The fluorescence intercalator displacement (FID) assay was performed in lithium cacodylate buffer solution (10 mM LiAsO₂Me₂) containing NaCl or KCl (100 mM) and adjusted to pH 7.2 with HCl. The fluorescence spectra were recorded with a HORIBA Jobin-Yvon Fluoromax-3® spectrofluorimeter in the wavelength range from 510 to 750 nm, in a 3-ml quartz cell (path length 1 cm), using the following experimental parameters: excitation wavelength, 501 nm; increment, 1 nm; optical slit widths, 3.0/3.0 nm; integration time, 0.1 s. The fluorescence spectrum of the buffer solution was recorded and was systematically subtracted from all following spectra. The solutions of DNA (0.25 µM) were mixed with thiazole orange (TO) (final concentration 0.5 μM) and reference fluorescence spectra were recorded. Increasing concentrations (from 0 to 10 molar equivalents, i.e. from 0 to 2.5 µM) of ligands to be tested were then added. After a 3-min shaking and equilibration period, a fluorescence spectrum was recorded for each ligand addition step. The degree of TO displacement (percentage of ligand-induced decrease of TO fluorescence) is calculated from the fluorescence area FA as: where FA₀ is the fluorescence area of the reference spectrum. The percentage of displacement is then plotted as a function of the concentration of added compound.

Helicase assays were carried out by incubating 100 nM of S. cerevisiae Pif1 and 2 nM of Cy5-labeled G4 DNA or 1 nM forked DNA substrate at 35°C, in a buffer containing 20 mM Tris pH 7.5, 50 mM NaCl, 100 µg/ml bovine serum albumin, 2 mM DTT, 5 mM Mg²⁺ and 4 mM ATP, and the indicated amount of the G4 binding ligands Phen-DC₃, Phen-DC₆ or NMM. Reactions were pre-incubated at 35°C for 10 min and started by addition of 4 mM ATP. Reactions (10 µl) with G4-DNA substrates were stopped by addition of 2 µl stop buffer (17% Ficoll, 50 mM EDTA, 3 mg/ml Proteinase K) and further incubated at 35°C for 10 min. Reaction products were loaded on a 8% polyacrylamide non-denaturing gel and resolved by electrophoresis at 4°C and 10V/cm in TBE 1X buffer containing 10 mM NaCl (for G4-DNA formed in NaCl) or 10 mM KCl buffer (for G4-DNA formed in KCl). Reactions containing the forked DNA substrate (10 µl) were stopped by addition of 2 µl of 17% Ficoll, 50 mM EDTA and 150 nM unlabelled fD20 oligonucleotide. Reaction products were loaded on a 10% polyacrylamide non-denaturing gel and resolved by electrophoresis at 4°C and 10 V/cm in TBE 1× buffer. Gels were scanned with a Storm PhosphorImager (Molecular Dynamics) at 635 nm and quantified using ImageQuant software (GE Healthcare). Two independent lots of NMM were purchased from Frontier Scientific and assayed with similar outcome.

The genotypes of the S. cerevisiae strains (S288C background) used here are reported in the Supplementary Table S1. The rad51Δ strain (AND1239-5C) containing CEB1-WT-1.7 was obtained after sporulation of the diploid obtained by crossing AND1202-11A (CEB1-WT-1.7 pif1::KanMX) with the BY4742 rad51::KanMX strain (31). The nucleotide sequences of the natural human CEB1-1.8 (see Supplementary Figure S6) and the synthetic CEB1-WT-1.7 and CEB1-Gmut-1.7 minisatellites, each containing 42 motifs, have been previously reported (8,32). The CEB1 arrays were inserted on the yeast chromosome VIII, upstream of the ARG4 promoter (32).

Cells taken from a fresh patch were suspended at a density of 2 × 10⁵ cells/ml into 5 ml of rich Yeast–Peptone–Dextrose (YPD) medium containing the indicated concentration of Phen-DC₃ and Phen-DC₆ and 1% DMSO, or in YPD containing only 1% DMSO (control treatment). Cells were grown for eight generations at 30°C with agitation, plated as individual colonies on YPD plates and grown at 30°C. For the induction in synthetic complete (SC) medium with NMM and Phen-DC₃, the final concentration of DMSO reached in the liquid culture and in the control treatment was 0.4%. For the NMM treatment in SC medium, cells were grown overnight in presence of the drug, diluted to 2 × 10⁵ cells/ml in the same media and grown for an additional eight generations at 30°C, thus undergoing approximately four additional generations in presence of the drug compared to the control condition or the Phen-DC₃ treatment.

Colonies grown from treated cultures were inoculated in 96-well megaplaque in YPD for 24–48 h at 30°C. Pools of 4–16 colonies were made right before DNA extraction. DNA was digested with ApaI and SpeI to probe for the CEB1 minisatellite, or with AluI to probe for the DAN4 and FLO1 minisatellites (8). Digestion products were migrated in a 0.8% agarose–TBE 1× gel and analyzed by Southern blot using a radiolabeled probe corresponding to the minisatellite of interest. Blots were analyzed using a Storm PhosphorImager (Molecular Dynamics) and quantified using ImageQuant software (GE healthcare). To account for secondary rearrangements (occurring after plating) in the analysis, each visible band was quantified and normalized to the mean of the intensity of the parental bands. Bands with signal intensity two times lower than expected for a primary rearrangement were removed from the counting. In numerous instances pools of colonies were depooled for verification and sequencing.

Additional experimental protocols are reported in Supplementary Material and Methods.

Our screening and initial analysis of a large series of potential G4-binding ligands was performed with two quadruplex-forming oligonucleotides: 22AG [5′-AG₃(T₂AG₃)₃-3′] mimicking a human telomeric sequence and TBA [5′-G₂T₂G₂TGTG₂T₂G₂-3′], the thrombin binding aptamer sequence (27). The bisquinolinium-dicarboxamide derivatives Phen-DC₃ and Phen-DC₆ (illustrated in Figure 1A) were among the most selective G4 ligands.

Since G4 can fold differently depending on primary nucleotide sequence (33), we wished to evaluate the capacity of the Phen-DC compounds to bind to the quadruplex conformation adopted by the more complex human minisatellite CEB1 motif used in the present study. To this end, we folded a single-stranded oligonucleotide representing the prevalent 39 nt CEB1 repeat motif (G4-CEB1) (8) and assayed its interaction with the Phen-DC and NMM compounds using the G4-FID assay (34), whose principle is illustrated in Figure 1B. This method is based on the competitive displacement of the fluorescent intercalator probe TO by a putative G4 ligand. TO being highly fluorescent when bound to DNA and virtually non-fluorescent when free in solution, the decrease of TO fluorescence as function of increasing ligand concentration thus serves to evaluate binding affinity of the ligand for G-quadruplex DNA (expressed as DC₅₀: concentration required to decrease the fluorescence signal of 50%). As shown in Figure 1C, fluorimetric titration of TO by the G4-CEB1 substrate showed a strong fluorescence increase, indicating that TO efficiently bound to the G4-CEB1 conformation. The titration curve adopted a regular shape that could be fitted with a 1/1 stoichiometry model, giving a K_d value in the micromolar range (K_d = 7.1 × 10⁻⁷ M). This behavior as well as the K_d value are similar to results obtained with the TBA and the human telomeric G4-DNA (27), thereby indicating that, in our experimental conditions, G4-CEB1 accommodates the probe at a single site corresponding to one of the two external G-quartets. The displacement of the TO by Phen-DC compounds was then examined in K⁺ (Figure 1D) and Na⁺ (Supplementary Figure S1) conditions. Both Phen-DC₃ and Phen-DC₆ isomers display a strong ability to displace TO (DC₅₀ = 0.4–0.5 µM) with complete displacement (100% quenching) being reached at a low ligand/quadruplex ratio (5 and 10 for Phen-DC₃ and Phen-DC₆, respectively). Considering previous results obtained with the telomeric quadruplex and the low concentration of G4-DNA (0.25 µM) used in the test, these data unambiguously reveal a strong interaction between Phen-DC compounds and G4-CEB1 (Figure 1D, Supplementary Figure S1). In sharp contrast, the NMM was much less efficient in displacing TO since saturation was not reached even at high ligand/quadruplex ratio (>40). This result indicates that NMM exhibits a much lower affinity for G4-CEB1 than our Phen-DC compounds (Figure 1D).

Recently, we showed that the yeast Pif1 helicase efficiently unwinds the G4-CEB1 substrate (8). To investigate the ability of the two Phen-DC derivatives to interfere with such G4 unwinding enzymatic activities, we tested their effect on G4-CEB1 unwinding by Pif1 in vitro. The G4-CEB1 substrates were made in two conditions: i.e. in presence of either Na⁺ or K⁺ as a monovalent cation (see Supplementary Materials and Methods) (8). This led to intermolecular G4-CEB1 structures with different gel migration properties, likely corresponding to a different folding of the G4-DNA. As shown in Figure 2A, Pif1 efficiently unwound the G4-CEB1 substrates formed in either conditions but was rapidly inhibited by low concentrations of Phen-DC₆ [k_i(K⁺) = 250 nM; k_i(Na⁺) = 300 nM]. Phen-DC₃ showed comparable activity (Supplementary Figure S2). Figure 2.

Inhibition of G4-CEB1 unwinding by Phen-DC₆ and NMM in vitro. (A) Pif1 unwinding of G4-CEB1 or forked DNA substrate (fD20) in presence of increasing concentrations of Phen-DC₆. The G4-CEB1 substrates were formed in presence of Na⁺ or K⁺ as indicated. Quantifications (mean ± SD) from three independent experiments are shown. Open triangle indicates boiled substrate. Dash indicates absence of Pif1. Position of G4-CEB1 (G4) and unwound substrate (ss) are indicated. (B) Same as in (A), using NMM as a G4 ligand.

To test the effect of the Phen-DC compounds in vivo and their specificity towards G4-DNA substrates, we examined their capacity to trigger the instability of the G4-prone human CEB1 minisatellite inserted in the yeast genome (38) (Figure 3A). In WT yeast cells, the synthetic 1.7 kb CEB1 array (CEB1-WT-1.7) composed of 42 wild-type motifs of 39 nt (Figure 3B) is rather stable (8). It is highly destabilized in the absence of the Rad27/FEN1 endonuclease or of the Pif1 helicase, but in the latter case, only when the CEB1 repeats contains G4-forming sequences (8). Mutation of the G4-forming motifs in the synthetic CEB1 array (CEB1-Gmut-1.7, Figure 3B) stabilizes CEB1 in pif1Δ cells but not in rad27Δ cells (8). Thus, this system provides a well-defined genetic assay to monitor the formation of G4 structures in vivo. Figure 3.

Phen-DC compounds trigger G-quadruplex-dependent CEB1 instability. (A) Schematic representation of the CEB1 minisatellite insertion in the chromosome VIII. (B) Sequences of CEB1-WT and CEB1-Gmut motifs. G-runs involved in G-quadruplex formation are boxed. Point mutations introduced in CEB1-Gmut motif and preventing G-quadruplex formation in vitro are shown in bold. (C) Southern blot analysis of WT cells carrying CEB1-WT-1.7 (strain AND1212-10D) or CEB1-Gmut-1.7 (AND1227-5C) after control (1% DMSO) or 500 µM Phen-DC₃ treatment. The number of colonies analyzed and the rearrangement frequencies are indicated below each gel. Size markers (kb) are indicated on the right. The position of the parental minisatellite alleles (42 repeats, 1.7 kb) is indicated by an asterisk.

The occurrence of G4-dependent CEB1 rearrangements in treated wild-type cells revealed that some G4 stabilized by the Phen-DC compounds escape from the unwinding activity of Pif1. This is consistent with the observations that these compounds inhibit G4-CEB1 unwinding by the Pif1 helicase in vitro (Figure 2), but whether or not Pif1 remove all or a fraction of the G4 that form in CEB1 remained to be evaluated. To address this question, we examined the frequency of CEB1-WT-1.7, CEB1-1.8 and CEB1-Gmut-1.7 rearrangements triggered by Phen-DC₃ or Phen-DC₆ in the absence of Pif1 (pif1Δ cells) and in a strain carrying a helicase-dead allele of PIF1 [pif1-K264A (39)].

In our assays, the background of CEB1 rearrangements in WT cells is low (<1%) and roughly negligible compared to the rapid and strong effect of the Phen-DC compounds (Table 1), but this is not the case in PIF1 mutant cells, in which the G4-prone CEB1 minisatellite is substantially unstable [see ref. (8) and present study]. Thus, to take into account CEB1 rearrangements which may pre-exists before the treatment, we estimated the effect of the G4-ligands in two steps. First, we measured the frequency of pre-existing rearrangements in the starting cell patches and then, as in WT cells, divided the assay culture in two samples: one being treated by the Phen-DC compounds and the other incubated in 1% DMSO alone (control treatment). Thus, upon subtraction of the pre-existing events (see Supplementary Materials and Methods for details on the procedure used for correction), we could estimate the frequency of rearrangements generated during the drug treatment only (eight generations). The pif1Δ cultures treated with 200 µM Phen-DC₆ or 500 µM Phen-DC₃ exhibited a high frequency of CEB1-WT-1.7 rearrangements, 20.9% and 36.8%, respectively, instead of 12.1% in the control treatment (P-values versus control treatment are 4.1e⁻⁴ and 2.7e⁻¹⁶, respectively) (Figure 4, Table 2). Similarly, the frequency of CEB1-1.8 rearrangements raised from 5.2% to 12.2% (P = 1.1e⁻³) and 2.7% to 7.3%, (P = 5.5e⁻³) in the pif1Δ and pif1-K264A cells, respectively. Thus, taking into account the pre-existing events, estimated at 9.8% (CEB1-WT-1.7) and 2.1 % (CEB1-1.8) in pif1Δ cell cultures, and at 1.3% (CEB1-1.8) in the pif1-K264A cell cultures, the treatments yielded a 3- to 12-fold increase of CEB1 rearrangements. Finally, as in WT cells, the Phen-DC compounds did not stimulate instability of the CEB1-Gmut-1.7 minisatellite (Figure 4). Altogether, the above results demonstrate that the Phen-DC compounds are able to stimulate CEB1 rearrangements in the presence or absence of Pif1, and in this latter case, uncover the in vivo formation of a higher number of G4 than in WT cells. Table 2.

Frequency of CEB1 rearrangements in pif1Δ cells

To determine the nature of the rearrangements induced upon Phen-DC treatment in WT and pif1Δ cells, we used the naturally polymorphic CEB1-1.8 minisatellite in which the presence of multiple base substitutions along the 42 motifs allows a precise analysis of the contribution of the parental motifs in the sequenced size variants (32). As for the synthetic CEB1-WT-1.7 minisatellite, the treatment of WT cells grown in YPD with 500 µM Phen-DC₃ stimulated the rearrangements of the CEB1-1.8 allele ∼14-fold (29/768 versus 5/1824 in the untreated cells, P = 1.8e⁻¹¹) and yielded variants of diverse size. Thus, the presence of base polymorphisms located outside the triplets of G involved in CEB1 quadruplex formation does not interfere with the drug target. The sequences of 14 and 13 CEB1-1.8 contractions obtained upon treatment of WT and pif1Δ cells are reported in Supplementary Figure S6. They were different from each other. In both strains, simple deletions with one chimerical motif, double deletion with two chimerical motifs, and complex events with multiple chimerical motifs and various internal reshuffling of the parental polymorphic markers occurred (Figure 5A and B), suggesting that the underlying mechanisms of rearrangements were similar in WT and pif1Δ cells. No de novo mutagenic events were observed. Thus, the nature of the rearrangements obtained after treatment with Phen-DC₃ is diverse and reflects the same general pattern of events as in the untreated pif1Δ (8) and rad27Δ (32) mutants. Figure 5.

Sequence and size of CEB1 rearrangements produced upon treatment with Phen-DC₃. (A) Schematic representation of the natural CEB1-1.8 allele and of 14 rearrangements produced upon treatment of WT cells (ORT2914) with 500 µM Phen-DC₃. Nucleotide sequences are reported in Supplementary Figure S6. Chimerical motifs combining the polymorphisms of more than three parental motifs (called ‘Mix’ in Supplementary Figure S6) are shown as a white box. Sequences of the two red motifs in C7 have not been determined. (B) Schematic representation of 13 rearrangements produced upon treatment of pif1Δ cells (ORT4841) with 500 µM Phen-DC₃. (C) Size distribution of CEB1-WT-1.7 and CEB1-1.8 rearrangements in WT (AND1212-10D and ORT2914), pif1Δ (AND1202-11A and ORT4841), pif1-K264A (ORT5087-5E) or rad27Δ (AND1218-1A and ORD6713-8D) cells, with or without treatment with 500 µM Phen-DC₃. Datasets are from this study, except the size distribution of rearranged CEB1-WT-1.7 alleles in rad27Δ cells, obtained from ref. (8) (strain AND1218-1A). The n indicates the number of variants analyzed. The vertical gray lines indicate the size of the parental alleles. For each distribution, dark bar: median; gray bar: mean; box: distance between the first and third quantiles; whiskers: extreme values; white circles: outliers. P-values were obtained by comparing the samples with a Mann–Whitney–Wilcoxon statistical test.

In comparison to the rearrangements observed in the absence of Pif1 alone, we noted that the rearrangements produced upon treatment of Pif1-deficient cells yielded a larger number of CEB1 fragments of small size, corresponding to short variants of 15 or less repeat units (Figure 4). A comparative analysis of the size of the CEB1-WT-1.7 and CEB1-1.8 rearrangements obtained in pif1Δ or pif1-K264A cells with or without treatment is reported in Figure 5C. Strikingly, in all cases, despite the existence of pre-existing rearrangements, the mean size of the rearrangements was shorter in Phen-DC₃-treated versus untreated Pif1-deficient cells (P ≤ 9.6e⁻³). We also examined the size of the CEB1 variants produced in the rad27Δ cells. This cell exhibits a high level of instability for CEB1-WT-1.7 (30%) (8) and CEB1-1.8 (41/158, 25.9%) that is comparable to the instability measured in pif1Δ-treated cells (Table 2). The mean size of the CEB1 rearrangements in rad27Δ is much higher than in both pif1Δ and pif1-K264A treated cells (P ≤ 2.2e⁻²) (Figure 5C). Hence, the increase in short alleles upon treatment of Pif1-deficient cells is not only a consequence of a high level of instability, but rather a specific effect of Phen-DC₃ treatment.

In previous studies, we demonstrated that in rad27Δ and pif1Δ cells the formation of CEB1 rearrangements depends on the process of homologous recombination, since it was abolished in the absence of the RAD51 or RAD52 genes (8,32). Similarly, we examined the behaviour of the CEB1-WT-1.7 array in a rad51Δ strain upon treatment with 200 µM Phen-DC₃ (Supplementary Figure S7). Compared to the WT strain (4.2% rearrangements, Table 1), we observed no rearrangements (0/384) in the absence of Rad51 (P = 2.6e⁻⁵). Therefore, Phen-DC treatment also leads to CEB1 rearrangements through the homologous recombination pathway.