Proliferation, migration and invasion of human glioma cells exposed to paclitaxel (Taxol) in vitro.

Paclitaxel (Taxol), an anti-cancer drug derived from Taxus species, was tested for its anti-migrational, anti-invasive and anti-proliferative effect on two human glioma cell lines (GaMg and D-54Mg) grown as multicellular tumour spheroids. In addition, the direct effect of paclitaxel on glioma cells was studied using flow cytometry and scanning confocal microscopy. Both cell lines showed a dose-dependent growth and migratory response to paclitaxel. The GaMg cells were found to be 5-10 times more sensitive to paclitaxel than D-54Mg cells. Paclitaxel also proved to be remarkably effective in preventing invasion in a co-culture system in which tumour spheroids were confronted with fetal rat brain cell aggregates. Control experiments with Cremophor EL (the solvent of paclitaxel for clinical use) in this study showed no effect on tumour cell migration, cell proliferation or cell invasion. Scanning confocal microscopy of both cell lines showed an extensive random organization of the microtubules in the cytoplasm. After paclitaxel exposure, the GaMg and the D-54Mg cells exhibited a fragmentation of the nuclear material, indicating a possible induction of apoptosis. In line with this, flow cytometric DNA histograms showed an accumulation of cells in the G2/M phase of the cell cycle after 24 h of paclitaxel exposure. After 48 h, a deterioration of the DNA histograms was observed indicating nuclear fragmentation. ImagesFigure 3Figure 6

taxel seems to block mitosis mainly by stabilizing spindle microtubules, while abnormal binding of microtubule polymers is seen at higher concentrations as a result of paclitaxel's stabilizing and promoting characteristics (Jordan et al, 1993). This is in contrast to agents, such as colchicine and vinblastine, that inhibit microtubule assembly. Tissue culture studies have shown that the cell kinetic effects of paclitaxel result in an increase in the fraction of cells in the G2 or M phase of the cell cycle (Schiff and Horwitz, 1980). Paclitaxel has been shown to be a potent cytotoxic agent against a range of human malignant cell types using both cell culture and animal xenograft models (Hruban et al, 1989;Rowinsky et al, 1990). Clinical trials have demonstrated that paclitaxel is an active agent in the treatment of many cancers, e.g. epithelial and ovarian cancers (McGuire et al, 1989), breast cancers (Holmes et al, 1991) and lung cancers (Chang et al, 1993).
The in vitro sensitivity of cancer cells to chemotherapeutic agents including paclitaxel is usually determined by growth inhibition and clonogenic assays. However, it is well known that cells growing under different environmental conditions may respond differently to specific therapeutic procedures. Factors such as nutrient supply, oxygen tension, pH and other microenvironmental conditions may affect the cellular sensitivity to paclitaxel (Carlsson et al, 1983;Mueller-Klieser, 1987). The three-dimensional arrangement of multicellular spheroids provides a cellular micromilieu that includes different environmental conditions in different parts of the spheroid, and it has been shown that spheroids from both normal and malignant tissues maintain several biochemical and morphological characteristics similar to those present in the corresponding tissue in vivo (Sutherland et al, 1986;Sutherland, 1988). Thus, testing paclitaxel on spheroids may provide additional information concerning paclitaxel effects in vivo.
The present study describes the effects of paclitaxel and its solvent (Cremophor EL) alone on glioma spheroids in vitro. Our model is suitable for exploring the effect of the drug on growth, directional cell migration and invasion. For this purpose, spheroids, from two permanent human glioma cell lines  were confronted with normal brain cell aggregates in an in vitro co-culture assay. In addition, the direct effect of paclitaxel on glioma cells was studied using flow cytometry and scanning confocal microscopy.
MATERIAL AND METHODS Drug Paclitaxel (Taxol) (Bristol-Myers Squibb Co., Princeton, NJ, USA) was stored at 4°C. Paclitaxel was prepared 1 h before use. The solution was diluted in complete Dulbecco's modified Eagle medium (DMEM) (see below) to a final concentration of 0.001-0.1 jIg ml-'.
Cremophor EL (Sigma Chemicals, St Louis, MO, USA) was stored at 4°C. It was prepared 1 h before use and the solution was diluted in complete growth medium to a final concentration of 0.01-0.1 jIg ml-'. To study the direct effect of paclitaxel on the cell cultures, crystalline paclitaxel was used. Crystalline paclitaxel (lyophilized powder) was prepared 1 h before use by dissolving it in ethanol (96%). The solution was further diluted in complete DMEM to a final concentration of 0.001-0.1 jg ml-'. The direct effect of paclitaxel was then studied using flow cytometry and scanning confocal microscopy (see below for details).
Cell lines and cell culture conditions Multicellular tumour spheroids from two human glioma cell lines were used. The human D-54Mg cell line was kindly supplied by Dr DD Bigner, Duke University Medical Center, Durham, NC, USA (Bigner et al, 1981). The GaMg was obtained from a 42year-old woman and histologically identified as a glioblastoma (Akslen et al, 1988).

Monolayer growth
For both cell lines, 104 cells in 2 ml of complete DMEM were seeded into 3.5-cm dishes (Nunc). The cells were then prepared for scanning confocal microscopy and for flow cytometry experiments.

Tumour spheroids
Tumour spheroids were initiated by the agar overlay culture method described by Yuhas et al (1977). Briefly, spheroids were formed by seeding 5 x 106 cells in 20 ml of complete DMEM into 80-cm2 agar-coated tissue culture flasks (Nunc, Roskilde, Denmark). After 10 days in culture, spheroids with diameters between 200 and 250 jm were selected for further experiments.
Tumour cell migration Spheroids from both GaMg and D-54Mg cell lines were placed individually into uncoated 16-mm multiwell dishes (Nunc) that were filled with 1 ml of complete DMEM. The spheroids were then continuously exposed to varying concentrations of paclitaxel (0.001-0.1 jig ml-'). There were 12 spheroids in each treatment group. One group was used as control (grown in complete DMEM without drugs). The spheroids plated within 2 h, and the cellular outgrowth from one spheroid was defined as a colony. Two orthogonal colony diameters were measured daily by phasecontrast microscopy over a 4-day period (96 h), and the migratory capacities of the glioma cells were then determined by calculating the colony areas from the diameter measurements. The experiments were done in triplicate.

Spheroid growth
Both the GaMg cells and the D-54Mg cells were continuously exposed to varying concentrations of paclitaxel (0.001-0.1 jig ml-'). For each experiment, 12 spheroids were transferred individually into 16-mm multiwell dishes (Nunc). The dishes were base coated with 0.5 ml of 0.75% DMEM agar and filled with 1 ml of complete DMEM. One group was used as control. After treatment, the dishes were incubated at 37°C. The diameters of the spheroids were measured daily in a phase-contrast microscope over a 15-day period and the spheroid volume was calculated. The experiments were done in triplicate.

Brain aggregates
Fetal rat brain cells were obtained from 18-day-old fetuses of inbred Wistar rats. The brains were dissected out under aseptic conditions and the meningeal coverings were removed. The brains were placed in sterile Petri dishes containing calcium-and magnesium-free phosphate-buffered saline (PBS; Sigma). The tissue was cut into small pieces, washed in PBS and dissociated by serial trypsination (0.025% trypsin; Whittaker Bioproducts, Walkersville, MD, USA) into a single cell suspension. The brain cell aggregates were produced by seeding 6 x 106 cells in 1 ml of growth medium into 16-mm multiwell dishes (Nunc), base coated with 0.5 ml of a non-adherent 0.75% medium agar substrate. During a 4-day period, immature brain cell aggregates were formed. After 20 days in tissue culture, the cellular differentiation in the aggregates was complete, resulting in a defined three-layered structure containing mature astrocytes, oligodendrocytes and neurons with myelinated axons and synapses present in a well-developed neuropil. Thus, the aggregates show a close resemblance to differentiated brain tissue in vivo (Trapp et al, 1982;Bjerkvig, 1986;Bjerkvig et al, 1986).

Tumour cell invasion
Three days before confrontation with brain aggregates, the GaMg and the D-54Mg tumour spheroids were continuously exposed to 0.01 and 0.04 jig ml-' paclitaxel respectively. This was done to obtain an effective growth inhibition, according to the protocol described above for spheroid growth. There were five parallels in each of the experiments. The spheroids were then transferred to 96-well multiwell dishes with a sterile Pasteur pipette and confronted individually with the brain aggregates. By using a sterile syringe and a stereomicroscope, the tumour spheroids and British Journal of Cancer (1997) 75(12), 1744-1752 the brain aggregates were placed in close contact with each other. The co-cultures were incubated for a 96-h period at 37°C and then fixed for light microscopic examination and morphometric analyses. The reduction of brain aggregate volume as a percentage of the initial brain volume before confrontation was used to quantify the invasive process (see below). The experiments were done in duplicate.

Light microscopy
The co-cultures were fixed in 2% glutaraldehyde in 0.1 M sucroseadjusted sodium cacodylate buffer (300 ± 25 mOsm). After 24 h, the specimens were washed in the same buffer without glutaraldehyde and post-fixed for 1 h in 1% osmium tetroxide before serial dehydration in increasing gradients of ethanol up to 96%. Embedding of the co-cultures in Epon 812 was performed using graded mixtures of epon propylenoxide. The specimens were polymerized for 48 h at 60°C. The co-cultures were then sectioned as described below.

Morphometry
Cavalieri's principle for direct estimation of volume from systematically sampled sections was used to determine the amount of brain tissue remaining after co-culture with tumour cell spheroids. Serial semi-thin sections (1.5 jum) were prepared from the specimens using a Reichert-Jung Microtome 2040 (Vienna, Austria). Every 15th section was sampled, starting with one of the ten first sections, which was randomized. The sections were then stained with toluidine blue for light microscopic examination. The brain tissue and the tumour tissue were easily distinguishable by the difference in morphology. The areas of tumour tissue and remaining brain tissue were then measured in each section by morphometry, using an Image Analysis System (Kontron, Eching, Germany).
The area of tumour tissue and brain tissue measured on each slide multiplied by the distance between every sampled section (15 x 1.5 gm) gave an estimate of the volume between the sampled sections. The total volume of tumour tissue and of remaining brain tissue was calculated by summation of all the individual volumes. The coefficient of error in this method for estimating volume of sections is below 5% (Gundersen and Jensen, 1987).

Control experiments
To see if Cremophor EL alone had any effect on cell growth and migration, both GaMg and D-54Mg spheroids were treated with 0.02 and 0.1 jg ml' Cremophor EL respectively. Migration and growth measurements were performed as described above. The experiments were done in triplicate.
In addition, to see if Cremophor EL alone had any effect on cell invasion, GaMg and D-54Mg spheroids were also exposed to 0.02 and 0.1 jIg ml-' Cremophor EL before confronting them with the normal rat brain cell aggregates. Exponentially growing monolayers of GaMg and D-54Mg cells They were then washed for 3 x 5 mmn in PBS before incubating were exposed to 0.03 jig ml and 0.06-0.1 jig ml paclitaxel them with a monoclonal antibody against 0-tubulin (2 jig ml-') for respectively. These doses were chosen based on the results obtained before an incubation with fluorescein isothiocyanate (FITC)-conjugated sheep IgG against mouse immunoglobulins (20 jg ml-', both antibodies were from Boehringer, Mannheim, Germany). Thereafter, the monolayers were washed for 3 x 5 min in PBS and then incubated with 0.5% RNAase in PBS for 30 s. The cells were then washed in PBS before incubation with propidium iodine (5 jig ml-1 in PBS). After a brief wash in PBS, the cells were mounted using a Vectashield mounting medium (Vector Laboratories, Burlingame, CA, USA). The cells were then observed using a Biorad MRC-1000 scanning confocal microscope (Biorad, Hemel Hempstead, UK).

Flow cytometry
Exponentially growing monolayers of GaMg and D-54Mg cells were exposed to 0.03 jg ml-' and 0.1 jg ml-1 paclitaxel, respectively, for 24 and 48 h. The cells were then fixed in ice-cold 96% ethanol and stored at 4°C until use. Before DNA measurements, the cells were incubated for 15 min with 0.5% pepsin at 37°C. The isolated nuclei were washed in 0.9% sodium chloride and treated for 1 min with ribonuclease (1 mg ml-' in physiological saline). Staining of DNA was obtained by adding propidium iodine (50 jig ml-1 in physiological saline) to the nuclei. The cellular DNA content was measured using a FACsort flow cytometer (Becton Dickinson, Palo Alto, CA, USA). Human peripheral blood lymphocytes were used as a standard diploid (2c) reference. The DNA histograms were obtained by gating a two-parameter forward and side scatter cytogram to a one-parameter DNA histogram. Each histogram was obtained by counting a total of 10 000 cell nuclei.

Tumour cell migration
The directional cell migration from the spheroids was determined for both cell lines exposed to increasing concentrations of paclitaxel. Both the GaMg and the D-54Mg spheroids showed a dosedependent response after treatment with paclitaxel. This effect was a rather rapid effect that progressed during the first 4 days after plating (Figure 1). For the GaMg cells, the outgrowth area was reduced by about 65% and 90% when continuously exposed to 0.003 and 0.01 gg ml-' paclitaxel respectively (Figure 1).
For the D-54Mg cells, a 50% reduction in outgrowth area was obtained by adding 0.04 gg ml-1 paclitaxel. An 85% reduction was obtained by 0.1 ,Ig ml-1 paclitaxel (Figure 1). The GaMg cells were found to be more sensitive to paclitaxel than the D-54Mg cells (Figure 1).

Spheroid growth measurements
For both cell lines, spheroids continuously exposed to paclitaxel also showed a dose-dependent inhibition of growth. This effect was clearly apparent after 3 days. For both cell lines, a paclitaxel exposure of 0.01 jg ml-caused a 50% reduction in spheroid growth (Figure 2).
At higher concentrations, there was a difference in sensitivity between the two cell lines. For instance, 0.04 jig ml-' paclitaxel caused a complete growth inhibition for the GaMg cells, while the same dose for the D-54Mg cell line caused a 75% growth reduction.
A fivefold higher paclitaxel concentration was needed to obtain complete growth inhibition of the D-54Mg cell line (0.1 jIg ml-') compared with the GaMg cells (0.02 jg ml-').

Tumour cell invasion
In co-cultures without paclitaxel, a replacement of brain tissue by invading GaMg cells was observed in the sections. In the confrontation zone, the outer layer of glial cells was lost, and the brain aggregate volume was reduced to 54 ± 2.5% of the initial volume ( Figure 3A and Figure 4). For D-54Mg cells, the control co-cultures presented a relative poorly defined border between normal and tumour tissue. The brain aggregate volume was reduced to 57 ± 2% of the initial volume ( Figure 3E and Figure 4). The tumour spheroids were exposed to paclitaxel 3 days before confronting them with the brain aggregates. As shown in Figure 2, this period was long enough to induce a growth inhibition. The cocultures were allowed to grow for 96 h.
After 4 days, treated GaMg spheroids showed a severe reduction in tumour volume compared with the control experiments, and no tumour cell invasion was observed. It is, therefore, shown that paclitaxel strongly inhibits the invasive process ( Figure 3C and Figure 4). These experiments were also performed with D-54Mg spheroids. Again, paclitaxel prevented tumour cell invasion ( Figure 3F and Figure 4).

Control experiments
Tumour cell migration and proliferation Cremophor EL alone had no effect on GaMg and D-54Mg tumour spheroid growth and migration ( Figure 5).

Tumour cell invasion
In the co-cultures in which the GaMg and D-54Mg spheroids were treated with Cremophor EL before confrontation, a similar reduction in brain volume was observed as shown for the control experiments (Figure 3 and Figure 4). This suggests that the Cremophor EL alone did not cause any inhibitory effect on the invasion process.

Scanning confocal microscopy
Both the GaMg and D-54Mg cells showed an extensive random organization of the microtubules in the cytoplasm. Most of the cells had a single nucleus with a large number of microtubules radiating out from it ( Figure 6A and D). After exposure to 0.03 ,ug ml-1 paclitaxel for 48 h, the GaMg cells were more contracted and the microtubules were more strongly evident in the cytoplasm ( Figure  6B and C). Of the GaMg cells, 30% also expressed a fragmentation of the nuclear material (Figure 6B), indicating a possible induction of apoptosis. In comparison, the D-54Mg cells were less affected by the paclitaxel exposure and several cells expressed a similar morphology to the control experiments. However, in these cells the microtubules were strongly expressed also ( Figure 6E). At 0.1 jIg ml-1 paclitaxel, a nuclear fragmentation was also observed in the D-54 Mg cell population. Some cells also showed a strong condensation of the cytoplasm, which may be explained by the strong polymerization of the microtubules commonly seen in paclitaxeltreated cells ( Figure 6F). comparison, the cell cycle distribution was 54%, 27% and 19% for the GaMg controls ( Figure 7A and B). The percentages of cells in the G1, S and G2/M phases of the cell cycle for the D-54Mg cells after 24 h were 30%, 34% and 36% respectively. In comparison, the cell cycle distribution was 57%, 22% and 21% for the D-54Mg controls (data not shown). After 48 h of paclitaxel exposure, a deterioration of the DNA histograms for both cell lines was evident. At 48 h, fragmented nuclei with an incomplete DNA content were evident to the left of the G, peak. This was accompanied by a decline in the G, and G2 peaks. This probably indicates a fragmentation of the nuclear material consistent with apoptotic cell death ( Figure 7C).

DISCUSSION
In the present work, it is shown that paclitaxel strongly inhibits glioma cell growth, migration and invasion in vitro. The assay systems that have been used involve the use of three-dimensional spheroids of both normal and malignant origin. Such cultures provide a cellular microenvironment, which to some extent reflects the in vivo situation . Therefore, based on the dramatic cellular effects as seen after paclitaxel treatment, the drug should in theory be effective for the treatment of gliomas in vivo. Recent reports have, however, shown that when paclitaxel is administered by i.v. infusion, the cerebrospinal fluid levels reach only 0.12-8.3% of those present in concomitant plasma samples. Thus, the drug penetrates the blood-brain barrier poorly (Glantz et al, 1995). Despite this, there is at present some evidence that paclitaxel can accumulate in brain tumour tissue, suggesting that the drug may have a place in brain tumour therapy (Heimans et al, 1994). In line with this, several reports have indicated that the drug can be effective against experimental brain tumours, when it is delivered either from biodegradable polymer implants (Walter et al, 1996) or from liposome-encapsulated vehicles (Riondel et al, 1992). Based on the fact that there may be a clinical potential for using paclitaxel in the treatment of gliomas, we studied the effect of the drug on three known biological parameters (cell growth, migration and invasion) that all play an important role in the progression of the disease (Lund-Johansen et al, 1990). Several critical cellular functions, such as mitosis, cell movement and maintenance of cell structure, are performed by cytoskeletal elements in which the microtubular system plays a major role. Paclitaxel represents a mitotic inhibitor in which the mechanism of action is to enhance the rate and yield of microtubular assembly preventing microtubular depolymerization (Schiff et al, 1979). Several reports have shown that such effects induce a block of cells in the G2/M phase of the cell cycle (Gangemi et al, 1995;Chang et al, 1996;Wahl et al, 1996) and our results support these findings (Figure 7). The reduced growth as seen in the spheroids can therefore be explained by an inhibition of cell division, which has also been shown by other authors (Teng et al, 1977;Crossin and Camey, 1981;Otto and De Assua, 1983). In addition, the scanning confocal microscopy indicated that both the GaMg and the D-54Mg cells underwent a nuclear fragmentation as a result of paclitaxel exposure ( Figure 6B). Such a fragmentation may imply that paclitaxel also induces apoptosis in the glioma cells. That the nuclear material is deteriorating is further supported by Effects of Taxol on gliomas 1751 the nuclear DNA content with a reduction in G1 and G2/M cells after 48 h of paclitaxel exposure (Figure 7). Similar observations have also been made in fibroblasts by Wahl et al, 1996). The outgrowth assay measures the ability of the cells organized in a three-dimensional structure to migrate and to proliferate. Cell migration is the dominating process that occurs during the first days after plating (Storme et al, 1981). Notably, paclitaxel caused an inhibition of cell migration for both cell lines (Figure 1). This effect was dose dependent, resulting in a complete migratory arrest of the GaMg cells after 0.02 ,ug ml-' paclitaxel exposure. The D-54Mg cells were also inhibited but were at least 5-10 times less sensitive to paclitaxel than the GaMg cells. The biological mechanisms explaining this difference are at present not completely understood. Other groups have also observed differences in sensitivity to paclitaxel between cell lines, and it has been suggested that this may be partly caused by a variability in the expression of the multidrug resistance gene (MDR]) (Helson et al, 1993;Berkova and Page, 1995). In theory, the results may suggest that the D-54Mg cell's resistance to paclitaxel is not related to its inability to target microtubules, but to the inability of the D-54Mg cells to undergo apoptosis (Gangemi et al, 1995). However, as indicated by the flow cytometric experiments, a deterioration of the DNA histograms was observed for the D-54Mg cells exposed to paclitaxel, which indeed indicates that the D-54Mg cells are able to undergo apoptosis. In line with our results, other groups have also shown that human glioma cell lines are sensitive to paclitaxel .
The scanning confocal microscopy revealed a large variation in the cellular morphology after paclitaxel exposure within the cell lines. This was most evident for the GaMg cells and suggests that paclitaxel may induce multiple biological effects within a heterogeneous tumour cell population ( Figure 6A-C). The influence of tumour heterogeneity on the efficacy of paclitaxel is caused not only by cell clones exhibiting inherent differential sensitivities, but also by the presence of subpopulations of cells differing in metabolic and physiological states. Within spheroids, steep gradients can exist for cellular oxygen levels, nutrients, pH and glucose concentrations (Mueller-Klieser, 1987). All these factors may affect glioma cell growth, migration and invasion and will probably interfere with the direct biological effects induced by paclitaxel.
Paclitaxel reduces cellular motility through interference with the microtubules, a fundamental part of the cytoskeleton. Thus, the present study confirms other studies (Stracke et al, 1993;Verschueren et al, 1994;Sollott et al, 1995), but contradicts the findings of Silbergeld et al (1995) who found that paclitaxel increased the motility of glioma cells.
In line with the inhibition of migration, we also showed that paclitaxel reduced the invasive capacity of the tumour cells (Figure 3), implying that the drug is active in inhibiting tumour cell invasion in vivo. Notably, the paclitaxel-treated glioma cells were not able to infiltrate the brain cell aggregates. This implies that the treated tumour cells were not able to move in the coculture system. It is well known that complex interactions take place between normal tissue and tumour cells. This occurs in a microenvironment in which growth factors and other cellular components are exchanged between the two cell populations in a paracrine fashion. Such factors were not able to induce the invasive process. Thus, our data support the findings of other groups showing an inhibition of MDCK and prostate tumour cell invasion into Boyden chambers and Matrigel (Stearns and Wang, 1992;Dugina et al, 1995). Interestingly, Cremophor EL (the solvent for paclitaxel for clinical use) had no effect on cell growth, migration and invasion in our assay systems at the concentrations used. This indicates that the cellular effects as shown in our three-dimensional assays were not caused by any interference with the paclitaxel solvent. These observations may contradict other findings showing cytotoxic effects of Cremophor EL on tumour cells (Liebmann et al, 1993;Nygren et al, 1995). However, the results that were obtained in these studies were based on experiments performed on monolayer cell cultures. It is well known that such cultures, in general, are more sensitive to therapy compared with multicellular spheroids that have been used in the present study (Bjerkvig, 1992). At present, it is therefore not clear to what extent Cremophor EL affects a malignant three-dimensional tissue structure. In conclusion, we have shown that paclitaxel inhibits the invasive growth of human glioma cell in vitro, using a threedimensional co-culture system involving both tumour and normal brain cells. The effects observed by paclitaxel imply that the drug is strongly active towards this tumour type. Thus, future studies should probably focus on new administration routes of paclitaxel to gliomas.