The extracellular matrix protein TGFBI enhances the cytotoxic response of cancer cells to paclitaxel by affecting integrin signals that stabilize microtubules. Extending the implications of this knowledge, we tested the more general hypothesis that cancer cell signals which increase microtubule stability before exposure to paclitaxel may increase its ability to stabilize microtubules and thereby enhance its cytotoxicity. Toward this end, we carried out an siRNA screen to evaluate how genetic depletion affected microtubule stabilization, cell viability, and apoptosis. High content microscopic analysis was carried out in the absence or presence of paclitaxel. Kinase knockdowns that stabilized microtubules strongly enhanced the effects of paclitaxel treatment. Conversely, kinase knockdowns that enhanced paclitaxel-mediated cytotoxicity sensitized cells to microtubule stabilization by paclitaxel. The siRNA screen identified several genes that have not been linked previously to microtubule regulation or paclitaxel response. Gene shaving and Bayesian resampling used to classify these genes suggested three pathways of paclitaxel-induced cell death related to apoptosis and microtubule stability, apoptosis alone, or neither process. Our results offer a functional classification of the genetic basis for paclitaxel sensitivity and they support the hypothesis that stabilizing microtubules prior to therapy could enhance antitumor responses to paclitaxel treatment. Cancer Res; 71(17); 5806–17. ©2011 AACR.

Taxanes (1, 2) are microtubule-stabilizing drugs used in the standard management of two of the most common solid tumors, breast and ovarian cancer, in addition to several other cancers (3, 4). In each of these cancers, only a fraction of patients will benefit. Understanding mechanisms of response to primary taxane therapy is an essential step in developing novel combination therapies that will improve response rates. Improving response rates to an established therapy is likely to improve survival outcomes for patients.

The taxane, paclitaxel, binds to specific pockets within β-tubulin in α/β-tubulin dimers (5), that are incorporated within preformed microtubule polymers (6, 7). Paclitaxel binding increases microtubule polymerization and stability and induces mitotic arrest of cycling cancer cells followed by apoptosis. Increasing the fraction of unstable microtubules in the cell by mutations outside the paclitaxel-binding sites (8, 9), overexpression of β3-tubulin (10, 11) or overexpression of depolymerizing microtubule-associated proteins (12) results in a disproportionate increase in free tubulin dimers and a consequent decrease in paclitaxel binding sites, leading to profound paclitaxel resistance.

Loss of the extracellular matrix protein TGFBI results in decreased focal adhesion kinase activation, decreased microtubule stability, and an increase in unstable microtubules resulting in significant paclitaxel resistance (13). Ovarian cancers which lack TGFBI are intrinsically resistant to single-agent paclitaxel therapy. Based on these findings we proposed a model whereby the proportion of baseline stable microtubules determines the availability of binding sites for paclitaxel and consequently the magnitude of paclitaxel-induced microtubule stabilization. In this study, we tested the hypothesis that modulating baseline microtubule stability by perturbing signaling pathways would alter paclitaxel response and could be utilized therapeutically.

Cell culture and siRNA transfection

All cell lines were obtained from the American Tissue Type Culture Collection. Cells were maintained in RPMI-140 medium supplemented with 10% fetal bovine serum and incubated at 37°C and 5% CO2. Human skin fibroblasts were a kind gift from Dr Karl Morten. A549 and A549-T12 lung carcinoma cells were a kind gift from Dr Susan Horwitz. Cell lines were authenticated using short tandem repeat profiling.

Transient knockdown experiments for the screen were carried out by transfection using a pool of 4 individual siRNA duplexes targeting individual genes (siGenome Smart pool, Dharmacon), or individual duplexes, using Dharmafect 4 (Dharmacon) according to the manufacturers' instructions. In summary, 0.1 μL of 20 μmol/L siRNA was mixed with 0.1 μL of Dharmafect 4 and 10 μL of Optimem medium (Invitrogen) for 20 minutes in a well of 384-well plate and overlaid with 40 μL of cell suspension (1,750 cells) in antibiotic-free cell line–specific media. This gave a final siRNA concentration of 40 nmol/L per well. The catalogue numbers and sequences of the siRNAs used are listed in Supplementary Table S1. siControl nontargeting siRNA number 4 (Dharmacon) was used as a transfection control in all experiments. Pools of 4 siRNAs targeting Tau, MAP2, and stathmin were obtained from Dharmacon (siGenome library).

For the purpose of the screen, the first 3 columns of every plate were reserved for transfection controls as follows: siControl, siTOX (Dharmacon) as control for transfection efficiency and cells only.

Immune-staining microscopy techniques, image acquisition, and image analysis

SKOv3 ovarian cancer cells were transfected using siRNAs in individual wells of black 384-well plates (Greiner, 1,750 cells per well) for 48 hours then either treated with paclitaxel (Sigma) 8 nmol/L or diluent only then fixed using 4% paraformaldehyde (PFA) in PBS for 3 to 5 minutes. Cells were washed to remove PFA and were incubated in absolute ethanol at −20°C for at least 1 hour before being washed with Tris-buffered saline, 0.04% SDS (Sigma), and 0.2% Triton X-100 (Sigma). Blocking was carried out using 1.5% bovine serum albumin in Tris-buffered saline. Nuclear DNA was stained using Hoechst 33258 dye (Invitrogen), and stable microtubules were revealed by rabbit anti–Glu-tubulin antibody (Millipore) and counterstained using anti-rabbit tagged with alexa-594 (Invitrogen). Image acquisition was obtained as previously described using the In Cell Analyzer 1000 (GE Healthcare; ref. 14). For quantification of α-tubulin intensity, mouse anti–α-tubulin antibody (Sigma) was counterstained with anti-mouse antibody tagged with alexa-488.

Statistical analysis

Analysis of data for the screen from individual cells (n = ∼4 million) was automated using the R programming language and statistical computing environment. In brief, each cell analyzed was annotated to individual wells representing a gene knockdown. Background subtracted mean pixel intensity values were calculated for each cell, and the mean of all cells per well was used as an estimate of the magnitude of microtubule stability for each well. The median of triplicate values was obtained for each gene with and without paclitaxel and expressed as a percentage of the mean value for microtubule stability following transfection with nontargeting siRNAs. The values in each plate were centered on 100 using median scaling. The same method was used for analyzing caspase and viability data.

Robust siRNA clusters were obtained using bagged gene shaving (15). We used 256 bootstrap resamples and obtained each resample first by ranking the individual cells in each sample, and then applying a weighted ranking across each siRNA using weights obtained from the Bayesian bootstrap distribution. Aggregated clusters obtained from the adjacency matrix were required to contain siRNA with at least 50% similarity within the cluster.

Cytotoxicity assays

For measuring apoptosis, caspase 3/7 activity was estimated 48 hours following treatment by adding equal volume of the Caspase-Glo 3/7 Assay reagent (Promega). Luminescence was read following 1 hour of incubation using a luminescence plate reader. Cell viability was estimated by adding 5 μL of the CellTiter blue Kit reagent (Promega) according to the manufacturer's instructions. Fluorescence was measured 4 hours later using a plate reader.

For colony forming assays, SKOv3 cells (2 × 105 cells/well) or Hey cells (9 × 104 cells/well) were seeded in 6-well plates. Twenty-four hours later, cells were transfected using either nontargeting siRNA or siRNAs targeting ILK or RAPGEF4 for 48 hours, then cells were reseeded in different 6-well plates at the density of 250 cells/well for SKOv3 or 200 cells/well for Hey. After 2 days, cells were treated with paclitaxel 1 nmol/L for 48 hours and maintained in the corresponding antibiotic-free medium for either 7 days (SKOv3 cells) or 4 days (Hey cells) then fixed and stained using Coomassie Blue and individual colonies were counted.

For measuring cell kinetics, 5-bromo-2′-deoxy-uridine incorporation was measured using an ELISA based assay (Roche Applied Sciences) following the manufacturer's instructions as a measure of the extent of cell cycling.

Reagents.

Paclitaxel and nocodazole were obtained from Sigma. Antibodies for Tau, MAP2, and stathmin were obtained from Millipore and for p-STAT3 from Cell Signaling. All chemicals were obtained from Sigma unless otherwise indicated.

A hypothesis-driven screen to investigate the influence of baseline microtubule stability on paclitaxel cytotoxicity

To test the hypothesis that modulating signaling-induced microtubule stabilization can directly influence paclitaxel-induced microtubule stabilization and cytotoxicity, we conducted an siRNA screen to identify candidate genes which upon knockdown enhance paclitaxel-induced microtubule stability, apoptosis, and cell death. We chose the kinome library from Dharmacon, as microtubule-associated proteins are strongly regulated by phosphorylation (16–21). SKOv3 ovarian cancer cells were transfected with either nontargeting control siRNA or a pool of 4 siRNAs targeting each of the 779 genes included in this library.

To measure microtubule stability in a high-throughput manner, we carried out triplicate immunofluorescence experiments in 384-well plates using an antidetyrosinated α-tubulin (also called Glu-tubulin) antibody, a specific marker for stable microtubules (22, 23), that was counterstained with an alexa-594–tagged antibody. Automated image capture was carried out, followed by automated image analysis to extract background-subtracted mean pixel intensity values from the cytoplasm of individual cells as a measure of microtubule stability. When microtubule stability was measured following exposure to increasing concentrations of paclitaxel (Supplementary Fig. S1) the Glu-tubulin–based assay yielded a wide (4.3-fold) dynamic range compared with the narrower range (2.5-fold) of α-tubulin–based assays. This difference may relate to the relatively small fraction of stable microtubules (Glu-tubulin) in cancer cells compared with total microtubules (α-tubulin) prior to paclitaxel treatment (Supplementary Fig. S1). The use of automated microscopy and image analysis for individual cells yielded a consistent and reproducible linear relationship between paclitaxel dose and microtubule stability (Supplementary Fig. S1) in contrast to the use of other high-throughput methods based on measuring total fluorescence from individual wells or measuring well luminescence in assays that utilize peroxidase-conjugated antibodies (data not shown). To verify the fidelity of the technique in identifying changes in microtubule stability, we tested the effect of depletion of the microtubule destabilizing protein stathmin (12) and the microtubule polymerizing protein MAP2 (24). As expected, depletion of stathmin resulted in a significant increase in baseline microtubule stability and paclitaxel-induced microtubule stabilization whereas depletion of MAP2 resulted in the opposite effect. In addition, depletion of Tau, which has previously been shown to compete with paclitaxel for the binding site on β-tubulin (24–26), resulted in a marked increase in paclitaxel-induced microtubule stabilization (Supplementary Fig. S2). In addition, the above results were also reproduced by measuring total microtubule polymer formation by using an antibody for α-tubulin (Supplementary Fig. S2).

Microtubule stability was measured following 1 hour of paclitaxel treatment as this time point highly correlated with cytotoxicity after 48 hours of treatment (Pearson correlation coefficient = 0.97, P < 0.001; Supplementary Fig. S3). Later estimation of microtubule stability was less reproducible because of easier detachment of cells following arrest in mitosis (not shown). In addition to measuring microtubule stability, we also measured caspase activity and cell viability with or without paclitaxel treatment. To avoid the bias introduced by lifting off of mitotic cells when measuring apoptosis or viability, we utilized assays that did not require washing of media prior to estimation of cytotoxicity (see Materials and Methods). All three outcome measures were optimized to maximize their dynamic range. Our assay development experiments showed that decreasing the proportion of cycling SKOv3 cells significantly decreases paclitaxel-induced apoptosis and this may decrease the dynamic range of the cytotoxicity tests and increase the false negative rate in siRNA experiments (Supplementary Fig. S4). As transfection itself can decrease proliferation rate and increase the proportion of noncycling cells, we conducted detailed assay development experiments to optimize our transfection conditions to avoid the bias introduced by shifting cell kinetics. We tested multiple starting cell densities (1,000, 1,250, 1,500, 1,750, and 2,500 cells per well), 8 transfection reagents, 9 nontargeting siRNAs, and 4 siRNAs that were known to induce cell death and measured cell viability using CellTiter blue kit reagent (Promega) following 48 hours of transfection (Supplementary Fig. S5). These experiments showed that using 1,750 cells, and Dharmafect 4 transfection reagent minimized the inhibitory effect of transfection on cell proliferation when compared with nontransfected cells and resulted in the least variability across replicates (Supplementary Fig. S5). Apoptosis was measured following 48 hours as this gave a detection range of 9-fold (Supplementary Fig. S6), and viability was measured after 72 hours as this resulted in a 2.6-fold detection range (Supplementary Fig. S7A) compared with only 1.6-fold detection range after 48 hours (Supplementary Fig. S7B). A dose of 8 nmol/L of paclitaxel was chosen to maximize the chance of detecting sensitizing siRNAs by providing at least an 80% window for detecting enhanced taxane effects. Using 8 nmol/L of paclitaxel resulted in only 2-fold increase in microtubule stability, a 2-fold increase in apoptosis, and a 15% decrease in cell viability. This dose was also chosen because cytotoxicity measurements tended to plateau at higher concentrations (Supplementary Fig. S7).

Thus, 3 kinome screens were conducted in parallel to measure the effect of siRNA transfections on microtubule stability, cell viability, and apoptosis with or without paclitaxel (Fig. 1A). Each of the 3 screens was conducted in triplicate. To estimate data precision, we measured the mean correlation coefficients among the 3 replicates for microtubule stability, cell viability, and apoptosis and found that the average Pearson correlation coefficients were 0.77, 0.86, and 0.96, respectively (Fig. 1B). This high reproducibility across replicates is a strong indicator of the high precision of data (27) obtained in this screen.

Figure 1.

A hypothesis-driven screen to investigate the influence of baseline microtubule stability on paclitaxel cytotoxicity. A, an outline of the siRNA screens conducted. The 779 siRNA pools used in each screen were divided into three 384-well plates (320 pools for plates 1 and 2 and 139 pools in plate 3), and 3 replicates were used for each plate with and without paclitaxel giving a total of 54,384-well plates for the entire experiment. B, shown in the lower diagonal of each plot are dot plots representing the correlation between the measures obtained for each pair of the 3 replicates used in the screen following paclitaxel treatment. The corresponding Pearson correlation coefficient for each pair is shown in the upper diagonal of each plot.

Figure 1.

A hypothesis-driven screen to investigate the influence of baseline microtubule stability on paclitaxel cytotoxicity. A, an outline of the siRNA screens conducted. The 779 siRNA pools used in each screen were divided into three 384-well plates (320 pools for plates 1 and 2 and 139 pools in plate 3), and 3 replicates were used for each plate with and without paclitaxel giving a total of 54,384-well plates for the entire experiment. B, shown in the lower diagonal of each plot are dot plots representing the correlation between the measures obtained for each pair of the 3 replicates used in the screen following paclitaxel treatment. The corresponding Pearson correlation coefficient for each pair is shown in the upper diagonal of each plot.

Close modal

Baseline microtubule stability strongly correlates with paclitaxel-induced microtubule stabilization

To test the hypothesis that increasing baseline microtubule stability results in enhanced paclitaxel-induced microtubule stabilization, we calculated the median microtubule stability values of 3 replicates prior to treatment and correlated this with the median value following treatment and found a highly statistically significant positive correlation (Pearson r2 = 0.66, P < 0.001). This indicated that on a global scale, baseline microtubule stability predicted taxane-induced microtubule stabilization (Fig. 2).

Figure 2.

Baseline microtubule stability strongly correlates with paclitaxel-induced microtubule stabilization. Seven hundred seventy nine siRNA pools targeting individual genes were used to transfect SKOv3 cells followed by immunofluorescence as described in Materials and Methods. The mean cytoplasmic pixel intensity values for each cell were extracted and background subtracted using the In-Cell Investigator (GE Healthcare) software from a total of 64,800 images including a total of approximately 4 million cells. The mean value of all cells in a well, which represent a single gene, was used to calculate the percentage microtubule stability following knockdown of a gene compared with the mean value obtained for nontargeting siRNA-transfected cells. In this plot, each circle represents the median of triplicate percentage values obtained following siRNA treatment alone, representing baseline microtubule stability, or siRNA and paclitaxel treatment.

Figure 2.

Baseline microtubule stability strongly correlates with paclitaxel-induced microtubule stabilization. Seven hundred seventy nine siRNA pools targeting individual genes were used to transfect SKOv3 cells followed by immunofluorescence as described in Materials and Methods. The mean cytoplasmic pixel intensity values for each cell were extracted and background subtracted using the In-Cell Investigator (GE Healthcare) software from a total of 64,800 images including a total of approximately 4 million cells. The mean value of all cells in a well, which represent a single gene, was used to calculate the percentage microtubule stability following knockdown of a gene compared with the mean value obtained for nontargeting siRNA-transfected cells. In this plot, each circle represents the median of triplicate percentage values obtained following siRNA treatment alone, representing baseline microtubule stability, or siRNA and paclitaxel treatment.

Close modal

Enhanced microtubule stabilization is a significant mechanism for enhanced paclitaxel-mediated cytotoxicity

We next sought to identify a group of kinases that regulate sensitivity of cancer cells to paclitaxel. We selected siRNAs that (i) induced a 2-fold enhancement of cell growth inhibition when combined with paclitaxel compared with the cytotoxic effect of siRNA alone (normalized paclitaxel- and siRNA-treated value divided by normalized siRNA-treated value of more than 2-fold) or (ii) increased paclitaxel-induced cytotoxicity following siRNA by 2.5-fold when compared with paclitaxel-induced cytotoxicity following transfection with nontargeting siRNA. To identify enhancers of paclitaxel-mediated microtubule stability, we selected siRNA pools that were in the upper quartile of all siRNAs tested (microtubule stability >116.1% compared with paclitaxel-treated cells transfected with nontargeting siRNAs). Using these criteria, we identified 61 siRNA pools that enhanced paclitaxel-induced apoptosis and 50 siRNA pools that enhanced paclitaxel-induced cell death (Supplementary Tables S2 and S3). Importantly, 34.4% of the genes that sensitized cells to paclitaxel-induced apoptosis (Fig. 3A and Supplementary Table S2) and 46% of the genes that sensitized cells to paclitaxel-mediated cell death (Fig. 3B and Supplementary Table S3) also enhanced paclitaxel-induced microtubule stability. The probability of these proportions occurring by chance alone is 0.03 and less than 0.001, respectively, based on 100,000 random samplings without replacement. This analysis confirmed that enhancing microtubule stability following paclitaxel treatment is strongly associated with enhanced paclitaxel-induced cell death. Interestingly, several genes that enhanced apoptosis and microtubule stability such as CHUK (also known as inhibitor of κB kinase A [IKKalpha]) or enhanced cytotoxicity such as CARD14, are known to activate the NF-κB) pathway (28, 29). This pathway is constitutively active in SKOv3 cells (30) and its inhibition has been previously reported to enhance paclitaxel-mediated cell death (31–33).

Figure 3.

Enhanced microtubule stabilization is a significant mechanism for enhanced paclitaxel-mediated cytotoxicity. SKOv3 cells were transfected and treated as described in Fig. 1 and Materials and Methods. After this, the following parameters were measured: (i) the fold change in apoptosis or cytotoxicity following siRNA pool transfection alone compared with nontargeting siRNA (fold no drug), (ii) the fold change in apoptosis or cytotoxicity following siRNA pool transfection and paclitaxel compared with nontargeting siRNA transfection and paclitaxel (fold drug treated), and (iii) the ratio of (ii) over (i). siRNA pools were identified that either increased the fold drug treated over 2.5 or increased the ratio over 2. In the figure, each circle represents the median of triplicate values of the indicated ratio. Data for apoptosis and cell viability is presented in (A) and (B), respectively. The blue lines indicate the cutoffs used in the screen for identifying hits. The green circles indicate genes that also enhanced paclitaxel-induced microtubule stability.

Figure 3.

Enhanced microtubule stabilization is a significant mechanism for enhanced paclitaxel-mediated cytotoxicity. SKOv3 cells were transfected and treated as described in Fig. 1 and Materials and Methods. After this, the following parameters were measured: (i) the fold change in apoptosis or cytotoxicity following siRNA pool transfection alone compared with nontargeting siRNA (fold no drug), (ii) the fold change in apoptosis or cytotoxicity following siRNA pool transfection and paclitaxel compared with nontargeting siRNA transfection and paclitaxel (fold drug treated), and (iii) the ratio of (ii) over (i). siRNA pools were identified that either increased the fold drug treated over 2.5 or increased the ratio over 2. In the figure, each circle represents the median of triplicate values of the indicated ratio. Data for apoptosis and cell viability is presented in (A) and (B), respectively. The blue lines indicate the cutoffs used in the screen for identifying hits. The green circles indicate genes that also enhanced paclitaxel-induced microtubule stability.

Close modal

Measuring microtubule stability identifies three distinct mechanistic classes for modulating paclitaxel-induced cytotoxicity

We next sought to classify siRNA pools targeting individual genes according to their similarities in the mechanism of modulating paclitaxel response. To do this, we first used gene shaving to obtain classes of genes that behave similarly in modulating cell viability and apoptosis with or without paclitaxel and paclitaxel-induced microtubule stability (15). We next used Bayesian resampling to identify classes that were statistically robust to permutation analysis. This method identified predominantly 3 robust classes of genes (Fig. 4A, Supplementary Tables S4–S6) in which paclitaxel-induced microtubule stability either (i) significantly correlated with paclitaxel-induced apoptosis (Pearson r2 = 0.5, P < 0.001, n = 208 siRNA pools, Fig. 4B) and to a lesser extent with paclitaxel-induced loss of viability (Pearson r2 = −0.2, P < 0.001, n = 208; Fig. 4B), (ii) significantly correlated with both paclitaxel-induced apoptosis (Pearson r2 = 0.5, P < 0.001, n = 132 siRNA pools, Fig. 4B) and loss of cell viability (Pearson r2 = −0.7, P < 0.001, n = 132 siRNA pools, Fig. 4B), or (iii) paradoxically, inversely correlated with paclitaxel-induced apoptosis (Pearson r2 = −0.6, P < 0.001, n = 105 siRNA pools, Fig. 4B) and loss of cell viability (Pearson r2 = 0.6, P < 0.001, n = 105 siRNA pools, Fig. 4B). Interestingly, the latter class was also characterized by a poor correlation between paclitaxel-induced apoptosis and paclitaxel-induced loss of cell viability (Pearson r2 = −0.3, Fig. 4B). To explain the paradoxical inverse correlation between microtubule stability and apoptosis in the third gene cluster, we tested the hypothesis that depletion of genes that result in inhibition of apoptosis may result in resistance to paclitaxel regardless of the effect on microtubule stability. To test this possibility, we identified all siRNA pools that resulted in increased paclitaxel-induced microtubule stabilization and decreased paclitaxel-induced apoptosis (n = 34). Examination of the known function of these genes revealed that several of them were either genes with known critical function in inducing apoptosis or genes that played a role in recognition of cell damage. For example, receptor interacting serine threonine kinase 2 (RIPK2, also known as CARD3) is a known apoptosis inducer (34). Depletion of RIPK2 resulted in decreased apoptosis irrespective of paclitaxel administration. Also, loss of the cell-cycle checkpoint protein ATM (35), resulted in enhanced microtubule stability but decreased paclitaxel-induced apoptosis and improved cell survival following paclitaxel treatment. We, therefore, reasoned that loss of these genes may result in a significant decrease in apoptosis that is independent of paclitaxel effect. Indeed, the mean normalized apoptosis induction following transfection with these siRNAs (n = 34) was less than half the mean of the rest (n = 745) of the genes (65% vs. 140%, respectively, P < 0.001, t test, Fig. 4C).

Figure 4.

Measuring microtubule stability identifies 3 distinct mechanistic classes for modulating paclitaxel-induced cytotoxicity. A, cluster co-occurrence matrix showing the results of gene clustering (from 1 to 779) following permutation analysis based on the similarity between genes in modulating paclitaxel-induced cytotoxicity, apoptosis, and microtubule stabilization. In this figure, red signifies high similarity and dark blue signifies no similarity. This analysis identified 3 main gene clusters. B, the pairwise correlation between the fold change in microtubule stability, apoptosis, and cell viability following siRNA and paclitaxel treatments. Note the good correlation between microtubule stability and apoptosis and viability for cluster 2, the good correlation between microtubule stability and apoptosis for cluster 1, and the paradoxical inverse correlation between microtubule stability and cell viability and apoptosis for cluster 3. In cluster 3, circles in magenta represent genes that upon depletion increase microtubule stability but paradoxically decrease apoptosis and improve cell viability following paclitaxel treatment. C, a box plot is presented to compare the fold change in apoptosis following siRNA-mediated depletion of genes shown in magenta in cluster 3 (n = 34) with the rest of the genes (n = 745). Note that for this cluster of 34 genes, siRNA depletion results in a significant decrease in apoptosis irrespective of paclitaxel treatment.

Figure 4.

Measuring microtubule stability identifies 3 distinct mechanistic classes for modulating paclitaxel-induced cytotoxicity. A, cluster co-occurrence matrix showing the results of gene clustering (from 1 to 779) following permutation analysis based on the similarity between genes in modulating paclitaxel-induced cytotoxicity, apoptosis, and microtubule stabilization. In this figure, red signifies high similarity and dark blue signifies no similarity. This analysis identified 3 main gene clusters. B, the pairwise correlation between the fold change in microtubule stability, apoptosis, and cell viability following siRNA and paclitaxel treatments. Note the good correlation between microtubule stability and apoptosis and viability for cluster 2, the good correlation between microtubule stability and apoptosis for cluster 1, and the paradoxical inverse correlation between microtubule stability and cell viability and apoptosis for cluster 3. In cluster 3, circles in magenta represent genes that upon depletion increase microtubule stability but paradoxically decrease apoptosis and improve cell viability following paclitaxel treatment. C, a box plot is presented to compare the fold change in apoptosis following siRNA-mediated depletion of genes shown in magenta in cluster 3 (n = 34) with the rest of the genes (n = 745). Note that for this cluster of 34 genes, siRNA depletion results in a significant decrease in apoptosis irrespective of paclitaxel treatment.

Close modal

Secondary validation identifies novel mechanisms of modulating microtubule stability and paclitaxel cytotoxicity

To investigate genetic specificity of siRNA pools, we used 4 individual siRNAs to knockdown each of 35 genes that enhanced paclitaxel effect on microtubules and also enhanced cytotoxicity, as detected by measuring cell viability or apoptosis. We measured the effect of knockdown on microtubule stability and viability with or without paclitaxel treatment in 384-well plates in triplicates. Thirty-four out of the 35 genes tested had at least one siRNA that resulted in a significant increase in microtubule stability. Twenty-nine genes were validated using at least 2 independent siRNAs (Fig. 5A and B, Table 1). Importantly, there was a strong positive correlation between baseline microtubule stability and paclitaxel-induced microtubule stability following siRNA transfections (correlation coefficient = 0.87, Fig. 5C). We next examined the correlation between paclitaxel-induced microtubule stabilization and paclitaxel-induced cell death and found a highly significant correlation (P < 0.001, Fig. 5D). Taken together, these results further validate the notion that baseline microtubule stability is a strong determinant of paclitaxel-mediated microtubule stabilization and identify previously unrecognized roles of a group of genes in modulating microtubule stability.

Figure 5.

Secondary validation identifies novel mechanisms of modulating microtubule stability and paclitaxel cytotoxicity. A, SKOv3 cells were transfected as described in Materials and Methods followed by treatment using diluent or paclitaxel as in the primary screen followed by immunofluorescence to measure microtubule stability. The bar plot shows the number of siRNAs per gene that significantly (P < 0.05) enhanced paclitaxel-induced microtubule stability. B, an example of increased percentage of paclitaxel-induced microtubule stability following depletion of PIK3C2A using 4 individual siRNAs compared with nontargeting siRNA-transfected (control), paclitaxel-treated cells. Shown is the mean + SD of triplicate measurements for each siRNA used. C, same experiment as in A. Shown is the correlation between baseline microtubule stability following siRNA transfection alone and microtubule stability following siRNA transfection and paclitaxel treatment. D, in parallel to the secondary validation described in (A), an additional secondary validation was carried out to measure cell viability 72 hours following paclitaxel treatment in a similar way to that used in the primary screen. Shown is the correlation between the percentage of microtubule stability and the percentage of cell viability following siRNA transfection and paclitaxel treatment.

Figure 5.

Secondary validation identifies novel mechanisms of modulating microtubule stability and paclitaxel cytotoxicity. A, SKOv3 cells were transfected as described in Materials and Methods followed by treatment using diluent or paclitaxel as in the primary screen followed by immunofluorescence to measure microtubule stability. The bar plot shows the number of siRNAs per gene that significantly (P < 0.05) enhanced paclitaxel-induced microtubule stability. B, an example of increased percentage of paclitaxel-induced microtubule stability following depletion of PIK3C2A using 4 individual siRNAs compared with nontargeting siRNA-transfected (control), paclitaxel-treated cells. Shown is the mean + SD of triplicate measurements for each siRNA used. C, same experiment as in A. Shown is the correlation between baseline microtubule stability following siRNA transfection alone and microtubule stability following siRNA transfection and paclitaxel treatment. D, in parallel to the secondary validation described in (A), an additional secondary validation was carried out to measure cell viability 72 hours following paclitaxel treatment in a similar way to that used in the primary screen. Shown is the correlation between the percentage of microtubule stability and the percentage of cell viability following siRNA transfection and paclitaxel treatment.

Close modal
Table 1.

Shown is the fold change in paclitaxel-induced microtubule stability following siRNA depletion of the indicated genes by 4 individual siRNAs compared with paclitaxel-induced microtubule stability following nontargeting siRNA transfection

siRNA number 1siRNA number 2siRNA number 3siRNA number 4
Gene symbolGene nameFold changePFold changePFold changePFold changeP
AATK Apoptosis-associated tyrosine kinase 1.65 0.025 1.09 0.025 0.92 0.973 1.42 0.008 
BMP2K BMP2 inducible kinase 1.53 0.030 1.01 0.356 1.07 0.173 1.45 0.021 
CDK5RAP3 CDK5 regulatory subunit associated protein 3 1.12 0.231 1.30 0.012 0.93 0.953 0.97 0.813 
CHEK1 CHK1 checkpoint homolog (S. pombe1.19 0.142 1.86 <0.001 1.45 0.006 1.60 0.018 
CHEK2 CHK2 checkpoint homolog (S. pombe1.58 0.054 1.21 0.104 0.97 0.819 1.15 0.066 
CHUK Conserved helix-loop-helix ubiquitous kinase 1.77 0.013 1.10 0.145 1.18 0.053 1.59 0.003 
COPB2 Coatomer protein complex, subunit β2 (βprime) 1.81 0.001 1.74 0.013 1.64 0.011 1.76 <0.001 
CSF1R Colony stimulating factor 1 receptor 1.67 0.023 1.34 0.033 1.17 0.080 1.12 0.068 
DAPK1 Death-associated protein kinase 1 1.19 0.102 2.13 0.004 1.54 0.020 1.17 0.074 
EDN2 Endothelin 2 0.86 0.986 1.58 0.005 1.15 0.006 1.21 0.122 
ERBB4 v-erb-a erythroblastic leukemia viral oncogene homolog 4 (avian) 1.72 0.016 1.31 0.018 1.95 0.035 1.12 0.132 
ERN1 Endoplasmic reticulum to nucleus signaling 1 1.04 0.264 1.12 0.112 1.96 0.018 1.35 0.014 
FER fer (fps/fes related) tyrosine kinase 1.28 0.008 1.52 0.027 1.32 0.001 0.96 0.693 
FLT4 fms-related tyrosine kinase 4 1.49 0.017 0.99 0.548 1.26 0.045 1.58 0.001 
GALK2 Galactokinase 2 1.14 0.017 1.22 0.078 2.01 0.008 2.43 0.008 
GRK4 G protein-coupled receptor kinase 4 1.66 0.012 2.00 0.026 1.27 0.028 1.19 0.095 
IKBKB Inhibitor of κ light polypeptide gene enhancer in B-cells, kinase beta 2.01 0.023 1.28 0.025 1.17 0.021 1.30 0.010 
ILK Integrin-linked kinase 1.58 0.005 1.11 0.148 1.80 0.001 1.38 0.001 
PANK4 Pantothenate kinase 4 1.02 0.363 1.02 0.399 1.52 0.004 2.42 0.001 
PDGFRB Platelet-derived growth factor receptor, β polypeptide 0.92 0.760 2.44 0.012 1.16 0.016 1.30 0.067 
PI4K2B Phosphatidylinositol 4-kinase type 2 β 1.18 0.076 1.49 0.002 1.33 <0.001 1.25 0.001 
PIK3C2A Phosphoinositide-3-kinase, class 2, α polypeptide 1.42 0.042 1.79 0.008 1.83 0.011 1.99 0.005 
PIK3R1 Phosphoinositide-3-kinase, regulatory subunit 1 (α) 1.07 0.264 1.12 0.151 1.55 0.016 1.07 0.208 
PKM2 Pyruvate kinase, muscle 1.09 0.190 1.35 0.001 1.25 0.073 1.27 0.051 
RAPGEF3 Rap guanine nucleotide exchange factor (GEF) 3 1.16 0.065 1.53 0.036 1.64 0.006 1.00 0.526 
RAPGEF4 Rap guanine nucleotide exchange factor (GEF) 4 0.98 0.653 1.03 0.310 1.11 0.013 1.22 0.001 
RFP Ring finger protein. Also known as tripartite motif containing 27 (TRIM27) 0.87 0.979 1.23 0.073 1.51 0.002 1.66 0.008 
SGK2 Serum/glucocorticoid regulated kinase 2 1.47 0.025 1.02 0.305 1.50 0.013 1.14 0.038 
SNF1LK SNF1-like kinase 1.30 0.001 1.04 0.271 1.08 0.114 1.00 0.524 
STK22B Serine/threonine kinase 22B. Also known as testis specific serine kinase 2 (TSSK2) 1.04 0.245 1.54 0.001 1.05 0.363 1.37 0.010 
STK24 Serine/threonine kinase 24 1.44 0.002 1.08 0.216 1.25 0.031 0.98 0.755 
TAF1L TAF1 RNA polymerase II, TATA box binding protein (TBP)-associated factor, 210 kDa-like 0.96 0.611 0.91 0.750 1.88 0.039 1.20 0.141 
TBK1 TANK-binding kinase 1 1.75 0.001 1.57 <0.001 1.44 0.003 1.28 0.023 
TESK2 Testis-specific kinase 2 1.20 0.093 1.32 0.021 1.38 0.002 0.87 0.994 
TLR4 Toll-like receptor 4 1.05 0.184 2.61 0.008 1.46 0.013 1.24 0.001 
siRNA number 1siRNA number 2siRNA number 3siRNA number 4
Gene symbolGene nameFold changePFold changePFold changePFold changeP
AATK Apoptosis-associated tyrosine kinase 1.65 0.025 1.09 0.025 0.92 0.973 1.42 0.008 
BMP2K BMP2 inducible kinase 1.53 0.030 1.01 0.356 1.07 0.173 1.45 0.021 
CDK5RAP3 CDK5 regulatory subunit associated protein 3 1.12 0.231 1.30 0.012 0.93 0.953 0.97 0.813 
CHEK1 CHK1 checkpoint homolog (S. pombe1.19 0.142 1.86 <0.001 1.45 0.006 1.60 0.018 
CHEK2 CHK2 checkpoint homolog (S. pombe1.58 0.054 1.21 0.104 0.97 0.819 1.15 0.066 
CHUK Conserved helix-loop-helix ubiquitous kinase 1.77 0.013 1.10 0.145 1.18 0.053 1.59 0.003 
COPB2 Coatomer protein complex, subunit β2 (βprime) 1.81 0.001 1.74 0.013 1.64 0.011 1.76 <0.001 
CSF1R Colony stimulating factor 1 receptor 1.67 0.023 1.34 0.033 1.17 0.080 1.12 0.068 
DAPK1 Death-associated protein kinase 1 1.19 0.102 2.13 0.004 1.54 0.020 1.17 0.074 
EDN2 Endothelin 2 0.86 0.986 1.58 0.005 1.15 0.006 1.21 0.122 
ERBB4 v-erb-a erythroblastic leukemia viral oncogene homolog 4 (avian) 1.72 0.016 1.31 0.018 1.95 0.035 1.12 0.132 
ERN1 Endoplasmic reticulum to nucleus signaling 1 1.04 0.264 1.12 0.112 1.96 0.018 1.35 0.014 
FER fer (fps/fes related) tyrosine kinase 1.28 0.008 1.52 0.027 1.32 0.001 0.96 0.693 
FLT4 fms-related tyrosine kinase 4 1.49 0.017 0.99 0.548 1.26 0.045 1.58 0.001 
GALK2 Galactokinase 2 1.14 0.017 1.22 0.078 2.01 0.008 2.43 0.008 
GRK4 G protein-coupled receptor kinase 4 1.66 0.012 2.00 0.026 1.27 0.028 1.19 0.095 
IKBKB Inhibitor of κ light polypeptide gene enhancer in B-cells, kinase beta 2.01 0.023 1.28 0.025 1.17 0.021 1.30 0.010 
ILK Integrin-linked kinase 1.58 0.005 1.11 0.148 1.80 0.001 1.38 0.001 
PANK4 Pantothenate kinase 4 1.02 0.363 1.02 0.399 1.52 0.004 2.42 0.001 
PDGFRB Platelet-derived growth factor receptor, β polypeptide 0.92 0.760 2.44 0.012 1.16 0.016 1.30 0.067 
PI4K2B Phosphatidylinositol 4-kinase type 2 β 1.18 0.076 1.49 0.002 1.33 <0.001 1.25 0.001 
PIK3C2A Phosphoinositide-3-kinase, class 2, α polypeptide 1.42 0.042 1.79 0.008 1.83 0.011 1.99 0.005 
PIK3R1 Phosphoinositide-3-kinase, regulatory subunit 1 (α) 1.07 0.264 1.12 0.151 1.55 0.016 1.07 0.208 
PKM2 Pyruvate kinase, muscle 1.09 0.190 1.35 0.001 1.25 0.073 1.27 0.051 
RAPGEF3 Rap guanine nucleotide exchange factor (GEF) 3 1.16 0.065 1.53 0.036 1.64 0.006 1.00 0.526 
RAPGEF4 Rap guanine nucleotide exchange factor (GEF) 4 0.98 0.653 1.03 0.310 1.11 0.013 1.22 0.001 
RFP Ring finger protein. Also known as tripartite motif containing 27 (TRIM27) 0.87 0.979 1.23 0.073 1.51 0.002 1.66 0.008 
SGK2 Serum/glucocorticoid regulated kinase 2 1.47 0.025 1.02 0.305 1.50 0.013 1.14 0.038 
SNF1LK SNF1-like kinase 1.30 0.001 1.04 0.271 1.08 0.114 1.00 0.524 
STK22B Serine/threonine kinase 22B. Also known as testis specific serine kinase 2 (TSSK2) 1.04 0.245 1.54 0.001 1.05 0.363 1.37 0.010 
STK24 Serine/threonine kinase 24 1.44 0.002 1.08 0.216 1.25 0.031 0.98 0.755 
TAF1L TAF1 RNA polymerase II, TATA box binding protein (TBP)-associated factor, 210 kDa-like 0.96 0.611 0.91 0.750 1.88 0.039 1.20 0.141 
TBK1 TANK-binding kinase 1 1.75 0.001 1.57 <0.001 1.44 0.003 1.28 0.023 
TESK2 Testis-specific kinase 2 1.20 0.093 1.32 0.021 1.38 0.002 0.87 0.994 
TLR4 Toll-like receptor 4 1.05 0.184 2.61 0.008 1.46 0.013 1.24 0.001 

NOTE: P value is the result of 1-sided t test comparison of triplicate values of paclitaxel-induced microtubule stability following siRNA depletion of an individual gene compared with triplicate values of paclitaxel-induced microtubule stability following transfection using nontargeting control siRNA.

Modulating microtubule stability directly influences microtubule mass and cytotoxicity in ovarian cancer cell lines

Paclitaxel induces a significant increase in microtubule stability and in total microtubule mass resulting in the induction of the characteristic feature of paclitaxel-induced bundles (36) which perturbs the ability of the cell to form an organized mitotic spindle. To investigate whether enhanced paclitaxel-induced microtubule stability also resulted in an enhanced paclitaxel-induced microtubule polymerization, we knocked down 7 genes (ILK, FER, RAPGEF4, AATK, CHEK1, FLT4, and SGK2) using siRNA pools and measured the effect of knockdown on microtubule stability and total microtubule mass following paclitaxel treatment at 10 and 100 nmol/L using both α- and Glu-tubulin staining. As expected, knockdown of each of the 7 genes enhanced Glu-tubulin formation following paclitaxel treatment at each of the 2 concentrations used (Fig. 6A). In addition, gene depletion also resulted in an increase in microtubule mass for at least one of the concentrations tested for each of the 7 genes (Fig. 6B). Depletion of these genes resulted in enhanced paclitaxel-induced loss of cell viability in short-term assays and a significant decrease in the ability of cancer cells to form colonies in long-term assays in SKOv3 and Hey ovarian cancer cell lines (Fig. 6C and D). Similar results were obtained following depletion of RAPGEF4 in the paclitaxel-resistant A549-T12 lung carcinoma cells that are known to have microtubule level of resistance to paclitaxel (12). We first confirmed that paclitaxel-induced microtubule stabilization was significantly reduced in A549-T12 cells compared with the parent paclitaxel-sensitive A549 cells (Supplementary Fig. S8). RAPGEF4-specific siRNA treatment of the A549-T12 cells resulted in a significant increase in paclitaxel microtubule stabilization and partially restored paclitaxel sensitivity (Supplementary Fig. S8). These results confirm the strong mechanistic effect of modulating microtubule stability on microtubule polymer mass and paclitaxel-mediated cytotoxicity.

Figure 6.

Modulating microtubule stability directly influences microtubule mass and cytotoxicity in ovarian cancer cell lines. A and B, SKOv3 cells were transfected using the indicated siRNAs for 48 hours, then treated with paclitaxel for 1 hour using the indicated concentrations, then fixed and probed using anti–Glu-tubulin counterstained with alexa-594 (A) or anti–α-tubulin counterstained with alexa-488 (B). Images were collected and segmented as previously described. Shown is the mean background (BG)-subtracted intensity ± SEM for each stain. C and D, cells were transfected using either nontargeting siRNA or siRNAs targeting ILK (C) or RAPGEF4 (D) for 48 hours, then treated with paclitaxel 1 nmol/L for 48 hours, and cells were maintained for either 4 days (Hey cells) or 7 days (SKOv3 cells), then fixed and stained using Coomassie Blue, and individual colonies were counted. Shown is the mean number of colonies ± SD of counts from 2 independent experiments.

Figure 6.

Modulating microtubule stability directly influences microtubule mass and cytotoxicity in ovarian cancer cell lines. A and B, SKOv3 cells were transfected using the indicated siRNAs for 48 hours, then treated with paclitaxel for 1 hour using the indicated concentrations, then fixed and probed using anti–Glu-tubulin counterstained with alexa-594 (A) or anti–α-tubulin counterstained with alexa-488 (B). Images were collected and segmented as previously described. Shown is the mean background (BG)-subtracted intensity ± SEM for each stain. C and D, cells were transfected using either nontargeting siRNA or siRNAs targeting ILK (C) or RAPGEF4 (D) for 48 hours, then treated with paclitaxel 1 nmol/L for 48 hours, and cells were maintained for either 4 days (Hey cells) or 7 days (SKOv3 cells), then fixed and stained using Coomassie Blue, and individual colonies were counted. Shown is the mean number of colonies ± SD of counts from 2 independent experiments.

Close modal

Pharmacologic approaches for overcoming resistance to chemotherapy in general and to antimitotic agents in particular have often focused on enhancing general mechanisms of cytotoxicity such as apoptosis (37–39), rather than on modulating drug-specific targets. Methods used to enhance cytotoxicity often lack specificity and such approaches, while important, are likely to induce unwanted off-target effects. In spite of the detailed knowledge of the mechanism of action of paclitaxel, identifying mechanism-specific methods for modulating its cytotoxic effect via modulating microtubules or mitosis has remained elusive. This work confirms and generalizes the concept that modulating microtubule stability in cancer cells is an effective method for enhancing paclitaxel cytotoxicity. This study has also discovered several targets that could be exploited for therapeutic benefit.

In contrast to several previously conducted siRNA screens that have focused on measuring a single general endpoint in multiple cell lines, we have conducted a comprehensive analysis of the effect of siRNA depletion on multiple related but independent outcomes, focusing on the paclitaxel-specific endpoint of microtubule stability as well as the more general parameters of cell viability and apoptosis. This approach revealed a previously underestimated level of complexity in regulating cell fate following paclitaxel treatment. Discordant results in the 3 assays after knockdown of certain groups of kinases indicate that promoting cell death through one mechanism does not necessarily predict outcomes for other seemingly parallel mechanisms of cell death. This may, in part, explain the inconsistency in the results obtained between different studies when evaluating a novel target for therapy.

Our work identified several gene clusters that influence paclitaxel response in strikingly different ways. More than a third of the genes that sensitized cells to paclitaxel-induced apoptosis and nearly half of the genes that sensitized cells to paclitaxel-mediated cell death also enhanced paclitaxel-induced microtubule stability (P = 0.03 and <0.001, respectively). This analysis confirmed that enhancing microtubule stability following paclitaxel treatment is strongly associated with enhanced paclitaxel-induced cell death. Although it is intuitive to hypothesize that enhancing microtubule stability following paclitaxel treatment should be followed by higher apoptosis and loss of cell viability, not all siRNAs that increased microtubule stability enhanced sensitivity to paclitaxel in this screen. An important limitation of short-term assays as those conducted in this screen, is the undue sensitivity of the assay to cell population kinetics. As paclitaxel-induced apoptosis is dependent on passage through mitosis, decreasing cell population doubling time due to suboptimal transfection conditions may result in false negative results. We have, therefore, extensively optimized assay development conditions prior to the screen to avoid this potential pitfall. However, the total elimination of this possibility is difficult to confirm.

How microtubule stabilization results in enhanced paclitaxel-induced cytotoxicity is likely to be multifactorial. However, our findings suggest that induction of apoptosis following increased mitotic arrest is likely to be the predominant mechanism. Consistent with previous findings (40), our work reveals the importance of the effect of modulating apoptosis on cell fate such that a negative effect on apoptosis may negate a therapeutic advantage of enhancing microtubule stability. In contrast to targeting the intrinsically promiscuous kinases that share and phosphorylate many substrates, targeting microtubule-specific molecular targets may result in a tighter concordance between the magnitude of microtubule stability and cytotoxicity.

Another possibility for the effect of enhancing microtubule stability on increasing paclitaxel-induced cell death is interfering with vesicular transport of key survival and transcription factors to the nucleus. To investigate this hypothesis, we tested whether microtubule stabilization or depolymerization decreased the nuclear transport of a constitutively activated transcription factor in SKOv3 cells, STAT3 (41, 42). This revealed that microtubule depolymerization using Nocodazole resulted in a significant decrease in nuclear accumulation of phospho-STAT3, in contrast to paclitaxel used at concentrations up to 100 nmol/L (Supplementary Fig. S9). Whether depletion of motor proteins involved in nuclear transport of transcription factors may result in sensitization to paclitaxel by targeting several transcription factors at the same time will remain to be tested.

The identification of molecular predictive markers is required for personalization of therapy (43). A significant obstacle to identifying clinically relevant predictive markers for taxane response is that, in clinical practice, most tumors are treated by a combination of paclitaxel and other chemotherapies (3, 4). Therefore, using standard endpoints for measuring response, such as radiological monitoring of tumor size or measuring the level of tumor markers, cannot distinguish the effect of individual drugs. Similarly, it is not possible to use early indicators of cytotoxicity, such as apoptosis markers, as endpoints in such studies. In this work, we showed that baseline microtubule stability is an important determinant of paclitaxel-induced microtubule stabilization and cytotoxicity. This presents a testable hypothesis that microtubule stability could be measured either as a predictive marker of paclitaxel response or as a surrogate endpoint for paclitaxel response that could be utilized to identify clinically relevant predictive molecular biomarkers. Measuring the magnitude of microtubule stability at presentation is feasible by measuring Glu-tubulin levels in tumors (44). In addition to measuring baseline microtubule stability, it is possible to hypothesize that a certain threshold of microtubule stability is required for paclitaxel to exert its cytotoxic effect in patients. Measuring paclitaxel-induced microtubule stability in mouse models and in specifically designed early-phase clinical trials will help in testing the relevance of this method in predicting paclitaxel response.

Mechanisms regulating microtubule dynamics are most probably shared between cancer and noncancer cells. For example, depletion of the microtubule-associated protein Tau, which is thought to compete with paclitaxel for binding sites in β-tubulin, in human skin fibroblasts grown in culture results in a significant increase in paclitaxel-induced microtubule stabilization (Supplementary Fig. S10). However, we have shown that paclitaxel-induced apoptosis is dependent on cell cycling (Supplementary Fig. S4). As many nontumor cell types are noncycling, the concept of specific targeting of microtubules to enhance paclitaxel-induced stabilization ensures the minimization of toxicity. Importantly, our study identified enhancers of microtubule stabilization following the administration of substoichiometric concentrations of paclitaxel (8 nmol/L) to mimic the concentrations that would be expected in the vicinity of ovarian cancers in patients. In contrast to poorly vascularized cancer tissue, nontumor cycling cells such as those in bone marrow are probably exposed to much higher concentrations of paclitaxel (above 50 nmol/L) that result in near stoichiometry and full microtubule stabilization (45, 46). Therefore, we hypothesize that enhancers of paclitaxel-induced microtubule stability may only increase cell death in cancer cells and not normal bone marrow cells as the former is exposed to much lower doses of paclitaxel. Future experiments in animal models of cancer will address this hypothesis. In addition, prioritization of targets for testing in preclinical and clinical models will depend on the pattern of expression and activity of the identified targets in ovarian and other cancers and the availability of inhibitors for the identified targets. In those instances where kinases are overexpressed or activated, selective enhancement of microtubule stability might be achieved within cancer cells, increasing rates of response. In contrast to many earlier studies, the goal would be to enhance primary sensitivity, rather than reverse mechanisms of acquired resistance.

In summary, this work shows that modulation of baseline microtubule stability is a key determinant for paclitaxel-induced microtubule stabilization and cytotoxicity. This finding has important predictive and therapeutic implications that could be tested in preclinical and clinical models of paclitaxel therapy.

No potential conflicts of interest were disclosed.

We would like to thank Dr. Gordon Mills and Catherine Nguyen for technical help.

A.A. Ahmed was funded by Cancer Research UK, the University of Oxford, and the Camilla Samuel Fund. A.A. Ahmed and all other authors were funded by the Zarrow Foundation, Mr. and Mrs. Stuart Zarrow, the Ovarian Cancer Research Fund Program Project Development Grant, the National Foundation for Cancer Research, and the University of Texas MD Anderson Cancer Center Ovarian Cancer Specialized Program of Research Excellence (P50 CA083639).

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.

1.
Schiff
PB
,
Horwitz
SB
. 
Taxol assembles tubulin in the absence of exogenous guanosine 5′-triphosphate or microtubule-associated proteins
.
Biochemistry
1981
;
20
:
3247
52
.
2.
Schiff
PB
,
Horwitz
SB
. 
Taxol stabilizes microtubules in mouse fibroblast cells
.
Proc Natl Acad Sci U S A
1980
;
77
:
1561
5
.
3.
Martin
M
,
Rodriguez-Lescure
A
,
Ruiz
A
,
Alba
E
,
Calvo
L
,
Ruiz-Borrego
M
, et al
Randomized phase 3 trial of fluorouracil, epirubicin, and cyclophosphamide alone or followed by Paclitaxel for early breast cancer
.
J Natl Cancer Inst
2008
;
100
:
805
14
.
4.
McGuire
WP
,
Hoskins
WJ
,
Brady
MF
,
Kucera
PR
,
Partridge
EE
,
Look
KY
, et al
Cyclophosphamide and cisplatin versus paclitaxel and cisplatin: a phase III randomized trial in patients with suboptimal stage III/IV ovarian cancer (from the Gynecologic Oncology Group)
.
Semin Oncol
1996
;
23
:
40
7
.
5.
Nogales
E
,
Wolf
SG
,
Khan
IA
,
Luduena
RF
,
Downing
KH
. 
Structure of tubulin at 6.5 A and location of the taxol-binding site
.
Nature
1995
;
375
:
424
7
.
6.
Horwitz
SB
. 
Taxol (paclitaxel): mechanisms of action
.
Ann Oncol
1994
;
5 Suppl
6
:
S3
6
.
7.
Parness
J
,
Horwitz
SB
. 
Taxol binds to polymerized tubulin in vitro
.
J Cell Biol
1981
;
91
:
479
87
.
8.
Schibler
MJ
,
Cabral
F
. 
Taxol-dependent mutants of Chinese hamster ovary cells with alterations in alpha- and beta-tubulin
.
J Cell Biol
1986
;
102
:
1522
31
.
9.
Cabral
F
,
Abraham
I
,
Gottesman
MM
. 
Isolation of a taxol-resistant Chinese hamster ovary cell mutant that has an alteration in alpha-tubulin
.
Proc Natl Acad Sci U S A
1981
;
78
:
4388
91
.
10.
McCarroll
JA
,
Gan
PP
,
Liu
M
,
Kavallaris
M
. 
BetaIII-tubulin is a multifunctional protein involved in drug sensitivity and tumorigenesis in non-small cell lung cancer
.
Cancer Res
2010
;
70
:
4995
5003
.
11.
Kavallaris
M
,
Kuo
DY
,
Burkhart
CA
,
Regl
DL
,
Norris
MD
,
Haber
M
, et al
Taxol-resistant epithelial ovarian tumors are associated with altered expression of specific beta-tubulin isotypes
.
J Clin Invest
1997
;
100
:
1282
93
.
12.
Martello
LA
,
Verdier-Pinard
P
,
Shen
HJ
,
He
L
,
Torres
K
,
Orr
GA
, et al
Elevated levels of microtubule destabilizing factors in a Taxol-resistant/dependent A549 cell line with an alpha-tubulin mutation
.
Cancer Res
2003
;
63
:
1207
13
.
13.
Ahmed
AA
,
Mills
AD
,
Ibrahim
AE
,
Temple
J
,
Blenkiron
C
,
Vias
M
, et al
The extracellular matrix protein TGFBI induces microtubule stabilization and sensitizes ovarian cancers to paclitaxel
.
Cancer Cell
2007
;
12
:
514
27
.
14.
Ahmed
AA
,
Lu
Z
,
Jennings
NB
,
Etemadmoghadam
D
,
Capalbo
L
,
Jacamo
RO
, et al
SIK2 is a centrosome kinase required for bipolar mitotic spindle formation that provides a potential target for therapy in ovarian cancer
.
Cancer Cell
2010
;
18
:
109
21
.
15.
Broom
BM
,
Sulman
EP
,
Do
KA
,
Edgerton
ME
. 
Bagged gene shaving for the robust clustering of high-throughput data
.
Int J Bioinform Res Appl
2010
;
6
:
326
43
.
16.
Chen
PW
,
Lin
SJ
,
Tsai
SC
,
Lin
JH
,
Chen
MR
,
Wang
JT
, et al
Regulation of microtubule dynamics through phosphorylation on stathmin by Epstein-Barr virus kinase BGLF4
.
J Biol Chem
2010
;
285
:
10053
63
.
17.
Manna
T
,
Thrower
DA
,
Honnappa
S
,
Steinmetz
MO
,
Wilson
L
. 
Regulation of microtubule dynamic instability in vitro by differentially phosphorylated stathmin
.
J Biol Chem
2009
;
284
:
15640
9
.
18.
Kojima
Y
,
Miyoshi
H
,
Clevers
HC
,
Oshima
M
,
Aoki
M
,
Taketo
MM
. 
Suppression of tubulin polymerization by the LKB1-microtubule-associated protein/microtubule affinity-regulating kinase signaling
.
J Biol Chem
2007
;
282
:
23532
40
.
19.
Wang
JZ
,
Grundke-Iqbal
I
,
Iqbal
K
. 
Kinases and phosphatases and tau sites involved in Alzheimer neurofibrillary degeneration
.
Eur J Neurosci
2007
;
25
:
59
68
.
20.
Yamamoto
H
,
Yamauchi
E
,
Taniguchi
H
,
Ono
T
,
Miyamoto
E
. 
Phosphorylation of microtubule-associated protein tau by Ca2+/calmodulin-dependent protein kinase II in its tubulin binding sites
.
Arch Biochem Biophys
2002
;
408
:
255
62
.
21.
Drewes
G
,
Ebneth
A
,
Mandelkow
EM
. 
MAPs, MARKs and microtubule dynamics
.
Trends Biochem Sci
1998
;
23
:
307
11
.
22.
Gundersen
GG
,
Bulinski
JC
. 
Distribution of tyrosinated and nontyrosinated alpha-tubulin during mitosis
.
J Cell Biol
1986
;
102
:
1118
26
.
23.
Gundersen
GG
,
Kalnoski
MH
,
Bulinski
JC
. 
Distinct populations of microtubules: tyrosinated and nontyrosinated alpha tubulin are distributed differently in vivo
.
Cell
1984
;
38
:
779
89
.
24.
Amos
LA
,
Schlieper
D
. 
Microtubules and maps
.
Adv Protein Chem
2005
;
71
:
257
98
.
25.
Rouzier
R
,
Rajan
R
,
Wagner
P
,
Hess
KR
,
Gold
DL
,
Stec
J
, et al
Microtubule-associated protein tau: a marker of paclitaxel sensitivity in breast cancer
.
Proc Natl Acad Sci U S A
2005
;
102
:
8315
20
.
26.
Kar
S
,
Fan
J
,
Smith
MJ
,
Goedert
M
,
Amos
LA
. 
Repeat motifs of tau bind to the insides of microtubules in the absence of taxol
.
EMBO J
2003
;
22
:
70
7
.
27.
Ahmed
AA
,
Vias
M
,
Iyer
NG
,
Caldas
C
,
Brenton
JD
. 
Microarray segmentation methods significantly influence data precision
.
Nucleic Acids Res
2004
;
32
:
e50
.
28.
Schmid
JA
,
Birbach
A
. 
IkappaB kinase beta (IKKbeta/IKK2/IKBKB)–a key molecule in signaling to the transcription factor NF-kappaB
.
Cytokine Growth Factor Rev
2008
;
19
:
157
65
.
29.
Hacker
H
,
Karin
M
. 
Regulation and function of IKK and IKK-related kinases
.
Sci STKE
2006
;
2006:re13
.
30.
Ganta
S
,
Amiji
M
. 
Coadministration of Paclitaxel and curcumin in nanoemulsion formulations to overcome multidrug resistance in tumor cells
.
Mol Pharm
2009
;
6
:
928
39
.
31.
Annunziata
CM
,
Stavnes
HT
,
Kleinberg
L
,
Berner
A
,
Hernandez
LF
,
Birrer
MJ
, et al
Nuclear factor kappaB transcription factors are coexpressed and convey a poor outcome in ovarian cancer
.
Cancer
2010
;
116
:
3276
84
.
32.
Zhang
D
,
Qiu
L
,
Jin
X
,
Guo
Z
,
Guo
C
. 
Nuclear factor-kappaB inhibition by parthenolide potentiates the efficacy of Taxol in non-small cell lung cancer in vitro and in vivo
.
Mol Cancer Res
2009
;
7
:
1139
49
.
33.
Jin
X
,
Qiu
L
,
Zhang
D
,
Zhang
M
,
Wang
Z
,
Guo
Z
, et al
Chemosensitization in non-small cell lung cancer cells by IKK inhibitor occurs via NF-kappaB and mitochondrial cytochrome c cascade
.
J Cell Mol Med
2009
;
13
:
4596
607
.
34.
Krieg
A
,
Le Negrate
G
,
Reed
JC
. 
RIP2-beta: a novel alternative mRNA splice variant of the receptor interacting protein kinase RIP2
.
Mol Immunol
2009
;
46
:
1163
70
.
35.
Paulsen
RD
,
Cimprich
KA
. 
The ATR pathway: fine-tuning the fork
.
DNA Repair (Amst)
2007
;
6
:
953
66
.
36.
Manfredi
JJ
,
Parness
J
,
Horwitz
SB
. 
Taxol binds to cellular microtubules
.
J Cell Biol
1982
;
94
:
688
96
.
37.
Gimenez-Bonafe
P
,
Tortosa
A
,
Perez-Tomas
R
. 
Overcoming drug resistance by enhancing apoptosis of tumor cells
.
Curr Cancer Drug Targets
2009
;
9
:
320
40
.
38.
Wilson
TR
,
Johnston
PG
,
Longley
DB
. 
Anti-apoptotic mechanisms of drug resistance in cancer
.
Curr Cancer Drug Targets
2009
;
9
:
307
19
.
39.
Ashkenazi
A
. 
Directing cancer cells to self-destruct with pro-apoptotic receptor agonists
.
Nat Rev Drug Discov
2008
;
7
:
1001
12
.
40.
Gascoigne
KE
,
Taylor
SS
. 
Cancer cells display profound intra- and interline variation following prolonged exposure to antimitotic drugs
.
Cancer Cell
2008
;
14
:
111
22
.
41.
Rosen
DG
,
Mercado-Uribe
I
,
Yang
G
,
Bast
RC
 Jr.
,
Amin
HM
,
Lai
R
, et al
The role of constitutively active signal transducer and activator of transcription 3 in ovarian tumorigenesis and prognosis
.
Cancer
2006
;
107
:
2730
40
.
42.
Huang
S
,
Chang
IS
,
Lin
W
,
Ye
W
,
Luo
RZ
,
Lu
Z
, et al
ARHI (DIRAS3), an imprinted tumour suppressor gene, binds to importins and blocks nuclear import of cargo proteins
.
Biosci Rep
2010
;
30
:
159
68
.
43.
McDermott
U
,
Settleman
J
. 
Personalized cancer therapy with selective kinase inhibitors: an emerging paradigm in medical oncology
.
J Clin Oncol
2009
;
27
:
5650
9
.
44.
Zhuang
SH
,
Hung
YE
,
Hung
L
,
Robey
RW
,
Sackett
DL
,
Linehan
WM
, et al
Evidence for microtubule target engagement in tumors of patients receiving ixabepilone
.
Clin Cancer Res
2007
;
13
:
7480
6
.
45.
Gianni
L
,
Kearns
CM
,
Giani
A
,
Capri
G
,
Vigano
L
,
Lacatelli
A
, et al
Nonlinear pharmacokinetics and metabolism of paclitaxel and its pharmacokinetic/pharmacodynamic relationships in humans
.
J Clin Oncol
1995
;
13
:
180
90
.
46.
Marchetti
P
,
Urien
S
,
Cappellini
GA
,
Ronzino
G
,
Ficorella
C
. 
Weekly administration of paclitaxel: theoretical and clinical basis
.
Crit Rev Oncol Hematol
2002
;
44
Suppl
:
S3
13
.