1..

Introduction

Aspiration biopsy cytology of thyroid tumors is currently one of the most employed diagnostic methods for the differential diagnosis of malignant and benign lesions in the clinical medicine.^{1 2} The cytological diagnosis of the thyroid lesions is based on the cellular atypia. These include the nuclear morphology such as the nuclear size, shape, and the chromatin pattern.^{3 4 5} Since evaluation of the nuclear chromatin pattern is of a particular importance in the diagnosis of the thyroid papillary carcinoma (PC), significant work has been done to develop reliable method for differentiation between chromatin patterns of the PC and other thyroid lesions, including hyperplasia, adenoma and follicular carcinoma.^{6 7 8 9 10 11 12 13 14 15 16 17} The majority of the studies have approached this issue by looking for changes of the local deoxyribose nucleic acid (DNA) concentration. However, characteristic chromatin patterns of the thyroid tumors are described not only by the local DNA concentration, but also by the DNA conformation and/or by the three-dimensional relationship between different chromosomes. This requires investigation of the structural and topological changes of the DNA on the angstrom scale. Fo¨rster resonance energy transfer (FRET) is a great tool for such distance measurements.¹⁸

We have reported on the cell cycle-related structural and topological changes in interphase nuclei of culture cells examined by the microscopic FRET method.^{19 20} The FRET is a nonradiative process when the excitation energy of the excited fluorescent donor is transferred to the acceptor molecule.^{18 21} Since the FRET efficiency depends on the inverse sixth power of the donor-acceptor distance, the FRET can be used as a highly sensitive ruler for measurements of intermolecular distances on the scale of 10–100 Å.^{22 23 24} When used in microscopy, the microscopic FRET of samples specifically labeled by a proper donor-acceptor pair makes possible to extract information on distances below the diffraction limit, which are normally inaccessible by the optical microscopy. Therefore measurements of the distance change between segments of DNA molecules related to the structural and topological changes of the DNA are possible.^{19 20 25 26}

In this paper, we present a novel medical application of our previous microscopic FRET work to the clinical cytological diagnosis of thyroid tumor cells.^{19 20 24 26} We describe in more detail difference in the DNA organization between the papillary carcinoma and the follicular lesions of the thyroid by means of the microscopic FRET measurement between the AT-specific dye Hoechst 33258 (Ho) and the GC-specific dye 7-aminoactinomycin D (7-AAD). Results are interpreted in terms of the spatial proximity of the AT- and the GC-rich DNA regions. In order to quantitatively evaluate the FRET images, the two-dimensional maps of the corrected FRET signal were classified by the texture analysis.^{27 28}

2..

Materials and Methods

3..

Results

Representative fluorescence intensity images of a nucleus from a single papillary carcinoma cell double-stained by Ho and 7-AAD taken with the F, D, and A filter sets are shown in Figs. 1(a), 1(b), and 1(c), respectively. The I_cFRET image calculated from the intensity images (a)–(c), and representing the corrected FRET signal is shown in Fig. 1(d). The I_cFRET image exhibits considerable spatial heterogeneity of the fluorescence energy transfer that indicates uneven spatial distribution of the donor-acceptor distances. Consequently, the image reports on an uneven separation of the AT- and GC-rich DNA segments in these locations. The bright and dark regions in the I_cFRET images correspond to high and low efficiency of the energy transfer, respectively.

Figure 1

Nuclear fluorescence intensity and I_cFRET images of a single cell double-stained by Ho and 7-AAD. (a) The fluorescence intensity image taken with the FRET filter set; (b) the fluorescence intensity image acquired with the donor filter set; (c) the fluorescence intensity image collected with the acceptor filter set; and (d) the I_cFRET image corrected by the Gordon’s equation [Eq. (1)].

Typical I_cFRET images of nuclei from single cells extracted from the papillary carcinoma, adenomatous goiter, and the follicular adenoma are presented in Figs. 2(a), 2(c), and 2(e), respectively. Visually inspected, the images reveal different distribution of the FRET efficiency in the nuclei. The I_cFRET image of the papillary cell exhibits highly heterogeneous FRET pattern whereas, the image of the adenomatous goiter cell is homogeneous. The I_cFRET image of the follicular adenoma cell shows granular spatial distribution of the FRET signal. This heterogeneity, however, seems to be lower than that of the PC cell, Fig. 2(a). Moreover, the cell nuclei seem to display a different total FRET signal, as judged by the overall brightness of the I_cFRET images.

Figure 2

Typical results obtained on a single cell of the papillary carcinoma [(a) and (b)], adenomatous goiter [(c) and (d)] and the follicular adenoma [(e) and (f)]. The I_cFRET images [(a),(c),(e)] were calculated from the Gordon’s equation [Eq. (1)]. The radar charts show results of the quantitative texture analysis. The CV and the SVar are texture parameters characterizing the image heterogeneity, the ASM characterizes image homogeneity, and the DVar quantifies image contrast. The characteristic shape of the radar chart reflects spatial distribution of the FRET efficiency in the cell nucleus. All radar charts are presented in the same scale.

To quantitatively support conclusions of the visual examination, we have performed the texture analysis of the I_cFRET images and quantification of the total FRET signal. Results of the texture analysis and values of the total FRET signal are shown in radar charts in Figs. 2(b), 2(d), and 2(f). The total FRET reports on the extent of the three-dimensional condensation of the AT- and the GC-rich DNA segments in the nuclei. The calculated texture coefficients provide a quantitative evaluation of the spatial heterogeneity, homogeneity, and contrast of the I_cFRET images. In particular, the CV and the SVar are texture parameters characterizing the image heterogeneity, the ASM characterizes image homogeneity, and the DVar quantifies image contrast. Mean values of the texture parameters averaged over 20 cells in each case were used for the case summary statistics, where we have compared characteristics of the PC, AG, and FA. Figure 3 shows distribution of the average total FRET signal for the individual cases. In parenthesis are given characteristic values calculated by averaging the plotted values over the cases in each group. Although the PC cases exhibit somehow higher values of the total FRET, there was not found significant difference of total FRET signal between benign and malignant tumor groups after averaging these values over the whole group. Neither the ASM, describing image homogeneity, nor the DVar, describing the image contrast, revealed significant difference between the tumor groups, Figs. 4(b) and 4(d). Fortunately, values of the CV and SVar, which describe image heterogeneity, exhibited significant difference (CV, p<0.02; SVar, p<0.04) between the groups, suggesting possibility to distinguish the malignant papillary carcinoma from the adenomatous goiters and follicular adenomas, Figs. 4(a) and 4(c).

Figure 3

Scatter graph of the total FRET signal in the PC, the AD, and the FC cells. Each data point represents a mean value calculated from 20 nuclei of one case. Means calculated over all cases in each group are given in parenthesis.

Figure 4

Scatter graphs of the texture parameters calculated from the I_cFRET images of cells from the PC, AG, and FA. Each circle represents mean value of the CV, ASM, SVar, and DVar in one case. Mean calculated over all cases in each group are given in parenthesis. (^*: PC versus AG and FA: CV, p<0.02; SVar, p<0.04.)

The results of the quantitative analysis suggest that the papillary carcinomas and follicular tumors have on average similar distance between the AT- and the GC-rich DNA segments. The papillary carcinomas, however, exhibit more heterogeneous distribution of the FRET efficiency in the nucleus than the adenomatous goiters and the follicular adenomas.

4..

Discussion

The characteristic nuclear chromatin pattern of the papillary thyroid carcinomas is regarded as a criterion for the cytological diagnosis.^{4 32 33} Previously, we have quantitatively investigated the chromatin pattern of thyroid tumor cells by the texture analysis. Difference in the spatial distribution of the nuclear DNA between the thyroid carcinoma and benign tumor cells have been reported.^{11 16 17} The chromatin pattern of tumor cells originates from the DNA conformation, which depends on chromosome territories, the DNA concentration and condensation, the transcription activity, and the spatial correlation between the heterochromatin and the euchromatin. Specifically, the spatial correlation between the heterochromatin and the euchromatin in the nuclei is an important diagnostic marker in the cytological specimens.⁴ The transcriptionally inactive heterochromatin contains the AT-rich DNA, whereas the transcriptionally active euchromatin consists mainly of the GC-rich DNA.³⁴ To extend our previous studies, we have used FRET technique to analyze the DNA organization on its molecular level, i.e., on the scale below the diffraction limit of the conventional optical microscopes.^{19 20 26}

The unique property of FRET is its ability to measure molecular distances on the scale of tens of angstroms. An improved sensitivity of CCD cameras makes measurements of FRET under microscopes possible.^{21 24 35} At present, the microscopic FRET is widely utilized for number of applications in biology and medicine.^{21 23 36} Nevertheless, only a few FRET studies of the DNA organization in the interphase nuclei were reported.^{25 37} Recently we have studied the cell-cycle dependent changes of the spatial distribution and the organization of the AT- and the GC-rich nuclear DNA regions in culture cells. Experiments were based on the microscopic FRET measurements using both the fluorescence intensity and the fluorescence lifetime imaging.^{20 26 31} We have shown a difference between the microscopic FRET measurements with and without the Gordon’s correction.²⁶ It was shown that the uncorrected FRET images were influenced not only by the donor-acceptor distance, but by the concentration of the donor as well. The corrected FRET efficiency was affected only by a variation of the distance between the AT- and the GC-rich DNA.

In this study, we have employed the corrected FRET measurement for estimation of the DNA topology unaffected by the local difference in the DNA condensation. Because all tumor cells were measured in the G_0/1 phases, we can eliminate an influence of the cell-cycle related change of the DNA organization. We have not found significant difference of the overall FRET efficiency between the benign and the malignant tumor groups. This suggests that the papillary carcinomas and the follicular tumors have on average similar distance between the AT- and the GC-rich DNA segments. Since the FRET signal calculated by the Gordon’s equation is not rigorously equal to the real FRET efficiency, we have to note that the high average I_cFRET values found in the few PC cases might not be related to an extremely large change of the DNA topology. On the other hand, the quantitative analysis of as few as 20 cells of the each case reveals that the papillary carcinomas exhibit more heterogeneous distribution of the total FRET signal in the nucleus than the adenomatous goiters and the follicular adenomas. As shown in Fig. 4, some papillary carcinoma cases exhibit overlapped parameters with adenomatous goiters, which can be troubling in the clinical application. In this study, however, we have demonstrated that the corrected FRET measurement of as few as 20 cells can give valuable information for detection of the tumorous changes in the DNA topology, that has been so far inaccessible by the conventional steady-state fluorescence microscopy. Due to a limed number of cases available for this study we do not propose to use the corrected FRET method as an exclusive tool for detection of the tumorous changes in the DNA topology. This would certainly require careful examination of much larger ensemble of cases. However, characteristic chromatin feature of the papillary carcinoma cells revealed by the corrected FRET method can add significant information to the ordinary cytological findings such as nuclear size, shape, and conventional chromatin patterns.

In conclusion, we have demonstrated that the corrected FRET imaging is a helpful tool for the clinical cytological diagnosis of human tumors by giving additional information on the characteristic chromatin patterns.