The extracellular matrix protein TGFBI induces microtubule stabilization and sensitizes ovarian cancers to paclitaxel

Abstract

The extracellular matrix (ECM) can induce chemotherapy resistance via AKT-mediated inhibition of apoptosis. Here, we show that loss of the ECM protein TGFBI (transforming growth factor beta induced) is sufficient to induce specific resistance to paclitaxel and mitotic spindle abnormalities in ovarian cancer cells. Paclitaxel-resistant cells treated with recombinant TGFBI protein show integrin-dependent restoration of paclitaxel sensitivity via FAK- and Rho-dependent stabilization of microtubules. Immunohistochemical staining for TGFBI in paclitaxel-treated ovarian cancers from a prospective clinical trial showed that morphological changes of paclitaxel-induced cytotoxicity were restricted to areas of strong expression of TGFBI. These data show that ECM can mediate taxane sensitivity by modulating microtubule stability.

Figure 1

Loss of TGFBI Is Sufficient to Induce Paclitaxel Resistance (A) Volcano plot shows log fold change in gene expression in the paclitaxel-resistant cell line SKOV-3TR compared to the sensitive parental line SKOV-3 and plotted against the likelihood of differential expression. Note that negative log2 expression ratios indicate underexpression in SKOV-3TR. Data points represent the probability value for differential gene expression and data shown is from four replicate experiments. (B) Relative expression levels of TGFBI in different cell lines using real time PCR. (C) Immunocytochemistry of stained sections from embedded cell pellets using anti-TGFBI antibody. Scale bars, 10 μm. (D) Western blotting of culture medium from SKOV-3TR, mock-transfected SKOV-3, and TGFBI siRNA-transfected SKOV3-K cell lines probed with anti-TGFBI antibody. (E) Effect of stable KD of TGFBI (SKOV3-A, SKOV3-K, and SKOV3-AK) on paclitaxel-induced apoptosis measured by FITC-annexin V and 7-AAD staining at 48 hr following paclitaxel treatment (150, 300, 600, 1200, and 2000 nM) compared to SKOV-3, mock-transfected SKOV-3 (mtSKOV3), and SKOV-3TR cells. Filled triangle indicates increasing paclitaxel dose across each group of bars. (F) Effect of stable KD of TGFBI on caspase 3/7 activation 48 hr following paclitaxel treatment. (G and H) Transient TGFBI-KD in OVCAR3 and TR175 lines induces paclitaxel resistance. Caspase 3/7 activation was estimated 48 hr following transfection using either a pool of 4 siRNAs targeting TGFBI or nontargeting scrambled controls (sc). OVCAR3 cells (G) or TR175 cells (H) were treated with paclitaxel for 48 hr. Immunoblot confirming knockdown of TGFBI protein is shown in (G). Error bars show mean ± SD.

Figure 2

Loss of TGFBI Causes Defective Paclitaxel-Induced Microtubule Polymerization (A–F) Overexpression of TGFBI in SKOV-3TR by pCSMT-TGFBI sensitizes microtubules to the polymerizing effect of paclitaxel. Arrowheads indicate paclitaxel-induced bundles (PIBs). Green, tubulin; blue, Dapi-stained DNA. Scale bars, 10 μm. (G) Paclitaxel induces Glu-tubulin formation. Cells were serum starved for 24 hr then treated with paclitaxel in serum-free medium (SFM) at 0, 4, 16, 75, 300, and 1200 nM concentrations for 1 hr. Lysates were collected for fluorescence immuno-blotting with anti-Glu-tubulin and anti-alpha tubulin. Bars represent the fold increase in Glu-tubulin fluorescence intensity values normalized for alpha-tubulin intensity values. Filled triangle indicates increasing paclitaxel dose across each group of bars. (H) A fluorescence immunoblot of soluble (sol) and insoluble (insol) tubulin fractions. (I) SKOV-3TR and SKOV3-K cells have increased soluble tubulin. Graph shows the percentages of soluble tubulin in relation to total tubulin in the different cell lines. Quantitative measurements were performed using fluorescence immunoblotting with anti-alpha tubulin antibody as described in Supplemental Experimental Procedures. Results shown are from two independent experiments. Horizontal bars indicate median values.

Figure 3

Loss of TGFBI Induces Mitotic Abnormalities (A and B) Stable KD of TGFBI in SKOV-3 cells results in abnormal mitotic spindle formation and centrosome amplification. Magenta, tubulin; blue, Dapi-stained DNA; yellow, gamma tubulin. (C) Proportion of abnormal mitotic cells at 9–30 days in SKOV-3TR, stable knockdown, and mock transfected (mt) cell line pools. (D) Proportion of interphase cells showing centrosome amplification. Number of cells counted is shown above corresponding bars. (E) Proportion of abnormal mitotic cells after 48 hr following transient knockdown of TGFBI using a pool of four siRNAs. Error bars show mean ± SD.

Figure 4

TGFBI Induces Microtubule Stabilization (A) Model of TGFBI induction of microtubule stabilization. (B) Microtubule stabilization demonstrated with anti-Glu tubulin antibody. SKOV-3 cells were plated for 90 min on rTGFBI or polylysine-coated glass slides (20 μg/ml) before immunofluorescence. Scale bars, 10 μm. (C) Cells were adhered to polylysine, fibronectin, or rTGFBI-coated wells (20 μg/ml) for 90 min before lysates were collected for immunoblotting using anti-phosphorylated FAK (P397). Also shown are immunoblots for lysates of cells treated in suspension with or without paclitaxel at 3 μM. (D) Percentages of SKOV-3 cells showing Glu-tubulin following adhesion to rTGFBI, fibronectin, or polylysine. (E) 48 hr following transfection using either FAK siRNA or nontargeting siRNAs, SKOV-3 cells were plated on rTGFBI coated glass slides for 90 min and the percentage of cells showing Glu-tubulin formation was estimated by immunofluorescence. (F) SKOV-3 cells were either treated with the Rho A inhibitor C3 toxin in SFM or with SFM alone for 4 hr before plating on rTGFBI-coated glass slides and estimation of Glu-tubulin formation. Error bars show mean ± SD.

Figure 5

rTGFBI Sensitizes Resistant Cells to the Effect of Paclitaxel (A) Cells were adhered to rTGFBI-coated wells (20 μg/ml) or noncoated wells for 24 hr prior to paclitaxel treatment for 48 hr. Shown is the percentage of apoptotic cells as measured using FITC-annexin V and 7-AAD staining. (B) Cells were serum starved for 24 hr then treated with paclitaxel in serum-free medium (SFM) at 0, 4, 16, 75, 300, and 1200 nM concentrations for 1 hour before lysates were collected for fluorescence immunoblotting using anti-Glu-tubulin and anti-alpha tubulin. Bars represent the fold increase in glu-tubulin fluorescence intensity values normalized for alpha tubulin intensity values. Filled triangle indicates increasing paclitaxel dose across each group of bars. (C) The slope and 95% confidence intervals for the linear regression of Glu-tubulin formation following paclitaxel treatment in SKOV-3 and mtSKOV3 (SKOV3-WT), SKOV3-A and SKOV3-K (SKOV3-KD), and SKOV3-KD following plating on rTGFBI. (D) SKOV-3TR cells were either plated on plastic or rTGFBI as in (A) and either treated with paclitaxel alone (SKOV3-TR) or with paclitaxel and verapamil 3.3 μM (TR + V and TR + V + rTGFBI) for 48 hr before caspase 3/7 activity was estimated. Also shown is the data for the parental sensitive line (SKOV-3) plated on plastic. (E and F) A2780 Cells (E) or PE0188 cells (F) were either plated on plastic or on rTGFBI (20 μg/ml) before paclitaxel treatment for 48 hr. Shown is the fold increase in caspase 3/7 activity. (G) Cells were either pretreated with 50 μg/ml of rTGFBI in SFM or SFM alone for 2 hr followed by paclitaxel treatment for 1 hr, washing, and incubation in full media for 48 hr. Shown is the percentage of apoptotic cells measured by FITC-annexin V and 7-AAD staining. A+rTGFBI+anti-alphaVbeta3; SKOV3-A cells pretreated with anti-alphaVbeta3 in SFM before treatment with rTGFBI and paclitaxel, A+rTGFBI+FAK-KD; SKOV3-A cells were transfected with siRNA targeting FAK 48 hr prior to rTGFBI and paclitaxel treatment, A+rTGFBI+C3-toxin; SKOV3-A cells were pretreated with the Rho A inhibitor, C3 toxin, in SFM for 4 hr before the application of rTGFBI and paclitaxel. Error bars show mean ± SD.

Figure 6

The ECM Protein TGFBI Sensitizes Ovarian Carcinoma Cells In Vivo to the Effect of Paclitaxel Correlation between expression of TGFBI and ECM genes in (A) CTCR-OV01 study and (B) independent ovarian cancer data set from Spentzos et al. (2005). Graphs show proportion of ECM-related genes increasing as a function of coexpression with TGFBI. (C) TGFBI expression in pretreatment biopsies of advanced ovarian carcinoma using real-time PCR. Resistant cases (n = 5; magenta) are compared to sensitive cases (n = 11; green). (D) Three-dimensional reconstruction of Z stack images obtained following double immunofluorescent staining of a representative ovarian carcinoma tissue section. Arrowheads indicate amplified centrosomes in a single cell confirmed by examination in all three planes. Anti-γ tubulin, green; chromosomal DNA, Hoechst 33258 red. Scale bar, 10 μm. (E) Centrosome amplification is associated with paclitaxel resistance. Centrosome counting was performed on samples from which adequate frozen tissue was available (n = 10; samples common between [C] and [E] are indicated by squares). In (C) and (E), horizontal bars indicate median values. (F) Immunohistochemistry for TGFBI of representative posttreatment ovarian cancer sample. Paclitaxel-induced morphological changes colocalize with focal TGFBI expression (brown staining, green box 1) but not in areas of low TGFBI expression (magenta box 2). Scale bars for main subfigure and boxes, 500 μm and 50 μm, respectively.

Figure 7

Proposed Model of Modulation of Paclitaxel Resistance by TGFBI via Effects on Microtubule Stability