMA by ChemProteoBase profiling. To identify the molecular target of CMA, we performed ChemProteoBase profiling. HeLa cells were treated with 5 μ M CMA for 18 h, and the resulting cell lysates were subjected to 2D-DIGE. The proteomic variation of 296 spots that matched on all gel images was quantified, followed by hierarchical cluster analysis and calculation of cosine similarities of compounds in the   (Fig. 2). Similarity analysis showed that Dp44mT ranked in the highest position among 143 compounds contained in the ChemProteoBase (cosine similarity = 0.71, Table 2). These data suggest that CMA is an iron chelator.
CMA forms 2:1 chelator-iron complex. To confirm whether CMA forms an iron complex, we performed qualitative analysis using High Performance Liquid Chromatography (HPLC). When CMA was mixed with Fe(II) or Fe(III) ions at half the concentration of CMA, the peak of CMA shifted completely (Fig. 3a). Furthermore, we carried out absorption spectroscopic analysis on the reactions of CMA with Fe(II) and Fe(III) ions, respectively. The binding of CMA to Fe(II) ion induced a new intense absorption with a maximum at 535 nm. The intensity increased linearly upon incremental concentration of Fe(II) ion and saturated at 0.5 equiv. with respect to CMA ( Supplementary Fig. S3). This result clearly demonstrates a rapid formation of Fe(II)-(CMA) 2 complex, which is identical to that obtained by HPLC analysis (Fig. 3a: left). Moreover, upon mixing of CMA and Fe(III) ion of 0.5 equiv. with respect to the chelator, a new broad absorption in the range from 360 nm to 420 nm increased immediately ( Supplementary Fig. S4). This spectral pattern, which is similar to that observed in the formation of Fe(III)-(Dp44mT) 2 complex 28 , and the similarity between Fig. 3a (left) and Fig. 3a (right) in HPLC analysis suggest that CMA is also able to bind to Fe(III) ion and form Fe(III)-(CMA) 2 complex. We next analyzed the reactivity of Fe(II)-(CMA) 2 complex to O 2 and H 2 O 2 , respectively, by using absorption and ESI-mass spectroscopic measurements. The intense absorption at 535 nm characteristic of the ferrous

CMA acts as a specific iron chelator in cells.
If CMA acts as an iron chelator and causes the inhibition of cell growth, the addition of exogenous iron ion should abolish the biological activity of CMA. As shown in Fig. 4a, CMA-induced cell growth inhibition was completely canceled by Fe(II) and Fe(III) ions, but not by other metal ions, including Zn(II), Mn(II), Cu(II), and Mg(II). To confirm whether CMA binds to Cu or Zn ions to inhibit cancer cell growth, we assessed the cytotoxic activity of CMA in the presence of TM (tetrathiomolybdate) as a Cu chelator 29 or TPEN [N,N,N′ ,N′ -tetrakis(2-pyridylmethyl)ethylenediamine] as a Zn chelator 30 . As shown in Supplementary Fig. S5, the activity of CMA was not reduced at all even if the concentrations of TM and TPEN were higher than that of CMA. Iron is involved in many cellular processes, including DNA synthesis, cell cycle regulation, energy generation, ROS generation, and hypoxia-inducible factor-1 (HIF-1) and WNT pathways 20 . When HL-60 cells were treated with 1 μ M CMA for 1.5 h, endogenous ROS level was significantly decreased (Fig. 4b). In addition, pretreatment with 1 μ M CMA also inhibited increased ROS level induced by H 2 O 2 or antimycin A, an inhibitor of mitochondrial electron transport (Fig. 4b). Furthermore, CMA as well as DFO, a known iron chelator, markedly increased the expression of HIF-1α protein ( Fig. 4c and Supplementary Fig. S6). The expression levels of HIF-1α after treatment of CMA increased 10.9-fold (6 h, P < 0.01), 15.8-fold (12 h, P < 0.001), 15.7-fold (18 h, P < 0.001), and 12.3-fold (24 h, P < 0.001) ( Supplementary Fig. S7a). The expression level of HIF-1α after treatment of DFO increased 13.4-fold (24 h, P < 0.001) ( Supplementary Fig. S7a). CMA as well as DFO significantly decreased the expression of ferritin, which is an intracellular iron storage protein (Fig. 4c). The expression rates of ferritin light chain after treatment of CMA decreased to 69.3% (12 h, P < 0.05), 32.7% (18 h, P < 0.001), and 12.6% (24 h, P < 0.001) (Supplementary Fig. S7b). The expression rate of ferritin light chain after treatment of DFO decreased to 11.0% (24 h, P < 0.001) (Supplementary Fig. S7b). The expression rates of ferritin heavy chain after treatment of CMA decreased to 74.3% (6 h, P < 0.001), 75.6% (12 h, P < 0.001), 54.5% (18 h, P < 0.001), and 46.9% (24 h, P < 0.001) (Supplementary Fig. S7c). The expression rate of ferritin heavy chain after treatment of DFO decreased to 39.0% (24 h, P < 0.001) (Supplementary Fig. S7c). These results suggest that CMA acts as a specific iron chelator, resulting in the inhibition of cancer cell growth. Proteomic and transcriptomic analyses of CMA-treated cells. Our ChemProteoBase predicts drug targets by comparing just the spot patterns without requiring any protein information. However, we have identified about 90% of the protein spots used in the profiling so far. Such protein information enables the investigation of the mode of action of the compound on a proteomic level. Keyword based analysis of CMA-induced proteomic changes showed that metabolic pathway-especially glycolysis and oxidative phosphorylation-related proteins and oxidative stress-related proteins underwent changes ( Supplementary Fig. S8). In particular, most of glycolysis-related proteins were upregulated and oxidative phosphorylation-related proteins were downregulated ( Supplementary Fig. S8). As shown in Table 3, glycolysis-related proteins, such as fructose-bisphophate aldolase A (ALDOA), phosphoglycerate kinase 1 (PGK1), phosphoglycerate mutase 1 (PGAM1), and triosephosphate isomerase (TPI1), were significantly upregulated in HeLa cells treated with CMA as well as Dp44mT. We further performed transcriptome analysis and found that all these proteins were upregulated at the mRNA level ( Table 3), suggesting that these proteomic changes are mainly regulated by gene expression but not by protein modification. Also, iron-related proteins, such as procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2 (PLOD2) and prolyl 4-hydroxylase subunit α -1 (P4HA1), were markedly upregulated in HeLa cells treated with CMA as well as Dp44mT (Table 3).

Discussion
In this study, we showed that CMA acts as an iron chelator, resulting in the inhibition of cancer cell growth. CMA binds to both Fe(II) and Fe(III) ions and forms 2:1 chelator-iron complex (Fig. 3, Supplementary Fig. S3, and Supplementary Fig. S4). Considering the structure of CMA (Fig. 1a) and the property of the iron ion, the complex may have an octahedral configuration with the iron ion in the middle, in which three nitrogen atoms from pyridine and oxime groups of CMA could be ligands. CMA is able to capture an Fe(II) ion instantly to form a stable Fe(II) complex with extremely low reactivity to O 2 and H 2 O 2 (Fig. 3b,c). Therefore, the observed inhibition of the growth of various cancer cells (Table 1) could be due to CMA lowering the total concentration of iron ions below what is necessary for cell proliferation and cell cycle and preventing the generation of ROS through reaction with O 2 or H 2 O 2 . In fact, CMA decreased intracellular ROS level (Fig. 4b), suggesting that CMA inhibits the Fenton reaction, in which Fe(II) ion reacts with hydrogen peroxide to produce the hydroxyl radical, the most reactive ROS. Thus, CMA is a chelator that inactivates the iron ions.
The inhibitory effect of CMA on cell growth was completely canceled by iron ions but not by other metal ions (Fig. 4a), suggesting that CMA acts as a specific iron chelator in cells, and the CMA-iron complex itself has no biological effect, at least against cell proliferation. Because iron is involved in many cellular processes, such as  Table 3. Proteomic and transcriptomic changes of the spots that changed significantly in the proteomic analysis of CMA-treated HeLa cells. In proteomic changes, the mean ratios of identified spots between control and compound-treated cells are listed. Non-repeated-measures ANOVA and Dunnett's test for post hoc analysis were performed. In transcriptomic changes, the ratio between control and compound-treated cells are listed. If there are more than one probes corresponding to the protein, the mean ratio was calculated. *P < 0.05; **P < 0.01; ***P < 0.001.
DNA synthesis, energy metabolism, and cell cycle progression, CMA must affect various iron-related proteins and processes directly or indirectly. Because CMA arrested cell cycle at the G1 phase, the main target of CMA may be ribonucleotide reductase, which is a key enzyme in DNA synthesis for the conversion of ribonucleotides into deoxyribonucleotides, and has two iron atoms in its active site 31 . Our ChemProteoBase can not only predict the molecular targets of bioactive small molecules by comparing protein variation patterns but can also explore their biological effects by using information from each of the studied proteins. Keyword based analysis of CMA-induced proteomic changes showed that glycolysis-related proteins tend to be upregulated and oxidative phosphorylation-related proteins tend to be downregulated ( Supplementary Fig. S8). This could be due to the accumulation of HIF-1α consequent to CMA treatment ( Fig. 4c and Supplementary Fig. S7a). HIF-1α is known to upregulate glycolysis and downregulate oxidative phosphorylation to help cells survive in a low oxygen (hypoxic) environment 32,33 . In normoxic conditions, HIF-1α undergoes hydroxylation by HIF prolyl hydroxylases. The hydroxylation allows the binding of von Hippel-Lindau protein (pVHL) to HIF-1α and subsequently results in the ubiquitination of the protein that leads to proteasomal degradation 34,35 . Because HIF prolyl hydroxylases require iron as a cofactor, CMA may inhibit the activity of HIF prolyl hydroxylases, leading to the accumulation of HIF-1α . ALDOA, PGK1, PGAM1, and TPI1, as well as P4HA1, which were significantly upregulated by CMA, have been reported to be HIF-1 target genes 36,37 .
A number of iron chelators have been designed and studied as anticancer agents in various types of cancer 25,38 . Importantly, every chemical property of a chelator, including binding potency and selectivity for metal ion, redox potential, charge, solubility, lipophilicity, pH dependence, and kinetics, can influence the biological outcome 39 . For example, Dp44mT can bind to copper as well as iron 29 . Treatment with Dp44mT causes efflux of iron from cells, but has the opposite effect on copper, which accumulates within lysosomes due to their acidic pH and the ionization characteristics of Dp44mT. Dp44mT forms a redox-active copper complex that promotes ROS generation, which degrades the lysosomal membrane and subsequently leads to apoptosis 29 . Although further investigations are needed, especially to understand in more detail the mechanism of action and in vivo antitumor efficacy, CMA can be used to understand iron metabolic pathways and can be administered as an anticancer agent in cancer chemotherapy. In addition, because iron is also involved in the pathophysiology of a number of diseases including malaria, HIV AIDS, fungal infections, and neurodegenerative diseases 38 , development of new iron chelators would be advantageous in a number of clinical scenarios. In conclusion, we have demonstrated that CMA acts as an specific iron chelator that binds to both Fe(II) and Fe(III) ions, resulting in the inhibition of cancer cell growth. Our study also shows that ChemProteoBase is useful for target prediction of bioactive small molecules, especially nonprotein-targeting small molecules.
Cell proliferation assay. Cell proliferation was determined by using a Cell Count Reagent SF (Nacalai Tesque, Kyoto, Japan) according to the manufacturer's instructions. Briefly, cells were seeded in a 96-well culture plate, cultured overnight, and exposed to compound for 72 h. Then, 1/20 volume of WST-8 solution was added to each well, and the plates were incubated at 37 °C for 1 h. Cell proliferation was measured as the absorbance at 450 nm on a microplate reader (Perkin Elmer).
Cell cycle analysis. Cells were treated with compound for the indicated times. Cells were then harvested, washed with PBS and fixed in 70% ethanol. Cells were washed twice with PBS and incubated with 50 μ g/mL propidium iodide in PBS containing 2 μ g/mL RNase A (Nacalai Tesque) for 30 min. The DNA content of the cells was analyzed on a Cytomics FC500 flow cytometer (Beckman Coulter).
ChemProteoBase profiling. ChemProteoBase profiling was performed as described previously 9 . Briefly, HeLa cells were treated with 5 μ M CMA for 18 h. Proteome analysis of the cell lysate was performed by using the 2D difference gel electrophoresis (2D-DIGE) system (GE Healthcare), and images of the gels were analyzed with Progenesis SameSpots (Nonlinear Dynamics). Of more than 1000 spots detectable in each 2D gel, 296 variational spots were found to be common between gels of reference-compound-treated cells and were selected as Scientific RepoRts | 6:38385 | DOI: 10.1038/srep38385 described previously 9 . Next, the volume of each spot was normalized by using the average of the corresponding control values from DMSO-treated control cells. From the normalized volumes of the 296 spots, cosine similarity between compounds was calculated, and hierarchical clustering analysis was performed using Cluster 3.0 (clustering method; centroid linkage with the means of uncentred correlation). The predictive dendrogram was visualized using Java Treeview 1.1.3.
For absorption and ESI-mass spectroscopic measurements, Fe(II)-(CMA) 2 complex was generated by mixing CMA and FeSO 4 at a molar ratio of 2:1 in MOPS buffer (20 mM MOPS, 100 mM NaCl, pH 7.4, 5% DMSO) degassed of dissolved O 2 . The reactions of the corresponding Fe(II)-(CMA) 2 complex with O 2 and H 2 O 2 were carried out by introduction of O 2 into the solution of the ferrous complex with gentle bubbling and addition of H 2 O 2 (20 eq./Fe(II) ion) to the solution, respectively. The spectral changes were analyzed by using V630 spectrometer (Jasco). The purple powder of Fe(II)-(CMA) 2 complex was obtained from the solution under Ar atmosphere or anaerobic conditions after the absorption analysis. The formulation and purity of the isolated complex were identified by ESI mass spectrophotometer (QSTAR Elite, AB SCIEX). The sample for the ESI mass spectroscopic measurement was prepared in MeOH. MS (ESI, m/z) 303.04 [Fe(CMA) 2 ] 2+ . ROS assay. HL-60 cells were seeded in 24-well plates, cultured overnight, and treated with compounds.
Transcriptome analysis. HeLa cells were seeded in 10-cm culture dishes, cultured overnight, and treated with 5 μ M CMA or 5 μ M Dp44mT for 18 h. Then, cells were harvested and total RNA from the cells was prepared using Isogen (Nippon Gene, Japan) according to the manufacturer's instructions. Gene chip analysis was performed using a GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix). The DNA microarray data of CMA-or Dp44mT-treated HeLa cells have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE78052.