Cell membrane softening in human breast and cervical cancer cells

The use of mammary and cervical tumour samples obtained from the University Hospital Leipzig was approved by the ethics committee of the medical department of the Leipzig University. All patients in this study signed a consent document prior to surgery. After tumour classification by the pathology department, the remaining vital tissue was processed for cell isolation and culture.

Tumour and healthy tissues were obtained from patients by surgical excision performed at Leipzig University Hospital. Figure 1 shows tissue samples of a human breast tumour and how cells were extracted.

Zoom In Zoom Out Reset image size

In order to dissociate tissue samples (figure 1(a)) and form a cell monolayer, samples were cut into smaller pieces of about 1 mm³ (figure 1(b)). Digestion was performed in a gentleMACS C tube (Miltenyi Biotec, Germany) containing 5 ml Dulbecco's Modified Eagle/HAM's F-12 medium (DMEM/HAM's F12, Biochrom, FG 4815) supplemented with 20 μg ml⁻¹ DNase I (AppliChem, A3778,0010). For breast tissue, 1.6 mg ml⁻¹ collagenase P (Roche, 11213857001) was added. For cervical tissue, 0.25 mg ml⁻¹ collagenase 1A (Sigma-Aldrich, C9722) and 0.25 mg ml⁻¹ pronase (Roche, 10165 921001) was added. To enhance the dissociation process, gentleMACS C tubes were processed in a gentleMACS Dissociator (Miltenyi Biotec, Germany) and subsequently incubated at 37 °C and 5% CO₂ in a humid atmosphere for 30–60 min. The last two steps were repeated until at least 80% of the tissue were dissociated.

The obtained cell suspension was centrifuged in two steps (at 40 g for 1 min and 300 g for 10 min) to separate epithelial cells from clusters, debris, and other cell types such as fibroblasts [25]. The remaining cell pellet was re-suspended in DMEM/HAM's F12 supplemented with 10% fetal calf serum (FCS, Biochrom, S 0615) and 1% antibiotic-antimycotic solution (Sigma-Aldrich, A5955) and cultured in six-well plates with the same cell culture medium (figure 1(c)). After 24 h, the medium was exchanged with serum-free HuMEC medium (Life Technologies, 12752-010) for breast cancer cells or with defined keratinocyte serum-free medium (Life Technologies, 10744-019) for cervical cells. Both media were supplemented with 1% antibiotic-antimycotic solution. Cell medium was renewed every 2 to 3 days.

Since passaging of cells can influence their biomechanical behaviour in vitro, primary cell membranes were obtained from cells which had not been passaged before and were only in culture for 5 to 10 days.

Non-malignant MCF-10A breast epithelial cells (ATCC, CRL-10317) were cultured in 1:1 DMEM/Ham's F12 medium supplemented with 100 ng ml⁻¹ cholera toxin (Sigma-Aldrich, C8052), 20 ng ml⁻¹ human epidermal growth factor (Sigma-Aldrich, E9644), 10 μg ml⁻¹ insulin (Sigma-Aldrich, I9278), 500 ng ml⁻¹ hydrocortisone (Sigma-Aldrich, H0888) and 5% horse serum (PAA Laboratories, B15-021) and incubated at 37 °C and 5% CO₂ in a humidified atmosphere. Cells were cultured in 75 cm² flasks (TPP). Culture medium was changed every 2 to 3 days.

For passaging or extraction of lipids for MS, cells were rinsed with phosphate buffered saline (PBS, Life Technologies, 18912-014) and detached by applying 1 ml 0.025% trypsin/EDTA solution (Biochrom, L2143). Afterwards, 5 ml culture medium was added to inhibit trypsin. The trypsin containing medium was removed by centrifugation at 100 g for 4 min and the cell pellet was re-suspended in 1 ml culture medium. The cell suspension was either extracted according to the Bligh and Dyer method to extract the lipids [26], or cultured again in a new flask with fresh medium under the same conditions as already described above.

Malignant MDA-MB-231 (ATCC, HTB-26) breast epithelial cells were cultured in 90% DMEM, supplemented with 10% FCS at 37 °C and 5% CO₂ in humidified atmosphere. Cells were cultured in 75 cm² flasks. Cell medium was renewed every 2 to 3 days. Passaging and lipid extraction were performed in the same way as described above for MCF-10A cells.

GPMVs are bilayer spheres (1–100 μm) which are produced from biological cells by the vesiculation mechanism. During vesiculation, the membrane dissociates from the actin cortex and small pieces of the plasma biomembrane separate from the cytoskeleton. A higher hydrostatic pressure inside the cells, induced by a contraction of the cytoskeleton, leads to the formation of GPMVs [27–29].

Cells were grown to approximately 90% confluency in 75 cm² tissue culture flasks prior to the vesiculation process. After removal of the cell culture medium, cells were rinsed twice with GPMV buffer. The GPMV buffer, composed of 150 mM NaCl (Sigma-Aldrich, S7653), 10 mM 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES, Sigma-Aldrich, H3375) and 2 mM CaCl₂ (Sigma-Aldrich, C5080) was prepared with ultra-pure water (Milli-Q system, Integral 5, Merck Millipore, r > 18 MΩ cm).

In order to trigger vesiculation, 2.5 ml GPMV buffer including 25 mM PFA (Sigma-Aldrich, P6148) and 4 mM 1,4 di-thiothreitol (DTT, Roth, 6908) was added to the washed cell monolayer [13, 30]. Mono- and divalent cations, disulfide reducing agents such as DTT and hypertonic medium potentiate the vesiculation process [14]. The pretreatment of cells, e.g. with chemical agents that destroy the cytoskeleton, is not necessary. Disrupting the microfilaments and microtubules is not accompanied by detectable effects on vesiculation [31, 32]. Membrane blebbing occurred during shaking for 120 min at 37 °C, 5% CO₂ and 60 cycle min⁻¹. Afterwards, the content of the flask was decanted into a 15 ml centrifuge tube containing 4 ml GPMV buffer solution and cooled on ice for 45 min The upper ¾ of the solution in the centrifuge tube contain most of the GPMVs and were transferred directly onto a microscope slide and immediately imaged via phase contrast microscopy (Leica, DM IRB, Leica Microsystems, Germany) with a 100× immersion objective.

Based on vesicle edge R(θ, φ, t) and displacement u(θ, φ, t) from its spherical form, bending rigidity κ can be calculated by Fourier analysis of the thermal shape fluctuation [15, 30, 33–35].

$$\begin{eqnarray}&&R\left(\theta ,\varphi ,t\right)={R}_{0}\left(1+u\left(\theta ,\varphi ,t\right)\right)={R}_{0}\left(1+\displaystyle \sum _{l=2}^{{l}_{{\rm{max}}}}\displaystyle \sum _{m=-l}^{l}{a}_{lm}(t){Y}_{lm}\left(\theta ,\varphi \right)\right),\end{eqnarray} \tag{ 1 }$$

R₀ is the average radius and R(θ, φ, t) the current radius of the vesicle. The displacements are decomposed in spherical harmonic eigenfunctions Y_lm(θ, φ) and their time-dependent amplitudes a_lm(t). l denotes the azimuthal, m the magnetic quantum number, θ the polar and φ the azimuthal angle. States with l = 0 do not conserve the volume and states with l = 1 correspond to translations of the sphere (equation (1)).

The assumption that single oscillating modes are independent (hydrodynamic theory) allows for the calculation of time-dependent amplitudes a_lm from each spherical harmonic. According to the equipartition theorem, mean square amplitudes of the spherical harmonic modes behave as described by Gracià et al [33]:

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

σ' = σ_eff R²/κ introduces the effective tension σ_eff, κ denotes the bending rigidity, T the temperature and k_B the Boltzmann constant. The sum in equation (2) depends only on l, allowing the study of vesicle membrane fluctuations by observing only the equatorial plane (θ = π/2). By employing Fourier transformation, displacements u(φ,t) of the vesicle can be expressed as amplitudes V_q(t) [33]:

$$\begin{eqnarray}&&u(\varphi ,t)=R\displaystyle \sum _{q=0}^{{q}_{{\rm{max}}}}{V}_{q}(t){{\rm{e}}}^{{\rm{i}}q\varphi }.\end{eqnarray} \tag{ 3 }$$

Substituting equations (1) and (2) into (3) yields mean square values of the fluctuation amplitudes:

$$\begin{eqnarray}&&{\langle {\left|{{V}}_{{q}}\right|}^{2}\rangle }_{t}=\displaystyle \frac{{k}_{B}T}{\kappa }\displaystyle \sum _{l\geqslant q}^{{l}_{\rm max}}\displaystyle \frac{{N}_{lq}^{2}{\left[{P}_{l}^{(q)}\left({\rm cos}\pi /2\right)\right]}^{2}}{\left(l+2\right)\left(l-1\right)\left[l\left(l+1\right)+\sigma ^{\prime} \right]}.\end{eqnarray} \tag{ 4 }$$

In equation (4), q is the mode number and N_lq P^(q)_l (cos(π/2)) are fully normalized associated Legendre functions. The sum in equation (4) is rapidly converging for moderate tensions and was calculated up to l_max = 400 [33].

As seen from equations (2) and (4) three regimes depending on q can be observed in the experimentally measured variance. Low-wave number modes are dominated by non-linear tension effects and are ignored in our fluctuation analysis. Higher order modes are noise dominated and were also not considered. The intermediate regime is bending dominated and was used to determine the bending rigidity κ in our experiments. In this regime the measured κ is nearly constant, viz. independent of the mode number q. Here the statistical variance (equation (S3)) of the fluctuation amplitude V_q scales ∼q⁻³ in agreement with Gracià et al [33].

In our experiments, vesicles and their contours were observed by phase contrast microscopy (figure 2(a)). Representative sequences of n images (≈10 000) per vesicle were recorded with an iXon DV887 EMCCD camera (Andor Technology, UK) with frame rates between 90 and 150 fps, depending on the binning size. An acquisition time of approximately 1 min was used, which guaranteed that GPMVs were able to move through the majority of their available configurations [33, 36]. A self-written MATLAB (The MathWorks, Natick, MA, USA) gradient based edge detection algorithm was employed to detect the contours of individual vesicles with subpixel resolution (figure 2(b)). After applying the edge detection, each point of the edge R(φ_i) was rescaled using the average radius.

Zoom In Zoom Out Reset image size

The Fourier coefficients V_q from equation (4) were obtained via a fast Fourier transformation and their mean square values were calculated by averaging |V_q|² over the number n of taken images (ensemble average).

The mode number independent κ (figure 3(a)) and the $\langle | {V}\left({q}\right){| }^{2}\rangle$ ∼ q⁻³ behaviour (figure 3(b)) demonstrate that our analyzed modes (7 ≤ q < 16) are bending dominated in agreement with [37, 38]. Following Pecreaux et al and newer publications of Yoon et al and Gracià et al [33, 37, 38] we used

$$\begin{eqnarray}&&\langle {\left|V\left(q\right)\right|}^{2}\rangle =\displaystyle \frac{{k}_{B}T}{2\sigma }\left[\displaystyle \frac{{R}_{0}}{q}-\displaystyle \frac{1}{\sqrt{\displaystyle \frac{\sigma }{\kappa }+{\left(\displaystyle \frac{q}{{R}_{0}}\right)}^{2}}}\right]\end{eqnarray} \tag{ 5 }$$

to fit our data and verify the $\langle | {V}\left({q}\right){| }^{2}\rangle$ ∼ q⁻³ behaviour and thus bending rigidity domination (figure 3(b)).

Zoom In Zoom Out Reset image size

The MALDI-TOF soft-ionization MS technique allows analysis of samples without significant fragmentation of the analytes of interest [24]. About 600 μl of the obtained cellular lipid extracts (∼5 × 10⁶ cells) were dried under vacuum and subsequently re-solubilized by addition of 10 μl matrix solution (2,5-dihydroxybenzoic acid (DHB, 0.5 M in methanol, Sigma-Aldrich, 39319)) [39]. The lipid moiety of cells was extracted according to the Bligh and Dyer method [26]. Then, 1 μl of the sample/matrix mixture was transferred onto a gold-coated MALDI target. To overcome assignment problems (interferences between differences in the fatty acyl compositions of lipids and the simultaneous generation of H⁺ and Na⁺ adducts), selected analyses were repeated with the same DHB matrix solution that has been saturated with caesium chloride (CsCl, ICN, 150589) as described in [40]. All MALDI-TOF mass spectra were recorded on a Bruker Autoflex MS device (Bruker Daltonics, Germany). The system utilizes a pulsed nitrogen laser emitting at 337 nm. All measurements were performed in the reflector mode using delayed extraction conditions. Positive ion spectra were exclusively recorded. Raw data were processed using the software 'Flex Analysis' version 2.2 (Bruker Daltonics). The intensities of the observed peaks are given relative to the intensity of the most intense peak. Further information regarding MALDI MS analysis of lipids is available in [24].

Tumour samples of breast and cervical cancer patients were analyzed and compared with measurements from two different breast epithelial cell lines (non-malignant MCF-10A and malignant MDA-MB-231). For both tumour types, corresponding normal epithelial cells from the same patient were used as reference. This approach allows a comparison of plasma membranes obtained from cancer and non-malignant cells. Most of the observed GPMVs had diameters between 10 μm and 30 μm. Based on the vesicle edges observed via phase contrast microscopy, bending rigidities were calculated by Fourier fluctuation analysis. For our calculation, only flaccid vesicles with a low excess area compared to the minimal spherical shape were analyzed to avoid membranes under mechanical tension.

Figures 4(a) and (b) show bending rigidity measurements of GPMVs obtained from primary breast and cervical cells, presented as relative frequency count over κ. Vesicles obtained from two breast epithelial cell lines, non-malignant MCF-10A and malignant MDA-MB-231, were investigated as references. These cell lines shown in figure 4(c) constitute an established model system to compare the behaviour of cancerous and healthy cells [41]. Vesicles from healthy (N = 36) and malignant (N = 44) primary breast cells as well as healthy (N = 34) and malignant (N = 31) ones from primary cervical cells were measured. For primary breast and cervical cancer cells we found a significantly increased number of GPMVs with lower bending rigidities in contrast to GPMVs obtained from normal tissue of the same patient. These results are confirmed by the established model cell lines (figure 4(d)). Here 75 vesicles from MCF-10A and 79 from MDA-MB-231 were analyzed.

Zoom In Zoom Out Reset image size

The rigidity of primary mamma and cervix cell membranes are characterized by a twofold decreased median for cancer cells compared to normal cells (figures 4(a) and (b)). We found κ = 3.2 × 10⁻¹⁹ J for GPMVs from healthy and κ = 1.7 × 10⁻¹⁹ J for GPMVs from malignant primary breast cells (mean deviation from the median MD = 2.5 × 10⁻¹⁹ J and MD = 1.6 × 10⁻¹⁹ J, respectively), κ = 6.2 × 10⁻¹⁹ J for GPMVs from healthy and κ = 3.7 × 10⁻¹⁹ J for GPMVs from malignant primary cervical cells (MD = 3.3 × 10⁻¹⁹ J and MD = 3.1 × 10⁻¹⁹ J, respectively). The same behaviour was observed for membranes obtained from breast cell lines resulting in membrane softening by a factor of 3. A median rigidity of κ = 4.5 × 10⁻¹⁹ J was measured for vesicles from MCF-10A cells and κ = 1.3 × 10⁻¹⁹ J for GPMVs from MDA-MB-231 cells (MD = 1.4 × 10⁻¹⁹ J and MD = 1.2 × 10⁻¹⁹ J, respectively).

To investigate a potential correlation between membrane rigidity and lipid composition of cancer cells compared to their non-malignant counterparts, positive ion MALDI-TOF MS was employed for all investigated samples. Relative amounts of the individual lipids can be easily determined by comparison of the peak intensities. Two significant effects were found. First, we observed a 30%–40% decreased level of SM 16:0 in primary cancer cells in contrast to normal ones. Changes were normalized to the base peak corresponding to the most abundant membrane lipid, PC (16:0/20:4), see figure 5.

Zoom In Zoom Out Reset image size

Second, we measured an increased proportion of PC with shorter fatty acyl chain lengths in malignant cells compared to non-malignant ones. Apart from the chain lengths of fatty acids, an increased proportion of fatty acids with a higher degree of desaturation was detected in non-malignant compared to malignant cells of the same patient (figure 5).