New 1,1-diarylethylene FOXM1 inhibitors potently reduce intracellular FOXM1 and suppress high-grade serous ovarian carcinoma cell viability

As an oncogenic transcription factor highly upregulated in numerous cancer types and responsible for promoting multiple hallmarks of the disease, Forkhead Box M1 (FOXM1) is a promising target for treatment against high-grade ovarian serous carcinoma (HGSC), a lethal and aggressive cancer that overexpresses FOXM1 and its transcriptional pathway. Several FOXM1 inhibitors have been established, but none have advanced to clinical trials. In this study, we report that our recently developed 1,1-diarylethylene FOXM1 inhibitors (NB compounds) potently and selectively inhibit HGSC cell viability, more so than a panel of other reported FOXM1 inhibitors. The NB compounds decreased FOXM1 expression through proteasomal degradation, which resulted in decreased expression of its downstream target cyclin B1 (CCNB1), without affecting other FOX family members. Moreover, the NB compounds exhibited robust anti-cancer effects by promoting apoptosis, suppressing colony formation more potently than other FOXM1 inhibitors, and synergizing with carboplatin to inhibit the viability of HGSC cells. Our data demonstrate that the NB compounds are promising FOXM1 inhibitors that may serve as a novel therapeutic strategy for HGSC and other cancers whose aggressive cancer biology is driven by FOXM1 and its oncogenic transcriptional pathway. FOXM1 inhibitors on CAOV3 and OVCAR4


Introduction
High-grade serous ovarian carcinoma (HGSC) comprises over 60% of ovarian cancers 1 and accounts for up to 90% of ovarian cancer-related deaths 2 . Hence, effective treatments for HGSC, especially for relapsed disease, are needed. Forkhead Box M1 (FOXM1) is a master transcriptional regulator of cell cycle and mitotic progression and has numerous oncogenic functions in multiple cancers, including ovarian cancer [3][4][5][6][7][8] . We have previously shown that FOXM1 is widely overexpressed in HGSC 9 and other cancers 10 , driven by multiple molecular mechanisms such as copy-number gains and disruption in p53 and Rb signaling [8][9][10] . Mounting evidence supports the multifunctional oncogenic role of FOXM1 in promoting ovarian cancer cell proliferation, invasion and metastasis, chemotherapy resistance, cancer cell stemness, genomic instability, and altered cellular metabolism 8 . The FOXM1 transcriptional pathway is aberrantly activated in over 85% of high-grade serous ovarian carcinoma (HGSC) cases 9,11 , second only to TP53 mutations (100% of HGSC cases 12 ), suggesting that FOXM1 may be a promising target for HGSC therapy.
Recognizing the oncogenic functions of FOXM1 prompted the development of small molecule inhibitors (SMIs) targeting FOXM1 in pursuit of clinical translatability. Among FOXM1 inhibitors, thiostrepton and Forkhead Domain Inhibitor 6 (FDI-6) are the most extensively studied. Thiostrepton is a thiazole antibiotic that promotes FOXM1 degradation [13][14][15] , but its exact mechanism of action (MOA) is presently debated [15][16][17] . FDI-6 was identi ed through a high-throughput screen (HTS) for SMIs that disrupt the interaction between the DNA binding domain of FOXM1 and its consensus promoter sequence 18 . However, as other forkhead box (FOX) family members share this sequence 19,20 , this approach may limit the selectivity of FDI-6 and disregards FOXM1 functions related to promoter binding to non-consensus sequences and super-enhancer regions [21][22][23][24] . Thus, inhibiting the DNA binding domain function alone may not be su cient to prevent these other critical interactions. Other reported FOXM1 inhibitors include monensin 25 , N-phenylphenanthren-9-amine 26 , and RCM-1 27 , all of which are much less characterized and appear to work by various mechanisms with less de ned e cacy and selectivity.
They also decreased migration and invasion phenotypes and suppressed epithelial-to-mesenchymal transition (EMT) and metastasis-related gene expression in TNBC cells 29 .
HGSC has a highly similar molecular pro le to TNBC, including TP53 mutations, Rb pathway dysregulation, homologous combination DNA repair de ciency, and extensive somatic copy number alterations 30 . Moreover, both TNBC and HGSC show similar functional FOXM1 copy number gains, which translate to increased FOXM1 gene and protein expression 10 . Therefore, we sought to investigate the anticancer effects of these novel FOXM1 inhibitors in HGSC. Here, we elucidate the capacity of the NB compounds to promote the death of HGSC cells, explore the underlying mechanisms of their actions, and provide a systemic and thorough characterization of the potency and e cacy of the NB compounds in comparison to reported FOXM1 inhibitors.
Cell Culture and Reagents CAOV3 and OVCAR5 cell lines were a gift from Dr. Anirban Mitra (Indiana University). CAOV3, OVCAR5, OVCAR8 (National Cancer Institute Division of Cancer Treatment and Diagnosis Cell Line Repository), COV318 (Sigma), and FUOV1 (MD Anderson) cell lines were cultured in DMEM (Corning) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin-streptomycin (pen-strep, Gibco Stocks of all cell lines were authenticated using short tandem repeat analysis at the DNA Services Facility, University of Illinois at Chicago. All cell line stocks were con rmed to be mycoplasma-free by PCR analysis using Myco nd ™ Mycoplasma PCR Detection Kit (Clongen Laboratories, LLC).

Western blotting
Whole cell protein extracts were prepared using radio-immunoprecipitation assay (RIPA) buffer (1X PBS, 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate [SDS]) supplemented with protease and phosphatase inhibitors (Sigma-Aldrich). Brie y, cells were lysed for 15 minutes in RIPA buffer and then sonicated, and the resulting extracts were centrifuged at 4 °C for 10 minutes at 16,100 x g to remove cell debris. Protein concentrations were measured using the Pierce ™ BCA Protein Assay Kit (Fisher Scienti c).

Apoptosis Assays
Cells were seeded into 6-well plates and treated with the designated compound(s) until the desired time point. The spent medium containing oating cells as well as trypsinized cells from the wells were collected, centrifuged at 500 x g to form a pellet, and washed with PBS. Cells were then prepared for apoptosis analysis using the Muse ® Caspase-3/7 Kit (Luminex Corporation) according to the manufacturer's instructions. Cellular apoptosis was analyzed by ow cytometry using the Guava ® Muse ® Cell Analyzer (Luminex Corporation).
All assays were performed in three independent trials and analyzed using GraphPad Prism 7 software.

Colony Formation Assays
For two-dimensional, anchorage-dependent colony formation assays, cells in single-cell suspension were seeded in triplicate wells of 6-well plates, treated with the designated compound(s), and allowed to form colonies for 7-14 days, depending on the cell line. Following incubation, cells were xed in methanol, stained with crystal violet, washed with water, and air-dried overnight. Pictures of the wells were captured and colonies were counted using Count and Plot Histograms of Colony Size (countPHICS) software 34 , a macro written for ImageJ. The threshold for colony size was automatically determined by countPHICS.
For three-dimensional, anchorage-independent (soft agar) colony formation assays, cells in single-cell suspension were seeded into a 0.4% SeaPlaque ™ Agarose (Lonza Bioscience) liquid solution and placed on top of a solidi ed layer of 0.8% SeaPlaque ™ Agarose solution in triplicate wells of 6-well plates. The top layer with cells was allowed to solidify at room temperature for at least 30 minutes. Cells were then treated with the designated compound(s) and allowed to form colonies for 14 days. Following incubation, cells were stained with crystal violet and washed with water. Pictures of the wells were taken, and colonies were counted using countPHICS 34 . The threshold for colony size was automatically determined by countPHICS.

Cell Viability Assays
Cells were seeded into 96-well plates and treated in serial dilutions with the designated compound(s) for 72 hours. Medium was then removed from the wells, and the cells were frozen at -80 °C. Cell viability was analyzed using the CyQUANT ™ Cell Proliferation Assay, for cells in culture (Invitrogen) according to manufacturer's instructions. Fluorescence intensity was measured using POLARstar OPTIMA microplate multi-detection plate reader (BMG LabTech) with settings according to manufacturer's instructions.
For cell viability assays examining synergistic interactions, the same method was followed as described above. Concentrations of each compound were decided based on the diagonal method to measure drug synergy 35 . Synergy was analyzed using CompuSyn software (ComboSyn, Inc.).
All assays were performed in three independent trials with three technical replicates per independent trial and analyzed using GraphPad Prism 7 software.

NB compounds decrease FOXM1 protein expression levels in HGSC cells
To assess whether the NB compounds function as FOXM1 inhibitors in HGSC cells, HGSC cell lines CAOV3 and OVCAR4 were treated with the NB compounds for 6 hours, 24 hours, and 48 hours and examined for FOXM1, p-FOXM1 (Thr600), Cyclin B1 (CCNB1), FOXA1, and FOXO3A protein expression.
Phosphorylation of FOXM1c at the threonine 600 residue (threonine 585 residue on FOXM1b) is widely considered critical for FOXM1 function 36,37 . Once activated, FOXM1 promotes the transcription of numerous cell cycle genes, including CCNB1 38-40 . FOXA1 is a FOX family member suggested to have oncogenic functions in epithelial ovarian cancer cells 41,42 , and FOXO3A is a FOX family member with tumor suppressor functions 43 . Like FOXM1, both FOXA1 and FOXO3A are strong transcriptional regulators that recognize the canonical RYAAAYA forkhead binding motif and therefore served as FOX family control proteins.
NB-55 decreased FOXM1, p-FOXM1, and CCNB1 protein expression at 6, 7, and 8 µM concentrations in CAOV3 cells ( Figure 1A). However, there was a narrow window of activity, with almost all CAOV3 cells dying by 24 hours when treated with 8 µM NB-55 and by 48 hours with 7 µM NB-55 treatment. Interestingly, at six hours treatment, p-FOXM1 and CCNB1 showed reduced expression while FOXM1 levels remained stable. This observation suggests that the NB compounds may also promote the dephosphorylation of FOXM1, which would be a novel MOA for the NB compounds, in addition to FOXM1 degradation. FOXA1 increased in expression at 48 hours, and FOXO3A temporarily increased in expression at 24 hours. NB-55 treatment of OVCAR4 cells ( Figure 1B) demonstrated a similar pattern for FOXM1, p-FOXM1, and CCNB1 expression, though CCNB1 protein expression was not detectable by western blot at six hours. There was a slight decrease in FOXA1 at 24 hours, which appeared to rebound at 48 hours, and FOXO3A did not appear to change considerably.
NB-73 induced a prominent decrease in FOXM1, p-FOXM1, and CCNB1 expression in CAOV3 cells at 1 and 2.5 µM concentrations in a time-and dose-dependent manner ( Figure 1C). FOXA1 expression increased at 48 hours NB-73 treatment, and FOXO3A expression was largely unaffected by NB-73 treatment. In OVCAR4 cells ( Figure 1D), FOXM1, p-FOXM1, and CCNB1 demonstrated a similar time-and dose-dependent decrease in expression at 24-and 48-hour treatment. FOXA1 expression decreased at 24 and 48 hours, and FOXO3A expression was largely unaffected by NB-73 treatment.
NB-115 also induced a prominent decrease in FOXM1, p-FOXM1, and CCNB1 expression at 1 and 2.5 µM concentrations in a time-and dose-dependent manner in CAOV3 cells (Figure 2A). FOXA1 expression increased at both 24 and 48 hours, and FOXO3A expression slightly increased at 24 hours. In OVCAR4 cells ( Figure 2B), FOXM1, p-FOXM1, and CCNB1 demonstrated a similar time-and dose-dependent decrease in expression at 24-and 48-hour treatment. FOXA1 expression increased at 24 hours and then decreased at 48 hours. FOXO3A expression was largely unaffected by NB-115 treatment.
For comparison, we also examined changes in protein expression in CAOV3 and OVCAR4 cells treated with FDI-6 or thiostrepton, two of the most established FOXM1 inhibitors. Given the reported mechanisms of thiostrepton [15][16][17] and FDI-6 18 , we anticipated that thiostrepton would promote degradation of overall FOXM1 protein while FDI-6 would inhibit downstream CCNB1 expression without promoting the degradation of FOXM1 or p-FOXM1 protein. In CAOV3 cells ( Figure 2C), FDI-6 did not induce a reduction of FOXM1 or p-FOXM1 expression at 24 hours and 48 hours, while CCNB1 expression decreased equally at both 24 and 48 hours. FOXA1 expression decreased at 48 hours, while FOXO3A expression did not change. Thiostrepton, on the other hand, induced a decrease of FOXM1 and p-FOXM1 expression at 24 hours and CCNB1 expression at 48 hours in a time-dependent manner. FOXA1 expression decreased at 48 hours, and FOXO3A expression temporarily increased 4-fold at 24 hours. In OVCAR4 cells ( Figure 2D), FDI-6 did not induce a reduction of FOXM1 or p-FOXM1 expression at 24 and 48 hours, while CCNB1 expression transiently decreased at 24 hours. FOXA1 was largely unchanged, and FOXO3A expression increased slightly at 24 hours. Thiostrepton induced a decrease of FOXM1, p-FOXM1, and CCNB1 expression at 24 and 48 hours in a time-dependent manner. FOXA1 expression decreased markedly at 24 and 48 hours, while FOXO3A expression increased at 24 hours and then decreased at 48 hours.

NB compounds promote FOXM1 proteasomal degradation in HGSC cells
To determine the mechanism mediating reduced expression of FOXM1 by NB compounds, we co-treated OVCAR4 or COV318 cells with the NB compounds and the proteasome inhibitor MG132. We found that MG132 was toxic in CAOV3 cells, so COV318 cells served as a more suitable HGSC model for this experiment. In both OVCAR4 ( Figure 3A) and COV318 ( Figure 3B) cells, FOXM1 expression markedly decreased upon treatment with 2.5 µM NB-73 or NB-115 but was rescued when proteasomal degradation was inhibited. This observation is consistent with our previously reported observations in breast cancer cells, where FOXM1 protein in breast cancer cell extracts treated with NB-73 was more susceptible to degradation by pronase, suggesting the NB compounds destabilize FOXM1 protein, and MG132 reversed the downregulation of FOXM1 by NB-73 in MDA-MB-231 cells 28 .
It has been previously reported that proteosome inhibition decreases FOXM1 expression 16 . Indeed, we did observe this phenomenon in some of the HGSC cell lines we tested. In fact, it appeared that this effect was dependent on both the cell line and MG132 concentration (Supplementary Figure S1). We also found that treatment with 1 µM MG132 for 24 hours was toxic in many, but not all, HGSC cell lines. It has been previously reported that the upregulation of HSP70, which is caused by the inhibition of its degradation, is responsible for the degradation of FOXM1 by proteasome inhibitors 17 . HSP70 purportedly binds to the FOXM1 DNA binding domain, inhibiting the ability of FOXM1 to promote its own transcription 17 . However, our previous studies were not able to con rm this positive autoregulatory loop in several cell types, including HGSC cells 10 . While we also observed an increase in HSP70 expression with proteosome inhibition, this did not consistently correlate with a decrease in FOXM1 expression (Supplementary Figure  S1). From our data, we cannot conclude whether the decrease in FOXM1 expression by proteasome inhibition is caused by the upregulation of HSP70, but our results suggest that the concentration of MG132 is critical when studying this phenomenon and FOXM1 expression generally.
Interestingly, after six hours treatment, NB-55 caused a decrease in p-FOXM1 expression with minimal change in FOXM1 expression in both CAOV3 and OVCAR4 cells ( Figure 1A-B), suggesting that NB-55 may promote a decrease in phosphorylated FOXM1 prior to FOXM1 degradation. We further explored the potential ability for NB compounds to modulate FOXM1 phosphorylation by examining their effects on p-FOXM1 and FOXM1 expression one hour after treatment (Supplementary Figure S2). In CAOV3 cells, the ratio of p-FOXM1 to FOXM1 expression remained ~1 regardless of treatment. In OVCAR4 cells, p-FOXM1 expression does not change while FOXM1 expression varies marginally, potentially as a result of slight loading differences. Thus, the NB compounds appear to primarily function by reducing overall FOXM1 protein levels.

NB compounds promote HGSC cellular apoptosis
To investigate the ability for the NB compounds to promote HGSC cell death, we examined the apoptotic activity of CAOV3, OVCAR4, and FT282 C11 cells treated with NB-73 and NB-115 for 24 hours. In CAOV3 cells ( Figure 4A), NB-73 demonstrated a dose-dependent increase in caspase 3/7 activity, a common measure of cellular apoptosis. NB-115 did not show this dose-dependent increase, possibly because maximum effect was already reached at 1 µM NB-115. In OVCAR4 cells ( Figure 4B), a dose-dependent increase in caspase 3/7 activity was observed for both NB-73 and NB-115. While HGSC cells treated with NB-73 and NB-115 demonstrated increased caspase 3/7 activity, FT282 C11 did not ( Figure 4C), further demonstrating the selectivity of NB-73 and NB-115 for HGSC cells. The pan-caspase inhibitor Q-VD-OPh decreased caspase 3/7 activity in CAOV3 cells ( Figure 4D) and rescued CAOV3 cells from cell death ( Figure 4E). Visually, CAOV3 cells showed remarkable cell death when treated with NB-73 and NB-115 ( Figure 4G,I) compared to DMSO control treatment ( Figure 4F). The visual cell death induced by NB-73 and NB-115 was prevented when cells were co-treated with Q-VD-OPh ( Figure 4H,J). Representative apoptosis pro les generated by ow cytometry for can be found in Supplemental Figure 3.

NB compounds inhibit HGSC colony formation
To examine the ability of the NB compounds to inhibit two-dimensional, anchorage-dependent HGSC colony formation (i.e., clonogenic growth), CAOV3, OVCAR4, and OVCAR5 cells were treated with the most potent NB compounds, NB-73 and NB-115, and allowed to form colonies. NB-73 decreased colony formation at concentrations between 50 and 200 nM in CAOV3, OVCAR4, and OVCAR5 cells, with 200 nM NB-73 yielding virtually no colonies ( Figure 5A). NB-115 decreased colony formation at concentrations between 25 and 200 nM in CAOV3 cells and between 12.5 and 200 nM in OVCAR4 cells, with 200 nM NB-115 yielding virtually no colonies in both cell lines ( Figure 5B). OVCAR5 cells needed higher concentrations of NB-115 to yield the same suppressive effect on colony formation as in CAOV3 and OVCAR4 cells. In OVCAR5 cells, concentrations between 50 and 200 nM NB-115 colony formation; however, >200 nM NB-115 would be necessary to fully suppress colony formation in OVCAR5.
For comparison, we also examined the ability of thiostrepton and FDI-6 to inhibit two-dimensional, anchorage-dependent colony formation in CAOV3 and OVCAR4 cells. Thiostrepton decreased colony formation at concentrations between 25 and 200 nM in CAOV3 cells, with ~10% colonies compared to control at 200 nM thiostrepton, and between 0.25 and 2 µM in OVCAR4 cells ( Figure 5C), with 2 µM yielding virtually no colonies. Unexpectedly, FDI-6 did not suppress colony formation in either CAOV3 or OVCAR4 cells, even at 100 µM concentrations ( Figure 5D).
To examine the ability of the NB compounds to inhibit three-dimensional, anchorage-independent HGSC colony formation, OVCAR5 and OVCAR8 cells in soft agar were treated with NB-73 or NB-115 and allowed to form colonies. CAOV3 and OVCAR4 cells were not able to form colonies in soft agar and were, therefore, not used in the analysis. NB-73 ( Figure 5E The synergy results are detailed in Table 1. The dose-effect curves can be found in Supplementary Figure  S5

NB-73 and NB-115 show superior selectivity and cytotoxicity in HGSC cells among a panel of tested FOXM1 inhibitors
To determine the potency and e cacy of the NB compounds compared to other reported FOXM1 inhibitors, we examined the effect of a panel of FOXM1 inhibitors on cell viability of CAOV3 and OVCAR4 cells. Potency was de ned by the IC 50 value, and e cacy was de ned by the loss of cell viability compared to control cells treated with solvent only. The IC 50 value for NB-55 ( Figure 6B) was in the low micromolar range compared to the mid-nanomolar IC 50 values for NB-73 ( Figure 6C) and NB-115 ( Figure   6D). FDI-6 ( Figure 6E) demonstrated a low micromolar IC 50 value and somewhat low e cacy, and thiostrepton ( Figure 6F) had a high nanomolar IC 50 value. Monensin had a low nanomolar IC 50 value but was not particularly e cacious in OVCAR4 cells and demonstrated a broad dose-response curve in CAOV3 cells ( Figure 6G). N-phenylphenanthren-9-amine ( Figure 6H) had a low micromolar IC 50 value, though at least 2-fold higher than that observed for FDI-6. RCM-1 ( Figure 6I) had no effect on HGSC cell viability, consistent with observations in breast cancer cell lines (data not shown). The IC 50 values and e cacy rates are summarized in Table 2, which is strati ed by the most potent and e cacious compounds (NB-73 and NB-115) to the least potent and e cacious compounds (N-phenylphenanthren-9amine and RCM-1).
To explore the selectivity of the NB compounds for FOXM1 and for HGSC cells versus non-malignant cells, we used the immortalized fallopian tube epithelium (FTE) cell line FT282 C11 to serve as a normal cell control highly relevant to HGSC 31,32 . Notably, in characterizing FT282 C11 cells for FOXM1, p-FOXM1, CCNB1, FOXA1, and FOXO3A expression, FT282 C11 cells had markedly lower FOXM1, p-FOXM1, CCNB1, and FOXA1 protein expression and markedly higher FOXO3A protein expression compared to CAOV3 and OVCAR4 cells ( Figure 6A). NB-55 ( Figure 6B) showed no selectivity for HGSC over FT282 C11 cells. In contrast, NB-73 ( Figure 6C) and NB-115 ( Figure 6D) showed an average of 10-fold and 7-fold selectivity, respectively, for HGSC cells. Due to their widespread use in the FOXM1 eld, we also investigated the effects of thiostrepton and FDI-6. FDI-6 ( Figure 6E) showed a moderate selectivity for HGSC cells at an average of 1.5-fold, and thiostrepton ( Figure 6F) showed no selectivity between HGSC and FT282 C11 cells. The IC 50 values and e cacy rates are summarized in Table 3, which is strati ed by most selective compound (NB-73) to least selective compound (NB-55 and thiostrepton) for HGSC cells.

Discussion
In this study, we demonstrated that the NB compounds promote several anti-cancer effects in HGSC cells that re ect favorably on their therapeutic potential for the treatment of HGSC cancers. The NB compounds promoted FOXM1 proteasomal degradation in HGSC cells, an effect that appears to be selective for FOXM1 as other FOX proteins were not similarly affected, and they potently and e caciously decreased HGSC cell viability, in large part by promoting apoptosis. The NB compounds have shown similar activities in TNBC cells 28 , and consistent with observations in TNBC cells 28 , the 1,1-diarylethylene diamine methiodide salts NB-73 and NB-115 were more potent FOXM1 inhibitors and suppressors of HGSC cell viability than the 1,1-diarylethylene monoamine methiodide NB-55. The NB compounds also suppressed two-dimensional, anchorage-dependent and three-dimensional, anchorage-independent HGSC colony formation at nanomolar concentrations. As the colony formation assay requires colonies to form from a single cell, these results suggest the NB compounds may exert anti-cancer effects on tumor cells growing at metastatic sites in addition to cancer cells in the primary tumor.
Ample evidence indicates that FOXM1 promotes chemotherapy resistance, including platinum-based chemotherapy resistance, in ovarian cancer 8 . Here, we found that the NB compounds synergized with the platinum-based chemotherapy agent carboplatin, a mainstay treatment for ovarian cancer and several other cancers, to inhibit HGSC cell viability, offering the potential for combination therapy in the clinic. In breast cancer cells that have become resistant to NB compound treatment, transcriptional processes had become rewired such that breast cancer cells became sensitive to lapatinib, an anti-EGFR therapy used in breast cancer 45 . Therefore, cancer treatment with the NB compounds shows promise in being used in combination as well as consecutively with existing chemotherapy.
Curiously, FOXA1 and FOXO3A expression appeared to show some changes in response to treatment with FOXM1 inhibitors. However, these expression changes were generally modest and transient, and the observations were not consistent between cell lines. The tumor suppressor FOXO3A is well-known to downregulate FOXM1 expression through several mechanisms 7 , but FOXM1 affecting FOXO3A expression has not been reported. Similarly, the relationship between FOXM1 and FOXA1 expressions has not been studied.
Thiostrepton required 48 hours in CAOV3 cells to downregulate CCNB1 expression and did not differentiate between FOXM1-high HGSC cells and FOXM1-low immortalized FTE cells in reducing cell viability. While thiostrepton was able to suppress colony formation in HGSC cells, the effective concentrations exhibited a broad range and were highly dependent on the HGSC cell line, despite FOXM1 overexpression. As a thiazole antibiotic, thiostrepton is known to have pleiotropic effects, such as disruption of mitochondrial protein synthesis 46 . Therefore, although thiostrepton did downregulate FOXM1 expression and, consequently, CCNB1 expression, our results suggest that the ability of thiostrepton to downregulate FOXM1 expression may not be the primary mechanism through which it exerts its biological effects.
In contrast to the NB compounds decreasing intracellular FOXM1 protein levels, we observed that FDI-6 did not substantially reduce FOXM1 protein expression in HGSC cells and only modestly inhibited the downstream CCNB1 expression. This observation is consistent with the reported inhibitory effects of FDI-6 on FOXM1 function without inducing FOXM1 protein degradation 18 . Consequently, FDI-6 may not be very effective at inhibiting cancer phenotypes. Indeed, FDI-6 demonstrated limited e cacy (~60%) when treating HGSC cells for 72 hours and did not affect HGSC colony formation over the course of two weeks. However, FDI-6 was recently used as a lead compound to create a proteolysis-targeting chimera (PROTAC), which promotes the degradation of FOXM1 protein 47 , an approach that may provide a better mechanism for long-term, therapeutic effect. Furthermore, FDI-6 was not particularly potent, affecting HGSC cell viability in the low micromolar range. A more potent molecule based on FDI-6 was recently generated 48 , which may offer more opportunity for further study of this class of FOXM1 inhibitor.
The other FOXM1 inhibitors we explored were either not e cacious (monensin and RCM-1) or considerably less potent (N-phenylphenanthren-9-amine) than the NB compounds. The disparity between the dose-response curve produced in CAOV3 and OVCAR4 cells treated with monensin suggests that pleiotropic effects are exerted to produce the biological response on cell viability. This observation is not surprising, as monensin is well-known to block Golgi transport and protein secretion 49,50 and is extensively used for this purpose in laboratory research. As an ionophore that selectively transports sodium cation across lipid membranes, monensin has also been shown to induce mitochondrial ROS production and disrupt calcium homeostasis 51 . Surprisingly, RCM-1 had no effects on HGSC (or breast cancer) cell viability. However, it was reported that RCM-1 did affect cell viability in several other cancer cell lines 52 . RCM-1, which has structural similarity to FDI-6, was initially shown to translocate nuclear FOXM1 into the cytoplasm of U2OS cells, resulting in the proteasomal degradation of FOXM1 27 . Our results suggest that further validation as to how RCM-1 affects cancer cell viability is warranted. Aside from RCM-1, which had no effect on HGSC cell viability, N-phenylphanenthren-9-amine was the least potent compound out of all the FOXM1 inhibitors we tested.
We were limited in our comparison of the NB compounds to the most widely used and accessible reported FOXM1 inhibitors. Although many natural compounds have been shown to inhibit FOXM1 53-55 , their anti-cancer effects may not be due to inhibition of FOXM1 given their likely pleiotropic nature. Other FOXM1 inhibitors, such as 7-di uoromethoxyl-5,4'-di-n-octyl-genistein (which was synthesized from the natural compound genistein) 56 , are not currently commercially available. Additionally, several new FOXM1 inhibitors have been recently reported 47,48,57,58 , demonstrating the high interest in therapeutically targeting FOXM1 for cancer treatment. Our study establishes the NB compounds as novel FOXM1 inhibitors that merit further investigation for their clinical translatability to treat aggressive cancers that overexpress FOXM1, such as HGSC.       Effect of NB compounds, thiostrepton, and FDI-6 on two-dimensional, anchorage-dependent and threedimensional, anchorage-independent HGSC colony formation.
For two-dimensional, anchorage-dependent colony formation, CAOV3, OVCAR4, and OVCAR5 cells were treated with NB-73, NB-115, thiostrepton, or FDI-6 in triplicate wells and allowed to form colonies for 7-14 days, depending on the cell line. For three-dimensional, anchorage-independent colony formation, OVCAR5 and OVCAR8 cells were treated with NB-73 or NB-115 and allowed to form colonies in soft agar for 14 days. Following incubation, cells were stained with crystal violet. Colonies were quanti ed using countPHICS (count and Plot Histograms of Colony Size) software 34 , where the threshold for colony size was automatically determined. Graphs were generated using GraphPad Prism 7 software. For twodimensional, anchorage-dependent colony formation, graphs are shown of CAOV3 (red), OVCAR4 (blue),  Three independent trials were performed with three technical replicates per trial. Values represent mean ± SEM. Curves were created, when applicable, using nonlinear regression with least squares (ordinary) t in GraphPad Prism 7 software.