2.1.

Cells and Viruses

Murine fibroblast cell lines (NIH/3T3, long-term in vitro) and mouse embryonic fibroblast cells (MEF, primary cells) were grown at $37°\mathrm{C}$ in Roswell Park Memorial Institute (RPMI) medium supplemented with 10% newborn calf serum (NBCS) and the antibiotics penicillin, streptomycin, and neomycin.

Clone 124 of TB cells chronically releasing the Moloney murine sarcoma virus (MuSV-124) was used to prepare a virus stock that contained an approximately 30-fold excess of MuSV particles over Moloney murine leukemia virus (MuLV) particles.³⁸ MuLV and MuSV used in this research were grown on NIH/3T3 cells. The virus concentration was determined by counting the number of foci (ffu-focus-forming units).

2.2.

Cell Infection and Determination of Malignant Transformation

Monolayers of NIH/3T3 and MEF cells were grown in $9\text{-}{\mathrm{cm}}^2$ tissue culture plates and treated with $0.8\mu \mathrm{g}∕\mathrm{ml}$ of polybrene (a cationic polymer required for neutralizing the negative charge of the cell membrane) for $24\mathrm{h}$ before infection with the virus. Free polybrene was then removed, and both types of cells were incubated at $37°\mathrm{C}$ for $2\mathrm{h}$ with the infecting virus (MuSV-124) at various concentrations in RPMI medium containing 2% of NBCS. The unabsorbed virus particles were removed, fresh medium containing 2% NBCS was added, and the monolayers were incubated at $37°\mathrm{C}$ . After various time intervals, the cell cultures were carefully examined for the appearance of malignant transformed cells by the following methods in parallel:

2.3.

Sample Preparation for Fourier Transform Infrared Microscopy Measurements

Since ordinary glass slides exhibit strong absorption in the wavelength range of interest, zinc sellenide crystals, which are highly transparent to IR radiation, were used. Cell cultures were washed with a physiological saline solution and picked up from the tissue culture plates after treatment with trypsin (0.25%) for $1\min$ . The cells were pelleted by centrifugation at $1000\mathrm{rpm}$ for $5\min$ . Each pellet was washed twice with saline and resuspended in $100\mu \mathrm{l}$ of saline. The number of cells was counted with a hematocytometer, and all tested samples were pelleted again and resuspended in an appropriate volume of saline to give a concentration of $1000\text{cells}∕\mu \mathrm{l}$ . A drop of $1\mu \mathrm{l}$ of each sample was placed on a certain area on the zinc sellenide crystal, air dried for $1\mathrm{h}$ , and measured by FTIR microscopy. The radius of such $1\text{-}\mu \mathrm{l}$ drop was about $1\mathrm{mm}$ , producing a monolayer of cells with about $10\mu \mathrm{m}$ . Figure 1 displays characteristic sites for measurements as observed by a light microscope for normal [Fig. 1a] and completely transformed [Fig. 1b] murine fibroblast cell line (NIH/3T3).

Fig. 1

Photomicrograph of (a) normal NIH/3T3 cell line and (b) completely transformed fibroblast cell line (NIH/MuSV).

2.4.

Fourier Transform Infrared Microspectroscopy and Data Acquisition

Measurements on cell cultures were performed using the FTIR microscope IR scope 2 with a liquid-nitrogen-cooled mercury-cadmium-telluride (MCT) detector, coupled to the FTIR spectrometer (Bruker Equinox model 55/S, OPUS software). To achieve high signal-to-noise ratio (SNR), 128 co-added scans were collected in each measurement in the wavenumber region $800\text{to}4000{\mathrm{cm}}^{-1}$ . The measurement site was circular with a diameter of $100\mu \mathrm{m}$ and spectral resolution of $4{\mathrm{cm}}^{-1}$ was used. To reduce cell amount variation and guarantee proper comparison between different samples, the following procedures were adopted.

2.5.

Statistical Analysis

The obtained parameters (biomarkers) were classified using the cluster analysis according to Ward’s method⁴⁰ and the discriminant classification function (DCF) method.^{41, 42} The differences were considered significant at $P<0.05$ .

3.1.

Spectral Differences between Normal and Malignant Cell Lines

The main objective of this research is to identify and study early changes during malignant transformation using FTIR-MSP. As a first step, it was important to find spectral biomarkers that can discriminate between normal and completely malignant cells.

We analyzed FTIR spectra of 50 different samples of both normal (NIH/3T3) and completely transformed murine fibroblast cell lines (NIH/3T3/MuSV). Two regions with significant and consistent differences were identified at $3000\text{to}2820{\mathrm{cm}}^{-1}$ and $1145\text{to}1000{\mathrm{cm}}^{-1}$ . For an effective comparison, these regions were cut from the entire spectra, baseline corrected and normalized.

The results in the region $3000\text{to}2820{\mathrm{cm}}^{-1}$ , presented in Fig. 2a , show four prominent absorbance bands: near $2852{\mathrm{cm}}^{-1}$ (due to the symmetric stretching of the methylene chains in membrane lipids); at $2923{\mathrm{cm}}^{-1}$ (due to the antisymmetric $\mathrm{C}{\mathrm{H}}_2$ stretch); at $2958{\mathrm{cm}}^{-1}$ (due to antisymmetric stretching of the methyl groups of both lipids and proteins); and at $2871{\mathrm{cm}}^{-1}$ (arising from the symmetric $\mathrm{C}{\mathrm{H}}_3$ stretching mode).¹ The average absorption intensities of normal and transformed fibroblast cell lines are distinctive at $2852\text{-}{\mathrm{cm}}^{-1}$ and $2958\text{-}{\mathrm{cm}}^{-1}$ bands [Fig. 2a]. It was found that the best discriminating values were obtained by deriving the intensity ratio of these two vibrational modes (i.e., ${\mathrm{A}}_{2958}∕{\mathrm{A}}_{2852}$ or $v_{\mathrm{as}}$ $\mathrm{C}{\mathrm{H}}_3∕v_{\mathrm{s}}\mathrm{C}{\mathrm{H}}_2$ ).

Fig. 2

FTIR spectra in the regions: (a) $2820\text{to}3000{\mathrm{cm}}^{-1}$ , (b) $1000\text{to}1145{\mathrm{cm}}^{-1}$ , and (c) the second derivative at $1000\text{to}1145{\mathrm{cm}}^{-1}$ of the normal murine fibroblast cell line (NIH/3T3) and of the completly transformed murine fibroblast cell line (NIH3T3/MuSV). Spectra are the average of 50 samples and five measurements of each sample after baseline correction and normalization.

The dimensionless ratio eliminates artifact, which may arise due to the baseline contribution underneath each band. Table 1 summarizes the statistical values of the previous ratio for the normal and malignant cell line. The $t$ -value of the two groups is 11.25 (Table 1). Therefore, this ratio may be considered as a satisfactory biomarker to follow the progress of malignant transformation.

Table 1

Statistical analysis of the biomarkers derived from FTIR spectra of normal and transformed murine fibroblast cell lines.

In the second region at $1145\text{to}1000{\mathrm{cm}}^{-1}$ [Fig. 2b], there are plenty of overlapping vibrational modes associated with absorbance of macromolecules such as proteins, nucleic acids, carbohydrates, and phospholipids. The bands at 1082 and $1056{\mathrm{cm}}^{-1}$ correspond to absorbance of the ${\nu }_s$ $\mathrm{P}{\mathrm{O}}_2^{-}$ of phosphodiesters of nucleic acids¹ and the O–H stretching coupled with C–O bending of C–OH groups of carbohydrates, respectively.³⁹ Other bands at 1121 and $1015{\mathrm{cm}}^{-1}$ can be clearly seen in the second derivative spectra [Fig. 2c]. Previous works have shown that ${\mathrm{A}}_{1121}$ arises from RNA absorbance, whereas the $1015{\mathrm{cm}}^{-1}$ shoulder is due to DNA.^{43, 44, 45}

From this region it is possible to derive two additional spectral biomarkers with outstanding statistical characteristics: ${\mathrm{A}}_{1121}∕{\mathrm{A}}_{1015}$ ratio (assigned as RNA/DNA ratio) and the wavenumber shift due to ${\nu }_s$ $\mathrm{P}\mathrm{O}_2{}^{-}$ (relative to $1082{\mathrm{cm}}^{-1}$ ). Even though the variability of these biomarkers is high due to overlapping absorbance, the average values of normal are still significantly different compared to malignant cells (Table 1).

3.2.

Early Stages of Malignant Cell Transformation

Both primary cells (MEF) and murine fibroblast cell lines (NIH/3T3) were infected with MuSV $(1\mathrm{ffu}∕\text{cell})$ and examined at various postinfection times for morphological and spectral changes. Figure 3 shows the expanded spectra of both cell cultures in the two wavenumber regions. This figure clearly demonstrates the gradual spectral variations following cell infection. Dramatic changes are observable in the case of MEF transformation, where the band at $1056{\mathrm{cm}}^{-1}$ decreases gradually and the band at $1082{\mathrm{cm}}^{-1}$ is shifted systematically to higher wavenumbers versus infection time [Fig. 3d]. Thus, it is possible to determine the first spectral signs of malignancy according to the alterations in the calculated values of the previously discussed biomarkers. The observed spectral changes in malignant cells compared to control cells are summarized in Table 2 . As can be seen in Table 2, the first morphological changes confirmed by microscopical observations and growth on soft agar appear considerably later than the first spectral signs. For example, the first spectral identification is possible on the first day ( ${\mathrm{A}}_{1121}∕{\mathrm{A}}_{1015}$ biomarker), while morphologically it can be discerned on the third day in the case of NIH/3T3 cell transformations (Table 2). In the case of MEF primary cells, the spectral changes induced by cell transformation were even more significant compared to those induced in the NIH/3T3 transformation [Fig. 3b]. Also in MEF primary cells, the first spectral signs appeared significantly earlier than the morphological changes (on the third day compared to the seventh day).

Fig. 3

FTIR spectra of (a) NIH/3T3 and (b) MEF cells at various intervals of postinfection time in the region $2820\text{to}3000{\mathrm{cm}}^{-1}$ . FTIR spectra of (c) NIH/3T3 and (d) MEF cells in various intervals of postinfection time in the region $1000\text{to}1145{\mathrm{cm}}^{-1}$ . Spectra are the average of (a) and (c) 20 transformations of NIH/3T3 cell lines and (b) and (d) 12 transformations of MEF cells.

4.1.

Cluster Analysis

Cluster analysis was used to classify the infected cells at each postinfection day into cancerous or normal groups. For this classification we utilized a vector array of spectral biomarkers, which were set as follows:

The results presented in Fig. 4 show that cluster analysis can indeed classify the infected cells into the cancerous group already at the first postinfection day in case of NIH/3T3 cells and at the fifth postinfection day in case of MEF transformation. We note that in both cases, the classifications were significantly earlier than the morphological identification of the malignant cells.

Fig. 4

Cluster analysis of (a) NIH/3T3 and (b) MEF cells at various intervals of postinfection. Cluster analysis is based on the average values of three biomarkers. Each postinfection day of transformation is represented using array of an average value of three biomarkers: ( ${\mathrm{A}}_{2958}∕{\mathrm{A}}_{2852}$ , ${\mathrm{A}}_{1121}∕{\mathrm{A}}_{1015}$ , and the shift of ${\nu }_s$ $\mathrm{P}{\mathrm{O}}_2^{-}$ ).

4.2.

Discriminant Classification Function

Discriminant classification function (DCF) is a statistical tool that enables us to improve discrimination between malignant stages by representing an adequate quantitative follow up of transformations versus time. DCF generates a classification score for each postinfection day, which is a linear combination of previously derived array of biomarkers with weight coefficients,^{37, 38} as can be seen in the following equation:

The weight coefficients were determined empirically in such a way that they nullify the average classification score of NIH/3T3 array (normal cell score) and yield 100 score (cancerous score) for the average NIH/MuSV array. The same weights were applied also to the MEF transformation. From the DCF analysis, we obtained a classification of the transformations that correspond to a sigmoid fit (Fig. 5 ). The abnormality can be distinguished as early as the first day in the case of the NIH/3T3 transformation [while first morphological identification is possible on the third day, Fig. 5a]. Similarly, Fig. 5b shows that the abnormality in the case of MEF transformation is apparent on the third postinfection day (while the first morphological identification is possible on the seventh day). In both cases, the infected cells reach an upper level plateau after full transformation (score of 100 for NIH/3T3 and score of 137 for cancerous MEF, Fig. 5).

Fig. 5

Discriminant classification function of (a) NIH/3T3 and (b) MEF cells at various intervals of postinfection. Each postinfection day of transformation is represented using an array of average values of three biomarkers: ( ${\mathrm{A}}_{2958}∕{\mathrm{A}}_{2852}$ , ${\mathrm{A}}_{1121}∕{\mathrm{A}}_{1015}$ , and the shift of ${\nu }_s$ $\mathrm{P}{\mathrm{O}}_2^{-}$ ).

Table 2

First signs of malignant transformation.

5.1.

${\mathrm{A}}_{2958}∕{\mathrm{A}}_{2852}$ ( ${\nu }_{\mathit{as}}$ $\mathrm{C}{\mathrm{H}}_3∕{\nu }_s\mathrm{C}{\mathrm{H}}_2$ )

The phospholipids/lipids/triglycerides and proteins absorb in the wavenumber regions from $2800{\mathrm{cm}}^{-1}\text{to}3000{\mathrm{cm}}^{-1}$ .^{1, 43} Previous studies^{46, 47} with rat fibroblast cell lines showed that lipids have more predominant absorbance relative to other biomolecules, including proteins, at the 2800- to 3000- ${\mathrm{cm}}^{-1}$ region. Also, changes in the absorbance due to ${\nu }_s$ $\mathrm{C}{\mathrm{H}}_2$ and ${\nu }_{\mathit{as}}$ $\mathrm{C}{\mathrm{H}}_3$ vibrational modes of lipids during carcinogenesis were found.^{48, 49} Our results showed remarkable increment in this ratio in malignant cells compared to normal cell cultures. Similar behavior of this biomarker was observed in leukemia,⁵⁰ cervical, colon, and colorectal cancer,⁵¹ as well as in murine fibroblast cell lines and rabbit bone marrow primary cells transformed by MuSV or H-Ras.⁵⁰ In addition, the $\mathrm{C}{\mathrm{H}}_3∕\mathrm{C}{\mathrm{H}}_2$ ratio was found to increase as a function of the progress in malignant lymphoma grade.⁵²

Lipids are considered as important components of the cellular membrane, which significantly affect its permeability and metabolites transportation during carcinogenesis,⁵³ and they also form an influential source of energy that might be essential for malignant metabolism. Moreover, the evidence that transformed cells differ in their average cellular volume compared to the normal cells⁵⁴ (Fig. 1) may also contribute to the observed changes seen in the previous ratio.

5.2.

${\mathrm{A}}_{1121}∕{\mathrm{A}}_{1015}$ (RNA/DNA) and Wavenumber Shift due to ${\nu }_s$ $\mathrm{P}{\mathrm{O}}_2^{-}$

The region between $900\text{to}1200{\mathrm{cm}}^{-1}$ has many overlapping bands that correspond to the nucleic acid absorbance.^{43, 44, 45} Differences in DNA isolated from cancer and normal cells/tissues using FTIR spectroscopy have been the basis of a number of studies for diagnosis of cancer.^{55, 56} A statistical comparison of the FTIR spectra of DNA obtained from prostate cancer and from normal prostate tissues of healthy younger men revealed a broad array of differences in base structures (e.g., N–H and C–O) as well as in vertical base-stacking interactions and in the phosphodiester-deoxyribose backbone.⁵⁵ Also, structural disorders in the pancreatic tumor DNA were detected in the phosphodiester-deoxyribose spectral region.⁵⁶

Our results showed significant increment in the ${\mathrm{A}}_{1121}∕{\mathrm{A}}_{1015}$ ratio and ${\nu }_s$ $\mathrm{P}{\mathrm{O}}_2^{-}$ peak shift to higher wavenumbers in malignant cells compared to normal cell cultures. The utilization of these spectral indicators was widely reported in previous studies. ^{7, 8, 9, 10, 12, 13, 19, 20, 34, 35, 36, 37, 57} The same tendency was observed for ${\mathrm{A}}_{1121}∕{\mathrm{A}}_{1015}$ ratio in melanoma,⁵⁸ leukemia,^{12, 13} cervical,^{19, 20} and colon,^{7, 8, 9, 10, 57} cancers as well as murine fibroblast cell lines transformed by MuSV or H-Ras^{34, 35, 36} and lymphoma.^{51, 59} In the case of ${\nu }_s$ $\mathrm{P}{\mathrm{O}}_2^{-}$ , it was found that the phosphodiester group shifted to higher wavenumbers in various malignant cell cultures such as human primary fibroblast, mouse primary fibroblast, murine fibroblast cell line (NIH/3T3), etc.³⁷ This shift has been also seen in breast cancer,⁶⁰ cancerous stomach tissues,⁶¹ and neoplastic human gastric cells.⁶²

It was suggested that the pivotal role of these biomarkers stems from nucleic acid absorbance. The change in the absorbance and conformation of nucleic acids (which can cause a shift in the absorbance of the phosphodiesters group) during carcinogenesis arises from a sharp increment of the proliferation and metabolic activity in the transformed cells and from the high levels of retrovirus DNA and RNA production in the infected or transformed cells. Also, these changes could arise from a variation in the nuclear volume of the transformed cells as previously reported.^{63, 64} Cell transformation progression in time can be well described by a sigmoid function that was obtained using a discriminant classification function. In a short interval of postinfection time (meaningfully shorter than the cell cycle), there are no detectable spectral differences between infected and control cell cultures. Then the spectral values of the infected cell cultures gradually approach the spectral values of the fully transformed cells (as can be seen in Fig. 5). After obtaining first signs of morphological transformation (as confirmed by microscope and growth on soft agar), all spectral indicators showed identical values to those of the fully transformed cells.

The results presented in this study prove the superiority of FTIR spectroscopy over the conventional technique used for detection of malignant cells in culture. Thus, FTIR-MSP in tandem with proper statistical tools may offer a promising technique for the detection of early stages of malignancy and for monitoring their progression.