Bromodeoxyuridine (BrdU) cell cycle analysis using flow cytometry is of clinical interest for making treatment decisions or for predicting response and survival, through proliferation rate (labeling index or S-phase fraction) assessment or Tpot calculation. Thymidylate synthase expression was tested in vitro, in vivo, and clinically as a prognostic factor for 5-fluorouracil (5FU) sensitivity. However, results were still controversial. Moreover, we had reported that 5FU sensitivity was related to the labeling index of untreated cell cultures.
We used six human cancer cell lines that exhibited a wide range of 5FU sensitivity. Cell cycle analysis was performed using flow cytometry monovariate propidium iodide (PI) analysis and bivariate distributions of BrdU incorporation versus DNA content. 5FU sensitivity was assayed using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) colorimetric assay.
In all cell lines, 5FU exposure resulted in a statistically significant G1/S accumulation. No statistically significant relationship was seen between G0/G1 delay determined by monovariate analysis and 5FU sensitivity. However, 5FU sensitivity was statistically correlated to the labeling index and G1/S subpopulation assessed with bivariate analysis using BrdU incorporation versus DNA content.
5-fluorouracil (5FU) resistance has been studied extensively. Many mechanisms have been implicated, such as pharmacokinetic resistance, decreased accumulation of activated metabolites, and altered effects on thymidylate synthase (TS; 1,2). Cell cycle disturbance and apoptosis induction also result in 5FU resistance (3–7). TS protein or mRNA expression predictive value was tested clinically in many tumoral pathologies (8–11). However, although the prognostic value of TS was accepted in colorectal tumors, its significance remains debated (9, 12–14). In head and neck pathologies, TS expression was unrelated to 5FU treatment outcome (15, 16). Other parameters were tested to predict 5FU response, but none were universally used or accepted (15, 17). More recently, cell cycle parameters (18) and apoptosis-related factors were implicated in the 5FU response (19, 20).
The study of human tumor proliferation has been facilitated by the development of monoclonal antibodies (mAbs) that recognize halogenated pyrimidines, such as iododeoxyuridine (IdU) or bromodeoxyuridine (BrdU), followed by cytometric analysis. Pyrimidine analogs are incorporated into DNA during S-phase. The cell kinetic information generated by this approach is more informative than that obtained from in vitro incubation with DNA precursors such as tritiated thymidine or by single-parameter DNA analysis (21). The most widely used proliferation parameters are labeling index (LI) and potential doubling time (Tpot). Tpot is defined as the time that would be required by a tumor to double in cell number in the absence of cell loss (22). It can be calculated in vivo by flow cytometric analysis of a tissue biopsy obtained after a single infusion of a thymidine analog, such as IdU or BrdU, which can be detected using a specific mAb, avoiding the use of radioactive isotopes (22, 23). Proliferation index (S-phase fraction or LI) and DNA ploidy cytometric analysis are of clinical interest for making treatment decisions or for predicting response and survival (24–28). As we previously described in vitro, basal LI values are related to 5FU cytotoxicity (18). The present study was designed to compare cell cycle cytometric analysis using monovariate propidium iodide (PI) or bivariate BrdU versus PI labelling, either before or after 5FU exposure. In vitro LI and G1/S subpopulation evaluated using BrdU incorporation are related to 5FU sensitivity, both before and after 5FU treatment.
MATERIALS AND METHODS
Materials and Chemicals
Cell culture materials were purchased from Costar (Dutscher, Brumath, France). Culture media and additives were obtained from Life Technologies (Gibco BRL, Cergy-Pontoise, France), except for fetal calf serum (FCS), which was obtained from Costar. Anti-BrdU mAb was provided by Dako (Trappes, France). All other chemicals were purchased from Sigma (St. Quentin Fallavier, France) and were of molecular biology grade.
CAL51 human breast adenocarcinoma, PANC3 pancreas carcinoma, and CAL27 and CAL33 human head and neck carcinoma cell lines were kindly provided by Dr. J.L. Fischel (Centre Antoine Lacassagne, Nice, France). FaDu and KB, head and neck carcinoma cell lines, were obtained from Pr. A. Hanauske (Munich University, Germany) as part of the EORTC Preclinical Therapeutic Models Group exchange program. All cell lines were grown in 75-cm2 plastic tissue culture flasks in RPMI 1640 medium supplemented with 10% heat-inactivated FCS, penicillin (100 IU/ml), and streptomycin (100 μg/ml) in a 37°C, 5% CO2 atmosphere. Exponentially growing cells were exposed to equitoxic 5FU concentrations (IC50) for 24 h and analyzed immediately.
5FU-induced cytoxicity was assessed using MTT assays according to a procedure previously reported (18). Briefly, cells were seeded at the initial density of 2 × 104 cells per milliliter in 96-well microtitration plates. Seventy-two hours after plating, cells were exposed to 5FU concentrations ranging from 0.08 to 4 × 104 μM, for 24 h. After 5FU exposure, 50 μl of 0.5% 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) solution was added to each well and incubated for 3 h at 37°C to allow MTT metabolization. The formazan crystals were dissolved by adding 50 μl per well of 25% sodium dodecyl sulfate solution and vigorous pipetting. Absorbance was measured at 540 nm using a Multiskan MCC/340 plate reader (Labsystem, Cergy-Pontoise, France). Results were expressed as relative absorbance to untreated controls. 5FU concentrations yielding 50% growth inhibition (IC50) were calculated using a median-effect algorithm (29) and were expressed as the mean values of five independent experiments.
PI Cell Cycle Distribution Analysis
Cell cycle distribution was measured before and after 5FU exposure at an IC50 value. Cell suspensions were washed with phosphate-buffered saline (PBS), resuspended in 0.1% sodium citrate, 0.1% Triton X100, and 50 μg/ml PI, and stored for 24 h at 4°C. After centrifugation at 1,500 rpm for 5 min, the cells were resuspended in PBS containing 250 μg/ml RNAse. Monovariate distributions of cell number versus DNA content (PI) were analyzed using an Orthocyte flow cytometer (Ortho Diagnostic Systems, Roissy, France) equipped with a xenon lamp and a filter set for excitation at 488 nm. PI fluorescence intensity was recorded through 575-nm high pass filters. At least 20,000 events were collected in each final gated histogram. Cell cycle analysis was performed using Dean and Jett's algorithm (Multicycle, Phoenix Flow Systems, San Diego, CA).
BrdU Cell Cycle Analysis
Samples were processed using flow cytometry according to the method reported by Marchal et al. (30). First, 200 μM BrdU was added directly to the culture medium for 20 min followed by two washes with PBS. Cell suspensions were prepared by trypsination and resuspended in cold PBS. While being vortexed, the samples were fixed by the addition of 2 ml cold 70% ethanol for storage at -20°C. Single nuclei suspensions were prepared by resuspending fixed cells in 0.1 N HCl and by incubating for 15 min in 2 N HCl. After three washes with PBS, nuclei were labeled with 40 μg/ml anti-BrdU mouse mAb in PBS containing 0.5% normal rabbit serum, 0.5% Tween 20. After 1 h of incubation at room temperature, samples were washed with PBS, resuspended in the PBS/serum/Tween 20 solution, and stained with 20 μg/ml fluorescein isothiocyanate (FITC)-conjugated rabbit anti-mouse immunoglobulin serum. After 1 h of incubation at room temperature, nuclei were washed twice with PBS and 25 μg/ml PI solution in PBS was added. At least 50,000 events were collected in each final gated histogram.
Bivariate distributions of BrdU content (FITC) versus DNA content (PI) were analyzed. FITC and PI fluorescence intensities were, respectively, recorded through a 520/530-nm band pass and 575-nm high pass filters. The data were analyzed using Multi2D software (Phoenix Flow Systems). LI corresponded to percentage of BrdU-positive cells. The G1/S subpopulation, corresponding to BrdU-positive cells containing G1 DNA, was determined (Fig. 1).
Unless indicated, all data are mean values ± SD calculated from at least four independent experiments. Spearman's rank correlation was used to test for the relationship between the different parameters. The Mann-Whitney U test was used to test for the significance level between independent variables. All experimental values (and not the mean value) were used for statistical analysis.
A wide range of 5FU sensitivity was found within the cell lines, with IC50 values ranging from 0.16 to 22.62 mM, respectively, for the CAL51 and PANC3 cell lines (Table 1).
Table 1. Cell Cycle Fractions in Untreated Control and After 5FU Exposure (IC50) Assessed by Cell Number Analysis Versus DNA Content (PI)†
Significant variations versus untreated control (P < 0.05).
Results were mean values of four experiments ± SD.
As demonstrated by PI analysis, after 5FU exposure, all cell lines were able to delay the G1 phase (Fig. 2). G0/G1 accumulation ranged from 1.1 to 1.3-fold, respectively, for the PANC3 and CAL51 cell lines. Nevertheless, only CAL51, KB, and FaDu cell lines, the most 5FU-sensitive lines, displayed a statistically significant higher accumulation in G1 phase (Table 1, Fig. 2). The S-phase fraction ranged from 33% to 48% in control cells and was unchanged (KB, CAL33, CAL27, and PANC3 cell lines) or slightly decreased after 5FU exposure (S-phase fraction decreased by a factor of 0.77 for the FaDu cell line). The G2/M phase was reduced more markedly, almost completely disappearing for the KB cell line (Table 1, Fig. 2). Using PI analysis, no statistically significant correlation was seen between 5FU cytotoxicity and cell cycle subpopulation, neither before nor after 5FU treatment.
BrdU Cell Cycle Analysis After 5FU Exposure
BrdU analysis showed that all cell lines were delayed at the G1/S boundary (Fig. 3) instead of at the G1 phase as displayed by PI analysis. Compared with the LI in untreated controls, LI increases for all cell lines by a factor of 1.4–2.9 after 5FU exposure (Table 2). The G1/S fraction increased significantly after 5FU treatment in all cell lines, with values ranging from 2.8 to 8.0-fold, respectively, for the CAL51 and KB cell lines. The LI increase was the consequence of G1/S cell accumulation, with the G1/S fraction ranging from 53% to 69% of the BrdU-labeled cells. The G1/S fraction was the most prevalent subpopulation in BrdU-labeled cells (Table 2, Fig. 3). LI data were correlated to the G1/S cell fraction for untreated controls (r = 0.706, P = 0.0121) and 5FU-treated cells (r = 0.923, P = 0.0001; Fig. 4). The G2/M phase was abolished almost completely, as demonstrated by PI analysis (Fig. 3).
Table 2. LI and G1/S Population in Untreated Control and After 5FU Exposure*
Analysis was assessed by bivariate analysis of BrdU content versus DNA content. G1/S population was determined as described in Figure 1. Results were mean values of four experiments ± SD.
52 ± 4
14 ± 1
74 ± 4
39 ± 6
33 ± 3
8 ± 1
95 ± 1
64 ± 2
41 ± 4
12 ± 2
87 ± 7
48 ± 7
41 ± 4
8 ± 1
59 ± 11
40 ± 7
32 ± 2
11 ± 1
63 ± 4
39 ± 6
33 ± 1
7 ± 1
47 ± 3
25 ± 1
5FU sensitivity was statistically correlated to LI (r = 0.61, P = 0.0004 and r = 0.67, P = 0.0016, respectively, for controls and 5FU-treated cells; Fig. 5). 5FU sensitivity was also correlated to the G1/S subpopulation before (r = 0.68, P = 0.0033) and after 5FU treatment (r = 0.50, P = 0.0294; Fig. 5).
Comparison of S-Phase Subpopulation as Determined by BrdU or PI Cell Cycle Analysis
These two analytical methods for S-phase evaluation were closely correlated (r = 0.92, P = 0.010; Fig. 6) and no statistically difference was noted between LI values and S-phase fraction by PI analysis (data not shown). This correlation was, however, demonstrated exclusively for untreated cells and not after 5FU exposure (Fig. 6A,B, respectively). LI was actually overestimated widely in BrdU analysis, leading to values of up to 90% for the FaDu or KB cell lines (Table 2).
Could 5FU-Induced Cell Cycle Arrest be Rescued by BrdU?
In order to test if BrdU pulse (20-min exposure) could lead to cell progression from late G1 to early S-phase, some control experiments were performed using the KB and CAL51 cell lines, respectively, displaying the lowest and the highest increase in LI and in the G1/S subpopulation after 5FU exposure. Using BrdU analysis, no difference was seen when BrdU was added during the last 20 min of 5FU exposure or after the 5FU exposure with the washing phase between 5FU and BrdU incubation (Table 3). PI experiments with BrdU incubation did not show any influence of BrdU on cell cycle distribution (Table 3). In conclusion, BrdU incubation was neither responsible for the cell progression from G1 to early S-phase nor for the increase of LI or the G1/S phase fraction.
With regards to clinical response to 5FU-based therapy, only TS expression was evaluated as a prognostic factor (8–11). Although TS expression analysis has demonstrated potential prognostic significance in patients with advanced colorectal cancer (12, 31), its clinical usefulness remains in question. Prospective studies are still required to confirm its importance in adjuvant chemotherapy. TS enzymatic activity can be altered by genomic polymorphism or mutation, phosphorylation level, and subcellular localization. These do not necessarily modify TS protein level but can affect TS prognostic value (32–35). TS is regulated throughout the cell cycle phase (36, 37).
Recent studies demonstrate that cell cycle kinetic studies have clinical value in predicting 5FU sensitivity. S-phase fraction has clinical utility for patients with breast cancer (28) and Tpot can predict patients with a low probability of achieving long-term local control with conventional fractionation of radiotherapy in head and neck carcinoma (26, 28, 38–40). However, standardization and quality control of these parameters have to be improved before Tpot can be implemented for routine use in community settings (28, 41, 42).
Cell cycle distribution or kinetics parameters can also be used to predict 5FU sensitivity because fluoropyrimidine treatment leads to cell cycle arrest (3, 43). In our panel of cell lines, 5FU equitoxic treatment results in G0/G1 accumulation as demonstrated by PI analysis (Table 1, Fig. 2). Cell cycle phase subpopulation was not statistically correlated to 5FU sensitivity. Conversely, significant G0/G1 accumulation induced after 5FU treatment was only observed for the most 5FU-sensitive cell lines (Table 1). However, BrdU analysis showed that a cell cycle delay took place specifically at the G1/S interface as previously reported (44, 45; Table 2, Fig. 3).
As BrdU is a thymidine analog that can be used by cells for DNA synthesis, it can lead to progression from late G1 to early S-phase. Therefore, the influence of BrdU incubation during or after 5FU exposure was assessed both by PI and BrdU analysis (Table 3). Using either PI or bivariate BrdU analysis, no BrdU influence on cell cycle distribution was seen. Therefore, the G1/S subpopulation cannot be considered to be derived from late G1 during BrdU incubation. In fact, inhibition of TS by 5FU was almost irreversible and BrdU could only rescue new synthesized TS molecules. A 20-min BrdU incubation was not enough long to initiate new DNA synthesis.
The LI was statistically correlated to 5FU sensitivity as we have previously described (18), both before and after 5FU exposure (Fig. 5A1,A2). The G1/S subpopulation was also strikingly correlated to 5FU sensitivity (Fig. 5B1,B2). In addition, considering the G1/S interface (Fig. 1), cytometric analysis with bivariate staining analysis (BrdU versus PI) compared with the single-parameter analysis of DNA content (PI) did not provide the same information. Consequently, these two methods will not have the same usefulness for the determination of prognostic indicators. In fact, BrdU analysis showed that G1 cells, as determined by PI analysis, was composed of BrdU-positive and BrdU-negative cell subpopulations (Figs. 1, 3). These subpopulations cannot be distinguished using PI analysis alone. This was the consequence of the cell fraction that was able to incorporate BrdU and maintain its G1 DNA content. This fraction was considered to be in G1 by PI analysis, whereas it was included in the labeled subpopulation by BrdU analysis. Thus, cells performing repair synthesis of DNA, with no net DNA increase, would be included in the LI score, but could still have a DNA content that was indistinguishable from G1 cells. The number of BrdU-positive cells within this peak increased with duration of 5FU exposure and was related to p21 mRNA expression induction (data not shown). On the other hand, differences in molecular mechanisms gave rise to the existence of these two subpopulations. For example, cyclin D is required for progression through early to mid-G1. Following this, expression and activity of cyclin E increase at the G1/S phase border and result in S-phase progression. Cyclin D and E cooperate temporally by successive pRb phosphorylation (46). Synchronization of cells, caused by a delay in G1/S, may be related to programmed cell death initiation. Induced delay in G1/S was representative of apoptosis involvement and of 5FU treatment sensibility. Actually, chromatin condensation and endonuclease-mediated DNA fragmentation were detected using fluorescent microscopy in the most 5FU-sensitive cells (data not shown). The amount of cells displaying DNA fragmentation was also related to 5FU sensitivity (data not shown). 5FU exposure leads to an LI increase. However, LI modifications were only related to G1/S phase accumulation, and not to a real enhanced S-phase fraction. In conclusion, cell cycle parameters using BrdU analysis could be used as predictive markers for 5FU response to complement the predictive value of TS.