Many methods in flow cytometry rely on staining DNA with a fluorescent dye to gauge DNA content. From the relative intensity of the fluorescence signature, one can then infer position in cell cycle, amount of DNA (i.e., for sperm selection), or, as in the case of flow karyotyping, to distinguish individual chromosomes. This work examines the staining of murine thymocytes with a common DNA dye, Hoechst 33342, to investigate nonlinearities in the florescence intensity as well as chromatic shifts.
Murine thymocytes were stained with Hoechst 33342 and measured in a flow cytometer at two fluorescence emission bands. In other measurements, cells were stained at different dye concentrations, and then centrifuged. The supernatant was then used for a second round of staining to test the amount of dye uptake. Finally, to test for resonant energy transfer, we measured fluorescence anisotropy at two different wavelengths.
The fluorescence of cells stained with Hoechst 33342 is a nonlinear process that shows an overall decrease in intensity with increased dye uptake, and spectral shift to the red. Along with the spectral shift of the fluorescence to the longer wavelengths, we document decreases in the fluorescence anisotropy that may indicate resonant energy transfer.
Flow cytometers typically measure scattering and fluorescence intensities from cells and other particles with an accuracy of 1–2%. Accuracies of this order are necessary for measurements such as cell cycle (1, 2), sperm sex selection (3), and flow karyotyping (4–7). In all of these techniques, the measured fluorescence is the result of binding of fluorescent dyes to DNA. A critical question for flow cytometric measurements, therefore, is whether the fluorescence intensity accurately represents the cellular dye content. A natural extension of this inquiry is to investigate to what degree nonlinearities in the fluorescent intensity versus dye concentration affect the estimation of DNA content.
As early as 1985, Watson et al. (8) and Smith et al. (9) noticed anomalous chromatic effects when chicken thymocytes were stained with Hoechst 33342. In particular, the fluorescence intensity in two different wavelength regions was noted to shift from violet to green. The color shift has since generated considerable interest, since this spectral shift of Hoechst 33342 fluorescence can be exploited for the isolation of hematopoietic stem cells (10, 11), and as a marker for apoptotic cells (9, 12, 13). These studies showed that hematopoietic stem cell populations are consistently bluer and dimmer than their more committed counterparts when stained with this dye. Studies characterized this spectral shift and the decrease in fluorescence intensity at high dye concentrations, and putatively assigned the spectral shift and decrease in fluorescence as being the result of secondary, weaker binding to alternative sites (14–16).
The binding of Hoechst 33258 and its analog Hoechst 33342 to DNA in situ has been shown to display the following characteristics: 1) The fluorescence intensity centered at 460 nm increases with increasing dye concentration, up to approximately 1 nmol/ml, after which the fluorescence intensity decreases (15); 2) Upon binding to DNA, the fluorescence yield increases approximately 60-fold, resulting in increased fluorescent intensity (17); 3) The emission spectrum of bound Hoechst 33342 is centered at 460 nm, but the emission spectrum of unbound dye is red-shifted 50 nm to 510 nm (17); and 4) Hoechst 33258 binds with highest affinity to (AT)4 base pairs with a particular preference for AATT sequences. Measured binding constants for this family of dyes to (AT)4 sequences is on the order of 2 × 108/mol (16, 18–21).
While binding of the dye molecules to alternative sites can explain some of these phenomena, it cannot account for the decrease in blue (460 nm) fluorescence intensity at higher concentrations. If in fact the red spectral shift is due only to secondary binding to a weaker site, there is no reason to expect the blue fluorescence to decrease. On the contrary, one would expect the blue fluorescence to increase until all of the blue binding sites are filled, whereupon the fluorescence intensity should plateau. Fluorescence self-quenching through reabsorption of the blue fluorescence emission is also unlikely, because of the relatively large Stoke's shift (352-nm excitation, 461-nm emission). To account for the observed decreases in blue fluorescence (20%), the intracellular dye concentration would have to be on the order of 0.2 mol.
In light of these observations, we have measured Hoechst 33342 binding to murine thymocytes to investigate the spectral shift of the fluorescence and the incorporation density of the bound dye. In one set of experiments, we stained murine thymocytes with different amounts of Hoechst 33342, and measured the blue and red fluorescence in a flow cytometer. The remaining sample was centrifuged and the supernatant was used for a second round of staining to determine the concentration of dye in the supernatant. These measurements confirmed the decrease in blue fluorescence with increasing dye concentration, but also revealed that the ratio of the blue and red fluorescence is an accurate indicator of intracellular dye concentration.
We hypothesized that the decrease in blue fluorescence could be indicative of resonant energy transfer (RET) between dye molecules bound to the DNA. To test this assertion, we measured the fluorescence anisotropy of both the red and the blue fluorescence from Hoechst 33342 bound to murine thymocyte DNA. Previous work has shown that RET is accompanied by a decrease in fluorescence anisotropy (22). Measurements of both the blue and red fluorescence of Hoechst 33342 show that the fluorescence anisotropy decreases with increasing dye concentration.
MATERIALS AND METHODS
Measurements were performed using a high-speed cell sorter shown schematically in Figure 1, which was designed in the laboratory of Ger van den Engh, and which is the basis for a research instrument sold commercially (Cytopeia, Seattle, WA). This cell sorter has been described previously (17, 23). Briefly, the sorter has three independent light paths, allowing excitation by up to three laser sources. Each of these light paths can accommodate several photomultiplier tubes. For the purpose of the experiments reported here, the central light path has been adapted to focus light onto the entrance slits of a scanning monochromator. The monochromator is scanned continuously with a potentiometer that is part of a voltage divider circuit attached to its lever arm to measure the wavelength associated with a given event. In this manner, the emission spectra of bound or unbound dye can be constructed from multiple events (17). The preamplifiers used to measure the fluorescence emission are equipped with a baseline restoration circuit that subtracts the background signal from unbound dye from the signals generated by the cells. This technique, combined with a very small measurement volume and a low quantum yield from unbound dye, makes the contribution of unbound dye to the measured fluorescence intensities negligible. To record fluorescence anisotropy, a polarizer (03PTA401; Melles Griot, Irvine, CA) was placed in the optical pathway, directly in front of a photomultiplier. The polarizer was rotated by hand over an angle of 240°, while a potentiometer in a voltage-divider circuit produced a voltage that varied linearly with the angle of the polarizer. This voltage was digitized and recorded each time a measurement was performed. The data from these measurements were fitted to a theoretical curve to calculate the fluorescence anisotropy (24).
Measurements were performed at room temperature using sterile 0.9% NaCl sheath fluid and a nozzle with a 70-μm aperture. The cytometer was configured to measure emission spectra or fluorescence anisotropy in one of the light paths, and red and blue fluorescence in a separate light path. The red emission was measured using a 680-nm longpass filter and a red-sensitive PMT (Hamamatsu R928); the blue fluorescence channel measured light in a band from 418–480 nm (PMT Hamamatsu 931B, Middlesex, NJ). Further details of the experimental setup are shown in Figure 1. The cytometer uses two argon ion lasers (Coherent model 305C, Santa Clara, CA) to produce up to 300 mW of multiline UV light to excite the dye.
Thymus tissue was harvested from 8–12-week-old C57BL/6 mice (Charles River Laboratories, Wilmington, MA), suspended in 2 ml of phosphate buffered saline (PBS) and homogenized using a Ten Broeck (Fisher Scientific, Pittsburg, PA) tissue grinder. The Institute for Systems Biology's Institutional Animal Care and Use Committee approved all animal protocols. The homogenized mixture was filtered though 40-μm mesh and the cells were enumerated using a hemacytometer. Cells were then centrifuged and resuspended at a concentration of 106/ml in PBS. For staining, the cells were permeabilized with 0.025% Triton-X (ICN Biomedicals, Inc., Irvine, CA) before the addition of Hoechst 33342 (Molecular Probes, Eugene, OR). Cells were allowed to equilibrate for 30 min at 37°C before being put on ice. For measurements using the supernatant for staining, we initially stained 2 ml (2 × 106 cells) at a given concentration of dye. After equilibration, 400 μl of sample was removed for measurements in the flow cytometer. The remaining sample was centrifuged, and 1 ml of the supernatant removed. To this aliquot, an additional 106 thymocytes were added from a concentrated stock. Once again, the cells were allowed to equilibrate for 30 min at 37°C before a small portion (400 μl) was removed for measurements in the cytometer. The remainder was once again centrifuged, and 400 μl of the supernatant was removed for use in a third round of staining.
Figure 2 shows the emission spectra of unbound Hoechst 33342 and of Hoechst 33342 bound to DNA measured in a cell sorter using the spectrometer shown in Figure 1. The unbound spectrum is collected from a 5-μmol solution of Hoechst 33342 in PBS, while the bound spectrum is from murine thymocytes stained with the same dye concentration. It is difficult to compare the amplitudes of these two spectra directly, because the bound dye spectrum is computed from the measured pulse heights of passing nuclei, whereas the unbound spectrum is a quasi-dc measurement. However, one may note qualitatively that the fluorescence intensity from Hoechst 33342 increases (by noting the relative sizes of the laser line at ≈350 nm for the two spectra) and that the spectrum is blue-shifted some 50 nm upon binding to DNA. Also shown in Figure 2 are the regions of the spectrum that are collected by the red and blue fluorescence PMTs (Fig. 1).
Figure 3 shows the fluorescence intensity from the blue and red region of the spectrum versus the ratio of Hoechst 33342 dye molecules to DNA base pairs for a solution of 106 thymocytes per milliliter. Because the bound fraction is not known, the data are plotted as the number of dye molecules per DNA base pair. The approximate maximal blue fluorescence occurs when there are 0.1 dye molecules per DNA base pair. At this point, the blue fluorescence plateaus before decreasing somewhat, while the red fluorescence increases up to a concentration of approximately one dye molecule per DNA base pair before saturating. Note also that the graph of both the blue and the red fluorescence intensity increases fairly linearly for dye molecules/DNA base pair ratios below these points.
Referring to Figure 3, one can make the following qualitative observations. First, the character of the blue fluorescence intensity changes at ratios above 0.1 dye molecules/base pair. Second, the red fluorescence intensity also undergoes a similar saturation at ratios above one dye molecule/base pair, although the red fluorescence intensity lacks the subsequent decreases seen in the blue fluorescence, and all attempts to measure at ratios higher than three dye molecules/base pair inevitably led to the loss of cell integrity.
To determine the bound/free dye ratio, the supernatants of the first experiments were incubated with fresh, unstained thymocytes. A comparison of the amplitude and shape of the fluorescence intensity curve for the supernatants gives an indication of the bound dye fraction. Figure 4 represents the intensity of the red and blue fluorescence channels as a function of the staining concentrations for the supernatant (i.e., the free dye that did not bind to the DNA) from the first series of experiments. For comparison, the original staining curves are also shown in each panel. It is important to note that neither the blue fluorescence nor the red fluorescence increase monotonically, which limits their utility for such a comparison. However, the shift of the concentration-dependent fluorescence intensity curve gives a good estimate of the bound dye fraction.
The fluorescence intensity from both the blue and red channels is not monotonic. Therefore, it is difficult to estimate the intracellular dye concentration from a simple measurement of either the blue or the red fluorescence intensity. Figure 5 shows that the ratio of the blue fluorescence to red fluorescence varies linearly with the log of the concentration. We have fit these data using a linear function that minimizes the sum of the squares of residuals. Using this fit, the blue/red ratio may be approximated by R = (–0.215 ± 0.0178)log([H33342]) + (0.458 ± 0.043), where R is the ratio of the blue fluorescence to the red fluorescence. The absolute values of the slope and the intercept hold true only for particular PMT gains, but do not affect the linearity of the relationship between the fluorescence ratio and the logarithm of the dye concentration. This graph demonstrates that this ratio is a much better indication of the intracellular dye concentration than either the blue or the red fluorescence alone. It is unclear whether this relationship between the fluorescence ratio and the concentration of dye molecules is coincidental or is a consequence of the underlying mechanism of the chromatic shift.
Fluorescence Anisotropy Measurements
To test our assertion that the decrease in blue fluorescence, as well as the increase in red fluorescence, may be due to RET, we measured the polarization of both the red and blue fluorescence as a function of the dye concentration. For a polarized, anisotropic source, the intensity of the signal is expected to vary with the polarizer angle as
with the polarization, P, defined as
The dimensionless parameters α and β are the intensities one would expect with the pass band oriented parallel to and perpendicular to, respectively, the dominant polarization axis. The maximal value of P for fluorescence is 0.5, while for scattering it can be as high as 1 (24). Figure 6 depicts a typical measurement of fluorescence anisotropy for a Hoechst 33342 to DNA ratio of 0.3 along with a least-squares fit to equation 1. Each measurement consisted of 10,000 individual measurements that were binned and averaged, before fitting them to the theoretical fluorescence anisotropy. For the data shown, the parameters α and β are determined to be α = 982.7 ± 11.24, and β = 1964.9 ± 13.73. As a result, the measured fluorescence anisotropy is calculated to be P = 0.325 ± 0.0081. Results for the complete set of polarization measurements are shown in Figure 7. In both the red and blue fluorescence channels, the fluorescence anisotropy decreases with increasing dye concentration.
While only a few studies have looked at Hoechst 33342 binding, there is a considerable body of knowledge regarding the binding of its analog, Hoechst 33258, and its affinity for AT-rich portions of the DNA. At low concentrations, the consensus of these studies is that Hoechst 33258 binds along the DNA minor groove to four contiguous AT base pairs with a high affinity for AATT sequences followed by TAAT, ATAT, TATA, and finally TTAA (21). Hoechst 33258's preference for these sequences arises from the narrower minor groove in these regions, allowing tighter conformation, and from steric interference between the dye molecule and GC sequences (25). The association constants for the five (AT)4 sequences listed above range from 5.2 × 108/mol to 0.027 × 108/mol (21). All of these inquiries were performed at low concentrations ([H33258] ≈1 pmol/ml) and a shift in the fluorescence spectrum was not noted. At higher concentrations, these dye molecules have been noted to bind to other DNA sequences with lower affinity (26).
From Figure 4, we note that the maximum of the blue fluorescence occurs between 0.05 and 0.1 dye molecules per base pair for the original solution and at 0.3 molecules per base pair in the supernatant. A similar shift can be seen in both the red and blue fluorescence at concentrations below this point. We use the shape of the concentration-dependent fluorescence intensity curve to estimate the concentration of the dye in the supernatant when it is used to stain cells a second time. From this analysis, we conclude that for cells stained with a 1 nmol of Hoechst 33342 per milliliter of solution (≈0.1 dye molecules per DNA base pair), approximately 0.2 nmol of dye is left in the supernatant, and the remainder (0.8 nmol) is bound to DNA. For a solution of thymocytes with 1 × 106 cells/ml, the number of DNA base pairs in solution is 6 × 1015. Using the DNA base pair concentration to estimate the number of potential binding sites, the calculated affinity for this binding is on the order of 4.4 × 105/mol. This affinity likely overestimates the number of potential minor groove binding sites. Crystallographic studies have shown that the Hoechst dye molecule touches five or six base pairs when it binds to the minor groove (18). As a result, the number of binding sites that can be occupied in this fashion (without overlap of the dye molecules) is approximately 1.6 nmol. The calculated binding affinity using this estimate for the total number of available binding sites is 5 × 106/mol, consistent with low affinity binding reported previously (21). The number of preferred (AT)4 binding sites is much smaller; for a random distribution of DNA base pairs, the five (AT)4 binding sites would comprise only 2% of the total. Our data are consistent with lower affinity binding to random base pair sequences, as there is a notable change in the blue fluorescence at staining concentrations of 1 nmol/ml (0.1 dye molecules/DNA base pair). Additionally, measurements of the supernatant show that a larger number of dye molecules are bound to the DNA than could be bound by the (AT)4 sites alone.
At concentrations higher than 0.1 dye molecules per DNA base pair, even more dye molecules appear to bind to the DNA. The red fluorescence intensity saturates at concentrations greater than 0.5 dye molecules per base pair in the original staining and at concentrations greater than 3 dye molecules per base pair in the supernatant. This observation implies that the DNA could be sequestering as many as 2.5 molecules per base pair. The mechanism for this binding is unclear, but the saturation of the red fluorescence at dye concentrations above one dye molecule per DNA base pair suggests that dye molecules may be binding to the DNA base pairs in a one-to-one ratio. One may also observe that it seems counterproductive to stain cells with four to five times more dye molecules than there are DNA minor-groove binding sites, but the anomalous chromatic shifts that are used to enrich for hematopoietic stem cells occur at these concentrations (10). These findings are consistent with the notion of secondary binding sites for Hoechst analogs to DNA sequences other than the preferred (AT)4 sites (16, 21, 25, 26).
Resonant Energy Transfer
One possibility for the chromatic shift from blue to red at high dye/DNA base pair ratios is RET. This phenomena is described in detail elsewhere (27, 28), and has been exploited as a spectroscopic ruler. The RET effect is characterized by the nature of the dipole–dipole interaction, the relative orientation of the dipoles, and the degree of overlap of the donor molecule's emission spectrum with the acceptor molecule's excitation spectrum. A characteristic distance, known as Förster's radius, quantifies the degree of overlap of the emission and excitation spectra and the relative orientation of the interacting dipoles. At this separation, one-half of all excited molecules transfer their energy to the donor molecules via this mechanism. There is no requirement that the molecules have to be different; homologous RET can occur between molecules with small Stoke's shifts. If the excitation and emission spectra of a donor and acceptor dye molecule are known, then Förster's radius can be calculated using in Å (28). In the preceding, κ2 is a geometric factor that relates the relative orientation of the interacting dipoles [κ2 = 2/3 for randomly oriented dipoles], n is the index of refraction of the media [1.4 for cells], Q is the quantum efficiency of the donor molecule, and J(λ) is the overlap integral defined as
F(λ) is the emission spectrum of the donor molecule normalized to unit area, ε(λ) is the acceptor molecule's extinction coefficient in units [mol–1cm–1], and for our purposes, λ is in [nm]. The emission and extinction spectra for Hoechst are readily obtained (http://www.molecularprobes.com), and one can calculate that Förster's radius for this dye is approximately 15 Å. The calculation above assumes a random orientation of the interacting dipoles. When bound to DNA, the dipoles are not distributed randomly, but are distributed about the DNA axis. The turns of the DNA helix, however, effectively randomize the orientation of the bound molecules; computer simulations have shown that the geometrical term predicted for minor groove binders is somewhat smaller than that of a random distribution (29).
In a free dye solution such as the ones typically used to stain DNA, the molecules are well outside the critical distance for energy transfer to take place. At a 10 nmol per ml, the average spacing between molecules is 7.32 × 10–8 m, or 732 Å. When bound to DNA in the cell nuclei, however, the concentrations can be much higher. We have shown previously that for an initial staining concentration of 3 nmol/ml, approximately 0.5 nmol remains in the supernatant, and, as a result, there are approximately 1.5 × 109 dye molecules sequestered in each cell nucleus. For a nucleus approximately 3 μm in radius, the volume is 1.13 × 10–16 m3. The 1.5 × 109 dye molecules, therefore, each occupy a volume of 7.53 × 10–26 m3, so that the average distance between molecules is 2.62 × 10–9 m, or 26.2 Å. At first glance, this would seem to rule out the possibility of RET, as the efficiency of energy transfer over this distance is . The volume calculation above, where the dye molecules are distributed throughout the nucleus, provides an upper bound on the separation between dye molecules. If dye molecules are distributed stochastically with a probability that obeys Poisson statistics, at an average spacing of 26.2 Å, 17.1% of the dye molecules will have another dye molecule within Förster's radius. Dye molecules bound to the nuclear DNA will be even closer together. DNA base pairs are separated by 3.4 Å, so for an average occupation of 0.25 dye molecules per base pair, one could expect the equilibrium spacing to be on the order of 14.2 Å. These estimates provide an upper and lower bound to the average separation between dye molecules. In either case, there is considerable evidence that the distances between dye molecules are sufficiently small that RET is expected to have an effect.
RET and Fluorescence Depolarization
If RET is taking place, there will be measurable changes in the fluorescence polarization. Many flow cytometers use highly polarized light sources, such as lasers, to induce fluorescence. The fluorescence resulting from a linearly polarized light source will itself be somewhat polarized because the laser cannot excite all excitation dipoles with equal efficiency (24). For a perfectly polarized excitation source (P = 1), the fluorescence emission will have a maximum fluorescence anisotropy P = 0.5. Molecular motion, RET, and other factors serve to reduce this number further so that in practice the fluorescence anisotropy is somewhat less than P = 0.5. Previous studies have shown that the fluorescence anisotropy decreases with increasing dye concentrations, with RET as the likely mechanism for the decrease (22). With this in mind, we measured the fluorescence anisotropy of the red and blue portions of the emission spectra and found that the fluorescence anisotropy decreased by approximately 15% in both the red and blue portions of the spectrum. There are two hypotheses for the decrease in the fluorescence polarization that are consistent with these results. One explanation for the fluorescence depolarization is an increase in the mobility of the bound fluorophores so that the emission dipole of the bound dye molecule is less well correlated with the excitation dipole as a result of molecular motions. Free dye molecules in solution are predicted by the Perrin equation (28) to show very little fluorescence polarization. This scenario, in contrast, would require the dye molecules to be bound in such a way that their fluorescence polarization is only somewhat less than is expected for a tightly bound molecule. A more likely explanation for the fluorescence depolarization is energy transfer between closely spaced molecules as this effect becomes apparent at distances on the same order as Förster's radius.
We have measured the red and blue fluorescence intensity from murine thymocytes stained with Hoechst 33342 in an attempt to understand the mechanisms underlying the blue-to-red shift seen in these cells with increasing dye concentration. With increased Hoechst 33342 concentration, the overall blue fluorescence plateaus and then decreases above 1 nmol/ml, while the red fluorescence increases up to approximately 10 nmol/ml before saturating. The nonlinear nature of the fluorescence makes it difficult to use the fluorescence intensity as an indicator of the intracellular dye concentration. However, we have shown that the ratio of the blue fluorescence to the red fluorescence is an accurate indicator of the intracellular dye concentration from 0.1–30 nmol/ml. Because the there is an intimate connection between the blue-red fluorescence intensity ratio and the intracellular dye concentration, one would expect a shift in blue/red ratio with dye uptake. Figure 8 shows the Hoechst 33342 blue fluorescence intensity versus red fluorescence intensity for the data presented in Figure 3. Flow cytometric bivariate dot plots of the Hoechst blue versus red fluorescence should follow the curve shown in this figure as the intracellular dye concentration increases. A fit to these data is also shown to allow smooth interpolation between different dye concentrations. As dye moves across the cellular membrane, the blue/red fluorescence ratio should follow this curve. These binding patterns are indeed observed with real cell samples (10). A popular method for enriching and isolating hematopoietic stem cells uses this property, and the measurements presented in this work elucidate the mechanisms responsible for the different red/blue fluorescence ratios.