6α-Acetoxyanopterine: A Novel Structure Class of Mitotic Inhibitor Disrupting Microtubule Dynamics in Prostate Cancer Cells Free

Prostate cancer is the second most diagnosed cancer in men in developed countries and the fifth most common cause for cancer-related deaths (1). Metastatic castrate-resistant prostate cancer (mCRPC) remains an incurable disease despite the recent approvals of new and more efficacious therapeutics (2). This highlights the need for additional prostate cancer therapies, for example, drugs that inhibit proven targets, yet remain unchallenged by known resistance pathways (2). The majority of anticancer drugs currently in use are natural products, natural product derivatives, or compounds developed on the basis of a natural product pharmacophore (3). Numerous organisms of various biota independently developed a great diversity of natural products throughout evolution as antipredatory defenses based on targeting microtubule dynamics (4). Microtubules are crucial for cellular processes, such as cell division, mobility, intracellular organelle transport, and endothelial cell biology (angiogenesis; ref. 5). Microtubules are highly dynamic protein structures, which continuously undergo growth through polymerization and shrinkage through depolymerization of α- and β-tubulin heterodimers (microtubule dynamics; ref. 6). In mitosis, microtubules play a vital role in the assembly of the spindle apparatus and proper segregation of chromosomes to both daughter cells (6). This critical function has been successfully exploited for the development of anticancer drugs and is exemplified by the semisynthetic taxanes docetaxel and cabazitaxel, which are the FDA-approved mainstay therapies of mCRPC (2). Vinca alkaloids (e.g., the natural products vinblastine, vincristine, and semisynthetic vinorelbine) are another class of microtubule inhibitors currently used as gold standards in chemotherapy for Hodgkin disease, non-Hodgkin lymphoma, and breast cancer (6). At clinically relevant doses (low nanomolar), both drug classes kinetically stabilize the microtubules, without changing microtubule polymer mass (7–9). This leads to mitotic arrest at the metaphase–anaphase transition, chromosome missegregation, genome instability, and ultimately mitotic catastrophe and cell death (6). At higher concentrations (micromolar), taxanes stabilize microtubules through binding to the taxane-binding site on β-tubulin (10). Vinca alkaloids bind to the vinca domain on β-tubulin and induce microtubule depolymerization (11). Another class of microtubule destabilizers includes the natural products colchicine, combretastatins, and 2-methoxyestradiol (2ME2), which all interact with the colchicine domain located at the intradimer interface between α/β-tubulin heterodimers, leading to microtubule depolymerization (12). Since the development of vinblastine and paclitaxel, other natural product–derived microtubule inhibitors have reached the market. Sagopilone (epothilone analogue, stabilizer) has shown very promising results in a phase II clinical trial in mCRPC (13). The FDA-approved microtubule-depolymerizing agent eribulin has shown activity in a phase II clinical trial (14). Despite their initial therapeutic success, resistance to these agents develops. Known resistance mechanisms induced by microtubule inhibitors include drug efflux through increased expression of multidrug resistance transporters, tubulin isoform expression, and mutations (6). We have previously reported the identification and cytotoxicity analysis of eight C₂₀ diterpenoid alkaloids that feature the unique anopterine scaffold from the Australian endemic rainforest plant Anopterus macleayanus (15). The anopterine analogues induced cell death in a panel of prostate cancer cell lines at nanomolar concentrations (15). Here, we present the mechanism-of-action studies of the most potent anopterine analogue, 6α-acetoxyanopterine (6-AA) and provide evidence that the anopterine scaffold represents a novel structure class of microtubule inhibitors with a mechanism of interaction that is distinct to taxanes, vinca alkaloids, and 2-methoxyestradiol. This is the first time that such activity has been described for this structure class.

Vinblastine, colchicine, nocodazole, verapamil, 2-methoxyestradiol (all Sigma-Aldrich), vincristine, doxorubicin, MNL8237, BI2536 (all Selleckchem), paclitaxel (Cytoskeleton Inc.), and 6-AA (15) were resuspended in DMSO (Sigma-Aldrich), and stock solutions were stored at −20°C.

LNCaP and PC-3 cells were obtained in 2010 from the ATCC and cultured in phenol red–free RPMI1640 medium supplemented with 5% FBS (Thermo Fisher Scientific). Cells were genotyped in December 2013 by short tandem repeat (STR) analysis at the DNA Diagnostic Center (Cincinnati, OH). HeLa cells were obtained in 1993 and authenticated by STR fingerprinting in 2015. HeLa-H2B-GFP (pBOS-H2BGFP, BD Biosciences) and HeLa-EB1-GFP [EB1-GFP (JB131), Addgene] were made by B. Gabrielli (16) and cultured in DMEM medium supplemented with 10% FBS. Nonadherent human T-cell acute lymphoblastic leukemia (T-ALL) cell line CCRF-CEM (CEM) and its drug-resistant sublines CEM/VCR-R (17, 18), CEM/2ME2-14.4R, and CEM/2ME2-28.8R (19) were a kind gift from the Kanamatsu laboratory in 1989 and validated by STR profiling in May 2016 by the Molecular Genetics Facility at the Garvan Institute of Medical Research (New South Wales, Australia). CCRF-CEM (CEM) and its drug-resistant sublines were cultured in RPMI1640 medium supplemented with 5% FBS. All cell lines were kept at 37°C in an atmosphere containing 5% CO₂, maintained in log phase growth, and were routinely screened for mycoplasma.

LNCaP (4,000 cells/well), HeLa (2,000 cells/well), and PC-3 (3,000 cells/well) cells were seeded in 96-well plates and allowed to attach for 24 hours. Cell viability as a function of metabolic activity was measured by alamarBlue endpoint assay (Thermo Fisher Scientific) after 72 hours of treatment (20). Nonadherent T-ALL cell lines (20,000 cells/well) were simultaneously seeded in 96-well plates and treated with the indicated compounds for 72 hours. Cell proliferation as a function of cell confluence was evaluated using live-cell imaging (IncuCyte, Essen BioScience) as described before (20).

LNCaP cells were plated in 6-well plates (150,000 cells/well), allowed to attach for 24 hours, and treated with the indicated compounds for 24 hours. Cell-cycle experiments were conducted as described previously (21). Samples were analyzed on a FACSCanto (BD Biosciences). DNA histograms were analyzed using ModFit LT software (Verity Software House).

PC-3 cells (30,000 cells/well) were seeded in 96-well ImageLock plates (Essen BioScience) and allowed to attach for 24 hours. Wounds were generated with a 96-well mechanical Woundmaker (Essen BioScience); cells were treated with the indicated compounds for 16 hours. Wound closure was measured using a live-cell imaging system (IncuCyte) according to the manufacturer's instructions (Essen BioScience).

LNCaP cells were seeded and allowed to attach for 24 hours in a 6-well plate (150,000 cells/well) then treated with 6-AA (10 nmol/L) or vinblastine (3.25 nmol/L) for 24 hours before RNA extraction and processing as described previously (22). For gene expression profiling, 3 to 4 repeats of each treatment were analyzed on a custom 180K Agilent oligo microarray (ID032034, GPL16604). Raw data were processed using the Agilent Feature Extraction Software (v10.7) and the LIMMA package in R as described before (22). Differential expression between treatment and vehicle control (DMSO) was classified on the basis of P_adj ≤ 0.05 and an average fold change of ≥1.5. Expression data are “Minimum Information About a Microarray Experiment” (MIAME) compliant and have been submitted to Gene Expression Omnibus (GEO) with the accession number GSE81277. Differentially expressed genes were examined using GeneGo MetaCore (Thomson Reuters) and GOrilla (Gene Ontology enRIchment anaLysis and visuaLizAtion tool; ref. 23) for functional annotation and network analysis.

RNA samples were generated as described above and processed as reported previously (21). qRT-PCR was performed with SYBR Green PCR Master Mix (Thermo Fisher Scientific) using a 7900HT Fast Real-Time PCR System (Applied Biosystems). Changes in mRNA expression levels were calculated on the basis of the ΔΔC_t method, normalized relative to RPL32 expression, and expressed as fold change relative to control (DMSO). Primer sequences are listed in Supplementary Table S1.

LNCaP cells were seeded, grown for 24 hours in a 6-well plate (150,000 cells/well), and treated with the indicated compounds for 24 hours before cell lysis in modified RIPA buffer (21). Cell lysates (30 μg/lane) were separated by SDS-PAGE, and proteins were blotted to PVDF membrane (Millipore). Primary antibodies used were PARP (46D11, 9532, Cell Signaling Technology, 1:1,000), phospho-Ser10 histone H3 (ab5176, Abcam, 1:1,000), and β-actin antibody (sc-47778, Santa Cruz Biotechnology, 1:2,500). Primary antibodies were probed with the appropriate horseradish peroxidase–conjugated secondary antibody (GE Healthcare) and visualized on a ChemiDoc XRS system (Bio-Rad) by chemiluminescence (ImmobiLion).

Optical 96-well plates (ibidi) were coated with poly-L-ornithine (Sigma-Aldrich) as described previously (24). LNCaP cells (5,000 cells/well) were seeded and allowed to attach for 24 hours. After the indicated times and concentrations of treatment, cells were fixed with ice-cold methanol for 3 minutes, incubated in blocking buffer [2% BSA (Sigma Aldrich) in TBS with 0.1% Triton X-100] for 30 minutes at room temperature before immunostaining for 1 hour at room temperature with primary antibodies against phospho-Ser10 histone H3 (ab5176, Abcam, 1:1,000) and α-tubulin (DM1A, ab7291, Abcam, 1:500), pericentrin (ab448, Abcam, 1:100), and Aurora A kinase (610398, BD Biosciences, 1:100). Cells were washed twice in TBS-T and incubated with the appropriate secondary, fluorophore-labeled antibody for 1 hour at room temperature in the dark. DNA was counterstained with 1 μg/mL DAPI (4′,6-diamidino-2-phenylindole, Sigma Aldrich). Methanol fixation and permeabilization with Triton X-100 were performed to wash out soluble tubulin subunits (25, 26).

Images were acquired on a Cytell or INCell 2200 automated imaging system (GE Healthcare) at ×10, ×20, or ×40 magnifications. Image segmentation and quantitation of approximately 4,500 cells per treatment were performed with CellProfiler software (Broad Institute, Cambridge, MA; ref. 27). For detailed segmentation method, see Supplementary Fig. S1. As the total tubulin level is not affected by known microtubule-targeting agents (28, 29), and most of soluble tubulin subunits were extracted during wash steps, mean α-tubulin staining intensity was correlated to α-tubulin polymer mass (29, 30). Analysis of spindle organization (∼800 cells/treatment) and the distance between spindle poles (∼120 cells/treatment) was performed using Fiji software (31).

LNCaP and HeLa cells were seeded as described above in poly-L-ornithine–coated plates. Cells were treated with the indicated concentration of compounds for 8, 24, or 72 hours. For washout experiments, media were carefully removed after 8 or 24 hours of treatment as indicated, and cells were left to recover in fresh media until completion of the experiment.

The CellEvent Caspase-3/7 reagent (Thermo Fisher Scientific) was used to quantitatively evaluate apoptosis. CEM (10,000 cells/well) and CEM/VCR-R (10,000 cells/well) cells were simultaneously seeded and treated in optical 96-well plates precoated with poly-L-ornithine. After 24 hours, cells were processed with CellEvent Caspase-3/7 reagent following the manufacturer's instructions. DNA was stained with Hoechst 33342 (Thermo Fisher Scientific). Images were acquired on an INCell 2200 imaging system (GE Healthcare) at ×20 and ×40 magnifications. Image analysis of 3,000 cells per treatment was performed using CellProfiler software. To quantify apoptotic cells, all nuclei (based on DAPI staining) were scored for caspase 3/7-positive and -negative staining.

HeLa-H2B-GFP cells (2,000 cells/well) were seeded and allowed to attach for 48 hours in optical 96-well plates, then treated with the indicated compounds. Images were immediately acquired with Cell R software on an Olympus IX81 live-cell microscope using a 10× objective at 37°C, 5% CO₂. Images were captured every 5 minutes for 24 hours. Quantification of the time in mitosis and cell fate of approximately 120 cells per treatment was performed using Fiji software (31).

HeLa-EB1-GFP cells (20,000 cells/well) were seeded and allowed to attach for 48 hours in a 35-mm 4-chambers glass bottom dish (Cellvis). After treatment with the specified compounds for 2 hours, cells were imaged on a Nikon Ti-E Motorized Inverted Microscope. Images (60×) were taken every second up to 2 minutes. EB1 comets were tracked and analyzed using Imaris software (v8.2).

Microtubule assembly was studied using the CytoDYNAMIX Screen Kit (BK006P; Cytoskeleton Inc.) according to the manufacturer's instructions, and polymerization was monitored using a FLUOstar Omega plate reader (BMG LABTECH) at 340 nm every 1 minute for 30 minutes at 37°C.

All data points were performed in technical duplicates or triplicates, and experiments were repeated independently at least twice and reported as the mean ± SD of the biological replicates. For all experiments, one-way ANOVA with Dunnet multiple comparisons test was used (ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001) unless stated otherwise. IC₅₀ and statistical significance were analyzed using GraphPad Prism 6 software (GraphPad Software).

In our previous study, we showed that 6-AA (Fig. 1A) inhibited cell growth of five prostate cell lines (IC₅₀ = 3.1–11.5 nmol/L; ref. 15). We have further tested 6-AA in HeLa (cervical cancer) and PC-3 cells (metastatic prostate cancer), where 6-AA was also potently cytotoxic (IC₅₀: LNCaP = 3.1 nmol/L, HeLa = 3.2 nmol/L, PC-3 = 5.1 nmol/L; Fig. 1B; Supplementary Table S2) with a potency comparable with the microtubule inhibitor vinblastine (IC₅₀: LNCaP = 3.2 nmol/L, HeLa = 11.6 nmol/L, PC-3 = 2.5 nmol/L; Supplementary Table S2). 6-AA inhibited cell proliferation in a concentration-dependent manner in LNCaP (Fig. 1C) and HeLa (Supplementary Fig. S2A) cells and significantly inhibited cell migration of PC-3 cells at 1.25 nmol/L (Supplementary Fig. S2B).

Cell-cycle analysis by flow cytometry showed that 6-AA led to a significant increase in the population of LNCaP cells in the G₂–M phase in a concentration-dependent (Fig. 1D) and time-dependent manner (Supplementary Fig. S2D). At 5 nmol/L, the G₂–M arrest of LNCaP cells reached significance (P < 0.01) as early as 14 hours after treatment with 6-AA and preceded the induction of significant levels of cell death (24 hours), as indicated by the increased number of cells in the sub G₀–G₁ phase (Fig. 1E and Supplementary Fig. S2D). Caspase-dependent cleavage of PARP, a marker of apoptosis (32), was detected in 6-AA–treated LNCaP cells after 24 hours, similar to the microtubule inhibitors paclitaxel and nocodazole (Fig. 1F; refs. 33, 34). Consistent with this, 6-AA and vinblastine-treated LNCaP (Fig. 1C) and HeLa cells (Supplementary Fig. S2C) showed membrane blebbing, chromatin condensation (visible in HeLa-H2B-GFP, Supplementary Fig. S2C), and cell disintegration, which are typical signs of apoptosis (35, 36).

For target discovery, we performed microarray analysis of LNCaP cells treated for 24 hours with 6-AA (10 nmol/L) or vinblastine (3.25 nmol/L; Supplementary Fig. S3A). Pathway analysis (MetaCore) of the 6-AA and vinblastine datasets identified a strong overlap of networks enriched in differentially expressed genes that were shared between both treatments (Fig. 2A), with mitosis as deregulated core network and the top three process subnetworks being “cell cycle–mitosis,” “cytoskeleton–spindle microtubules,” and “cell cycle–G₂–M.” Several critical regulatory genes of mitotic entry, spindle formation, and mitotic exit were upregulated (Supplementary Table S3), such as mitotic kinases (AURKA, AURKB, and PLK1), mitotic checkpoint components (BUB1, CDC20, and CENPF), mitotic kinesins (KIF18B and KIF20A), and other important mitotic regulators (CCNB1, CDK1, NEK2, CDC25B, BIRC5, INCENP, and NUF2). Furthermore, GOrilla analysis revealed that the genes deregulated by 6-AA and vinblastine corresponded to the “microtubule binding” function (Supplementary Fig. S3C). 6-AA–mediated differential expression of CCNB1, CDC25B, PLK1, HMMR, PTTG1, and CDKN3 was confirmed by qRT-PCR (Fig. 2B).

The observed accumulation of cells in G₂–M and deregulation of mitotic pathways prompted us to investigate whether 6-AA is an inhibitor of mitosis. Similar to the mitotic inhibitors paclitaxel and nocodazole, Western blotting revealed that 6-AA strongly increased the level of phospho-Ser10 histone 3 protein (PHH3) in LNCaP cells, a marker of mitosis (Fig. 3A; ref. 37).

Fluorescence microscopy of PHH3 demonstrated that, like vinblastine, 6-AA caused a concentration-dependent accumulation of PHH3-positive LNCaP cells after 24 hours (Fig. 3B), which was detectable as early as 8 hours (1.25–10 nmol/L, Supplementary Fig. S4A). Consistent with this, these cells displayed the morphologic hallmarks of mitosis (cell rounding and condensation of chromatin, Supplementary Fig. S1).

Notably, like vinblastine (38), the 6-AA–mediated mitotic block and growth inhibition (up to 2.5 nmol/L) was reversible when cells were treated for 8 hours, then washed with fresh media, and left to recover for 16 hours (Supplementary Fig. S4). However, treatment with 6-AA for longer than 8 hours before inhibitor removal was associated with a visibly reduced rate of proliferation and viability, which was also observed for vinblastine (Supplementary Fig. S4B, S4C, and S4F).

A detailed analysis of 6-AA–mediated inhibition of mitosis and its impact on cell fate was performed using time-lapse microscopy of HeLa cells expressing H2B-GFP (Fig. 3C). HeLa cells normally progressed through mitosis with a mean time of 67.4 ± 32.4 minutes, followed by cytokinesis (87.1% bipolar division leading to two daughter cells). Like vinblastine (10 nmol/L: 398.8 ± 270.3 minutes), 6-AA caused a significant, concentration-dependant increase in the duration of mitosis (Fig. 3C), prolonging the time spent in mitosis from 216.3 ± 192.0 minutes (0.62 nmol/L) to 687.8 ± 207.7 minutes (10 nmol/L). Close inspection of the images further revealed that cells treated with increasing concentrations of 6-AA failed to properly congress and align their chromosomes at the metaphase plate and displayed misaligned chromosomes and patterns of chromosome alignment that are evidence of monopolar and multipolar spindles. These phenotypes were also seen in vinblastine-treated cells (Supplementary Videos S1 and S2). Evaluation of the cellular fate (number of daughter cells and cell death) revealed a concentration-dependent decrease in bipolar divisions and concomitant increase in asymmetric cell divisions, leading to more than two daughter cells (mostly 3–4) when cells were treated with 0.62 to 5 nmol/L of 6-AA as well as a substantial rise in cell death (Fig. 3C), culminating at 10 nmol/L 6-AA with 100% cell death. Similar effects were observed with vinblastine (Fig. 3C). Anaphases with completed karyokinesis and cytokinesis (into three or more daughter cells) were considered as asymmetric divisions (39). Furthermore, daughter cells derived from these asymmetric divisions commonly merged together into a single multinuclear cell (Fig. 3D; Supplementary Videos S1 and S2; ref. 16). Immunofluorescence microscopy of LNCaP cells treated with 6-AA (2.5 nmol/L) or vinblastine (10 nmol/L) also showed multinucleated cells (data not shown).

We next tracked HeLa-H2B-GFP cells treated with 6-AA (2.5 nmol/L) for 144 hours by time-lapse microscopy to study the fate of the cells derived from asymmetric cell division. We observed that most of the daughter cells failed to divide further and died in an interphase-like state (Fig. 3D, left) or during the next cell cycle (Fig. 3D, right). These results are consistent with a study by Ganem and colleagues (40) and could partially explain the long-term cell-killing effect of transient treatment with 6-AA.

Given the almost identical transcriptional and phenotypic effects of 6-AA and vinblastine, we investigated whether 6-AA interfered with the mitotic spindle organization, in particular microtubule dynamics, the major target of vinblastine (8). Immunofluorescence microscopy of PHH3 and α-tubulin showed that LNCaP cells treated for 24 hours with 2.5 nmol/L of 6-AA exhibited a range of microtubule spindle abnormalities, including bipolar spindles with misaligned chromosomes, monopolar and multipolar spindles, whereas vehicle control (DMSO) showed cells with normal metaphase plates and bipolar anaphases (Fig. 4A). Phenotypic scoring of the 6-AA–induced spindle abnormalities revealed that, at 2.5 nmol/L of 6-AA, 23% of cells displayed bipolar spindles with misaligned chromosomes, and 35.8% and 41.3% of cells had monopolar and multipolar spindles, respectively. The latter defect visibly increased in a dose-dependent manner (Fig. 4A), which was consistent with our findings in HeLa cells (as shown above). A similar frequency of these spindle defects was observed in vinblastine-treated LNCaP cells (Fig. 4A). Measurement of the distance between spindle poles in cells with bipolar spindles revealed that 6-AA produced a dose-dependent reduction of the spindle length, similar to vinblastine (Fig. 4A). These defects were also observed in PC-3 cells (Supplementary Fig. S5). Seven structural analogues of 6-AA induced a similar phenotype (Supplementary Fig. S6A and S6B).

During normal mitosis, each of the two spindle poles contains one centrosome, formed by a pair of centrioles fixed in pericentriolar material, containing pericentrin (41). Immunofluorescence microscopy of vehicle-treated LNCaP cells in mitosis showed two similar-sized areas of pericentrin staining at both spindle poles, whereas 6-AA (5 nmol/L) or vinblastine (20 nmol/L) caused a smaller, more punctuated pericentrin staining and frequently two discrete spots per pole (Fig. 4B), and in some cases, pericentrin staining was absent from the spindle pole (data not shown). These results demonstrate that 6-AA disrupts normal spindle organization in mitosis.

Inhibition of mitotic kinases Aurora kinase A (Aurora A) and Polo-like kinase-1 (Plk1) has been shown to cause defective spindle organization (42, 43). Immunofluorescence microscopy revealed that 6-AA–treated cells displayed strong Aurora A staining at the spindle poles, which was absent at this localization with the Aurora A inhibitor MNL8237 (Supplementary Fig. S7A). The Plk1 inhibitor BI2536 exclusively induced the formation of monopolar spindles (42), often with dispersed chromosomes at the cell periphery (Supplementary Fig. S7B). Together, these distinct phenotypes strongly suggest that 6-AA does not target the mitotic kinases Aurora A or Plk1.

To determine whether 6-AA inhibits mitotic spindle organization by directly targeting microtubule formation, we performed a microtubule assembly assay with purified components in a cell-free system (44). 6-AA inhibited in vitro tubulin polymerization in a concentration-dependent manner, similar to the microtubule-depolymerizing compounds vinblastine and nocodazole, whereas the microtubule-stabilizing molecule paclitaxel increased polymerization (Fig. 5A). Anopterine, a 6-AA analogue, also inhibited tubulin polymerization (Supplementary Fig. S6C).

In LNCaP cells, high doses of 6-AA (5–80 × IC₅₀) significantly reduced the cellular tubulin polymer mass (45) in a dose-dependent manner and visibly disrupted cellular microtubule networks in LNCaP cells, similar to the effect of vinblastine, whereas paclitaxel caused an increase in tubulin polymer mass (Fig. 5B), which was consistent with previous studies (46).

Inhibitor washout experiments revealed that disruption of the cellular microtubule network by 6-AA and vinblastine was reversible in LNCaP cells (Supplementary Fig. S8; ref. 47), whereas with colchicine, the destabilizing effect was irreversible (48).

We studied the effect of 6-AA on microtubule dynamics in HeLa cells expressing the end-binding protein (EB1-GFP) using real-time spinning disk microscopy. EB1 remains attached to the plus end of the growing microtubules but dissociates during the depolymerization phase (49) and can be used to track polymerizing microtubules and measure changes to microtubule dynamics (50). 6-AA visibly disrupted directional microtubule growth, as exemplified by maximum intensity projections of the EB1 comet time-lapse sequences (Fig. 5C; Supplementary Videos S3–S5). Analysis of the time-lapse videos showed that 6-AA significantly shortened the displacement length of the EB1 tracks in a concentration-dependent manner when compared with the vehicle control (Supplementary Fig. S9).

Taken together, 6-AA directly and reversibly inhibits microtubule dynamics (low concentration) and induces tubulin depolymerization (high concentration), leading to a decrease of cellular tubulin polymer mass and disrupted microtubule networks.

To address whether the activity of 6-AA is affected by resistance mechanisms that human T-ALL cell lines developed in response to chronic treatment with the microtubule inhibitors 2-methoxyestradiol (CEM/2ME2-14.4R and CEM/2ME2-28.8R cells) or vincristine (CEM/VCR-R cells), we measured the resistance factor of 6-AA in these cell lines. The 2ME2-resistant cell lines have acquired βI-tubulin mutations in the colchicine-binding pocket, whereas the vincristine-resistant CEM/VCR-R cells have acquired multiple microtubule alterations (e.g., mutations in βI-tubulin, altered expression of β-tubulin isotypes; ref. 18) and overexpression of P-glycoprotein (P-gp), thereby rendering cells multidrug resistant to non-microtubule–targeting inhibitors, such as doxorubicin (51), daunorubicin, actinomycin D (17), and tetracycline (52).

Consistent with previous work, CEM/2ME2-14.4R and CEM/2ME2-28.8R cells displayed resistance factors (RF) for 2ME2 of 18.7 and 30.6, respectively when compared with the parental CEM line. In contrast, the potency of 6-AA remained largely unaffected (RF: CEM/2ME2-14.4R = 2.3, CEM/2ME2-28.8R = 1.7; Fig. 6A; Supplementary Table S4).

Previous studies have shown that CEM/VCR-R cells are highly resistant to vincristine and vinblastine (17). Consistent with this, the RF for vincristine was 12,336 in CEM/VCR-R cells. Yet, 6-AA was 52 times more potent than vincristine, and CEM/VCR-R displayed an RF of only 45.3 (Fig. 6A, Supplementary Table S4). Similarly, and as previously shown (51), the RF for doxorubicin, a non-microtubule–targeting drug, was 24.9 in these cells, suggesting that the multidrug transporter P-gp confers resistance to doxorubicin (Fig. 6A; Supplementary Table S4). Analysis of cells positive for caspase-3/7 activity by quantitative fluorescence microscopy demonstrated that 6-AA induced apoptosis in both CEM and CEM/VCR-R cells in a concentration-dependent manner (Fig. 6B). A 50- to 100-fold higher dose of 6-AA was required to induce a similar level of apoptosis (∼30%) in CEM/VCR-R cells, whereas a 200-fold higher dose of vincristine was required.

To determine to what extent P-gp activity was responsible for the observed resistance of CEM/VCR-R cells to vincristine, 6-AA, and doxorubicin, the multidrug efflux pump was inhibited with verapamil (53). Cotreatment with verapamil partially restored sensitivity of CEM/VCR-R cells to vincristine (RF = 63.6, Fig. 6C; Supplementary Table S5), suggesting that previously described mechanisms (e.g., microtubule alterations; ref. 18) caused the remaining resistance. In comparison, verapamil almost completely resensitized CEM/VCR-R cells to 6-AA and doxorubicin as demonstrated by RFs of 2.1 and 1.1, respectively. This strongly suggested that P-gp activity was solely responsible for resistance to 6-AA and doxorubicin and, more importantly, that 6-AA potency was not affected by the reported microtubule alterations (18).

In summary, the absence of resistance that is mediated by microtubule alterations provides evidence that 6-AA interacts with tubulin with a mechanism that is distinct to vinca alkaloids and 2ME2.

In this work, we investigated the mechanism of action of 6-AA and discovered that this compound is a novel potent mitotic inhibitor in prostate and cervical cancer cells. The natural product 6-AA and seven analogues were previously identified by our group from the Australian endemic tree Anopterus macleayanus and are C₂₀ diterpenoid alkaloids featuring the unique anopterine scaffold (15). This is the first report that the anopterine diterpenoid scaffold is associated with inhibition of microtubule dynamics, representing a novel structure class of microtubule-targeting agents.

Depending on their effect on microtubules at high, micromolar concentrations, microtubule-targeting agents are categorized into microtubule-stabilizing compounds (e.g., taxanes, epothilones) and microtubule-destabilizing compounds (e.g., vinca alkaloids, nocodazole, colchicine; ref. 4). Our data show that 6-AA is a reversible microtubule-destabilizing molecule that directly interacts with tubulin (Fig. 5). Using washout experiments to investigate the reversibility of inhibition and relationship of treatment period with cell fate, we showed that LNCaP cells treated with 6-AA for a short period of time (8 hours) could exit the mitotic block and resume proliferation. Longer treatment (24 hours) with 6-AA or vinblastine showed that LNCaP and HeLa cells could partially recover and start to proliferate again. Gajate and colleagues (48) reported that cells treated with the bicyclic colchicine analogue MCT for a short period of time (before induction of the apoptotic response) resumed proliferating; however, once the treatment period exceeded the tolerable threshold (cell line specific) and apoptosis was triggered, the cell proliferation capacity could not be restored by the removal of the inhibitor. Furthermore, Towle and colleagues showed that after 12 hours of treatment of U-937 cells (histiocytic lymphoma) with microtubule-targeting agents at concentrations that induced complete mitotic block, and 10 hours of washout, the mitotic arrest was reversible or irreversible depending on the compound (38). Small changes in chemical structures between analogues (e.g., eribulin vs. ER-076349, vincristine vs. vinblastine, colchicine vs. colcemid) led to profound differences in mitotic block reversibility.

Most microtubule-targeting agents of this class bind to the vinca or colchicine sites on microtubules (4). The vinca site is located on the β-subunit of tubulin dimers (11). Studies have shown that vinblastine can also bind at the interdimer interfaces to both α- and β-tubulin (54, 55). The relatively large colchicine site is positioned between the α- and β-tubulin subunits within the same dimer (12). Point mutations located within these bindings pockets or sites that alter the tubulin conformation have been frequently observed and confer resistance through changed interaction with microtubule inhibitors of the same pharmacophore class (cross-resistance; refs. 18, 19). The minimal resistance of 6-AA in the 2ME2-resistant leukemia cell lines (CEM/2ME2-14.4R and CEM/2ME2-28.8R, Fig. 6) suggests a tubulin-binding mechanism of 6-AA that is different to 2ME2. Notably, CEM/2ME2-28.8R cells express class I β-tubulin with several mutations (S25N, D197N, A248T, and K350N), with A248T and K350N residing in the colchicine pocket, whereas CEM/2ME2-14.4R cells contain only the S25N and D197N mutations (19). On the basis that both cell lines lack significant cross-resistance to colchicine, we cannot dismiss the possibility that 6-AA might bind reversibly to the colchicine site.

A large body of different experiments presented here show that at low nanomolar concentration, 6-AA acts phenotypically and functionally indistinguishable from the vinca alkaloid vinblastine (7, 8). The antiproliferative effect of both compounds was due to inhibition of microtubule dynamics leading to mitotic catastrophe.

Vincristine/vinblastine-resistant leukemia CEM/VCR-R cells possessed noticeable resistance to 6-AA (Fig. 6). In contrast to parental CEM cells or the 2ME2-resistant sublines, CEM/VCR-R cells display the classic multidrug resistance phenotype mediated through high expression of P-gp that render cells resistant to inhibitors that do not target microtubules but are known substrates of P-gp (e.g., anthracyclines; ref. 17). Inhibition of P-gp with verapamil almost completely restored 6-AA sensitivity of CEM/VCR-R cells to levels seen in parental CEM cells (Fig. 6), demonstrating that 6-AA was not affected by the remaining resistance mechanisms like the reported microtubule alterations that render CEM/VCR-R cells resistant to vinca alkaloids. In light of our discovery that 6-AA is a P-gp substrate, future studies with our panel of anopterine derivatives will address whether modifications of the chemical structure could reduce its affinity for P-gp.

Altogether, our results strongly suggest that 6-AA interacts with tubulin in a way that is distinct to both vinca alkaloids and 2ME2. However, we cannot rule out that 6-AA is binding in the colchicine or vinca alkaloid–binding pockets, with a distinct orientation in the site and interacting with different amino acids. Further experiments are warranted to investigate the exact binding of 6-AA with tubulin by molecular docking, detailed analysis of the crystal structure of the complex between tubulin and 6-AA, and site-directed mutagenesis of key amino acids combined with binding studies.

No potential conflicts of interest were disclosed.

Conception and design: C. Levrier, M.C. Sadowski, M. Kavallaris, R.A. Davis, C.C. Nelson

Development of methodology: C. Levrier, M.C. Sadowski

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Levrier, M.C. Sadowski, A. Rockstroh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Levrier, M.C. Sadowski, A. Rockstroh, B. Gabrielli, M. Lehman

Writing, review, and/or revision of the manuscript: C. Levrier, M.C. Sadowski, B. Gabrielli, M. Kavallaris, R.A. Davis, C.C. Nelson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.C. Sadowski

Study supervision: M.C. Sadowski, R.A. Davis, C.C. Nelson

Other [Supply of compound (6-AA)]: R.A. Davis

The authors thank the Australian Research Council for access to MetaCore. C. Levrier would like to thank Griffith University for a Ph.D. scholarship (GUIPRS) and CTx for a PhD Top Up scholarship.

The authors acknowledge the National Health and Medical Research Council (NHMRC) for financial support (grant APP1024314; to R.A. Davis). This work was supported by funding from the Australian Government Department of Health and The Movember Foundation and the Prostate Cancer Foundation of Australia through a Movember Revolutionary Team Award (to M.C. Sadowski, A. Rockstroh, M. Lehman, and C.C. Nelson). B. Gabrielli was supported by an NHMRC Senior Research Fellowship. M. Kavallaris is funded by the Australian Research Council Centre of Excellence in Convergent Bio-Nano Science and Technology (CE140100036) and NHMRC Program grant (APP1091261). The Translational Research Institute is supported by a grant from the Australian Government. This work was also supported by the Australian Research Council grant LE150100161.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.