Energy Transfer between Fluorescein Isothiocyanate and Propidium Iodide – A Problem in the Estimation of Tpot with the Bromodeoxyuridine–DNA Flow Cytometry Technique?

Energy transfer in flow cytometry can occur when two fluorochromes are bound in close proximity (generally within 100 Å) and the emission spectrum of one fluorochrome overlaps significantly with the excitation spectrum of the other. The latter criterium is fullfilled for the fluorochromes fluorescein isothiocyanate and propidium iodide and also the former when they, e.g., are used in bromodeoxyuridine – DNA flow cytometry methods. In the present growth kinetic study using this method, we show that energy transfer does take place between fluorescein isothiocyanate and propidium iodide which results in a detected increase in DNA content with 2–3%. Despite the erroneous increase in the obtained DNA content values, this does not seem to have any influence on the calculation of DNA synthesis time and potential doubling time where the DNA content, based on the relative movement principle of the labelled cells, is used.


Introduction
Bromodeoxyuridine (BrdUrd) is a thymidine analog that is used in various applications in the study of cell proliferation, including the elucidation of growth kinetic parameters [3,6,12,16,22]. In one application, cells in vitro or in vivo are pulse-labelled with BrdUrd. The cells are then either fixed directly after labelling or after a period of continued proliferation in an environment free of BrdUrd. In order to analyse the incorporated BrdUrd, it is labelled with primary monoclonal BrdUrd antibodies and secondary fluorescein isothiocyanate (FITC)-conjugated antibodies. In order to determine the cell cycle position of unlabelled and of BrdUrd-labelled cells, the total cell population is then stained with the DNA intercalating dye propidium iodide (PI). Since BrdUrd can only be detected in singlestranded DNA and PI can only bind to double-stranded DNA, the DNA is subjected to partial denaturation before the staining procedure. Thus, part of the DNA will be single-stranded and part of it double-stranded. The cells are analysed by flow cytometry (FCM), most often using an argon ion laser with excitation line at 488 nm by which both fluorochromes are excitable. The maximum emission of FITC fluorescence is at 520 nm, with a range of 480-600 nm and that of PI is at 630 nm, with a range of 550-700 nm [14]. The quantity of emitted FITC and PI fluorescence is proportional to the amount of incorporated BrdUrd and cellular DNA contents, respectively. In addition to being excited at 488 nm, PI has an excitation maximum at 520 nm [14] which is within the emission range of the FITC fluorescence and corresponds to its maximum emission. Thus, in any experiment using the combination of the fluorochromes PI and FITC, the latter may contribute to the excitation of the former, a phenomenon which is called energy transfer [4,11]. Energy transfer can occur when the separation between the two molecules is within 100 Å and can be observed as a loss of characteristic fluorescence from the donor molecule and an increase in the characteristic fluorescence of the acceptor molecule [4,7]. There are other conditions that have to be fulfilled for energy transfer to occur, e.g., the steric orientation of the dyes [11,18]. Considering the fact that both FITC and PI are bound to DNA when applied to detect incorporated BrdUrd and DNA content, there is a probability that both the distance and steric considerations are fulfilled for energy transfer to occur.
In the present investigation, we show that energy transfer indeed takes place between FITC and PI when using the BrdUrd/FCM method to study growth kinetic parameters. We discuss the impact of this on the interpretation of growth kinetic data, since energy transfer between FITC and PI would result in an increase in observed PI fluorescence. The consequence of that would be that BrdUrd-labelled cells containing FITC molecules as a means of localizing BrdUrd, would seemingly have a higher DNA content than unlabelled cells. An important step in the measurement of the DNA synthesis time (T S ) and potential doubling time (T pot ) with the BrdUrd/FCM method is the estimation of relative movement (RM), i.e., the increase in DNA content, of the BrdUrd-labelled cells a certain time after BrdUrd labelling. RM is used in the formulas to calculate T S , which in turn is included in the expressions of T pot . Thus, the aim was to study whether the energy transfer phenomenon has any impact on the T pot measurement.

Cultured cell lines, bromodeoxyuridine labelling, and experimental designs
Chinese hamster ovary (CHO) cells, V79 hamster fibroblast cells, MCF-7 and SK-BR-3 human breast adenocarcinoma cell lines were maintained in serial passages in RPMI 1640 medium containing 10% foetal calf serum (FCS) and antibiotics (penicillin and streptomycin). Insulin (10 µg/ml) was also added to the MCF-7 growth medium. CHO and V79 cells were subcultured twice a week, while MCF-7 and SK-BR-3 cells were sub-cultured once a week. The cultures were incubated at 37 • C in a water-saturated atmosphere consisting of air with 5% CO 2 .
For the experiments, a number of replicate cultures, consisting of plateau phase cells seeded at a low density, were set up. When the cells reached exponential growth, BrdUrd (1 mM dissolved in phosphatebuffered saline) was added to achieve a final concentration of 5 µM. Labelling and subsequent cell handling were carried out under subdued light. After 30 min of incubation at 37 • C, the BrdUrd containing medium was aspirated, the cells were rinsed twice with medium (37 • C) containing 0.5% FCS (CHO, V79) or 1% FCS (MCF-7, SK-BR-3). Finally, growth medium (37 • C) of the respective cell line was added and the cultures were further post-incubated at 37 • C. In order to evaluate the effect of different BrdUrd concentrations, an experiment was designed in which CHO cells were labelled with 1, 5, 10, 20, or 50 µM BrdUrd for 30 min. At various times after labelling (sampling times), cultures were trypsinized, the cell numbers determined, and the cells pelleted at 500g for 5 min at 4 • C. The pelleted cells were fixed by resuspension in ice-cold 70% ethanol to a final concentration of 1−2 × 10 6 cells/ml and the samples were stored at -20 • C until analysis. Some cultures were fixed directly after labelling. For some applications, the entire population of CHO cells was labelled for 48 hours with 5 µM BrdUrd before fixation.

Preparation of cells for simultaneous flow cytometry-mediated quantification of DNA content and incorporated bromodeoxyuridine
The method was originally described by Dolbeare et al. [5] and further developed and modified by Schutte et al. [13] and van Erp et al. [19]. The preparation procedure has been described in detail previously [9]. Briefly, after washing the cells, the DNA of the nuclei was partly denatured and then the nuclei were incubated with monoclonal anti-BrdUrd antibodies (diluted 1 : 10) followed by incubation with FITCconjugated anti-mouse antibodies (diluted 1 : 20). Finally, nuclear isolation medium containing 10 µg/ml PI [17] was added for staining of the nuclear DNA.
In order to more accurately determine the DNA content of various cell populations, PI-stained nuclei from a human diploid tumour, used as an internal DNA reference, was added to the PI-and FITC-stained nuclei before FCM analysis in some experiments.
In order to evaluate whether there was an interaction between FITC and PI, the secondary biotin-conjugated anti-mouse antibody (diluted 1 : 200) was used in some experiments. This antibody is identical to the FITCconjugated one, with the exception of the conjugate.
In one experiment, the BrdUrd-labelled nuclei were not stained with PI before FCM analysis and thus only the green fluorescence was measured. After FCM analysis, the nuclei were stained with PI and run through the FCM again to study the impact of PI on the green fluorescence signal.

Flow cytometric analysis
The FCM analysis and handling of the data have previously been described in detail [9]. Two different flow cytometers were used, either an Ortho Cytoron Absolute flow cytometer (Ortho Diagnostic Systems, Raritan, NJ, USA) equipped with a 15 mW air-cooled argon-ion laser, or a Cytofluorograph System 50-H (Ortho Instruments, Westwood, MA, USA) equipped with a 4 W argon-ion laser (output, 200 mW) (Lexel Corp, Palo Alto, CA, USA). The excitation line of 488 nm was used for both flow cytometers. PI fluorescence was separated from FITC fluorescence by a 560 nm dichroic mirror. The FITC fluorescence was quantified in the wavelength interval of 515-545 nm defined by a bandpass filter. The PI fluorescence beyond 600 nm (Cytofluorograph System 50-H) respective 620 nm (Ortho Cytoron Absolute) was limited by a lowpass filter. Otherwise the filter settings were the same. The computerized analysis of stored data was performed either with Multi 2D R and MultiCycle R software programs (Phoenix Flow Systems, San Diego, CA, USA) (Ortho Cytoron Absolute flow cytometer) or with Model 2140, DG MP 200 (Ortho Instruments and Data General, Westboro, MA, USA) (Cytofluorograph System 50-H). The obtained fluorescence signals were the same irrespective of which flow cytometer was used.
Because of the partial spectral overlap between the PI and FITC fluorescence emissions (PI into FITC), a proportional reduction of the PI fluorescence measured in the FITC fluorescence optics was done electronically. The FITC fluorescence into the PI fluorescence was blocked by the 600 and 620 nm lowpass filters in the PI fluorescence optics for respective flow cytometer.

Interpretation and obtainable parameters of flow cytometry-mediated data
The interpretation of FCM data has previously been described in detail [1,9].
In some experiments, narrow windows were set in the middle of the DNA distribution of the G 1 and G 2 phases [8], in order to estimate the fraction of BrdUrdlabelled cells in the middle of each phase (Fig. 1).
In order to investigate if energy transfer could affect growth kinetic parameters, T S and T pot were calculated. Fig. 1. Cytogram with mid-G 1 (Region 1) and mid-G 2 (Region 2) phase windows, which were set after determination of the mean DNA contents of G 1 and G 2 phase cells in the corresponding DNA histogram of the total cell population. CHO cells were labelled with 5 µM BrdUrd for 30 min and then allowed to progress through the cell cycle for 4 hours before fixation and processing to detect Br-dUrd and DNA contents with FCM. The BrdUrd-labelled cells in G 1 and G 2 phases are seen to the right of the windows in the respective phase. Note that the BrdUrd-labelled divided cells in G 1 phase, appearing as a narrow and vertical cluster, does not lean to the right (i.e., show any increase in DNA content) with increasing BrdUrd content. This implies that no leakage of FITC into the PI detector occurs.
The formulas by Johansson et al. [8] to estimate RM and T S (T SJ ) were used. T pot was calculated according to Steel [15] and LI according to Johansson et al. [10].

Results
In an experiment studying the fraction of BrdUrdlabelled cells in mid-G 1 and mid-G 2 phases, we serendipitously found that the peaks representing unlabelled and BrdUrd-labelled cells in G 1 phase did not coincide with respect to relative mean DNA content. To study the fraction of BrdUrd-labelled cells in mid-G 1 and mid-G 2 phases, narrow windows were set manually in the BrdUrd cytogram (Fig. 1), the positions of which were determined in the DNA histogram of the total cell population. We consistently observed that the G 1 and G 2 phase windows were found to the left of the BrdUrd-labelled cells in G 1 phase, as well as to left of the BrdUrd-labelled cells in G 2 phase. The implication of this is that BrdUrd-labelled cells seemed to have a higher relative DNA content than unlabelled cells.
We then incubated CHO cells for 48 hours with 5 µM BrdUrd to obtain a cell population where all cells were labelled. These cells were firstly not subjected to the denaturation procedure but directly stained with PI. Comparing the DNA histograms of those cells with a population of unlabelled cells, there was no difference in the relative DNA content of G 1 phase cells of unlabelled and BrdUrd-labelled populations (not shown). Neither was there any difference in the DNA content of the same cell populations when they were subjected to denaturation before staining with PI without the antibodies (not shown).
Thereafter we proceeded to compare the unlabelled and BrdUrd-labelled cells in a cell population that had been pulse-labelled with BrdUrd for 30 min, i.e., the population contained both unlabelled and BrdUrdlabelled cells. The cells were prepared for the detection of DNA and BrdUrd contents with PI and with FITC. The DNA histograms of unlabelled and BrdUrdlabelled cells were compared concentrating on the positions of, i.e. the mean DNA value of, G 1 phase cells (Fig. 2), since this is the cell cycle phase from which the most reliable data can be obtained from a statistical point of view. A ratio between the peak values of BrdUrd-labelled and unlabelled G 1 phase cells was calculated for the cell lines CHO, V79, MCF-7, and SK-BR-3. In all samples, it was found that the BrdUrdlabelled G 1 phase cells were found to have a slightly higher DNA content than the unlabelled cells, giving The ratio between the mean DNA content of BrdUrd-labelled G 1 phase cells and that of unlabelled G 1 phase cells was 1.03 ± 0.001 (n = 62). a quota of 1.03 ± 0.001 (mean ± standard deviation (SD), n = 62). The quota was the same in CHO cells incubated with 1, 5, 10, 20, or 50 µM BrdUrd.
To test whether the increase in relative DNA content in BrdUrd-labelled cells was due to energy transfer, we compared the relative DNA contents of BrdUrdlabelled G 1 phase cells that after the incubation with the primary BrdUrd antibody were incubated either with the secondary FITC-conjugated antibody or with a secondary biotin-conjugated antibody (Fig. 3). In these experiments we included separately PI-stained human diploid tumour cells as an internal DNA reference. The results show that there is a shift in the G 1 peak to a higher relative DNA content in cells that were incubated with FITC-conjugated antibodies, in relation to those that were incubated with biotinconjugated antibodies. In this experiment too, the ratio was 1.03, when comparing mean DNA contents of FITC-conjugated and biotin-conjugated G 1 phase cells, which is equivalent to the ratio between the mean DNA contents of BrdUrd-labelled and unlabelled G 1 phase cells.
A characteristic of energy transfer is that the fluorescence of the donor molecule decreases as the fluorescence of the acceptor molecule increases. In order to test this, FITC-labelled nuclei were analyzed in the flow cytometer before and after staining with PI. The results showed that the green fluorescence signal decreased about 6 times after staining with PI. This net green fluorescence corresponds very well to any BrdUrd experiment in general.
When studying growth kinetics, a number of formulas exist to calculate T S and T pot . In the calculation of these parameters, the mean relative DNA contents of G 1 , G 2 , and BrdUrd-labelled S and G 2 + M phase cells are included. It is possible, that the increase in DNA content caused by energy transfer may affect the outcome of T S and T pot . In order to partly exclude such a possibility, we compared T S using the mean relative DNA content of the G 1 phase cells of the total cell population, of the unlabelled cells, and of the BrdUrdlabelled cells (Table 1). We found no difference in T S and T pot .

Discussion
We have been using the BrdUrd-DNA FCM method to study growth kinetics of experimental tumours and cells in culture for almost a decade. When evaluating DNA histograms of unlabelled and BrdUrd-labelled cells, obtained after detection of DNA and BrdUrd contents, we have been confounded by the fact that there seemed to be a difference in DNA content, based on the PI fluorescence, between these two populations although they were part of the same total cell popu- Table 1 DNA synthesis time (T SJ * ) and potential doubling time (Tpot * * ) in MCF-7 cells calculated with the mean DNA content of G 1 phase cells obtained in different ways Mean ± SD 17.5 ± 1.2 39.6 ± 3.8 1 7 .6 ± 1.2 39.9 ± 3.7 1 7 .1 ± 1.0 38.5 ± 3.2 * T SJ , T S calculated according to Johansson et al. [8]. * * Tpot = T SJ /LI J , LI J calculated according to Johansson et al. [10]. 1 The mean DNA content of the G 1 phase cells obtained from the total cell population. 2 The mean DNA content of the G 1 phase cells obtained from the BrdUrd-labelled cell population. 3 The mean DNA content of the G 1 phase cells obtained from the unlabelled cell population.
lation. The BrdUrd-labelled cells always had a DNA content that seemed to be greater than the unlabelled cells after DNA denaturation and BrdUrd detection using FITC-conjugated secondary antibodies. We have observed this phenomenon in all cell lines we have worked with. We envisaged several causes for this observation, including energy transfer, and we will discuss those below.
Before doing so, it should be mentioned that a discussion on a change in DNA content may be based on the entire cell population and/or on cells in a selected cell cycle phase. We have mainly based our comparative studies on cells in the G 1 phase, simply because that is the phase where the most reproducible and accurate results may be obtained, using either manual or computer-based region settings. In some experiments we have included an internal DNA reference that was not affected by the preparation procedure to ascertain that the differences found were indeed not caused by the preparation.
The incorporated BrdUrd molecule as such could have a positive impact on the binding of PI, somehow promoting the binding of more PI molecules. This was not the case since the G 1 DNA content of a cell population totally labelled with BrdUrd was the same as that of a totally unlabelled cell population, when stained with PI without denaturation. The incorporated BrdUrd molecules could affect DNA denaturation in a negative manner, resulting in a slightly lower degree of DNA denaturation in BrdUrd-labelled cells than in unlabelled cells. This was not the case either, since the G 1 DNA content of a cell population totally labelled with BrdUrd was the same as a totally unlabelled cell population when stained with PI after the DNA denaturation step. The presence of primary and secondary antibodies could promote the binding of more PI molecules. The fact that the use of the secondary biotin-conjugated antibody did not result in an increase in the DNA content rules out this possibility.
Neither was there a change in DNA content when only the primary antibody was used. Thus, it is our notion that the difference in DNA content seen in unlabelled and BrdUrd-labelled cells, where BrdUrd is detected with the use of FITC-conjugated antibodies and DNA with PI, is caused by energy transfer. In addition to the increase in PI-derived red fluorescence, we found that the FITC-derived green fluorescence decreased, which also is characteristic of energy transfer. Notably, the decrease in the green fluorescence signal was much larger than the increase in the red fluorescence signal. There may be other physical factors than energy transfer involved in this phenomenon, e.g., bleaching and/or quenching as described by Shapiro [14].
Although the increase in DNA content was persistent in several cell lines, it was not of a large magnitude. The ratio between the mean DNA content of BrdUrd-labelled G 1 phase cells and that of unlabelled G 1 phase cells, where BrdUrd is detected with the use of FITC-conjugated antibodies and DNA with PI, was 1.03 ± 0.001. In an attempt to see whether this increase was dependent on the amount of incorporated BrdUrd, we labelled CHO cells with BrdUrd concentrations ranging from 1-50 µM. There was no difference in the magnitude of the shift in the mean DNA content of BrdUrd-labelled G 1 phase cells relative to that of unlabelled G 1 phase cells (not shown). These results may imply that even if there is an increase in the amount of incorporated BrdUrd molecules, only a maximum amount of them impart a positioning of the secondary FITC-conjugated antibody with respect to distance and steric orientation so that an interaction with PI can occur. Presumably that occurs around the boundaries of native and denatured DNA.
Although our discussion has been based on the G 1 phase cells, it is our assumption that a shift also takes place in the BrdUrd-labelled S and G 2 phase cells, when BrdUrd is detected with the use of FITCconjugated antibodies and DNA with PI. It is possible to measure the DNA content of unlabelled and BrdUrdlabelled G 2 phase cells although with a lower degree of accuracy and reproducibility than can be done with the G 1 phase cells. However, it is not possible to measure the mean DNA content of a population of BrdUrdlabelled cells moving through S phase by other means than by first detecting them with antibodies towards BrdUrd.
It should be mentioned that the observed effect of an increase in the DNA content of BrdUrd-labelled cells subjected to denaturation and labelled with the FITC antibodies and PI could also be the result of spectral overlap. However, we have taken careful measures to prevent the spectral overlap of FITC fluorescence leaking into the PI detector, by the optical filter settings as described in Section 2. A means to detect that no further compensation for spectral overlap is needed is that the cluster of BrdUrd-labelled divided cells in G 1 as they appear in the cytogram would be leaning to the right, the more FITC the higher DNA content. In addition, the coefficient of variation of the G 1 peak would be increased. We see neither of these phenomena, thus indicating that there is no leakage of FITC fluorescence into the PI detector (Fig. 1).
In our experiments we are not using the maximum laser effect which may imply that the FITC and PI fluorochromes are not maximally excited thus permitting the additional excitation of PI by the FITC emission fluorescence. By using a higher laser effect, the risk of energy transfer may diminish. However, commercial flow cytometers are not equipped with high effect lasers, of technical and economic reasons. Furthermore, the current laser power used is sufficient to produce excitation, resulting in fluorescence light intensities which are easily detected by the optics.
In the calculations of T SJ , the RM of the moving cohort of BrdUrd-labelled cells is used [8]. In the calculation of RM, the mean DNA contents of BrdUrdlabelled S and G 2 + M, G 1 , and G 2 phase cells are used [2]. The mean DNA contents of G 1 and G 2 phase cells are usually determined in the DNA histogram of the total cell population while that of the BrdUrd-labelled cells is determined in the DNA histogram of BrdUrdlabelled cells. In the present study, we have used the relative mean DNA contents of G 1 phase cells obtained in the DNA histogram of the total cell population and in the DNA histograms of unlabelled cells and of BrdUrd-labelled cells, for the calculations of T SJ . Fortunately, the differences in mean relative G 1 values were not reflected in the obtained T SJ values. However, the greatest source of error due to energy transfer is probably in the determination of the mean DNA content of BrdUrd-labelled cells in S and G 2 + M phases. When the mean DNA contents of G 1 and G 2 phase cells are determined in the DNA histogram of the total cell population, it will be influenced by the mean DNA contents of unlabelled cells and BrdUrd-labelled cells in G 1 and G 2 phases, and also by the relative numbers of unlabelled and labelled cells. Presumably, the mean DNA contents of G 1 and G 2 phase cells of the total cell population will not deviate significantly from either the mean DNA contents of G 1 and G 2 of the unlabelled or labelled cells. On the other hand, the mean DNA content of the BrdUrd-labelled cell population is only based on a population where energy transfer may occur. This may imply that T SJ [8] and any other formula to calculate T S [2,20,21] where the mean DNA content of the BrdUrd-labelled S phase cells is used, underestimates T S , which will result in an underestimation of T pot . We are currently trying to theoretically solve the degree of underestimation that might take place. A means to solve this experimentally would be to use fluorochromes with excitation and emission spectra that do not overlap.
In conclusion, our data show that energy transfer takes place between FITC and PI when a secondary FITC-conjugated antibody is used to detect incorporated BrdUrd and PI is used to stain total cellular DNA. This reasoning does not only hold for the detection of BrdUrd and DNA, but in any application using these two fluorochromes bound in close proximity. Fortunately, energy transfer between FITC and PI in the Br-dUrd/FCM method does not seem to have any impact on the calculation of T S and T pot .