Quantitative Measure of Cytotoxicity of Anticancer Drugs and Quantitative Measure of Cytotoxicity of Anticancer Drugs and Other Agents Other Agents

Many anticancer drugs act on cancer cells to promote apoptosis, which includes impairment of cellular respiration (mitochondrial O 2 consumption). Other agents also inhibit cellular respiration, sometimes irreversibly. To investigate the sensitivity of cancer cells to cytotoxins, including anticancer drugs, we compare the proﬁles of cellular O 2 consumption in the absence and presence of these agents. Oxygen measurements are made at 37 (cid:2) C, using glucose as a substrate, with [O 2 ] obtained from the phosphores- cence decay rate of a palladium phosphor. The rate of respiration k is deﬁned as (cid:2) d [O 2 ]/ dt in a sealed container. Different toxins produce different proﬁles of impaired respiration, implying different mechanisms for the drug-induced mitochondrial dysfunction. The decrease in the average value of k over a ﬁxed time period, I, is proposed as a characteristic value to assess mitochondrial injury. The value of I depends on the nature of the toxin, its concentration, and the exposure time as well as on the cell type. Results for several cell types and 10 cytotoxins are presented here.

Although the process of drug-induced tumor cell death remains poorly understood, many anticancer agents induce apoptosis through similar mechanisms such as the release of cytochrome c from mitochondria into cytosol and the consequent activation of caspases. The activated caspases then attack the permeabilized mitochondria, disrupt the mitochondrial electron transport chain, collapse the mitochondrial inner membrane potential, and diminish the mitochondrial structural integrity [1,2]. Interferences in the apoptotic scheme reduce cancer cells' sensitivity to therapy [3,4]. Because most anticancer drugs and other cytotoxic agents impair mitochondria directly or indirectly, their cytotoxicity can be easily assessed by measuring their effects on mitochondrial respiration. It is proposed here that measurement of cellular respiration in the presence and absence of an agent is a useful way to characterize cytotoxicity. In the case of an anticancer drug, this measurement becomes an index of the cell's sensitivity to treatment by the drug.
A variety of agents are studied here. The anthracycline antibiotic doxorubicin intercalates with DNA and produces DNA breaks by stimulating topoisomerase II cleavable complex formation [5]. In the cell, the quinone moiety of doxorubicin is reduced to the semiquinone radical, generating reactive oxygen species that directly damage cell organelles and induce apoptosis [6]. Dactinomycin is an important anticancer chromopeptide that intercalates between DNA base pairs and inhibits transcription [7]. Platinum (Pt) 1 com-pounds, including cisplatin, carboplatin, oxaliplatin, and nedaplatin, bind to DNA and promote apoptosis [8]. The latter three Pt agents differ from cisplatin (the parent compound) in the different ligands occupying the Pt coordination spheres. These structural variations yield unique antitumor activity and toxicity profiles such as the reactivity with DNA and induction of apoptosis [8]. Transplatin, the geometric isomer of cisplatin, is expected to be less cytotoxic than cisplatin. Tirapazamine (3-amino-1,2,4-benzotriazine-1,4-dioxide) is a promising agent that is effective in hypoxic tumor environments [9]. The activated compound results from a two-electron reduction of tirapazamine, which is readily reoxidized to its parent form in aerobic conditions [10]. A highly cytotoxic hydroxyl radical (HO) is released and produces DNA breaks, chromosomal aberrations, and eventually cell death [9][10][11]. Not considered as an anticancer drug, cyclosporine A (CsA) is an important immune suppressor. Its cellular targets include cyclophilin and calmodulin, modulating calcium fluxes through mitochondrial ion channel [12,13]. CsA decreases the possibility of mitochondrial transition permeability (MTP) induced by Ca 2+ accumulation in the cytosol [14,15]. In addition, it blocks Ca 2+ efflux from mitochondria, causing calcium flood in the mitochondrial matrix that induces serious cytotoxicity and further injures mitochondrial respiration [16]. Finally, caffeine is known to activate Ca 2+ channels on the plasma membrane and endoplasmic reticulum, rapidly accumulating Ca 2+ in neighboring mitochondria, and this may impair their function. Caffeine also potentiates the cytotoxicity of anticancer drugs.
The purpose of this work is to define a characteristic cytotoxicity parameter that will be useful in pharmacology and toxicology. The parameter is the extent of inhibition of respiration, where inhibition is defined in terms of the ratio of the average rates of respiration in the presence and absence of the cytotoxic agent. Both rates must be measured on cells from the same batch because rates depend on previous history and other uncontrollable factors. It is expected that the rate of respiration in the presence of the agent will depend on the agent's nature, its concentration, and the incubation time, as well as on the cell type, so that the inhibition will depend on all of these factors. For simplicity, the experiments discussed in this article are limited to two cell types: Jurkat and HL-60.
Respiration rate is obtained from the decrease in dissolved oxygen concentration in a closed vessel. Dissolved oxygen concentration is measured from the decay rate of phosphorescence from a palladium phosphor, present in the solution, using a homemade instrument [17][18][19][20]. In previous studies, we used this instrument to explore the mitochondrial perturbation during anticancer therapy in vitro as well as enzyme reactions involving O 2 [21][22][23][24]. The phosphorescence method of oxygen analysis has several advantages compared with electrochemical and other methods. It is totally noninvasive and does not affect the system being measured because no oxygen is consumed by the measurement. It allows measurement of [O 2 ] down to nanomolar concentrations. It can be used continuously for long periods of time with no deterioration (as for electrodes) or loss of accuracy, so that slow processes such as decrease in respiration rates can be followed. Furthermore, it requires only a single calibration, with no necessity to check the instrument during or after long runs.

Reagents
Pd(II) complex of meso-tetra-(4-sulfonatophenyl)-tetrabenzoporphyrin (sodium salt, palladium [Pd] phosphor) was obtained from Porphyrin Products (Logan, UT, USA). Its solution (2.5 mg/ml = 2 mM) was prepared in distilled water (dH 2 O) and stored at À20°C in small aliquots. Doxorubicin HCl (3.45 mM) was purchased from Bedford Laboratories (Bedford, OH, USA). Dactinomycin (actinomycin D, MW 1255.43) was purchased from Merck (Whitehouse Station, NJ, USA); its solution was made fresh in dH 2 O, and its final concentration was determined by absorbance at 440 nm using an extinction coefficient of 24 19) and remaining reagents were purchased from Sigma-Aldrich.

Cells
Human promyelocytic leukemia (HL-60) and T-cell lymphoma (Jurkat) cells were purchased from American Tissue Culture Collection (Manassas, VA, USA). The cells were cultured in medium plus 10% fetal bovine serum, 1% l-glutamine, and 1% penicillin/streptomycin. Tu183 cells were obtained from Edward J. Shillitoe (State University of New York, Upstate Medical University). They were derived from a squamous cell carcinoma of the tonsil and were highly resistant to therapy. The cells were cultured in Dulbecco's modified Eagle's medium nutrient mixture F-12 (Invitrogen, Carlsbad, CA, USA) plus 10% fetal bovine serum, 1% penicillin/streptomycin, and 0.2% primosin. For harvesting, the cells were incubated at 37°C in 2.5 ml of 0.05% (w/v) trypsin plus 0.53 mM ethylenediaminetetraacetic acid (EDTA) for 5 min and then collected. Cell counts and viabilities were determined by light microscopy using a hemocytometer under standard trypan blue staining conditions.
Cells were suspended at 0.5 to 1.0 Â 10 6 cells/ml in medium plus 2 lM Pd phosphor and 0.5% fat-free bovine serum albumin.
For each condition, 1.0 ml of the cell suspension was placed in a 1-ml glass vial (8-mm clear vials, Krackler Scientific, Albany, NY, USA). The vial was sealed with a crimp top aluminum seal (using a Wheaton hand crimper, Fisher Scientific) and placed in the instrument for O 2 measurement at 37°C. Mixing was accomplished with the aid of parylene-coated stirring bars (1.67 Â 2.01 Â 4.8 mm, V&P Scientific, San Diego, CA, USA).

O 2 measurement
The O 2 detection system was built to measure the phosphorescence of Pd phosphor as described previously [21,22]. DASYlab (Measurement Computing, Norton, MA, USA) was used for data acquisition. The data were analyzed by a C ++ language computing program [25] that calculated phosphorescence lifetime () and decay constant (1/s). Because oxygen quenches the phosphorescence, the decay rate 1/s is a linear function of [O 2 ] as 1=s ¼ 1=s 0 þ k q ½O 2 ; where 1/s 0 , the decay rate in the absence of oxygen, was 10,087 ± 156 s À1 and the value of the quenching rate constant, k s , was 96.1 ± 1.2 lM À1 s À1 [21].

Index of respiration inhibition
Here we propose an index for the efficiency with which a chemical agent inhibits cellular respiration.
and use hki as a measure of cytotoxicity. Because k = Àd[O2]/dt, To do this, we fit all of the measurements to a suitable analytic function and calculate [O 2 ] at the end points from the parameterized function. A simple and convenient fitting function is a secondorder polynomial, [O 2 ] = a + bt + ct 2 . This is appropriate because it includes the linear function as a special case, c = 0. With this function, k = À(b + 2ct) and hki = Àb À c(t i + t f ). It is expected that the quadratic will be concave upward, corresponding to a gradual decrease in respiration rate, so that b will be negative and c will be positive. More generally, a power series, [O 2 ] = a + bt + ct 2 + dt 3 . . ., can be adopted. If the parameters c and d are small, [O 2 ] versus t will be a line; if not, it will be curved. The cubic term makes it possible to represent a k that is nonmonotonic in time. Although we have not found any cases where the fit is improved significantly by going past the t 2 term, this must be checked under specific circumstances. The average value of k for the cubic is hki ¼ For all toxin-treated conditions, we choose t f À t i as a fixed time interval. Invariably, t i is longer than the preceding incubation time by the time required for the sample preparation. In our experiments, this time can vary from a few minutes to a half-hour, depending on the number of sampling conditions, and is different for different samples. However, this does not affect the calculation of hki given that hki is an average over the same time period for all samples.
After the calculation of hki from fitting of the experimental data to a polynomial, we propose to calculate an index, I, defined as where hki is the average value of negative slope in the toxin-treated condition and hki 0 is the average value of negative slope in the untreated condition. The meaning of I is the average inhibition of cellular respiration due to the toxin treatment during the period when measurements are made .If I depends on dosage D, experimentally determined I values can be fitted to the function where I max and A are constants for a particular drug and cell line. When drug dosage reaches the concentration A, I equals half-maximal inhibition (I max ) of cellular respiration. When a polynomial is used to fit experimental [O 2 ] for t i < t < t f , it must be remembered that the polynomial has no validity outside this region (e.g., [O 2 ] certainly does not increase rapidly for later t as a quadratic does). Because the polynomial cannot be used to extrapolate to t = 0, it may be necessary to use the initial [O 2 ] (e.g., 225 lM at 37°C). This information is added to the experimental points before fitting.

Dactinomycin and doxorubicin
We first evaluated the impact of a prolonged exposure to dactinomycin or doxorubicin on cellular respiration. HL-60 cells (10 6 cells/ml) were suspended in medium, 2 lM Pd phosphor, and 0.5% bovine serum albumin without or with 10 lM dactinomycin.  Table 1).
In and measurements began at 50 min, so that incubation time was 50 min.
Linear fits were made to the data for each condition (dashed lines for untreated and dactinomycin-treated cells, solid lines for doxorubicin-treated cells and cells treated with both drugs). The slopes were À0.929 lM O 2 /min for untreated cells (r 2 = 0.984), À0.732 lM O 2 /min for dactinomycin-treated cells (r 2 = 0.996), À0.833 lM O 2 /min for doxorubicin-treated cells (r 2 = 0.987), and À0.826 lM O 2 /min for the mixture (r 2 = 0.995). Because deviations from linearity were not evident, calculation of average k was unnecessary. For the three conditions, respiration decreased by 21, 10, and 11%, respectively. The results show that for Jurkat cells, dactinomycin is more potent than doxorubicin and the effect of combining the two drugs is less than additive.
The advantages of this protocol include the ability to compare multiple conditions (the same drug at different dosages or the same dosage for various drugs) synchronously. Also, this protocol requires fewer cells per condition, allowing lower rates of O 2 consumption and, thus, longer measurement times that may permit detection of deviations from linearity in [O 2 ] versus t. Based on the above experience, we adopted the latter protocol in most of our experiments.
We previously reported [21] that [O 2 ] versus t plots for cells treated with doxorubicin were actually well fit by two lines; that is, k was constant and equal to the value for untreated cells for approximately 150 min and then decreased to a lower value. The results of Fig. 1B may show this discontinuity in slope. However, this does not have a major effect on our conclusions given that the k values from linear fits, reported above, are average values for the full time of measurement. In contrast to doxorubicin, dactinomycin produced nonlinear [O 2 ] versus t plots that were concave upward; that is, k decreased gradually with time [22]. Therefore, dactinomycin data for long periods of time should be analyzed using a polynomial model.    Table 1 for t i = 90 and 150 min only. This is sufficient because if a single quadratic is used to fit all of the data for [O 2 ] versus t, hki will be a linear function of t i so long as t f À t i is kept constant. respectively. We also fitted to Eq. (5) with I max = 1 (i.e., 100% inhibition) and found the best value of A (8.45 M). However, the low value of r 2 (0.817) indicated that the latter fitting was not suitable. The results presented so far indicate that dactinomycin is approximately twice as potent as doxorubicin for Jurkat cells; that is, to get comparable inhibition of respiration, doxorubicin must be used at twice the concentration of dactinomycin. We next investigated the effect of these two drugs on HL-60 cell respiration. We  tion by 13.3%, whereas doxorubicin inhibited respiration by 22.0%; after 180 min incubation, dactinomycin inhibited cell respiration by 47.6%, whereas doxorubicin inhibited respiration by 48.8%. Thus, dactinomycin is twice as potent as doxorubicin for HL-60 cells as well as for Jurkat cells.
There is an apparent decrease in the respiration rate for untreated cells with time (k = 0.645 at 30 min and 0.435 at 180 min), possibly due to problems with maintenance of the cell culture. Day-to-day variations are likely even more important than intraday variations. This underlines the importance of always comparing measured k for treated cells with k for untreated cells from the same batch measured over the same time period, as we did here.

Pt compounds
We next investigated the effects of Pt compounds on the respiration of Jurkat cells. Fig. 3A shows results of one experiment in which 10 6  which was the same as that for individual drugs within statistical error. Furthermore, the rate of respiration for drug-treated cells was not statistically different from the rate for untreated cells within statistical error. The plots for the drug-treated cells showed no systematic deviation from linearity, confirming that none of the drugs inhibited respiration during the 100 min when [O 2 ] was measured. In contrast, the addition of NaCN inhibited cellular respiration 82.5% during the 100 min incubation. On performing the same experiments with all of the Pt compounds but changing the cell line to Tu183 and HL-60 cells, similar results were observed; that is, there was no inhibition occurring in Pt drug-treated cell respiration after approximately 2 h exposure.
The next experiment was performed to confirm and extend the above observations. Jurkat cells (0.5 Â 10 6 cells/ml) were incu-   [23]. For 40 lM cisplatintreated Jurkat cells, respiration rates appear to be significantly lower than those for untreated cells after 12 h, implying that the inhibition of mitochondrial O 2 consumption occurred earlier for the higher concentration. The decrease in respiration is more marked after 15 h. The inhibition I may be calculated as 1 minus the ratio of k for treated cells to k for untreated cells (with the average of two values being used for untreated cells). Then, for the incubation times in Table 3 Table 3. There is no statistically significant trend in the values of hki/hk o i, although they are all significantly lower than unity. Thus, the effect of tirapazamine at these concentrations is too small to be accurately measured.
To study the effect of higher concentrations of tirapazamine on HL-60 cell respiration, 10 6 cells/ml were incubated with various concentrations of the drug for 3 h. Fig. 4B shows O 2 consumption  Values of hki were calculated according to Eq. (3) with d = 0 and t f À t i = 60 min. The results are shown in Table 3  From the k values obtained from the linear fitting, we calculated I = 1 À k/k 0 . Results are shown in Fig. 5B, with the best fit to Eq. (5) shown as a solid line and with I max = 0.272 and A = 57.9 nM (r 2 = 0.994). When I max = 1 (i.e., 100% inhibition) was assumed, the best value of A turned out to be 1150.3 nM, but r 2 = 0.598, so this fitting cannot be adopted.

Caffeine
We investigated the effect of pure caffeine (powder formulation without vehicle) on HL-60 cellular respiration. The cells (10 6 cells/ ml) were suspended in medium plus 2 lM Pd phosphor and 0.5% albumin with the addition of 0.1, 0.5, 1.0, or 2.0 mM caffeine. Caffeine was dissolved in the medium and added to the cell suspension immediately prior to O 2 measurement. Measurements were made alternately on the four samples. Results are shown in Fig. 6A. Here k values were obtained from the slopes of the best linear fits because going to quadratic fits produced no improvement. For the four concentrations of caffeine, we found k = 0.843, 0.726, 0.727, and 0.676 lM O 2 /min (r 2 = 0.966, 0.941, 0.929, and 0.946, respectively). From the four k values, linear extrapolation to 0 concentration produced k 0 = 0.8090 ± 0.0364. This value of k 0 was used to calculate the inhibitions for the four concentrations, which were À0.042, 0.102, 0.101, and 0.165, respectively. Fig. 6B shows results of two separate experiments to test whether the effect of caffeine on HL-60 respiration was immedi-ate. After O 2 measurement proceeded for approximately 40 min, caffeine was injected into the vials and O 2 measurement was continued. The results for 1 mM caffeine are shown by empty and filled triangles in Fig. 2B, with best-fit lines (before and after caffeine addition) shown as dashed. The k value before addition was 0.883 lM O 2 /min and was reduced instantly to 0.787 lM O 2 /min after addition (11% inhibition). For 3.0 mM caffeine (data not shown), the k value before addition was 1.028 lM O 2 /min and was reduced instantly to 0.568 lM O 2 /min after addition (45% inhibition).
All of the inhibitions from the experiments of Figs. 6A and 6B are plotted versus caffeine concentration in Fig. 6C. Here I = 1 À k/k o . The best-fit line is also shown; it comes very close to passing through (0,0) as expected. There is no sign of saturation (i.e., leveling off in the plot of I vs. [caffeine]) in this range of concentrations, so no attempt was made to fit to Eq. (5).

Discussion
The use of respiratory monitoring is proposed here to assess the extent of drug-impaired cellular function and how it depends on parameters such as incubation time and dosage. To obtain respiration rates, we measure [O 2 ] using phosphorescence decay (other methods could also be used [26,27]). Although different drugs attack different cellular targets such as mitochondria, DNA, and thiols, their inhibitory effects on the mitochondrial function sometimes appear to be similar. Clearly, answering the question of why a particular drug gives a specific cellular respiration pattern requires more understanding of the drug-cell interaction than is currently available. However, differences in oxygen consumption profiles (Figs. 1-6) indicate different mechanisms of action for dif- ferent drugs. The diverse profiles make it important to characterize the drug effect by a single parameter such as the inhibition I.
To measure the extent of inhibition of mitochondrial oxygen consumption, various cell lines were exposed continuously to a variety of cytotoxic drugs. For testing, the cells and the cytotoxic compounds were sealed in a closed vessel that also contained growth medium, sufficient glucose (as a respiratory substrate), and the Pd phosphor. Untreated cells consume O 2 at a constant rate, whereas drug-treated cells exhibit a diversity of O 2 consumption profiles (linear or nonlinear).
Oxygen concentration is fit to a linear (Pt compounds and CsA) or quadratic (dactinomycin, doxorubicin, and tirapazamine) function of time. Fitting to a higher order, such as cubic or polynomial, can be applied on the condition that the fit is significantly improved (i.e., r 2 significantly increased). The rate of respiration k is the negative of the slope of a plot of [O 2 ] versus time as calculated from the fitting function. The time average respiration rate is proposed as a characteristic parameter. It is obtained by integrating k over a well-defined time interval. Then the inhibition I is calculated as 1 À hki/hki 0 , where hki 0 is the average value of k for cells not treated by drug. It is important to obtain hki and hki 0 from measurements on the same batch of cells. For the most meaningful comparison, the measurements of [O 2 ] should be done simultaneously (i.e. alternately) on treated and untreated cells.
When I has been measured for a series of drug concentrations D, it is convenient to fit the results to the two-parameter function, I = I max D / (D + A). The value of I provides a simple and convenient characterization of drug cytotoxicity. (Other measures of drug cytotoxicity continue to be suggested [28][29][30].) It also helps to predict the dosing necessary to shut down respiration of malignant cells and, thus, may be useful in assessing the efficiency of anticancer drugs.