1.

Introduction

The gold standard of clinical diagnosis of diseases, particularly those in discriminating of normal and cancerous tissues, remains to be histology. In the most common implementation, tissue of interest is fixed, paraffin-imbedded, thinly sliced, and labeled with hematoxylin and eosin for optical microscopic examination. Pathologist then determines the pathological state of the tissue specimens by qualitative examination and semiquantitative scoring. In the case of cancer, a positive identification is often attributed to a higher nuclear-to-cytoplasm (N/C) ratio characteristic of tumor cells. Particularly, N/C ratio analysis has been applied to characterize oral smears,¹ nuclear-to-cytoplasmic change is indicative of dysplastic change in oral tissues,² and high CDK1 N/C ratios correlate to poor prognosis in colorectal cancer patents.³ Furthermore, two-color, two-photon imaging found larger N/C ratio in colon adenocarcinoma specimens.⁴ In addition to tissue-based histology, cytological procedures, such as Pap smear, and analysis based on bodily fluids, such as sputum, urine, and peritoneal fluid, are heavily used in the clinical setting. Despite the wide use of N/C ratio, significant variations among observers were found in nuclear ratios and interpersonal variations in N/C ratio determination, a persistent problem in histological analysis.^5,6

Uncertainty of using N/C ratio for disease diagnosis is due to the fact that traditionally, N/C ratio analysis is two-dimensional (2-D) in nature. Since the tissue slices are on the order of a few microns, diagnosis of such specimens is a 2-D representation of the actual specimen. Depending on the physical location and orientation of the tissue section, simulations have shown that geometric factors, such as nuclear position, cell shape, and orientation, can significantly affect 2-D N/C ratios.⁷ Therefore, a more precise definition of N/C ratio would result in reduced ambiguity in clinical differentiation of tumor from normal cells. While techniques such as automatic digital extraction of morphological features, three-dimensional (3-D) reconstruction of serial histological images, and 3-D N/C ratio obtained from optical coherence tomography have been developed for improved cancer identification,^8–10 a comparison of the efficacy between 2-D and 3-D N/C ratio analysis through dual-color fluorescence analysis is absent. In this study, we imaged both normal and lung adenocarcinoma cells by 3-D multiphoton microscopy. By labeling the nuclei and cytoplasm with fluorescent dyes of different spectral properties, we acquired serial images in three dimensions for determination of 2-D and 3-D N/C ratios.

2.

Materials and Methods

3.

Results

Figure 1(a) shows the representative multiphoton images of the Beas2B, CL1-0, and CL1-5 cells in three dimensions. Individual 2-D images at different focal depths can then be extracted [Fig. 1(b)]. Qualitatively, the nucleus is largest for CL1-5, followed by CL1-0, and Beas2B. The total size of the cells also follows the same trend. Furthermore, 2-D N/C ratio at each depth was determined. The results show the uncertainty in the precise determination of 2-D N/C ratio value. Specifically, images for all three cells at the depths of 10 and $15\mu \mathrm{m}$ are similar in appearance; however, the 2-D N/C ratios can differ from 0.36 to 0.26 (Beas2B), 0.54 to 0.42 (CL1-0), and 0.96 to 0.72 (CL1-5).

Fig. 1

(a) Representative multiphoton images of the Beas2B, CL1-0, and CL1-5 cells in three dimensions. (b) The 2-D images of the three cells along with 2-D N/C ratio at different focal depths. Blue pseudocolor: Hoechst 33342 labeled nucleus; red pseudocolor: CellTracker^™ Orange CMTMR labeled cytoplasm.

The ambiguity that exists in a precise definition of N/C ratio in two dimensions is further demonstrated in Fig. 2 where we plotted the nuclear area, cytoplasmic area, 2-D N/C ratio, and total cell area as a function of axial position. In the three representative cells, the axial location of maximum nuclear size and 2-D NC ratio do not correspond and can differ in the 4 to $8\mu \mathrm{m}$ range. Furthermore, depending on the axial position chosen to calculate the 2-D N/C ratio, the values can differ by a factor of 2.

Fig. 2

Representative axial profiles of nuclear area (N), cytoplasmic area (C), total cell area (N+C), and 2-D N/C ratio ($500\times$) for Beas2B, CL1-0, and CL1-5 cell lines.

To quantify different values of N/C ratio in two dimensions, we applied three definitions of 2-D N/C ratios: (1) $N_{\text{max}}$: largest nuclear area. N/C ratio was calculated at an axial position at which the nuclear area is the largest; (2) $N/C_{\text{max}}$: after determining the N/C ratio at different depths, the largest N/C ratio value was selected; and (3) $N_{\text{midpoint}}$: midpoint of nuclear axial position was determined and N/C ratio calculated. In addition, the nuclear and cytoplasmic volumes were calculated by summing over the nuclear and cytoplasmic areas at each axial position. Finally, by summing nuclear and cytoplasmic areas at each focal plane, 3-D N/C ratio can be determined. In addition to showing the average value and standard deviations of these parameters, we also determined the percentage of data deviating from the average values (Table 1).

Table 1

Statistical analysis of the physical characteristics of Beas2B, CL1-0, and CL1-5 cell lines in two and three dimensions (20 cells in each cell line were analyzed).

4.

Discussion

In this study, we studied 2-D N/C ratios determined from different criteria and 3-D volumetric N/C ratio for normal lung epithelial cell line Beas2B and two lung adenocarcinoma cell lines (CL1-0 and CL1-5) cell lines. We found that depending on the criterion chosen, 2-D N/C ratios can vary drastically (Table 1). Particularly, the variance of the data accumulated from 20 cells varies significantly (21% to 86%), suggesting that 2-D N/C ratio is critically affected by an individual cell’s geometrical orientation at the growth site. On the other hand, the 3-D volumetric N/C ratio is unambiguous in definition and shows a lower variance (28% to 38%). In addition to higher 3-D N/C ratios for cancer cells, we found that the nuclear volumes for CL1-0 ($2768{\mu \mathrm{m}}^3$) and CL1-5 ($3271{\mu \mathrm{m}}^3$) are significantly larger than that of Beas2B ($1787{\mu \mathrm{m}}^3$) where the cytoplasmic volume of CL1-5 ($5239{\mu \mathrm{m}}^3$) is the largest, followed by Beas2B ($4756{\mu \mathrm{m}}^3$), and CL1-0 ($3800{\mu \mathrm{m}}^3$). Therefore, the more metastatic CL1-5¹¹ is the largest cell type in nucleus and cytoplasm, although it has a smaller 3-D N/C ratio than CL1-0. Since the 3-D N/C ratio for Beas2B and CL1-5 statistically overlap, our results suggest that in addition to N/C ratio analysis in three dimensions, other metrics such as nuclear, cytoplasmic, and total cell size should be included for distinguishing between normal and cancerous cells in the clinical setting. If spatial resolution is defined as the size of the focal spot defined by first minima of the Airy pattern, then our use of an NA 0.75 objective, at 780 nm, would result in nominal lateral and axial resolutions are 1.27 and $5.55\mu \mathrm{m}$, respectively. Since images were acquired at different depths, the use of objectives with different NAs would affect the N/C ratio results. In the case of low NA, although a larger axial depth is sampled, the corresponding decrease in lateral resolution may introduce ambiguity in lateral determination of the size of cellular components. The approach described in this work may be applied to pathological analysis in cytology.