Drug Targeting the Actin Cytoskeleton Potentiates the Cytotoxicity of Low Dose Vincristine by Abrogating Actin-Mediated Repair of Spindle Defects

Combination therapy has shown promise for cancer treatment by reducing individual drug dosage while maintaining or increasing overall treatment efficacy (1–9). Although there has been significant effort to identify synergistic drug combinations (10–13), few successful combinations have been found and they are often effective in specific tumor types (10). Because antimicrotubule (anti-MT) chemotherapeutics are used as first-line and/or relapse treatment in up to half of all patients with cancer, and have toxic and often long-lasting side effects at therapeutic doses (14), identifying compounds that synergize with anti-MTs especially at low concentrations (15) represents a significant advance for current therapies.

To improve the efficacy of anti-MT treatment, we established a synergistic combination regime designed on the interplay between the microtubule network and the actin cytoskeleton in promoting cell growth (16, 17). This regime combines the anti-MT drugs vincristine (VCR) or paclitaxel with compounds targeting the actin-associated protein tropomyosin 3.1 (Tpm3.1). Tropomyosins are dimers that form head-to-tail polymers along both sides of actin filaments (18) and determine the functional characteristics of individual actin filaments in an isoform-dependent manner (19, 20). The functional impacts of the tropomyosins extend to the control of both actin-binding protein interactions and myosin motor activity (19–21). Tpm3.1 is highly upregulated and is the major isoform in cancer cells (22, 23). Anti-Tpm3.1 prototype compounds, TR100 and ATM3507, have shown anticancer activity in vitro and in vivo without adverse impact on cardiac structure and function, which is the major side-effect of antiactin drugs (16). This drug combination shows strong synergy in neuroblastoma both in vitro and in vivo (17).

Our previous study in neuroblastoma showed that anti-Tpm3.1 compounds enhance the G₂–M-phase arrest and mitotic spindle organization defects caused by vincristine at low concentrations (17). Although the function of Tpm3.1 in mitosis is unclear, the exacerbation of vincristine-induced spindle defects suggests a potential cooperation between the microtubule network and Tpm3.1-containing actin filaments in regulating mitosis. This hypothesis is supported by (i) the well-established interaction between the microtubule network and actin cytoskeleton during cell division (24–30), (ii) detection of actin at the spindle, particularly its dynamic association with spindle poles and the cortex during mitotic progression (26), and (iii) evidence that a specific tropomyosins isoform localizes to the cortex and another localizes to the mitotic spindle to promote chromosome segregation in Drosophila (31).

On the basis of the mechanical roles and the cross-linking of these two major cytoskeletal structures in a cell, we proposed that the synergy of our drug combination may be universal across different cancer types, and the mechanism behind the synergy may reflect a fundamental biological mechanism. We therefore evaluated the synergy in a wide range of cancers and initiated a mechanistic study in HeLa cells to understand how anti-Tpm3.1 compounds potentiate the antimitotic effect and the treatment efficiency of vincristine at low concentrations.

TR100 and ATM3507 (ATM; refs. 32, 33) were synthesized by SYNthesis. ATM was provided by Novogen. TR100, ATM, vinorelbine (VNB, Sigma-Aldrich), vincristine (VCR, Sigma-Aldrich), and MLN8237 (MLN, Selleck Chemicals) were dissolved in DMSO (Sigma-Aldrich) and stored at −20°C. Thymidine (Sigma-Aldrich) was dissolved in cell culture media and stored at 4°C.

Lung carcinoma (A549, Calu-6, H-460, H-1975, H-1650, and H-1299), colorectal adenocarcinoma (HT-29), ovarian adenocarcinoma (OVCAR-3), pancreatic carcinoma (MIA PaCa-2), breast adenocarcinoma (MDA-MB-231 and MCF7), gastric carcinoma (KATO-III and OACP4C), esophageal carcinoma (KYSE-410 and KYAE-1), pharyngeal carcinoma (Detroit 562), and tongue squamous carcinoma (SCC9) cell lines were from the ACRF Drug Discovery Centre for Childhood Cancer (ACRF DDCCC, Sydney, Australia). Prostate cancer cell lines [DU145 (HTB-81) and LNCaP (CRL-1740)] and HeLa human epithelial carcinoma (CCL-2) were obtained from the ATCC. HeLa cells were tested for Mycoplasma contamination using a PCR Mycoplasma Test Kit (Applichem, APPA3744,0020) in 2017. Other cell lines were tested using the MycoAlert Mycoplasma Detection Kit (Lonza, LT07-318) in 2014–2016 when the experiments were conducted. The results were negative. HeLa cells stably expressing α-tubulin-GFP and H2B-mCherry (RRID:CVCL_L802) were a gift from Anthony A. Hyman's laboratory, Max-Planck Institute for Cell Biology and Genetics, Dresden, Germany (MPI-CBG). No Mycoplasma testing was performed for this cell line in our laboratory.

Cells were grown at 37°C, 5% CO₂, and 95% humidity using the media indicated: RPMI with 10% (v/v) FBS (H-460, H-1975, H-1650, H-1299, DU145, LNCaP, OVCAR-3, OACP4C, and KYSE-410); RPMI with 20% (v/v) FBS (KATO-III); RPMI/F12 (1:1) with 10% (v/v) FBS (KYAE-1); DMEM with 10% (v/v) FBS (HeLa, A549, Calu-6, HT-29, MDA-MB-231, and MCF7); DMEM with 10% (v/v) FBS and 2 mmol/L l-Glutamine (MIA PaCa-2); DMEM/F12 (1:1) with 10% (v/v) FBS and 0.5 mmol/L NaP (SCC9); DMEM with 10% (v/v) FBS, 2 mmol/L l-Glutamine, 1% MEM nonessential amino acids, and 1 mmol/L NaP (Detroit 562); and DMEM with 10% (v/v) FBS, geneticin (400 μg/mL, Thermo Fisher Scientific), puromycin (0.5 μg/mL, Sigma-Aldrich; HeLa cells expressing α-tubulin-GFP and H2B-mCherry). Two HeLa cell lines were cultured for no more than 30 passages and 10 passages for other cell lines.

Primary antibodies: mouse anti-Tpm3.1 (γ-9d, 2G10.2, described in ref. 34); rabbit anti-histone H3 (phosphor S10, Abcam, ab5176); rabbit anti-PARP (Abcam, ab191217); mouse anti-GAPDH (Merck, MAB374); mouse anti-α-tubulin (DM1A, Sigma-Aldrich, T9026); rabbit anti-EB1 (Sigma-Aldrich, E3406); rabbit anti-pericentrin (Abcam, ab4448); mouse anti-dynactin/p150^Glued (BD Biosciences, 610474); mouse anti-dynein (Merck, MAB1618); and rabbit anti-NuMA (Novus Biologicals, NB500174). Secondary antibodies for Western blotting: rabbit anti-mouse (Cell Signaling Technology, 58802s) and donkey anti-rabbit (Abcam, ab16284). Secondary antibodies for immunofluorescence (Invitrogen): goat anti-mouse Alexa 488 (A-11001); donkey anti-mouse Alexa 647 (A-37571); goat anti-rabbit Alexa 488 (A-11034); and donkey anti-rabbit Alexa 647 (A-31573). Secondary antibody for cell-cycle analysis: goat anti-rabbit IgG FITC (Abcam, AB6717). Phalloidin Atto 488 (Atto Tec, AD 488-81) was used for actin staining.

Animal studies were performed by Jubilant Life Sciences, approved by the Institutional Animal Care and Use Committee (IAEC-JDC-2016-80), and carried out in accordance with the Guide for the Care and Use of Laboratory Animals, 8th Edition, 2010 (National Research Council of the National Academies). Calu-6 cells growing in exponential phase were inoculated subcutaneously on the dorsal right flank in athymic female nude mice. When tumors reached approximately 119 mm³ size, 64 mice were randomized into eight groups, 8 mice per group. Formulations: ATM (30% dexolve-7 in water, sterile filtered): vinorelbine (sterile PBS). Each group received intravenously either vinorelbine (5 or 10 mg/kg), ATM (25 or 50 mg/kg), or combinations at pH 8.0 on days 1, 5, 9, 13, and 17. Vehicle control groups received either PBS (for vinorelbine) or 30% dexolve-7 (for ATM). Mice were monitored on days 22, 26, 29, 33, 36, 40, 43, 47, 50, and 54. Time to reach maximal tumor size (2,500 mm³) was plotted as percent survival out to 68 days and median survival (days) determined via Mantel–Cox survival analysis.

Cells were seeded at 750 cells/well in 384-well plates (DU145 and LNCaP), or the following cells/well in 96-well plates: 500 (KYSE-410 and H-1299), 600 (H-460), 700 (KATO-III and Detroit 562), 800 (A549), 900 (MIA PaCa-2), 1,000 (KYAE-1), 1,200 (Calu-6 and H-1650), 1,300 (HT-29), 2,000 (H-1975, OVCAR-3, MDA-MB-231, OACP4C, and SCC9), and 6,000 (MCF7). After 24 hours, drugs were added (Tecan HP D300 Digital Dispenser) in technical triplicate under minimum light, with final DMSO concentration 0.4%. The drug combination was tested using a 6 × 6 concentration matrix with 2-fold dilution steps based on the IC₅₀ values of each drug. After 72 hours, cell viability was measured by adding 10% (v/v) Alamar Blue Reagent (Bio-Rad). Fluorescence intensity (excitation 555 nm, emission 585 nm) of each well was determined (SpectraMax M5 Plate Reader, Molecular Devices) at 0 hour (background) and 6 hours. Viability of control cells was set to 100% and fractional viability reduction of drug-treated cells was calculated for the Bliss independence drug interaction analysis. This work was done by the ACRF DDCCC.

Synergistic drug interactions were determined by applying the Bliss independence model that assumes independence of drug mechanisms (35). The predicted additive Bliss value for each combination in a 6 × 6 concentration matrix was calculated on the basis of the Bliss independence model using the fractional viability reduction data as described previously (35–37). The maximum synergy value of 1 was set to 100 for normalization. The difference between the experimentally observed value and the predicted additive value was then normalized to 100 as reported Bliss scores. Scores that were more than 0 demonstrated synergy, less than 0 indicated antagonism, and equal to 0 demonstrated an additive effect. The highest individual Bliss score for the combination matrix in each cell line is presented in Fig. 1A and B. This work was done by the ACRF DDCCC.

Cells were seeded in 96-well plates (1,000 cells/well), grown for 24 hours, and treated with drugs at designated concentrations for 72 hours. Final DMSO concentration was 0.5%. Cell viability was determined using MTS Reagent (Promega) as described previously (38). The optical density value of each well was measured using the VersaMax Plate Reader.

Cells were seeded in 96-well plates (1,000 cells/well), grown for 24 hours, and treated with drugs for 72 hours. Nine to 16 concentrations of each drug were selected from 0 μmol/L to IC₈₀ to create an 81- to 208-point combination matrix. Cell viability was measured using MTS reagent. Isobologram analysis was performed as described previously (38). For the fixed-ratio isobologram, cells were treated with two drugs (6–8 concentrations each) at fixed ratios for 72 hours, and cell viability measured. The combination index (CI) was calculated using CalcuSyn Software (Biosoft; ref. 38).

Cells were seeded in 6-well plates (10⁵ cells/well), grown for 24 hours, and then treated with drugs for 24 hours. Apoptosis induction and drug impact on cell-cycle distribution were determined as described previously (38).

Cells were seeded in 6-well plates (10⁵ cells/well), grown for 24 hours, and then treated with drugs for 24 hours. Lysates for Western blotting were prepared as described previously (38). Samples were run on a 10% SDS-PAGE gel, transferred to a Polyvinylidene Difluoride (PVDF) Membrane (Thermo Fisher Scientific), and probed with antibodies for PARP (1:2,000) and GAPDH (1:5,000). Immunoblotting was performed using Chemiluminescent Substrates (Thermo Fisher Scientific) and images captured with ImageQuant LAS 4000 (GE Healthcare).

HeLa cells were plated in culture flasks (5 × 10⁵ cells/flask), grown for 24 hours, and double thymidine block performed as described previously (39). Nine hours post release from the block, mitotic cells were collected by mitotic shake-off.

Mitotic cells were either lysed in 37°C immunoprecipitation lysis buffer (Thermo Fisher Scientific) and kept at 37°C or lysed in 4°C immunoprecipitation lysis buffer and kept on ice. Lysates were centrifuged at 13,000 rpm, 20 minutes, 37°C or 4°C. One milligram total protein of each supernatant (1 mL) was incubated with Protein G Sepharose Beads (GE Healthcare) and mouse lgG (sc-2025, Santa Cruz Biotechnology) for 1 hour, 37°C or 4°C on a rotator. Purified supernatants were incubated/rotated with fresh protein G beads and Tpm3.1 antibody or lgG-negative control at 37°C or 4°C for 2 hours. Beads were washed 4× with 37°C or 4°C immunoprecipitation lysis buffer. Proteins were eluted in Laemmli buffer at 95°C, 5 minutes, separated by SDS-PAGE, and transferred to a PVDF membrane. The membrane was divided, incubated with antibodies [α-tubulin (1:3,000), Tpm3.1 (1:3,000), EB1 (1:3,000), dynactin (1:2,000), and dynein IC (1:2,000)], and bound antibody detected using chemiluminescent substrates. Images were captured using ImageQuant LAS 4000. Coimmunoprecipitations (co-IP) with EB1 antibody followed the same procedure but in unsynchronized cell lysates.

Cells were seeded on glass-bottom dishes (0.5 × 10⁵ cells/dish; FluoroDish 35 mm, Coherent Scientific), grown for 24 hours, then cotransfected with GFP-tagged EB1 and mCherry-tagged Tpm3.1 constructs using Lipofectamine 3000 according to the manufacturer's instruction (Thermo Fisher Scientific). Serial Z sections (0.5 μm) of metaphase cells (maintained at 37°C, 5% CO₂) were acquired. To visualize the colocalization of Tpm3.1 and actin in metaphase, cells were transfected with a mCherry-tagged Tpm3.1 construct, fixed, and stained with Phalloidin (Atto 488, 1:250) as described previously (23). Immunostaining using the anti-Tpm3.1 antibody (1:2,000) was as described previously (34). Chromosomes were stained with DAPI in PBS and imaging performed at room temperature.

Cells were seeded onto glass coverslips in 6-well plates (0.5 × 10⁵ cells/well), grown for 24 hours, and then treated with DMSO (1%), ATM (3 μmol/L), TR100 (3 μmol/L), vincristine (1.5 nmol/L), or drug combinations for 15 hours. Cells were fixed in −20°C methanol, 5 minutes, blocked with 5% BSA (in PBS), and then incubated with primary antibodies [α-tubulin (1:2,000), pericentrin (1:3,000), NuMA (1:3,000), dynein (1:2,000), and dynactin (1:2.000); ref. 40] in 2.5% BSA (in PBS), 4°C, O/N. Cells were washed with PBS, incubated with secondary antibodies (1:500) in 2.5% BSA, 30 minutes, room temperature, washed with PBS, and stained with DAPI. Coverslips were mounted with Aqua-Poly/Mount Media (Polysciences) O/N at room temperature, then 0.3 or 0.5 μm serial Z sections were acquired at room temperature.

Serial Z sections (0.1 or 0.3 μm) of cells stained with antibodies for NuMA and α-tubulin were acquired at room temperature. Because of variability in brightness across the images, images were normalized prior to quantification. Minimum contrast was set to 0. Rather than using the brightest pixel in the volume for the maximum, which can be subject to fluctuation, setting the maximum value to one that spanned 99.998% of the summed pixels yielded the most consistent normalized images. After setting the contrast, images were converted to 8-bit with scaling applied. 3D Objects Counter was used to quantify the number of NuMA foci (41). Threshold was manually determined on a subset of images and this threshold was used across all images.

HeLa cells expressing α-tubulin-GFP and H2B-mCherry were seeded on glass-bottom 6-well plates (0.25 × 10⁵ cells/well; No. 1.5, MatTek Corporation), grown for 24 hours, and then treated with DMSO (1%), ATM (3 μmol/L), TR100 (3 μmol/L), vincristine (1.5 nmol/L), or drug combinations. Time-lapse live cell imaging was performed using a Nikon Eclipse Tie Microscope with plan-apochromat 20 × DIC objective and Texas Red filter. Imaging chamber was maintained at 37°C, 5% CO₂. Imaging commenced 1 hour post-drug treatment and lasted 72 hours.

HeLa cells expressing α-tubulin-GFP and H2B-mCherry were seeded on glass-bottom dishes (0.5 × 10⁵ cells/dish; FluoroDish 35 mm, Coherent Scientific), grown for 24 hours, and then treated with DMSO (1%), ATM (3 μmol/L), TR100 (3 μmol/L), vincristine (1.5 nmol/L), or drug combinations. To determine spindle phenotypes, cells were treated for 24 hours, then 0.3- or 0.5-μm serial Z sections of metaphase cells were acquired. To visualize the formation of mitotic spindles, cells were treated for 6 hours, then 1.8- or 2.0-μm serial Z sections of interphase or prophase cells were acquired at 2-minute intervals. Cells were imaged in a chamber maintained at 37°C, 5% CO₂.

HeLa cells expressing α-tubulin-GFP and H2B-mCherry were seeded on coverslips in 6-well plates (0.5 × 10⁵ cells/well), grown for 24 hours, and then treated with DMSO (1%), ATM (3 μmol/L), vincristine (1.5 nmol/L), or drug combination for 8 hours. Cells on coverslips were fixed for 5 minutes in −20°C methanol, mounted O/N at room temperature, then 0.3-μm serial Z sections were acquired. Distances between spindle poles were measured in three dimensions.

All imaging except the time-lapse live cell imaging for cell fate analysis was carried out using a Zeiss LSM 880 microscope with Airyscan and plan-apochromat 63×/1.40 oil DIC M27 objective. Raw images were processed using ZEN software. Fiji software was used for post-acquisition processing and analysis.

All data with error bars are shown as the mean ± SEM. The unpaired Student t test was performed with Prism 7.0 (GraphPad Software Inc.) to determine the statistical significance of experimental data as indicated in the figures. Differences between indicated groups were presented with P values designated with asterisks (*, P < 0.05; **, P < 0.01; ***, P < 0.005; and ****, P < 0.001). P values less than 0.05 were considered statistically significant.

Synergy between the anti-Tpm3.1 compound ATM and vinca alkaloid anti-MT drugs were evaluated using the Bliss independence model (35) in 19 cell lines representing prostate, lung, gastric, breast, head/neck, esophageal, colon, ovarian, and pancreatic cancer. Synergy was observed in all cell lines with variable Bliss synergy scores from 23 to 82 (Fig. 1A). Comparison of two vinca alkaloids, vincristine and vinorelbine (VNB), in combination with ATM in six lung cancer lines resulted in virtually identical synergy for vincristine and vinorelbine in each line, but the extent of synergy again varied with Bliss synergy scores from 21 to 81 (Fig. 1B). Synergy was also seen in a lung cancer (Calu-6) xenograft model (Fig. 1C–E). Specifically, vinorelbine (10 mg/kg) alone increased the median survival of mice by 26 days [137% increase in life span (ILS)] compared with the control group. The addition of ATM at a concentration (50 mg/kg) that did not show significant impact alone, effectively prolonged the survival of vinorelbine-treated mice by 38 days (200% ILS) as compared with the control. Therefore, the drug synergy in such a range of cancer types both in vitro and in vivo implicates the existence of a fundamental biological mechanism operating in a wide range of cancer types.

HeLa cells were selected as a study model to investigate the mechanism underlying the synergy. The combination of vincristine with two anti-Tpm3.1 compounds, ATM and TR100, exhibited antiproliferative synergy (72-hour treatments; Fig. 1F; Supplementary Fig. S1A–S1D), as determined by isobologram analysis (42), indicating that HeLa cells are highly sensitive to these drug combinations. Moreover, both ATM and vincristine showed concentration-dependent impact on reducing the IC₅₀ of vincristine and ATM, respectively (Fig. 1G and H). To quantitatively evaluate the levels of synergy and determine an optimal drug combination ratio for subsequent mechanistic studies, we performed a fixed-ratio isobologram analysis (Fig. 1I; Supplementary Fig. S1E–S1G), which assesses combinatorial effects using the Chou–Talalay CI at different fraction affected (Fa) levels (42). The results demonstrated a strong (0.1 < CI < 0.3) or very strong (CI < 0.1) synergy at a wide range of Fa levels (0.45–0.95; Fig. 1L; Supplementary Fig. S1H) for all tested concentration ratios.

To further clarify whether the observed synergy is cytotoxic or cytostatic, we measured apoptosis/necrosis induction using Annexin V and 7AAD and evaluated the level of cytotoxic synergy using the fixed-ratio isobologram method (24-hour treatments). Single drug treatments slightly increased the population of dead cells undergoing early (Annexin V⁺/7AAD⁻) and late (Annexin V⁺/7AAD⁺) apoptosis/necrosis; whereas, the drug combinations strongly promoted cell death induction (Fig. 1J; Supplementary Fig. S2A–S2C). Cytotoxic synergy was strong or very strong at most Fa levels (Fig. 1M; Supplementary Fig. S2D). To further evaluate the synergy in causing apoptosis, we detected the cleavage of PARP-1 by caspases that results in 89 kDa and 24 kDa fragments. This cleavage pattern of PARP-1 is considered to be a hallmark of apoptosis (43). Western blot analysis results showed that the combination treatment (24 hours) caused a robust activation of PARP-1 cleavage as seen by the increased levels of the 89 kDa band (Supplementary Fig. S2E), while single drugs barely showed any impact. Together, the data supports a cytotoxic rather than cytostatic mechanism of synergy.

We next studied the impact of drug combinations on the cell cycle of HeLa cells. Our previous work in neuroblastoma cells showed that the drug combinations induce a G₂–M-phase arrest (17). To correlate the drug impact on cell-cycle progression with the observed synergy, HeLa cells were treated using the same conditions described in the cell death analysis. As a microtubule-destabilizing agent, vincristine caused a concentration-dependent mitotic arrest with a correspondingly decreased G₀–G₁-phase cell population (Supplementary Fig. S3A). ATM and TR100 alone did not significantly disrupt cell-cycle progression (Supplementary Fig. S3B and S3C), even at high concentrations (10 μmol/L) that strongly suppressed cell proliferation. However, they effectively sensitized HeLa cells to vincristine-induced mitotic arrest (Fig. 1K; Supplementary Fig. S3D–S3F). This result suggests that the addition of anti-Tpm3.1 compounds enhances the defects caused by vincristine during mitotic spindle formation, which may be responsible for the drug synergy.

Given the dramatic combinational impact of ATM+VCR on mitosis, we determined whether combining ATM with other types of antimitosis drugs has a similar synergistic effect. We evaluated the antiproliferative synergy between ATM or vincristine and an Aurora-A inhibitor MLN8237 (MLN), which induces mitotic delay and spindle formation defects by inhibiting the phosphorylation of Aurora A (Thr288) and causing a loss of phospho-Aurora A at the spindle poles (44). The fixed-ratio isobologram analysis with CI calculation demonstrated that the MLN+ATM combination resulted in little synergy at the two different concentration ratios designated on the basis of their IC₅₀ ratios (Supplementary Fig. S4A and S4B). Synergy was only seen at high Fa levels (0.5–0.95), which was considerably lower than that seen with the ATM+VCR combinations (Supplementary Fig. S4E). The MLN+VCR combination displayed a significant antagonistic effect (Supplementary Fig. S4C, S4D and S4F). These results indicate that the antimitotic synergy between ATM and vincristine is mechanistically dependent on their specific and cooperative impacts (anti-Tpm3.1 and microtubule destabilizing) on mitosis progression.

To understand how Tpm3.1 inhibition enhances the antimitotic effect of vincristine, we visualized the localization of Tpm3.1 in mitotic cells using live-cell imaging of HeLa cells expressing Tpm3.1-mCherry and EB1-GFP. As a microtubule plus-end–tracking protein, the localization of EB1 on microtubules shows the spindle structure, indicating the mitotic stages of the cell. Imaging revealed that Tpm3.1 is enriched at the cell cortex of metaphase cells (Fig. 2A; Supplementary Fig. S5). The cortical localization of Tpm3.1 is also seen with actin filaments in metaphase cells (Fig. 2B). This cortical enrichment raises the possibility of Tpm3.1 being involved in interactions to aid in spindle formation (24–26, 45, 46).

To test the potential interaction of Tpm3.1 with the microtubule network we conducted co-IP analysis using a Tpm3.1 antibody and extracts of synchronous mitotic HeLa cells. To distinguish interactions that are dependent on the presence of intact microtubules, we performed co-IP under two conditions: high temperature (37°C) that supports the stability of polymerized microtubules, and low temperature (4°C) that causes microtubule depolymerization (47). Both α-tubulin and EB1 coimmunoprecipitated with Tpm3.1 at 37°C (Fig. 2C), indicating that microtubules potentially interact with Tpm3.1 during mitosis. The Tpm3.1 antibody also pulled down the microtubule-associated proteins dynein (intermediate chain) and dynactin (p150^Glued), all of which regulate mitotic spindle formation and positioning via the movement of astral microtubules and their cortical interactions to generate pulling forces (Fig. 2C; refs. 45, 48). Importantly, substantially reduced interactions were detected at 4°C (Fig. 2C), highlighting the importance of a polymerized microtubule structure for the Tpm3.1-associated interactions during mitosis. The Tpm3.1–microtubule interaction was further supported by co-IP using EB1 antibody at the two temperatures (Fig. 2D).

We next investigated the correlation between the combined drug impact on mitosis and the observed cytotoxic synergy. Long-term live-cell imaging (72 hours) was performed to visualize mitotic defects during the first mitosis and the fate of individual HeLa cells (expressing α-tubulin-GFP and H2B-mCherry) following drug treatments. DMSO control and ATM/TR100 (3 μmol/L)-treated cells spent a similar time in prometa-metaphase; whereas, vincristine treatment (1.5 nmol/L) caused a significant mitotic delay (Fig. 3A–C and E; Supplementary Fig. S6A and S6D). However, the prometa-metaphase duration was remarkably prolonged in cotreated cells with a 10-fold increase over vincristine treatment (Fig. 3D and E; Supplementary Fig. S6C and S6D). Moreover, vincristine failed to trigger cell death during the first cell cycle, but caused 21% of mother cells to undergo multi-polar cell division (Fig. 3F). The addition of ATM, which showed no impact on cell fate at the concentration tested, strongly sensitized cells to vincristine treatment with 70% of mother cells dying during the first mitosis (Fig. 3F). Of the ATM+VCR-treated mother cells that divided, 95% underwent a multi-polar division (Fig. 3F and G) and a similar result was seen with the TR100+VCR combination (Supplementary Fig. S6E and S6F). We also frequently observed failure of cytokinesis in cotreated cells that were able to survive the first mitosis (Supplementary Fig. S6H). Furthermore, the majority of single drug–treated daughter cells from the first cell division were alive at the end of the imaging; whereas 75% of ATM+VCR-treated daughter cells (49% for TR100+VCR) died during the second cell cycle as indicated by morphology change (Fig. 3H; Supplementary Fig. S6G). Therefore, we conclude that targeting Tpm3.1 potentiates the vincristine-induced mitotic defects, which correlates with the drug synergy in suppressing cell proliferation and causing cell death.

The high frequency of multi-polar division observed in cotreated cells suggests that a high incidence of multi-polar spindle formation may be linked to the prolonged mitotic arrest. We performed live-cell imaging to visualize and compare the mitotic spindle phenotypes caused by individual drugs versus combinations in HeLa cells expressing α-tubulin-GFP and H2B-mCherry. Vincristine (24-hour treatment, 1.5 nmol/L) caused spindle multi-polarity in 26% of prometa-metaphase cells (Fig. 3I). Although 74% of total mitotic spindles were still bipolar, 94% of such bipolar spindles had chromosomes attached to one of the poles away from the metaphase plate (Fig. 3I and K; Supplementary Fig. S7), indicating that mitotic defects also occurred during bipolar spindle formation. ATM or TR100 alone (24-hour treatment, 3 μmol/L) had a minor impact on mitotic spindles (Fig. 3I and K; Supplementary Figs. S6I and S7); however, when used in combination they increased the rate of vincristine-induced spindle multi-polarity by 3.4-fold, with 87% cotreated cells having multi-polar spindles (Fig. 3I and K; Supplementary Figs. S6I and S7). In addition, the anti-Tpm3.1 compounds increased the extent of multi-polarity when used in the combination with vincristine (Fig. 3J; Supplementary Fig. S6J). Together, these results illustrate that Tpm3.1 inhibition is highly effective in promoting spindle multi-polarity in the presence of a drug that destabilizes microtubules.

To understand the mechanism that drives these different spindle phenotypes, we visualized the mitotic spindle formation in drug-treated HeLa cells expressing α-tubulin-GFP and H2B-mCherry. Upon 6-hour treatment (1.5 nmol/L vincristine and/or 3 μmol/L ATM) interphase cells were selected for live-cell imaging. Although 95% of vincristine-treated prophase cells had only two microtubule organizing centers (MTOC; seen as microtubule asters), all of them exhibited supernumerary MTOCs upon nuclear envelope breakdown (NEBD, defined by the onset of condensed chromosomes released into the cytoplasm, visualized with H2B-mCherry; Fig. 4A–D). However, during a prolonged prometa-metaphase transition, 80% of vincristine-treated cells were able to cluster the additional MTOCs into two poles to create a bipolar spindle (Fig. 4C and D). Similarly, all cotreated cells had two MTOCs in prophase and multiple MTOCs after NEBD; however, those MTOCs subsequently formed stable multi-polar spindles during prometa-metaphase (Fig. 4A–D). This data indicates that vincristine induces a transient multi-polarity that can be resolved over time via a MTOC clustering mechanism and that this mechanism is compromised when Tpm3.1 is simultaneously inhibited.

We also observed that supernumerary MTOCs occurred as early as 1 hour after ATM+VCR treatment. Most cells cotreated for 1 to 5 hours were able to cluster extra MTOCs into bipolar spindles during mitotic arrest (Fig. 4E), suggesting that vincristine has already impacted spindles, but the addition of ATM was not effective enough to disrupt the clustering mechanism at such early time points. The declustering of extra MTOCs and multi-polar spindle formation were consistently seen after cotreating cells for 6 hours and thereafter (Fig. 4E), indicating that the combination needs approximately 6 hours to collaboratively disrupt spindle formation and show synergy as demonstrated by spindle multi-polarity. In single drug–treated cells, the bipolar spindles formed at early time points (before 8 hours) were slightly shorter compared with the DMSO control (Fig. 4F and G; Supplementary Fig. S8); whereas, the drug combination resulted in a 66% (type I) to 44% (type II) decrease depending on the formation stage of the bipolar spindles (Fig. 4F–H). It is notable that the 8-hour treatment is long enough for the drug combination to affect mitosis while still allowing the capture of bipolar spindles formed at early time points before anaphase onset. The very short spindles (2–4 μm) in cotreated cells were more likely to be pseudometaphase spindles, where the majority of chromosomes were organized by either one of the two poles (type I in Fig. 4H). With fewer misattached chromosomes and a clearer metaphase plate, the spindle length increased gradually (4–8 μm; type II in Fig. 4H), which might be a result of the progression of a prolonged metaphase/pseudometaphase. Overall, the results suggest that the drug combination causes a decrease in cortical pulling forces.

We then investigated whether centrosome amplification (CA) is involved in multi-polar spindle formation, which is reported to cause transient spindle multi-polarity that can be corrected via centrosome clustering in cancer cells (49, 50). Fixed cell imaging showed that upon NEBD (indicated by chromosomes stained with DAPI) only two centrosomes (pericentrin or PCNT staining) formed in vincristine- or combination-treated HeLa cells with supernumerary microtubule asters (α-tubulin staining; Fig. 5A and B). During prometa-metaphase/pseudometaphase, the majority of vincristine-treated (74%) or cotreated cells (96%) still exhibited two centrosomes (Fig. 5A and B). It is notable that cells were treated for 15 hours, which is long enough to generate multi-polar spindles within the first round of mitosis. Therefore, CA is not the primary cause for the spindle multi-polarity.

It is well-documented that NuMA is capable of tethering microtubule minus-ends and forming asters independently of centrosomes (51–54). To determine whether extra microtubule asters are organized by NuMA, we costained for NuMA, α-tubulin, and chromosomes in drug-treated cells as described above for PCNT. Fixed cell images show that in all drug-treated cells NuMA colocalized with each microtubule aster (Fig. 5D), supporting its association with the formation of acentrosomal MTOCs (aMTOC). In comparison with the control, vincristine increased the number of NuMA aggregates (seen as foci) with small to medium sizes upon NEBD, which then partially formed large foci and focused at two spindle poles in prometa-metaphase with several “free” aggregates (not involved in MTOCs) in the cytoplasm (Fig. 5C–F). In contrast, ATM+VCR further enhanced the fragmentation of NuMA with significantly more small and medium sized foci upon NEBD, which failed to cluster into large aggregates and resulted in persistence of multiple MTOCs during prometa-pseudometaphase (Fig. 5D–F). We conclude that inhibiting Tpm3.1 potentiates vincristine-induced aMTOC amplification via disrupting the assembly of NuMA aggregates and their localization around centrosomes, which ultimately results in enhanced spindle multi-polarity.

Because NuMA acts in a complex with dynein/dynactin to (i) assemble MTOCs (51–54) and (ii) generate the core pulling forces at the cortex for spindle positioning and anaphase onset (45), and (iii) dynactin and dynein co-immunoprecipitated with Tpm3.1 in mitotic cell extracts (Fig. 2C), this suggests a role for dynein/dynactin in the synergy mechanism. Fixed cell images of cells costained with NuMA and dynactin (p150^Glued) or dynein intermediate chain (dynein IC) show that upon cotreatment dynactin and dynein accumulated with every NuMA aggregate after NEBD and exhibited the same distribution defects as seen with NuMA (Fig. 6). These observations indicate that the declustering of NuMA aggregates induced by ATM+VCR consequentially disrupts the localization of these motor proteins and may account for spindle multi-polarity and the delay or failure of anaphase onset.

Our results demonstrate that the synergy of vincristine in combination with anti-Tpm3.1 compounds is very strong across a wide range of cancer types in vitro and a xenograft model of lung cancer. This is consistent with the drug synergy observed in neuroblastoma both in vitro and in vivo (17), and indicates a fundamental mechanism underlying the synergy of this combination regime. Our study reveals that the mechanism of drug synergy is primarily due to the suppression by anti-Tpm3.1 drugs of an intrinsic rescue response that normally corrects spindle defects and promotes mitotic progression in the presence of vincristine alone at low concentrations.

Vinca alkaloids cause mitotic arrest via subtly altering the dynamics of tubulin addition and loss at the ends of spindle microtubules as well as disrupting chromosome movement (55, 56). Vinca alkaloids at low clinically effective concentrations suppress microtubule dynamics with little or no impact on microtubule polymerization and spindle organization (56). High concentrations of vincristine cause the amplification of MTOCs, which leads to multi-polar spindle formation and consequent failure of mitosis (57, 58). The finding that low concentrations of vincristine initially lead to similar MTOC amplification upon NEBD provides insight into the difference between the impact of low and high concentrations of vincristine. Upon vincristine treatment at low concentrations, the cell subsequently responds to this perturbation by clustering supernumerary MTOCs to reconstitute a bipolar spindle via a Tpm3.1-dependent rescue mechanism. This rescue mechanism fails at high vincristine concentrations. Our finding provides a potential explanation for the observation that microtubule-destabilizing drugs are not very effective at causing multi-polar spindles in cancer cells (59).

Although centrosomes are the classic MTOCs and CA is often responsible for spindle pole fragmentation and spindle multi-polarity (49, 50, 60), they are not present at the vincristine-induced extra MTOCs upon NEBD. We found that NuMA associates with such aMTOC amplification. Under normal conditions, after NEBD, NuMA accumulates at spindle poles and recruits the cytoplasmic dynein/dynactin to form a complex that acts to mechanically stabilize the spindle architecture (51, 61, 62). Although the NuMA–dynein–dynactin complex is concentrated at poles with centrosomes, it is not involved in microtubule nucleation or formation of centrosomal MTOCs (51, 63); rather the complex tethers parallel microtubule bundles into a pole (i.e., MTOC) independently of centrosomes (51–54). Our observation of vincristine-induced MTOC amplification without CA provides further evidence to support the acentrosomal microtubule organizing function of the NuMA–dynein–dynactin complex.

The NuMA–dynein–dynactin complex has been shown to gradually accumulate at the cell cortex to generate core pulling forces especially during anaphase (45). Our observation that the drug combination causes the NuMA complex to accumulate at a large number of dispersed microtubule asters in prometa-metaphase/pseudometaphase may result in a reduced or delayed cortical enrichment of the NuMA complex and consequently decreased cortical pulling forces. This hypothesis is supported by the largely delayed anaphase onset and the division failure in cotreated cells. Moreover, the inability of cotreated cells to achieve bipolar attachment of chromosomes, activate separase, and satisfy the spindle assembly checkpoint also likely promotes the delay in anaphase onset. In addition, the cortical enrichment of Tpm3.1 may also contribute to a mechanical impact on cortical force generation. The impact of anti-Tpm3.1 compounds on spindle length in both the presence and absence of vincristine supports this model. This finding is consistent with the previous proposal that cortical actin promotes spindle lengthening as cells progress through mitosis (27).

The clinical implication of this study relates to the potential efficacy of low-dose vincristine treatment. Treatment of patients with vinca alkaloids is fraught with side effects including extensive neuropathic pain (64, 65). Combination therapy allows a reduction in the dose levels of single drugs, potentially reducing toxicity and resistance (66). Vincristine has been used with other chemotherapeutics in the clinic, most of which are DNA damage inducers (1, 2, 67). Recent preclinical studies also identify several kinds of drugs that synergize with vincristine (68–72); however, those combinations are used mostly to treat blood cancers or childhood cancers. Unlike those examples of synergy, which result from simultaneously targeting two distinct pathways, the synergy between vincristine and anti-Tpm3.1 compounds is based on the structural impact of these drugs on the mechanical roles and the cross-linking of the microtubule network and actin filaments during mitosis. Such synergy is not limited to certain types of cancer, indicating a potentially broad application of this combination regime.

In summary, our work shows that the combination of vincristine and anti-Tpm3.1 compounds exhibits a strong cytotoxic, but not cytostatic synergy in a wide range of cancer types. Tpm3.1 inhibition sensitizes cancer cells to low-dose vincristine treatment via eliminating a rescue response induced by vincristine to resolve its disruption in spindle assembly, indicating a previously undescribed role of Tpm3.1 or Tpm3.1-containing actin filaments in regulating mitosis. This is consistent with the recent observation that actin has a direct and dynamic association with spindle poles in mitosis (26). We conclude that the combination therapy using these agents is promising for the treatment in which vincristine is currently used in the clinic and makes a strong case to test this combination regimen in a clinical setting.

E.C. Hardeman reports receiving a commercial research grant from and has ownership interest (including patents) in TroBio Therapeutics Pty Ltd. P.W. Gunning reports receiving other commercial research support from and has ownership interest (including patents) in TroBio Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Wang, J.H. Stear, N.S. Bryce, T.P. Cripe, E.C. Hardeman, P.W. Gunning

Development of methodology: Y. Wang, J.H. Stear, M. Carnell

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Wang, J.H. Stear, A. Swain, X. Xu, J. Stehn, E.C. Hardeman, P.W. Gunning

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Wang, J.H. Stear, A. Swain, X. Xu, M. Carnell, J. Stehn, E.C. Hardeman, P.W. Gunning

Writing, review, and/or revision of the manuscript: Y. Wang, J.H. Stear, A. Swain, X. Xu, N.S. Bryce, I.B. Alieva, V.B. Dugina, T.P. Cripe, J. Stehn, E.C. Hardeman, P.W. Gunning

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Wang, J.H. Stear, E.C. Hardeman

Study supervision: E.C. Hardeman, P.W. Gunning, J. Stehn

We thank Greg Arndt, Anna Mariana, and Tim Failes (ACRF Drug Discovery Centre, Children's Cancer Institute, Kensington, Australia) for drug synergy screening services; Renee Whan, Alex Macmillan, and Elvis Pandzic [Biomedical Imaging Facility, University of new South Wales (UNSW), Sydney, Australia] for imaging assistance; Chris Brownlee and Emma Johansson Beves (Flow Cytometry Facility, UNSW, Sydney, Australia) for cell sorting support; Anthony A. Hyman (MPI-CBG) for providing HeLa cells stably expressing α-tubulin-GFP and H2B-mCherry; Andrew Burgess (ANZAC Research Institute, Concord Repatriation Hospital, Concord NSW, Australia) for advice on cell fate mapping; and Anna Akhmanova (Faculty of Science, Utrecht University, Utrecht, the Netherlands) for providing tagged EB1, 2, and 3 constructs and advice about the results. P.W. Gunning and E.C. Hardeman were supported by grants from the Australian Research Council (ARC grant DP160101623), the Australian National Health and Medical Research Council (NHMRC grant APP1100202, APP1079866), and The Kid's Cancer Project. I.B. Alieva was supported by the Russian Foundation for Basic Research (RFBR grant 18-29-09082).

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.