Texture Analysis of Fluorescence Lifetime Images of Nuclear DNA with Effect of Fluorescence Resonance Energy Transfer

Abstract

Background:

Fluorescence lifetime imaging microscopy (FLIM) is becoming an important tool in cellular imaging. In FLIM, the image contrast is concentration insensitive, whereas it is sensitive to the local environment and interactions of fluorophores such as fluorescence resonance energy transfer (RET).

Methods:

Fluorescence microscopy, lifetime imaging, and texture analysis were used to study the spatial distribution of fluorophores bound to nuclear DNA. 3T3-Swiss albino mice fibroblast nuclei were labeled with Hoechst 33258 (Ho), an AT-specific dye, and 7-aminoactinomycin D (7-AAD), a GC-specific dye. Ho is a RET donor to the 7-AAD acceptor.

Results:

Texture analysis of 50 alcohol-fixed nuclei quantitatively showed changes of spatial distribution of apparent donor lifetimes. RET increased the spatial heterogeneity in the phase and modulation lifetime images. In most of the doubly stained cells (about 80%), the phase and modulation lifetime distributions were spatially homogeneous. In about 20% of the cells, we noticed that lower phase and modulation lifetimes caused by RET were correlated with regions of high Ho intensity in the nuclei.

Conclusions:

The spatial lifetime heterogeneity of Ho in presence of 7-AAD seems to be caused by RET between closely spaced strands in the three dimensionally condensed regions of DNA.

Fluorescence staining of DNA is widely used in spectroscopy (1–4) and in microscopy (5,6). Fluorescent dyes bound to DNA are also widely used in DNA sequencing (7,8), for genetic analysis by fluorescence in situ hybridization (9–12), and in analysis of DNA structure using resonance energy transfer (RET; 13–16). Time-resolved studies of probes bound to DNA have shown that the probe lifetimes often increase upon binding to DNA (17,18) and that the decay times are sensitive to RET (19,20).

During the past 10 years, a new method of cellular imaging has become available, fluorescence lifetime imaging microscopy (FLIM; 21–26). In FLIM, the image contrast is independent of the local intensity of the stained cell; instead, it is based on the fluorescence lifetime in each region of the cell. Lifetime-based contrast is valuable because the lifetimes of probes often depend on their local environment or the presence of a specific analyte. For instance, the lifetimes of the calcium-specific probes Quin-2 and Calcium Green are known to increase upon binding to calcium. For cells labeled with Quin-2 or Calcium Green, the lifetime images reveal the spatial distribution of calcium and are mostly independent of the local probe concentrations (24,25).

In this study, we extend the concept of FLIM to imaging of cell nuclei from 3T3 fibroblasts. The nuclei were labeled with either Hoechst 33258 (Ho) or with both Ho and 7-aminoactinomycin D (7-AAD). Ho binds specifically to AT-rich regions of DNA and 7-AAD binds preferentially to GC-rich regions of DNA (27,28). Ho is a RET donor to 7-AAD. The chosen pair of Ho and 7-AAD shows a good spectral match required by RET (29). The Förster distance of this donor-acceptor pair is 40 Å (28). We measured intensity and lifetime images of Ho in the absence and presence of the 7-AAD acceptor. These images were subjected to texture analysis (30,31) to evaluate the effect of RET from Ho to 7-AAD on the angular second moment, sum variance, and difference variance in the intensity and lifetime images. The effects of RET on these parameters of texture analysis were then interpreted in terms of spatial proximity of Ho and 7-AAD in the nuclear DNA.

MATERIALS AND METHODS

Cells and Culture Conditions

Mouse fibroblasts (3T3-Swiss albino) were grown at 37°C in the bottom-glass dishes (Mat Tek, Ashland, MA) containing Dulbecco’s modified Eagles’s medium with 10% calf serum.

Fixation and Cell Staining

The cells in the dishes were fixed in 70% ethanol (4°C, at least 30 min). After rinsing, the cells were stained by 0.4 μM Ho and 0.8 μM 7-AAD, which were used as donor and acceptor, respectively. Ho and 7-AAD were obtained from Molecular Probes (Eugene, OR). All the experiments were carried out at room temperature in 10 mM Tris-HCl buffer, pH 7.5, containing 100 mM NaCl.

Measurement Condition of Fluorescence Intensity and Lifetime Imaging

The principles of operation of the homodyne FLIM technique were described previously (21). The apparatus was built around the Axiovert 135TV inverted fluorescence microscope (Carl Zeiss, Thornwood, NY). The picosecond light pulses at 335 nm from the frequency-doubled synchronously pumped DCM dye laser were directed to the microscope using a quartz optical fiber. The epi-illumination was accomplished by the Zeiss C-Apochromat water immersion objective, 40×/1.2 w, with a 1.6× Optovar insert and the Zeiss FT 395 dichroic beam splitter. Fluorescence from the sample was isolated by the 450DF65 interference filter (Omega Optical, Brattleboro, VT) and quantitized using a C5825 high-speed modulated image intensifier (Hamamatsu, Bridgewater, NJ) and a PXL scientific-grade, slow-scan, cooled CCD camera (Photometrics, Tucson, AR). The gain of the intensifier was modulated at 113.202 MHz using an output of the PTS 300 synthesizer (Programmed Test Sources, Littleton, MA). The synthesizer was phase locked to the master-oscillator of the pumping Ar⁺ laser mode-locker. For each lifetime measurement, eight images with the detector phase equally spaced over 360° were acquired and analyzed by a method described elsewhere (25). The illumination level of the sample was decreased by a neutral-density filter to the level that the acquisition time for the images was in the order of seconds. No detectable photobleaching was observed under these conditions.

Measurement of Fluorescence Intensity and Lifetime Imaging

The positions of 50 cells on a bottom-glass dish were recorded and assigned numbers. The cells were stained by 0.4 μM Ho and their fluorescence intensity and lifetime imaging were measured. Then 0.8 μM 7-AAD was added to the same dish for fluorescence RET measurement. The fluorescence intensity and lifetime images of the same 50 cells double stained with Ho and 7-AAD were measured individually.

Texture Analysis Using Co-Occurrence Matrix

The program for texture analysis was written in C language. The program calculated 14 texture parameters from the co-occurrence matrices (30), which were constructed from gray-level images after reduction from 16 to 8-bit gray levels and normalization to the same average values. The following three parameters were selected as independent descriptors of the Ho staining pattern. The angular second moment (Eq. 1) shows spatial homogeneity, the sum variance (Eq. 2) reveals spatial heterogeneity, and difference variance (Eq. 3) describes the contrast of the Ho staining pattern. These values were calculated using:

$$ASM=\sum _{i=1}^{Ng}\sum _{j=1}^{Ng}\{P(i,j){\}}^2$$ (1)

$$SVar=\sum _{k=2}^{2Ng}{\left(k-\sum _{k=2}^{2Ng}k\cdot P_{x+y}(k)\right)}^2\cdot P_{x+y}(k)$$ (2)

$$DVar=\sum _{k=0}^{Ng}{\left(k-\sum _{k=0}^{Ng}k\cdot P_{x-y}(k)\right)}^2\cdot P_{x-y}(k)$$ (3)

where P_x+y(k) and P_x–y(k) are defined as follows (30).

$$P_{x+y}(k)=\underset{i+j=k}{\sum _{i=1}^{Ng}\sum _{j=1}^{Ng}P}(i,j)k=2,3\dots ,2Ng$$ (4)

$$P_{x-y}(k)=\underset{|i-j|=k}{\sum _{i=1}^{Ng}\sum _{j=1}^{Ng}}P(i,j)k=0,1,\dots ,Ng-1$$ (5)

The term P(i, j) is the (i,j)th entry in a normalized gray-tone spatial-dependence matrix (co-occurrence matrix) and Ng is the number of distinct gray levels in the quantized image. These three parameters (ASM, SVar, and DVar) are calculated from every nuclear image of the 50 cells, which were stained by Ho in the absence and presence of 7-AAD.

RESULTS

The nuclei stained with Ho, the AT-specific dye, were strongly fluorescent and the images were detectable without significant nonspecific background fluorescence from the cytoplasm. intensity images of Ho-stained nuclei are shown in Figures 1 and 2. The nuclei displayed regions of higher and lower intensity. Given the base specificity of Ho, the higher intensity regions probably correspond to regions of higher AT base pair concentrations in the nuclei. In contrast to the intensity images, the phase and modulation images were flat and within resolution of our microscopy images did not display significant spatial variations. This result is consistent with the known behavior of Ho, which prefers AT-base pairs and then displays the same lifetimes whenever bound to DNA.

FIG. 1.

Nuclear images of fluorescence intensity and lifetime in cell stained with Ho in the presence or absence of 7-AAD. After the addition of 7-AAD, the lifetime images became slightly more heterogeneous and the cell showed weak correlation between fluorescence intensity and lifetime.

FIG. 2.

Fluorescence intensity and lifetime images of another cell stained with Ho in the presence and absence of 7-AAD. After the addition of 7-AAD, the lifetime image becomes strongly heterogeneous. This cell shows more obvious correlation between fluorescence intensity and lifetime than the cell in Figure 1.

Addition of the acceptor 7-AAD to the nuclei resulted in a decrease in the intensity and lifetime of the DNA-bound Ho (29; Figures 3–5). Following the addition of 7-AAD, 80–90% of the cells displayed a flat lifetime image by both phase and modulation (Fig. 1). This result indicates that, on average, the 7-AAD molecules are found and distanced comparable to the Forster distance R₀ from Ho. In interphase cells, which were used in this study, the chromosomes with AT- and GC-rich regions are three dimensionally intermingled (32). For this three-dimensional distribution of chromosomes and our acceptor concentration, the 7-AAD molecules are found and distanced comparable to the Forster distance R₀ from Ho as described in Murata et al. (29). Additionally, a flat lifetime image suggests that there is no significant spatial difference in the AT to GC base content in the nuclei to within the resolution of the microscopy images.

FIG. 3.

Histograms and co-occurrence matrices of fluorescence intensity images from the cell shown in Figure 1. The cell was stained with Ho in the presence and absence of 7-AAD. Co-occurrence matrices were constructed from images normalized to the same average fluorescence intensity.

FIG. 5.

Histograms and co-occurrence matrices of phase lifetime images from the cell shown in Figure 1. Cell was stained with Ho in the presence and absence of 7-AAD. Co-occurrence matrices were constructed from images normalized to the same average phase lifetime.

The remaining 10–20% of the nuclei displayed spatially heterogeneous phase and modulation lifetimes (Fig. 2). More specifically, the phase and modulation lifetimes are lower in regions of higher Ho intensity. Apparently in this subpopulation of cells, the extent of RET from Ho to 7-AAD is higher in regions of locally higher AT concentrations.

To obtain a quantitation measurement of spatial heterogeneity in the nuclei, we used texture analysis (30,31). For this purpose, the intensity phase and modulation images were normalized (Figs. 3–5). The intensity histograms show that 7-AAD decreased the Ho intensity, which is a good indication for the occurrence of RET. Visual examination of the normalized intensity histogram shows no discernible difference in the absence or presence of 7-AAD (Fig. 3). The presence of 7-AAD results in a modest decrease in the range of intensities, as seen from the decrease in length of the ASM scattergram from NW to SE (Fig. 3). There also appears to be a minor decrease in contrast, as seen from a slightly decreased width in the perpendicular direction.

Different results were observed for the effects of 7-AAD on the phase and modulation lifetimes (Figs. 4, 5). The unnormalized histograms show a decrease in the phase and modulation lifetimes due to 7-AAD (Figs. 4, 5). The normalized histograms are more informational and show an increased width of the lifetime distributions in the presence of 7-AAD (Figs. 4, 5). The ASM scattergrams are somewhat elongated (NW to SE) in the presence of 7-AAD, which also reflects a wide distribution of lifetimes. Examination of the perpendicular width suggests a minor increase in contrast for this single nucleus.

FIG. 4.

Histograms and co-occurrence matrices of modulation lifetime images from the cell shown in Figure 1. The cell was stained with Ho in the presence and absence of 7-AAD. Co-occurrence matrices were constructed from the images normalized to the same average modulation lifetime.

In order to test our visual impressions, we averaged the results of the texture analysis for all 50 nuclei for the intensity (Fig. 6), modulation lifetime (Fig. 7), and phase lifetimes (Fig. 8). For the intensity images (Fig. 6), the presence of 7-AAD decreased the intensity (Fig. 6A), increased the homogeneity (Figs. 6B,C), and decreased the contrast (Fig. 6D). Distinct results were found for the modulation and phase images (Figs. 7, 8). The presence of 7-AAD decreased the lifetime (Figs. 7A, 8A), but increased the spatial heterogeneity (Figs. 7B,C, 8B,C), and contrast (Figs. 7D, 8D). Taken together, the effects of RET on the three parameters (ASM, Svar, and DVar) suggest that RET quenching of Ho is not spatially heterogeneous and that regions of higher DNA density, both AT and GC pairs, are present in select regions of the nuclei.

FIG. 6.

Scattergrams of the texture analysis results collected from the fluorescence intensity images of 50 different cells. Closed circles represent results from individual cells. Error bars indicate mean value with SE. A: Original fluorescence intensity data. B–D: Texture analysis results of normalized images.

FIG. 7.

Scattergrams of the texture analysis results collected from the modulation lifetime images of 50 cells. Closed circles represent the results from individual cells. The error bars indicate mean values with SE. A: Original lifetime data. B–D: Texture analysis results of normalized images. The presence of 7-AAD reduced the average Ho modulation lifetimes from 2.3 to 1.4 ns.

FIG. 8.

Scattergrams of the texture analysis results collected from the phase lifetime images of 50 cells. Closed circles represent results from individual cells. Error bars indicate mean values with SE. A: Original lifetime data. B–D: Texture analysis results of normalized images. The presence of 7-AAD reduced the average phase lifetime from1.8 to 0.6 ns.

Figure 9 shows correlation between RET efficiency and contrast (the difference variance) of the modulation life-time image. The contrast of the lifetime image seems to increase with decreasing mean efficiency of RET. Because the cells with lower RET efficiency are expected to have less condense DNA and larger average donor-to-acceptor distance, some donors seem to be partiallly quenched.

FIG. 9.

Relation between the efficiency of the energy transfer and the difference variance of the modulation lifetime. The data points represent 50 nuclei double stained by Ho and 7-AAD. F₀ represents the average fluorescence intensity with Ho only; F represents the average fluorescence intensity with Ho and 7-AAD

We have also found that RET efficiency depends on the cell cycle. Cells in G₂/M phase show slightly higher RET efficiency than the cells in G_0/1 phase (Fig. 10). At this moment, we do not know whether the increased energy transfer is due to a higher density of the DNA condensation centers or to spatial redistribution of the AT- and GC-rich regions of DNA. Such cell cycle-dependent redistribution could change base pair composition of the centers in favor of GC base pairs and increase RET efficiency.

FIG. 10.

Relation between cell cycle and energy transfer efficiency. F₀ is an average fluorescence intensity of cells stained with Ho only; F is an average fluorescence intensity of cells stained with Ho and 7-AAD.

DISCUSSION

We interpreted the changes in intensity and lifetime images after addition of 7-AAD in terms of the local AT- and GC-rich DNA concentration. On the other hand, we must also consider lifetime heterogeneity of donor decay. it is well known that RET between donors and acceptors bound to DNA results in increased lifetime heterogeneity in the phase and modulation values, depending on acceptor concentration (19,20). This effect would result in a range of decay times due to the nonexponential donor decays. However, irrespective of the lifetime heterogeneity of the donor decays, the phase and modulation values, measured at a single light modulation frequency, results in a unique apparent phase and modulation lifetime (33). For a multiexponential or nonexponential decay, the apparent phase lifetime is always shorter than the apparent modulation lifetime (29,34,35), as seen for Ho in Figures 7 and 8. Unfortunately, we cannot recover the true values of nonexponential fluorescence decay using FLIM at a single light modulation frequency. However, our goal was to search for variations of the distance between the Ho donor and the 7-AAD acceptor in individual nuclei. We can study such changes by quantification of spatial variations of apparent donor lifetimes in the presence and absence of the acceptor, which were shown to be directly related to the nonexponential lifetime decay and RET efficiency.

In conclusion, the results of the texture analysis revealed higher heterogeneity of the lifetime images in the presence than in the absence of 7-AAD. In about 20% of the cells, we could visually find a spatial lifetime variation that is represented by lower regions in the nuclear lifetime images in the presence of the acceptor (Fig. 2). In these regions, which correspond to the AT-rich DNA regions as seen from the intensity images (Fig. 2), the RET efficiency is higher because lifetime images of donor-only stained nuclei are flat. Such a result indicates that the average distance between Ho and 7-AAD is small in these regions of the nuclei. It is well known that the AT-rich regions of DNA are more condensed than the GC-rich regions (32). It seems probable that the spatial lifetime heterogeneity of Ho in the presence of 7-AAD is due to the closely spaced DNA strands in these three dimensionally condensed regions, rather than due to the RET between more closely spaced donors and acceptors in the same DNA helix.