Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

When a new drug is brought into clinical development, the choice of dosing strategy is of great importance. There may be differences in both drug efficacy and toxicity between giving the drug, e.g., as one high, single dose and giving it continuously over several days. It could be useful to study the effects of different anticancer drug exposures in vitro, to gain information to use in the development of clinical dosing strategies. Several approaches have been used to study the time dependence of cytotoxic drug effect, and the methods and models used are diverse. Some investigators discuss the shape and characteristics of the concentration-effect curves in a more descriptive way (Knowles and Hwang, 1995; Karlsson et al., 1998) and few in vitro studies model both exposure time and drug concentration (Kalns et al., 1995). Others use a more mechanistic approach taking, for example, cell cycle specificity of the drug and growth characteristics of the cells into account (Ozawa et al., 1989; Gardner, 2000).

Levasseur et al. (1998) have recently suggested a method to use in vitro data from a large number of different drug concentrations and exposure times to simultaneously model both the concentration-effect curves and the dependence of its different parameters on exposure time. They present data and models for seven cytotoxic drugs, including doxorubicin, cisplatin, and methotrexate, and also some cautious interpretation of the results.

CHS 828 is a novel cytotoxic agent that has shown promising preclinical behavior and is currently in early clinical trials (Hjarnaa et al., 1999; Jonsson et al., 2000, 2001; Ahlgren et al., 2001). There have been some in vivo observations of a schedule-dependent drug effect. Weekly administration of CHS 828 has been shown to be more effective than daily administration in a 7-week treatment of MCF-7 xenografts in nude mice (Hjarnaa et al., 1999). On the other hand, giving the same amount of drug divided into five daily doses instead of giving it all on 1 day increases the effect of CHS 828 substantially in a 6-day hollow fiber rat model (Jonsson et al., 2000; L. Friberg, S. B. Hassan, E. Jonsson, R. Larsson, and M. O. Karlsson, unpublished data). When bringing CHS 828 into clinical studies the issue of drug scheduling is of major importance. Giving the drug daily for 5 days every cycle, as one single dose every cycle or once weekly, may give different treatment results in the clinic. This question is being currently addressed in ongoing clinical trials.

In this study, the time dependence of CHS 828 effect was studied in vitro, with the method suggested by Levasseur et al. (1998). The method was slightly modified for our purposes, and this study also includes data on the cytotoxic drug paclitaxel, topotecan, and etoposide for which the optimal clinical dosing schedule is still a matter of debate.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Cell Lines. The cell lines used were the histocytic lymphoma cell line U937 GTB, kindly provided by Prof. K. Nilsson (Department of Genetics and Pathology, Uppsala University, Uppsala, Sweden), the human breast cancer cell line MDA 231, kind gift from Dr. Jonas Bergh (Department of Oncology, Uppsala University), and the myeloma cell line RPMI 8226/S, kindly provided by Dr. W. S. Dalton (Department of Medicine, Cancer Center Division, University of Arizona, Tucson, AZ). The cells were grown in culture medium RPMI 1640 (HyClone, Northumbria, UK), supplemented with 10% heat-inactivated fetal calf serum (Sigma Chemical, St. Louis, MO), 2-mM glutamine, 50 µg/ml streptomycin, and 60 µg/ml penicillin. Growth and morphology were monitored weekly.

Primary Human Tumor Cells and Normal Lymphocytes. Peripheral blood was obtained from one patient with chronic lymphocytic leukemia (CLL). Leukemic cells were isolated from the blood by density gradient centrifugation on 1.077 g/ml Ficoll-Paque (Amersham Pharmacia Biotech AB, Uppsala, Sweden) (Larsson et al., 1992). Cell viability was determined with trypan blue dye exclusion test, and the proportion of tumor cells was found to be 75% judged by microscopic examination of May-Grünwald-Giemsa-stained cytospin preparations by a trained hematologist. The cells were cryopreserved in fetal calf serum containing 10% dimethyl sulfoxide (DMSO; Sigma Chemical) by initial freezing for 24 h in 70°C followed by storage in liquid nitrogen, which does not appear to affect drug sensitivity (Nygren et al., 1992). Peripheral blood from one healthy donor was collected and normal mononuclear cells were prepared as described above.

Drugs and Reagents. CHS 828 was provided by Leo Pharmaceutical Products (Ballerup, Denmark) dissolved in DMSO and was tested at 18 concentrations obtained by 2-fold serial dilution in phosphate-buffered saline (PBS; Hyclone), by using 10 µM as maximum concentration. Paclitaxel, topotecan, and etoposide were obtained from the hospital pharmacy and dissolved according to manufacturer's instructions and were tested at 18 concentrations obtained by 2-fold serial dilution from a maximum of 11.7 µM for paclitaxel and 169.9 µM for etoposide and 235.5 µM for topotecan. Ninety-six-well microtiter plates (Nunc, Roskilde, Denmark) were prepared with 20 µl/well of drug solution at 10 times the desired concentration, with the aid of a pipetting robot (Propette; PerkinElmer Instruments, Norwalk, CT). Each concentration was prepared in triplicate. The plates were stored frozen at 70°C for up to 2 months until further use. Under these conditions, no apparent change in drug activity was observed (Larsson et al., 1992). PBS was used throughout for washing procedures. Fluorescein diacetate (FDA; Sigma Chemical) was dissolved in DMSO and kept frozen (20°C) as stock solution protected from light.

Exposure. To develop and evaluate the assay procedure some initial experiments were performed using the 8226/S cell line. To determine the possible loss of cells in the washing procedure, four 96-well microtiter plates were seeded with 200 µl of cell suspension (20 × 10 ³ cell/well) in culture medium with the aid of a programmable pipetting robot. The plates were incubated at 37°C for 1 h before washing. One plate was washed once, the second plate was washed twice, the third plate was washed four times, and the last plate was served as a control, unwashed plate. The plates were washed as follows: they were centrifuged at 200g for 5 min and the supernatant was aspirated with the aid of multireagent microtiter plate washer (Dynatec Laboratories, Billingshurst, West Sussex, UK). Sterile PBS (180 µl) was added to every well with the aid of a programmable pipetting robot followed by a new centrifugation. After the last washing step, 180 µl of fresh medium was added. The cell number in each plate after completing the washing procedure was determined with the fluorometric microculture cytotoxicity assay procedure (FMCA). The fluorescence signals in the different plates were compared as a measurement of difference in cell number.

Fluorometric Microculture Cytotoxicity Assay Procedure. The FMCA procedure is based on measurements of fluorescence generated from hydrolysis of FDA to fluorescein by cells with intact plasma membrane and has been described previously in detail (Larsson et al., 1992). At the end of the 72-h incubation, the plates were centrifuged and the medium was aspirated followed by one wash with PBS and addition of 100 µl/well of FDA dissolved in PBS (10 µg/ml). The plates were incubated for 40 min and the generated fluorescence from each well was measured at 538 nm in a 96-well scanning fluorometer (Fluoroscan II, Labsystems Oy, Helsinki, Finland). The fluorescence is proportional to the number of the viable cells in the well.

Data Analysis. Data analysis was performed using nonlinear mixed effect models. For each model, data from all experiments using the same drug and cell line was included in the analysis (1-3 experiments with a total sum of 5-23 plates, each including 18 drug concentrations tested in triplicate). The models were selected from Table 1on the basis of a graphical examination of the concentration-effect curves and the pattern of change of the parameters with time together with comparison between the objective function values and the precision of the parameter estimates. A mixed effects model estimates two types of variability. In the present case these two levels represented plate-to-plate variability in parameter values, when such variability was estimated to be present, and residual variability for each observation compared with the predicted curve. Parameter variability was described by a log normal distribution and residual variability by an additive error model. The first order method implemented in the program NONMEM (Beal and Sheiner, 1992) was used for the analysis, and the program Xpose (Jonsson and Karlsson, 1999) was used for model diagnostics. Figures were prepared with GraphPad Prism (GraphPad Software, San Diego, CA).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Under our experimental conditions, the loss of cells from the washing procedure was very small (<2%) after both a two- and four-time washing procedure (data not shown). It was also found that washing two times was not enough to remove all CHS 828 because the cytotoxic effect was further decreased after addition of two washes as shown in Fig. 1. An extra 2-h incubation step between the last washes did not change the pattern of the concentration-response curve. The remaining volume in the well after aspiration of the supernatant in each washing step was found to be 20 µl (data not shown). Thus, the dilution factor for two washes was around 10³ and for four washes around 10⁵. Based on this, the four-wash procedures excluding the extra 2-h incubation step was chosen for all experiments.

View larger version (17K): [in this window] [in a new window]   Fig. 1.   Cytotoxic effect of a 2-h CHS 828 exposure induced on 8226/S cells by using different washing procedures to remove the drug. Effect is expressed as SI defined as cell survival in percentage of control. One typical experiment of two. , two washes; , two washes + 2-h extra incubation time; , four washes; ; four washes + 2-h extra incubation time; and , 72-h continuous exposure.

Figure 2 displays the concentration-effect curves of CHS 828 on U937 GTB, MDA 231, 8226/S, CLL, and normal mononuclear cells and of paclitaxel, etoposide, and topotecan on the U937 GTB cells. The CHS 828 curves had a similar shape for all three cell lines, CLL, and normal mononuclear cells. The curves were steep at short (<24-h) and long (>48-h) exposures, whereas the curves became shallow at intermediate exposure times (24-36 h). The maximum effect did not appear to change with exposure time. The U937 GTB cell line and CLL cells were more sensitive to CHS 828 than MDA 231, 8226/S, and normal mononuclear cells. Response to paclitaxel increased with increasing the exposure time and the concentration-effect curves became steeper as exposure time increased. For etoposide and topotecan the effect increased with exposure time up to 24 h, but longer exposure times did not seem to increase the effect further. The plateau level of the upper part of the concentration-effect curves of topotecan appeared to be time-dependent and a second drop in the cell viability was observed at concentrations above 117 µM at shorter exposure times (<24 h).

View larger version (22K): [in this window] [in a new window]   Fig. 2.   Concentration effect curves of CHS 828 on U937 GTB, MDA 231, 8226/S, CLL and normal mononuclear cells and of paclitaxel, etoposide and topotecan on U937 GTB tumor cells after different exposure time. One representative experiment. Effect is expressed as SI (%) of control. Different symbols represent different exposure times in h. +, 1; , 2; , 4; , 6; , 12; *, 24; , 30; , 36; , 38; and , 72.

Table 1 presents the structural models used to describe the concentration-effect curves and models for change of the parameters with time. Figure 3 shows a schematic representation of single Hill concentration-effect curve (A) and a double Hill concentration-effect curve (B) simulated from eqs. 1a and 1b in Table 1.

View larger version (14K): [in this window] [in a new window]   Fig. 3.   Schematic representation of single Hill concentration-effect curve (A) and a double Hill concentration-effect curve (B) simulated from eqs. 1a and 1b in Table 1. The parameters used to construct the simulation for A were Econ = 106, B = 10.7, IC50 = 0.903 µM,  = 8.83. The parameters used for construction of B were Econ = 100, GROW = 18.5, IC50a = 0.101 µM, a = 0.698, Econb = 72.25, IC50b = 132.8 µM, b = 5.65, B = 1.

The single Hill model (Table 1, eq. 1a) was used to describe the concentration-effect relationship of CHS 828. The concentration-effect curves of mononuclear cells were not modeled due to the small amount of data but the curves showed trends similar to all other cell types with CHS 828. The same model was used for paclitaxel and etoposide relationships. For topotecan, a double Hill model (Table 1, eq. 1b) was used to describe the concentration-effect relationship. An extra parameter (GROW) was added to describe an increase in cell survival after exposure to low drug concentrations.

Table 2 presents the parameter estimates of IC₅₀, slope, and plateau parameters for all different drugs and cell lines together with the relevant model equations. The data contained enough information to estimate the interindividual variability in Econ in all data sets and in IC₅₀ for U937 GTB and MDA 231 on CHS 828 and for U937 GTB on the three standard drugs. The data contained enough information to estimate the interindividual variability in slope for U937 GTB on CHS 828 and for U937 GTB on paclitaxel and topotecan. The estimated IC₅₀ values (left), slopes  (middle) of the concentration-effect curves as a function of exposure time of all drugs combinations, and the goodness of fit plots from individual predicted and observed SI % (right) are displayed in Fig. 4. The pattern of change of IC₅₀ for CHS 828 with exposure time had a sigmoid shape for all cell lines tested, as well as for CLL. This was best described by eq. 2c in Table 1. At exposure times less than 24 h, drug potency was leveled then sharply decreased at intermediate times (24-36 h) and finally plateaued at higher times (48-72 h). The absolute value of the slope parameter () of the concentration-effect curves decreased up to 24 h (Table 1, eq. 3b.I, for U937 GTB and 8226/S cell line and up to 28 h for CLL (Table 1, eq. 3b.II) then increased again after this exposure time. The absolute values of the slope parameter estimate were constant (Table 1, eq. 3a) for the MDA 231 cell line.

View larger version (26K): [in this window] [in a new window]   Fig. 4.   Pattern of change of IC50 (left), slope (middle) with time, and goodness of fit plots from individual predicted and observed SI % (right) for CHS828 on U937GTB, MDA 231, 8226/S, and CLL and for paclitaxel, etoposide, and topotecan on U937 GTB tumor cells. Curves are best fits of eqs. 2 and 3 in Table 1. Data points are the individual parameter estimates for every exposure time in each experiment obtained from fitting of eq. 1a or 1b in Table 1 to the observed SI. Different symbols represent different experiments. Shown is line of identity () for Goodness of fit plots.

Equation 2c in Table 1 was also used to describe the patterns of IC₅₀ change for etoposide and topotecan but the sigmoidicity was not very pronounced for etoposide, with low T₅₀ (4.12 h) and shallow slope (2.99), respectively. Equation 2a in Table 1 was used to describe the IC₅₀ in the second part of the topotecan concentration-effect curve. The absolute value of the slope parameter () of the concentration-effect curves for topotecan was increased and then leveled (Table 1, eq. 3b.I), whereas the slope parameter was constant for etoposide (Table 1, eq. 3a).

The appropriate model used to describe the pattern of change of IC₅₀ with time for paclitaxel was the IC_xⁿ T model (Table 1, eq. 2b). IC₅₀ declined with exposure time and n was fixed to 1 because its estimate was very close to 1. The absolute value of the slope parameter estimate was increased with exposure time for paclitaxel (Table 1, eq. 3b.I).

For topotecan the pattern of change of Econ_b with time was an exponential decrease and was best described by eq. 4a in Table 1. The goodness of fit plots showed that the models give an adequate description of the data.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

This work was performed based on previous results published by Levasseur et al. (1998). We have shown a feasibility of using the fluorescence-based nonclonogenic cytotoxicity assay FMCA instead of the sulforhodamine B protein dye assay for determining the cell survival, and to use NONMEM instead of SAS/ProcNLIN for data modeling. Based on the initial experiments presented, we chose four washing steps instead of two, which was used in the original reference. Four washings is a time-consuming procedure, but seems to give negligible cell loss and minimizes the risk of remaining drug continuing to exert effect after washing. This may be especially important for CHS 828, which is extremely potent at long exposure times. Based on our extensive washing procedure, we have chosen not to take possible dilution artifacts into account in the model.

The concentration-effect relationships for CHS 828, etoposide, and paclitaxel were all well described by single Hill models. For topotecan, a double Hill model was used to take into account the drop in cell viability occurring at concentrations above 117 µM. The biological significance of this second drop is unclear, and the effect may be nonspecific and caused by concentrations far above those obtainable in vivo. The log IC₅₀ decreased linearly with the logarithm of the exposure time for paclitaxel, which is similar to what Levasseur et al. (1998) described for most drugs tested. In the IC₅₀ⁿT equation, n was estimated to be fairly close to 1. This might be interpreted as a situation where the cytotoxic effect of the drug is simply dependent on the concentration times time product (area under the concentration-time curve), and is independent on the schedule used. The patterns of IC₅₀ change for topotecan and etoposide in our study was best described by a sigmoid relationship. However, the sigmoidicity was not very pronounced for etoposide with low T₅₀ and a shallow slope. The T₅₀ was estimated to be around 4 h, demonstrating that the log IC₅₀ decreased rather linearly with log time before 24 h. The decrease was somewhat greater than for paclitaxel. The effect did not get significantly better with exposure times over 24 h for neither etoposide nor topotecan. For all three standard drugs, one main mechanism of action has been proposed: microtubule stabilization for paclitaxel, topoisomerase II inhibition for etoposide, and topoisomerase I inhibition for topotecan. Some investigators have suggested additional mechanisms, such as phosphorylation of Bcl-2, Bcl-xL, and other proteins for paclitaxel (Blagosklonny and Fojo, 1999), DNA interaction for etoposide (Gordaliza et al., 2000), and nuclear factor-B activation for camptothecins (Huang et al., 2000). The slight variation in slope and sigmoidicity in the IC₅₀ versus time relationship might be explained by a minor influence of these possible additional mechanisms of action.

The most commonly used dosing schedules for topotecan and etoposide are daily short infusions for 3 to 5 days, whereas a single 3- or 24-h infusion is recommended for paclitaxel (FASS, 2000). All these three drugs are suggested to act preferentially on dividing cells in specific phases of the cell cycle (S or M phase), and therefore should theoretically be more efficient if given over longer times to enable more cells to get through the cell cycle. There are ongoing discussions if more protracted schedules would be more efficient. The optimal clinical dosage regimens of these drugs are, however, still not determined. Clinical studies have suggested that a more extended administration of etoposide might be more effective (Joel and Slevin, 1994), whereas this has still not been demonstrated convincingly for paclitaxel (Rowinsky, 1997) or topotecan (O'Reilly, 1999). These clinical data appear roughly to be consistent with the present findings.

Unlike for the standard drugs, the drop in IC₅₀ for CHS 828 with exposure time had a clearly sigmoid shape for all cells tested, and the steep increase in drug potency was occurred between 20 and 30 h. At these intermediate time points the concentration-effect curves had a characteristic shallow shape. In the study by Levasseur et al. (1998) only trimetrexate induced a sigmoid-type decrease of IC₅₀ with time. The pattern was in some aspects similar to what was observed for CHS 828, with a decrease in IC₅₀ at exposure times between 10 and 48 h and shallower curves at the intermediate concentrations. The mechanism for this seems unclear, but the pattern was suggested to argue for long-time exposure of trimetrexate for higher drug potency.

A shallow concentration-response relationship could be interpreted as a sign of a heterogeneous cell population or multiple cellular targets (Levasseur et al., 1998). Response heterogeneity due to difference in cell cycle phase of the cells may be less likely in the case of CHS 828, because the same activity pattern was observed in primary cells from CLL and normal mononuclear cells, which have been shown not to proliferate significantly under these experimental conditions. In addition, because CHS 828 showed a similar pattern of activity in cell cultures of a single cell type disease such as CLL as in tumor cell lines, the phenomenon is possibly not due to a heterogeneous cell population response. Another possible explanation of a sigmoid drop in IC₅₀ may be drug remaining bound inside the cells after the washing step. This may not be likely for CHS 828, because the washing procedure with an intermediate 2-h incubation period did not change the cytotoxicity pattern, which suggests a fast CHS 828 equilibrium between the intracellular and extracellular fluid. The data may instead suggest two different mechanisms of action of CHS 828, one of lower potency independent of exposure time and one, very potent, acting only at longer exposure times. The identity of the molecular targets or pathways influenced by CHS 828 remains to be identified, but these data suggest that both short and long time exposures of CHS 828 should be explored in the search for the mechanism of action of the drug. Recent studies have shown that CHS 828 induces an increase in extracellular acidification rate in tumor cells that could be explained by mitochondrial respiration inhibition causing a rapid decrease of ATP level (Ekelund et al., 2001). However, this alone does not explain the cytotoxic action of CHS 828 so additional mechanisms of drug action must be present. Treatment with 3-aminobenzamide produces a significant right shift of the concentration-response curve resembling that typical for the short-term exposure curves presented here. This effect of 3-aminobenzamide was independent of poly (ADP-ribose) polymerase. Whether this represents the pharmacological association of two different and schedule-dependent mechanistic pathways remains to be elucidated (Lövborg et al., 2000).

The results described for CHS 828 fit with the observations made in the in vivo hollow fiber model, where a protracted scheduling was more effective than one single dose (Jonsson et al., 2000). One may speculate that the single dose only uses the low-potency effect of short drug exposure, whereas dividing the dose into five separate doses makes the drug exposure long enough to be influenced by the high-potency effect part. In the ongoing clinical phase I trials, CHS 828 given as one single dose seems to be less toxic than giving the same dose divided into 5 days (Ravaud et al., 2001). If these dosing schedules result in two mechanistically different effects, it would be important to elucidate which one gives the best therapeutic index and produces the best effect on different tumor types.

The present approach represents a descriptive way of showing the concentration-time-effect relationships for CHS 828 as well as paclitaxel, topotecan, and etoposide. It may provide information useful in the design of interesting dosing schedules for in vivo and clinical studies. As suggested by Gardner (2000) more mechanistic models on schedule dependence may be even more useful. Further studies on the mechanistic background for the time-dependent behavior of CHS 828 are ongoing. This will hopefully lead to results sufficient to create a mechanistic model that may provide insight into the mechanisms of CHS 828 schedule dependence.