Pharmacological targeting of membrane rigidity: implications on cancer cell migration and invasion

Soraphen A was isolated by Gerth et al as described previously [19] and dissolved in ethanol. The invasive mammary carcinoma cell line MDA-MB-231 was obtained from Cell Line Services (Eppelheim, Germany) and cultured in DMEM media (PAA, Coelbe, Germany) supplemented with 10% fetal calf serum (FCS) and 1% penicillin/streptomycin (P/S). T24 bladder carcinoma cells were kindly provided by Barbara Mayer (Department of Surgery, University of Munich, Germany) and recently authenticated by the DSMZ (Braunschweig, Germany). T24 cells were maintained in McCoy's 5A medium (PAA) supplemented with 10% FCS, 2 mM L-glutamine and 1% (P/S).

Phospholipids were extracted, separated by reversed phase liquid chromatography and detected by ESI tandem mass spectrometry as described in [20]. In brief, water, methanol, chloroform and saline (final ratio: 14:34:35:17) were successively added to the cells. The organic layer was evaporated and the residual dissolved and diluted in methanol. The extracted phospholipids were separated on an Acquity UPLC BEH C8 column (1.7 μm, 1 × 100 mm, Milford, MA) using an AcquityTM Ultraperformance LC system (Waters, Milford, MA). The chromatography system was coupled to a QTRAP 5500 Mass Spectrometer (AB Sciex, Darmstadt, Germany) with an electrospray ionization source. Both fatty acid anion fragments were detected through multiple reaction monitoring. The most intensive transition was selected for quantification. Mass spectra were processed using Analyst 1.6 (AB Sciex, Darmstadt, Germany). The reported method was optimized to compare phospholipid profiles between samples but not for absolute quantification.

Cells were detached from flask and re-suspended in culture medium. Cell suspension in tube was connected to the optical stretcher setup [21], a dual beam fiber laser trap (figure 1(a)). Whole deformation measurement was recorded as an image series (30 fps). The microscope stage temperature was held constant at 23 °C. A minimum of 300 cells were recorded for each experiment. Cell numbers used for analysis after sorting out rotating cells are shown in figures 4(c) and (d).

Zoom In Zoom Out Reset image size

An edge detection Matlab algorithm (The MathWorks Inc., Natick) constructs deformation data from the images and corrects for small angle rotations in the trap (figure 1(b)). Relative deformation is the ratio of elongation and initial cell diameter d₀ along laser axis,  = (d(t)−d₀)/d₀, and plots are shown as creep deformation J(t) = /σ₀, with σ₀ being the optically induced stress linearly dependent on the applied laser power P_Stretch (figure 1(c)) [22, 23]. Creep deformation J is shown as median, appropriate for the non-Gaussian distribution of the data, and bootstrapping was used to estimate a 95% confidence interval. Additionally, a two-sample Kolmogorov–Smirnov test was used to control if the data are from different distributions at the end of the stretch phase.

Prior to vesiculation, cells were grown to 90% confluence in 75 cm² tissue flasks. The culture medium was removed and rinsed with a buffer solution composed of 150 mM NaCl (Sigma-Aldrich, S7653), 10 mM 2-(4-(2-hydroxyethyl)-1-piperazinyl)-ethanesulfonic acid (HEPES, Sigma-Aldrich, St.Louis, MO, USA, H3375) and 2 mM CaCl₂ (Sigma-Aldrich, C5080). Buffer ingredients, di-thiothreitol (DTT) (Roth, 6908) and paraformaldehyde (PFA) (Sigma-Aldrich, P6148) powder were dissolved in ultra-pure water (Milli-Q system Integral 5, Merck Millipore, Billerica, MA, USA, R > 18 MΩ cm). Afterwards, cells were incubated in a shaker (37 °C, 5% CO₂ and 60 cycle min⁻¹) for 120 min in 2.5 ml buffer, containing 25 mM PFA and 4 mM DTT. After shaking, the content was settled on ice for about 45 min. In order to observe GPMVs, the upper 3/4 of the solution was pipetted on an object plate for imaging (optical phase contrast microscope, 100× oil immersion objective, LEICA DM IRB).

The theoretical background of the fluctuation analysis was introduced by Helfrich in 1973 [24]. He proposed the curvature energy per unit area of bilayers:

$$\begin{eqnarray}&&w=\displaystyle \frac{1}{2}\kappa {\left({c}_{1}+{c}_{2}+{c}_{0}\right)}^{2},\end{eqnarray} \tag{ 1 }$$

where c₀ is the spontaneous curvature, κ the bending rigidity and c₁ and c₂ principle curvatures. More than 10 years later, Schneider extended Helfrich's model to quasi-spherical vesicles [25]. Schneider described thermally excited membrane fluctuations using displacements u(θ, φ, t) of a vesicle from its spherical form and decomposed them in spherical harmonic eigenfunctions Y_lq(θ, φ). The radius of the vesicle can be expressed as:

$$\begin{eqnarray}&&r(\theta ,\varphi )=R\left(1+\displaystyle \sum _{l=2}^{{l}_{\rm max}}\displaystyle \sum _{q=-l}^{l}{a}_{lq}(t){Y}_{lq}\left(\theta ,\varphi \right)\right)\end{eqnarray} \tag{ 2 }$$

a_lq(t) are time-dependent amplitudes, θ the polar angle, φ the azimuthal angle, l the azimuthal, and q the magnetic quantum number. R describes the averaged radius of the sphere. States with l < 2 correspond to spherical translations and violate volume conservation. Provided the thermally excited modes follow the equipartition theorem and the vesicle is quasi-spherical, dimensionless mean square amplitudes a_lq from each spherical harmonic can be expressed as:

$$\begin{eqnarray}&&{\langle {\left|{a}_{lq}\right|}^{2}\rangle }_{t}=\displaystyle \frac{1}{n}\displaystyle \sum _{i=1}^{N}{\left|{a}_{lq}\right|}^{2}\\ &&\quad \quad \quad =\displaystyle \frac{{k}_{B}T}{\kappa (l-1)(l+2)\left[l(l+1)\right]+\sigma ^{\prime} }\end{eqnarray} \tag{ 3 }$$

T denotes the temperature, k_B the Boltzmann constant, and σ' = σ_eff R²/κ introduces the effective tension. The expected value is calculated as ensemble average.

The analysis of thermally excited membrane fluctuations is based on sequences of snapshots obtained by optical microscopy. A representative ensemble of n images (≈10 000) per vesicle was recorded with an iXon camera (Andor, UK) while the vesicles were observed by phase contrast microscopy. We worked with acquisition times of about 1 min and rates between 90 and 150 frames per second to guarantee that vesicles were able to constitute most of its available configurations. For typical giant vesicles 1 min is much longer than the recurrence or the relaxation times of the excitation [26].

Equation (3) depends only on l, which allows studying membrane fluctuations by observing just one plane of the vesicle. In the experimental setup we obtained information from the equatorial plane only (θ = π/2). In order to detect the vesicles contour R(φ), a Matlab (The MathWorks Inc., Natick) gradient based edge detection algorithm with subpixel resolution [21] was used. The shape of each single vesicle is described with 256 radii R(φ_i). φ_i denotes the azimuthal angle in the image plane. Fourier analysis of the observed 2D vesicle contour fluctuation was correlated to Schneider's 3D theory [27, 28]. The relative deformations u(φ, t) of the vesicle can be expressed as amplitudes V_q(t) in the equatorial plane using Fourier transformation:

$$\begin{eqnarray}&&u(\varphi ,t)=R\displaystyle \sum _{q=0}^{{q}_{\rm max}}{V}_{q}(t){{\rm{e}}}^{{\rm{i}}q\varphi }.\end{eqnarray} \tag{ 4 }$$

The mean square values of the amplitudes V_q(t) were obtained from equations (2) and (4):

$$\begin{eqnarray}&&{\langle {\left|{V}_{q}\right|}^{2}\rangle }_{t}=\displaystyle \frac{{k}_{B}T}{\kappa }\displaystyle \sum _{l\geqslant q}^{{l}_{\rm max}}\displaystyle \frac{{N}_{lq}^{2}{\left[{P}_{l}^{(q)}\left({\rm cos}\;\pi /2\right)\right]}^{2}}{\left(l+2\right)\left(l-1\right)\left[l\left(l+1\right)+\sigma ^{\prime} \right]}.\end{eqnarray} \tag{ 5 }$$

$\langle | {V}_{q}{| }^{2}\rangle$ was estimated by averaging V_q over the number of taken images n per vesicle. ${N}_{lq}{P}_{l}^{\left(q\right)}$ (cos(π/2)) denote the fully normalized associated Legendre functions in the image plane and q the mode number. We realized the Fourier transformation from equation (4) with a discrete transformation to obtain discrete coefficients. The measured mean square values can be introduced in equation (5) to obtain κ and σ_eff by a two-parameter fit. The sum in equation (5) is rapidly converging for moderate tensions [28].

As seen from equation (5) the effective tension σ_eff dominates the fluctuations for low-wave-number modes which are typically ignored in the fluctuation analysis [28]. These modes (q < 7) lead to systematically too high bending rigidities. The tension dominated regime is followed by an intermediate one used to calculate the bending rigidity. In this regime, κ is practically independent of the wavenumber and mean square amplitude of the contour fluctuations $\langle | h\left({q}_{x}\right){| }^{2}\rangle$ scales as ${{q}_{x}}^{-3}$ [29]. High-wavenumber modes (q ≥ 16) are noise dominated and were also ignored. In the present work we used vesicles with observable flicker and only modes 7 ≤ q < 16 to determine the bending rigidity. The fits in figure S1 show the typical behavior of the bending dominated regime ($\langle | h\left({q}_{x}\right){| }^{2}\rangle \sim {{q}_{x}}^{-3})$ for the analyzed modes in agreement with [29, 30].

Engelhardt further suggested a simplification for practically tension free vesicles [27]:

$$\begin{eqnarray}&&\langle \left|\left.{V}_{q}^{2}\right|\right.\rangle =\displaystyle \frac{{R}^{2}{k}_{B}T}{\kappa }{\displaystyle \sum _{l\geqslant q}^{{l}_{\rm max}}\left(\displaystyle \frac{{N}_{lq}{P}_{l}^{\left(q\right)}\left(0\right)}{(l+2)(l-1)}\right)}^{2}.\end{eqnarray} \tag{ 6 }$$

R denotes the average radius of each vesicle obtained from all images n per dataset. A two-sample Kolmogorov–Smirnov test was performed to control if the data are from different distributions.

2 × 10⁵ MDA-MB-231 or T24 cells, respectively, were seeded on a six-well plate and stimulated with Soraphen A for 2 h. For MDA-MB-231 Boyden chambers with 8 μm inserts and for T24 cells 5 μm inserts were used. The migration time was 4 (MDA-MB-231) or 16 h (T24), respectively. The invasion time was 24 h for both cell lines. Non-migratory or non-invading cells were removed from the upper insert compartment with a cotton bud. Migrating and invading cells, respectively, were visualized by crystal violet staining and counted.

Cell death rate was determined with propidium iodide (PI)—exclusion assays. In brief, cells were treated for 24 h with increasing concentration of Soraphen A, washed, stained with 50 μg ml⁻¹ Propidium iodide, and analyzed by flow cytometry using a FACSCalibur (Becton Dickinson, Heidelberg, Germany).

Invasion and migration assays were performed three times. Data are expressed as means ±S.E.M and analyzed using one-way-ANOVA +Tukey HSD post hoc or Student's t test. Values of p < 0.05 were considered as significant.

To investigate whether modulations of lipid composition of plasma membranes influence the biomechanical properties of cancer cells and thus their metastatic potential, in the first step, the breast cancer cell line MDA-MB-231 was treated for 6 h with the ACC1 inhibitor Soraphen A, and a detailed lipidome analysis was performed with ESI tandem mass spectrometry. Extending first reports regarding the effect of Soraphen A on the phospholipid content after long incubation times of 72 h [15], the ACC inhibitor does not alter the total phospholipid content (measured as sum of all species detected). One exception was phosphatidylserine (PS), whose levels were slightly but non-significantly increased (figure 2(a)). Of note, desaturation in all analyzed phospholipid species increased after treatment with Soraphen A (figures 2(b)–(e)).

Zoom In Zoom Out Reset image size

However, the composition of membrane phospholipids, including phosphatidylcholines (PC), phosphatidylethanolamines (PE), phosphatidylserines (PS), and phosphatidylinositols (PI) is strongly altered by Soraphen A treatment after the short treatment time of only 6 h (figures 3(a)–(d)). The proportion of phospholipids is shifted towards those with longer fatty acid chains, whereas the proportion of phospholipids with shorter fatty acid chains significantly declined after incubation with Soraphen A.

Zoom In Zoom Out Reset image size

Since phospholipids are the main components of cellular membranes, we assumed that Soraphen A induced changes in phospholipid composition might affect rigidity of membranes and deformation of whole cells. Hence, by evaluating membrane fluctuations [31], bending rigidity of cellular membranes was determined as described in [27, 32]. In order to exclude effects such as cytoskeletal restrains, membrane rigidity was determined using GPMVs [33]. GPMVs are derived from viable cells by vesiculation—an energy, pH, and temperature-dependent procedure. The process is chemically induced by PFA and potentiated with mono- and divalent cations as well as 1,4-di-thiothreitol (DTT) [33, 34]. A contraction of the cytoskeleton induces an increased hydrostatic pressure inside the cell and pieces of cellular membrane separate from the cytoskeleton [35]. It should be noted, that cellular membranes in living cells are dynamic systems where cell signaling, exo-, and endocytosis constantly occur, and it is reported that those processes can also be affected by membrane mechanics [36, 37]. To generate GPMVs, the cell membrane is decoupled from the underlying cytoskeleton during vesiculation, which also disrupts these fundamental cellular processes and may inhibit some membrane enzymes [38]. We cannot exclude that these processes are influenced during vesiculation and thus have an impact on bending rigidity. Nevertheless, cells are not homogenized during vesiculation and GPMVs are representatives of the cell surface [38]. Most of the observed vesicles had radii between 5 and 15 μm. GPMVs contain numerous membrane lipids and polypeptides in contrast to artificially biomimetic membranes and are devoid of cortical actin assembly [34]. Hypotonic lysis removes cytosolic ingredients. This allows interaction-free measurement of rigidity of cellular membranes via Fourier analysis of thermal vesicle shape fluctuation. Fourier analysis or flicker spectroscopy of the fluctuating vesicle contour is a sensitive method to estimate the bending rigidity κ of thermally excited vesicle membranes. The aggressive cancer cells lines MDA-MB-231 (breast cancer cells) and T24 (urinary bladder carcinoma cells) were incubated with 1 μM Soraphen A for 2 h. 45 treated and 48 untreated vesicles obtained from MDA-MB-231 cells were measured and 30 treated and 28 untreated vesicles for T24 cells, respectively. Notably, Soraphen A did not apparently affect the efficiency of the vesiculation process.

As presented in figures 4(a) and (b), bending rigidity of cell membranes obtained from Soraphen A treated MDA-MB-231 cells (κ = 2.2 × 10⁻¹⁹ J) showed an increased median compared to untreated GPMVs (κ = 1.0 × 10⁻¹⁹ J). The same behavior was observed for membranes obtained from T24 cells (κ = 4.4 × 10⁻¹⁹ J for treated and κ = 2.3 × 10⁻¹⁹ J for untreated vesicles). The results indicate that the inhibition of ACC1 and the resulting modulation of phospholipid composition enhances the rigidity of cellular membranes.

Zoom In Zoom Out Reset image size

In addition, optical stretcher measurements were performed to determine whole cell deformation and thus stiffness of cells in the presence of an intact cytoskeleton [21]. Cells were pumped in a microfluidic system, serially stopped in the focus of the phase contrast microscope, and trapped by the laser with 100 mW. A single cell deformation measurement is composed of a 1 s trap of the cell, following a 2 s stretch phase with a laser power of 800 mW, and another trap phase for 2 s to record relaxation. The setup is illustrated in figure 1(a). As shown in figure 4(c), cellular stiffness of MDA-MB-231 cells was increased after incubation with 1 μM Soraphen A for 2 h. 101 treated and 137 control cells were measured. In contrast, the ACC1 inhibitor did not alter median creep deformation J of T24 cells (157 treated and 185 control cells) (figure 4(d)). The median of the creep deformation J was analyzed at the end of the stretch phase (t = 3 s) as most significant parameter (figures 4(e) and (f)). Soraphen A has affected membrane rigidity and cytoskeleton in the breast epithelial cell line, where as in the urinary bladder cell line only membrane rigidity was affected. The stage of the optical stretcher is held at 23 °C during all measurements. The optical stretcher heats the cells inside the capillary during the measurement up to 44 °C for 2 s in the center of the trap [39, 40]. Although we cannot exclude phase transition induced changes in bending rigidity of whole cell membranes during the stretching process, this should not have an impact on the measured creep deformation. This is because the contribution of membrane rigidity to creep deformation is significantly lower compared to the underlying intact cytoskeleton in these whole cell experiments.

To investigate whether Soraphen A has a functional impact on metastatic potential of cancer cells, Boyden chamber assays were performed. Soraphen treated cells were detached and plated into uncoated (for analysis of migration) or matrigel coated (for invasion assays) Boyden chamber inserts and allowed to migrate or invade towards a 10% FCS + 100 ng ml⁻¹ EGF gradient. Despite the different behavior of T24 and MDA-MB-231 cells in the optical stretcher experiments, Soraphen A inhibits migration and invasion capacity of both cell lines to a similar extent (figures 5(a)–(d)) without significantly affecting cell death (figure 6). The strongest effect of Soraphen A on cell invasion and migration was observed for cells treated with 1 μM Soraphen A.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size