Open Access
Add to BibSonomy
Abstract
Gross chromatin imbalance and high DNA content are distinct features of various types of cancer cells. However, severe inflammation can also produce similar symptoms in cells. In this study, normal, inflammatory, and carcinoma hepatic cells were stained with 4’,6-diamidino-2-phenylindole (DAPI) and investigated by hyperspectral microscopy. DAPI is a DNA-sensitive fluorochrome. Therefore, the differences in the cellular DNA of the samples can be revealed by the corresponding fluorescence. Our experimental results demonstrate that although chromosomal disorder and high DNA content both occur in severely inflammatory and carcinoma hepatic cells, there is still a slight difference in their DNA, making their fluorescent intensity and even their spectral shapes distinguishable. Based on these spectral features, we developed a method for the precise identification of normal, inflammatory, and carcinoma hepatic cells in the field of view. The identification accuracy for these three types of cells was 99.8%. We believe that examination that combines DAPI staining with hyperspectral microscopy is a potential method for the identification and investigation of various types of cancer tissues.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Most cancer cells contain aneuploidy and large-scale structural rearrangement of chromosomes. [1,2] Aneuploidy is usually caused by the persistent loss and gain of whole chromosomes, while chromosomal rearrangements are caused by improper repair of DNA damage. [3] These features result in differences in DNA content between normal and cancer cells. One of the conventional methods for detecting DNA content and ploidy is flow cytometry combined with 4’,6-diamidino-2-phenylindole (DAPI) staining. DAPI is a DNA-specific and DNA-sensitive stain. Therefore, DNA content and ploidy can be calculated by examining the fluorescence intensity of DAPI stained nuclei via flow cytometry. [4–6] With this method, the differences in DNA content and DNA ploidy between normal and cancer cells can be clearly shown. For example, Kitayama et al. found that the frequency of aneuploidy in gastric cancer with moderate or poor differentiation was apparently higher than that in highly differentiated gastric cancer. [7] This method was also tested in screening for early detection of lung cancer, and it could distinguish cancer cells with high DNA content from normal cells. [8] However, this method cannot precisely identify cancer cells from inflammatory cells, [9] because severely inflammatory cells also have many aneuploid as well as high DNA content. [10,11] When inflammatory and cancer cells are stained with DAPI, their fluorescence intensities are similar. Therefore, a method that can only obtain information on the fluorescence intensity of cells is not suitable for the precise identification of normal, inflammatory, and cancer cells.
Hyperspectral microscopy can obtain both the spectral and spatial information of the sample. [12–14] As an emerging non-contact optical imaging modality, it has great potential for biological identification. [15–18] Especially, hyperspectral microscopic imaging technology has already been proposed to study cancer cells. [19–22] In 2020, we first used hyperspectral microscopy to replace flow cytometry and investigated normal and carcinoma cells stained with DAPI. The results showed the fluorescence intensity and spectral shapes of the normal and carcinoma cells were different. Based on these spectral features, normal and carcinoma cells can be identified precisely with an accuracy of 99.3%. [23] However, to the best of our knowledge, hyperspectral microscopy has not yet been applied for the investigation and classification of inflammatory and carcinoma cells.
In this study, the fluorescence characteristics of normal, inflammatory, and carcinoma hepatic cells stained with DAPI were studied using hyperspectral microscopy. The experimental results demonstrate that the fluorescence intensity and spectral shapes of three types of hepatic cells are distinguishable, indicating that their DNA contents are different. A linear discriminant analysis (LDA) [24] model based on fluorescent spectra was used for identification, and the accuracy was 99.8%. In particular, to avoid the fluorescent intensity fluctuation introduced by changes in the environment, we used the normalization spectra to train the model, and the accuracy was also up to 99.8%.
2. System and samples
2.1 Hyperspectral microscopy imaging system
The hyperspectral microscopic system used in the experiment is shown in Fig. 1. The system consisted of an inverted fluorescence microscope (ECLIPSE Ti-U, Nikon), a liquid crystal tunable filter (LCTF, VariSpecVIS, CRI Inc.), and a 16-bit CMOS camera (ORCA-Flash 4.0 LTC11440-42U, HAMAMATSU) with a spatial space cell size of 6.5 µm (H) 6.5 µm (V). In this experiment, a xenon lamp combined with a narrow band UV filter at 360 nm was used as the excitation light source for exciting the fluorescence of the samples. Accordingly, a dichroic mirror with a cut-off wavelength of 450 nm was added to the system to separate the excitation light and fluorescence of the samples. When the fluorescence passes through the LCTF, it is filtered and finally captured by the CMOS camera. By continually changing the voltage of the LCTF, a set of 2D images of the sample at a single wavelength can be obtained, enabling the acquisition of morphological and spectral information of the samples. In the spatial domain, the pixel resolution was set to 512 × 512. In the spectral domain, the scanning band was set from 450 nm to 650 nm with a wavelength scanning step of 2 nm. So, we can obtain enough spectral features for the identification. The scan duration was set to 0.1 s. For each scanning, we obtained 101 spectral images (10.1 s) corresponding to the range of 450–650 nm. In addition, a color CCD was employed in the system to capture the color images of the samples.
2.2 Samples
In this experiment, normal, inflammatory, and carcinoma hepatic cells were selected for further investigation. Normal hepatic cells (BNCC 338070 LX-2) and carcinoma hepatic cells (BNCC 341818 HepG2) were provided by the BeNa Culture Collection Company (BNCC Inc.). Lipopolysaccharide (LPS) was used to activate the inflammation of hepatic cells (BNCC 338070 LX-2). Hence, we can obtain two types of pure samples (normal and carcinoma cells), and one type of LPS-reactive sample that contains normal and inflammatory cells. All samples were first washed with a phosphate buffer solution (PBS) (NaCl 137 mM, KCl 2.7 mM, Na2HPO4 10 mM, KH2PO4 7.4 mM, pH 7.4) and then fixed in 4% paraformaldehyde (PFA) for 20 min. The samples were then washed three times for 10 min each with PBS containing 0.1% Tween (PTw). Finally, PTw-diluted DAPI (1 µg/1000 ml) was used to stain the samples for 10 min. In this experiment, 18 slides were sealed for identification (Table 1). Before image acquisition, the slides were kept at 4 °C in the dark.
Table 1. Slides samples of cell
3. Method and results
3.1 Hyperspectral dataset acquisition and analysis
The standard spectra of normal and carcinoma hepatic cells can be easily extracted from the corresponding slide samples containing only normal or carcinoma hepatic cells. However, the standard spectra of inflammatory hepatic cells cannot be extracted from LPS-reactive slide samples directly because these samples contain both normal and inflammatory hepatic cells. To obtain standard spectral data for three different types of hepatic cells, we first performed spectral data extraction for each cell in all unmixed slide samples and obtained 27 sets of image groups (three sets for each slide sample). Figure 2 demonstrates the spectral data extraction process for each cell in the field of view.
Fig. 2. Flow chart of fluorescent spectral extraction of carcinoma hepatic cells. the interior of the dashed box describes the fluorescent spectral extraction of carcinoma hepatic cells. Each set of images group contains 101 fluorescence images from 450 nm to 650 nm with a wavelength scanning step of 2 nm. The average spectra of all pixels for each nucleus were used as the characteristic spectra of the cell. Hence, the standard spectra of carcinoma hepatic cells for each cell in all the 9 sets of image groups were obtained. The same processes were performed in normal and LPS-reactive samples.
The spectra of all cells within a certain field of view of an LPS-reactive sample are shown in Fig. 3(a). One can see that the spectra of the LPS-reactive cells can be divided into two categories based on their fluorescence intensity, and the boundary of these categories is clear. In particular, the lower-intensity spectra are in accordance with the spectra extracted from the samples of normal hepatic cells, as shown in Fig. 3(a) and Fig. 3(b). Hence, the higher-intensity spectra should belong to cells with higher DNA content than normal cells. Although not all inflammatory cells contain higher DNA content, in the LPS-reactive samples, the increase in the DNA of the cells can only be caused by inflammation. Based on this analysis, we conclude that the lower-intensity spectra belong to the normal hepatic cells, while the higher-intensity spectra belong to the inflammatory hepatic cells, especially in severe cases. This observation can be used to distinguish the normal and severely inflammatory hepatic cells in LPS-reactive samples and extract their spectra.
Fig. 3. The comparing of the fluorescent intensity spectral curves of severely inflammatory and normal hepatic cells.
The averaged fluorescent spectral curves of normal, inflammatory, and carcinoma hepatic cells for different individual cells and corresponding normalized results are shown in Fig. 4(a) and Fig. 4(b). The non-normalized spectral curves demonstrate that there is a significant difference in the fluorescence intensity between these three cell types, and the normalized spectral curves demonstrate that their spectral shapes are slightly different. The cells can be identified using their non-normalized spectra. However, given that the equipment and environmental conditions have a significant influence on the fluorescence intensity, the normalized spectra make the identification of the cells more stable and adaptable. Hence, in this experiment, non-normalized spectra and normalized spectra were used for identification respectively, and both results were discussed.
Fig. 4. (a) Average fluorescence intensity spectral curves of normal, severely inflammatory, and carcinoma hepatic cells with individual fluctuations. (b) Normalized spectral curves of normal, severely inflammatory, and carcinoma hepatic cells with individual fluctuations. The average data were obtained from the data of 300 nuclei.
3.2 Analysis of DNA ploidy
Flow cytometry examination combined with DAPI staining is one of the most effective ways to detect the cellular DNA content/ploidy type. This examination is based on the conclusion that fluorescence intensity is proportional to the DNA content. In this experiment, the fluorescence intensity can be calculated as the integral area of the spectrum; then, we can calculate the DNA content and DNA ploidy. Here, we first obtained the spectrum of a normal cell with a diploid and calculated its integral area, as shown in Fig. 5(a) and Fig. 5(b). This integral area is defined as the fluorescent intensity Id of the diploid. Therefore, the DNA ploidy type of each cell can be calculated using the following formula:
where N is the number of DNA-ploidy types and I is the fluorescence intensity of a cell stained with DAPI. Based on this calculation, we obtained a histogram of DNA ploidy for the three types of hepatic cells, as shown in Fig. 5(c).Fig. 5. (a) Gray image of normal cells, (b) and the average of the spectrum curves of each point in the normal cell in the blue dashed box of (a). The integral of the spectrum from 450 nm to 650 nm is the area S of the shaded portion of the curve of the cell. (c) statistical results of DNA (rel. units) for the three types of hepatic cells. According to the “C-index” scale, first suggested by the American Swift, diploid was designated as “2C” and haploid DNA was designated as “1C” [25].
It is evident that almost all normal nuclei were diploid (2C), while the inflammatory and carcinoma nuclei fell in the class over 3C and most were aneuploid. On average, the DNA content of carcinoma cells was much higher than that of normal cells. This is why the flow cytometry can be used for the diagnosis of cancer based on the statistical data of cellular DNA content. However, it can easily lead to incorrect diagnosis if the sample contained inflammatory cells. Therefore, flow cytometry is not suitable for the precise identification of normal, inflammatory, and carcinoma cells. In particular, classical flow cytometry can only obtain information on the total fluorescence intensity of a cell, and it cannot capture the difference in the spectral shape of these three kinds of samples.
3.3 Model training and testing
We extracted 1800 standard spectra covering the band from 450 nm to 650 nm of three types of hepatic cells for the analysis (600 fluorescent spectra of each type). Half were used as training samples, while the rest were used as test samples. Then, the linear discriminant analysis (LDA) method was employed for model training and identification. The corresponding results are listed in Table 2.
Table 2. The identification results in basis of non-normalized fluorescent intensity spectral data and Linear Discriminant Analysis (LDA) method
In this experiment, the accuracy (ACC) of identification was defined as follows:
Here, Ture identification is that the predicted conditions correspond to the actual conditions. The ACC was calculated to be 99.8% according to the data in Table 2. Considering that the normalized spectra data were more stable, we normalized all the training data and test data. Then, it was also analyzed based on the LDA method. The results showed that the accuracy was also 99.8%, as shown in Table 3.
Table 3. The identification results in basis of normalized fluorescent intensity spectral data and Linear Discriminant Analysis (LDA) method
The sensitivity (SEN) is used to measure the proportion of positives identified correctly in all positives and the specificity (SPEC) is used to measure the proportion of negatives identified correctly in all negatives. They were defined by the formulas as below [26]:
TP present the number of true positives while TN present the number of true negatives. FN and FP are the number of false negatives and the number of false positives, respectively.
In the diagnosis, carcinoma hepatic cells are usually regarded as positives while normal and inflammatory hepatic cells are regarded as negatives. The SEN and SPEC for carcinoma hepatic cells identification based on non-normalized spectra were 99.67% and 99.83% respectively, and the SEN and SPEC for carcinoma hepatic cells identification based on normalized spectra were also 99.67% and 99.83%, respectively.
3.4 Automatic identification of cells in FOV
As a method that can obtain both spatial and spectral information, hyperspectral imaging can not only be used for the identification of samples but can also be used to digitally stain samples and label different types of targets with different colors. Figure 6 shows the identified and labeled results of the mixed samples. Although the cells in the mixed samples are indistinguishable in morphology, they can be precisely identified by their spectra and labeled by pseudo-color marking. Compared with the immunofluorescence labeling method, this method can label three types of cells with one staining (DAPI).
Fig. 6. Image processing for automatic identification of cells in the FOV. The first line is the original single wavelength at 486 nm for mixed samples (mix of normal and severely inflammatory cells, mix of carcinoma and severely inflammatory cells, mix of three kinds of cells respectively). The second line is the corresponding pseudo-color image of the identification result. The cells colored with red, green, and blue are identified as normal, severely inflammatory, and carcinoma cells, respectively.
4. Conclusion
Our study demonstrates that the fluorescent spectra of DAPI-stained normal, inflammatory, and carcinoma cells were distinguishable. The fluorescence of DAPI, a DNA-specific fluorochrome, is very sensitive to cellular DNA. Although chromosome disorder and high DNA content both occur in severely inflammatory and carcinoma hepatic cells, the small difference in their DNA can still be revealed by the fluorescence intensity and its spectral shape. By using hyperspectral microscopy, we precisely identified normal, inflammatory, and carcinoma hepatic cells with a corresponding accuracy of 99.8%. Compared with flow cytometry, hyperspectral microscopy can obtain more spectral and spatial information, which enable the precise identification of each cell in the field of view. We believe that examination that combines DAPI staining with hyperspectral microscopy is a potential method for the dentification and investigation of various types of cancer tissues.
Funding
Key-Area Research and Development Program of Guangdong Province (2020B090922006); National Natural Science Foundation of China (61935010, 61975069); Guangdong Project of Science and Technology Grants (2018B030323017); Guangzhou Science and Technology Project (201903010042, 201904010294).
Disclosures
The authors declare that there are no conflicts of interest related to this study.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. H. Rajagopalan and C. Lengauer, “Aneuploidy and cancer,” Nature 432(7015), 338–341 (2004). [CrossRef]
2. N. J. Ganem, Z. Storchova, and D. Pellman, “Tetraploidy, aneuploidy and cancer,” Curr. Opin. Genet. Dev. 17(2), 157–162 (2007). [CrossRef]
3. S. L. Thompson and D. A. Compton, “Chromosomes and cancer cells,” Chromosome Res. 19(3), 433–444 (2011). [CrossRef]
4. J. Kapuscinski, “DAPI: a DNA-specific fluorescent probe,” Biotech. Histochem. 70(5), 220–233 (1995). [CrossRef]
5. Y. Umebayashi and F. Otsuka, “Prognostic significance in malignant melanoma of nuclear DNA content measured by a microfluorimetric method,” Arch. Dermatol. Res. 287(8), 718–722 (1995). [CrossRef]
6. R. Bernander, T. Stokke, and E. Boye, “Flow cytometry of bacterial cells: comparison between different flow cytometers and different DNA stains,” Cytometry 31(1), 29–36 (1998). [CrossRef]
7. Y. Kitayama, S. Nakamura, H. Sugimura, and I. Kino, “Cytophotometric and flow cytometric DNA content of isolated glands in gastric neoplasia,” Gut 36(4), 516–521 (1995). [CrossRef]
8. M. Takahama and A. Kagaya, “Hematoporphyrin/DAPI staining: simplified simultaneous one-step staining of DNA and cell protein and trial application in automated cytological screening by flow cytometry,” J. Histochem. Cytochem. 36(8), 1061–1067 (1988). [CrossRef]
9. E. Deinlein, H. Schmidt, J. F. Riemann, R. Grässel-Pietrusky, and O. P. Hornstein, “DNA flow cytometric measurements in inflammatory and malignant human gastric lesions,” Virchows Arch. A: Pathol. Anat. Histol. 402(2), 185–193 (1983). [CrossRef]
10. R. Porschen, U. Robin, A. Schumacher, S. Schauseil, F. Borchard, K. J. Hengels, and G. Strohmeyer, “DNA aneuploidy in Crohn’s disease and ulcerative colitis: results of a comparative flow cytometric study,” Gut 33(5), 663–667 (1992). [CrossRef]
11. D. Lothschütz, M. Jennewein, S. Pahl, H. F. Lausberg, A. Eichler, W. Mutschler, R. G. Hanselmann, and M. Oberringer, “Polyploidization and centrosome hyperamplification in inflammatory bronchi,” Inflammation Res. 51(8), 416–422 (2002). [CrossRef]
12. M. E. Martin, M. B. Wabuyele, K. Chen, P. Kasili, M. Panjehpour, M. Phan, B. Overholt, G. Cunningham, D. Wilson, R. C. DeNovo, and T. Vo-Dinh, “Development of an advanced hyperspectral imaging (HSI) system with applications for cancer detection,” Ann. Biomed. Eng. 34(6), 1061–1068 (2006). [CrossRef]
13. A. A. Gowen, C. P. Odonnell, P. J. Cullen, G. Downey, and J. M. Frias, “Hyperspectral imaging – an emerging process analytical tool for food quality and safety control,” Trends Food Sci. Technol. 18(12), 590–598 (2007). [CrossRef]
14. A. A. Gowen, Y. Feng, E. Gaston, and V. Valdramidis, “Recent applications of hyperspectral imaging in microbiology,” Talanta 137, 43–54 (2015). [CrossRef]
15. S. V. Panasyuk, S. Yang, D. V. Faller, D. Ngo, R. A. Lew, J. E. Freeman, and A. E. Rogers, “Medical hyperspectral imaging to facilitate residual tumor identification during surgery,” Cancer Biol. Ther. 6(3), 439–446 (2007). [CrossRef]
16. S. J. Leavesley, N. Annamdevula, J. Boni, S. Stocker, K. Grant, B. Troyanovsky, T. C. Rich, and D. F. Alvarez, “Hyperspectral imaging microscopy for identification and quantitative analysis of fluorescently-labeled cells in highly autofluorescent tissue,” J. Biophotonics 5(1), 67–84 (2012). [CrossRef]
17. L. Zhi, D. Zhang, J. Q. Yan, Q. L. Li, and Q. L. Tang, “Classification of hyperspectral medical tongue images for tongue diagnosis,” Comput. Med. Imaging Graph. 31(8), 672–678 (2007). [CrossRef]
18. Q. Liu, K. Sun, J. Peng, M. Xing, L. Pan, and K. Tu, “Identification of bruise and fungi contamination in strawberries using hyperspectral imaging technology and multivariate analysis,” Food Anal. Methods 11(5), 1518–1527 (2018). [CrossRef]
19. M. E. Paoletti, J. M. Haut, J. Plaza, and A. Plaza, “Deep learning classifiers for hyperspectral imaging: A review,” ISPRS J. Photogramm. 158, 279–317 (2019). [CrossRef]
20. R. Pike, G. Lu, D. Wang, Z. G. Chen, and B. Fei, “A minimum spanning forest-based method for noninvasive cancer detection with hyperspectral imaging,” IEEE Trans. Biomed. Eng. 63(3), 653–663 (2016). [CrossRef]
21. L. Ma, M. Halicek, and B. Fei, “In vivo cancer detection in animal model using hyperspectral image classification with wavelet feature extraction,” Proc. SPIE 11317, 113171C (2020). [CrossRef]
22. H. Akbari, L. V. Halig, D. M. Schuster, A. Osunkoya, V. Master, P. T. Nieh, G. Z. Chen, and B. Fei, “Hyperspectral imaging and quantitative analysis for prostate cancer detection,” J. Biomed. Opt. 17(7), 0760051 (2012). [CrossRef]
23. K. Liu, S. Lin, S. Zhu, Y. Chen, H. Yin, Z. Li, and Z. Chen, “Hyperspectral microscopy combined with DAPI staining for the identification of hepatic carcinoma cells,” Biomed. Opt. Express 12(1), 173–180 (2021). [CrossRef]
24. M. Pohar, M. Blas, and S. Turk, “Comparison of logistic regression and linear discriminant analysis: a simulation study,” Metodoloski zvezki 1(1), 143–161 (2004). [CrossRef]
25. H. Swift, “The constancy of desoxyribose nucleic acid in plant nuclei,” Proc. Natl. Acad. Sci. U. S. A. 36(11), 643–654 (1950). [CrossRef]
26. T. Ince, S. Kiranyaz, L. Eren, M. Askar, and M. Gabbouj, “Real-time motor fault detection by 1-D convolutional neural networks,” IEEE Trans. Ind. Electron. 63(11), 7067–7075 (2016). [CrossRef]
