Open Access
Add to BibSonomy
Abstract
Drug potency assessment plays a crucial role in cancer chemotherapy. The selection of appropriate chemotherapy drugs can reduce the impact on the patient's physical condition and achieve a better therapeutic effect. Various methods have been used to achieve in vitro drug susceptibility assays, but there are few studies on calculating morphology and texture parameters quantitatively based on phase imaging for drug potency assessment. In this study, digital holography microscopy was used to get phase imaging of ovarian cancer cells after adding three different drugs, namely, Cisplatin, Adriamycin, and 5-fluorouracil. Based on the reconstructed phase imaging, four parameters of ovarian cancer cells changed with time, such as the average height, projected area, cluster shade, and entropy, were calculated. And the half-inhibitory concentration of cells under the effect of different drugs was calculated according to these four parameters. The half-inhibitory concentration, which can directly reflect the drug potency, is associated with the morphological and texture features extracted from phase images by numerical fitting. So, a new method for calculating the half-inhibitory concentration was proposed. The result shows that the morphological and texture feature parameters can be used to evaluate the sensitivity of ovarian cancer cells to different drugs by fitting the half-inhibitory concentration numerically. And the result provides a new idea for drug potency assessment methods before chemotherapy for ovarian cancer.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
As the one of most common cancers in women worldwide, ovarian cancer is considered one of the most challenging tumors to treat due to its biological and molecular complexity, difficult early detection, high morbidity and low survival rate of late-stage [1–3]. Currently, the primary treatment for ovarian cancer is surgical resection combined with postoperative chemotherapy [4,5]. The choice of chemotherapy drugs mainly depends on the clinician's experience and clinical response rate [6]. In the course of chemotherapy, the resistance of cancer cells to chemotherapeutic drugs is also an essential reason for the poor prognosis of patients [7,8]. Therefore, the implementation of personalized in vitro drug susceptibility testing and drug potency assessment is significant for improving the prognosis and survival rate of patients. Many studies have been carried out to determine drug susceptibility in vitro. For example, labeling factors that characterize cell death (BCL2 family, caspase, etc.) [9,10]; judging cell activity by assaying cell metabolites (lactate, ATP, etc.) [11,12]; performing 5-diphenyltetrazolium bromide (MTT) assay to determine cell half-inhibitory concentration (IC50), and so on [13]. Among them, IC50 is a parameter used to measure the effectiveness of drugs and is commonly used to describe the response of cells to different drugs [14]. Several methods have been proposed to assess the potency of different drugs on certain cells by predicting IC50 values. For example, Gregory et al. used random forests to build a regression model between the gene expression signatures of cell lines and the corresponding IC50 values for each drug [15]; Michael et al. developed machine learning models by inputting chemical properties of drugs and molecular characteristics of cell lines to predict IC50 values [16]; James et al. applied a Bayesian multitasking approach to predict drug response based on genomic, epigenomics, and proteomic datasets of human breast cancer cell lines [17]; Wang et al. used a similarity-regularized matrix factor approach, which predict the IC50 values of unknown drugs by the similarity between medicinal chemical structures and gene levels of cell lines [18]; Minjae et al. constructed a one-dimensional convolutional neural network model called DeepIC50 to predict cellular responses to different drugs by multiple features including mutation status and drugs’ molecular fingerprints [19]. The above methods mostly complete the evaluation of drug potency by establishing the mapping relationship between the features at the molecular level or the gene level and the IC50 value, which requires a large number of data sets and complex experimental determinations. There are still many challenges in dynamism and practicality.
It is well known that when chemotherapeutic drugs act on cancer cells, they react non-specifically with various subcellular components through different signaling pathways [20]. Different drugs have different mechanisms of action, including binding to DNA to prevent their replication, affecting RNA transcription, and affecting ribosomes to prevent the production of proteins, etc. Thereby, the purpose of inhibiting or killing tumor cells is achieved by affecting the proliferation and division of cells [21]. Morphological and textural features of cells are generally considered to directly characterize cellular function [22]. Thus, the effects of chemotherapeutic drugs on cellular function can be reflected by morphological or textural features. Recently, some studies have used optical microscopy to observe stained cells or use phase contrast microscopy to observe living cells, and employ the morphological parameters, such as membrane area, nuclear-cytoplasmic ratio, and nuclear optical density, to evaluate cell morphological changes [23,24]. Furthermore, the texture feature such as uniformity and kurtosis of tumors can be obtained by contrast-enhanced computed tomography [25] or dynamic contrast-enhanced magnetic resonance imaging [26]. These features are used to quantify the heterogeneity of tumors and analyze the response of tumor cells to chemotherapy. Therefore, these results show that the morphological and texture characteristics of cells can act as quantitative markers of cell viability and efficacy of chemotherapy drugs, and the degree of cell response to different chemotherapeutics can be assessed by analyzing morphological and texture parameters. It is possible to use optical methods to extract cell images and analyze the morphological and texture features of cells so as to conduct drug potency assessment and drug sensitivity studies.
Digital holographic microscopy (DHM) is a non-label and non-contact phase imaging technique that can meet the needs of real-time dynamic observation [27]. Therefore, it has been widely used in biomedical research to observe the dynamic changes of living cells in real-time [28–30], such as calculating the phase height of pancreatic tumor cells treated with paclitaxel [31], monitoring the cell death process and studying its mechanism through observing the change of cell morphology [32], and tracking cell growth through each cycle [33] by DHM.
In this study, we aimed to establish a novel method to evaluate the effectiveness of different drugs on ovarian cancer cells. Cell morphology and texture parameters calculated from phase images are used to obtain IC50, which is commonly used to evaluate drug efficacy by fitting quantitatively. In this method, four parameters are selected, such as ${S_{Cell}}$ (cell projected area), ${h_{ave}}$ (average height), $Clu$ (cluster shade), and $Ent$ (entropy) to quantitatively and dynamically describe the response degree of ovarian cancer cell A2780 at each moment when three drugs, Cisplatin (DDP), Adriamycin (Adr), and 5-fluorouracil(5-Fu) are added. Then we used the numerical fitting method to relate these four parameters with the biomedical parameter IC50 and evaluate the effectiveness of the three drugs accurately.
2. Materials and methods
2.1 Cells culture and drug preparation
A2780 is an epithelial ovarian cancer cell line isolated from human female ovarian cancer patients who have not received any treatment. Due to its weak adhesion to matrix membrane, A2780 has strong migration and invasion ability, resulting in a high degree of malignancy. The A2780 cells used in this research were purchased from ATCC (American Type Culture Collection) and preserved in the Obstetrics and Gynecology Laboratory of Peking University People's Hospital. A2780 cell lines were seeded in a 100 mm glass Wilson dish at a density of 1×107 cells per dish and cultured in the full medium [1640 (RPMI Medium 1640 basic 1X, GIBCO, China) supplemented with L-glutamine, 15 mM HEPES, and 10% fetal bovine serum (Gibco 10099-141, Australia)] at 37°C in 5% CO2 for 24h period.
Cisplatin (DDP), Adriamycin (Adr), and 5-fluorouracil(5-Fu) are chemotherapy drugs commonly used in the treatment of ovarian cancer. DDP inhibits cell mitosis by binding to DNA and destroying its function. Adr is an antibiotic that can inhibit the synthesis of DNA and RNA. 5-Fu can be incorporated into RNA to interfere with the synthesis of proteins, thereby inhibiting the malignant proliferation of tumors. The sensitivity of A2780 ovarian cancer cells to these three drugs was determined in this experiment. For distinguishing the effects of the three drugs on the cells in a short period, the three drugs were dissolved in the medium to prepare a solution with a concentration of 100µM. Before the experiment, the experimental groups were filled with a drug solution, and the control group was added with a fresh medium without a drug.
2.2 Determination of the half-maximal inhibitory concentration (IC50) by absorbance-based commercially available kit
The drug's IC50 of cancer cell viability is widely used to measure its effect on tumor. Ovarian cancer cell line A2780 viabilities were determined by Cell Counting Kit-8 (CCK-8; Dojindo, Kumamoto, Japan) assay following treatment with different drugs (DDP, Adr, 5-Fu) for 48 h. Briefly, cells were seeded in 96-well tissue culture plates (100 µl/well) at a density of 5×103 cells/well. After the cell monolayer reached 50-60% confluence, the culture medium of each well was refreshed with 100 µl medium combined with the above drugs at a concentration of 5µM, 10µM, 20µM, 40µM, 60µM, 80µM, and 100µM respectively. The control group was added with a fresh culture medium without drugs. After 48 h, the cells were incubated with CCK-8 solution (10µl/well) for 2 h to determine the number of live cells. A Multiskan Spectrum spectrophotometer (Thermo Fisher Scientific, Inc.) was used to measure absorbance at 490 nm. The IC50 was calculated using the following equation: IC50= (mean OD of specific treatment group)/ (mean OD of a negative control group) %.
where, $O{D_{treatment}}$and$O{D_{control}}$are the absorbance of the treatment group and control group, respectively. The cell viability curve was plotted with drug concentration as abscissa and cell viability rate as ordinate. The concentration corresponding to a 50% cell viability rate is the IC50 value.2.3 Cell observation by Zernike phase microscopy and fluorescence imaging
The cell morphology of the control group and the experimental group before adding drugs and after adding different drugs for 2 hours were qualitatively observed under an optical microscope.
Immunofluorescence staining was used to observe the distribution of F-actin in cells after adding different drugs. The attached cells were washed briefly with PBS three times and fixed with 4% paraformaldehyde for 20 min. F-actin was treated with phalloidin-Rhodamine (1:500 solution for 1 h) (Invitrogen, Carlsbad, California) and labeled red. Nuclei were treated with DAPI (1:1000 solution for 5min) (Invitrogen) and labeled blue. Fluorescence imaging was performed using a laser scanning fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
2.4 Digital holography microscopy and data collection
2.4.1 Digital holography microscopy and data collection
The experimental setup is based on an off-axis Mach-Zehnder holographic interferometer, and the light source is a solid-state laser with a power of 100mW and a wavelength of 532nm (MSL-U-532, China). As shown in Fig. 1, the beam is divided into a reference beam and an object beam by the polarizing beam splitter (PBS). Then the two beams are filtered and expanded by spatial filters (SF) separately to produce a collimated plane wave. The reference beam passes through the attenuator and half-wave plate (HWP) before expansion and collimation so as to compensate for the attenuation of the sample to the object beam and to adjust the polarization component, which is the same as the object beam to the maximum ensuring better interference image. After traversing the object, the object beam is captured by an imaging system composed of the objective lens (20×, NA = 0.4) and field lens (FL), and then spreads to the camera surface. The reference beam is directly reflected by the mirror and interferes with the object beam on the camera surface to produce an interference pattern, which is a digital hologram. A 2048×2048-pixel hologram was recorded by a CCD camera (2048×2048 pixels, 5.5µm, PointGrey, Canada).
2.4.2 Digital hologram recording and phase reconstruction
We prepared 4 dishes of A2780 cells, one dish was used as a control without any treatment, and the other three dishes were added with DDP, Adr, and 5-Fu, respectively. The recorded dish of cells was placed on the experimental platform at room temperature, and the other dishes of cells were kept in a constant temperature box at 37°C to maintain the normal physiological activity of the cells. The hologram of the cells’ changes over time after being added to the drug was recorded. The field of view was recorded per second for a total of 2 hours, and the control group was recorded in the same way. About 25 single cells could be separated from the recorded field of view of each dish for subsequent calculation and analysis. The experiment was repeated three times under the same experimental conditions to reduce experimental errors and increase the stability of experimental results.
The phase distribution was reconstructed from the recorded holograms through an angular spectrum algorithm [34]. Automatic focusing was used to eliminate the reconstruction error caused by defocusing. Tilt aberration caused by the off-axis was corrected by a straight-line fitting on the holographic plane. Using the background mask with the cell region removed and the Zernike polynomial fitting method, the inherent high-order aberration of the system itself was used to compensate precisely. Finally, an offset procedure was applied so that the background of the phase image was close to zero.
2.5 Calculation of cellular parameters and IC50 value fitting from label-free acquired DHM images
2.5.1 Single cell segmentation
Before calculating parameters, every single cell should be separated, then we calculated the parameters of each cell and calculated their average value to reduce the random error.
Firstly, a mask was made to extract single cell division into polygon areas, and then the threshold method was adopted for segmenting cells in each polygon areas. Segmentation results retained the largest connected region and filled the void to realize the aim of de-noising. Finally, after obtaining the cell mask, it was multiplied with the original phase image to obtain the phase image of every single cell.
2.5.2 Single cell segmentation
We selected four parameters including area, average height, cluster shade, and entropy, to describe the morphological and texture features of cells comprehensively.
As shown in Eq. (2), the area of a cell should be the number of pixels occupied by a single cell multiplied by the ratio of the pixel size and the system magnification. When the system is fixed, the pixel size and magnification are both constant. For the convenience of calculation, the area of a cell is represented by the number of pixels in this paper.
The phase value is determined by the refractive index difference between cells and culture medium and the height of the sample, which is shown in Eq. (3). Since the refractive index difference is almost unchanged, it can be considered as a constant. Therefore, the phase value can be directly used to characterize object height. The average height, which can be calculated by Eq. (4), is the average value of the phase values of each pixel in the single-cell region.
The texture feature parameters are derived from the statistical method of the gray level co-occurrence matrix (GLCM) [35]. GLCM describes texture by studying the spatial correlation characteristics of the gray level. The calculation method of GLCM is shown in Eq. (5).
Cluster shade is a measure of matrix skewness and is used to describe the symmetry of an image. The larger the cluster shade, the more asymmetrical the cell. The calculation method is shown in Eq. (6).
Entropy is a measure of the amount of information in an image, which is used to indicate the degree of non-uniformity or complexity of texture in an image. The higher the entropy, the more homogenous the cell. The calculation method is shown in Eq. (9).
In order to compare the changes of parameters with different drugs, as shown in Eq. (10), all parameters are normalized.
where, $par{a_{nor}}$ is the parameter after normalization, $par{a_t}$ is the parameter after t minutes of adding a drug, $par{a_0}$ is the parameter when t = 0.In addition, the parameter change rate is defined as Eq. (11), which is used to characterize the change degree of the cellular parameter before and after adding the drug.
2.5.3 IC50 value fitting and verification
By fitting the curves of the four parameters with time mentioned in Section 2.5.2 respectively, four functions can be obtained, that is ${S_{Cell}} = {f_1}(t)$, ${h_{ave}} = {f_2}(t)$, $Clu = {f_3}(t)$, $Ent = {f_4}(t)$.
Taking the cell area${S_{Cell}}$as an example, if the fitting is a first-order function, the coefficient of variable $\textrm{t}$ in the expression will be different when different drugs are added. That is ${S_{Cell\_DDP}} = {a_1}t + {a_0}$, ${S_{Cell\_5Fu}} = {a_2}t + {a_0}$, ${S_{Cell\_Adr}} = {a_3}t + {a_0}$. At this point, the intensity of change of cell area with time can be measured by the coefficients${a_1}$, ${a_2}$, and${a_3}$.Similarly, three groups of coefficients${b_1}$, ${b_2}$, ${b_3}$; ${c_1}$, ${c_2}$, ${c_3}$ and ${d_1}$, ${d_2}$, ${d_3}$ can be obtained after fitting the average height ${h_{ave}}$, cluster shade $Clu$ and entropy $Ent$ with time.
These four groups of coefficients are given a certain weight and added together, as shown in Eq. (12), to obtain a compound parameter, which is used to describe the influence of these four parameters on IC50 comprehensively.
With A as an independent variable and known IC50 values corresponding to different drugs as dependent variables, quadratic curve fitting was performed and Eq. (13) was obtained.
where, m, $n$ and $k$ are the coefficients of the fitting formula.According to Eq. (13) and Eq. (12), given the values of four parameters${S_{Cell}}$, ${h_{ave}}$, $Clu$ and $Ent$ at a certain time ${t_0}$ and the current time${t_0}$, the IC50 value can be calculated to judge the type of drugs added.
In order to prove the effectiveness of the proposed method, three more experiments were carried out. In order to obtain as many cell phase images as possible, the field of view scanning was adopted. Ten fields of view were scanned for each dish of cells in each experiment, with a 4-minute cycle for a total of 2h.
3. Results
3.1 IC50 value of A2780 cells to different drugs determined by absorbance-based kit
The cell viability with multiple drug concentrations under different drugs was shown in Fig. 2. When the viability was 50%, the corresponding concentration values of DDP, Adr and 5-Fu were 7.896µM, 0.2707µM, and 33.46µM, respectively, which was the IC50 value of the three drugs. The larger the IC50 value was, the weaker the drug's potency on cancer cells was. It can be seen that Adr has the strongest cytotoxicity, followed by DDP, and 5-Fu has the weakest cytotoxicity.
Fig. 2. The curve of cell viability under different drugs determined by absorbance-based kit. (The concentration values correspond to the intersection of the dotted line with the viability of 50% and the curves are the determined IC50 values of each drug.)
3.2 Changes of cell morphology by Zernike phase microscopy and fluorescence imaging after adding different drugs
Figure 3 shows the A2780 cells under the optical microscope before and after the addition of different drugs for 2h. There is no significant change in morphology seen for a single cell, but the cell quantity could qualitatively reflect the inhibitory effect of drugs on cells’ proliferation. The number of cells still increased greatly after the addition of 5-Fu, indicating that 5-Fu had the weakest cytotoxicity to A2780 cells. In contrast, the number of cells did not change significantly after the addition of Adr. It indicated that Adr had the most obvious inhibitory effect on the proliferation of A2780 cells. These results show that traditional optical microscopy can only qualitatively reflect the differences in the drugs’ effects on cells, but cannot quantitatively calculate and analyze drug potency.
Fig. 3. The Zernike phase contrast images of A2780 cells after the addition of different drugs for 0 h and 2 h.
We examined the F-actin cytoskeleton in A2780 cells treated with different drugs. As shown in Fig. 4, all images were captured under the same condition. The F-actin fibers of Adr-treated cells were short and irregular, while the other groups had long and regular fibers, indicating that Adr has the strongest cytotoxicity to A2780 cells. The changes in the cytoskeleton observed by immunofluorescence staining also provided the possibility to assess the drug potency by extracting the texture features of the cells.
Fig. 4. Immunofluorescent images demonstrated the F-actin structure and distribution of A2780 cells treated by different drugs, with F-actin and DAPI stained in red and blue.
3.3 Change process and change rate of cellular parameters after adding different drugs
Three dishes of A2780 cells added three different drugs, 5-Fu, DDP, and Adr, respectively, while the control group did not do any treatment. Four groups of cells were placed in the same environment, and each group was recorded for 2h. Figure 5 shows the phase maps of four groups of a single cell at 0, 20, 40, 60, 80, 100, and 120min. From Fig. 5, it can be seen qualitatively that the cell morphology change in the control group was the least, followed by the cells added with 5-Fu and DDP, and the most dramatic change was in the cells added with Adr, which was consistent with the IC50 value.
We quantitatively measured the changes of ${S_{Cell}}$, ${h_{ave}}$, $Clu$ and $Ent$ with time after adding 5-Fu, DDP, and Adr, as well as the change rates of these four parameters. The parameter value at each time point is the average value of this parameter calculated from every single cell. As shown in Eq. (10) and Eq. (11), the parameter value itself and its change rate are normalized, so they are all dimensionless quantities. And it's worth saying, the curve in the Fig. 6 reflect the change of the parameter at that time corresponding to 0min, so they start from the same point. As shown in Fig. 6, the changes of ${S_{Cell}}$, ${h_{ave}}$, $Clu$ and $Ent$ with time in the experimental group adding Adr were the most drastic, followed by the experimental group adding DDP. The changes of these four parameters in the experimental group adding 5-Fu showed no significant difference from the control group, which changed slowly or almost remained unchanged. In addition, for each parameter, the changing trend was consistent with the addition of different drugs, and the difference was only in the change rate. For example, the ${S_{Cell}}$ and $Ent$ gradually decreased with time, while the ${h_{ave}}$ and $Clu$ gradually increased, which was consistent with our expectation.
Fig. 6. Temporal evolution of (a)cell projected area; (c)average height; (e)cluster shade;(g) entropy. (b)(d)(f)(h) are the change rate of these four parameters. Purple, red, blue, and green represent cells without any treatment, cells adding with 5-Fu, DDP, Adr, respectively.
3.4 IC50 value fitting and verification results calculated from quantitative phase imaging
First-order polynomial, power function, and exponential function were selected to fit the curves of the four parameters ${S_{Cell}}$, ${h_{ave}}$, $Clu$ and $Ent$ changing with time. For the same parameter, the sum of the root mean square error (RMSE) of our experimental data groups’ fitting results was calculated. The function corresponding to the smallest sum of the RMSE was taken as the final fitting function of this parameter. The mean of coefficients which are independent with t in the fitting function expressions of the three groups of experimental data under each parameter was calculated as constant values in the final IC50 fitting results. The results of fitting and calculation of related parameters are shown in Fig. 7 and Table 1.
Fig. 7. The curve fitting results of (a)cell projected area; (b)average height; (c)cluster shade; (d)entropy. Dotted lines represent fitted function curves and dots represent original data points.
Table 1. The fitting function and the related coefficient
Considering the influence of each parameter on the final fitting result and the magnitude of the coefficient in each function expression, different weights are given to different items, and the compound parameter $A$ is defined by Eq. (14) to comprehensively describe the influence of each parameter on the IC50 value.
Three groups of data points from 30min to 120min were selected and substituted into Eq. (14) for calculation, and the average $\overline {{A_{DDP}}} $, $\overline {{A_{5 - Fu}}} $, $\overline {{A_{Adr}}} $ were calculated, respectively. The three averages were used as independent variables, and the IC50 values corresponding to the three different drugs DDP,5-Fu, and Adr were used as dependent variables, parabolic fitting was performed to obtain Eq. (15).
Combined with Eq. (14) and Eq. (15), given a certain time t (t > 30min) of a dish of cells and the average of each cell's four parameters corresponding to this time, the IC50 value can be calculated, thereby judging the type of drugs added in this group. The fitting result is shown in Fig. 8.
Fig. 8. (a)IC50 fitting result. The dots are the calculation results after substituting the data points at each moment into Eq. (14) and Eq. (15). The dotted line is the experimental measurement result of IC50 value (regarded as the standard value), and different colors represent different drugs. (b) The average and standard deviation of the fitting results and their comparison to standard values.
In order to prove the accuracy of the fitting results, the experiment was repeated three times, and the IC50 value was calculated according to the same steps. The verification results are shown in Fig. 9.
Fig. 9. Verification results of three replicate experiments for the proposed method. ((a), (b), and (c) are the verification results of three repeated experiments, respectively, and different colors represent different drugs. The standard bar represents the IC50 value determined by absorbance-based kit, and the calculation bar is the IC50 value calculated by the proposed method.)
As shown in Fig. 9, the calculation result is close to the standard value. Therefore, the IC50 value can be calculated by the proposed method to judge the type of the added drug.
4. Discussion
Cisplatin, Adriamycin, and 5-fluorouracil are three drugs commonly used in chemotherapy of ovarian cancer. The related study found that changes in the genome of cells [36] and transcriptional dysregulation caused by aberrant transcription factors [37] can lead to differences in the response of cells to different drugs. Therefore, evaluating the sensitivity of ovarian cancer cells to these three drugs is significant for clinical guidance of medication and implementation of individualized chemotherapy. Recently, there have been several methods for drug potency assessment by in vitro drug sensitivity test, and quantitative phase imaging is an extension of the existing in vitro cytotoxicity detection methods, which have been used in the evaluation of the cytotoxicity of medical organic nanoparticles [38], detection of the effects of different bacteria on T lymphocytes based on cell morphological changes [39], and tracking of cell death with drugs [40]. Digital holographic microscopy has the characteristics of non-invasive, no label, no need to destroy the biological activity of the tested sample, and the sample preparation is convenient. At the same time, when used for drug potency assessment, morphological parameters and texture parameters of cells can be quantitatively calculated from phase images, the changing trend of cells under different drugs can be analyzed from the time-scale, and accurate evaluation results can be obtained in a short time. However, this technique also has some limitations. For example, some slow-acting drugs may not have enough cytotoxicity to be detected within a limited time.
In this research, morphological parameters ${S_{Cell}}$, ${h_{ave}}$ and texture parameters$Clu$, $Ent$ of A2780 cells at each time after the addition of different drugs were quantitatively calculated by the phase distribution of every single cell, and the curves of each parameter changing with time were plotted to quantitatively compare the sensitivity of cells to the three drugs. At the same time, these four parameters are linked with IC50 value, which can directly reflect the drug potency through the numerical fitting. It can reflect the difference in cell sensitivity to different drugs to realize the preliminary assessment of drug potency in the process of cancer chemotherapy.
Specifically, under the action of drugs, cells will be induced to apoptosis, and the specific mechanism remains to be studied, which may be achieved by blocking the cell cycle [41] or stimulating cells to produce reactive oxygen species (ROS) [42]. During the process of cell apoptosis, the cells shrink and become round and gradually separate from the culture medium. Some previous research on real-time apoptosis detection by DHM found that these morphological changes of apoptosis were shown in quantitative phase images as increased cell height, decreased projection area, and significantly decreased total volume of cells over time [43,44]. In addition, studies have shown that during the drug-induced cell apoptosis process, apoptotic bodies will be formed, accompanied by the reorganization of the actin filament structure and the fragmentation of the nucleus [45]. These changes will lead to an increase in cell asymmetry and inhomogeneity, which is manifested in texture feature parameters’ change. To be specific, the cluster shade increases, and the entropy decreases. These parameters’ change trend is consistent with our experimental results. The drug's effectiveness can be reflected by the speed of drug-induced apoptosis, which is reflected in the rate of parameter change with time. IC50 is a parameter directly used to measure drug cytotoxicity. Therefore, calculating the IC50 value by fitting the rate of morphological and texture characteristic parameters’ change with time, which can evaluate the sensitivity of cells to different drugs and realize drug potency assessment, is feasible and effective.
In the process of calculating IC50 by morphological parameters and texture characteristic parameters, in the beginning, the characteristic parameters did not change significantly because it took a certain amount of time for cells to absorb drugs, and the reaction between drugs and cells might not have started. However, if the dosed time is too long, the apoptosis process of cells may have ended, and there will be no noticeable changes in morphology or texture. Morphological and texture parameters after adding different drugs tend to be constant, so it is impossible to assess drug potency by calculating IC50 of parameters’ change rate. Therefore, it is reasonable to select data points between 30min and 120min after adding drugs to calculate IC50 value.
With a view to clinical applications, due to individual differences in drug resistance of cancer cells, it is necessary to judge the degree of drug resistance of each patient to various drugs. The cancer cells can be isolated from the patients and proliferated by subculture, and then by adding different drugs, the IC50 value could be calculated through the change of morphological and texture feature parameters. It could judge the sensitivity of the cancer cells of different medicine and assess the drug potency initially, thus providing guidance for clinical drug choice and laying the foundation for the implementation of personalized diagnosis and treatment.
5. Conclusion
In this study, the phase images of ovarian cancer cells A2780 were obtained by DHM, and the morphological parameters${S_{Cell}}$, ${h_{ave}}$ and texture feature parameters$Clu$, $Ent$ of ovarian cancer cells were measured each minute from 0 to 120 minutes after adding DDP, Adr, and 5-Fu. The change of the parameters with the time of each group added with different drugs has the same trend, which shows that the average height increases, the projected area decreases, the cluster shade increases, and the entropy decreases. The rate of change is related to the type of drug added. The cells added with Adr changed the fastest, added with DDP was the second, and added with 5-Fu was the slowest, which was consistent with the IC50 value. The IC50 value can be calculated from these four parameters at any time within 30∼120min, which directly reflects the potency of the drug. And it is verified by repeated experiments, which proves the effectiveness of the proposed method. Therefore, this study proposed a method to determine IC50 value from the quantitative phase image, which can evaluate the sensitivity of cancer cells to different drugs by directly fitting the IC50 value through morphological and texture parameters. It may help to establish a new method for drug potency assessment of ovarian cancer cells based on quantitative phase imaging, which has certain guiding significance for the selection of clinical chemotherapy drugs.
Funding
Capital’s Funds for Health Improvements and Research (2020-2Z4088).
Acknowledgments
The authors thank the funding support of Beijing Municipal Health Commission for the research related to this article.
Disclosures
The authors declare that there are no conflicts of interest related to this article.
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. C. Stewart, C. Ralyea, and S. Lockwood, “Ovarian cancer: an integrated review,” Semin. Oncol. Nurs. 35(2), 151–156 (2019). [CrossRef]
2. V. Tarhriz, M. Bandehpour, S. Dastmalchi, E. Ouladsahebmadarek, H. Zarredar, and S. Eyvazi, “Overview of CD24 as a new molecular marker in ovarian cancer,” J. Cell. Physiol. 234(3), 2134–2142 (2019). [CrossRef]
3. L. J. Havrilesky, G. Broadwater, D. M. Davis, K. C. Nolte, J. C. Barnett, E. R. Myers, and S. Kulasingam, “Determination of quality of life-related utilities for health states relevant to ovarian cancer diagnosis and treatment,” Gynecol. Oncol. 113(2), 216–220 (2009). [CrossRef]
4. V. Rojas, K. M. Hirshfield, S. Ganesan, and L. Rodriguez-Rodriguez, “Molecular characterization of epithelial ovarian cancer: Implications for diagnosis and treatment,” Int. J. Mol. Sci. 17(12), 2113 (2016). [CrossRef]
5. M. Barani, M. Bilal, F. Sabir, A. Rahdar, and G. Z. Kyzas, “Nanotechnology in ovarian cancer: diagnosis and treatment,” Life Sci. 266, 118914 (2021). [CrossRef]
6. P. E. Colombo, M. Fabbro, C. Theillet, F. Bibeau, P. Rouanet, and I. Ray-Coquard, “Sensitivity and resistance to treatment in the primary management of epithelial ovarian cancer,” Crit. Rev. Oncol. Hematol. 89(2), 207–216 (2014). [CrossRef]
7. A. Chandra, C. Pius, M. Nabeel, M. Nair, J. K. Vishwanatha, S. Ahmad, and R. Basha, “Ovarian cancer: Current status and strategies for improving therapeutic outcomes,” Cancer Med. 8(16), 7018–7031 (2019). [CrossRef]
8. J. A. Ledermann, F. A. Raja, C. Fotopoulou, A. Gonzalez-Martin, N. Colombo, and C. Sessa, “Newly diagnosed and relapsed epithelial ovarian carcinoma: ESMO clinical practice guidelines for diagnosis, treatment and follow-up,” Ann. Oncol. 24, vi24–vi32 (2013). [CrossRef]
9. A. N. Hata, J. A. Engelman, and A. C. Faber, “The BCL2 family: Key mediators of the apoptotic response to targeted anticancer therapeutics,” Cancer Discov. 5(5), 475–487 (2015). [CrossRef]
10. A. Komoriya, B. Z. Packard, M. J. Brown, M. L. Wu, and P. A. Henkart, “Assessment of caspase activities in intact apoptotic thymocytes using cell-permeable fluorogenic caspase substrates,” J. Exp. Med. 191(11), 1819–1828 (2000). [CrossRef]
11. V. M. C. Quent, D. Loessner, T. Friis, J. C. Reichert, and D. W. Hutmacher, “Discrepancies between metabolic activity and DNA content as tool to assess cell proliferation in cancer research,” J. Cell. Mol. Med. 14(4), 1003–1013 (2010). [CrossRef]
12. J. R. Doherty and J. L. Cleveland, “Targeting lactate metabolism for cancer therapeutics,” J. Clin. Invest. 123(9), 3685–3692 (2013). [CrossRef]
13. B. Wu, J. S. Zhu, Y. Zhang, W. M. Shen, and Q. Zhang, “Predictive value of MTT assay as an in vitro chemosensitivity testing for gastric cancer: One institution’s experience,” World J. Gastroenterol. 14(19), 3064–3068 (2008). [CrossRef]
14. A. Eatemadi, M. Darabi, L. Afraidooni, N. Zarghami, H. Daraee, L. Eskandari, H. Mellatyar, and A. Akbarzadeh, “Comparison, synthesis and evaluation of anticancer drug-loaded polymeric nanoparticles on breast cancer cell lines,” Artif. Cells Nanomedicine Biotechnol. 44, 1–10 (2015). [CrossRef]
15. G. Riddick, H. Song, S. Ahn, J. Walling, D. Borges-Rivera, W. Zhang, and H. A. Fine, “Predicting in vitro drug sensitivity using random forests,” Bioinformatics 27(2), 220–224 (2011). [CrossRef]
16. M. P. Menden, F. Iorio, M. Garnett, U. McDermott, C. H. Benes, P. J. Ballester, and J. Saez-Rodriguez, “Machine learning prediction of cancer cell sensitivity to drugs based on genomic and chemical properties,” PLoS One 8(4), e61318 (2013). [CrossRef]
17. J. C. Costello, L. M. Heiser, E. Georgii, M. Gönen, M. P. Menden, N. J. Wang, M. Bansal, M. Ammad-Ud-Din, P. Hintsanen, S. A. Khan, J. P. Mpindi, O. Kallioniemi, A. Honkela, T. Aittokallio, K. Wennerberg, J. J. Collins, D. Gallahan, D. Singer, J. Saez-Rodriguez, S. Kaski, J. W. Gray, and G. Stolovitzky, “A community effort to assess and improve drug sensitivity prediction algorithms,” Nat. Biotechnol. 32(12), 1202–1212 (2014). [CrossRef]
18. L. Wang, X. Li, L. Zhang, and Q. Gao, “Improved anticancer drug response prediction in cell lines using matrix factorization with similarity regularization,” BMC Cancer 17(1), 1–12 (2017). [CrossRef]
19. M. Joo, A. Park, K. Kim, W. J. Son, H. S. Lee, G. Lim, J. Lee, D. H. Lee, J. Ahn, J. H. Kim, T. Ahn, and S. Nam, “A deep learning model for cell growth inhibition IC50 prediction and its application for gastric cancer patients,” Int. J. Mol. Sci. 20(24), 6276 (2019). [CrossRef]
20. L. Amable, “Cisplatin resistance and opportunities for precision medicine,” Pharmacol. Res. 106, 27–36 (2016). [CrossRef]
21. F. S. Liu, “Mechanisms of chemotherapeutic drug resistance in cancer therapy-a quick review,” Taiwan. J. Obstet. Gynecol. 48(3), 239–244 (2009). [CrossRef]
22. A. A. Zhikhoreva, A. V. Belashov, V. G. Bespalov, A. L. Semenov, I. V. Semenova, G. V. Tochilnikov, N. T. Zhilinskaya, and O. S. Vasyutinskii, “Morphological changes in the ovarian carcinoma cells of Wistar rats induced by chemotherapy with cisplatin and dioxadet,” Biomed. Opt. Express 9(11), 5817 (2018). [CrossRef]
23. J. Lobo, E. Y. S. See, M. Biggs, and A. Pandit, “An insight into morphometric descriptors of cell shape that pertain to regenerative medicine,” J. Tissue Eng. Regen. Med. 10(7), 539–553 (2016). [CrossRef]
24. A. A. Gamal El Din, M. A. Badawi, S. E. Abdel Aal, N. A. Ibrahim, F. A. Morsy, and N. M. Shaffie, “DNA cytometry and nuclear morphometry in ovarian benign, borderline and malignant tumors,” Open Access Maced J Med Sci 3, 537–544 (2015). [CrossRef]
25. M. Ravanelli, D. Farina, M. Morassi, E. Roca, G. Cavalleri, G. Tassi, and R. Maroldi, “Texture analysis of advanced non-small cell lung cancer (NSCLC) on contrast-enhanced computed tomography: prediction of the response to the first-line chemotherapy,” Eur. Radiol. 23(12), 3450–3455 (2013). [CrossRef]
26. A. Ahmed, P. Gibbs, M. Pickles, and L. Turnbull, “Texture analysis in assessment and prediction of chemotherapy response in breast cancer,” J. Magn. Reson. Imaging 38(1), 89–101 (2013). [CrossRef]
27. N. Benzerdjeb, C. Garbar, P. Camparo, and H. Sevestre, “Digital holographic microscopy as screening tool for cervical cancer preliminary study,” Cancer Cytopathol. 124(8), 573–580 (2016). [CrossRef]
28. B. Kemper and G. Von Bally, “Digital holographic microscopy for live cell applications and technical inspection,” Appl. Opt. 47(4), A52–A61 (2008). [CrossRef]
29. D. Bettenworth, A. Bokemeyer, C. Poremba, N. S. Ding, S. Ketelhut, P. Lenz, and B. Kemper, “Quantitative phase microscopy for evaluation of intestinal inflammation and wound healing utilizing label-free biophysical markers,” Histol. Histopathol. 33, 417–432 (2018). [CrossRef]
30. T. Cacace, V. Bianco, and P. Ferraro, “Quantitative phase imaging trends in biomedical applications,” Opt. Lasers Eng. 135, 106188 (2020). [CrossRef]
31. B. Kemper, D. Carl, J. Schnekenburger, I. Bredebusch, M. Schäfer, W. Domschke, and G. von Bally, “Investigation of living pancreas tumor cells by digital holographic microscopy,” J. Biomed. Opt. 11(3), 034005 (2006). [CrossRef]
32. N. Pavillon, J. Kühn, C. Moratal, P. Jourdain, C. Depeursinge, P. J. Magistretti, and P. Marquet, “Early cell death detection with digital holographic microscopy,” PLoS One 7(1), e30912 (2012). [CrossRef]
33. M. Mir, Z. Wang, Z. Shen, M. Bednarz, R. Bashir, I. Golding, S. G. Prasanth, and G. Popescu, “Optical measurement of cycle-dependent cell growth,” Proc. Natl. Acad. Sci. U. S. A. 108(32), 13124–13129 (2011). [CrossRef]
34. Z. He, X. Sui, G. Jin, and L. Cao, “Distortion-correction method based on angular spectrum algorithm for holographic display,” IEEE Trans. Ind. Inf. 15(11), 6162–6169 (2019). [CrossRef]
35. X. Yang, S. Tridandapani, J. J. Beitler, D. S. Yu, E. J. Yoshida, W. J. Curran, and T. Liu, “Ultrasound GLCM texture analysis of radiation-induced parotid-gland injury in head-and-neck cancer radiotherapy: an in vivo study of late toxicity,” Med. Phys. 39(9), 5732–5739 (2012). [CrossRef]
36. W. Yang, J. Soares, P. Greninger, E. J. Edelman, H. Lightfoot, S. Forbes, N. Bindal, D. Beare, J. A. Smith, I. R. Thompson, S. Ramaswamy, P. A. Futreal, D. A. Haber, M. R. Stratton, C. Benes, U. McDermott, and M. J. Garnett, “Genomics of Drug Sensitivity in Cancer (GDSC): A resource for therapeutic biomarker discovery in cancer cells,” Nucleic Acids Res. 41(D1), D955–D961 (2012). [CrossRef]
37. L. Garcia-Alonso, F. Iorio, A. Matchan, N. Fonseca, P. Jaaks, G. Peat, M. Pignatelli, F. Falcone, C. H. Benes, I. Dunham, G. Bignell, S. S. McDade, M. J. Garnett, and J. Saez-Rodriguez, “Transcription factor activities enhance markers of drug sensitivity in cancer,” Cancer Res. 78(3), 769–780 (2018). [CrossRef]
38. K. M. Eder, A. Marzi, Á Barroso, S. Ketelhut, B. Kemper, and J. Schnekenburger, “Label-free digital holographic microscopy for in vitro cytotoxic effect quantification of organic nanoparticles,” Cells 11(4), 644 (2022). [CrossRef]
39. K. L. Vom Werth, T. Wörmann, B. Kemper, P. Kümpers, S. Kampmeier, and A. Mellmann, “Investigating morphological changes of T-lymphocytes after exposure with bacterial determinants for early detection of septic conditions,” Microorganisms 10(2), 391 (2022). [CrossRef]
40. T. Vicar, M. Raudenska, J. Gumulec, and J. Balvan, “The quantitative-phase dynamics of apoptosis and lytic cell death,” Sci. Rep. 10(1), 1566 (2020). [CrossRef]
41. X. Zhao, J. Fan, P. Wu, C. Wei, Q. Chen, Z. Ming, J. Yan, and L. Yang, “Chronic chemotherapy with paclitaxel nanoparticles induced apoptosis in lung cancer in vitro and in vivo,” Int. J. Nanomedicine 14, 1299–1309 (2019). [CrossRef]
42. P. Liu, H. Yang, W. Chen, J. Zhao, and D. Li, “Silver nanotriangles and chemotherapy drugs synergistically induce apoptosis in breast cancer cells via production of reactive oxygen species,” J. Nanoparticle Res. 21(1), 1–11 (2019). [CrossRef]
43. M. Mugnano, P. Memmolo, L. Miccio, S. Grilli, F. Merola, A. Calabuig, A. Bramanti, E. Mazzon, and P. Ferraro, “In vitro cytotoxicity evaluation of cadmium by label-free holographic microscopy,” J. Biophotonics 11(12), e201800099 (2018). [CrossRef]
44. A. Khmaladze, R. L. Matz, T. Epstein, J. Jasensky, M. M. Banaszak Holl, and Z. Chen, “Cell volume changes during apoptosis monitored in real time using digital holographic microscopy,” J. Struct. Biol. 178(3), 270–278 (2012). [CrossRef]
45. C. Díaz-Ruiz, B. Montaner, and R. Pérez-Tomás, “Prodigiosin induces cell death and morphological changes indicative of apoptosis in gastric cancer cell line HGT-1,” Histol. Histopathol. 16, 415–421 (2001). [CrossRef]
