Screening a living biobank identifies cabazitaxel as a strategy to combat acquired taxol resistance in high-grade serous ovarian cancer

Abstract

The anti-mitotic agent taxol (paclitaxel) remains a cornerstone of ovarian cancer treatment. To tackle drug resistance and toxicity, second-generation targeted anti-mitotic agents and combination strategies are being explored but have yet to demonstrate meaningful clinical benefits. A limitation is the lack of a platform to compare strategies in models that capture disease heterogeneity. To overcome this, we screen 83 patient-derived ex vivo ovarian cancer models that exhibit extensive intra- and inter-patient heterogeneity, testing four distinct approaches to enhance taxol sensitivity. Inhibitors of the HSET kinesin or the Mps1 spindle assembly checkpoint kinase show minimal impact on the taxol sensitivity landscape. By contrast, Bcl-xL inhibition exerts a global anti-proliferative effect. Inhibition of the MDR1 drug efflux pump restores taxol sensitivity in models characterized by ABCB1 overexpression. These MDR1-driven resistant models also respond to cabazitaxel, which is a poor MDR1 substrate, highlighting a potential therapeutic option for ovarian cancers with acquired taxol resistance.

Keywords: ABCB1; MDR1; centrosome amplification; chromosome instability; drug resistance; intrinsic apoptosis pathway; ovarian cancer; paclitaxel; spindle assembly checkpoint; taxol.

Graphical abstract

Figure 1

Analyzing the taxol sensitivity landscape of ovarian cancer (A) Workflow for sample processing, OCM validation, and drug sensitivity screen. (B) Carboplatin and taxol dose-response curves for 83 OCMs. (C) Bar charts showing GI₅₀ (ranked ordered) and AUC values for carboplatin and taxol. (D) xy plot showing mean AUC values for carboplatin and taxol, with linear regression, 95% confidence bands, and Spearman r correlation. p < 0.0001. (E) Dose-response curves for carboplatin (C) and taxol (T) for selected OCMs, showing mean and standard error. (F) Exemplar CFA images for selected OCMs following a 6-day exposure to carboplatin or taxol. https://creativecommons.org/licenses/by/4.0/ (A) adapted from Nelson et al. 2020 (license at https://creativecommons.org/licenses/by/4.0/). Values throughout derived from at least three biological replicates. See Figures S1 and S2; Table S1.

Figure 2

Global analysis of taxol modulation strategies (A) Mechanism of action of taxol and inhibitors targeting HSET, Mps1, Bcl-xL, and MDR1. (B) Heatmap of AUC values for OCMs treated with taxol plus HSETi, Mps1i, Bcl-xLi, or MDR1i, ranked by taxol sensitivity. (C) xy plots showing sensitivity to taxol versus taxol + inhibitor. Solid black lines show x = y, solid-colored lines show linear regression, r represents Pearson correlation. p < 0.0001 (all panels). (D) Box-and-whisker plots with interquartile ranges showing residuals based on the linear regression model (upper graph; values represent standard deviation as a measure of variance), and the y-x difference (lower graph). (E) xy plots showing log2-transformed AUC ratios for Bcl-xLi/Mps1i, Bcl-xLi/MDR1i, and Mps1i/MDR1i. Spearman r is used to measure correlation. Values derived from three biological replicates; p < 0.01 (left and center) and p < 0.0001 (right). See Figures S3, S4, and S5.

Figure 3

Mps1 inhibition sensitizes a subset of OCMs to taxol (A) Mitotic OCM.74-1 cells treated ± Mps1i showing Mad2 and centromeres (ACA). Insets show higher magnification of kinetochores. Bar 10 μm (B) xy graph plotting AUC for taxol versus taxol plus Mps1i, as shown in Figure 2C, highlighting OCMs selected for further analysis. (C) Dose-response curves for highlighted OCMs. Mean and SD from three biological replicates. (D) Exemplar CFA of selected OCMs after 6-day exposure to taxol ± Mps1i. (E) Time-lapse analysis of GFP-H2B-expressing OCM.99 and OCM.149 either untreated (Ctrl) or exposed to taxol (T, 4 nM), Mps1i (Mi, 2 μM), or the combination (TMi). Time in mitosis was measured from nuclear envelope breakdown (NEBD) to anaphase onset, ≥200 cells analyzed per condition. Lines represent median and interquartile range. One-way ANOVA, n.s. p > 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. (F) Bar graph quantifying mitotic abnormalities. (G) Cell fate profiling of OCM.74-5, either control or treated with taxol (1 nM), Mps1i (2 μM), or the combination. Horizontal bars represent a single cell (50 cells per condition), with colors indicating cell behavior. Numbers in colored boxes show the percentage of cells with indicated behavior. Data in (A), (E), (F), and (G) derived from one biological replicate. See Figure S6.

Figure 4

Bcl-xL inhibition has a broad-ranging sensitization effect (A) Graph showing apoptosis in RKO cells expressing tetracycline (Tet)-inducible Bcl-xL following exposure to taxol, taxol + Tet, and taxol + Tet + Bcl-xLi. (B) xy graph plotting AUC for taxol against taxol plus Bcl-xLi, as shown in Figure 2C, highlighting OCMs selected for further analysis. (C) Left: dose-response curves for highlighted OCMs. Data are mean and SD from three biological replicates. Right: exemplar CFA images after 6-day exposure to taxol ± Bcl-xLi. (D) Graph plotting effect of 100 nM Bcl-xLi alone with OCMs ranked by sensitivity, showing mean and SD from three biological replicates. (E) Graph showing duration of the first mitosis for OCMs 99 and 322. Each symbol represents an individual cell, with >30 cells analyzed. The bars show mean and SEM from one experiment. Paired t test. ∗∗∗p < 0.001. (F) xy graph plotting the response (mean normalized absorbance) of 17 OCMs to 100 nM Bcl-xLi against median time spent in mitosis. Pearson r is used to measure the correlation. p > 0.05. (G) Bar graph quantitating clonogenic potential of OCM.322 exposed to Bcl-xLi (10, 25, 50, and 100 nM) ± Mps1i (2 μM). Bars show mean and SD from three biological replicates. (H) Cell fate profiling of OCMs 99 and 322, untreated or treated with 100 nM Bcl-xLi for 5 days. Horizontal bars represent a single cell (50 cells per condition), with colors indicating cell behavior. Numbers in colored boxes show the percentage of cells with the indicated behavior. Data in (A), (E), (F), and (H) show one biological replicate. See Figures S5 and S7.

Figure 5

Inhibition of MDR1i-mediated drug efflux re-sensitizes a subset of OCMs to taxol (A) CFA of RKO cells expressing tetracycline-inducible MDR1 exposed to a taxol titration ±1 μg/mL tetracycline (Tet) and 250 nM MDR1i (MDi, elacridar). (B) xy graph plotting AUC for taxol against taxol plus MDR1i, as shown in Figure 2C, highlighting selected OCMs. (C) CFA of OCMs 149 and 246 exposed to taxol ±250 nM MDR1i. Red/green boxes highlight doses referred to in the text. (D) Taxol dose-response curves for OCM.149 ± MDR1i (left), with proliferation measured by time-lapse imaging of cells expressing GFP-Histone H2B (GFP-H2B), imaging every 4 h for 136 h. Data are from three biological replicates. Mean GI₅₀ values for OCM.149 (right). Paired t test ∗∗p < 0.01. (E) Immunoblot of MDR1 in selected OCMs. Bub3 is a loading control. (F) Mitotic OCM.149 cell stained for Aurora A and Phospho-Histone H3 Ser10 following treatment with 4 nM taxol and MDR1i. Bar 10 μm. (G) Bar graph quantifying the number of spindle poles in OCM.149 following exposure to 4 nM taxol (T), MDR1i (MDi), or the combination (T + MDi). At least 100 cells were analyzed per condition. (H) Time-lapse analysis of OCM.149 expressing GFP-H2B either untreated (Ctrl) or exposed to taxol (T, 4 nM), MDR1i (MDi), or the combination (T + MDi). Time in mitosis was measured from nuclear envelope breakdown (NEBD) to anaphase onset, with the median time indicated. At least 141 cells analyzed per condition. One-way ANOVA, n.s. p > 0.05, ∗∗∗∗p < 0.0001. (I) Bar graph quantifying mitotic abnormalities. (J) Cell fate profiling of OCM.149 either untreated (control) or treated with taxol (4 nM), MDR1i, or the combination for 6 days. Horizontal bars represent a single cell (50 cells per condition), with colors indicating cell behavior. Numbers in colored boxes show the percentage of cells with the indicated behavior. Data in (A) and (E)–(J) derived from one biological replicate. See Figures S5, S7, and S8.

Figure 6

Acquired taxol resistance correlates with ABCB1 overexpression and accumulated taxol exposure Multi-variable graph plotting AUC for taxol on the x axis against AUC for taxol + MDR1i on the y axis. Bubble size represents ABCB1 expression levels (normalized read count from RNA sequencing), and color represents patient exposure to taxol (mg) prior to sample collection. For six OCMs, taxol exposure was not known (nk). OCMs in the bottom right quadrant are sensitized to taxol by the MDR1i. AUC values derived from at least three biological replicates. Lines show linear regression with 95% confidence intervals and r represents Pearson correlation; p < 0.0001. See Figure S9; Table S2.

Figure 7

OCMs re-sensitized to taxol via MDR1i are sensitive to cabazitaxel monotherapy (A) xy graph plotting AUC values for taxol against taxol + MDR1i, as shown in Figure 2C, highlighting 36 OCMs screened with cabazitaxel. (B) Heatmap showing AUC values of OCMs rescreened with taxol, taxol + MDR1i, or cabazitaxel, rank ordered by Taxol response. Data are average from three biological replicates. (C) xy graphs plotting AUC values for taxol + MDR1i against either Taxol (left) or cabazitaxel (middle). Superimposing these two graphs (right) highlights ten OCMs (blue) sensitive to cabazitaxel monotherapy. Lines show linear regression with 95% confidence intervals and the values represent Spearman r correlations; p < 0.0001 (left and center). (D) CFA of OCM.339 after exposure to taxol ± MDR1i or cabazitaxel. (E) Dose-response curves for OCM.149 exposed to taxol ± MDR1i or cabazitaxel with proliferation measured by time-lapse imaging of cells expressing GFP-H2B, imaging every 4 h for at least 144 h. Graphs show the mean and SEM from at least three biological replicates. Values show GI₅₀. (F) Time-lapse analysis of GFP-H2B-expressing OCMs 149 and 339, untreated (Ctrl) or exposed to 4 nM cabazitaxel (Cabazi). Time in mitosis represents nuclear envelope breakdown (NEBD) to anaphase onset, with the mean indicated. At least 250 cells analyzed per condition. Mann-Whitney test ∗p < 0.05, ∗∗∗∗p < 0.0001. (G) Bar graph quantifying mitotic abnormalities. (H) Images of GFP-H2B-expressing OCM.149 after 72 h, showing micronuclei (arrow heads) in control cells and highly abnormal nuclei (arrows) in cabazitaxel-treated cells. Bar 20 μm. Data in (F)–(H) derived from one biological replicate. See Figure S10.