Determining the Cytotoxicity of Oxidized Lipids in Cultured Caco-2 Cells Using Bioimaging Techniques

Fish lipids are comprised of considerable quantities of polyunsaturated acids and are prone to oxidation, producing reactive oxygen species and hydroperoxides. This study aimed to evaluate the biochemical and structural alterations in Caco-2 cells following exposure to 100 μg/mL methyl linoleate or fish oil, and then radiated for 24, 48 or 72 h. Electron spin resonance spectroscopy detected free radicals in the lipid membrane, Raman microscopy observed biochemical alterations and atomic force microscopy identified changes in morphology, such as the breakdown of DNA bonds. The study showed that bioimaging and biochemical techniques can be effective at detecting and diagnosing cellular injuries incurred by lipid peroxidation.


Introduction
Epidemiological studies suggest polyunsaturated fatty acids (PUFAs) may be protective against colorectal neoplasia, especially those extracted from fish lipids [1]. However, these fatty acids are also known to generate lipid peroxides, which accelerate the oxidation of proteins and contribute to the development of many degenerative diseases, such as diabetes, atherosclerosis and cancer [2][3][4]. Mechanistically speaking, research has found, through the use of electron spin resonance (ESR) spectroscopy, that the damage incurred by lipid peroxides is due to free radicals inserting themselves into lipid bilayers to compromise the integrity of cell membranes, ultimately leading to cell death. Necrosis and apoptosis, two forms of cell death, differ according to variations in their structural and biochemical properties; the former causes cells to swell and rupture [5,6], whereas the latter leads to cell shrinkage and DNA fragmentation. In addition, apoptosis is associated with stimulating cysteine aspartic acid-specific proteases recognised in form of caspases [7,8]. Notably, early investigations have found PUFAs to promote apoptosis and prevent cellular proliferation in vitro [9,10].
Traditionally, cellular toxicity has been assessed by various techniques, such as fluorescence microscopy and electrophoresis [11,12]. Each experiment that employs these techniques must take into careful consideration the amount of cells and materials needed, as well as logistics pertaining to the Molecules 2020, 25, 1693 2 of 16 staining procedure. Additionally, the entire process can be rather time-intensive and still fail to deliver desired results, sometimes taking up to three days to complete and not detecting anything remarkable. As such, new methods are being explored to overcome these challenges. One that has proven to be especially advantageous, in particular, is bioimaging. This requires minimal sample handling and no expensive chemicals, is efficient and can observe biological processes and compositional changes in tissues at the subcellular level.
Examples of two bioimaging techniques that have proven quite promising, as of late, are Raman microscopy (RM) and atomic force microscopy (AFM). RM uses a laser-based microscopic device that is able to distinguish between biological compounds (e.g., nucleic acids, proteins, lipids and carbohydrates) in human cells and tissues, viruses and bacteria, by way of the intensity and frequency of their molecular vibrations [13][14][15][16][17]. It has been utilized extensively to observe cell death in cancer studies [18,19]. AFM is a scanning probe that has been used to explore microbial surfaces, producing 3D images with molecular resolution [20,21]. With a tip the size of an atom, it can measure intramolecular and intermolecular forces associated with biological systems [21][22][23][24], calculate mechanical and biological response to mechanical stimuli by measuring force-distance curves [25][26][27], and provide high-resolution images [28,29].
For our study, Caco-2 cells isolated from human colorectal adenocarcinoma, were used to explore the cytotoxicity of PUFAs because they have been used extensively in related nutritional, pharmacological and toxicological investigations [30]. We aimed to evaluate changes in their molecular structure and morphology using RM, AFM and ESR.

Results and Discussion
Results showed that microspectroscopy allowed direct scanning of morphological variations at the molecular level, details from the Raman spectra analysis provided valuable information on cell damage resulting from endogenous or exogenous toxic compounds.

ESR Spectroscopy Images
ESR is considered as the first direct and sensitive method free radicals detecting for the study of lipid oxidation [31][32][33]. An increased production of reactive oxygen species (ROS) was detected following the oxidized lipid treatments for 4 and 24 h. Every molecule with unpaired electrons has a unique spectrum because of its nuclear spin and magnetic field. In physiological systems, ROS's are constantly produced by mitochondria during respiration, but they are quenched by antioxidants. Excessive amounts of ROS, however, will not be fully quenched and can lead to cellular damage.
ESR spectra of Caco-2 cells incubated with 24 h UV-radiated methyl linoleate for 4 or 24 h are shown in Figure 1. All samples were stored at −20 • C for one week before measurements. We previously established that free radicals are stable at −20 • C [34]. All samples showed an isotropic 3 lines spectrum with hyperfine coupling, typical of a nitroxide radical. For untreated cells, the signal had a g = 2.0053, a N = 15.46. In contrast, cells incubated with 24 h UV-radiated ML for 4 h, the signal had a g = 2.0091, a N = 14.76; cells with a 24 h incubation had a g = 2.0063, a N = 15.76. The change in a N values shows a shift in free radicals moving from lipids to protein sites [34,35]. The intensity of spectra for the oxidized ML samples was significantly higher than that of the Caco-2 cells, indicating that free radicals attacked the cell membrane and regenerated radicals. Of note, the intensity of spectra for Caco-2 cells treated with oxidized ML for 48 and 72 h was very low compared to that of the control, as well as other treatment times, so the results not shown. A decrease in the ESR signal from samples incubated with oxidized ML for 24 or 72 h, compared with 4 h incubation, suggests that free radicals are attacking other molecules and forming products, such as aldehydes [34,36].  Figure 2 shows a 50 μm × 50 μm tapping-mode from topographical images of untreated and treated Caco-2 cells. The untreated cells have smooth surfaces, lacking irregularity or pores ( Figure  3A), while cells incubated with 72 h UV-radiated fish oil or ML present a dramatic change in cell structure especially in the cell membrane and nucleus ( Figure 3B). There was a correlation between changes in cell height and morphology: cell height increased for untreated and treated (with 72 h UVradiated fish oil or ML) cells by 600 nm and 2000 nm, respectively, indicating disruption of the cell membrane and aggregation. A line profile analysis also showed dramatic differences between the samples: untreated cells had one peak and treated cells had many peaks, suggesting that lipid peroxidation products may be interacting with cellular components, leading to aggregation and crosslinking.  Figure 2 shows a 50 µm × 50 µm tapping-mode from topographical images of untreated and treated Caco-2 cells. The untreated cells have smooth surfaces, lacking irregularity or pores ( Figure 3A), while cells incubated with 72 h UV-radiated fish oil or ML present a dramatic change in cell structure especially in the cell membrane and nucleus ( Figure 3B). There was a correlation between changes in cell height and morphology: cell height increased for untreated and treated (with 72 h UV-radiated fish oil or ML) cells by 600 nm and 2000 nm, respectively, indicating disruption of the cell membrane and aggregation. A line profile analysis also showed dramatic differences between the samples: untreated cells had one peak and treated cells had many peaks, suggesting that lipid peroxidation products may be interacting with cellular components, leading to aggregation and crosslinking.

RM Images
Results show that information on all major cell components (lipids, proteins and nucleotides) is contained in the Raman spectra. Table 1 summarizes the assignment of wave numbers of some regions which, according to the literature, have specific biological significance [14,[37][38][39][40][41]. Band at 1449 cm −1 was allocated to C-H deformation vibrations in cell constituents [42][43][44][45] and was used to normalize the spectra, since it hardly changes within a certain cell type. Table 2 shows the comparative intensities of the bands.

Table 2.
Relative intensity values at selected regions of Raman spectra for Caco-2 cells treated with UV-radiated ML or fish oil (F). Data is expressed as mean ± SD for three baseline-corrected and normalized spectra of 128 co-added scans.   There were also changes in the protein bands, amide I at 1660 cm −1 , amide III in different conformations at 1238 cm −1 , 1258 cm −1 (β-sheet) and 1271 cm −1 (α-helices), CH2/CH3 bending vibration at 1449 cm −1 , CH2 twisted at 1340 cm −1 . The CH2 corresponding to the lipid content in the cell membrane contributed to the band at 1302 cm −1 and 1449 cm −1 band (CH2 scissor mode). Figure 5 shows a reduced amount of nucleic acids in the treated cells. This reduction is indicated by a decrease in the relative intensity in the band 720 and 782 cm −1 , corresponding to cytosine and thymine which decreased by 23%, 62% and 75% in samples treated with 24, 48 and 72 h UV-radiated fish oil, and 53%, 63% and 69% in samples treated with 24, 48 and 72 h UV-radiated ML. In addition, due to the break in the diphosphoester bonds, there was a decrease in the peak 1059 cm −1 by 20%, 50% and 45% in samples treated with 24, 48 and 72 h UV-radiated fish oil, and 38%, 47% and 44% in samples treated with 24, 48 and 72 h UV-radiated ML.  There were also changes in the protein bands, amide I at 1660 cm −1 , amide III in different conformations at 1238 cm −1 , 1258 cm −1 (β-sheet) and 1271 cm −1 (α-helices), CH 2 /CH 3 bending vibration at 1449 cm −1 , CH 2 twisted at 1340 cm −1 . The CH 2 corresponding to the lipid content in the cell membrane contributed to the band at 1302 cm −1 and 1449 cm −1 band (CH 2 scissor mode). Figure 5 shows a reduced amount of nucleic acids in the treated cells. This reduction is indicated by a decrease in the relative intensity in the band 720 and 782 cm −1 , corresponding to cytosine and thymine which decreased by 23%, 62% and 75% in samples treated with 24, 48 and 72 h UV-radiated fish oil, and 53%, 63% and 69% in samples treated with 24, 48 and 72 h UV-radiated ML. In addition, due to the break in the diphosphoester bonds, there was a decrease in the peak 1059 cm −1 by 20%, 50% and 45% in samples treated with 24, 48 and 72 h UV-radiated fish oil, and 38%, 47% and 44% in samples treated with 24, 48 and 72 h UV-radiated ML. There were also changes in the protein bands, amide I at 1660 cm −1 , amide III in different conformations at 1238 cm −1 , 1258 cm −1 (β-sheet) and 1271 cm −1 (α-helices), CH2/CH3 bending vibration at 1449 cm −1 , CH2 twisted at 1340 cm −1 . The CH2 corresponding to the lipid content in the cell membrane contributed to the band at 1302 cm −1 and 1449 cm −1 band (CH2 scissor mode). Figure 5 shows a reduced amount of nucleic acids in the treated cells. This reduction is indicated by a decrease in the relative intensity in the band 720 and 782 cm −1 , corresponding to cytosine and thymine which decreased by 23%, 62% and 75% in samples treated with 24, 48 and 72 h UV-radiated fish oil, and 53%, 63% and 69% in samples treated with 24, 48 and 72 h UV-radiated ML. In addition, due to the break in the diphosphoester bonds, there was a decrease in the peak 1059 cm −1 by 20%, 50% and 45% in samples treated with 24, 48 and 72 h UV-radiated fish oil, and 38%, 47% and 44% in samples treated with 24, 48 and 72 h UV-radiated ML.  A deconvolution technique was used for analysing the 1200-1400 cm −1 Raman region, which includes important bands pertaining to proteins, nucleic acids and lipids. The fitted peaks are the weak band 1367 cm −1 , assigned to symmetric CH 3 -stretched phospholipids; adenine, carbohydrates, vibrations and proteins of C-H deformation are at 1342 and 1320 cm −1 ; CH 2 -twisted phospholipids are at 1301 cm −1 ; peaks at 1284 (alpha-helix), 1258 and 1242 (beta-sheets) and 1231 cm −1 (random coil) as well as the C-C 6 H 5 -stretched phenylanaline at 1209 cm −1 . The in-plane deformation of =CH for lipids was donated to spectra ranging 1260 to 1300 cm −1 . Peak fitting in the 1200-1400 cm −1 region indicated a significant reduction in the spectra of treated cells compared to untreated cells, especially for samples treated with 72 h UV-radiated ML or fish oil ( Figure 6). The band of 1336 cm −1 corresponding to DNA bases adenine and guanine reduced from 38% to 21.7%, 20.4% and 17.8% (p < 0.03) in samples treated for 24 h with 24, 48 and 72 h UV-radiated fish oil, respectively. Similar decreases were also detected in cells incubated with oxidized ML for 24 h, decreasing by 13.6%, 21% and 18.2 % (p < 0.01) with 24, 48 and 72 h UV-radiated ML, respectively. Cellular membrane damage was investigated by monitoring peaks at 1231 cm −1 , which corresponded to a random coil and thereby indicated changes in protein conformation. It has been reported that protein strands correlate with alpha-helices and beta-sheets [46,47], and in our study, this conformational change was clearly seen by an increase in the amide I band from 78 in the untreated sample to 89 and 120 in samples treated with 24 and 72 h UV-radiated ML, respectively. Furthermore, the ratio of the 1302 cm −1 band to the amide III band assigned to the Raman intensity of lipid vibration at 1301 cm −1 and Raman intensity of amide III protein band decreased from 1.7 to 1.2, confirming structural damage to the cell membrane (Table 3). Similarly, the ratio of the 1336 cm −1 band to the amide III band also decreased.  The Raman band at 1743 cm −1 has been studied by FTIR and RM and is related to non-hydrogen bonded carbonyl C=O stretching in phospholipids. The intensity of 1743 cm −1 peak decreased from 0.2 in the untreated sample to 0.05, 0.04 and 0.01 for samples treated with 24, 48 and 72 h UV-radiated fish oil, and 0.04, 0.07 and 0.03 in samples treated with 24, 48 and 72 h UV-radiated ML respectively, suggesting oxidative damage indicative of apoptosis [48,49]. In addition, a substantial increase in the peak at 1604 cm −1 was observed; and this peak increased from 0.05 in the control samples to 0.12 and 0.086 in samples incubated with 100 μg/mL 72 h UV-radiated ML and fish oil, respectively (p < 0.05). According to previous studies, an increase in the band of 1602 cm −1 is symbolic of DNA fragmentation and damage to the cellular membrane, prominent features of apoptosis [46,50]. The Raman band at 1743 cm −1 has been studied by FTIR and RM and is related to non-hydrogen bonded carbonyl C=O stretching in phospholipids. The intensity of 1743 cm −1 peak decreased from 0.2 in the untreated sample to 0.05, 0.04 and 0.01 for samples treated with 24, 48 and 72 h UV-radiated fish oil, and 0.04, 0.07 and 0.03 in samples treated with 24, 48 and 72 h UV-radiated ML respectively, suggesting oxidative damage indicative of apoptosis [48,49]. In addition, a substantial increase in the peak at 1604 cm −1 was observed; and this peak increased from 0.05 in the control samples to 0.12 and 0.086 in samples incubated with 100 µg/mL 72 h UV-radiated ML and fish oil, respectively (p < 0.05). According to previous studies, an increase in the band of 1602 cm −1 is symbolic of DNA fragmentation and damage to the cellular membrane, prominent features of apoptosis [46,50].

PCA Findings
Principal component analysis (PCA) is a mathematical technique for reducing the dimensionality of datasets into less variables (principal components) in order to identify the most significant variations and patterns [51,52]. Component 1 of PCA explained 66.2% of treatment variance, component 2 explained 18.1% (Table 4). The peaks assigned to proteins (1002, 1660, 1238 and 1271 cm −1 ), lipids (1449 and 1340 cm −1 ) and nucleic acids (788, 828, 782 and 1095 cm −1 ) were responsible for the differences observed between untreated and treated cells. All Raman spectral characteristics that were responsible for the separation observed in the PCA plots are shown in Table 5 [53,54]. In order to compare untreated with treated cells, PC scores were plotted where each sample was placed as distinct by Raman data (Figure 7). 11

PCA Findings
Principal component analysis (PCA) is a mathematical technique for reducing the dimensionality of datasets into less variables (principal components) in order to identify the most significant variations and patterns [51,52]. Component 1 of PCA explained 66.2% of treatment variance, component 2 explained 18.1% (Table 4). The peaks assigned to proteins (1002, 1660, 1238 and 1271 cm −1 ), lipids (1449 and 1340 cm −1 ) and nucleic acids (788, 828, 782 and 1095 cm −1 ) were responsible for the differences observed between untreated and treated cells. All Raman spectral characteristics that were responsible for the separation observed in the PCA plots are shown in Table  5 [53,54]. In order to compare untreated with treated cells, PC scores were plotted where each sample was placed as distinct by Raman data (Figure 7).     The most striking peak intensities can be seen in Table 2 for the first two components. Separation on PC1 is largely influenced by the 1129 cm −1 band and less by 829, 1575 and 2935 cm −1 Raman bands attributed to nucleic acids, proteins and lipids, indicating variation in the cell contents between the samples. Additionally, the majority of the defining peaks are in PC1.
The PC1 vs. PC2 plots for untreated and treated cells were different, presumably due to the cytotoxicity of oxidized lipids. Untreated cells and cells with 24 h UV-radiated fish oil were clearly separate from treated cells (24,48  Cells treated with 24 and 72 h UV-radiated ML and 48 and 72 h UV-radiated fish oil were defined by high values in the C-N and C-C stretch, guanine, adenine (ring stretch) and Phe (1129, 1575 and 1606 cm −1 ), as well as low values for most of the variables assigned to DNA, lipid and proteins. Cells treated with 48 h UV-radiated ML showed a similar pattern, but to a lesser extent. Samples treated with 24 h UV-radiated fish oil were defined by a high content of DNA bands 829, 1202 and 829 cm −1 , and a low content DNA bands 1575 and 1606 cm −1 .

Sample Preparation
Cells were grown in 6-well plates until they were 80% confluent, then and incubated at 37 • C/24 h with lipids (100 µg/mL fish oil or ML) that had been UV-radiated for 24, 48 or 72 h. The cells were washed twice with PBS, scraped and centrifuged for 5 min at 300× g at 4 • C. The supernatant was removed and the pellet was stored at −80 • C until analysis.

ESR Spectroscopy Measurements
Collected cells were mixed with spin trap 2-methyl-2-nitrosopropane (MNP) in DMSO. This mixture was kept at −20 • C for one week before ESR measurement. ESR spectra were verified in a quartz tube using Jeol X-band spectrometer (Hmb, Germany), which operated at 9.71 GHz using a 100 kHz modulating magnetic field. First-derivative spectra were noted, developed into a PC by EW software (Scientific Services), then processed by Excel [34].

AFM Measurements
Cell suspensions were spin-coated onto pieces of mica and analysed by AFM (NTEGRA, NT-MDT, Moscow, Russia) in 24 h. A silicon cantilever (AC160TN, Olympus) was prepared with a sharp, conical tip and a curvature radius less than 10 nm. The nominal resonant frequency, fo, of the cantilever was 280 kHz, and the spring constant, k, was 42 N/m. Images were logged concurrently in topographic style and in phase style, the scan sizes ranging from 2 to 50 µm.

RM Measurements
Since the asymmetry of Raman bands in water molecules can interfere with sample spectra, our samples were freeze-dried. Raman spectra were recorded using a Renishaw System 2000 spectrometer equipped with a Leica microscope for both sample illumination and scattered light collection. A GaAlAs laser (wavelength 782 nm) was used for excitation, producing a maximum of 9 mW at the focal plane of a 50× objective, typically used to illuminate the sample. A grating of 1200 lines per mm was used to disperse the collected and laser-filtered Raman signal across a Peltier cool CCD camera. The spectral resolution was 4 cm −1 for all measurements. On-chip integration times ranged up to 128 s, and up to 16 spectra were added to improve the signal-to-noise ratio. The system's lateral spatial resolution was determined at 1.4 µm using a layered silicon sample, and the depth resolution was 7 µm using a thick polyethylene sample. The RS of every sample was taken as the average of RS from three dissimilar regions of three different experiments.

Statistical Analyses
Data analyses were conducted using SPSS (SPSS Inc., Chicago, IL, USA). One-way ANOVA and Tukey's post hoc tests were applied using significance at p < 0.05. Principal component analysis (PCA) was made for producing principal components (PCs), including a compact number of orthogonal changes which caused the most variance in the spectra. PC is connected to the original spectrum by the PC score, which characterizes a component's weight against the basic spectrum. In order to analyse the PCA data, a basic demonstration of PC coefficients with included variables was applied [53]. A positive or negative symbol indicates a PC coefficient that is a fourth to a half of the largest value, and a double positive or negative symbol indicates a PC coefficient that was higher than half of the absolute value. An empty space was assigned when the value was less than a fourth of the largest absolute value.

Conclusions
This study was the first to investigate the biomolecular modification of cells at the cellular level after exposure to oxidized lipids. Here, cytotoxicity following 24 h exposure of Caco-2 cells to 100 µg/mL oxidized lipids (ML or fish oil) was assessed. Analysis by ESR spectroscopy detected free radicals in the lipid membrane, and RM and AFM revealed changes in chemical composition and morphology, respectively. Such alterations were found to be primarily due to a breakdown of DNA and the cell membrane. Further studies are needed to monitor the conformational changes in DNA by RM and AFM.