1.

Introduction

Radiation therapy (RT) is a standard treatment used in $\sim 50\%$ of cancers worldwide. Its goal is to maximize tumor cell death while minimizing radiation-induced damage to healthy tissue.¹ Despite the tremendous progress made in improving the efficacy of RT, radiation resistance in cancer cells and radiation toxicity in nearby healthy tissue are recognized as major problems in RT.² The quest for solutions has led to improved technologies for delivering the RT dose,³ e.g., intensity-modulated RT (IMRT) and stereotactic ablative radiotherapy (SART). It has also led to increased efforts to gain a better understanding of the biological effects of radiation on cancer cells and the tumor microenvironment,^1–3 with the aim of identifying the molecular mechanisms of radiation resistance and toxicity of bystander cells. Ionizing radiation causes direct damage to DNA, including single strand and double strand breaks. In addition, it induces water radiolysis, producing free radicals such as reactive oxygen species (ROS) that can cause lipid, protein, and DNA base modification.^1,4–6 Increasing attention has been focused on the role of fatty acid metabolism in the radiation-induced response of cancer cells.⁷ Dysregulation in lipid metabolism of cancer cells⁸ has been implicated in increased radiation resistance⁹ in addition to cancer progression and aggressiveness.¹⁰ Hence quantitative methods to detect radiation-induced changes in lipid droplets (LDs) and lipid metabolism in cancer cells are required.

Previous studies have employed confocal fluorescence microscopy with lipid-conjugated or lipid-binding fluorophore probes to visualize changes in cellular LDs after radiation exposure of human cancer cells⁹ and murine cancer cells¹¹ of distinct origins including from the lung, breast, bladder, and colon. The fluorophore probes, such as Oil Red O and Nile Red, have an associated risk of altering the inherent biophysical properties of the lipids thus affecting the native dynamics of LDs.¹² Label-free imaging techniques based on Raman spectroscopy avoid such limitations. Raman spectroscopy involves the inelastic scattering of light due to vibrations of molecular bonds; it provides biochemical signatures of multiple macromolecules in a single acquisition. Over the past decade, Raman spectroscopy has been demonstrated as a formidable technique to investigate effects of ionizing radiation in cells and tissue.^13–15 However, since the spontaneous Raman signal is very weak, long acquisition times are involved for imaging thus making it impractical for high resolution imaging of subcellular organelles including LDs. In fact, there are only a limited number of studies reporting the application of spontaneous Raman microscopy to assess radiation-induced changes to LDs in cancer cells.^16–18 In contrast, coherent Raman imaging consisting of coherent anti-Stokes Raman scattering (CARS) and stimulated Raman scattering (SRS), relies on coherent Raman scattering and offers more than 1000-times faster imaging speed with subcellular resolution.¹⁹ SRS imaging greatly simplifies extraction of quantitative information and has significantly lower background compared to CARS imaging. SRS imaging of the significant high density of endogenous C-H chemical bonds (2800 to $3100{\mathrm{cm}}^{-1}$) has been demonstrated as an effective approach to image the biochemical response associated with cancer induced changes in DNA, lipids, and proteins of cells and tissue.^19–21 In particular, SRS imaging was used to investigate the lipid metabolism¹⁸ in cell models of prostate cancer¹⁰ and ovarian cancer,²² as well as to probe drug-cell interactions in prostate²³ and breast cancer cells.²⁴ Coherent Raman imaging has never been applied to measure the radiobiological response of cancer cells, to the best of our knowledge.

Here we expand upon our preliminary findings²⁵ and report the application of SRS imaging to detect the radiobiological response of human cancer cells for the first time. Human epithelial breast cancer (MCF7) cells were used as the model since their radiobiological response has been well characterized using confocal microscopy and Raman spectroscopy.^9,12,13 The SRS microscopy images were acquired in the C-H stretching region (2800 to $3100{\mathrm{cm}}^{-1}$) from live MCF7 cells at three time points: 24, 48, and 72 hrs after exposure to 0 Gy (no irradiation) and 30 Gy of X-ray ionizing radiation. We characterized the radiation-induced effects on the LD area per cell, lipid and protein content per cell, as well as the cellular area and morphology as a function of dose and time. Cell viability and confluency were measured using a live cell imaging system, and radiation-induced lipid peroxidation was assessed using BODIPY C11 staining and flow cytometry. This work thus substantially builds upon our previous report,²⁵ which only assessed the LD area per cell in the irradiated and control MCF7 cells at the different time points. Our results demonstrate that SRS imaging is a promising method for assessing the subcellular radiobiological response in MCF7 breast cancer cells.

2.

Materials and Methods

3.

Results

3.1.

Radiation-Induced Effects on Cell Confluency and Cell Viability

Cell confluency and death was measured every 4 hrs using an Incucyte S3 Live Cell Imaging system. We found that irradiation was not associated with any significant cell death up to 72 hrs after irradiation [Fig. 1(b)]. Cell confluency increased at similar rates in both control and irradiated cells up to 24 hrs but diverged between 24 and 72 hrs. While the confluency of control cells remained $>75\%$ confluency, the confluency of the irradiated population dropped $<75\%$. At 72 hrs, the confluency of control cells was close to 100% for control but only neared 90% for irradiated cells. Consistent with previous studies,³² these changes are likely associated with delays in proliferation and growth between the control and irradiated cells.

Fig. 1

(a) Confluency (%) and (b) cell death (%) as a function of time for irradiated and control MCF7 cells. These results are the mean values of three trials with error bars (± 1 SD) and were measured using the Incucyte S3 live cell imaging system.

3.2.

Radiation-Induced Lipid Changes in MCF7 Cells

To determine the changes in lipid metabolism, we imaged both unirradiated (0 Gy) and irradiated (30 Gy) MCF7 cells at the $2850{\mathrm{cm}}^{-1}$ (${\mathrm{CH}}_2$) band with contributions predominantly from lipids, and at the $2926{\mathrm{cm}}^{-1}$ (${\mathrm{CH}}_2$, ${\mathrm{CH}}_3$) band with overlapping contributions from proteins and lipids. Representative SRS images of unirradiated (0 Gy) and irradiated (30 Gy) MCF7 cells at 2850 and $2926{\mathrm{cm}}^{-1}$ and 72 hrs post exposure are shown in Figs. 2(a)–2(d). Figures 2(e)–2(h) shows the resulting spectrally unmixed images with the SRS images of lipid concentration showing darker nuclei, as expected. The threshold masks for the lipid images are shown in Figs. 2(i) and 2(k), and the composite lipid-protein images after spectral unmixing and LD segmentation are shown in Figs. 2(j) and 2(l).

Fig. 2

Representative SRS images of MCF7 cells taken at $2850{\mathrm{cm}}^{-1}$ (a,c) and $2926{\mathrm{cm}}^{-1}$ (b,d) for unirradiated (a,b) and irradiated (c,d). The corresponding spectrally unmixed lipid-rich (e,g) and protein-rich (f,h) images after applying MCR-ALS. The threshold masks (i,k) from the above lipid images and composite lipid-protein images (j,l).

To examine the time-dependent changes in the cellular lipids, we first generated segmentation masks of the LDs as described in Sec. 2.6 and evaluated the LD area per cell at the multiple time points. Figure 3 shows increasing LD area per cell with time, with a greater mean LD area per cell for irradiated than control, with this difference being most significant ($\mathrm{p}<0.001$) at 72 hrs for all three trials. The total lipid and protein intensities per cell were also measured and found to show a significant increase after 48 and 72 hrs in the 30 Gy-irradiated cells compared to the control, as can be seen in Fig. 3.

Fig. 3

(a) Average LD area per cell, (b) total lipid intensity per cell, and (c) total protein intensity per cell over 72 hrs. LD Area/cell is the average LD area per cell for an imaged region ($n=10$). (ns = no significance, *$\mathrm{p}<0.05$, **$\mathrm{p}<0.01$, ***$\mathrm{p}<0.001$, and ****$\mathrm{p}<0.0001$).

3.3.

Differences in the Cellular Morphology of Control and Irradiated Cells

To determine whether the increase in LDs correlated with cellular expansion, we next evaluated the area of the control and irradiated cells at the different time points. We observed a significant increased cell size in the irradiated cells compared to control at each time point [Fig. 4], with this difference in average size increasing with time. An example of this cellular enlargement in the irradiated MCF7 cells can be seen in the representative SRS image at $2926{\mathrm{cm}}^{-1}$ in Fig. 4 and is characteristic of senescent cells.¹¹

Fig. 4

Representative cell images at $2926{\mathrm{cm}}^{-1}$ for (a) 0 Gy and (b) 30 Gy at 72 hrs. (c) Mean cell area at each time point where the cell area was determined from protein images ($n=10$). (**$\mathrm{p}<0.01$, ***$\mathrm{p}<0.001$, and ****$\mathrm{p}<0.0001$).

Besides the larger average cell size for irradiated cells compared to control, other morphological differences were also observed in the SRS images of the irradiated cells including increased vacuole formation in the cytoplasm and multinucleation, a sign of mitotic catastrophe [Fig. 5].

Fig. 5

Some regions of MCF7 cells contained multi-nucleated (red circles) cells (a,b) and even more regions (c,d) contained cells containing vacuoles (arrows).

3.4.

Lipid Peroxidation Assay

To determine whether the radiation exposure caused lipid peroxidation, MCF7 cells were exposed to 0 Gy or 30 Gy, stained with C11-BODIPY and analyzed by flow cytometry over a three-day recovery period. The results for a representative trial are shown in Fig. 6 where the colored histograms refer to three populations of MCF7 cells as follows: the 0 Gy is the unirradiated control group stained with BODIPY; the 30 Gy is the irradiated group stained with BODIPY, and the control group is neither irradiated nor stained with BODIPY to establish a baseline relative to the stained samples. As seen in Figs. 6(a)–6(c), a significant increase of $\sim 53\%$ in the relative MFI for this trial is seen for the 30 Gy irradiated MCF7 group after 72 hrs compared to the two control groups suggesting an increase in the lipid peroxidation in irradiated MCF7 cells after 72 hrs. The average relative increase in MFI over all three trials was 67%.

Fig. 6

Flow cytometry data for MCF7 cells over a 3-day recovery period at (a) 24, (b) 48, and (c) 72 hrs for a single representative trial. Histograms show profiles for unstained (0 Gy;orange), control (0 Gy with BODIPY;red), and irradiated cells (30 Gy irradiated with BODIPY;blue). The $Y$-axis represents the normalized cell count as a percentage of the Max, and the $X$-axis represents the fluorescent intensity of the portion of C11-BODIPY dye that was oxidized after interaction with peroxyl radicals, emitting in the green region (490 to 510 nm) of the visible spectrum.

4.

Discussion

Human cancer cell lines have varying levels of intrinsic radiosensitivity.^13,33 Here we wanted to evaluate the effects of ionizing radiation on cells with limited radiosensitivity. Hence we investigated the MCF7 human breast cancer cell line since it is known to be relatively radiation resistant with the surviving fraction after 2 Gy (${\mathrm{SF}}_2$) value $>0.6$.¹³ It is important to note that the ${\mathrm{SF}}_2$ value determined from a long-term clonogenic survival assay, measures the potential for proliferation post-irradiation and is different from the measurements of cell viability. Consistent with the relatively radiation-resistant phenotype, our results [Fig. 1] show no difference in % cell viability in unirradiated versus irradiated cells over the 72-hr time frame. Irradiated cells did, however, show signs of reduced proliferative potential [Fig. 1(a)]. Interestingly, MCF7 cells contain a wild type p53 gene, which is responsible for G1 phase cell cycle arrest post-irradiation.^13,33 Therefore, irradiated MCF7 cells will remain viable after irradiation but will have a reduced clonogenic potential due to cell cycle arrest.

As seen in Fig. 3, we found a significant increase in LDs in the 30 Gy-irradiated cells at 72 hrs post-radiation. Cancer cells are known to have higher levels of LDs due to metabolic reprogramming.³⁴ LD accumulation has been seen in human breast, bladder, lung, neuroglioma, and prostate cancer cells in response to 6 Gy X-rays treatment.⁹ Adjemian et al. reported a similar increase in LDs in a panel of murine cancer cells irradiated at high ($>20\mathrm{Gy}$) X-rays dose and observed cellular features related to methuosis.¹¹ Their accumulation has been correlated to their survival after irradiation. It is also closely connected with iron metabolism. Free iron within the cell leads to the generation of ROS through the Fenton reaction (${{\mathrm{Fe}}_2}^{+}+{\mathrm{H}}_2{\mathrm{O}}_2\to {{\mathrm{Fe}}_3}^{+}+•\mathrm{OH}+{\mathrm{OH}}^{-}$). The hydroxyl radical produced causes lipid peroxidation, targeting unsaturated fatty acids. Iron chelation by desferoxamine (DFO) was able to induce LD accumulation and this, in turn, conferred higher survival ability to cells after radiation treatment, as shown by clonogenic assays.⁹ It has been suggested by Tirinato et al. that LDs could serve as an antioxidant system, being able to buffer the excess of ROS commonly found in cancer cells and prevent damage to other cellular components, such as proteins and nucleic acids. Earlier work using Raman spectroscopy found increased levels of proteins in MCF7 cells over 72-hr post irradiation¹³ that were attributed to the cellular radioadaptive response. Consistent with previous studies, we found an overall increase in the lipids and protein intensity per cell [Fig. 3] in the 30 Gy irradiated cells compared to the control cells over 72 hrs post-irradiation. Interestingly, our results of Figs. 3(b) and 4(c) indicate that the total lipid intensity per cell and the mean cell area, respectively, show significant increase at 24 hrs in irradiated cells compared to the control. On the other hand, the LD area per cell [Fig. 3(a)] and the overall protein intensity per cell [Fig. 3(c)] show comparable response in irradiated and control cells at 24 hrs. Hence, we hypothesize that lipid accumulation and endoplasmic reticulum (ER) expansion involving hypertrophy of ER, Golgi, and possibly mitochondria membranes, occur concurrently at 24 hrs post-irradiation. LD formation is induced by ER stress to maintain lipid, protein, and calcium homeostasis.³⁵ In particular, hypertrophy of the smooth ER where lipids are synthesized, occurs in the initial stages of ER stress followed by LD formation. We theorize that irradiation of cells by 30 Gy X-rays may result in initial ER stress and subsequent LD formation, as has been shown in other studies.³⁶ Moreover, Fig. 3(a) suggests that the irradiated cells may already be accumulating lipids from the extracellular fluid but have not yet begun an increase in production of LDs at 24 hrs.

Methuosis is a nonapoptotic mechanism characterized by massive vacuolation, accumulation of LDs, macropinocytosis, mitochondrial dysfunctions, and eventual implosive, non-autophagic cell death.³⁷ Vacuoles and LDs are generated by expansion of the ER and can be induced by iron chelators, such as DFO.³⁸ Exposure of cancer cells to ionizing radiation triggers an accumulation of unsaturated fatty acids that are subject to lipid peroxidation in the presence of free iron. As a compensatory effect, the induction of the ferritin heavy chain protein may sequester iron to prevent further damage from lipid peroxides but may trigger a methuosis-like state in the cell due to iron depletion. Consistent with earlier work, our SRS imaging study revealed methuosis-like features, such as cell area expansion [Fig. 4] and vacuolization [Fig. 5] in irradiated MCF7 breast cancer cells in the later stages (after 3 days) of recovery from irradiation. We also found an increase in lipid peroxidation in MCF7 cells after 72 hrs [Fig. 6] using the lipid peroxidation sensor BODIPY C11. A key difference from true methuosis is that cell death was not observed in this study. MCF7 cells continued to proliferate, albeit slower after irradiation, while displaying methuosis-like features. This may either be a contributing factor, or as a result of, their radiation resistance. Further studies on the effects of irradiation in the presence or absence of cellular iron, as well as the effects of this on lipid peroxidation and methuosis-like states in cancer cells, are required.

The radiation dose of 30 Gy investigated here is more relevant to high dose RT modalities such as IMRT and SART than conventional, 2 to 10 Gy dose regimens. Our future SRS imaging work will investigate radiation response at these lower doses, including radiation-induced effects on LDs composition at different time points. Although the dose rate used in this study was constant at $1\mathrm{Gy}/\min$ during the whole irradiation period, our methods can be extended to investigate samples irradiated at different dose rates. Moreover, the SRS imaging methodology and analysis techniques demonstrated here can be readily applied to normal cells and tissue for investigating the effect of ROS in radiation-induced normal tissue damage.⁶ Our future work will simultaneously probe changes in lipids, DNA, and proteins using multiple Raman bands for SRS imaging. The use of SRS imaging strategies utilizing stable isotope labeling and bio-orthogonal tags^18–21 will be explored to increase the molecular specificity and sensitivity of detection of radiation-induced effects.

5.

Conclusion

This pilot study demonstrated the potential of SRS imaging for investigating ionizing radiation-induced changes in cancer cells. The high spatial resolution imaging approach enabled quantitative visualization of the changes in lipids and morphological features in 30 Gy irradiated MCF7 cells compared to unirradiated control cells, albeit without the use of fluorescent labels. Future investigations of irradiated cancer as well as normal cells and tissue using SRS imaging have the potential to provide critical insights on radiation resistance in tumors, radiation toxicity in surrounding healthy tissue, and non-targeted bystander effects. This will aid in the path toward the development of personalized RT.