Identification and characterization of topoisomerase III beta poisons

Significance

Topoisomerase poisons that trap topoisomerase I and II (TOP1 and TOP2) have long been important agents in anticancer therapy and for elucidating the cellular functions of TOP1 and TOP2. However, poisons for topoisomerase III have never been reported. Topoisomerase III beta (TOP3B) is the only topoisomerase with a dual activity on RNA as well as DNA and has recently been reported to ensure genome stability, chromatin accessibility, R-loops resolution, and potentially the replication of positive-sense RNA viruses, including dengue and SARS-CoV-2. In our study, we designed and carried out a chemical screen for poisons that trap human TOP3B. Among the most active inhibitors, we report and characterize in vitro and cellular trapping of TOP3B by bisacridine and thiacyanine compounds.

Abstract

We designed and carried out a high-throughput screen for compounds that trap topoisomerase III beta (TOP3B poisons) by developing a Comparative Cellular Cytotoxicity Screen. We found a bisacridine compound NSC690634 and a thiacyanine compound NSC96932 that preferentially sensitize cell lines expressing TOP3B, indicating that they target TOP3B. These compounds trap TOP3B cleavage complex (TOP3Bcc) in cells and in vitro and predominately act on RNA, leading to high levels of RNA-TOP3Bccs. NSC690634 also leads to enhanced R-loops in a TOP3B-dependent manner. Preliminary structural activity studies show that the lengths of linkers between the two aromatic moieties in each compound are critical; altering the linker length completely abolishes the trapping of TOP3Bccs. Both of our lead compounds share a similar structural motif, which can serve as a base for further modification. They may also serve in anticancer, antiviral, and/or basic research applications.

Many metabolic processes in our cells generate torsional tensions, catenanes, or knots in the DNA, which require the activity of topoisomerases for their removal (1). All topoisomerases share a catalytic tyrosine residue, which transiently and reversibly breaks the backbone of nucleic acids to relieve torsional strains. During the transient intermediate state of the topoisomerase activity, the enzyme is covalently attached to one end of the disconnected nucleic acid backbone, forming the topoisomerase cleavage complex (TOPcc). A class of compounds termed topoisomerase poisons trap and stabilize TOPccs. Structural studies have shown that many topoisomerase poisons stabilize TOPccs by binding at the topoisomerase-DNA interface, and these are termed interfacial inhibitors (2). Various topoisomerase poisons specifically targeting human topoisomerases I and II (TOP1 and TOP2) are among the most potent and widely used anticancer agents. When replication or transcription machinery encounter stabilized TOPccs, detrimental DNA–protein cross-links and double-strand breaks are generated (3). Poisons of the bacterial topoisomerases (gyrase and topo IV) are also among the most potent and widely used antibacterial agents (4).

Among the 6 human topoisomerases, topoisomerase III beta (TOP3B) is the only member of the family that can act on RNA as well as DNA (5–7). The RNA topoisomerase activity of TOP3B is likely important given the prevalence of RNA topoisomerases in all domains of life (8). While TOP3B is not an essential gene, loss of TOP3B leads to abnormal neurodevelopment (5, 6, 9), increased R-loops, and genome instability (7, 10). Data from an engineered TOP3B self-trapping mutant also demonstrated that irreversible TOP3B cleavage complexes (TOP3Bccs) lead to DNA damage and reduce cell survival (11). Furthermore, tumors lacking TOP3B grow more slowly (12), making TOP3B a rational anticancer target. Currently, there are no known TOP3B poison or inhibitor, which could also act as antiviral agents, as a recent report showed that TOP3B is required for efficient replication of positive-sense RNA viruses, including dengue viruses, chikungunya viruses and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2 coronaviruses) (13). Furthermore, TOP3B poisons could potentially cause persistent damage to viral RNA and act as antiviral compounds, and they could serve as valuable tools to study the fundamental biological processes requiring TOP3B.

To this end, we designed and carried out a Comparative Cellular Cytotoxicity Screen for TOP3B poisons, followed by mechanistic study of the identified compounds in vitro and in cells. Here, we report two classes of compounds based on thiacyanine and bisacridine chemical scaffolds that induce TOP3Bccs in cells and in band shift assays with recombinant TOP3B, consistent with TOP3B poisons.

Results

High-Throughput Screen to Identify TOP3B Poisons.

To identify TOP3B poisons, we reasoned that such compounds should lead to deleterious DNA damage only in the presence of TOP3B. After generating TOP3B knockout (TOP3B-KO) cells from human colon carcinoma HCT116 cells, we carried out a Comparative Cellular Cytotoxicity Screen (Fig. 1A), where we searched for compounds to which HCT116 TOP3B-KO cells are more resistant than isogenic parental HCT116 cells. We labeled HCT116 parental cells and TOP3B-KO cells with different fluorescent proteins, thus allowing us to monitor the viability of each cell line in the same well via separate fluorescence microscopy channels. The parental Green Fluorescent Protein (GFP)-labeled HCT116 and mCherry-labeled TOP3B-KO cells were mixed and seeded in each well at a 1-to-1 ratio. After 24-48 h, we carried out media change with compounds from the drug library at different concentrations and treated the cells for 72 h. The entire plate was imaged at the end of drug treatment, and the sensitivity of each cell line to a particular compound was directly assessed using two distinct channels in fluorescence microscopy (Fig. 1A). With this strategy, we were able to assay the parental and TOP3B-KO cells in the same well, which reduced the amount of all required reagents by 50%, including the amount of compounds required. Furthermore, since the parental and TOP3B-KO cells were cultured in the same well, this strategy ensured that both cell lines underwent identical growth conditions. Our measurement of relative cytotoxicity in two cell lines eliminated potential variation in drug concentrations introduced by pipetting errors or different degrees of liquid evaporation in different wells.

Fig. 1.

Comparative Cellular Cytotoxicity Screen to identify TOP3B poisons. (A) A pair of isogenic HCT116 and HCT116-TOP3B-KO cells were constructed to express GFP and mCherry, respectively. Both cell lines were mixed at 1:1 ratio, seeded together, and allowed to attach for 48 h prior to treatment with drugs at indicated concentrations for an additional 72 h. On each cell culture plate, one column of dimethyl sulfoxide (DMSO)-treated samples served as control. At the end of drug treatment, the entire cell culture plate was imaged in multiple fluorescent channels to calculate the relative viability rates of the wild-type and TOP3B-KO cells, normalized to the DMSO-treated control wells on the same plate. The 4 potential outcomes are represented from left to right in the middle of the figure: 1) If TOP3B-KO cells show selective resistance, the compound potentially targets TOP3B; 2) if a compound is selectively toxic to TOP3B-KO cells, the compound potentially interferes with parallel pathways of TOP3B, so that in KO cells, inhibiting the parallel pathway leads to adverse effects; 3) if a compound shows no effect (same results as DMSO-treated samples), it does not interfere with any pathways related to TOP3B and is discarded; and 4) if a compound shows high toxicity to both HCT116 and TOP3B-KO cells, it is retested at lower concentrations. (B) Representative fluorescent microscopy images of a DMSO control and drug-treated samples are shown. After deconvoluting into individual fluorescent channels, the drug-treated sample was compared to the DMSO control sample in each fluorescent channel. The GFP channel conveys the effect of the drug on the HCT116 cells while the mCherry channel conveys the effect of the drug on TOP3B-KO cells. Each sample is normalized to the DMSO-treated control samples on the same plate.

Using Comparative Cellular Cytotoxicity Screen, we screened the 811 compounds of the NCI DTP Mechanistic Set VI Library. This small compound library had been curated to include drugs known to induce different growth inhibition patterns in the NCI-60 cell lines (14). Live-cell images of mixed HCT116 parental and TOP3B-KO cells were taken after 72-h treatment for each compound in both GFP and mCherry channels, as well as the light microscopy channel. We trained the image-processing software to quantify cell count and confluency for the parental and TOP3B-KO cells in their respective channel before normalizing to the control (dimethyl sulfoxide [DMSO]-treated) wells on the same plate (Fig. 1B). Images were manually inspected for the instances where the drugs strongly fluoresced in one of the channels. In such cases, the cell count and confluency of light microscopy images were quantified. The light microscopy images represented combined parental and TOP3B-KO cells, so that the levels of cells in the flooded fluorescence channel could be backcalculated. To avoid missing the window of efficacy for a particular compound, we screened each compound across a wide range of concentrations (from 32 nM to 20 µM, in fivefold serial dilutions). Combining all data, we constructed the survival curves of both HCT116 and TOP3B-KO cells to each of the compounds in the library.

Four types of responses to the drug treatments were anticipated: 1) TOP3B-KO cells are selectively resistant; 2) TOP3B-KO cells are selectively sensitive; 3) TOP3B-KO cells have the same response as the parental cells; and 4) compounds are too toxic even at the lowest concentration tested (Fig. 1A). For a particular compound, the relative viability of TOP3B-KO vs. HCT116 (WT) cells after treatment at each concentration was multiplied together to yield a “Resistance Factor” (RF) value (Fig. 2A). Therefore, if TOP3B-KO cells were more resistant to a given compound over a wide range of concentrations, the RF value for that compound would be higher. Conversely, if a compound was consistently more toxic to the TOP3B-KO cells across a wide range of concentrations, then its RF value would be <<1. A compound without differential effect for TOP3B-KO vs. WT cells would generate an RF value close to 1. Finally, the compounds that were too toxic in the initial screen were retested at lower concentrations. As RF can be extracted either from the cell numbers or confluency (RF_{Cell Count} or RF_Confluency), we plotted the log (RF_{Cell Count}) against the log (RF_Confluency) on a scatter plot (Fig. 2B and Dataset S1). As expected, nearly all data points were clustered along the X = Y line, indicating the cell count number agreed well with the confluency value, ensuring confidence in the automated quantification by the software. We selected the compounds with the highest RF values from this initial screen for further study.

Fig. 2.

Data analysis of the Comparative Cellular Cytotoxicity Screen. (A) Equation used to compute the Resistance Factor (RF) for TOP3B-KO cells. Either cell count or confluency from a set of samples can be used to obtain a specific RF for a given compound. (B) After the RF_{Cell Count} and the RF_Confluency are calculated for the entire NCI DTP Mechanistic Set VI Library, the log (RF_{Cell Count}) value is plotted against the log (RF_Confluency) value for all 811 compounds. The top right corner represents the compounds to which TOP3B-KO cells showed the highest resistance, and the top 8% hits (marked by the red circle) were selected for further testing.

We repeated the Comparative Cellular Cytotoxicity Screen with the top 8% hits from the initial screen (area within the red circle in Fig. 2B) with independently procured compounds to eliminate false-positive compounds prior to testing in the next stage. To eliminate the possibility that GFP and mCherry expression might create any bias in the screen, we also repeated the assays with cell lines where we switched fluorescent protein labeling. Of the compounds tested in this stage, 19 of them showed consistent differential effects on the HCT116 vs. TOP3B-KO cells (summary of the screen outlined in SI Appendix, Fig. S1 and select data shown in SI Appendix, Figs. S2 and S3).

Rapid Approach to DNA Adduct Recovery (RADAR) Assays Identify Compounds Inducing Cellular TOP3Bccs.

To determine whether the 19 compounds selected from the Comparative Cellular Cytotoxicity Screen induced cellular TOP3Bccs, we used a modified RADAR assay, which specifically detects proteins covalently attached to cellular nucleic acids (Fig. 3A) (11, 15). After drug treatment, the cells were lysed by a combination of chaotropic salts and detergent, ensuring any TOPcc was instantaneously denatured and thus unable to reverse/release its covalent bond to the nucleic acids. When TOP3B is covalently attached to DNA or RNA, it copurifies with the nucleic acids and the presence of TOP3Bcc can be detected with specific antibodies for TOP3B (11). Out of 19 compounds selected from the Comparative Cellular Cytotoxicity Screen, 6 induced clear signals of trapped TOP3Bcc in human epithelial kidney HEK293 cells transiently overexpressing Flag-tagged TOP3B (TOP3B OE) after 1- to 4-h treatments at 100 µM (Fig. 3 B and C).

Fig. 3.

Modified RADAR assay to demonstrate drug-induced cellular TOP3Bccs. (A) Schematic procedures of the modified RADAR assay. (B) Chemical structures of six hit compounds selected from the screen using modified RADAR assays. (C) HEK293 cells transiently transfected with Flag-tagged TOP3B (TOP3B OE) were treated with the indicated compounds at 100 µM for 1 h prior to analysis by modified RADAR. Anti-Flag antibody was used to probe the level of TOP3Bccs. The same samples probed with anti-dsDNA antibodies served as loading controls. The six compounds showed a spectrum of potency in inducing TOP3Bccs.

Among the 6 compounds active in the modified RADAR assays, NSC690634 and NSC96932 induced TOP3Bccs within 1 h, while it took the other compounds ~4 h to induce comparable levels of TOP3Bccs (Fig. 3C). As the well-characterized TOP1 and TOP2 poisons only require short treatments to generate high levels of TOPccs (16), NSC690634 and NSC96932 (Fig. 3) appeared most consistent with a direct TOP3B trapping mechanism.

Band Shift Assay Screens for Enhanced TOP3Bcc Induction by Recombinant Human TOP3B.

We previously reported an in vitro system to detect TOP3Bccs with either DNA or RNA substrates using recombinant TOP3B enzymes (11). As TOP3B forms covalent linkage with the 5′-end of nucleic acids (1), we used a single-strand DNA or RNA construct containing a short hairpin section with a fluorophore molecule attached at the 3′-end (Fig. 4A). In this assay, when the labeled construct becomes covalently linked to TOP3B at any binding site, its molecular weight becomes significantly higher than the original free DNA or RNA construct (Fig. 4A) (11). The two species could be unambiguously resolved on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and detected via the fluorophores on the oligo constructs, allowing us to measure the levels of TOP3Bccs. This assay also allowed us to test different effects of TOP3B on DNA vs. RNA, as DNA-TOP3Bcc and RNA-TOP3Bcc could be observed separately (11).

Fig. 4.

Band shift assays demonstrating the induction of DNA- and RNA-TOP3Bccs by NSC690634 and NSC96932 with recombinant TOP3B. (A) Simplified scheme of the band shift assay (11). A single-strand DNA or RNA oligo substrate (containing a short hairpin section) labeled on the 3′-end with a fluorophore is incubated with recombinant human TOP3B. A small population of the TOP3B forms TOP3Bccs with the DNA or RNA substrate. The resulting molecular weight of TOP3Bccs with the attached fluorophore is significantly higher than that of free DNA and RNA substrates, regardless of the positions of TOP3B binding sites. (B) The DNA-TOP3Bcc (Upper) and RNA-TOP3Bcc (Lower) are resolved from the free DNA and RNA on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and directly visualized by the fluorophores attached to the oligo constructs. The formation of DNA- and RNA-TOP3Bccs (100 to 180 kDa) is TOP3B-dependent and the TOP3Bccs were titrated with each of the six lead compounds to reach equilibrium. The effect of each compound is compared to the DMSO-treated control samples.

We reasoned that, if the compounds had a direct trapping effect on TOP3B, they should further elevate the level of TOP3Bccs, similar to other well-established TOP1 and TOP2 poisons (3). Among the 6 compounds showing robust TOP3B trapping in the modified RADAR assays, the two compounds inducing TOP3Bcc in cells within 1 h (NSC690634 and NSC96932) also showed increased TOP3Bccs in the band shift assay (Fig. 4B). As previously reported (11), in the absence of any drug, through undefined mechanisms, TOP3B forms several high-molecular-weight species (100 to 180 kDa) (Fig. 4B). While two compounds (NSC47147 and NSC697726) showed little to no effect, two compounds (NSC267229 and NSC316157) showed mild inhibitory effects on both DNA-TOP3Bcc and RNA-TOP3Bcc compared to DMSO-treated samples (Fig. 4B). This is likely related to the fact that both NSC267229 and NSC316157 are derivatives of anthracenedione compounds, a family of potent DNA intercalators including doxorubicin and mitoxantrone (Fig. 3B).

Importantly, compounds NSC690634 and NSC96932 increased the levels of TOP3Bccs in vitro. NSC690634 clearly enhanced the level of RNA-TOP3Bcc (100 to 180 kDa), while high concentrations of NSC690634 induced additional ultra-high-molecular-weight species (>200 kDa) with both DNA and RNA constructs. We note that the exact mechanism of formation for these high and ultra-high-molecular-weight species is not understood at present. Yet the band shift assay demonstrated a clear effect of the TOP3B poisons on the formation of TOP3Bcc in a dose-dependent manner. NSC96932 also enhanced the levels of RNA-TOP3Bcc but inhibited DNA-TOP3Bcc formation, which is potentially related to DNA intercalation by the acridine rings. These results indicate that NSC690634 and NSC96932 mainly act to stabilize RNA-TOP3Bccs with relatively mild effects on DNA-TOP3Bccs. Both compounds were selected for further characterization in cellular studies.

NSC690634 and NSC96932 Rapidly Induce TOP3Bccs in Human Cells.

NSC690634 belongs to the class of bisacridines, and NSC96932 to the class of thiacyanines (Fig. 3B). Comparative Cellular Cytotoxicity Assays confirmed that TOP3B-KO cells were more resistant to both NSC690634 and NSC96932 than the HCT116 parental cells (SI Appendix, Figs. S2 and S3), regardless of which fluorescent proteins were used for cellular labeling. We also tested the effect of NSC690634 and NSC96932 on HAP1 parental and TOP3B-KO cells, where HAP1 TOP3B-KO showed resistance to both compounds (SI Appendix, Figs. S2G and S3D). However, TOP3B-KO cells remain sensitive to NSC690634 and NSC96932 above certain threshold concentrations, indicating that these compounds likely have additional cellular targets beyond TOP3B.

Using HEK293 cells transiently over-expressing TOP3B to enhance TOP3Bcc signals (11), we tested the time- and dose-dependence of both compounds on TOP3Bcc formation by modified RADAR assays. The sensitivity of anti-Flag antibody and the abundance of exogenously expressed TOP3B in this system (11) allowed detection of TOP3Bccs after treatments as short as 10 min for both compounds (SI Appendix, Fig. S4 A and B). The level of TOP3Bccs induced by NSC690634 and NSC96932 increased in a time-dependent manner and leveled off after ~0.5 h. Similarly, both compounds induced TOP3Bccs in a dose-dependent manner, with TOP3Bccs readily detectable at 5 µM of either compound within 1 h (SI Appendix, Fig. S4 C and D).

Next, we tested whether NSC690634 and NSC96932 induced endogenous TOP3Bccs in different human cell lines. The three cell lines tested, HEK293, HCT116, and A549, all express comparable levels of endogenous TOP3B. Due to the lower sensitivity of anti-TOP3B antibody and the lower level of endogenous TOP3B, higher concentrations of either compound or longer treatments were required to detect endogenous TOP3Bccs. Both NSC690634 and NSC96932 induced comparable levels of endogenous TOP3Bccs in time- and dose-dependent manners in all human cell lines tested (Fig. 5 A–D, loading controls shown in SI Appendix, Fig. S5 A–D). As the level of TOP3Bcc increases, the level of free TOP3B is expected to decrease correspondingly. We performed immunoblotting of HEK293 cells treated with NSC690634 or NSC96932, and free cellular TOP3B decreased with increasing concentrations of both NSC690634 and NSC96932 as expected (SI Appendix, Fig. S5 E and F).

Fig. 5.

NSC690634 and NSC96932 induce endogenous TOP3Bccs in different cell lines. (A) Representative modified RADAR assays showing that NSC690634 (100 µM) induces endogenous TOP3Bccs in a time-dependent manner in the indicated human cell lines, immunodetection with anti-TOP3B antibodies. (B) Modified RADAR assay showing that NSC690634 (4 h) induces endogenous TOP3Bccs in a dose-dependent manner in three human cell lines, immunodetection with anti-TOP3B antibodies. (C) Same as Fig. 5A, but with NSC96932. (D) Same as Fig. 5B, but with NSC96932 (1 h). (E) HEK293 cells transiently over-expressing TOP3B were treated with NSC690634 or NSC96932 (100 µM for 1 h) prior to analysis by in vivo complex of enzyme (ICE) bioassays. Ultracentrifugation of cell lysates on a CsCl gradient enabled separation of DNA and RNA species. The separated fractions were quantified and 1.5 µg of DNA or RNA were blotted on a slot blot for independent detection of DNA- and RNA-TOP3Bccs through immunodetection probed with anti-Flag antibodies. The upper row contained 1.5 µg of nontreated control DNA or RNA samples. Representative blot of 3 independent experiments is shown. (F) Genomic DNA of HCT116 and HCT116-TOP3B-KO cells with or without treatment with NSC690634 (100 µM, 4 h) was collected and blotted on nitrocellulose membrane before immunoprobing by S9.6 antibody (7). RNase H-treated samples serve as negative controls.

We previously reported that TOP3Bccs induced by expression of the self-trapping toxic TOP3B-R338W mutant stimulated γH2AX signals in cells (11); therefore, we examined whether NSC690634 and NSC96932 might also lead to increased γH2AX signals. Indeed, cells treated with both NSC690634 and NSC96932 showed increased γH2AX signals in a time-dependent manner (SI Appendix, Fig. S6A).

Additionally, because it is known that the interaction between TDRD3 and TOP3B is important for the function of TOP3B in cells (6), we probed the TOP3B trapping effect of NSC690634 and NSC96932 using the band shift assay in the presence of TDRD3. Titration of these drugs with DNA- or RNA-TOP3Bcc showed similar accumulation of TOP3Bcc with or without TDRD3 (SI Appendix, Fig. S6B), indicating that the effect of these compounds on trapping TOP3B is unrelated to TDRD3. Together, these results demonstrate that the bisacridine compound NSC690634 and the thiacyanine compound NSC96932 trap TOP3B both in band shift assays and in cells.

NSC690634 and NSC96932 Mainly Trap RNA-TOP3Bccs in Cells.

Because TOP3B is the only RNA topoisomerase in human cells (8) and we have previously shown that TOP3Bccs form on both DNA and RNA (7, 11), we tested whether NSC690634 and NSC96932 induced TOP3Bccs on cellular DNA and/or RNA. Cellular TOP3Bccs on DNA and RNA can be distinguished by separating the two nucleic acid species in a CsCl gradient in invivo complex of enzyme (ICE) bioassay (11, 17). HEK293 cells transiently over-expressing TOP3B were treated with NSC690634 or NSC96932, and the DNA and RNA species in the cell lysate were separated using the ICE bioassay. Equal amounts of nucleic acids (1.5 µg each) from the DNA and RNA fraction were loaded on slot blots for independent immunodetection of TOP3Bcc on DNA vs. RNA. Results from three independent ICE bioassay experiments found that NSC690634 and NSC96932 induced TOP3Bcc mainly with RNA and to a lesser extent with DNA (Fig. 5E, loading control shown in SI Appendix, Fig. S7A). Using two additional purification methods for protein complexes cross-linked to cellular RNA, we verified that treatments with both NSC690634 and NSC96932 led to strong induction of RNA-TOP3Bccs (SI Appendix, Fig. S7 B and C). Additionally, with our modified RADAR assay, where both DNA- and RNA-topoisomerase cross-links were copurified (see details in Materials & Methods), cellular RNA and DNA can also be distinguished by selectively digesting away the RNA or DNA in the modified RADAR samples (11). The TOP3Bcc signals induced by NSC690634 and NSC96932 were greatly attenuated after cotreatment with RNase A/T1, while treatment with DNase I did not significantly reduce the levels of TOP3Bccs compared to the undigested sample (SI Appendix, Fig. S7D). Collectively, these results demonstrate that NSC690634 and NSC96932 mainly induce RNA-TOP3Bccs in cells, consistent with our band shift assays showing that NSC690634 and NSC96932 trap RNA-TOP3Bcc rather than DNA-TOP3Bcc (Fig. 4).

NSC690634 Specifically Traps TOP3B and Induces R-Loops in a TOP3B-Dependent Manner.

Next, we tested the specificity of these compounds for trapping TOP3Bccs and probed for the presence of cleavage complexes with other topoisomerases by modified RADAR assay. We observed that NSC690634 induced only TOP3Bccs, whereas long treatment with NSC96932 also induced cleavage complexes of TOP3A and TOP1 (SI Appendix, Fig. S8A). These results indicate that NSC690634 is specific for the trapping of TOP3B, and NSC96932 appears to be less specific, particularly after long periods of drug treatment.

Since TOP3B is important for resolving R-loops in cells (11), we examined the impact of TOP3B poison on cellular R-loops. We chose to focus on NSC690634 due to its higher specificity for trapping TOP3B than NSC96932. In agreement with previous reports (7, 10), R-loop levels were higher in the HCT116-TOP3B-KO cells than the parental HCT116 cells (Fig. 5F, loading control in SI Appendix, Fig. S8B). Importantly, treatment with NSC690634 markedly increased the R-loops in the parental HCT116 cells, but not in the HCT116-TOP3B-KO cells (Fig. 5F and SI Appendix, Fig. S8B). As a control, treatment of the genomic DNA samples with RNase H eliminated all signals, demonstrating the specificity of the S9.6 antibody for R-loop detection (Fig. 5F). Our combined results indicate that NSC690634 specifically stabilizes RNA-TOP3Bccs and increases R-loops in a TOP3B-dependent manner.

Structure–Activity Insights.

Because NSC690634 is a bisacridine with a 3-carbon linker between the two acridine moieties, we tested acridine compounds without or with longer carbon linkers for their effectiveness in trapping TOP3B (structures shown in Fig. 6A). Strikingly, modified RADAR assays showed that only NSC690634 with the 3-carbon linker induced TOP3Bccs (Fig. 6B), and alteration of the linker length inhibited formation of TOP3Bccs. In vitro band-shift assays also showed that neither the potent TOP2 poison m-AMSA nor o-AMSA, its inactive isomer (3), was able to induce TOP3Bccs with either DNA or RNA substrates (Fig. 6C). Consistent with earlier results, NSC690634 strongly induced RNA TOP3Bccs but had a relatively mild effect on DNA TOP3Bcc. Lengthening the linkers of the bisacridine compounds to a 6- or 8-carbon linker completely abolished formation of both DNA- and RNA-TOP3Bcc in vitro (Fig. 6C). These combined results indicate that the length of linker between the two acridine moieties is critical for trapping RNA-TOP3Bccs by bisacridines.

Fig. 6.

Structure–activity analysis of bisacridine analogs of NSC690634 demonstrates that linker length is critical for inducing TOP3Bccs. (A) Chemical structures of a series of bisacridines. The length of the carbon linker between the two acridine moieties increases from left to right. (B) HEK293 cells transiently over-expressing TOP3B (TOP3B OE) were treated with DMSO (control) or with the indicated bisacridines (100 µM for 1 h) prior to analysis by modified RADAR and probed with anti-Flag antibody. The same samples probed with anti-dsDNA antibody served as loading controls. Only NSC690634 showed a strong induction of TOP3Bccs while all the other structurally related compounds with different linker lengths failed to induce cellular TOP3Bccs. (C) In vitro band shift assays demonstrate that linker length is critical for inducing TOP3Bccs by the bisacridine NSC690634. Compounds without linkers did not trap TOP3B, whereas NSC690634 (3-carbon linker) strongly enhanced TOP3Bccs with RNA. Bisacridine compounds with longer carbon linkers not only failed to induce TOP3Bcc but they also destabilized TOP3Bccs with both DNA and RNA.

The thiacyanine NSC96932 shared a degree of structural similarity with NSC690634, in that they both contain a pair of nitrogen-substituted small aromatic rings joined by a 3-carbon linker. We also tested a series of thiacyanine compound structurally related to NSC96932 for their effects on trapping TOP3Bcc (structures shown in Fig. 7A). Analogous to the results with bisacridines, increasing the lengths of the linkers in thiacyanine compounds decreased trapping of TOP3Bccs in the modified RADAR assay (Fig. 7B). All 3 thiacyanine compounds with 3-carbon linkers were highly effective in trapping TOP3Bccs, whereas thiacyanine compounds with 5- or 7-carbon linkers were not. Among the thiacyanine compounds with 3-carbon linkers, NSC96932 was the most potent in inducing TOP3Bccs. The additional methyl group present on the carbon linker of NSC96932 might be important in efficiently trapping TOP3Bccs (Fig. 7B). Further structure–activity relationship studies are warranted.

Fig. 7.

Structure activity analysis of thiacyanine analogs of NSC96932 demonstrates that the linker length is critical for inducing TOP3Bccs. (A) Chemical structures of a series of thiacyanine analogs. The length of the carbon linker between the two indole ring moieties increases from Left to Right. (B) HEK293 cells transiently over-expressing TOP3B (TOP3B OE) were treated with the indicated thiacyanine analogs (100 µM for 1 h) prior to analysis by modified RADAR and probed with anti-Flag antibody. The same samples probed with anti-dsDNA antibody served as loading controls. Only the thiacyanine with 3-carbon linkers (NSC96932, NSC356711, and NSC93472) induced endogenous TOP3Bccs while the compounds with longer carbon-linker lengths failed to induce cellular TOP3Bccs.

Conclusions

We have designed a Comparative Cellular Cytotoxicity Screen allowing high-throughput screening for TOP3B poisons that trap TOP3Bccs. We have also modified and optimized the RADAR assay and developed an in vitro band shift assay to validate and study the mechanism of hit compounds from the screen. Here, we report and characterize two poisons of TOP3B, the bisacridine NSC690634 and the thiacyanine derivative NSC96932, both of which induced genomic DNA damage measured by γH2AX. TOP3B-KO cells showed higher resistance to NSC690634 and NSC96932; however, the differential cytotoxicity was observed only within a relatively modest concentration range, suggesting that the compounds likely impact other cellular targets at higher concentrations. Modified RADAR assays and in vitro band shift assays both show that NSC690634 and NSC96932 induce primarily RNA-TOP3Bccs, with modest effect on DNA-TOP3Bccs.

The family of bisacridine compounds has been reported to act as bactericidal, antiparasitic, and anticancer agents (18–21). The monomer m-AMSA (a.k.a. amsacrine, Fig. 6A) is a TOP2 poison used as an anticancer agent (3, 22, 23). A series of bisacridine compounds with different linker lengths have been proposed as antitumor agents, partially due to the propensity of these compounds to intercalate within DNA. Notably, most of the published studies focused on bisacridines with an optimized C6-C10 linker, which showed the strongest DNA intercalation (19, 21). Our preliminary structural activity relationship studies show that monomers like m-AMSA and o-AMSA could not trap TOP3B. While NSC690634 with a C3 linker showed strong trapping TOP3Bccs in cells and in band shift assays, bisacridine compounds with longer linkers (C6 to C8) failed to trap TOP3Bccs. These results suggest that NSC690634 traps TOP3Bcc via a different mechanism than simple DNA intercalation.

The cyanine compound family is also known to form aggregates on DNA (24). Coupled with their large extinction coefficients, these cyanine compounds are widely used as molecular probes and labels. It has been reported that cyanine dyes with long linkers preferentially form aggregates (24), while our results show that the short C3 linker in NSC96932 is critical for trapping TOP3B. This suggests that TOP3B trapping by NSC96932 is unlikely to rely on aggregation with DNA.

Although NSC690634 and NSC96932 belong to different chemical families, they share a certain degree of structural similarities: They both have small aromatic rings joined by a short C3 linker. Both compounds preferentially trap TOP3Bccs on RNA, and longer linkers in either compound family abolish their capacity to trap TOP3Bccs. Further structural and mechanistic studies on how TOP3Bccs are formed and how these compounds trap TOP3Bccs are warranted. Nonetheless, the compounds we report here can serve as a base for further structural modification. Our identification of TOP3B poisons has implications for basic research to study the biological functions of TOP3B as well as for developing potential anticancer and antiviral drugs.

Materials and Methods

Mammalian Cell Culture.

Human embryonic kidney HEK293 cells, human colorectal carcinoma HCT116 cells, and human lung carcinoma A549 cells (all obtained from American Type Culture Collection) were cultured in 1 × Dulbecco's modified Eagle's medium (Life Technologies), supplemented with 10% (v/v) Fetal Bovine Serum (FBS, Gemini), 100 U/mL penicillin, 100 µg/mL streptomycin, and 1 × GlutaMax (ThermoFisher). HAP1 cells (obtained from Horizon Discovery) were cultured in 1 × Iscove’s Modified Dulbecco’s medium (Life Technologies) supplemented with 10% (v/v) FBS (Gemini), 100 U/mL penicillin, and 100 µg/mL streptomycin. Cells were incubated at 37 °C with 5% CO₂ incubator. Human TOP3B-Myc-Flag cDNA open reading frame (ORF) was purchased from OriGene (RC223204) and transfected in HEK293 cells using Lipofectamine 3000 Reagent (ThermoFisher Scientific) according to the manufacturer’s protocol for 48 h before drug treatments.

Generation of HCT116-TOP3B-KO cells was as described previously, and HAP1-TOP3B-KO cells were generated using the same CRISPR constructs and selection process (7). HCT116 and HCT116-TOP3B-KO cells stably expressing GFP or mCherry were generated using lentivirus containing pFUGW-FerH-ffLuc2-eGFP (Addgene #71393) (25) or pFUGW-FerH-ffLuc2-mCherry (gift from Dr. Chi-Ping Day, Laboratory of Cancer Biology and Genetics, CCR, NCI, NIH) followed by Fluorescence activated cell sorting and subculturing into monoclonal populations.

Fluorescent Microscopy Imaging and Analysis.

All microscopy images were acquired using Cytation 5 (BioTek, Agilent), with 4× objective on cells seeded in 384-well black cell-culture plates with a clear bottom in GFP and Texas Red channels as well as a bright-field channel. Automated analysis of all images was done using Gen5 Image Prime Software (BioTek, Agilent).

Modified RADAR Assay for Detection of TOP3Bcc.

Our modified RADAR assays were carried out with the goal to purify both DNA and RNA species from the cell lysates, and the entire procedure was carried out under DNase- and RNase-free conditions. Briefly, after treatment with specified compound at indicated concentration and time, cells with or without prior transfection with the Flag-tagged TOP3B (TOP3B OE) were directly lysed with 400 µL of DNAzol (Invitrogen). Instead of the optional centrifugation step at this point, which would remove RNA and other insoluble tissue fragments, the entire sample was collected in order to retain both DNA and RNA species. The samples were then precipitated with half-volume of 100% ice-cold ethanol and incubated at −20 °C for 20 min. Instead of spooling the precipitated DNA as described in the manufacturer’s manual, which would mainly enrich DNA species, all nucleic acids were pelleted by centrifugation of the entire sample at 15,000 rpm for 15 min at 4 °C in order to retain both DNA and RNA species. The pellets were then washed with 75% ice-cold ethanol twice, air-dried, and resuspended in 100 µL RNase-free Tris-ethylenediaminetetraacetic acid (TE) buffer instead of 8 mM NaOH, as RNA is unstable under alkaline conditions.

To eliminate RNA-TOP3Bcc from TOP3Bccs, 1 µg nucleic acids samples were incubated with 1 µg/µL RNase A and 1 U/µL RNase T1 at 37 °C for 2 h, followed by adding 1/10 volume of 3 M sodium acetate, and 3 volume of 100% ice-cold ethanol. To eliminate DNA-TOP3Bcc from TOP3Bccs, 1 µg nucleic acids samples were incubated with 0.1 U/µL DNase I at 37 °C for 2 h, followed by adding 1/10 volume of 3 M sodium acetate, and 3 volume of 100% ice-cold ethanol. Samples were incubated at −80 °C for 1 h and the remaining nucleic acid polymers were pelleted by centrifugation at 15,000 rpm at 4 °C for 15 min. Nucleic acid pellets were washed with 75% ice-cold ethanol twice, air dried, and resuspended in 100 µL TE buffer.

For the detection of TOP3Bccs, 1 µg to 3 µg of nucleic acids samples per well were applied to nitrocellulose 0.45-μm membranes (Bio-Rad Laboratories, CAT#: 1620115) through a slot-blot vacuum manifold as described (11). TOP3Bcc signals were probed by anti-Flag antibody (mouse monoclonal, clone M2, Sigma) or anti-TOP3B antibody [rabbit monoclonal, (EP7779), Abcam]. Loading controls were probed with anti-ds DNA antibody (ab27156, Abcam).

ICE Bioassays.

The ICE bioassays were carried out as described with minor modifications (11, 17). Briefly, ~1 million HEK293 cells transiently over-expressing TOP3B were treated with the indicated concentration of TOP3B poisons for 1 h. The treated and nontreated control cells were pelleted and immediately lysed with 1 mL of 1% sarkosyl. After homogenization with a Dounce, cell lysates were gently layered on step gradients containing four different CsCl (Sigma-Aldrich, CAT#:746487-1KG) solutions (2 mL of each) of the following densities: 1.82, 1.72, 1.50, and 1.45. The gradients were prepared by diluting a stock solution of CsCl of density 1.88. Cesium sulfate (Sigma-Aldrich, CAT#:C5205-50G) was included in the bottom solution of density 1.82 to facilitate flotation of the RNA, and sodium thiocyanate (Sigma-Aldrich, CAT#: S7757-1KG) was included in topmost solution of density 1.45 to facilitate the complete removal of noncovalently bound proteins from the nucleic acid species. Samples were centrifuged at 30,700 rpm in a Beckman SW40 rotor for 24 h at 20 °C. Half-milliliter fractions were collected from the bottom of the tubes. Fractions containing DNA and RNA were pooled separately, quantitated, diluted with 25 mM sodium phosphate buffer (pH 6.5), and applied to nitrocellulose 0.45-μm membranes (Bio-Rad Laboratories, CAT#: 1620115) through a slot-blot vacuum manifold as described (11). TOP3Bccs were detected with an anti-Flag antibody (mouse monoclonal, clone M2, Sigma), and the loading control was carried out by staining with methylene blue stain (Molecular Research Center) following the manufacturer’s protocol.

Isolation of Cellular Covalent RNA–Protein Adducts from Cells Using TRIzol® Reagent.

RNA–protein adducts were isolated from cells using protein-crosslinked RNA extraction (XRNAX), as described previously (11, 26). Briefly, 10⁷ HEK293 cells transiently over-expressing TOP3B were treated with the indicated concentration of TOP3B poisons for 1 h. Cells were lysed in 1 mL TRIzol™ Reagent (Invitrogen, USA, CAT#:15596026) by pipetting the samples up and down several times followed by incubation at room temperature for 5 min. Then, 200 μL chloroform was added to the samples and mixed thoroughly by inverting the tubes. After incubation at room temperature for 3 min and centrifugation for 10 min at 7,000 × g at 4 °C, the aqueous phase was removed, and the interphase was transferred to a new tube. The interphase was gently washed twice with 1 ml low SDS buffer (50 mM Tris-Cl, 1 mM ethylenediaminetetraacetic acid (EDTA), 0.1% SDS), resuspended in low SDS buffer, centrifuged at 5,000 × g for 2 min at room temperature, and the supernatant was stored. Pellets were washed again with 1 mL of low SDS buffer, then twice more with 1 mL high SDS buffer (50 mM Tris-Cl, 1 mM EDTA, 0.5% SDS), and all the supernatants were stored following centrifugation. NaCl was added to a final concentration of 300 mM to each of the interphase eluates, along with 10 µg of RNase-free glycogen and 1 mL isopropanol before mixing by inversion. Samples were spun down for 15 min with 18,000 × g at −10 °C. Supernatant were discarded; pellets were washed with 70% ethanol, with residual ethanol removed; and the pellets were resuspended in nuclease-free water at 4 °C. 10× TURBO DNase Buffer (ThermoFisher Scientific, USA) was added to the resuspended samples to 1× concentration along with 10 μL TURBO DNase (ThermoFisher Scientific, USA) and incubated for 60 min at 37 °C with constant shaking at 700 rpm. After DNase treatment, samples were isopropanol precipitated in the presence of 300 mM NaCl and dissolved in diethyl pyrocarbonate (DEPC)-treated water. RNA purity and concentrations were estimated by spectroscopy on a NanoDrop 1000 Spectrophotometer (ThermoFisher Scientific, USA). Samples were slot-blotted on nitrocellulose membrane, and RNA-TOP3Bccs were detected using mouse monoclonal anti-FLAG M2 antibody (Millipore Sigma, St. Louis, MO, CAT#: F1804), and the loading control was carried out by staining with methylene blue stain (Molecular Research Center) following the manufacturer’s protocol.

Phenol–Toloul Extraction (PTex) to Purify Cellular Covalent RNA–Protein Adducts.

For isolation and detection of RNA-TOP3Bcc we performed the PTex method as described previously (27, 28). Briefly, 5 × 10⁶ HEK293 cells transiently over-expressing TOP3B were treated with the indicated concentration of TOP3B poisons for 1 h, harvested and suspended in 600 μL of phosphate buffered saline (PBS). Cell suspensions were mixed with 200 μL each of neutral phenol (Phenol solution Equilibrated with 10 mM Tris HCl, pH 8.0, 1 mM EDTA, Millipore Sigma, Catalog#: P4557-400ML), 1-Bromo-3-chloropropane (BCP, Millipore Sigma, Catalog #:B9673-200ML) and Toluene (Millipore Sigma, Catalog #: 32249-1L) for 1 min (21 °C, 2,000 rpm), followed by centrifugation (20,000 g, 3 min, 4 °C). The aqueous phase was carefully taken out and transferred to a 2-mL eppendorf tube containing 300 μL of solution D (5.85 M guanidine isothiocyanate; 31.1 mM sodium citrate; 25.6 mM N-lauryosyl-sarcosine; 1% 2-mercaptoethanol). Then, 600 μL phenol and 200 μL BCP were added to the samples, mixed, and centrifuged (20,000 g, 3 min, 4 °C). After phase separation, the upper aqueous and the lower organic phases were removed using a syringe with a blunt needle. The resulting interphase was mixed with 400 μL water, 200 μL ethanol, 400 μL phenol, and 200 μL BCP (1 min, 21 °C, 2,000 rpm) and centrifuged (20,000 g, 3 min, 4 °C). The upper aqueous and the lower organic phases were carefully removed, while interphase was precipitated with 9 volumes of ethanol (−20 °C, overnight). Samples were centrifuged (4 °C, 30 min, 20,000 g), pellets dried, and then solubilized in RNase-free DEPC-treated water. RNA purity and concentrations were estimated by spectroscopy on a NanoDrop 1000 Spectrophotometer (ThermoFisher Scientific, USA). Samples were slot-blotted on the nitrocellulose membrane and RNA-TOP3Bccs were detected using the mouse monoclonal anti-FLAG M2 antibody (Millipore Sigma, St. Louis, MO, CAT#: F1804), and the loading control was carried out by staining with methylene blue stain (Molecular Research Center) following the manufacturer’s protocol.

R-Loop Detection by Slot Blotting.

Genomic DNA from HCT116 and HCT116-TOP3B-KO cells treated with DMSO or NSC690634 (100 µM, 4 h) was extracted using the DNA–RNA immunoprecipitation sequencing (DRIP) protocol as described previously (29). Briefly, cells were lysed in TE buffer containing SDS and proteinase K (at 37 °C overnight) before being extracted with phenol/chloroform/isoamyl alcohol (25:24:1) and ethanol precipitated. Genomic DNA was resuspended in TE buffer and digested using a cocktail of restriction enzymes (HindIII, SspI, EcoRI, BsrGI, and Xbal; 30 U each), treated with RNase A (10 µg/mL) and shortcut RNase III (2 units, New England Biolabs) before purification again with phenol/chloroform/isoamyl alcohol (25:24:1). For control samples, 10 µg of genomic DNA was treated with 20 U of RNase H at 37 °C for 3 h. The resulting genomic DNA samples were spotted on nitrocellulose 0.45-μm membranes (Bio-Rad Laboratories, CAT#: 1620115), cross-linked and blocked with PBS-Tween (0.1%) buffer containing 5% nonfat milk. The membrane was probed with mouse S9.6 antibody (1:500 dilution, at 4 °C overnight) and developed using standard enhanced chemiluminescence techniques. The same samples probed with anti-dsDNA antibodies served as loading controls.

Recombinant Human TOP3B Production.

TOP3B was initially polymerase chain reaction (PCR) amplified from Human TOP3B-Myc-Flag cDNA ORF (CAT#: RC223204) using forward primer 5′-CGGGGTACCATGAAGACTGTGCTCATGG-3′ and reverse primer 5′-CCGCTCGAGTCATACAAAGTAGGCGGCCAG-3′ and cloned into Gateway entry vector pENTR3C (Invitrogen, CAT#: A10464). TOP3B was then subcloned by Gateway LR recombination (Thermo Fisher) into pDest-635 (22876-X01-635) for insect cell expression which includes an N-terminal His6 tag. Bacmid was prepared in DE77, a DH10Bac-derived strain (Bac-to-Bac system, Thermo Fisher), and after purification, bacmid DNA was verified by PCR amplification across the bacmid junctions. Bacmids were transfected in SF-9 cells using polyethylenimine (1 mg/mL with 5% glucose; Polysciences, CAT#: 23966), recombinant baculovirus stock was collected and titrated using ViroCyt (Beckamn). Two liters of Tni-FNL cells were set in a baffled 5-l Thomson Optimum Growth Flask in GIBCO Express 5 medium with 18 mM glucose at a cell density of 1 × 10⁶ cells/ml at 27 °C and 24 h later infected at an multiplicity of infection of 3. After 3 d of incubation at 21 °C, cell pellets were collected by centrifugation at 2,000 rpm for 11 min and flash-frozen on dry ice. The cell pellet was thawed by the addition of 200 mL of lysis buffer (20 mM HEPES, 300 mM NaCl, 1 mM Tris (2-carboxyethyl) phosphine (TCEP), and 1:100 v/v of Sigma protease inhibitor P8849) and homogenized by vortexing. The cells were lysed by performing two passes on an M-110EH-30 microfluidizer (Microfluidics) at 7,000 psi, clarified at 100,000 × g for 30 min at 4 °C using an optima L-90K ultracentrifuge (Beckman), filtered (0.45 μm), and applied to a f20-mL IMAC HP column (GE Scientific) that was pre-equilibrated with lysis buffer containing 50 mM imidazole on a Bio-Rad NGC. The column was washed with lysis buffer containing 50 mM imidazole, and proteins were eluted with lysis buffer containing 500 mM imidazole. After SDS-PAGE/coomassie staining, positive fractions were pooled, dialyzed to 20 mM HEPES, 50 mM NaCl, 1 mM TCEP, 0.5 mM phenylmethylsulfonyl fluoride, 1:1,000 v/v of protease inhibitor, and 50% glycerol, pH 7.2. Protein concentration was determined and stored at −80 °C for future use.

In Vitro Band Shift Assay.

The single-strand DNA oligo substrate GGGATTATTGAACTGTTGTTCAAACTTTAGAACTAGCCATCCGATTTACACTTTGCCCCTATCCACCCC-3’FITC or the corresponding RNA oligo substrate G​GGA​UUA​UUG​AAC​UGU​UGU​UCA​AAC​UUU​AGA​ACU​AGC​CAU​CCG​AUU​UACACUUUGCCCCU-3′ Cy5 was synthesized by Integrated DNA Technologies. First, 75 nM of substrate was combined with 180 nM of purified recombinant TOP3Β in 100 mM potassium glutamate (pH 7.0), 3 mM MgCl₂, 0.02% v/v Tween-20, and 1 mM dTT, with the indicated drug compound or DMSO (10% v/v) and incubated at 30 °C for 60 min before addition of SDS (0.2%) to the samples to stop the reaction. Then, equal volume of tris-glycine-SDS loading buffer was added to the samples and boiled for 5 min before the samples were resolved on 4 to 20% tris-glycine-SDS-PAGE and directedly visualized via the fluorescence of FITC or Cy5 on a typhoon scanner.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information, including Dataset S1.