Validation of a High Throughput Screening Assay to Identify Small Molecules that Target the Eukaryotic Replicative Helicase

Abstract

Mcm2–7 is the catalytic core of the eukaryotic replicative helicase, which together with CDC45 and the GINS complex unwind parental DNA to generate templates for DNA polymerase.

Being a highly regulated and complex enzyme that operates via an incompletely understood multi-step mechanism, molecular probes of Mcm2–7 that interrogate specific mechanistic steps would be useful tools for research and potential future chemotherapy. Based upon a synthetic lethal approach, we previously developed a budding yeast multivariate cell-based high throughput screening (HTS) assay to identify putative Mcm inhibitors by their ability to specifically cause a growth defect in an mcm mutant relative to a wild-type strain¹. Here, as proof of concept, we used this assay to screen a 1280-member compound library (LOPAC) for potential Mcm2–7 inhibitors. Primary screening and dose-dependent retesting identified twelve compounds from this library that specifically inhibited the growth of the Mcm mutant relative to the corresponding wild-type strain (0.9 % hit rate). Secondary assays were employed to rule out non-specific DNA damaging agents, establish direct protein-ligand interaction via biophysical methods, and verify in vivo DNA replication inhibition via fluorescence activated cell sorter analysis (FACS). We identified one agent (β-carboline-3-carboxylic acid N-methylamide, CMA) that physically bound to the purified Mcm2–7 complex (Kd_app 119 μM), and at slightly higher concentrations specifically blocked S-phase cell cycle progression of the wild-type strain. In total, identification of Mcm2–7 as a CMA target validates our synthetic lethal HTS assay paradigm as a tool to identify chemical probes for the Mcm2–7 replicative helicase.

Introduction

Not all drug targets are created equal. Simple single-subunit enzymes that are amenable to easy purification and structural determination have been highly favored drug discovery targets; as their discovery assays are fast, quantitative, and operationally simple, they are ideally suited to target-based biochemical screening. However, many biologically interesting targets are incompatible with such approaches as they often function in multi-protein complexes and require binding partners and a biologically relevant context for activity. One such example is the CMG (CDC45, Mcm2–7, GINS) complex, the eukaryotic replicative helicase that unwinds duplex DNA². This helicase consists of eleven distinct and essential subunits; a catalytic “core” consisting of the six member Mcm2–7 complex, and two activator components (CDC45, and the 4 subunit GINS complex³). The heart of the CMG is the donut-shaped Mcm2–7 complex that uses its intrinsic ATPase activity to fuel conformational changes that unwind duplex DNA.

In addition to unwinding DNA Mcm2–7 is also the major focus of S-phase regulation. DNA replication is restricted to S-phase through the carefully orchestrated cell-cycle dependent assembly of Mcm2–7 onto replication origins in G1, followed by the subsequent recruitment and assembly of CDC45 and the GINS complex during G1/S to complete and activate the CMG complex⁴. Moreover, Mcm2–7 is also intimately connected to the replication checkpoint⁵, a quality control system that monitors and controls DNA replication fidelity⁶. These various functions are clearly important for human health, as mcm mutations lead to genome instability^{7, 8} Meier-Gorlin syndrome⁹, deafness¹⁰, and immunodeficiency¹¹. Although the basic cellular roles of Mcm2–7 are becoming clear, the mechanistic details that fuel these activities are poorly understood. The development of specific small molecule inhibitors to dissect specific mechanistic steps in Mcm2–7 activity is of potential interest for both research as well as potential therapeutic applications (reviewed in ^{12, 13}).

Discovery of novel CMG inhibitors has been slow due to CMG’s complexity. A few Mcm2–7 inhibitors with limited specificity have been fortuitously identified^{14, 15}. Bench scale biochemical screens to identify Mcm2–7 helicase inhibitors have been somewhat more successful¹⁶, but the most specific Mcm2–7 inhibitor identified from this screen (ciprofloxacin) has very low potency (IC₅₀ of ~600 μM) and poor solubility. Furthermore, biochemical assays for Mcm2–7 inhibitors are severely limited by the purification difficulties and limited yield of either CMG or Mcm2–7 that would be required for large scale primary screening.

To overcome these shortcomings and difficulties, we recently established a cell-based screening assay¹ combining the benefits of both target-based and phenotypic screening. This 384 well plate assay is based upon synthetic lethality¹⁷, and identifies Mcm2–7 inhibitors by their ability to reduce the growth of a specific budding yeast mcm mutant (e.g., the mcm2DENQ mutation^{5, 18, 19}) relative to the wild-type control. To maximize the growth kinetic differences between the mutant and wild-type strains we identified a positive control compound that differentially impeded the growth of the mcm2DENQ reporter strain (aphidicolin). This compound was used to implement a multivariate analysis platform (linear discriminant analysis, LDA^20–22) that consolidates three different parameters of growth as represented in the Gompertz equation²³ into a single number that fully describes the extent and kinetics of cell expansion¹.

Here, we validated this assay in a pilot screen using the LOPAC test cassette consisting of 1280 biologically annotated small molecules. We identified twelve compounds that selectively inhibited the mutant yeast strain, of which five confirmed in dose-response (0.4% final hit rate). After a series of secondary assays to eliminate non-specific toxins and DNA damaging agents, we found one “hit”, β-carboline-3-carboxylic acid N-methylamide (CMA), that bound purified Mcm2–7 with an apparent Kd of ~ 120 μM and specifically and reversibly blocked S-phase progression in yeast cells. These results validate our screening approach and strongly implicate CMA as an Mcm2–7 inhibitor.

Methods

Pilot library screening.

The LOPAC library was obtained through Sigma Aldrich. Based upon our previously developed 384 well HTS assay¹, budding yeast strains UPY1540 (wild-type) and UPY1544 (mcm2DENQ) (Sup. Methods and Table S1) were diluted from an overnight culture to an Abs₆₀₀ of ~0.1 and 45 μl of each strain were seeded in parallel into duplicate 384 well clear-bottom microtiter plates for each strain (Greiner 781091). As detailed in the Sup. Methods, these strains contain deletions in key multidrug transporters that increase their sensitivity to small molecules. A master plate of LOPAC compounds, stored in assay ready 10 mM DMSO aliquots (2 μl) under temperature- and humidity-controlled conditions was thawed and reconstituted in 18 μl YPD using a multidrop bulk liquid dispenser to obtain a 10X treatment solution of 1 mM compound in 10% DMSO/YPD. Columns 1 and 24 were loaded with 10% DMSO/YPD and columns 2 and 23 with 1 mM aphidicolin in 10% DMSO/YPD using an electronic multichannel pipettor (ThermoFisher MATRIX). 5 μl of treatment solution were transferred to yeast plates with an Agilent Bravo liquid handler for a final compound concentration of 100 μM in 1% DMSO. Plates were covered with transparent film (Perkin Elmer [Waltham, MA] TopSeal-A, cat. no. 6005185) and incubated at room temperature for 24 hours in a Perkin Elmer Envision Plate Reader fitted with a plate stacker. Abs₆₀₀ reads were recorded every hour following shaking for 10 seconds prior to reading.

The % control value was calculated as:

$$\%\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}=\left(\mathrm{A}\mathrm{b}{\mathrm{s}}_{600}\left(\mathrm{w}\mathrm{e}\mathrm{l}\mathrm{l}\right)–\mathrm{A}\mathrm{b}{\mathrm{s}}_{600}\left(\mathrm{w}\mathrm{e}\mathrm{l}\mathrm{l}\mathrm{t}=0\right)\right)/\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}\left(\mathrm{A}\mathrm{b}{\mathrm{s}}_{600}\left(\mathrm{v}\mathrm{e}\mathrm{h}\mathrm{i}\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\right)–\mathrm{A}\mathrm{b}{\mathrm{s}}_{600}\left(\mathrm{v}\mathrm{e}\mathrm{h}\mathrm{i}\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{t}=0\right)\right)*100.$$

Duplicate wells at each time and treatment were averaged and plotted (mean ± SD) as a function of incubation time or drug concentration, as needed¹.

Dose-response.

Strains UPY1540 (wild-type) and UPY1544 (mcm2DENQ) (Sup. Methods and Table S1) were grown and plated as in the general LOPAC protocol above except that each plate contained two columns of the opposite strain to be treated with DMSO or 100 μM aphidicolin, respectively, as positive and negative controls to calculate Z-factors for differences in strain sensitivity to compounds. Compounds were repurchased from commercial vendors and dissolved in DMSO at the highest concentration that upon a 10-fold dilution with YPD did not lead to precipitation (Table S2). Ten point, two-fold serial dilutions were generated in triplicates on 384 well plates in DMSO using the Agilent Bravo, and aliquoted (3 μl) into assay-ready daughter plates. On the day of treatment, compounds were reconstituted with 27 μl of YPD, and aphidicolin and DMSO added manually as above. 5 μl aliquots were then added to yeast plates and plates incubated and read on the Envision reader as above. The growth curves presented in Fig. 6 were conducted similarly but utilized 96-well plates and were read using a BioTek Cytation 5 microplate reader.

Quantification of growth kinetics.

Curves were analyzed for three cell growth parameters as described¹ and subjected to linear discriminant analysis (LDA). LDA holistically quantifies growth curves of the wild-type and mcm2DENQ strains when exposed to a given drug. This analysis method accounts for the lag time before log phase, growth rate during log phase, and maximum saturation to generate an overall “LDA score.” Drugs that specifically decreased the growth of mcm2DENQ strain but did not significantly affect the wild-type strain were considered “hits.” The LDA scores from growth kinetics were calculated with an R script as previously described¹. For dose-responses, LDA as a mathematically derived, aggregated value resulted in different maximum responses for each compound, and for concentration response curves we used the slope of the growth curves (mu) instead to enable direct comparisons of compound potency.

Analysis of reactive oxygen species (ROS).

Our ROS assay was adapted from a previously established method²⁴. Freshly grown culture of strains UPY1540 (wild-type) and UPY1544 (mcm2DENQ) were treated for 24 h with either DMSO (1%), CMA (200 and 500 μM), 3 mM H₂O₂, or 0.3% methylmethane sulfonate (MMS). Aliquots (~5×10⁶ cells) were centrifuged for 20 seconds, and the pellet was resuspended in 500 μl PBS containing 5 μM dihydroethidium (DHE, Thermofisher Invitrogen D11347). After 30 minutes at room temperature, the cells were pelleted again and resuspended in 500 μl PBS. 30 μl of the resulting cell suspensions were plated in the wells of a clear bottom 384 well microplate (Greiner), spun down at 500 xg for 30 seconds, and imaged on a Perkin Elmer Opera Phenix HCS reader using a 20x air or 63x water immersion objective. Five imaging fields were acquired in the transmitted light and Cy3 channel (ex 561 / em 570–630 nm). Final cell numbers were ~10,000 cells/field at 20x and ~2,000 cells/field at 63x. Images were subsequently analyzed using Harmony 5.0 software (Perkin Elmer). Yeast cells were detected in inverted images from the transmitted light channel, refined by size and roundness, and analyzed for fluorescence intensity. Thresholds were set as the mean ± 2 SD from DMSO treated wild-type cells. Data were presented as the percentage of cells positive for Cy3. Cytological images of ROS shown in Fig. S4 were manually counted from ≥ 100 cells/sample using a Zeiss Axiophot fluorescence microscope equipped with a Zeiss #20 (red) filter set.

Prokaryotic DNA damage assay.

The 96 well assay UMU Chromotest kit (EBPI (ebpi-kits.com) was used following the manufacturer’s instructions. In brief: 1) The test strain in LB + ampicillin (100 μg/mL) (~ 0.1 Abs₆₀₀) was incubated in a 96 well microtiter plate with serial dilutions of the target drug for 2 hours. Each plate additionally included DMSO solvent controls (1 %), positive controls (10 μg/mL 4-Nitroquinoline 1-oxide, 4NQO), blank wells containing only growth media, and negative controls containing only growth media and cells; 2) To allow cell growth after compound exposure, the wells were diluted 10x into fresh LB + Ampicillin in a new plate, grown at 37° C for 2 hours, and then the Abs₆₀₀ was recorded and used to calculate cell growth (below); 3) To assay ß-gal activity, 30 μl of each well was mixed with assay reagent (30 μl/well of substrate (40 mM ortho-nitrophenyl-β-galactoside (ONPG) in 0.1M Na Phosphate buffer (pH 7.4)) and 120 μl/well of buffer (3.2 mM BME, 10 mM KCl, 1 mM MgSO₄, 0.1 % SDS, and 0.1M Na phosphate (pH 7.4)), incubated for 30–60 minutes at 37°C (depending upon the rate of color development), stopped by addition of 1M sodium carbonate, and the Abs₄₂₀ was taken and recorded to calculate enzyme activity (below).

Drug toxicity was calculated as the ratio of ß-gal activity to relative growth.

$$\mathrm{R}\mathrm{e}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}=\left({\mathrm{S}}_{600}-{\mathrm{B}}_{600}\right)/\left({\mathrm{N}}_{600}-{\mathrm{B}}_{600}\right)$$

Where S is the Abs₆₀₀ of the sample, B is the average Abs₆₀₀ of the blank wells and N is the average Abs₆₀₀ of the negative control wells. Similarly

$$\mathrm{\beta }\text{-gal}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{t}\mathrm{y}=\left({\mathrm{S}}_{420}-{\mathrm{B}}_{420}\right)/\left({\mathrm{N}}_{420}-{\mathrm{B}}_{420}\right)$$

$$\mathrm{D}\mathrm{r}\mathrm{u}\mathrm{g}\mathrm{t}\mathrm{o}\mathrm{x}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{t}\mathrm{y}=\mathrm{\beta }\text{-gal}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{t}\mathrm{y}/\mathrm{r}\mathrm{e}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}$$

For a compound to be considered genotoxic, the relative growth needed to exceed 0.5 and the drug toxicity needed to exceed 30% of the 4NQO positive control. Finally, the levels of DNA damage were normalized to the solvent control.

Eukaryotic DNA damage.

DNA damage was quantified as the faction of S. cerevisiae cells containing Rad52-YFP repair foci²⁵. Asynchronous cultures of the UPY1540 (wild-type) and UPY1544 (mcm2DENQ) strains expressing a Rad52-YFP reporter construct were grown in YPD + drug at 30° C for 4 hours, cells harvested by centrifugation, dispersed by sonication, spun down, and resuspended in 40% glycerol. Cells containing Rad52 foci were manually counted from ≥ 100 cells/sample using a Zeiss Axiophot fluorescence microscope equipped with a Zeiss #9 (green) filter set.

Thermal shift assays.

This assay was conducted largely as described²⁶. Purified Mcm2–7 (Fig. S1) was diluted to 0.2 mg/mL in H/0.3 (25 mM HEPES pH 7.6, 0.3 M potassium glutamate, 1 mM EDTA, 1mM EGTA, 5 mM Mg(OA)₂, 10% glycerol, 1 mM DTT0. 0.1 mM PMSF), 10x SYPRO Orange protein gel stain (Invitrogen), and test drug with a final [DMSO] of 0.4%. 20 μL of each sample (containing 5 μg of Mcm2–7) was loaded into a MicroAmp Fast 96-Well Reaction Plate (0.1 mL, Applied Biosystems) and developed using a Thermofisher Quantstudio 3 RT-PCR machine running a 25–95 °C temperature gradient in 1°C steps, with each step held for 21 seconds. Fluorescence was recorded using the fluorescein filter set (Ex 470 nm/Em 570 nm). The data from at least 3 independent assays were normalized, averaged, and the melting temperatures (T_ms) were determined by non-linear regression to fit our data to equation 1 (from²⁷) using GraphPad Prism. To measure apparent Kd, this assay was repeated over the indicated CMA concentration range in the presence of 2% DMSO, and the resulting data was fitted to the Hill equation (4.1 of reference²⁸).

Fluorescence Activated Cell Sorter analysis (FACS).

Cultures of the wild-type and mcm2DENQ strains (UPY1540 and 1544) were first synchronized in G1 using alpha factor²⁹. After 3 hours of incubation at 30° C, the cells were washed to remove alpha factor and released into pre-warmed YPD to resume cell cycle progression. At 45 minutes after release (corresponding to T=0, roughly corresponding to G2), the indicated treatment was added (CMA (250 or 500 μM), hydroxyurea (Sigma H-8627, 200 mM), or 0.25% DMSO). Incubation continued, and aliquots (5×10⁶ cells) were harvested at the indicated times and processed in two different ways: one sample was quantified for Rad52-YFP foci as described above, while a second sample was fixed in 95% ethanol, processed as described for FACS analysis³⁰, and analyzed using a Thermofisher Attune Nxt Flow Cytometer and plotted using FloJo (https://www.flowjo.com).

Results

HTS assay validation and multi-day assay variability.

To fully implement and validate the synthetic lethality assay for HTS, we first conducted a formal three-day variability assessment. Using HTS automation, two full 384 well microplates of both the wild-type and mcm2DENQ reporter strains (Sup. Methods) were both treated with the positive control aphidicolin (100 μM) on 3 separate days. Growth kinetics were monitored over 24 hours and analyzed by Gompertz equation fitting and Linear Discriminant Analysis (LDA) as described¹. LDA values were then used to calculate performance parameters. Our assay met accepted HTS criteria for inter- and intra-plate variability on all 3 days (Z-factors 0.76 – 0.81, % CV< 10%, Fig. S2)

Pilot library screening.

The LOPAC library was then used as a test cassette to validate this HTS under compound screening conditions using laboratory automation. The LOPAC library is a small (1280 member) collection of FDA-approved drugs and biologically active compounds whose individual members are readily available (Sigma). The library was screened at three concentrations (100, 30, and 10 μM) and analyzed for hit rate, selection criteria, and assay performance. The range of high compound concentrations was chosen due to an anticipated intrinsic resistance of yeast to small molecules. LDA was used as a measure of cytotoxicity. As expected for a yeast screen, the lower drug concentrations gave only minor responses (data not shown), and thus only the highest concentration was chosen for analysis.

As expected, our previously determined aphidicolin positive control¹ (red symbols in Fig. 1AB) was much less toxic to wild-type cells compared with the mcm2DENQ cells as expected. The primary hit rate of toxic compounds was similar for the two strains (Fig. 1A and 1B); the total number of hits that inhibited either or both strains by > 50% was 52 (4.1% hit rate). This result suggests that during HTS many non-selective compounds would be identified if we only tested for mutant strain toxicity, an approach sometimes used in synthetic lethal screening. We therefore determined selectivity for the mutant strain by calculating the difference between LDA values for the two strains and found that 77% of the primary hits for the mcm2DENQ strain were not selective for this mutant strain (Fig. 1C upper right-hand quadrant). In contrast, a small number of compounds (12) clustered with aphidicolin (Fig. 1C lower right-hand quadrant), indicating selectivity. We consider these 12 compounds to be putative Mcm2–7 inhibitors.

Fig. 1. Pilot library screening.

Wild-type and mcm2DENQ strains were treated in parallel with library compounds. Growth kinetics were captured by linear discriminant analysis (LDA) combining lag time, growth rate, and maximum expansion. A. Compound growth inhibition in the W303 wild-type strain (UPY1540). B. Compound growth inhibition in the W303 mcm2DENQ strain (UPY1544). C. Identification of compounds selectively affecting the mutant strain by plotting differential growth inhibition between strains (y-axis, LDA difference) against toxicity in the mutant (x-axis, LDA). Data in A-C are based on the 100 μM condition and are the average of two replicates. D. Dose-response curves for aphidicolin and five hits that showed both complete dose-response curves and selectivity for the mcm2DENQ mutation (green) vs. wild-type (red) in two different genetic backgrounds. Upper panel, SEY6210 (wild-type MRY548; mcm2DENQ UPY1355); lower panel, W303 (wild-type UPY1540; mcm2DENQ UPY1540). Data are based on triplicates from a single experiment. Note that the y-axis in D. is the slope of the growth curves (mu) instead of LDA, as LDA is a mathematically derived value, and magnitudes of response can vary widely between compounds, making direct comparisons of maximal response difficult. Inhibition curves were analyzed by a four-parameter logistic equation in Spotfire (TIBCO). Table shows EC50s, maximal growth inhibition, selectivity index (SI), and the maximally achievable concentrations of test agents in 1% DMSO/YPD.

To provide a preliminary characterization of our “hits”, authentic samples of all 12 positives were then repurchased from commercial vendors (Table S2) and tested in 10-point, two-fold dose response in two genetically distinct strain backgrounds (SEY6210 and W303). This yielded five compounds (5-(6H)-phenanthridinone, ellipticine, idarubicin, CMA, and SP600125) that showed both dose-dependent toxicity and selectivity for the 2DENQ mutation in both strains (42% confirmation rate, Fig. 1D and Table 1), which were followed up in a series of secondary assays (below).

Table 1.

List of compounds tested in secondary assays

Compound/PubChem CID	Structure	Source/purity	Primary Human target
Ellipticine/ 3213		Sigma, 324688, >99% (HPLC)	DNA intercalator Topoisomerase II49
Idarubicin/ 42890		Sigma, I1656, 100% (TLC)	DNA intercalator Topoisomerase II50
SP600125/ 8515		Sigma, S556, <98% (HPLC); MolPort-000-628-456, 98% (HPLC)	c-JNK51
6(5H)-Phenanthridinone/1853		Sigma, 299634, 99.1% (HPLC)	PARP52
N-Methyl- β -carboline-3-carboxamide/ 4375 (CMA)		Sigma, E006, 98.7% (HPLC); FG7142, Tocris Bioscience, >99%	GABAA receptor inverse agonist 53
Ethyl β-carboline-3-carboxylate/ 105078 (CEE)		Sigma, 244309, >99% (TLC)	GABAA receptor54
Norharmane/ 64961		Sigma, N6252, >99.3% (TLC)	MAO-A 55

Secondary Screening Rationale.

The primary synthetic lethal screen will identify all possible mechanisms that cause a differential growth defect between the wild-type and mcm2DENQ strains. While the mcm2DENQ mutant is hypersensitive to Mcm inhibition, at least some non-Mcm2–7 proteins cause synthetic lethality in combination with the mcm2DENQ allele (e.g., mutations in the replication checkpoint), and thus only some of our “hits” are expected to directly target Mcm2–7¹. Furthermore, as Mcm2–7 is an essential enzyme, various known Mcm mutants by themselves either block DNA replication³¹, cause DNA damage⁸, or block cell cycle progression³², we anticipate that true Mcm inhibitors identified in our screen will cause those effects in both the wild-type strain as well as in our sensitized mcm2DENQ reporter strain, albeit at lower compound concentrations in the latter. We therefore performed a series of secondary assays to eliminate compounds whose activity is inconsistent with selective Mcm2–7 inhibition, in particular those that 1) cause non-specific DNA damage, 2) are unable to physically interact with purified Mcm2–7 and 3) are unable to cause an S-phase arrest and block DNA replication in wild-type cells. Together, we anticipate that these secondary assays will collectively screen for Mcm inhibitors worthy of detailed study.

Elimination of non-specific DNA damage agents.

An anticipated source of false positives are agents that directly interact with and damage DNA instead of inhibiting Mcm2–7. As prokaryotes lack the Mcm complex, we reasoned those non-specific agents should cause DNA damage in both prokaryotes as well as eukaryotes, while Mcm2–7 specific inhibitors should only cause DNA damage in eukaryotes. We compared the DNA damage properties of our “hits” in both a prokaryote (Salmonella typhimurium) as well as in budding yeast to operationally distinguish between these possibilities.

Toward this end, we used a commercially available 96-well assay system that features a Salmonella test strain containing a translational fusion between the DNA damage-induced protein UmuC and ß-galactosidase; the amount of DNA damage is proportional to the amount of this fusion protein, which can be conveniently quantified using a standard ß-galactosidase assay (³³, Materials and Methods).

We tested our five LOPAC “hits” over the same concentration range used during library screening (10 μM −100 μM). Results are shown for each compound at the highest concentration that still allows >50% of the cell growth relative to the solvent control (Fig. 2A) as lethality invalidates the test results. Following normalization of ß-galactosidase levels to cell concentration, the positive controls 4-nitroquinolone N-oxide (4NQO³⁴) and methylmethane sulfonate (MMS³⁵) respectively caused a ~ 9- and 30-fold induction in ß-UMC-galactosidase levels (Fig. 2A). Two LOPAC “hits” generated high levels of DNA damage in S. typhimurium (ellipticine (5.3x) and idarubicin (23.7x)), while the three remaining compounds produced little or no demonstrable damage.

Fig. 2. DNA damage assessment.

A. S. typhimurium DNA damage assay. The test strain (TA1535/pSK1002) couples a reporter gene (lacZ) to a DNA damage promoter. Values indicate enzymatic levels of β-galactosidase activity in each treatment, normalized for cell number, relative to the solvent control. The average and range of 2 independent experiments are shown. B. DNA Damage in wild-type yeast (UPY1540). Values reflect the fraction of cells containing Rad52-YFP foci normalized to the levels observed in the absence of drug (1% Rad52-YFP). C. DNA Damage in the mcm2DENQ strain (UPY1544). Same as B) except the levels are standardized to the mcm2DENQ strain in the absence of drug (12% Rad52-YFP). All [drug] used are noted in parenthesis after the drug name (μM). The left y-axes show fold increases in DNA damage, the right Y-axis shows the corresponding raw percentages. The horizontal line on each graph reflects the level of damage observed in the reporter strain solvent control (DMSO 1%). Unless noted, results denote the average ± SEM of three independent experiments.

The three prioritized hits were then examined for DNA damage potential in both budding yeast reporter strains via a cytological assay. Rad52 is a key recombination factor, and the abundance of Rad52 nuclear foci is a common metric for DNA damage¹⁹. Logarithmically growing asynchronous cultures of the reporter strain with 100 μM test compound (except for idarubicin, which due to cytotoxicity was assayed at 10 μM) were visualized by microscopy after 4 hours of growth and scored for DNA damage foci. As Mcm2–7 inhibition leads to genomic instability, we anticipate that test compounds that increase the level of DNA damage foci in either the wild-type or mcm2DENQ test strains but not in S. typhimurium to be potential Mcm2–7 inhibitors.

In the wild-type strain, about 1% of the cells demonstrated DNA repair foci during unchallenged growth, while addition of known DNA damage agents (e.g., MMS, ellipticine, and idarubicin), yielded a 10–38-fold increase in cells with DNA damage foci (Fig. 2B). Among the three prioritized hits, one generated high levels of DNA damage (SP600125, 12-fold) while the other two caused a significant but smaller 2–3 fold increase (Fig. 2B).

We next assayed DNA damage in the mcm2DENQ strain. As previously reported ¹⁹, the mcm2DENQ mutation causes elevated DNA damage, with about 12% of the cells in the solvent control containing Rad52-YFP foci (Fig. 2C). The positive control (e.g., MMS, ellipticine and idarubicin) produced an additional 3–5.7 -fold increase in DNA damage after normalization (Fig. 2C). However, the remaining “hits” caused a range of DNA damage: CMA caused DNA damage comparable to that observed with ellipticine and idarubicin (5.8-fold); SP600125 caused moderate DNA damage roughly similar to that observed with the wild-type strain (~3-fold); and phenanthridinone caused only a slight increase in DNA damage as compared to the solvent control (~ 2-fold) (Fig. 2C).

In summary, ellipticine and idarubicin caused significant DNA damage in both S. typhimurium and S. cerevisiae and were thus eliminated as potential Mcm2–7 inhibitors. The three remaining candidates (CMA, 6(5H)-phenanthridinone, and SP600125) caused little DNA damage in S. typhimurium but preferentially caused DNA damage in either the wild-type or mcm2DENQ strains of yeast consistent with being eukaryotic-specific DNA replication inhibitors.

Identification of small molecules that bind Mcm2–7.

We anticipate that Mcm2–7 inhibitors will physically bind this complex. Toward this end, we developed a thermal shift assay to identify Mcm2–7 small molecule binders^{26, 27}. This assay leverages the observation that ligand binding often stabilizes proteins with a concomitant increase in their T_m (temperature at which 50% of the protein unfolds).

We first assayed the inherent thermal stability of the purified wild-type Mcm2–7 complex (Materials and Methods, Fig. 3A) and found that it denatured between ~40–50°C, with the T_m at 44.05 °C (Fig. 3A, Table 2). We used the non-hydrolyzable ATP analog ATPγS as a positive control³⁶which caused a large +5.6 °C T_m shift in Mcm2–7 (Fig. 3B, Table 2). This result supports the feasibility of our assay.

Fig. 3. Mcm2–7 ligand binding.

See Materials and Methods for details. Unless noted, all assays contain 0.4% (v/v) DMSO. T_ms shown are the average and SEM from at least 3 independent experiments; often the variance is too small to generate visible error bars. Melt curve for: A. Mcm2–7 (solvent control); B. Mcm2–7 ± 5 mM ATPγS; C. Mcm2–7 ± 200 μM SP600125 or 200 μM Phenanthridinone; and D. Mcm2–7 ± 200 μM β-carboline 3-carboxylic acid N-methylamide (CMA). The results from 2 different CMA sources (Sigma vs. Fisher) are shown. E. The thermal shift assay repeated over the indicated range of CMA concentrations to measure Kd app and the Hill coefficient. F. ATPγS and CMA cause a synergistic increase in the Mcm2–7 T_m. The melting curve of Mcm2–7 + ATPγS (5 mM) (left curve) and Mcm2–7 + ATPγS (5 mM) + CMA (200 μM) (right curve) are compared.

Table 2.

Thermal Shift Assay Results

Name	(conc μM)		Tm (°C)	95% CI	Δ Tm
No inhibitor	ND		44.05	43.95–44.15	-
ATPγS	5000		49.65	49.55–49.65	+5.6
CMA - Sigma	200		45.65	45.45–45.95	+1.6
CMA - Fisher	200		45.75	45.55– 45.95	+1.7
SP6001254	200		44.95	44.75–45.15	0.9
Phenanthridinone	200		44.45	44.35–44.55	0.4
ATPγS + CMA	5000 + 200		50.25	50.15–50.25	+6.2
CEE	200		45.05	44.95–45.25	+ 1.0
Norharmane	200		43.95	43.85–44.05	− 0.1

We next examined our 3 putative “hits”. To help ensure complete binding occupancy, we utilized compound concentrations 2-fold higher than that used in the primary screen (200 μM). Although both phenanthridinone and SP600125 caused a small Mcm2–7 T_m shift of ≤ +1.0 °C (Fig. 3C, Table 2), CMA routinely produced a larger temperature shift of ~ +1.7 °C that was independent of the compound’s source (Fig. 3D, Table 2), Sigma vs. Fisher). Given these data, we prioritized the further study of CMA relative to the other remaining hits.

We repeated the thermal shift assay over a broad CMA concentration range (Fig. 3E) to determine its binding affinity to Mcm2–7. We modeled the observed T_m as a function of [CMA] using non-linear regression and found that a sigmoidal binding curve (R² = 0.91) fit the data better than a hyperbolic equation (R² = 0.76). With the sigmoidal equation, the Kd app for CMA was 119 μM (93.2–147.1 μM), a value similar to the CMA concentration used during primary screening. The Hill coefficient was estimated to be 2.5, suggesting that Mcm2–7 likely contains multiple allosteric binding sites for CMA. Given that Mcm2–7 contains six ATP binding sites³⁷, and CMA is a flat heterocyclic molecule resembling a nucleobase, we were curious if CMA was functioning as an ATP analog. We repeated the thermal shift assay in the presence of both ligands to test this possibility. As anticipated for ligands binding distinct sites, the observed Tm shift was roughly additive (Fig. 3F), suggesting that CMA and ATP do not compete for Mcm2–7 binding.

Finally, to further explore CMA binding, we examined several analogs that either contain just the unsubstituted carboline ring structure (norharmane), or a substitution of the carboxamide side chain for an ethyl ester (ß-carboline-3 carboxylic acid ethyl ester, CEE) (Table 1). At a concentration of 200 μM, norharmane did not produce a T_m shift while CEE did reproducibly produce a +1.0 °C temperature shift (Fig. S3A). Although these data suggest that CEE may bind Mcm2–7, neither 200 μM norharmane nor CEE appear to be toxic toward our yeast reporter strains (Fig. S3B, Table 2), suggesting perhaps that a < +1° C T_m shift is too small to reliably predict stable ligand binding. As these analogs lack the carboxamide side chain found in CMA, these preliminary structure activity data suggest that the carboxamide side chain is critical for Mcm2–7 binding specificity.

CMA specifically blocks DNA replication and S-phase progression.

As conditional Mcm mutants block cell cycle progression and DNA replication under non-permissive conditions (^{31, 32}), we anticipated that a true Mcm2–7 small molecule inhibitor at an appropriate concentration should cause similar effects even in the wild-type strain. We therefore used fluorescence activated cell sorter analysis (FACS³⁰) to quantify the DNA content of individual cells in a population (Materials and Methods). In brief, a culture of the test strain was first synchronized in G1, and then released into fresh media. At 45 minutes after G1 release (denoted as T=0), the indicated concentrations of either DMSO (solvent control), hydroxyurea (control for S-phase arrest²⁹) or CMA were added to separate cultures. Aliquots were taken at indicated intervals following CMA addition and processed for both FACS analysis and Rad52-YFP DNA damage foci quantification.

In the wild-type solvent control at T=0, the cell population had a nearly complete 2N DNA content, and subsequently lost cell cycle synchrony by T=2 hours (Fig. 4A). In contrast, cultures containing hydroxyurea completed mitosis and arrested by T=1.5 hour with a 1N DNA content. As HU only slows but does not completely stop DNA replication³⁸, by T=4, the DNA content of this culture had slowly increased and reached a DNA content/cell midway between 1 and 2 N (Fig. 4A). These features are consistent with S-phase arrest due to inhibition of DNA replication.

Fig. 4. CMA causes S-phase arrest.

Cycling cultures of either the wild-type (A. UPY1540) or mcm2DENQ (B. UPY1544) reporter strains were synchronized in the G1 phase as described in Materials and Methods and released into YPD media. 45 minutes after release, the indicated treatment was added. Cell aliquots were taken at the indicated times post CMA addition and processed for both FACS analysis and Rad52-YFP DNA damage foci quantification (boxed numbers at each timepoint). Representative experiments are shown.

Like hydroxyurea, CMA blocked S-phase progression and DNA replication. Wild-type cells released into 500μM CMA divided with normal kinetics but arrested with 1N DNA content by 1.5 hours that slowly trended toward a higher DNA content by T=4 hours (Fig. 4A). At a lower CMA concentration (250 μM), cell cycle progression was slowed in G1/S relative to the solvent control, but the cells did not arrest (Fig. 4A).

The mcm2DENQ strain was similarly analyzed. As with the wild-type strain, exposure to 500 μM CMA caused an S-phase arrest (Fig. 4B). However, in contrast to the wild-type strain, the mcm2DENQ strain at 250 μM CMA arrested with a 1 N DNA content (Fig. 4B). This increased sensitivity to CMA can be explained by the unique checkpoint and replication defects of the mcm2DENQ mutant (below).

CMA differentially affects both cellular viability and DNA damage in the wild-type and mcm2DENQ strains.

CMA causes a differential growth defect in the mcm2DENQ strain relative to the wildtype strain (Fig. 1). In principle, this difference could be either caused by a reduction in growth rate, a reduction in cellular viability, or possibly both. The effect of DNA damage and cellular viability were assayed following CMA exposure to dissect these possibilities.

Aliquots from the same timepoints used in the FACS analysis (Fig. 4) were examined for DNA damage by measuring nuclear Rad52-YFP foci¹⁹. CMA caused little apparent DNA damage with wild-type cells (Fig. 4A, inset numbers) consistent with our observations using asynchronous cells (Fig. 2B). However, with the mcm2DENQ strain, the DNA damage heavily depended upon the CMA concentration in a counter-intuitive manner: at the high CMA concentration (500 μM), the levels of Rad52 foci were low and like that observed with the solvent control (10–15%). In sharp contrast, at the lower CMA concentration (250 μM), Rad52-foci increased over the course of the experiment and by T=4 hours peaked with 72% of cells containing Rad52 foci (Fig. 4B), a level of DNA damage like that observed using asynchronous cultures (Fig. 2C).

To determine the relationship between DNA damage and cellular viability, CMA was added to asynchronous populations of both strains in liquid culture, and viability was measured as a function of CMA exposure time by a colony formation assay (Sup. Methods).

We observed that cellular viability (Fig. 5) paralleled the level of DNA damage (Fig. 4). By 24 hours, 500μM CMA treatment (a concentration 5-fold above that used in initial screening) caused <40% reduction of cell viability for both wild-type and mcm2DENQ cells (Fig. 5). However, in the presence of 200 μM CMA (a concentration only 2-fold above that used during initial screening), viability of the mcm2DENQ strain was reduced to ~ 1% of the starting value following 24 hours of CMA incubation (Fig. 5).

Fig. 5. CMA toxicity.

A. To growing cultures of either the wild-type (UPY1550) or mcm2DENQ (UPY1544) reporter strains CMA was added to the indicated concentration. At the specified timepoints, the cells were diluted, counted in a hemocytometer to measure total cells, and plated on YPD media lacking CMA to measure viable cells. Together these data were used to calculate viability (Sup. Methods). The average and SEM from 4 independent experiments are shown. B. Strains and growth conditions were the same as A., except that DHE was added to the cultures at 2.5 μg/ml after 24 hours of growth, and the cells were scored for reactive oxygen (e.g., red fluorescence) using a Perkin Elmer Opera Phenix HCS reader as described in Materials and Methods. The means and SEMs of three independent experiments are shown.

We then wished to explore this apparent viability loss during CMA treatment – was apoptosis induced during early CMA exposure or were the cells metabolically active but otherwise unable to divide? To assay apoptosis, we examined the inappropriate production of reactive oxygen (ROS) using an established cytological assay²⁴. In brief, test cells were treated with DHE, a cell-permeant dye. Upon exposure to reactive oxygen, this compound is converted to ethidium, which is a red-fluorescent dye that remains in the cell via DNA intercalation. Red fluorescence is then a metric for ROS production, and we scored the fraction of red fluorescent cells in a test population both via high-content screening (Fig. 5B) as well as by standard fluorescence microscopy (Fig. S4, Materials and Methods). In addition, the ability of the cells to form microcolonies after each treatment was also assayed (Fig. S4).

As a positive apoptosis control, both wild-type and mcm2DENQ strains were treated with hydrogen peroxide¹⁹ or the potent DNA damage agent MMS. After 24 hours in the presence of either compound, approximately 70–100% of the cells demonstrated red fluorescence and were almost completely unable to form microcolonies following incubation on non-selective agar plates (Fig 5B, S.4, and data not shown). In contrast, both wild-type and mcm2DENQ strains similarly treated with DMSO demonstrated negligible ROS (3–7%) and efficient microcolony formation (~90%, Fig 5B and S.4).

We next examined the effects of CMA on ROS production. Except for the mcm2DENQ strain with 500 μM CMA (~16 % ROS, Fig 5B), 24 hours of continuous CMA exposure did not increase the level of ROS in either the wild-type nor mcm2DENQ strains above those observed for the solvent controls. Considering the possibility that apoptosis may only develop upon normal growth resumption, we also examined ROS production after 24 hours of growth in CMA followed by 8 hours of growth in the absence of CMA; this treatment did not increase ROS levels (Fig. S4). However, similar to our original viability data (Fig. 5A), we observed that microcolony formation was very efficient in the wild-type strain following CMA exposure, but microcolony formation was reduced in the mcm2DENQ strains, with very few microcolonies formed following 200 μM CMA exposure (Fig. S4). Given the lack of correlation between ROS levels and apparent loss of viability in the mcm2DENQ strain, apoptosis does not appear to be responsible for the viability loss under this condition.

Closer examination of the microcolony experiments demonstrated that most of the mcm2DENQ cells in the 200 μM CMA experiment that failed to form microcolonies arrested as oversized cells with a bud morphology consistent with a G2 arrest (>85%, Fig. S4). This result is in considerable contrast to the hydrogen peroxide control, in which the non-dividing cells do not correlate with any cell cycle stage (Fig. S4). These results strongly suggest that in the mcm2DENQ strains with 200 μM CMA, the cells may remain metabolically active yet fail to divide as they are irrevocably arrested in G2. Although prolonged G2 arrest may eventually result in apoptosis, this does not appear to be the primary cause of their apparent cell division defect following 200 uM CMA exposure.

In summary, three features should be noted. First, the effects of CMA were generally reversible, as the wild-type cells grew and formed colonies with little to moderate loss of viability once CMA was removed. Second, the viability of the mcm2DENQ strain largely correlated with the level of DNA damage. Third, DNA damage and loss of viability in the mutant strain only occurred at the lower CMA concentration in a manner that does not appear to correlate with apoptosis. These results will be further discussed below.

The mcm2DENQ mutant demonstrates unique growth sensitivity to CMA.

To further explore CMA specificity toward the Mcm2–7 complex, we tested its effect on the growth of several potentially informative isogenic replication and checkpoint mutants. Similar to our screening protocol, these test strains were grown in 96-well plates at room temperature and monitored hourly at OD₆₀₀ using a microplate reader (Materials and Methods). Each test strain was assayed over a range of CMA concentrations from 0–200 μM as indicated. For the wild-type strain, little growth inhibition was observed until the CMA concentration exceeded 100 μM (Fig. 6A). In contrast, the mcm2DENQ mutant demonstrated noticeable growth defects even at the lowest CMA concentration used (12.5 μM, Fig. 6A). These observations parallel the original screening results (Fig. 1).

Fig. 6.

The dose-dependent effect of CMA on the growth of various replication and checkpoint mutants. As described in Materials and Methods, growing cultures of wild-type (UPY1550), mcm2DENQ (UPY1544), mcm4 chaos3 (UPY1720.2), cdc45–27 (UPY1797) and Δmrc1 (UPY1723.1) were diluted to ~ 0.05 OD₆₀₀ in rich media (YPD) in the presence on the indicated concentration of CMA. Cultures were incubated at room temperature and the OD₆₀₀ of each culture was read hourly over a 24-hour period using a Bio-Tek Cytation 5 microplate reader. For each graph, the growth of the wild-type solvent control is shown, as well as the growth of the mutant (test strain) with the indicated treatment.

To examine if CMA also inhibits other mcm mutants, we tested its effect on a well-studied mutation in Mcm4 (chaos 3⁸). Typical for mcm alleles, including mcm2DENQ, this non-conditional Mcm4 allele causes a wide variety of genome stability defects. However, in our assay we found that the chaos3 mutant demonstrates wild-type CMA resistance (Fig 6C).

To more broadly examine if CMA inhibits mutants in other non-Mcm subunits of the CMG complex, we tested its effects on the growth of a temperature sensitive allele of CDC45, cdc45–27³⁹. Under the permissive conditions used in this assay (growth at room temperature), the cdc45–27 mutant also demonstrated wild-type growth resistance to CMA (Fig. 6D).

In addition to DNA replication defects, mcm2DENQ mutant has defects in the replication checkpoint cascade⁵. To test the effect of CMA on other checkpoint mutations, we tested a deletion mutant in the non-essential MRC1 gene. MRC1 is a mediator in the replication checkpoint cascade⁴⁰, and mrc1 mutations have an array of genome stability defects that are roughly similar to those observed in the mcm2DENQ mutant¹⁹. Similar to the chaos3 and cdc45–27 mutants, the mrc1 mutant demonstrated wild-type CMA resistance (Fig. 6E)

In summary, the mcm2DENQ mutant appears to be unique in its growth sensitivity to CMA. Although the numbers and types of additional mutations examined is far from comprehensive, these results suggest the tantalizing possibilty that growth sensitivity to CMA is mcm2DENQ specific.

Discussion

Our screening approach effectively identified candidate Mcm2–7 inhibitors.

Under actual screening conditions used in this study, we find that our differential screening approach, coupled with simple dose-dependent retesting, eliminates many non-selective compounds relative to conventional screens that identify putative candidates through growth inhibition of a single reporter strain. For example, in our screen, 52/1280 compounds were toxic to the mcm2DENQ strain (hit rate of 4.1%), as opposed to the 5/1280 that were differentially toxic to the mcm2DENQ reporter strain (a 90% reduction in non-selective hits). Moreover, we have successfully instituted a three-prong secondary screening approach (UMU-Chromotest, thermal shift assay, and FACS analysis), which eliminated non-selective cytotoxic and DNA damaging agents, and prioritized compounds that had cell cycle effects consistent with Mcm2 inhibition. Overall, our screening pipeline successfully identified and validated a likely Mcm2–7 inhibitor (CMA) that physically interacts with Mcm2–7. Although improvements in potency are likely needed to make CMA into an optimal molecular probe, it currently has better affinity and in vivo utility than previously identified Mcm2–7 inhibitors¹⁶.

Mcm2–7 is a CMA target.

Our secondary screens demonstrate that CMA physically binds to purified wild-type Mcm2–7 and blocks S-phase progression and DNA replication in a dose-dependent manner (Fig. 4A and B). Although somewhat lower CMA concentrations appear to be needed for in vitro Mcm2–7 binding (Kd app ~ 119 μM) than blockage of in vivo DNA replication (350–500 μM), this difference is likely due to the relative impermeability of yeast to small molecules. Budding yeast contain at least 16 known multidrug transporters that remove a host of exogenously added molecules from the cell⁴¹. In contrast, our reporter strains only eliminate three of these transporters, perhaps accounting for the higher CMA levels needed to observe in vivo activity.

Interestingly, CMA demonstrates positive Mcm2–7 binding cooperativity (Fig. 3E). This strongly suggests that CMA binds multiple Mcm subunits. As each Mcm subunit is an ATPase³⁷, one possibility is that CMA cooperatively binds the ATP active sites. However, addition of both the non-hydrolyzable ATP analog ATPγS and CMA in the thermal denaturation assay causes an additive T_m increase (Fig. 3F), suggesting that with at least some Mcm subunits the two compounds can bind simultaneously. These data suggest that CMA may bind to a previously unknown allosteric site in the Mcm2–7 complex.

CMA is a ß-carboline derivative, and ß-carbolines are known to have additional potential targets⁴². Although in humans CMA binds GABA_A receptors⁴³, specific ß-carbolines can also bind DNA⁴⁴ and act as topoisomerase inhibitors⁴⁵. However, as ß-carbolines are defined by their DNA intercalating triple ring structure, it is interesting to note that the two CMA analogs that we examined, (norharmane and CEE) have the same triple ring structure as in CMA, yet unlike CMA they do not cause an apparent growth defect in either the wild-type or mcm2DENQ strains (Sup. Fig. 1C). Given that these analogs do not cause a growth defect, these data support the likelihood that CMA functions by direct interaction with Mcm2–7 rather than indirect interaction through DNA binding. Finally, our data suggests that the carboxamide side chain of CMA is essential for Mcm2–7 inhibition, an observation that may serve as a starting point for future structure activity relationship studies to identify chemical modifications that improve compound potency.

How the mcm2DENQ allele contributes to our synthetic lethal screening approach.

Although our synthetic lethal approach uses the mcm2DENQ allele as a reporter, it is not obvious why this mutant is unusually sensitive to CMA – after all, Mcm2–7 is presumably equally essential to both wild-type and mcm2DENQ strains. However, Mcm mutants often show synthetic lethality with each other¹. Consistent with both an essential role of Mcm2 and the ability of its mutants to cause synthetic lethality, our results show that CMA dose-dependently inhibits growth of both wild-type and mcm2DENQ strains (Fig. 1D) but potentiates DNA damage and reduces cell viability specifically in the mcm2DENQ strain (Figs. 2, 4, and 5). Curiously, the latter two phenotypes are not observed at a high CMA concentration that completely inhibits DNA replication. This unanticipated and counter-intuitive finding appears to depend upon (error-prone) DNA replication, as detailed below.

Our results suggest that the mcm2DENQ allele makes unique and unanticipated contributions to the success of this HTS. This allele causes a variety of specific defects: partially reduced viability, replication-dependent DNA damage in the absence of fork collapse, and a defect in activating the DNA replication checkpoint^{5, 18, 19}. In aggregate, these properties lead to error-prone DNA replication with a reduced ability to recognize and respond to the concomitant DNA damage. We posit that at high CMA concentrations, Mcm2–7 is nearly completely inhibited, with little or no actual DNA replication occurring in either reporter strain. In this case, an absence of DNA replication leads to an absence of DNA damage and a retention of cellular viability. At lower CMA concentrations, Mcm2–7 is only partially inhibited. In the wild-type reporter, S-phase is slowed with a negligible increase in DNA damage or inviability. In the mcm2DENQ reporter, however, partial Mcm2–7 inhibition presumably aggravates the mcm2DENQ replication defects, resulting in massive DNA damage and a corresponding viability reduction. As the mcm2DENQ strain is insensitive to most DNA damaging agents¹, CMA was likely identified not because it causes general DNA damage but that it causes DNA damage in a mcm2DENQ-specific manner. This supposition is supported by analysis of CMA sensitivity in other replication or checkpoint mutants, suggesting that the choice of mcm2DENQ as a reporter in our HTS was somewhat fortuitous. Collectively, our data indicate that CMA is specifically cytotoxic to the mcm2DENQ mutant, supporting the feasibility of using synthetic lethal screening approaches to identify inhibitors to specific disease alleles.

Application of Molecular probes to the study of Mcm2–7 mechanism.

Mcm2–7 undergoes a variety of likely or demonstrated ATP-dependent structural transitions that fuel its ability to load onto replication origins, unwind DNA, and respond to numerous challenges during elongation⁴⁶. Despite numerous biochemical and structural studies, many likely Mcm2–7 structural transitions remain unobserved and unstudied, in part because tools to isolate Mcm2–7 in various transition states are unavailable. Since CMA appears to block elongation (Fig. 4), we posit that CMA could be the first such tool to investigate Mcm2–7 mediated structural transitions.

Leveraging synthetic lethality to identify small molecule inhibitors of complex targets.

As only a small number of human proteins have been successfully targeted by small molecule inhibitors⁴⁷, development of effective drug screens for non-traditional targets is an important area. Although the use of a synthetic lethal approach toward drug screening has been used to great effect previously (e.g., the identification of PARP inhibitors as a way to combat triple negative breast cancer⁴⁸), such screens are often designed to identify a new target rather than an inhibitor of a specific target. Our data support the feasibility of using synthetic lethality in concert with specific target alleles as an effective cell-based screening approach to identify small molecule inhibitors of difficult targets.

Abbreviations:

BED

ß-estradiol

CMA

β-carboline-3-carboxylic acid N-methylamide

CMG

CDC45, Mcm2–7 GINS complex

DMSO

dimethyl sulfoxide

DTT

dithiothreitol

HTS

high throughput screening

ONPG

ortho-nitrophenyl-β-galactoside

PMSF

phenyl methyl sulfonate

MMS

methylmethane sulfonate

4NQO

4-nitroquinolone N-oxide

YPD

yeast extract peptone dextrose

ROS

reactive oxygen species

DHE

dihydroethidium