Identification of Elongation Factor G as the Conserved Cellular Target of Argyrin B

Argyrins, produced by myxobacteria and actinomycetes, are cyclic octapeptides with antibacterial and antitumor activity. Here, we identify elongation factor G (EF-G) as the cellular target of argyrin B in bacteria, via resistant mutant selection and whole genome sequencing, biophysical binding studies and crystallography. Argyrin B binds a novel allosteric pocket in EF-G, distinct from the known EF-G inhibitor antibiotic fusidic acid, revealing a new mode of protein synthesis inhibition. In eukaryotic cells, argyrin B was found to target mitochondrial elongation factor G1 (EF-G1), the closest homologue of bacterial EF-G. By blocking mitochondrial translation, argyrin B depletes electron transport components and inhibits the growth of yeast and tumor cells. Further supporting direct inhibition of EF-G1, expression of an argyrin B-binding deficient EF-G1 L693Q variant partially rescued argyrin B-sensitivity in tumor cells. In summary, we show that argyrin B is an antibacterial and cytotoxic agent that inhibits the evolutionarily conserved target EF-G, blocking protein synthesis in bacteria and mitochondrial translation in yeast and mammalian cells.


Introduction
Natural products constitute a major resource for the identification of bioactive molecules. Indeed, most antibacterials in current use are natural products or semisynthetic derivatives thereof. The argyrins are natural peptides produced by myxobacteria and actinomycetes that have an intriguing antibacterial spectrum of activity [1,2,3]. This includes the intrinsically drug resistant organism Pseudomonas aeruginosa, but not other Gram negatives tested, such as Escherichia coli or Salmonella typhimurium, unless the cells are compromised in their outer membrane permeability barrier, presumably allowing access to the intracellular target [1,2]. It has been shown that argyrins inhibit bacterial protein synthesis [2], but the specific cellular efficacy target of these antibacterials has not been identified. Argyrins were also shown to be immunosuppressive [1,4,5] and anti-tumorigenic, with more recent investigations providing evidence that argyrin A inhibits the proteasome, induces apoptosis, and blocks angiogenesis by a p27-dependent mechanism [6,7,8,9]. To shed more light on the cellular target of this interesting class of natural products, we employed bacterial and yeast mutant selection and whole genome sequencing to identify the target of argyrin B and explored whether the mechanism of action is conserved in mammalian cells.

Argyrin B inhibits bacterial elongation factor G
We first confirmed the activity of argyrin B against P. aeruginosa PAO1 strain K767, and also observed activity against another intrinsically drug resistant Gram-negative pathogen, Burkholderia multivorans ( Table 1). P. aeruginosa mutants with decreased susceptibility to argyrin B were selected at a frequency of circa 5610 28 on solid medium containing 128 mg/ml argyrin B (minimum inhibitory concentration (MIC), 4-8 mg/ml). Sequencing the genome of one mutant identified a single mutation in the fusA1 gene, resulting in an amino acid substitution in elongation factor G (EF-G). EF-G mediates the translocation of mRNA and tRNA through the ribosome and is essential for protein synthesis [10,11]. The fusA1 gene from 5 additional mutants was then sequenced and found to contain point mutations encoding amino acid substitutions. The overall list of individual substitutions identified in EF-G were: P414S, S417L, S459F, P486S, T671A and Y683C (mutants CDR0052, CDR0054, CDR0055, CDR0056, CDR0057, CDR0058; Table 1). Argyrin B had variable solubility in MIC assays (visible precipitate above 16 mg/ ml), so a more sensitive strain was used to more accurately determine the change in susceptibility conferred by resistance mutations. P. aeruginosa strain Z61 is hypersusceptible to a wide range of antibiotics, due to mutations affecting cell permeability [12,13]. Strain Z61 plated on 2 mg/ml argyrin B (MIC 0.125-0.25 mg/ml), yielded mutants also having the S417L and S459F substitutions, and identified an additional L663Q alteration (mutants CDA0055, CDA0056, CDA0061; Table 1). While mutants with S417L and S459F substitutions were still partially sensitive, L663Q conferred a higher level of resistance to argyrin B (MIC.128 mg/ml, Table 1). This suggested that argyrin B was inhibiting bacterial growth by targeting EF-G. P. aeruginosa has two similar genes encoding elongation factor G proteins, fusA1 and fusA2 (designated here as EF-G1 or EF-G2). All mutations were found in fusA1, consistent with the observation that fusA1 is highly transcribed whereas fusA2 is not (genechip analysis, data not shown). In B. multivorans, argyrin B resistant mutants (frequency circa 10 29 -10 28 ) also had mutations in fusA1, mirroring the scenario observed for P. aeruginosa (Table 1). To exclude the possibility that selection of fusA mutations using argyrin B might reflect an indirect resistance mechanism rather than direct interaction of argyrin with this target protein, we tested if purified P. aeruginosa EF-G1 binds to argyrin B. Very tight binding was observed as measured with isothermal titration calorimetry (173 nM; stoichiometry of 1) as well as surface plasmon resonance (SPR, 176 nM). In contrast, EF-G1 S459F failed to bind argyrin B sufficiently to generate a K d value by SPR (data not shown).

Argyrin B binding to EF-G reveals a new allosteric binding site and novel mode of inhibition
Since argyrin B clearly bound to the P. aeruginosa EF-G protein encoded by fusA1, we sought to derive more detailed information regarding the target binding pocket and possible mode of inhibition. The structure of P. aeruginosa EF-G1 in complex with argyrin B was therefore determined to 2.9 Å . The overall domain structure of EF-G1 seen here was similar to that of previously reported bacterial EF-G protein structures [14,15,16,17]. Argyrin B bound at the interface of domains III and V, revealing a novel inhibitor binding site that is clearly distinct from that of the characterized EF-G inhibitor fusidic acid ( Figure 1A). Key binding interactions between argyrin B and domain III of EF-G1 are defined by hydrogen bonding interactions between the backbone amide of Ala489 and the N-methylated glycine of argyrin B, the hydroxyl group of Ser417 to the oxygen within the methoxytryptophan, as well as from the side chain of Lys448 to the glycine of argyrin B ( Figure 1B, C and D). Van der Waals interactions are also observed between the thiazole within the argyrin B structure and the backbone of Gln487 and Val488. Interactions between argyrin B and domain V of EF-G1 are defined by van der Waals interactions between the indole ring of the tryptophan portion of argyrin B and Met620 and Met685, as well as additional interactions between argyrin B to both Leu663 and Phe687 ( Figure 1B, C and D). Amino acid substitutions important for P. aeruginosa resistance are shown on the co-crystal structure and clearly line the binding pocket of argyrin B ( Figure 1B). Of particular interest, the argyrin B-bound protein displays a rotation of domains III and V relative to domains I and II as compared to previously determined structures [15,17]. A ratcheting of domain IV is observed, indicating that argyrin Bbound EF-G1 adopts a more elongated conformation than has been described previously for Thermus thermophilus EF-G in complex with GTP [15] or in complex with fusidic acid and the ribosome [17] ( Figure 1E). These structural data indicate that argyrin B directly binds to EF-G at a novel allosteric pocket causing EF-G to adopt an extended conformation that is unlikely to be compatible with ribosome binding.

Argyrin B targets mitochondrial elongation factor G in yeast
Argyrins are also active against eukaryotic cells [6,7,8,9]. To explore this further, we first determined toxic effects on yeast, which represent another genetically tractable system for mechanistic studies [18]. Argyrin B was inactive against wild-type yeast (BY4741) and a strain deleted for eight genes involved in efflux (strain CMB970, designated as BYD8) when tested in rich medium containing glucose as carbon source. However, with ethanol/ glycerol substituted as carbon source, argyrin B scored an IC 50 concentration of 4 mM on wild-type and 0.04 mM on the drug efflux-deleted strain BYD8 (Figure 2A). While glucose can be metabolized in yeast cells by anaerobic fermentation, energy generation from ethanol/glycerol is strictly dependent on oxidative phosphorylation, suggesting that argyrin B interfered with mitochondrial function. The observed IC 50 shift in the drug-efflux compromised strain also indicated that argyrin B was a substrate for drug efflux pumps in yeast. Consistent with argyrin B interference with mitochondria, incubating a strain with Cox4-GFP labeled mitochondria [19] in 36 the IC 30 concentration of argyrin B for 30 minutes followed by fluorescence microscopy analysis showed mitochondrial morphology defects (fragmentation) for cells grown on ethanol/glycerol but not glucose ( Figure 2B). Treatment with the microtubule toxin benomyl at 36 its IC 30 had no effect on mitochondria, indicating that the response to argyrin B was not a general stress response.
To identify the specific target of argyrin B in yeast, a mutagenesis strategy [20] similar to that described above for bacteria was used. On ethanol/glycerol medium, argyrin B abolished colony formation of the efflux-defective yeast strain at 0.1 mM. We therefore plated 10 7 chemically mutagenized yeast cells on this selection medium and analyzed resistant survivors by whole genome sequencing. The MEF1 gene encoding mitochondrial elongation factor 1 (mEF-G1) was a hot spot for mutations. Four individual mutations were identified resulting in four distinct amino acid changes (S473L, S514F, G517D, P667L, Figure 2C). Each mutation was introduced into a fresh BYD8 strain background, and these all showed a significantly shifted susceptibility ( Figure 2D) confirming that the mutations selected in MEF1 mediated resistance to argyrin B. The A514F/G517D double mutant (Strain 6, Figure 1C) was equally susceptible to the A514F single mutant, suggesting that G517D is likely a passenger mutation not contributing to argyrin B resistance. mEF-G1 is the closest homologue of bacterial EF-G in the yeast S. cerevisiae, which supports conservation of the argyrin B target in eukaryotic cells. While P667L represents a novel mutation, S473L and S514F correspond to P. areuginosa EF-G1 mutations S417L and S459F, respectively ( Table 2).

Argyrin B targets mitochondrial elongation factor G in mammalian cells
Given the anti-tumor activity of argyrins [6,7,8,9] we next explored the argyrin B-sensitivity of mammalian cells by testing it at multiple doses across a panel of 512 human cancer cell lines by measuring ATP levels (CellTiter Glo) after a three day incubation. Argyrin B inhibited cell viability in 18 cell lines with an IC 50 below 1 mM ( Figure 3A). Comparing the IC 50 and maximal activity profile of argyrin B to other cytotoxic agents revealed a strong correlation with another closely related natural argyrin (argyrin A), followed by mitochondrial electron transport inhibitors such as rotenone and antimycin A, but less so with other cytotoxic agents ( Figure 3B). Electron transport complexes consist of several mitochondrial-encoded components, which are translated by mitochondrial ribosomes [21]. Hence, correlation of argyrin B to mitochondrial electron transport inhibitors may be caused by inhibition of mitochondrial protein synthesis via mitochondrial elongation factor G1 (mEF-G1, encoded by human GFM1), the closest mammalian homologue to bacterial EF-G and yeast mEF-G1 ( Table 2). Supporting this, argyrin B treatment depleted mitochondrial-encoded cytochrome c oxidase subunit 2 (COX2), but did not affect the levels of nuclear-encoded succinate dehydrogenase flavoprotein subunit (SDHA) in two highly sensitive cell lines (HCT116 and RKO, Figure 3C). Furthermore, RKO cells became less sensitive to argyrin B under hypoxic culture conditions where energy metabolism switches from mitochondrial respiration to cytosolic glycolysis (data not shown). In line with mEF-G1 being the cellular target of argyrin B in mammalian cells, knock down of mEF-G1 by siRNA lowered the IC 50 of argyrin B circa 9.6-fold in HCT116 and 3.3-fold in RKO cells without altering sensitivity to the proteasomal inhibitor MG132 ( Figure 3D and E). The overall sequence homology between P. aeruginosa EF-G1 and human mEF-G1 is only 40%, but there is a striking conservation of amino acid residues shown to mediate binding to argyrin B and bacterial resistance ( Table 2). We therefore introduced corresponding changes in human mEF-G1 (S452L, S494F, L693Q) to see if these reversed argyrin Bsensitivity in mammalian cells. Stable overexpression of mEF-G1 WT did not alter the argyrin B-sensitivity in HCT116 and RKO cells, and mEF-G1 S452L and S494F showed only marginal effects. However, expression of mEF-G1 L693Q decreased susceptibility to argyrin B by 4.7-and 7-fold in HCT116 and RKO cells, respectively, without altering sensitivity to MG132 ( Figure 4A, B and C). It should be noted that the sensitive endogenous mEF-G1 is still expressed in these cells, which may decrease the magnitude of the susceptibility shift that can be mediated by the resistant mEF-G1 variants. Importantly, the rescue potential of the different mEF-G1 variants correlated well with their ability to bind argyrin B. Recombinant mEF-G1 WT, S452L and S494F still bound to argyrin B, with little binding detected with mEF-G1 L693Q ( Figure 4D).

Discussion
The data presented in this study strongly supports the notion that argyrin B inhibits bacterial protein synthesis and cell growth through binding of elongation factor G (EF-G) at a novel allosteric binding pocket. The detailed structural data elaborates a potential new mode of protein synthesis inhibition whereby EF-G (in complex with argyrin B) may be blocked from interacting with the ribosome. This may stand in contrast to the characterized inhibitor of EF-G, fusidic acid, which binds at the GDP/GTP binding pocket and interferes with the post translocation release of EF-G bound on the ribosome [22,23,24]. Supporting this mode of inhibition, argyrin B bound purified P. aeruginosa EF-G tightly, whereas fusidic acid binds EF-G in complex with the ribosome. Hence, argyrin B may constitute a new tool for the study of EF-G and ribosome function.
Argyrins also have an intriguing spectrum of antibacterial activity, which includes the Gram negative pathogen P. aeruginosa, for which a combination of membrane impermeability and active efflux mediates significant intrinsic resistance to antibacterial compounds. Although this activity is moderate (MIC = 4-8 mg/ml) it is nonetheless interesting since generally more susceptible Gram negatives such as E. coli and S. typhimurium, and even Gram positives such as Staphylococcus aureus, showed less susceptibility [2]. An S. typhimurium mutant defective in the outer membrane (increased permeability) was susceptible to argyrin, and these molecules inhibited in vitro transcription translation reactions, indicating that argyrins could inhibit those cells if there is sufficient cellular penetration. Consistent with this, we also observed substantially increased activity of argyrins against the efflux and membrane defective hypersusceptible P. aeruginosa mutant Z61. Intriguingly, we also demonstrated the previously unreported activity of argyrins against another notably intrinsically resistant Gram negative pathogen, B. multivorans (MIC = 1 mg/ml), whose EF-G is homologous to that of P. aeruginosa (Table 2). Based on these observations, the unique antibacterial spectrum of argyrins is determined in part by differences in cell penetration while other factors such as intrinsic target sequence variation likely also play a role, however this awaits further investigation.
In yeast and mammalian cells, argyrin B inhibited mitochondrial EF-G1, a close homologue of bacterial EF-G, consistent with the prokaryotic origin of mitochondria as predicted by the endosymbiosis theory [25]. The lack of inhibitory activity of argyrin B on glucose-grown yeast, which can utilize anaerobic fermentation, suggests an exclusive mitochondrial target in this context. We cannot exclude the possibility of additional targets of argyrin B in mammalian cells relating to the reported immunosuppressive activity [1,2,3,4,5], but our study identifies mEF-G1 as the major efficacy target for the suppression of mammalian cell growth. Since argyrin B-sensitive tumor lines showed an overall higher CellTiter Glow signal (data not shown), we hypothesize that susceptibility to argyrin B is dependent on carbon source utilization rates as well as the proliferative index, given that mitochondria need to divide and segregate as part of the cell division cycle. By inhibiting bacterial and mitochondrial EF-G, argyrin B blocks an important cellular machine, which is conserved in prokaryotic and eukaryotic cells, consistent with protein synthesis as a target of many natural products.

Isolation of argyrin
Argyrins A and B were isolated from culture broths of Actinoplanes sp. 86317. The strain was cultivated in a medium consisting of potato starch 2.5%, glycerol 2.0%, soybean meal 0.5%, cornsteep powder 1.5%, yeast extract 0.3% and CaCO 3 0.5% for 7 days. Two liters of culture broth were filtered and mycelia were extracted with 1 L of ethyl acetate. The extract (2.7 g) was purified by reversed phase chromatography (RP18 as stationary phase) using formic acid (0.1%) and acetonitrile amino-acid changes. (D) Validation of the identified mutations by introduction of the individual base-pair changes into wild-type cells and recording IC 50 curves in duplicates. *Except for the DMSO control, all argyrin B concentrations in the range tested resulted in full inhibiton of the wild-type strain and the IC 50 value from panel A is depicted. doi:10.1371/journal.pone.0042657.g002 Table 2. Key residues involved in resistance to argyrin B are conserved in EF-G homologues.
The amino acid sequence of P. aeruginosa EF-G1 was used to identify EF-G homologues in the indicated organisms by a BLAST search. EF-G sequences were then aligned using ClustalW, and residues conferring resistance to argyrin B are shown in bold.

Bacterial mutant selection experiments and fusA1 sequencing
The strains used in this study are listed in Table 1. For isolation of mutants with decreased susceptibility to argyrin B, cultures were grown to mid-log phase (OD 600 , approximately 0.6-0.8) in Mueller-Hinton broth, collected by centrifugation, and resuspended in fresh medium and plated on Mueller-Hinton agar containing argyrin B. Serial dilutions were also plated on Mueller-Hinton agar without compound for enumeration (colony forming units/ml, CFU/ml). Resistance frequencies were calculated as the number of CFU (mutants) on drug containing plates divided by the number of CFU plated. Whole genome sequencing was done as described previously [26]. The fusA1 gene from P. aeruginosa was generated for sequencing in two parts, using primer pairs PAEFF1/PAEFR1: GGCCATGCGTTGGCTGGTGGAC/ GTGACGTCCTTCATGCCGATC and EFF2/EFR1: GTTCA-AGAACAAGGGCGTTC/GGTGCCGACGTTGACGTGCGG. The fusA1 gene from B. multivorans was generated as a single band using BCEFGF1/BCEFGR1: CATTTCCGTTTCTAAGCGCC/ CAATCGTACCAACGTTCACGTG. Nucleotide sequencing of the products was done by Agencourt (Beverly MA).

Purification of P. aeruginosa EF-G1 and EF-G1 S459F
For wild type P. aeruginosa EF-G1 expression, the fusA gene was amplified from genomic DNA of P. aeruginosa PAO1 K767 using the primer pair EFG-F/EFG-R: AGCCATATGGCCCGTACTA-CACCCATCAACC/AAAAAGCTTATCAACCTTGTTTTTT-AACCAGCGC. The product was digested with NdeI and HindIII and ligated into similarly digested pET28a to generate pET28a-PaEFG-Nhis. This plasmid was then used for site directed mutagenesis to introduce the nucleotide change encoding EF-G1 S459F (pET28a-PaEFGS459F-Nhis). The plasmids were then transformed into an E. coli strain BL21 (DE3) for expression. A single colony was used to inoculate 10 ml LB broth containing 50 mg/ml kanamycin which was incubated at 37uC overnight, with shaking at 250 rpm. The overnight culture was then diluted 100-fold into 1 liter fresh LB broth containing 50 mg/ml kanamycin. After growing at 37uC with vigorous shaking for about 2.5 hours, expression of P. aeruginosa EF-G1 or EF-G1 S459F was then induced by adding 0.1 mM IPTG, with further culture incubation at 18uC for 24 hours. Cells were then pelleted by centrifugation at 5,000 rpm with Sorvall SLC-6000 rotor and stored at 220uC. Cell pellets were then resuspended in 20 ml of BugBuster solution (Novagen) and incubated at room temperature for 20 minutes. The cell lysate was centrifuged at 20,000 rpm in a Sorvall SA-300 rotor for 20 min at 4uC. The supernatant was then loaded onto a Ni-NTA agarose column with bed volume of 2.5 mL. The mixture in the column was then rotated on a platform mixer at 4uC for an hour. The column was washed with 10 column volumes (CV) of wash buffer (100 mM Tris-HCl pH 8, 200 mM NaCl, 10 mM imidazole), followed by 5 CV of wash buffer containing 40 mM imidazole. The bound his-tagged P. aeruginosa EF-G1 proteins were eluted with 5 CV of buffer containing 500 mM imidazole. The eluent was concentrated to about 2 ml using an Ultracel-30K centrifugal filter unit with a molecular weight cutoff of 30 kDa (Amicon). The resulting protein solution was centrifuged at 13,000 rpm at 4uC with a bench top centrifuge, and then filtered with an Anotop 10 filter to remove any possible large aggregates before being injected onto a HiLoad 16/60 Superdex 200 SEC column and purified by AKTA FPLC system. The chromatography was conducted using running buffer containing 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 50 mM NH 4 Cl, 10% Glycerol at a flow rate of 1 mL/min. Fractions of 2.5 mL each were collected. The protein in the fractions was monitored by UV absorbance at 280 nm. Fractions containing P. aeruginosa EF-G1 protein were pooled and stored at 280uC. Protein concentration was determined by Nano-drop spectrophotometer. The identity of P. aeruginosa EF-G1 proteins was confirmed by SDS-PAGE and molecular weight determination using an LC-ESI-MS system (Agilent 1100/Waters ZQ4000). Isothermal titration calorimetry (ITC) determinations For isothermal titration calorimetry, P. aeruginosa EF-G1 was dialyzed against 50 mM Tris-HCl pH 7.6, 10 mM MgCl 2 , 1 mM TCEP, 10 mM NH 4 Cl, 100 mM GDP, in a 3 mL dialysis cassette (10K MWCO) at 4uC overnight. DMSO and buffer were used to adjust to 44 mM EF-G1 and 2% DMSO in 0.5 mL. Argyrin B was diluted to a final concentration of 2.5 mM (2.0 mL) with the dialysis buffer and DMSO (2% final concentration). Both protein and compound solutions were degassed and loaded onto the VP-ITC titration calorimetry instrument according to the manufacturer's instruction. After initial injection of 2 mL, P. aeruginosa EF-G1 was injected into the cell containing argyrin B at the volume of 10 mL per injection with interval of 240 seconds. The stirring rate was 307 rpm and the titration temperature was 25uC. The data was analyzed using nonlinear least-squares curve fitting in MicroCal Origin Version 7.0. The standard one binding site model was used to obtain the thermodynamic parameters K (binding constant), enthalpy of binding (DH), entropy of binding (DS), and n, where n is the ratio of argyrin B to P. aeruginosa EF-G1 in the complex. Crystallography of the EF-G argyrin complex BL21 (DE3) cells harboring pET28a-PaEFG-Nhis were grown in Terrific Broth containing 100 mg/ml kanamycin. EF-G1 expression was induced using 1 mM IPTG. Cell pellets were resuspended and lysed using a microfluidizer, after which ultracentrifugation was performed. Purification was done using NiNTA resin followed by removal of the N-terminal Histidine tag with thrombin protease. The protein was further purified using a Mono Q column and was then concentrated and loaded onto a HiLoad Superdex 200 26/60 column, exchanging the protein into 25 mM Tris-HCl pH 7.5, 150 mM NaCl and 1 mM TCEP. For crystallization of the protein with argyrin B, 2 mg/ml P. aeroginosa EF-G1 was incubated with 200 mM argyrin B for 1 hour on ice followed by concentration of the protein to 11 mg/ml. Sitting drop vapor diffusion method was used for crystallization, with the crystallization well containing 100 mM Tris-HCl pH 7.5, 18% PEG 3350 and 200 mM sodium nitrate, and the drop containing a 1:1 volume of protein and crystallization solution. After crystals formed they were subsequently cryoprotected using the crystallization solution with the addition of 20% glycerol, followed by flash freezing directly into liquid nitrogen.

Data collection and structure determination
Diffraction data for the P. aeroginosa EF-G1/argyrin B complex were collected on a Dectris Pilatus 6M Detector at the Advanced Photon Source beamline 17-ID at a wavelength of 1 Å . The data were measured from a single crystal maintained at 100uK, and the reflections were indexed, integrated, and scaled using autoPROC [27,28]. The spacegroup of the complex is C2 with 1 molecule in the asymmetric unit. The structure was determined with PHASER [29] using molecular replacement methods with a starting model of Thermus thermophilus EF-G (PDB Code 2BM0) [30]. The starting model was broken into three search models, which included Domains I and II, Domain IV, and Domain V while Domain III was fit manually. Structure determination was achieved through iterative rounds of positional and simulated annealing refinement using BUSTER (BUSTER, version 2.8.0. Cambridge, United Kingdom: Global Phasing Ltd.), with model building using COOT [31]. Individual B-factors were refined using an overall anisotropic B-factor refinement along with bulk solvent correction. The argyrin B as well as the solvent molecules were built into the density in later rounds of the refinement. Data collection and refinement statistics are shown in Table 3. The structure of P. aeruginosa EF-G1 in complex with argyrin B contains protein residues Met1-Leu40, Thr62-Asn194, Lys198-Arg408, Phe411-Pro422, Leu435-Val479, Ile483-Gln705, 1 molecule of Argyrin B and 21 solvent molecules. The coordinates and structure factors have been deposited in the Protein Data Bank (PDB ID Code: 4FN5).

Yeast susceptibility determination
Yeast strains used in this study are BY4741 [32] and strain BYD8, derived from BY4741 but deleted for eight genes involved in drug resistance (efflux pumps: snq2, pdr5, yor1; transcription factors: pdr1, pdr2, pdr3, yap1, yrm1). Argyrin potency was determined as follows: 11 point serial dilutions (3.1 dilution factor) were prepared in 96 well plates with log phase growth yeast cultures in YPD (2% glucose, 2% BactoPeptone, 1% yeast extract) and YPEG (2% glycerol, 1% ethanol, 2% BactoPeptone, 1% yeast extract), giving a compound dilution range from 200 mM to 2.44 nM and a DMSO control. Starting OD 600 was 0.1, total volume per well 120 ml. Plates were incubated at 30uC with 770 RPM orbital shaking. The 16 h time point in YPD and the 42 h time point in YPEG represented late log phase and was used to calculate IC 50 values using a logistic regression curve fit algorithm.

Selection of drug resistant S. cerevisiae cells
Strain BYD8 was incubated in 2.5% ethylmethanesulfonate until only 50% of the cells formed colonies. A total of 2610 7 mutagenized BYD8 cells were plated on two 14 cm dishes with synthetic complete medium (0.7 g/l Difco Yeast Nitrogen Base w/ o amino acids, 0.79 g/l MPbio CSM amino acid mixture, 2% Glycerol, 1% Ethanol) containing 0.1 mM Argyrin B. After 3 days resistant colonies appeared. Resistance was confirmed by restreaking on 0.1 mM Argyrin B. Genomic DNA was extracted (YeaStar Genomic DNA Extraction Kit, Zymo Research) and the genome analyzed by 80 nucleotide paired end reads on an Illumina HighSeq 2000. Analysis of the obtained sequence showed an average of 1806coverage. Amplifying the mutated MEF1 gene and transforming it into non-mutagenized BYD8 cells validated identified mutations in MEF1. Integration by homologous recombination was selected for on plates containing 0.1 mM argyrin B. Presence of the corresponding point mutations were confirmed by direct sequencing of the MEF1 gene. Resistance was confirmed by recording dose-response curves in with serial dilutions of argyrin B as described above.

Mammalian cell line profiling and compound correlation analysis
A panel of 512 human cancer cell lines (http://www. broadinstitute.org/ccle) was tested for sensitivity to various pharmacological agents, including argyrin A, argyrin B, rotenone, antimycin A, topotecan, paclitaxel, 17-AAG, staurosporin and velcade. All cell lines were maintained in humidified incubators at 37uC and 5% CO 2 in DMEM or RPMI1640 supplemented with 10% FBS. Cells were seeded into 1536-well plates at a density of 250 cells per well, allowed to adhere for 12-24 hours, then treated for 3 days with pharmacological agents in 8-point dose response (semi-log from 8 mM to 2.5 nM), and cell viability was assessed using CellTiter Glo (Promega). Automatic curve fit was used to determine IC 50 and maximal activity (A max ) values, relative to 0.4% DMSO (0%) and 1 mM MG132 (2100%) [33]. Compound correlation analysis was performed by calculating a Pearson correlation for each compound pair (using either IC 50 or A max values) across the 512 cell lines. We required data values for a minimum of 50 cell lines in common to avoid spuriously high correlations.