1. Introduction
The plasma membrane is a dynamic, complexly organized cell structure that separates the internal contents of the cell from the external environment. In cancer cells, the membrane is the primary barrier for chemotherapeutic drugs, so its biophysical properties and biochemical composition are important factors that determine efficient drug uptake [
1]. On the other hand, chemotherapeutic drugs themselves can alter the microviscosity of the plasma membrane through direct interaction with the lipid bilayer or indirectly through lipid peroxidation [
2,
3,
4]. In addition, fluctuations in membrane microviscosity and changes in lipid profile accompany the response of cancer cells to cytotoxic chemotherapy, including the development of cell death and proliferation arrest. Meanwhile, the effects of chemotherapeutic agents on plasma membranes remain poorly investigated.
Taxol, or paclitaxel, is a chemotherapeutic agent of a plant alkaloids class, which has proven antineoplastic activity against a variety of cancers, including breast, brain, head and neck, lung, colon, cervical, and ovarian tumors [
5]. Its major mechanism of action is the stabilization of microtubules via polymerization of tubulin and preventing depolymerization, leading to mitotic arrest and cell death. However, biophysical and biochemical processes that occur before and after the binding of paclitaxel to microtubules are not well understood. Recent studies suggest that the plasma membrane is also the target of taxane drugs. Both membrane composition and fluidity can influence penetration into cells and, consequently, the biological action of paclitaxel. This paclitaxel–lipid interplay is largely determined by phospholipid properties. The higher degree of acyl chain unsaturation, smaller headgroup size, and longer acyl chain length increase paclitaxel incorporation into the lipid membrane [
6]. Paclitaxel, due to its lipophilic nature, actively interacts with the cell membrane, and this interaction may contribute to the drug retention and development of chemoresistance [
7]. At the same time, taxanes, including paclitaxel, can induce serious physicochemical alterations in membranes, specifically changes in membrane fluidity, conformation of receptors and enzymes, lipid packing density, etc. It is known that paclitaxel causes membrane liquefaction and, conversely, rehydration in a dose-dependent manner [
6]. In general, data on the effect of paclitaxel on membrane microviscosity and lipids are scarce and contradictory, and these studies were mainly performed on model membranes.
The aim of the present work was to study microviscosity and lipid changes in the plasma membranes of tumor cells during chemotherapy with paclitaxel.
Membrane microviscosity of live cells in cell monolayers and multicellular spheroids was measured using a two-photon fluorescence lifetime imaging microscopy (FLIM) with the fluorescent viscosity-sensitive dye BODIPY 2. FLIM of BODIPY-based fluorescent molecular rotors is an established technique for measuring viscosity at the microscopic level that allows both quantitative imaging of viscosity with high spatial resolution and dynamic measurements of viscosity in a living cell with high temporal resolution [
8,
9]. The fluorescence characteristics of molecular rotors, such as fluorescence intensity and lifetime, are strictly dependent on the viscosity of their immediate microenvironment. Of them, the lifetime-based readout seems more appropriate since it does not depend on the concentration of the rotor and its photobleaching, the intensity of the exciting light, and the configuration of the microscope. Thus, the fluorescence lifetime of a rotor can be directly converted to the viscosity of its medium using pre-recorded calibration curves. The methodologies for imaging of membrane microviscosity in cultured cancer cells, tumor spheroids, and animal tumors in vivo have been previously developed by our group and applied to follow the changes over the course of therapy with platinum drugs and 5-fluorouracil [
10,
11,
12].
In parallel, the membrane lipid composition was identified using time-of-flight secondary ion mass spectrometry (ToF-SIMS) with a focus on phosphatidylcholine, sphingomyelin, cholesterol, and unsaturated fatty acids content, the key components responsible for the regulation of viscosity. ToF-SIMS enables the detection and visualization of organic compounds in cells with submicron spatial resolution. High chemical specificity and sensitivity make ToF-SIMS a valuable tool for the analysis of lipids in the cell plasma membrane. Our previous studies have revealed that changes in lipid profile underlay viscosity changes in cell membranes induced by chemotherapeutic agents [
10,
11,
12].
3. Discussion
It is known that the plasma membrane plays a critical role in the response of tumor cells to chemotherapy, but the relationships between membrane lipid composition, its biophysical properties, and drug response are poorly explored. In the present study, we have identified the effects of antimicrotubule agent paclitaxel on the plasma membrane of cancer cells in cell monolayer and tumor spheroids using a combination of microviscosity imaging by FLIM and membrane lipid profile analysis by ToF-SIMS.
The cytotoxic effect of many chemotherapy agents is, at least partly, associated with their contribution to the molecular organization of membranes, specifically with disrupting the organization of lipids. The plasma membrane is one of the most important targets, in addition to microtubules, for taxane-based anticancer drugs, including paclitaxel. These drugs can modify the physicochemical properties of membranes, such as fluidity (the reciprocal of viscosity) [
13], the conformation of membrane-bound enzymes and receptors [
14], lipid packing density [
15], and lipid–lipid or lipid–protein interactions [
16].
The relationship between membrane viscosity and the response of tumor cells to paclitaxel has been especially poorly studied. The viscosity of membranes plays an important role in the rate of diffusion and absorption of the drug into the cell. There are several studies pointing to the direct effect of paclitaxel on the viscosity of membranes through direct interaction with lipids. For example, using the methods of differential scanning calorimetry and electron paramagnetic resonance spectroscopy, a decrease in membrane viscosity was found in model membranes when paclitaxel was incorporated [
17]. Zhao et al. also showed an increase in membrane fluidity with paclitaxel on model membranes [
18].
Some other studies suggest the direct interaction of paclitaxel with membrane lipids, which can cause their liquefaction. For example, Kang and Loverde report on the interactions between hydrophobic paclitaxel and a model cell membrane at the molecular level using molecular dynamics simulations of all paclitaxel atoms. It has been shown that the main taxane ring is localized in the outer hydrophobic zone, while the three phenyl rings prefer to be located closer to the hydrophobic core of the membrane [
19]. Süleymanogluit et al. found that paclitaxel liquefies the upper part of the acyl chains as it binds to the surface of phospholipid acyl chains and increases fluidity in the region of their head group, which also indicates a decrease in the viscosity of the lipid layer of cell membranes [
20]. Moreover, paclitaxel has been shown to form temporary pores in the membrane, which also favor its fluidification. These facts can explain the liquefaction of tumor cell membranes when exposed to paclitaxel.
Another possible reason for the lowering of the microviscosity of membranes in paclitaxel-treated cells is alterations in the cellular cytoskeleton. The main mechanism of action of paclitaxel is associated with the ability of the drug to stimulate the assembly of microtubules from dimeric tubulin molecules and, therefore, to stabilize their structure, which disrupts the mitotic function of cells. There are many studies that show changes in microtubules at doses of paclitaxel close to those used in our study [
21,
22,
23]. With regard to resistant cells, it was demonstrated that they had less polymerized tubulin and an increased growth rate of microtubules [
24].
The association of tubulin with the plasma membrane involves several levels of penetration into the bilayer: from integral membrane proteins which bind on the surface to microtubules attached by linker proteins to proteins in the membrane [
25]. In agreement with our study using paclitaxel, Aszalos et al. showed that depolymerization of microtubules using colcemid, colchicine, vinblastine, podophyllotoxin, and griseofulvin resulted in a decrease in the viscosity of the plasma membranes of CHO cells [
26]. Rémy-Kristensen et al. investigated the effects of microtubule network integrity on plasma membrane fluidity in L929 mouse fibroblasts [
27]. It was shown that in the cells treated with drugs that depolymerized microtubules, membrane fluidity increased, as assessed by fluorescence depolarization using a TMA-DPH probe. A slight decrease in microviscosity under the action of colchicine was also demonstrated on isolated membranes by Berlin et al. [
28]. These studies suggest that microtubule integrity contributes to the high lipid order of the plasma membrane, but to a lesser extent than other factors such as lipid composition and cholesterol content.
It is known that paclitaxel affects not only tubulin but also the actin cytoskeleton [
23]. There are data indicating the relationship between the submembrane actin cytoskeleton and the biophysical parameters of the membrane, including its microviscosity [
27,
29]. For example, the movements of integral membrane proteins are more limited than those of lipids, mainly due to interactions with the cytoskeleton through their intracellular domains [
29]. So, they anchor in the membrane and serve as stabilizers of its fluid properties. On the other hand, the cytoskeleton can form barriers directly under the plasma membrane, and the organization of the cytoskeleton under the membrane limits the area of the membrane where molecules, including anticancer agents, can diffuse. An attempt to evaluate the cytosol viscosity under the action of paclitaxel and to correlate it with changes in the cytoskeleton structure was made by Chen et al. Using the method of optical tweezers, it was shown that the viscosity of the cytoplasm of cancer cells treated with paclitaxel dramatically decreases in the first 3 h due to the destruction of actin filaments [
30]. In general, the effect of submembrane cytoskeleton organization on membrane microviscosity is not fully understood, but it can not be ruled out that alterations of actin structure also contributed to the fluidization of the membrane upon treatment with paclitaxel. Analysis of the cytoskeleton rearrangements in paclitaxel-treated cancer cells was beyond the scope of this study.
In our study, we noticed a trend towards a decrease in the microviscosity of tumor cell membranes in the first hours of incubation with paclitaxel and a dramatic drop in microviscosity after 24 h. In general, our results are in good agreement with the literature on membrane fluidification under the action of paclitaxel.
Fluidification of the plasma membrane is involved in the induction of apoptosis [
31,
32]. In our study, the recorded decrease in membrane viscosity is consistent with the presence of extensive areas of cell death, including late stages of apoptosis, detected by PI staining.
The viscous properties of cell membranes largely depend on the qualitative and quantitative composition of lipids in these membranes. Depending on the lipid profile, the effectiveness of paclitaxel may vary. Specifically, the addition of cholesterol decreases the penetration of paclitaxel due to a decrease in diffusion, which was demonstrated on the model membranes [
6,
33]. In a study by Pereira et al., it was also shown that the effects of paclitaxel on Langmuir monolayers strongly depend on the relative concentration of cholesterol in the membrane [
33]. To evaluate the role of phospholipid properties, Zhao et al. examined the penetration profiles of paclitaxel in model membranes of different compositions [
34]. Lipid chain length has been shown to affect drug–membrane interactions, and while shorter chain length phospholipids have shown a higher ability to incorporate paclitaxel, longer chains have hindered the penetration process due to higher packaging hence acting as a physical barrier. In addition, saturated phospholipids may pack tightly into each other, resulting in less incorporation of paclitaxel.
In our study, it was shown, for the first time, that paclitaxel treatment changes the composition of membrane lipids. Mass spectrometric analysis revealed a statistically significant increase in the signal of phosphatidylcholine in the first hour of exposure to paclitaxel. Xia Li and Ying-Jin Yuan showed, using the NPLC-ESI/MS(n) procedure, that changes in some phosphatidylcholine species with unsaturated fatty acid chains and phosphatidic acid species with saturated fatty acid chains are closely related to the change in cell membrane fluidity that occurs during apoptosis [
35]. In addition, we noticed an increase in the content of mono- and polyunsaturated fatty acids after treatment with paclitaxel, which correlated with decreased microviscosity. Menendez et al. have shown that polyunsaturated fatty acids (PUFA) are able to enhance the sensitivity of cancer cells to paclitaxel by enhancing peroxidative processes and regulating the expression of oncoproteins [
36]. The most potent PUFA that increased the toxicity of paclitaxel in breast cancer cell lines was gamma-linolenic acid. The most common event characterizing multidrug-resistant cells is the overexpression of ABC transporters such as P-glycoprotein (Pgp). In several studies, PUFAs are described as direct inhibitors of Pgp, and an increase in their concentration can suppress the activity of Pgp. For example, ω-3 and ω-6 PUFAs reduced Pgp transcription in colorectal cancer cells [
37]. Thus, an increase in the content of unsaturated fatty acids may be a part of the response of the tumor cell to drug stress.
Previous studies have revealed that different drugs induce different alterations to cell membranes. For example, platinum drugs (cisplatin, oxaliplatin) initially caused a decrease (minutes to 1 h) and then an increase (1–2 days) in the microviscosity of plasma membranes, the latter mainly associated with an increase in cholesterol content [
10,
11] but not with a partitioning of the drug into the cell membrane. The drug 5-fluorouracil, a cytotoxic agent of the anti-metabolite class, induced only short-term fluctuations in membrane viscosity immediately after the addition of the drug, whereas the development of resistance to 5-fluorouracil resulted in a steady increase in viscosity associated with an increase in the content of sphingomyelin and cholesterol [
12]. Unlike platinum drugs and 5-fluorouracil, paclitaxel has a different intracellular target, which causes completely different changes to membrane microviscosity.
Based on our results with platinum drugs, 5-fluorouracil, and paclitaxel, we can suggest that the microviscosity of the cell membrane is regulated not by one type of lipid but by a complex interplay between different lipids that constitute the membrane. Notably, among the observed alterations in lipid composition induced by these drugs, there were always those that correlated with the viscosity changes.
The plasma membrane is a highly heterogeneous structure that includes the bulk lipid bilayer and “lipid rafts.” It is known that the bulk liquid-phase plasma membrane contains less cholesterol and sphingomyelin and more phospholipids with unsaturated acyl chains compared to the lipid rafts. The lipid rafts are the cholesterol-rich, highly ordered lipid “islands” that act as organizing hubs for membrane-embedded proteins. A limitation of our study is that the ToF-SIMS technique measures the signal from the entire cell surface, including lipid rafts, while microviscosity is measured only in the hydrophobic lipid bilayer of the membrane, where the BODIPY 2 rotor is localized. We suppose that this could be a reason for discrepancies between microviscosity and lipids (especially cholesterol and sphingomyelin) changes in the treated cells.
The precise mechanisms of membrane fluidization upon paclitaxel treatment in cellular models have yet to be clarified.
4. Materials and Methods
4.1. Cell Culture
HeLa–Kyoto (human cervical cancer) cells were used. The cells were cultured in DMEM (Life Technologies, Carlsbad, CA, USA) containing 100 μg/mL penicillin, 100 μg/mL streptomycin sulfate, and 10% fetal bovine serum (FBS) at 37 °C in a humidified atmosphere with 5% CO2.
To generate tumor spheroids, HeLa–Kyoto cells were seeded into ultra-low attachment 96-well round bottom plates (Corning Inc., Corning, NY, USA), ~300 cells/well in 200 μL DMEM and cultured in standard conditions (37 °C, 5% CO2, 80% humidity). After 5 days, spheroid formation was verified using light microscopy.
4.2. Chemotherapy
The chemotherapeutic drug Paclitaxel (Bristol–Myers Squibb, Irving, TX, USA) was used at doses of 1.6 nM (IC50/2), 3.2 nM (IC50), and 6.4 nM (2× IC50) for monolayer HeLa–Kyoto cells. The IC50 concentration was determined in our previous experiments using the MTT assay. Spheroids were treated with 6.5 nM of paclitaxel. The cells were incubated with the drug from 10 min to 48 h, spheroids from 3 to 48 h, and microviscosity was measured immediately after the treatment. Untreated cells or spheroids were used as a control.
The protocol for the creation of a paclitaxel-adapted cell subline was adopted from Ref. [
38]. Briefly, a HeLa cell culture was continuously exposed to gradually increasing concentrations of paclitaxel. The initial concentration was 1/150 IC50. Each successive concentration was increased by 25% from the previous one and added only clear adaptation of the cells to the drug, i.e., after restarting cell proliferation without significant cell death in a plate (after 2–7 days). In ~4 months from first exposure, the cells were considered paclitaxel-resistant. Measurements of microviscosity in these cells were performed in 48 h after washing out the drug to avoid the immediate effects of paclitaxel.
4.3. Cell Viability Assay
A live/dead double staining kit (Sigma)—Calcein AM and propidium iodide (PI)—was used to stain live and dead HeLa–Kyoto cells, respectively, after chemotherapy according to the manufacturer’s protocol [
39]. Cells were stained after 24 h, spheroids after 3, 6, 24, and 48 h, and the percentage of dead cells (stained with PI) of the total number of cells was calculated. The fluorescence of calcein was excited using an argon laser at a wavelength of 488 nm, and the emission was measured in the range of 500–570 nm. PI fluorescence was excited at a wavelength of 543 nm, and the emission was measured in the range of 600–700 nm. One-photon fluorescence confocal images were obtained using an LSM 880 (Carl Zeiss, Göttingen, Germany) laser scanning microscope.
4.4. Fluorescent Molecular Rotor BODIPY 2 and FLIM Microscopy
BODIPY 2 (4,4-difluoro-4-bora-3a,4a-diaza-s-indacene) was used as a viscosity-sensitive probe [
8,
9]. For microscopic imaging, the cells were seeded on glass-bottom dishes for confocal microscopy (FluoroDishes, Life Technologies, Carlsbad, CA, USA) in complete DMEM media without phenol red (Life Technologies, USA). The membrane viscosity was examined at 10 min, 1, 3, 6, and 24 h after adding Paclitaxel. Before imaging, the culture media was replaced with ice-cold Hank’s solution without Ca
2+/Mg
2+, and cells were incubated at +4 °C for 3 min. Thereafter, Hank’s solution was replaced with an ice-cold BODIPY 2 solution (4.5 μM). The spheroids were stained with BODIPY 2 on day 5 after the growth phase when they had a compact heterogeneous structure and a size of ~330 μM. A total of 4–6 spheroids were carefully transferred to each glass-bottom dish in 1.5 mL DMEM without phenol red and placed in a CO
2 incubator for 1 h to allow their attachment. The membrane viscosity was examined at 3, 6, 24, and 48 h after adding paclitaxel. Before imaging, the culture media with paclitaxel was replaced with ice-cold Hank’s solution without Ca
2+/Mg
2+, and cells were incubated at +4 °C for 10 min. Afterward, Hank’s solution was replaced with an ice-cold BODIPY solution (8.9 μM). Two-photon excited FLIM images were acquired within 5–10 min after adding BODIPY2, while the probe was located in the plasma membranes.
A multiphoton tomography MPTflex (JenLab, Jena, Germany) with a tuneable 80 MHz, 200 fs Ti:Sapphire laser (MaiTai), a single-photon counting module SPC-150 and detector PMC-100-20 (Becker&Hickl, Berlin, Germany), as well as an LSM 880 (Carl Zeiss, Jena, Germany) laser scanning microscope equipped with a FLIM module, the SPC 150 TCSPC (Becker & Hickl GmbH, Berlin, Germany) and a Mai Tai HP femtosecond laser, 80 MHz, 140 fs (Spectra Physics, Milpitas, CA, USA) were used for fluorescence lifetime imaging microscopy (FLIM).
Using MPTflex, the images were acquired through a 40×, 1.3 NA oil immersion objective. In the case of spheroids, the images were acquired from a depth of ~20 μm. BODIPY 2 fluorescence was excited at a wavelength of 850 nm and detected in the range of 409–680 nm using a fixed pre-fitted emission filter. The average power applied to the sample was ~7 mW. The acquisition time was ~7 s per image. On the LSM 880 microscope, BODIPY 2 fluorescence was excited at a wavelength of 850 nm and detected in the range of 500 to 550 nm. A C Plan-Apochromat 40×/1.2 NA objective lens was used for image acquisition. The FLIM images were acquired at a laser power of 1–2%, with a photon collection time of 60 s.
Fluorescence lifetime analysis was performed in the SPCImage 8.3 software (Becker & Hickl, Germany). The collected amount of photons per the decay curve was at least 5000. FLIM images were obtained from 10 randomly selected fields of view in each culture dish. Fluorescence decays at each pixel of the whole image were fitted using a monoexponential model. The goodness of the fit χ2 ≤ 1.20 indicates that the model used provided a reasonable fit. The fluorescence lifetime of BODIPY 2 was measured in the plasma membranes of cells by manually selecting zones of the plasma membrane as regions of interest.
Experimentally measured lifetimes of BODIPY 2 (in ns) were converted to viscosity values (in cP) using previously obtained calibration plots [
40].
4.5. ToF-SIMS
HeLa–Kyoto cells (5 × 105) were seeded on confocal dishes containing clean and dry poly-L-lysine-coated cover glass and were incubated at 37 °C and in an atmosphere of 5% CO2 for 24–48 h. Then, paclitaxel (3.2 nM) was added to the culture medium. After 1 or 24 h incubation with paclitaxel, cells were washed three times with phosphate buffer (PBS), and then cells were incubated with 4% paraformaldehyde (PFA) for 45 min at room temperature for chemical fixation. Afterward, cells were washed three times with PBS. In total, three samples were prepared—control samples without drugs and cells incubated with paclitaxel for 1 h and 24 h. Fixed cells were stored in PBS for ~5–6 h due to shipping to the ToF-SIMS laboratory. Cells were washed with mQ water to remove excess salts. Drying was carried out under a gentle stream of argon at room temperature for 2 h.
Mass spectra were acquired with a ToF-SIMS 5 (ION-TOF Gmbh, Münster, Germany) equipped with a 30 keV Bi3+ liquid metal ion source. Twelve mass spectra were recorded for each sample in both positive and negative ion polarities, with an analysis area of 300 × 300 µm2, and the raster was 64 × 64 pixels. The primary ion dose density was 43 × 1011 ions/cm2 to maintain static SIMS conditions. A low-energy electron flood gun was activated for charge compensation in all experiments. Lipid ion yields were calculated from the intensity of the corresponding peak of interest normalized to the total ion count amount.
The areas within a sample with morphological and compositional anomalies, such as dead cells and contamination particles, were excluded from the analysis. Moreover, dead cells, if present in a sample, are detached during sample preparation for ToF-SIMS. So the lipids analysis was performed mostly on a viable cell population.
4.6. Statistics
The mean values (M) and standard deviations (SD) were calculated for the microviscosity values. Student’s t-test was used to compare data (p < 0.05 was considered statistically significant). The number of cells for mean value calculations was 50–70 in 7–10 fields of view.