Identification of ATP-Competitive Human CMG Helicase Inhibitors for Cancer Intervention that Disrupt CMG-Replisome Function

The replicative CMG helicase is an emerging target for anticancer intervention that presents an exploitable vulnerability in cancer cells derived molecularly by oncogene-driven mismanagement of CMG assembly and function (1). However, to date, no small chemical inhibitors of the human CMG helicase have been identified. The CMG is a multisubunit enzyme that performs the primary DNA melting and unwinding steps within replisomes during DNA replication in eukaryotic cells (2). The CMG helicase is composed of Cdc45, a mini-chromosome maintenance (MCM) heterohexameric ATPase core, and the GINS tetramer (Go-Ichi-Ni-San in Japanese for Sld5, Psf1, Psf2, and Psf3; refs. 2–4). Assembly and activation of the CMG occur in a stepwise manner, with an excess of MCM hexamers loaded onto DNA during the G1 phase (called Licensing; refs. 5, 6), followed by recruitment of Cdc45/GINS near the G1-S transition to a subset of these MCM hexamers (3, 4, 7, 8). The dynamics of this MCM-CMG conversion process are important for maintaining genome stability in human cells (1). Unused MCM complexes act as reserves that are converted to CMG helicases during replicative stress (RS) to facilitate recovery of DNA replication (1, 9–12). Reserve MCM complexes also modulate replication fork speeds to prevent DNA damage (13).

Oncogenic changes cause problems with CMG assembly and function (1). Cyclin E overexpression reduces the number of MCM hexamers that load onto DNA, resulting in reduced MCM reserves, loss of replication fork fidelity, and consequent DNA damage (1, 14). Elevated Myc overstimulates the conversion of MCM hexamers into CMG helicases, leading to DNA damage in genomic regions with excessive CMG activity due to increased replication fork density and reduction of unused MCM reserves (1, 15–17). Although these oncogenic events produce DNA damage that facilitates tumorigenesis and tumor heterogeneity, they also create a reduction in MCM/CMG functional fidelity in tumor cells that is likely exploitable with CMG inhibitors (1). Consistent with these concepts, oncogenic K-Ras also causes fork stalling and RS (18, 19), and cells expressing mutant K-Ras are selectively sensitive to the reduction of MCM licensing (20), suggesting that future MCM/CMG inhibitors will synergize with K-Ras mutations to suppress tumor growth. In addition, inhibition of MCM/CMG reserves can selectively sensitize tumor cells (including those with mutant K-Ras) to fork-stalling chemotherapy drugs (11, 12), suggesting that CMG inhibitors will have the potential to overcome chemo-resistance in the management of cancer.

We report here the design of an orthogonal chemical screening approach and its use in the identification of the first ATP-competitive inhibitors of the ATPase and helicase activities of the human CMG enzyme [CMG helicase inhibitors (CMGi)]. These CMGi have drug-like features and are members of the aminocoumarin class of compounds (21), specifically clorobiocin and coumermycin-A1 (CA1), whereas the closely related compound novobiocin is not effective at CMG inhibition. Consistent with this, biochemical analyses with synthetic compound derivatives show that CMGi require specific chemical elements for CMG inhibition. Modeling suggests CMGi can occupy multiple Mcm2–Mcm7 ATP-binding clefts and channels, the latter of which are likely used by ATP to access ATPase domains within the MCM ring. CMGi display distinct modes of action for cell growth inhibition and induction of DNA damage relative to other chemotherapy drugs, blocking MCM DNA binding and GINS recruitment during CMG assembly, and disrupting CMG-replisome costructural integrity during the S-phase. Tumor cells are selectively sensitive to low concentrations of CMGi that target the CMG helicase in vivo and in vitro, but not to novobiocin, strongly suggesting that differential effects of these aminocoumarins on tumor cells are derived from the sugar head group targeting CMG helicase inhibition. We discuss how our findings can explain why aminocoumarins that function as CMGi had side effects in humans when tested at high doses as potential oral antibiotics in the distant past (21), and more importantly, how our results support the CMG helicase as a tumor-specific vulnerability and novel anticancer target. These CMGi can serve as unique probes to investigate ATPase-dependent CMG/MCM functions in human cells and can be used to inform the development of a new class of anticancer compounds that target and disrupt CMGs/replisomes.

Human immortalized keratinocytes (HaCaT; CLS/Cytion, Germany; cat#300493; RRID:CVCL_0038), 143B cells (osteosarcoma, OS; ATCC, Manassas, VA; cat#CRL8303; RRID:CVCL_2270), and HEK293T cells (ATCC cat#CRL3216; RRID:CVCL_0063) were maintained in DMEM supplemented with 10% FBS (Peak Serum, Bradenton, FL; cat#PS-FB2). Human Psn1 cells (pancreatic ductal adenocarcinoma; ATCC cat#CRL3211; RRID:CVCL_1644) and H460 cells (nonsmall cell lung carcinoma; ATCC cat#HTB177; RRID:CVCL_0459) were cultured in RPMI1640 medium supplemented with 10% FBS. Human primary keratinocytes (ATCC cat#PCS200-011) were cultured in a serum-free medium as per provider instructions but were temporarily cultured in DMEM/10%FBS for viability analyses. All cells used in this report were verified by our group to be mycoplasma-free or by the provider (Psn1, primary keratinocytes). Short tandem repeat verifications for cell line authenticity were performed by the provider (Psn1) or by our group (HaCaT, 143B, H460, HEK293T). Primary keratinocytes were not short tandem repeat–verified and can differ in purchased batches used for experiments by others. HaCaT cells were synchronized in G₀ using serum starvation for 48 hours and released into the cell cycle by the addition of DMEM with 10% FBS (22). HaCaT cells were infected with a lentivirus that allowed regulatable expression of HA-tagged H-Ras61L by the Tet-On promoter [using 0.05-µg/mL doxycycline inducer (cat# BP2653-1), and transduced cells were pooled after selection with blasticidin (1 µg/mL, cat#R210-01; both drugs from Thermo Fisher Scientific, Waltham, MA)]. Novobiocin (cat#46531) and CA1 (cat#C9270) were obtained from Sigma-Aldrich (Sigma, United States). Etoposide (cat#S1225) and AT7519 (cat#S1524) were obtained from SelleckChem (Houston, TX). All stock solutions of inhibitors were stored at −20°C as 10-mmol/L suspensions in DMSO.

Cell viability determinations were performed using CellTiter-Glo Assays (Promega, Madison, WI; cat#G7572). Cells were seeded in 96-well plates at a density of ∼3 × 10³ cells/well and treated with drugs for 72 hours, after which the cells were processed for viability using CellTiter-Glo reagent according to the instructions of the manufacturer. Each drug concentration test was performed using four replicates, and results were averaged and plotted on graphs, ±1 SD.

Immunoblotting was performed using standard enhanced chemiluminescent and polyacrylamide gel techniques. Lysates from equal cell numbers were separated into Triton X100-soluble or -resistant (chromatin-bound) protein fractions as described (23, 24), and compared with whole-cell protein lysates. All cell lysates were supplemented with protease inhibitors [1-mmol/L phenylmethylsulfonylfluoride (PMSF), 1-mmol/L benzamidine, 0.15-µmol/L aprotinin, 4-µmol/L leupeptin, and 1-µmol/L antipain]. Immunoblots were assessed with the following antibodies (all used at 1:500–1:1,000 dilutions): from Santa Cruz Biotechnology (Santa Cruz, CA): anti-Mcm5 (sc165994, RRID:AB_2142526), anti-Mcm6 (sc55577, RRID:AB_831540), anti-Mcm7 (sc9966, RRID:AB_627235), anti-Orc2 (sc32734, RRID:AB_2157726), anti-Cdt1 (sc28262, RRID:AB_2076885), anti-Cdc6 (sc9964, RRID:AB_627236), antiphospho-Ser54-Cdc6 (sc12920-R, RRID:AB_668066), anti-DNA polymerase ε (sc12728, RRID:AB_675496), anti-DNA polymerase δ (sc17776, RRID:AB_675487), anti-γ-H2AX (sc517348, RRID:AB_2783871), anti-Sld5 (sc398784, RRID:AB_2940776); from Abcam (Boston, MA): anti-Psf1 (ab183524, RRID:AB_2922402), and anti-Psf3 (ab254855, RRID:AB_2940777), antiphospho-Ser53-Mcm2 (ab109133, RRID:AB_10863901), anti-DNA polymerase α (ab31777, RRID:AB_731976); from Proteintech (Rosemont, IL): anti-Psf2 (16247-1-AP, RRID:AB_2111895), anti-GAPDH (60004-1-lg, RRID:AB_2107436); from Cell Signaling (Danvers, MA): anti-RPA32 (2208s, RRID:AB_2238543), anti-RPA70 (2267s, RRID:AB_2180506), anti-Mcm3 (4003s, RRID:AB_2142261), anti-Lamin A/C (2032S, RRID:AB_2136278), anti-WRN (4666, RRID:AB_10692114), anti-BLM (2742, RRID:AB_2064649), antiphospho-Ser139-Mcm2 (12958, RRID:AB_2798069), antiphospho-Ser345-Chk1 (2348, RRID:AB_331212), anticleaved Parp (5625, RRID:AB_10699459), antiphospho-Thr320-PP1-α (2581, RRID:AB_330823), anti-Rb (9313, RRID:AB_1904119), antiphospho-Ser807/811-Rb (9308, RRID:AB_331472); from Sigma: anti-Flag (F3165, RRID:AB_259529), antibeta-actin (A5441, RRID:AB_476744), rabbit IgG (I5006, RRID:AB_1163659), anti-Flag M2 agarose (A2220, RRID:AB_10063035), and antimouse IgG agarose (A6531, RRID:AB_258295); from Biolegend (San Diego, CA): anti-HA (901501, RRID:AB_2801249) and anti-And1 (Ctf4; 630301, RRID:AB_2215084); from Bethyl Laboratory (Montgomery, TX): anti-Mcm10 (A300-131A, RRID:AB_2142119); from Invitrogen (ThermoFisher): anti-PP1-α (MA5-15589, RRID:AB_10980092); and from BD Pharmingen (Franklin Lakes, NJ): anti-Mcm4 (559544, RRID:AB_397267). Chicken polyclonal anti-Cdc45, rabbit polyclonal anti-Cdc6, and rabbit polyclonal anti-Mcm2 (used at 1:2,000 dilutions) were generated by our group and validated as described (8).

Verification of synchronization and determination of drug effects was performed by measuring the incorporation of bromo-deoxyuridine (BrdU; Sigma) into replicating foci within nuclei. At times indicated, cells were pulse-labeled with 15 µmol/L BrdU for 30 minutes, fixed with 4% paraformaldehyde, and analyzed by standard immunofluorescent techniques (22, 25) with an anti-BrdU monoclonal antibody (Roche, USA; clone no. BMC9318, RRID:AB_2313622). The average counts of three fields of 100 or more cells were used to determine the percentages of BrdU-labeled nuclei, ±1 SD.

The Topoisomerase II (Topo-II) drug screening kit (kDNA-based; TopoGEN, Inc., Buena Vista, CO; TG1009-1A) was used to assess in vitro Topo-II enzyme activity according to the manufacture protocol. Assays were performed in 20-μL reactions with 4 µL of 5X Assay Buffer [0.25-mol/L Tris-HCl (pH 8), 10-mmol/L ATP, 0.75-mol/L NaCl, 50-mmol/L MgCl2, 2.5-mmol/L dithiothreitol (DTT), 150-μg/mL BSA], 1-μL kDNA (0.2 μg), and 1-μL DMSO solvent or compounds at indicated concentrations, 1-μL Topo-II (2 units), and 13-μL water. Assays were incubated at 37°C for 30 minutes and stopped by the addition of 2-μL 10% SDS and Proteinase K (50 μg/mL; Zymo Research, Irvine, CA; cat#D3001-2). After incubating at 37°C for 15 minutes, samples were mixed with 1/10 volume of loading buffer and resolved on a 1% agarose gel.

Nuclear extracts were prepared from synchronized HaCaT cells after releasing into S-phase (at 20 hours postrelease) or HEK293T cells expressing stable Flag-Mcm2 protein (human). Cells from one 10-cm dish were resuspended in 300 μL of Buffer A [10-mmol/L Hepes–KOH pH 7.5, 10-mmol/L KCl, 1.5-mmol/L MgCl2, 0.34-mol/L sucrose, 10% glycerol, protease inhibitors (as for immunoblotting), and 0.1% Triton X100] and incubated on ice for 5 minutes. The nuclear pellet was obtained by centrifugation at 4,000 rpm at 4°C for 5 minutes and washed with Buffer A without Triton X100. The nuclear extract was obtained by resuspending the nuclear pellet in 300 μL of Buffer A containing 420-mmol/L potassium acetate and 0.01% Triton X100 and incubating at 4°C for 1 hour. The final concentration of potassium acetate in nuclear extracts was adjusted to 200 mmol/L for the coimmunoprecipitation assays. For the coimmunoprecipitation assays, 10 μg of indicated antibody (anti-Mcm2, anti-Psf1, or rabbit IgG), or 30 μL of mouse IgG agarose beads or anti-Flag M2 agarose beads, was incubated with 300 μL of nuclear extract at 4°C for 4 hours. Protein A/G agarose beads (30 μL, Santa Cruz, sc2003) were added and incubated for 1 hour with antibodies. Agarose beads were washed once in Buffer A containing 200-mmol/L potassium acetate and 0.01% Triton X100 and incubated with 15-µmol/L CA1 or DMSO (same concentration as in the CA1 sample) for 30 minutes. Beads were washed in the same buffer and analyzed by immunoblotting with indicated antibodies.

The human CMG helicase was purified following the established and validated protocol of Hurwitz and colleagues (26, 27), with minor modifications. A detailed description of the approach is shown here, to allow for comparisons with the methods described by the Hurwitz group.

High Five insect cells (2–3 million cells/mL, 1.5 L) were cultured in a shaking incubator at 27°C in suspension with ESF921 serum-free medium (Expression Systems, Davis, CA; cat#96-001-01) and then coinfected for 60 hours with 11 baculoviruses expressing human 6His2Flag-Cdc45, the human MCM hexamer (Mcm2, Mcm3, Mcm4, Mcm5, Mcm6, and Mcm7), and human GINS (GST-Sld5, Psf3, Psf2, and Psf1). Each virus was infected at a multiplicity of infection of ∼10 from individual virus stocks. Infected cells were harvested by centrifugation at 650 × g for 5 minutes at 4°C, washed with cold PBS, frozen on dry ice, and stored at −80°C until use. The cell pellet (∼20 mL) was thawed on ice, resuspended in 45-mL hypotonic buffer [20-mmol/L Hepes–NaOH (pH 7.5), 5-mmol/L KCl, and 1.5-mmol/L MgCl₂] with protease inhibitors (1-mmol/L PMSF, 1-mmol/L benzamidine, 0.15-µmol/L aprotinin, 4-µmol/L leupeptin, and 1-µmol/L antipain), and kept on ice for 10 minutes before lysing by Dounce homogenization (tight fitting, 60 strokes). The cell extract was adjusted to 0.42-mol/L potassium acetate and centrifuged at 43,000 × g for 1 hour at 4°C. The cleared lysate was mixed with 0.75-mL glutathione beads (cat#17-0756-05; GE Healthcare, Chicago, IL) preequilibrated with FEQ buffer [20-mmol/L Hepes–NaOH (pH 7.5), 0.42-mol/L potassium acetate, 5-mmol/L KCl, 1.5-mmol/L MgCl₂] and incubated by rotation at 4°C overnight. Following centrifugation at 290 × g at 4°C, the bound glutathione beads were washed four times (15 minutes each wash) with 40 mL of FW buffer [20-mmol/L Hepes–NaOH (pH 7.5), 0.42-mol/L potassium acetate, 1-mmol/L DTT, 1-mmol/L ethylenediaminetetraacetic acid (EDTA), 0.01% Nonidet P40 (NP40), and 10% glycerol (vol/vol) with protease inhibitors as above]. Bound proteins were eluted at 4°C three times (1 hour each elution) with 3.5-mL Q buffer [20-mmol/L Hepes–NaOH (pH 7.5), 1-mmol/L DTT, 1-mmol/L EDTA, 0.01% NP40, 10% glycerol (vol/vol), and protease inhibitors as above] containing 0.15-mol/L potassium acetate and 20-mmol/L reduced glutathione. The eluted fractions were combined and applied to a HiTrap Q-Sepharose Fast Flow column (HiTrap Q FF 1 mL; cat#17505301, Cytiva, Marlborough, MA), and preequilibrated 3X with 5 mL of Q buffer containing 0.15-, 0.75-, 0.15-mol/L potassium acetate. The Q-Sepharose FF column was washed with 10 mL of FW buffer containing protease inhibitors (as above). The proteins were eluted with 10 mL of Q buffer containing 0.75-mol/L potassium acetate. The eluted fraction was mixed with 0.15 mL of anti-Flag M2 affinity gel/beads and incubated while rotating at 4°C overnight. After centrifugation at 290 × g for 5 minutes at 4°C, the beads were washed three times with 10-mL PreScission buffer [50-mmol/L Tris-HCl (pH 7.5), 0.15-mol/L NaCl, 1-mmol/L DTT, and 1-mmol/L EDTA] for 15 minutes at 4°C and eluted at 4°C three times (1 hour each elution) with 0.2-mL PreScission enzyme buffer containing 0.2-mg/mL 3X-Flag peptides (cat#F4799, Sigma). The combined eluates were incubated with 20 Units of PreScission Protease (cat#270843; Cytiva) and 0.1-mL glutathione beads for 4 hours at 4°C. The supernatant was layered onto a 15% to 40% glycerol gradient [25-mmol/L Tris-HCl (pH 7.5), 50-mmol/L NaCl, 1-mmol/L DTT, 1-mmol/L EDTA, 0.01% NP40, protease inhibitors as above] in a 5-mL ultracentrifugation tube and centrifuged at 260,000 × g for 14 hours at 4°C. Glycerol fractions (0.15 mL each fraction) were collected from the bottom of the tube and stored at −80°C until use. Typically, fractions 6 to 9 contained complete hCMG (∼750–800 kDa) and comigrated with thyroglobulin.

Estimation of the hCMG protein amount isolated was determined by comparing it with BSA standards. We nominally achieved purification of 5- to 6-ng/µL hCMG enzyme from three glycerol fractions, or ∼7.5-fmol hCMG/µL. Across multiple preps, this varied from 5-fmol hCMG/µL to 15-fmol hCMG/µL, consistent with that reported by Hurwitz and colleagues (26, 27). The specific ATPase activity of our isolated hCMG enzyme was consistent with the hCMG ATPase activity obtained by Hurwitz and colleagues (26, 27). This prior hCMG analysis determined that the human helicase hydrolyzes ATP to ADP at a rate of ∼80 mol-ADP per minute per mol-hCMG in the presence of 500-µmol/L ATP. Using the ADP-sensing fluorescent polarization assay (described below), 15 fmol (in 2 µL) of hCMG elicits an ∼52% millipol (mP) change relative to the assay window (Results), equating to the production of ∼7- to 8-µmol/L ADP in 1 hour (80 × 15 fmol × 60 minutes = 72-pmol ADP in a 10-µL reaction or 7.2-µmol/L ADP). This activity represents an ∼1.5% ADP conversion rate by the hCMG and falls within the reliability range of Z′ = 0.6 to 0.8 for screening in the primary ATPase assay.

The simian virus-40 (SV40) large-T antigen (TAg) helicase was expressed in Sf9 cells by infecting the cells with 6His2Flag-TAg baculovirus. The Sf9 cells (300 mL) were grown at 27°C in suspension with ESF921 serum-free medium and infected with 6His2Flag-TAg baculovirus at a density of 2 to 3 million cells/mL. After 70 hours, the cells were harvested by centrifugation at 500 × g for 10 minutes at 4°C, washed with 30 mL cold PBS, and then frozen on dry ice and stored at −80°C until use. The frozen cell pellet was thawed on ice and resuspended with 8-mL hypotonic buffer (20-mmol/L Hepes–NaOH (pH 7.5), 5-mmol/L KCl, and 1.5-mmol/L MgCl₂) containing protease inhibitors (1-mmol/L PMSF, 1-mmol/L benzamidine, 0.15-µmol/L aprotinin, 4-µmol/L leupeptin, and 1-µmol/L antipain). The cell suspension was lysed by Dounce homogenization for 60 strokes, and the lysate was adjusted to 0.42-mol/L potassium acetate followed by centrifugation at 18,800 rpm (SW55Ti rotor) for 60 minutes at 4°C. The cleared lysate was mixed with 0.4 mL of anti-FLAG M2 affinity gel preequilibrated with FEQ buffer [20-mmol/L Hepes–NaOH (pH 7.5), 0.42-mol/L potassium acetate, 5-mmol/L KCl, and 1.5-mmol/L MgCl₂] and incubated and rotated overnight at 4°C. With centrifugation at 290 × g for 5 minutes at 4°C, the beads were washed three times with 10-mL FW buffer [20-mmol/L Hepes–NaOH (pH 7.5), 0.42-mol/L potassium acetate, 1-mmol/L DTT, 1-mmol/L EDTA, 0.01% NP40, and 10% glycerol (vol/vol) with protease inhibitors]. The bound TAg was eluted three times with 2-mL Q buffer [20-mmol/L Hepes–NaOH (pH 7.5), 1-mmol/L DTT, 1-mmol/L EDTA, 0.01% NP40, and 10% glycerol (vol/vol) with protease inhibitors] containing 0.15-mol/L potassium acetate and 0.1-mg/mL 3 × FLAG peptides. The eluted proteins were combined and applied to a Q-Sepharose Fast Flow column (HiTrap Q FF 1 mL, Cytiva) preequilibrated with Q buffer containing 0.15-mol/L potassium acetate. The Q column was washed with 5-mL Q buffer containing 0.3-mol/L potassium acetate and eluted with 5-mL Q buffer containing 1-mol/L potassium acetate. The elution was further applied to an Ultra15 centrifugal unit (cat#UFC905096, 50 K, Millipore, St. Louis, MO) and centrifuged at 3,800 × g at 4°C until the volume is 250 μL. After adding 15-mL storage buffer [20-mmol/L Tris-HCl (pH 8.0), 200-mmol/L NaCl, 2-mmol/L DTT, 2-mmol/L EDTA], the centrifugal unit was centrifuged again until the final volume is 250 μL. For storage, 250 μL of 100% glycerol was added to the final TAg protein solution and stored at −80°C.

The human papillomavirus (HPV) HPV16- or HPV18-E1 helicase domains (16E1HD or 18E1HD) were expressed and purified from Sf9 cells by infecting the cells with 2XFlag-GST-E1 baculoviruses. The cleared cell lysates were prepared for TAg and incubated with glutathione beads at 4°C overnight. The bound E1 proteins were eluted from beads with 20-mmol/L reduced glutathione in Q buffer, applied to a Q-Sepharose Fast Flow column, and eluted with 0.75-mol/L potassium acetate in Q buffer. Purified E1 proteins were obtained by further incubation with anti-Flag-M2 beads and elution with 3X-Flag peptides.

Helicase fork-unwinding assays were performed at 37°C for 1 hour in 20-µL reaction volumes using 2 to 4 µL of hCMG (∼15–30 fmol), in a buffer consisting of 25-mmol/L Hepes–NaOH (pH 7.5), 5-mmol/L NaCl, 0.5-mmol/L ATP, 10-mmol/L magnesium acetate, 1-mmol/L DTT, 0.1-mg/mL BSA, and ∼10 fmol of radio-labeled DNA forks (27). Reactions were terminated with 4 µL of 6X stop solution [50-mmol/L EDTA (pH 8.0), 40% (vol/vol) glycerol, 2% (wt/vol) SDS, and 0.3% bromophenol blue], loaded onto 10% (wt/vol) polyacrylamide gels, resolved at 150 V in 1X-TBE buffer (89-mmol/L Tris base, 89-mmol/L boric acid, and 2-mmol/L EDTA), dried on filter paper, analyzed by autoradiography, and quantified by PhosphorImager assessment. Assays using TAg helicase or HPV-E1 helicases were performed in the same conditions as those used for the hCMG helicase.

The DNA substrates were formed as described by Hurwitz and colleagues (27) by annealing two oligonucleotides: 10 pmol of M13-39–5′dT40 (5′-(T)₄₀GAT​TAA​GTT​GGG​TAA​CGC​CAG​GGT​TTT​CCC​AGT​CAC​GAC-3′) and 10 pmol of anti-M13-39–3′dT40 (5′-GTC​GTG​ACT​GGG​AAA​ACC​CTG​GCG​TTA​CCC​AAC​TTA​ATC(T)₄₀-3′) in the presence of 0.1 mol/L NaCl by heating for 5 minutes at 95°C, followed by slow cooling to room temperature. The oligonucleotide M13-39–5′dT40 was 5′ end-labeled with T4 polynucleotide kinase (cat#M0201S; New England Biolabs, Ipswich, MA) and [γ-³²P]-ATP before annealing. The annealed DNA substrates were resolved in 10% polyacrylamide gels in 1X-TBE buffer at 150 V for 30 minutes. Bands containing double-stranded DNA forks were cut from the gel and forks eluted after crushing the gel with 200-µL Tris-EDTA buffer (10 mmol/L Tris base, 1 mmol/L EDTA, pH8.0) containing 0.15-mol/L NaCl for 3 hours at 37°C. The eluted fraction was collected by centrifugation at 13,000 rpm (microfuge) at 4°C for 10 minutes and stored at 4°C until use.

The fluorescent polarization (FP) ADP-sensing assays were performed using the Transcreener ADP² FP assay kit (cat#3010-1 K, BellBrook Labs, Madison, WI). For hCMG inhibitor screening, assays were performed at 37°C for 1 hour in 10-µL reactions in a 384-well plate (cat#4514, Corning, Corning, NY) using 2 µL of hCMG (∼15 fmol), 25-mmol/L Hepes–NaOH (pH 7.5), 10-mmol/L NaCl, 0.5-mmol/L ATP, 10-mmol/L magnesium acetate, 1-mmol/L DTT, 0.1-mg/mL BSA, and DNA fork substrates. However, assays without DNA substrates can also be performed, because the hCMG does not require DNA to be present to hydrolyze ATP (27). Selected inhibitors or samples from a chemical library were added to the reactions when conducting the screening or testing inhibitor effectiveness. The NCI Diversity Set VI chemical library was obtained from the NCI (Bethesda, MD).

The window of sensitivity for the ADP-sensing assays is determined by setting up two 10-µL control samples without any added hCMG helicase (using reagents from the Transcreener kit above): a low-FP mixture (4-nmol/L ADP Alexa Fluor633 Tracer alone) and a high-FP mixture (4-nmol/L ADP Alexa Fluor633 Tracer plus patented anti-ADP² antibody). The amount of ADP² antibody used had to be adjusted to account for the use of 500-µmol/L ATP in our hCMG assays versus 10-µmol/L ATP in enzyme reactions typically assessed with standard kits prepared by BellBrook Labs. The ADP² antibody has a considerably higher affinity for ADP compared with ATP, but because it can bind to ATP to some extent, this must be offset by including more of the ADP² antibody in our high-ATP assays. This was done according to the manufacturer by performing a titration with increasing ADP² antibody, fixed 4-nmol/L ADP Alexa Fluor633 Tracer, and 500 µmol/L ATP to determine the ∼EC_80–85 for mP changes, which determined the optimal ADP² antibody concentration to use as 0.64 mg/mL (Supplementary Fig. S1A). Samples are read with a Perkin Elmer Envision II plate reader with optimized Cy5 (far-red) FP-compatible mirror and cubes (cat# 2100-8390, Perkin Elmer, Waltham, MA). The low-FP sample is the least polarized and gives a low mP reading, whereas the high-FP sample is the most polarized and gives the highest mP reading. The difference between the low-FP and high-FP values defines the FP assay window, which is normally in a range of 150 to 200 mP under ideal conditions for screening purposes. The hCMG helicase hydrolyzes ATP to ADP and decreases the mP reading within this window, with an ideal change (∆mP) in the FP window of at least 50% to be in a readable range for inhibitor screening (as per manufacturer). Potential hCMG chemical inhibitors reverse this effect and cause the mP readings to increase.

We performed a titration with increasing and decreasing concentrations of ADP and ATP, respectively, to assess the sensitivity of the FP assay in detecting ADP production (Supplementary Fig. S1B). Consistent with the stated manufacturer predictions, the assay can reliably detect 1% to 3% changes in ADP production (i.e., 5- to 15-µmol/L ADP production) in starting concentrations of ATP of 500 µmol/L, with a Z′ efficiency between 0.6 to 0.8 (determined following manufacturer protocols). We also note that ATP-gamma-S cannot be used as a positive control for inhibition of the CMG in this ADP-sensing assay, as it competes with the anti-ADP² antibody and alters the detection window by itself (Supplementary Fig. S1C).

Once the hCMG enzyme reactions are complete, 10 µL of 1X Stop Solution and Detection Buffer (buffer from BellBrook Labs Transcreener kit; containing 4 nmol/L ADP Alexa Fluor633 Tracer plus ADP² antibody) are added to each reaction, and to the high-FP control. The low-FP control receives 10 µL of 1X Stop Solution and Detection Buffer containing only 4 nmol/L ADP Alexa Fluor633 Tracer (no ADP² antibody). Samples are incubated at 25°C for 1 hour, then read in the plate reader. Potential positive inhibitors (hits) of the hCMG are verified to be incapable of altering the mP window on their own, by artificially raising the mP readings to seem more polarized as occurs when the hCMG is truly inhibited by a compound. For this, potential hCMG inhibitors are added to a 10-µL mixture containing 25-mmol/L Hepes–NaOH (pH 7.5), 10-mmol/L NaCl, 0.5-mmol/L ATP, 10-mmol/L magnesium acetate, 1-mmol/L DTT, 0.1-mg/mL BSA, and no hCMG and mixed with 10 µL of Stop Solution and Detection Buffer (with 4-nmol/L ADP Alexa Fluor 633 Tracer, without ADP² antibody). Plate readings from these tests of potential positive hits should be similar to that seen with the low-FP control, or such hits may instead be false positives.

The structure of the human CMG helicase determined at 3.3 Å by electron cryomicroscopy (cryo-EM) with bound ADP, ATP-gamma-S, or no nucleotide within individual ATP clefts and with bound DNA (PDB accession code 6XTX) was used for in silico docking with Autodock Vina Version 1.2.3 (RRID:SCR_011958; ref. 28). Estimated interaction/binding energies were calculated by Autodock. The conversion of protein and small chemical molecules into PDBQT format and grid box generation was achieved by using Autodock Tools Version 1.5.6. Each pair of MCMs forming an ATP cleft was extracted, and the grid box was derived with the active site in the center of the cleft. The docking configurations of clorobiocin were viewed and analyzed using Pymol Version 2.4.1 (RRID:SCR_000305).

All steps involved in the synthesis and nuclear magnetic resonance and high-performance liquid chromatography verification of synthetic chemical compounds tested in this report are shown in Supplementary Chemical Synthesis Methods.

All the statistical details of experiments can be found in the figure legends or Materials and Methods descriptions. Statistical analyses were performed with Prism (GraphPad, La Jolla, CA). Error bars represent mean ± SD. Statistical comparisons were analyzed with an unpaired two-tailed Student t test. Data were considered statistically different at P < 0.05.

All original data used in this report is available for examination by other scientists. Assistance with data analysis or methodology used in this report can also be obtained by communication with the corresponding author.

We developed a rigorous biochemical screening approach utilizing two rounds of primary screening with a commercially available ATP hydrolysis assay and a secondary orthogonal validation assay that measures DNA unwinding by the human CMG (hCMG; Fig. 1A). The hCMG helicase was purified using the established protocol of Hurwitz and colleagues in which all 11 hCMG subunits are coexpressed using baculoviral-based infections of insect cells followed by a multistep purification of the hCMG holoenzyme (Fig. 1A; refs. 26, 27). The quality of the hCMG enzyme obtained was verified by silver staining and immunoblotting, which showed that all 11 hCMG subunits were present at similar stoichiometries and purity compared with that obtained in previous studies (Supplementary Fig. S2A and S2B; refs. 26, 27). The hCMG isolated is active in DNA fork-unwinding (helicase) assays and is dependent on the binding and hydrolysis of ATP as indicated by a dose-dependent suppression of fork-unwinding activity in the presence of slow-hydrolyzable ATP-γ-S (Supplementary Fig. S2C). For fork-unwinding activity, the hCMG displays an ∼K_m of 690 µmol/L (ATP; below), in close agreement with the K_m [625 µmol/L (ATP)] for hCMG ATP hydrolysis activity described by others (26, 27). The specific ATPase activity of the isolated hCMG also closely matches that obtained by Hurwitz and colleagues (Materials and Methods; refs. 26, 27).

We optimized the FP Transcreener ADP² assay (BellBrook Labs, Madison, WI) for quantitative analysis of ADP production by the hCMG (Materials and Methods; Supplementary Fig. S1A–S1C). The assay relies on a patented anti-ADP² antibody that binds and polarizes an ADP-Tracer, changes to which due to competition with ADP produced by the hCMG are measurable in FP plate readers. This assay is highly sensitive and reliable (Z′ > 0.6; Supplementary Fig. S1B), being able to quantify small changes in ADP production (29–32). We titrated purified hCMG into the assay, which produced a dose-dependent increase in the ADP production (Fig. 1B). The hCMG concentration that produced ∼50% change in the assay window was used for screening. Because remarkable quantities of hCMG are required for screening, we performed primary chemical library screening with hCMG purified through the Flag enrichment step, and a repeat of primary screening with positive hits on a higher purity (but lower yield) hCMG after glycerol fractionation (Fig. 1A). The hCMG obtained after the Flag enrichment step is active in the primary assay and is dependent on the presence of intact hCMG helicase, as failure to express Mcm4 yields preparations devoid of ATPase activity (Fig. 1C). The latter indicates that a contaminating ATPase from insect cells is not present in our hCMG preparations during our screening.

We used the National Institutes of Health (NIH) Diversity Set VI library at 1-mmol/L concentrations for primary screening. Fewer than 3% of compounds were capable of partial or complete hCMG inhibition on initial testing. Many of these initial hits were intercalating agents or potentially reactive compounds based on medicinal chemistry assessment, whereas others interfered with assay analytes (e.g., anti-ADP² antibody or tracer). Such compounds were defined as false positives and were not assessed any further. An important limitation going forward was the need for fresh dry powder and/or a sufficient quantity of any potential CMG inhibitor to test compounds in additional assays. Based on these criteria, only a few compounds were assessed in a repeat of primary screening at 500 µmol/L with the limited amounts remaining from the NCI library. One compound, clorobiocin, drew attention due to its drug-like features and effective nature of hCMG inhibition (Fig. 1D). Clorobiocin did not interfere with far-red UV light readings or assay reagents. Clorobiocin is an aminocoumarin derived from Streptomyces roseochromogenes and is related to two similar chemicals, novobiocin and CA1 (21). It was difficult to continue working with clorobiocin, as we needed fresh dry chemical powder for further validation work, the NIH had none available, and we could not find a commercial source. However, CA1 and novobiocin were commercially available, and we tested both for their ability to inhibit the hCMG. CA1 was found to be a potent inhibitor of ATP hydrolysis by the hCMG, whereas novobiocin had a very little inhibitory effect on the hCMG (Fig. 1D). For practical reasons, these aminocoumarins became the primary focus of our CMG inhibitor analyses due to availability and intriguing structure–activity information as described below.

Validation with our secondary strand-displacement (helicase) assay showed that clorobiocin and CA1, but not novobiocin, were effective hCMG helicase inhibitors at 500-µmol/L concentrations (Fig. 1E). The in vitro IC₅₀ of CA1 for hCMG helicase inhibition was determined to be ∼15 µmol/L (Fig. 1F), which closely matches the IC₅₀ of CA1 for reducing viability of human cells (below). Using the FP assay and an ADP/ATP standard curve comparison, the IC₅₀ for hCMG ATP hydrolysis inhibition is ∼85 µmol/L (Fig. 1G). The hCMG has six distinct ATPase clefts and it is likely that CA1 does not target all of them with the same efficiency (below). It is therefore possible that CA1 targets a cleft(s) necessary for helicase activity at higher affinity, but more CA1 is necessary to inhibit remaining ATP sites.

Using yeast MCM complexes as the basis for screening, it has been suggested that ciprofloxacin might be an inhibitor of the human replicative helicase (33, 34). However, high concentrations of ciprofloxacin or other quinolones do not inhibit the purified hCMG helicase (Supplementary Fig. S3A). CA1 is known to inhibit the bacterial type-2 topoisomerase, gyrase (21). However, relative to that required to block CMG activity, inhibition of human Topo-II in decatenation assays in vitro requires ∼10-fold higher CA1 concentration (Supplementary Fig. S3B). To assess the selectivity of CA1 toward the CMG helicase, we tested whether CA1 could inhibit two related mammalian hexameric ring helicases with notable structural similarity to the CMG, the Large-T antigen (TAg) helicase from SV40 virus, and E1 helicases from HPVs. Purified TAg helicase and E1 helicases from HPV16 and HPV18 are sensitive to CA1, but require higher CA1 concentrations (TAg IC₅₀ ∼70 µmol/L; HPV16-E1 IC₅₀ ∼170 µmol/L; Supplementary Fig. S3C and S3D). These results indicate that CA1 displays in vitro selectivity toward the CMG helicase at low concentrations. Although it is possible CA1 might affect additional helicases (or other ATPases), the remainder of this report will focus on understanding the chemical warhead and mechanisms for CA1 as a CMG inhibitor, as CA1 and clorobiocin represent the first biochemically validated small chemical compounds that effectively inhibit ATPase and helicase activities of the hCMG (defined as CMGi).

We determined the mechanism of hCMG inhibition by CA1. The three aminocoumarins are comprised structurally of a noviose sugar “head” group joined to a coumarin group, and an amide group in two of the molecules (Fig. 2A; coumarin-amide domain referred to here as the “tail” of the molecule; ref. 21). Clorobiocin and CA1 contain 2-methylpyrrole ester modifications to the sugar (Fig. 2A; orange arrows) whereas novobiocin has a primary carbamate modification (Fig. 2A; blue open arrow). CA1 resembles a tail-tail dimer of clorobiocin but replaces the chlorine with a methyl. Because novobiocin has little inhibitory effect on the hCMG in the biochemical assays (only at high concentrations) and largely differs from the other compounds in the modification of its sugar, this structure–activity relationship (SAR) indicates that the sugar head groups of clorobiocin and CA1 (Fig. 2A; purple boxes) provide chemical specificity in mediating inhibition of the hCMG.

Biochemical and cocrystallographic data assessing how these aminocoumarins inhibit gyrase (a type-II topoisomerase) provide information on how CA1 can inhibit the hCMG (21, 35–37). Aminocoumarins inhibit ATP binding and hydrolysis of gyrase using a competitive mechanism, inserting the sugar head groups through a channel/groove into the ATPase cleft of the GyrB subunit, with the sugar situated in the region where the adenosine and ribose of ATP normally interact (21, 35). CA1 interacts with two GyrB ATPase domains at the same time using this mechanism (21, 35). We reasoned that CA1 might likewise inhibit hCMG ATPase and helicase activities by direct competition with ATP binding and hydrolysis. We performed hCMG helicase assays to determine the mode of hCMG inhibition by CA1 in increasing ATP concentrations (Fig. 2B). Michaelis–Menten kinetics and Lineweaver–Burke (double-reciprocal plot) analyses showed that CA1 inhibits hCMG helicase activity using a classic ATP-competitive mechanism, resulting in a substantial increase in the K_m for (ATP; 690 µmol/L without inhibitor to 1,550 µmol/L with CA1), without a change in the V_max of the hCMG. This notable shift in K_m explains why low micromolar concentrations of CA1 are capable of inhibiting the hCMG in the presence of high ATP concentrations, as occurs in a cellular environment (Fig. 2B; Supplementary Fig. S4).

We next used in silico docking of CA1 and clorobiocin to model how these compounds can interact with the ATPase domains of the hCMG to competitively block ATP binding and hydrolysis. The hCMG has six biochemically distinct ATPase domains formed between adjacent MCM subunit pairs (4, 38–40) and the cryo-EM structure of the hCMG has been determined (28). Docking software places CA1 and clorobiocin into channels leading to the ATPase clefts of three MCM ATPase domains with similar binding energies, notably clefts for Mcm3–Mcm7, Mcm4–Mcm6, and Mcm5–Mcm3 (Fig. 2C and D, CA1 docking; Supplementary Fig. S5A–S5D, clorobiocin docking). Consistent with cocrystallographic data for aminocoumarin-gyrase interactions (21, 35–37), the sugar head groups are inserted into the ATP-binding sites where adenosine and ribose from ATP are normally situated (“sugar-first” direction), whereas the coumarin group occupies channels leading to the ATP-binding sites. CA1 can be docked in either direction in these MCM ATPase clefts/channels due to its symmetry. It is quite possible that these channels are used by ATP or ADP for ingress and/or egress during enzyme function, suggesting that clorobiocin and CA1 act like a “cork in a wine bottle” to block ATP movement through the channels, consistent with their ATP-competitive mode of hCMG inhibition. Three MCM ATPase domains were not capable of in silico docking for either compound, specifically the clefts for Mcm7–Mcm4, Mcm6–Mcm2, and Mcm2–Mcm5. In the existing hCMG structure (28), the channels leading to these ATPase clefts are narrow relative to the MCM sites that can be docked, suggesting a steric hindrance to inhibitor binding (Supplementary Fig. S5E and S5F). The cryo-EM structure of the hCMG is in one enzymatic state, and it remains possible that these sites might also be subject to inhibition by clorobiocin or CA1 under different enzymatic states when these channels might be accessible.

We prepared several compounds to support the SAR and modeling results obtained with the CMGi identified in our screen. A compound referred to herein as methylbiocin (MBC) was synthesized that is identical to clorobiocin, except for the replacement of the chlorine with a methyl, and is thus half of the CA1 molecule (Fig. 2A; synthesis of all compounds and nuclear magnetic resonance purities shown in Supplementary Chemical Synthesis Methods; refs. 41–43). Derivatives of MBC were also synthesized: a noviose sugar-pyrrole compound (MBC-D1), a coumarin–benzamide “tail” compound (MBC-D2), and a compound comprised the coumarin–benzamide tail with the noviose sugar (MBC-D3). All compounds were initially tested at 500-µmol/L concentrations to assess for inhibition of hCMG helicase activity (Fig. 2E). The MBC compound and CA1 both potently inhibit the hCMG, whereas novobiocin and the sugar-pyrrole compound (MBC-D1) show only a small level of inhibition. Interestingly, the tail compound (MBC-D2) is somewhat effective at inhibiting the hCMG, but this inhibition of the hCMG is diminished when the noviose sugar is added (Fig. 2E, MBC-D3). The IC₅₀ for MBC was ∼59 µmol/L, which is lower than that for CA1 (Fig. 2F). The IC₅₀ for the other derivatives was not determined because these compounds elicited only partial hCMG inhibition at high micromolar concentrations. These SAR results demonstrate that inhibition of hCMG activity is derived from a chemical warhead comprised the coumarin moiety (“tail”) linked to the noviose–pyrrole (sugar “head”), which modeling suggests cooccupy the channel and ATP-binding site(s) to competitively limit ATPase function. CA1 contains two of these MBC warheads, which may partly explain why CA1 is a better inhibitor than MBC.

At present, we do not know if certain MCM ATPase clefts display preferences for inhibitor binding over other clefts, particularly in cells, as determining this is technologically challenging for an enzyme comprised six ATPase domains that are not easily separated for analysis. Going forward, we assess hCMG ATPase/helicase inhibition and chemical biology in cells with commercially available CA1 from a holo-enzyme perspective (ATP-competitive effects on all clefts combined) rather than a particular ATPase cleft.

We next investigated the molecular mechanisms by which CA1/CMGi inhibited cell growth, DNA replication, and CMG helicase assembly/function under nontumor conditions in human cells. For this, we used the immortalized, nontumor derived human keratinocyte HaCaT cell line, which can be efficiently synchronized in a quiescent state by serum withdrawal and released into the cell cycle after readding serum to study early- and late-G1 events, as well as the S-phase effects of CA1. This also allowed analysis of CA1 mechanisms at specific cell cycle times without influences of other synchronization protocols that may elicit checkpoint responses (e.g., double-thymidine blocks or nocodazole mitotic release). A cell viability analysis found that CA1, but not novobiocin, reduces HaCaT viability with an IC₅₀ of ∼15 µmol/L (Fig. 3A). This concentration of CA1 is similar to that required to inhibit in vitro hCMG helicase activity. However, cellular IC₅₀ determinations confer important requirements on a compound, including serum binding, the ability to cross membranes, stability in cells, and success at finding targets in larger complexes. Such caveats often result in higher concentrations of target-specific inhibitors being necessary to block cell growth versus in vitro determinations. Offsetting this, the number of CMG helicases (perhaps fewer) that must be inhibited to suppress cell growth could differ from the total enzyme activity inhibited in vitro at similar concentrations, resulting in cellular responses at lower levels of CA1. Although the concentrations of CA1 necessary for both in vitro CMG inhibition and in vivo HaCaT cell inhibition are a result of different requirements for each outcome, the similar range of sensitivity for both is consistent with the hCMG being a target of CA1 in human cells.

As shown previously (24), HaCaT cells are efficiently synchronized by serum deprivation for 2 days (Fig. 3B). Using BrdU labeling and immunofluorescent measurements to assess DNA replication in synchronized HaCaT cells, the start of G1 occurs at time 0 (serum readdition), the G1/S transition occurs at ∼15 hours postrelease into the cell cycle (when ∼50% of population is labeled), and the peak of S-phase occurs during the 18 to 20 hours window (Fig. 3B; ref. 24). CA1 inhibits DNA replication when added to HaCaT cells in early G1 (at time 0), whereas novobiocin has little effect (Fig. 3B), consistent with assembly and activity of hCMG complexes being required for G1 progression into the S-phase.

Studies of MCM assembly using yeast in vitro models have suggested that ATP binding and hydrolysis by most MCM ATPase clefts are required for efficient Mcm2–Mcm7 ring loading onto DNA (39, 40). We asked whether CMGi/CA1 could block chromatin/DNA binding of human MCM complexes in vivo due to a dependency on ATP utilization. Synchronized HaCaT cells were treated with CA1, novobiocin, or DMSO carrier in early-G1 (at release, 0 hours), middle-G1 (6 hours), or late-G1 (12 hours) to assess the effects of CMGi on different stages of MCM assembly (Fig. 3C). MCM loading onto chromatin in human cells is considerably inhibited by early-G1 CA1 treatment (Fig. 3D). GINS and Cdc45 loading onto chromatin are consequently blocked, whereas Orc2 chromatin binding is not affected. This suggests that the ORC complex, which contains ATPase domains required for its DNA binding and roles in MCM loading, is not itself a target of CA1.

Some MCM complexes are already loaded onto chromatin between early and middle G1 (3–10 hours after release), but an increase in MCM loading occurs around 12 hours as cells approach G1/S (Fig. 3E). Although exposure of cells to CA1 at 6 hours does not affect MCMs already loaded, the increase in MCM loading at 12 hours is inhibited by CA1 but not by novobiocin. Cdt1 also loads onto chromatin at higher levels when MCM loading increases. However, CA1 blocks this Cdt1 loading and promotes loss of total Cdt1 (Fig. 3E). Taken together with the previous experiment, these results indicate that CMGi inhibit an early step in the MCM loading process in human cells that requires efficient ATP binding and/or hydrolysis by Mcm2–Mcm7, but once loaded, MCMs are resistant to CMGi/CA1. The results also suggest that Cdt1 is sensitive to Mcm2–Mcm7 ATPase inhibition in human cells, consistent with yeast studies showing that MCM–Cdt1 interactions are adversely affected by defective Mcm2–Mcm7 ATPase sites (39, 40).

Orc2 and Orc4 affinity for chromatin is not affected by CA1 exposure in middle G1 (Fig. 3E), suggesting that the ATPases of ORC are not a target of CA1. Cdc6 protein is not affected in chromatin association by CA1 when measured using a polyclonal antibody (Fig. 3E). In contrast, analysis with a monoclonal antibody to Cdc6 suggests that one form of Cdc6 increases on chromatin in parallel with MCM elevation and is sensitive to CA1. Cdc6 contains an ATPase domain that is not required for MCM assembly on DNA but for removal of improperly loaded MCMs (39, 40). Although we cannot rule out the possibility that the ATPase site of Cdc6 may be affected by CA1, these prior studies suggest that it is unlikely that this would contribute to the inhibition of MCM assembly we have observed in the presence of CA1.

Synchronized HaCaT cells treated with CA1 in late-G1 (treated at 12 hours), but not novobiocin, fail to undergo DNA replication (Fig. 3F). Near the G1/S transition (15–18 hours) there is additional MCM loading onto chromatin, which coincides with GINS and Cdc45 being recruited to loaded MCM hexamers on chromatin (Fig. 3G). Treatment with CA1 during this late-G1 period has only a small effect, if any, on the remainder of MCM loading and does not affect Cdt1 dynamics, suggesting that Mcm2–Mcm7 ATPase functions in MCM/Cdt1 loading are no longer required at this later time. However, whereas Cdc45 is not appreciably affected, CA1 inhibits GINS recruitment to DNA-loaded MCM hexamers (Fig. 3G). Such results are consistent with yeast in vitro studies showing that certain ATPase sites of the Mcm2–Mcm7 ring are required for GINS binding (40). However, our results differ somewhat from another yeast study showing that GINS and Cdc45 are both dependent on ATP binding to the Mcm2–Mcm7 ring (44). Reasons for this difference may be that CA1 is less efficient at binding a particular MCM ATPase cleft involved in Cdc45 recruitment in human cells, or that the Cdc45 extraction conditions vary between experimental approaches.

There are other enzymes with ATPase domains that function in MCM/CMG assembly, including Dbf4-Cdc7 (DDK) and Cdk2 (2). We asked whether CA1 affected these and other enzymes in human cells. DDK phosphorylates two sites in Mcm2 (S53 and S139) to facilitate Cdc45 recruitment (45), both of which show no phosphorylation changes after extended exposure to CA1 (Fig. 4A). This agrees with our observation that Cdc45 is recruited to DNA-loaded MCM hexamers (Fig. 3G) and indicates that DDK is not a target of CA1. A pan-Cdk inhibitor (AT7519) that efficiently targets Cdk1 (Cdc2), Cdk2, Cdk3, Cdk4, Cdk6, and Cdk9 blocks phosphorylation of Cdk2 targets, including Rb and Cdc6 (S54P; refs. 25, 46), and the Cdk1 target PP1α (47). However, extended exposure to CA1 has no effect on these substrates (Fig. 4B). We conclude that CA1 does not target these kinases required for MCM/CMG assembly. Although we cannot exclude the possibility that other unknown CA1 targets exist in cells, particularly at higher concentrations, these results support that the effects of CA1 on MCM/CMG dynamics in cells are due primarily to the targeting of the Mcm2–Mcm7 ATPases.

We determined how CMGi exposure affected the dynamics of hCMG and replisome structure during S-phase. Synchronized HaCaT cells were treated with CA1, novobiocin, or DMSO once cells reached early S-phase (18 hours) and immunoblots were performed assessing chromatin-bound and total protein components of the replication machinery (Fig. 5A). Etoposide has no effect on hCMG helicase activity in vitro (Supplementary Fig. S3B) and was included to compare how inhibition of Topo-II affected replication dynamics. Treatment with CA1 and etoposide effectively suppressed DNA replication, whereas novobiocin/DMSO did not (Fig. 5B), confirming that hCMG and Topo-II activities are required for ongoing DNA replication. MCM association with chromatin/DNA was not affected by any compounds (Fig. 5C; left). However, GINS and Cdc45 chromatin association was notably suppressed by CA1, and not by novobiocin or etoposide. Total protein levels were slightly affected for Psf3 and Cdc45 but not for other subunits (Fig. 5C; right). We conclude from these results that DNA topological issues and DNA replication arrest due to Topo-II inhibition do not disrupt GINS/Cdc45 interactions with hCMG helicases. However, CMGi inhibition of ATP binding and/or hydrolysis by the Mcm2–Mcm7 ATPases results in GINS and Cdc45 dissociation from hCMG helicases during ongoing DNA replication. Two related interpretations are possible. CA1 could block GINS/Cdc45 from being recruited or cause dissociation after recruitment (the latter tested below).

Structural studies have shown that components of the human replisome interact directly with the hCMG helicase, mediated in part through GINS and Cdc45 (48, 49). We asked if CMGi-induced loss of Cdc45 and GINS from hCMGs resulted in disruption of replisomes in human cells. Treatment of the S-phase cells with CA1 but no other compounds caused a loss of DNA polymerase-α, DNA polymerase-δ, and DNA polymerase-ε from chromatin (Fig. 5C). Factors such as Ctf4 and Mcm10, which interact with the hCMG and facilitate DNA polymerase-α function on the lagging strand (48, 49) are also reduced on chromatin by CA1. Consistent with hCMG and replisome disruption, RPA (single-stranded binding protein) is also reduced on chromatin. Intriguingly, exposure to etoposide, which stops DNA replication, does not diminish replisome components or RPA on chromatin, except for a small change to DNA polymerase-α and DNA polymerase-δ (Fig. 5C). We conclude that although inhibition of Topo-II (and DNA replication) does not have a notable negative effect on replisome integrity, the structural cointegrity of replisomes and hCMGs is dependent on ATP binding and/or hydrolysis by the Mcm2–Mcm7 ATPases during ongoing DNA replication in human cells and is disrupted by CMGi.

We next determined if CMGi treatment of partially purified hCMG helicases and replisomes from human cells were sensitive to CA1 after such complexes had formed. Nuclear extracts were prepared from HaCaT cells enriched in the S-phase and subjected to immunoprecipitation using antibodies to Psf1 or Mcm2. Immunoprecipitated complexes were treated directly with CA1 or DMSO, followed by immunoblotting for associated proteins (Fig. 5D). Psf1 associates with Psf2, Psf3, Mcm2, and Mcm6, indicating that hCMG helicases were extracted from cells (Fig. 5D; middle). We could not examine Cdc45 in this experiment due to signal interference with IgG on immunoblots. Treatment with CA1 did not disrupt Psf1–Psf3 interactions, indicating that the GINS complex itself is not abrogated by CA1. However, Mcm2 and Mcm6 interactions with GINS are abolished by CA1. Mcm2 associates with Mcm6, Mcm7, and Cdc45, and CA1 treatment causes Cdc45 to dissociate from MCMs but does not disrupt MCM complexes (Fig. 5D; right). We performed a similar experiment using a different human cell line (HEK293T) expressing ectopic Flag–Mcm2 (Fig. 5E). Flag–Mcm2 interacts with endogenous Mcm7, Psf1, Cdc45, DNA polymerase-ε, and Ctf4 (Fig. 5E; right), indicating that Flag–Mcm2 forms complexes with hCMG and replisome components in human cells. CA1 does not disrupt MCM interactions but displaces Psf1, Cdc45, and Ctf4 from Flag–Mcm2. Interestingly, CMGi does not disrupt DNA polymerase-ε binding to Flag–Mcm2, suggesting that differences exist between replisome–hCMG interactions in vivo and in vitro. A possible explanation is that in vivo other factors may contribute to replisome disassembly, such as the ubiquitin ligase CUL2(LRR1), which contributes to disassembly (50).

To further support the previous findings with human cell-based testing of hCMG complexes, we expressed and purified hCMG enzyme using baculoviral methods and insect cells and assessed isolated hCMG complexes for CA1 sensitivity (Fig. 5F). hCMG containing GST–Mcm7 was bound to glutathione beads and then treated with CA1 or DMSO, followed by immunoblotting for proteins retained on beads. Consistent with previous findings, the GINS (Psf2/3) complex and, to a lesser extent Cdc45, are dissociated from isolated hCMGs specifically following CA1 exposure. However, CA1 does not disrupt MCM proteins (Fig. 5F). These experiments altogether demonstrate that CMGi disrupt the structural cointegrity of hCMG and replisome components after complexes have formed. MCM hexamers are not disrupted by CMGi, consistent with yeast studies using MCM ATPase mutants (39, 40). However, interactions of GINS, Cdc45, DNA polymerases, and cofactors with hCMG helicases depend on functional MCM ATPase domains during S-phase that are inhibited by CMGi exposure.

Oncogenic signals (e.g., Myc or Cyclin E overexpression) cause RS and reduce MCM/CMG reserve functionality (1). Because MCM/CMGs are required to recover from RS but are debilitated at the same time, vulnerabilities to CMGi likely exist in solid tumor cells driven by certain oncogenes. K-Ras-driven tumor cells contain RS (e.g., replication fork stalling) and have been shown genetically to be selectively sensitive to the reduction of MCM licensing, which is synthetically lethal (18–20). This predicts that K-Ras-driven tumor cells will be selectively sensitive to pharmacologic treatment with CMGi. We assessed the CMGi/CA1 effects on three tumor lines from malignancies that are K-Ras-driven and have limited treatment options or few effective targeted therapies [Psn1 pancreatic ductal adenocarcinoma (PDAC; K-Ras-G12R, also Myc amplified); H460 nonsmall cell lung carcinoma (NSCLC; K-Ras-Q61H); 143B OS (K-Ras-G12S/A59T)]. CMGi effects were compared with responses seen in immortalized nontumor HaCaT keratinocytes and primary human keratinocytes. In viability assays, all three tumor lines are ∼4–15 times more sensitive to CA1 exposure relative to HaCaT and primary cells, with tumor cell IC₅₀ estimates of ∼1–4 µmol/L (Fig. 6A, also compared with Fig. 3A, in which HaCaT shows IC₅₀ ∼15 µmol/L). Novobiocin has little effect until higher concentrations are tested. Using a larger cohort of tumor cells, the NCI (NCI60 tumor cell analyses) also found similar tumor cell sensitivities to CA1, but not novobiocin, using proliferation assays (CA1 GI₅₀ 0.5–5 µmol/L; discussed below).

We assessed the doubling times of the three tumor lines to that of HaCaT and primary cells (Fig. 6B). Two of the lines display faster growth rates than HaCaT (143B and Psn1), whereas H460 NSCLC cells are similar to HaCaT, and primary cells display the longest doubling times. This indicates that although a shorter doubling time, and perhaps reduced G1 and/or S-phase lengths that are known to reduce MCM loading (51), might explain in part the selective sensitivity of certain tumor cells to CA1/CMGi, differences in doubling times are not the single reason. As we discuss below, it is more likely that these K-Ras-driven tumor cells are sensitive to CA1/CMGi for numerous reasons, including the presence of RS, which requires efficient MCM/CMG functionality for recovery, or changes to the CMG helicase similar to that elicited by Myc or Cyclin E overexpression.

We next verified that at these effective CA1/CMGi concentrations for viability loss, the CMG helicase was indeed targeted in asynchronous tumor cells. Results show that in each cell type the CMG helicase components (MCM subunits, Cdc45, GINS/Psf1,3) were displaced from chromatin at concentrations between 1 and 2.5 µmol/L, with MCMs in some analyses being slightly more resistant (Fig. 6C). RPA was also lost from chromatin, indicating that CMG enzyme function was inhibited at these concentrations. In contrast, the ORC complex (assessed via Orc4 subunit), which contains multiple ATPase domains necessary for chromatin association, was not similarly sensitive to CA1. Exposure of all three tumor lines and HaCaT cells to a slightly higher 5 µmol/L concentration of CA1 shows a tumor cell-specific increase in DNA damage signals (gamma-H2AX surrogate) and Parp cleavage indicative of apoptosis (Fig. 6D). Increasing the CA1 dose to 15 µmol/L shows that HaCaT cells will eventually incur DNA damage and Parp cleavage at higher doses.

We next tested whether ectopic expression of a mutant Ras protein could acutely sensitize cells to CMGi. For other ongoing studies, we generated a lentiviral construct that allowed regulatable expression of HA-tagged oncogenic H-Ras61L (using the Tet-On promoter). HaCaT cells containing wt-Ras were infected with these lentiviruses, pooled, and tested ±H-Ras61L expression for effects on cell proliferation, DNA damage, and Parp cleavage in the absence or presence of CMGi (Fig. 6E). Ectopic expression of H-Ras61L alone has a small suppressive effect on proliferation of nontumor HaCaT cells. However, although the proliferation of uninduced HaCaT cells (with wt-Ras) is reduced at two tested CMGi concentrations, expression of H-Ras61L causes a considerable increase in sensitivity of HaCaT cells to CMGi, reducing proliferation by almost 70% compared with cells without H-Ras61L expression. Consistent with this, immunoblotting (Fig. 6E; right) shows that H-Ras61L expression in the presence of CMGi causes a dose-dependent increase in DNA damage, and at the higher CMGi concentration (15 µmol/L, the HaCaT IC₅₀ for CMGi) Parp cleavage is elevated.

We have identified the first small chemical compounds capable of inhibiting ATPase and helicase functions of the human replicative CMG helicase, as an important advance toward the development of effective CMGi with anticancer clinical applications. Common targets in anticancer chemotherapy regimens often include the DNA replication and repair machinery that functions at replication forks or damaged DNA sites. Clearly, the CMG helicase represents another such target that has yet to be drugged, and we show distinct mechanisms of action used by CMGi to abrogate DNA replication. CA1, clorobiocin, and the methyl-substituted synthetic derivative referred to here as MBC are effective CMGi, competing with ATP for binding and hydrolysis at one or more ATPase sites within the MCM ring. In human cells, CMGi inhibit/disrupt the human CMG at several steps of assembly or activity that studies of the eukaryotic CMG have shown require ATPase functions of the MCM ring (4, 39, 40, 44, 52, 53): (i) CMGi block MCM loading onto chromatin/DNA, but once loaded, MCM rings are no longer disrupted by CMGi exposure, which is likely due to such loaded MCM complexes retaining ADP in their ATPase sites until conversion to CMGs near G1/S (44, 53); (ii) recruitment of GINS to MCM rings is blocked by CMGi exposure during G1; and (iii) CMGi treatment in the S-phase not only inhibits CMG helicase activity but also causes GINS and Cdc45 to dissociate from MCMs. As a consequence, replisomes are also structurally disrupted by CMGi, indicating that CMGi elicit a “hit and run” form of DNA replication inhibition that requires new replisomes to be (re)assembled for the S-phase to proceed. This is distinct from that of etoposide, which also inhibits the replication apparatus but has little or no effect on these CMG and replisome complexes.

The viability of several K-Ras-driven tumor lines is selectively reduced in the presence of low micromolar (1–4 µmol/L) concentrations of CMGi/CA1. In agreement with our findings, the NCI has publicly available growth suppression data (NCI60 set) showing that many solid tumor cell lines, including K-Ras-driven types, are sensitive to similar concentrations (GI₅₀ ∼0.5–5 µmol/L for CA1) of the CMGi tested here (CA1, NSC107412; clorobiocin, NSC227186) but are largely insensitive to novobiocin (NSC2382; https://dtp.cancer.gov/dtpstandard/dwindex/index.jsp). The K-Ras-driven tumor lines tested here and other solid tumor lines tested by the NCI are ∼4–15 times more sensitive to CA1/CMGi compared with nontumor immortalized HaCaT keratinocytes and primary human keratinocytes. Importantly, whereas this shows K-Ras expressing tumor cells are sensitive to CMGi, the large number of tumor cell types assessed by the NCI suggests a broader CMGi activity against genetically different tumor cells. This suggests that other oncogene-induced issues may also contribute to CA1/CMGi sensitivities, and the K-Ras analysis herein is one initial example (discussed next). Selective sensitivity of tumor cells to CA1/CMGi suggests that a therapeutic window will exist for targeting the CMG helicase with future derivatives of these CMGi. However, we caution that this limited cell culture–based analysis is only a proof-of-concept, and actual toxicity windows for CMGi will be more accurately assessed using animal models or future human trials. The identification of CA1 as a CMGi now provides a probe compound for such animal-based tumor studies.

There are multiple potential reasons why tumor cells are more sensitive to CA1/CMGi compared with nontumor cells, although the CMG helicase (or MCMs) is being targeted in both cases. Oncogene-driven RS and DNA damage in tumor cells render the function of the CMG and reserve MCM complexes more critical, and therefore a liability, for efficient recovery and repair of such defects (1). More importantly, overexpressed oncogenes such as Myc and Cyclin E produce deficiencies in the CMG helicase by over-activating CMG helicases or suppressing MCM loading, respectively (1, 14–17, 54), thereby reducing unused MCM/CMG reserves. These adverse effects on MCM/CMG complexes not only contribute to the oncogene-induced RS (fork stalling, DNA replication fork rate changes) but also render tumor cells more dependent on remaining (or limited) unused reserve MCM/CMG complexes for survival. Recent studies have shown that oncogenic Ras similarly causes RS (fork stalling and other adverse DNA replication effects) and that suppression of MCM complexes by siRNA is selectively and synthetically lethal in tumor cells expressing oncogenic Ras (18–20). Our findings are consistent with such Ras-MCM genetic results, showing that K-Ras-driven tumor cells are pharmacologically sensitive to CMGi. Although it remains unknown whether mutated K-Ras causes deficiencies in MCM/CMG complexes, it is noteworthy that Myc and Cyclin E are downstream of Ras signaling, leading one to speculate that molecular changes in MCM/CMG management might also contribute to CMGi sensitivity in K-Ras-driven tumor cells. Compounding the oncogene-driven weaknesses in the CMG, p53 deficiencies have also been shown to be synthetically lethal with loss of MCM proteins (55), suggesting that altered p53 proteins or pathways may predispose to CMGi sensitivity in tumor cells.

These known oncogenic changes that adversely affect MCM/CMG regulation could be used in the future clinic in a personalized medicine approach to stratify patients with genetically defined solid tumors for treatment with CMG-targeted drugs. Our results predict that K-Ras-driven tumors (e.g., PDAC, NSCLC, and colorectal cancers) will be sensitive to CMGi, consistent with genetic synthetic lethality observations (18–20). Tumors with amplified Myc (e.g., breast cancers) or Cyclin E (∼25% ovarian cancers) are likely candidates for CMGi anticancer regimens. Osteosarcomas in particular often have complete loss of p53 expression, strongly suggesting that CMGi may elicit a tumor-selective inhibition for such malignancies. These particular oncogene and p53 changes cover a wide range of human malignancies, and CMGi could be used as a monotherapy based on results here, or in combination approaches with chemotherapy drugs that cause replication stress requiring efficient CMG function for survival as our previous genetic results have suggested (12). Indeed, targeted approaches using CMGi might prove useful in anticancer regimens in which existing chemotherapy or other targeted approaches have failed, as CMGi block DNA replication/repair and reduce cell viability through different mechanisms. Indeed, CMGi antitumor mechanisms are uniquely different from most other targeted therapeutic approaches, in that CMGi are targeting a biochemical weakness (MCM/CMG reduced functionality) rather than an overexpressed/amplified enzyme activity (e.g., many kinase targets) in tumor cells.

We cannot rule out that CA1, like other drugs (e.g., kinase inhibitors), might have off-target activity against other unknown or related enzymes, especially if doses are increased. However, our results demonstrate that the CMG helicase and MCM complex are indeed major targets because the known ATPase-dependent functions of MCM/CMG complexes described above are clearly inhibited by CA1 in cells. This is also supported by the fact that CA1 effects on CMG complexes in cells mirror experimental results using in vitro biochemical assays. Multiple other ATPases in cells are not affected by CA1 exposure, including ORC, Cdc6, and several kinases, showing that at the low concentrations tested, there is observable CA1 selectivity toward MCM/CMG complexes. As a corollary, such results also suggest that upstream growth factor pathways that regulate ORC, Cdc6, and these kinases are not sensitive to CA1. More intriguing, the specificity of chemicals that function as CMGi directly correlates with their effects on proliferation and viability. Only chemicals containing the noviose sugar with the methylpyrrole group (CA1, clorobiocin, MBC in vitro) are capable of inhibiting the CMG and cell proliferation/viability, whereas novobiocin (lacking the pyrrole) cannot appreciably inhibit the CMG or affect cell viability. In terms of NCI tumor cell testing, a cellular target(s) required for proliferation and sensitive to low concentrations of CA1 (∼0.5–5 µmol/L), but not targeted by closely related novobiocin, was not clear. Our results demonstrate that MCM/CMG complexes are likely a major target of CA1/CMGi in the tumor cell testing performed by us and the NCI. It has been suggested that HSP90 might be a target, but the high doses required to inhibit HSP90 (700–1,000 µmol/L) and its sensitivity to both CA1 and novobiocin suggest it is not the target (56, 57). Interestingly, at higher doses novobiocin (>100 µmol/L) can inhibit a different anticancer target, DNA Polymerase-theta, which also contains a helicase domain and is involved in DNA repair (58). Such results indicate that these aminocoumarins can display target specificity and distinct modes of action that could be used in specific anticancer regimens.

Some intriguing possibilities arise when considering the repurposing of CA1 or related aminocoumarins as anticancer CMGi agents. CA1 was originally investigated more than five decades ago as a potential antibiotic by Bristol Myers, although aminocoumarin use as antibiotic was supplanted by more effective quinolones (e.g., ciprofloxacin; ref. 21). Unfortunately, CA1 has poor pharmacokinetic and pharmacodynamic (PK/PD) characteristics, being a large molecule (FW > 1,100) with poor solubility and poor bioavailability (oral or intravenous; refs. 21, 59, 60). In the late 1960’s Bristol Myers instead developed a molecule called BL-C43 that was essentially CA1 cut in half (FW ∼550), retained the noviose sugar methylpyrrole and coumarin groups, but replaced the benzamide “tail” with a propionyl to increase solubility and bioavailability (21, 61). Except for the propionyl group, BL-C43 is identical to the MBC molecule generated and shown here to be a CMGi compound. BL-C43 has excellent PK/PD characteristics, with increased solubility, high oral (p.o.) bioavailability, stability, and considerable absorption into the blood. Testing showed that a 500-mg p.o. dose in humans could achieve 15- to 20-μg/mL blood levels (∼30–40 µmol/L) at 3 hours, and ∼16 µmol/L in blood at 10 hours (21, 61). Anticipating BL-C43 antibiotic use, a trial was performed on healthy subjects given 750 mg/day/p.o. for 14 days (50% more drug), which conceivably achieved ∼45–60-µmol/L daily blood peak levels based on the prior testing. Unfortunately, after 10 days subjects displayed unwanted side effects for an antibiotic, presenting with mild jaundice and rash, and one subject had symptoms of congestive heart failure, all of which reversed with no harm when the drug was stopped (21, 61). It is quite plausible from our findings that such high blood levels of BL-C43 could have been (excessively) inhibiting MCM/CMG complexes in healthy cells (e.g., in the liver) to produce these side effects, particularly given BL-C43’s substantial improvements in bioavailability. Intriguingly, given that K-Ras-driven tumor cells are sensitive to ∼4–15-fold lower levels of CMGi, it would seem feasible that far lower doses of BL-C43 could be used to achieve lower blood levels that might offer anticancer benefits without these side effects due to targeting of CMG helicase weaknesses selectively in tumor tissue.

Although BL-C43 has not been formally validated as a CMG-targeting agent, the BL-C43 trial illustrates the potential for the development of CMGi drugs with excellent PK/PD features. Some goals for future CMGi synthetic chemical development will include the design of molecules that can target the CMG with higher affinities, more stability, and defined selectivity toward the CMG, perhaps with PK/PD features learned from BL-C43. The identification of CA1 and MBC as effective CMGi offer a starting point to inform such medicinal chemistry projects while also providing useful molecular probes to further investigate anticancer mechanisms and the therapeutic potential of CMGi against other tumor types.

H.R. Lawrence reports grants from NCI during the conduct of the study, as well as has patents for Dimeric Helicase Inhibitors and Helicase Degraders pending. N.J. Lawrence reports grants from NIH during the conduct of the study, as well as has patents for Dimeric Helicase Inhibitors pending and Helicase Degraders pending. D.R. Reed reports other support from Eisai and Springworks outside the submitted work. M.G. Alexandrow reports grants from NIH, other support from Bristol Myers Squibb, and grants from Adolescent and Young Adult Award, NCI, and NCI during the conduct of the study, as well as has patents for Helicase Inhibitors for the Treatment of Medical Disorders, Dimeric Helicase Inhibitors, and Helicase Degraders pending. No disclosures were reported by the other authors.

S. Xiang: Investigation, writing–original draft. K.C. Craig: Investigation. X. Luo: Investigation. D.L. Welch: Investigation. R.B. Ferreira: Investigation. H.R. Lawrence: Methodology, writing–original draft. N.J. Lawrence: Methodology, writing–original draft. D.R. Reed: Funding acquisition, writing–original draft. M.G. Alexandrow: Conceptualization, supervision, funding acquisition, investigation, writing–original draft.

We thank Dr. Jennifer Binning (Moffitt Cancer Center) for providing plasmids expressing the HPV16 and HPV18 E1 helicases. This work was supported by the National Pediatric Cancer Foundation (nationalpcf.org; D.R. Reed, D.L. Welch) and research support to M.G. Alexandrow from the National Institutes of Health (R01 GM140140-01), Bristol Myers Squibb, and the Adolescent and Young Adult Program at the Moffitt Cancer Center. H.R. Lawrence was supported by an R50 Award from the NCI (2R50-CA211447-06A1). The CCSG grant awarded by the National Cancer Institute to the Moffitt Cancer Center (P30 CA076292) also supported services from the Chemical Biology Core.