Cancer cell migration on straight, wavy, loop and grid microfibre patterns

Cell migration plays an important role in physiological and pathological processes where the fibrillar morphology of extracellular matrices (ECM) could regulate the migration dynamics. To mimic the morphological characteristics of fibrillar matrix structures, low-voltage continuous electrospinning was adapted to construct straight, wavy, looped and gridded fibre patterns made of polystyrene (of fibre diameter ca. 3 μm). Cells were free to explore their different shapes in response to the directly-adhered fibre, as well as to the neighbouring patterns. For all the patterns studied, analysing cellular migration dynamics of MDA-MB-231 (a highly migratory breast cancer cell line) demonstrated two interesting findings: first, although cells dynamically adjust their shapes and migration trajectories in response to different fibrillar environments, their average step speed is minimally affected by the fibre global pattern; secondly, a switch in behaviour was observed when the pattern features approach the upper limit of the cell body’s minor axis, reflecting that cells’ ability to divert from an existing fibre track is limited by the size along the cell body’s minor axis. It is therefore concluded that the upper limit of cell body’s minor axis might act as a guide for the design of microfibre patterns for different purposes of cell migration.


Introduction
Cell migration refers to the polarisation and locomotion of living cells. It can occur spontaneously and also be driven by a variety of biological, chemical and physical signals [1][2][3]. Extensive efforts have been made to investigate the effects of physical signals on cell migration, including the role of surface topography and roughness that mimic characteristics of extracellular matrices (ECM) [4][5][6][7][8][9][10]. Since the physical architecture and recapitulation of native ECM are complex with multiple global patterns and local topographies, reductionist experimental approaches are valuable in providing mechanistic insights.
Natural ECM derived matrices and synthetic polymers can exist in the form of hydrogels, or other chemically or physically crosslinked polymeric scaffolds [11][12][13][14][15][16][17][18][19]. These different forms of material interfaces have been used to investigate cell migration. For example, cell-scale collagen bundles aligned on microfluidic devices were used to elicit cellextracellular matrix interaction in modelling local cancer invasion [20,21]. Meanwhile, synthetic polymeric materials provide alternative strategy in modelling the physical features of ECM where biochemical factors attributed by naturally derived matrices can be decoupled. For example, micron-scale features (such as grooves, pits or posts) can by fabricated by stereo-lithography techniques on rigid substrates, to study substrate topography effects on different cellular behaviours including migration [22][23][24][25][26]. Among the different material fabrication strategy, electrospinning could be considered as one of the most facile approaches which produce fibre structures mimicking the physical building blocks of fibrillar ECMs [27][28][29][30][31][32][33][34]. As linear tracks and straight fibre patterns have provided the foundation to recapitulate fibrillar morphology [35][36][37], it is of interest to evaluate how the features at a higher scale of fibre architecture, such as waviness, loops, and cross-junction grid patterns commonly found in ECM, could influence the cell migration dynamics.
This study adopted a low-voltage continuous electrospinning technique [38] to simplify typical ECM architectures into well-defined and repeatable patterns, and design cell migration assays consisting of individual polystyrene microfibre tracks of straight, wavy, looped or grid patterns. Microfibres (made of polystyrene) were deposited onto substrates (made of polystyrene coated glass slide), with the entire surfaces subjected to plasma treatment, thus cells had the freedom to explore different shapes in response to the well-defined fibril geometries. The combination of such fibre topography with ECM morpho-typical contours could potentially provide new insights into physically-guided cell migration behaviour, compared to established synthetic migration assays such as micro-patterned lanes [39] and channels [40]. Here, this work examines how different global fibre patterns influence the migration dynamics of green fluorescent protein (GFP) transfected MDA-MB-231, a human breast cancer adenocarcinoma cell line characterised by elevated SNX27 (a protein marker of cell mobility and proliferation) expression, and with the highest persistent 2D migration among all breast cancer cell lines [41][42][43][44].

Results and discussion
2.1. Global fibre pattern determines cell migration direction and cell shape, but not the speed Disease or age-associated extracellular matrix remodelling often leads to the creation of stiff collagen fibre bundles, which could impart biomechanical cues to cells with their rod-like nature (with several microns in diameter) and nano-scaled surface topography [45][46][47]. To create synthetic microfibres mimicking the physical attributes of individual stiff ECM fibre bundles, low-voltage continuous electrospinning patterning (LEP), developed in our previous work [38,48], was used to fabricate polystyrene fibres with required features. Made of polystyrene, such synthetic fibres could provide a chemically stable and robust framework to study cell migration. Figure 1(a) shows the basic fibre printing process. The fibres deposited on polystyrene-coated glass slides have an average diameter of 3 ± 1 µm. In figure 1(b), scanning electron images of single fibres show that the fibre has a circular cross-section and exhibits 'nano-wrinkle' surface features. In figure 1(c), the contact Young modulus of the fibre determined by atomic force microscopy (AFM) nanoindentation was shown to be around 200 MPa, within the range of contact moduli of collagen fibrils (∼0.1-10 GPa) measured by AFM with different buffer solutions [49,50]. Adjusting the stage collection speed and the application voltage, fibre contours that mimic morphological characteristics of the extracellular matrix, from straight fibres [51][52][53][54], to waves and loops [55][56][57][58][59][60], can be fabricated (figures 1(d)-(g)). To define the wavy fibre features, we fitted sinusoidal functions to the fibre contours. For loop fibres, as the loops were approximated circular (figure 1(e)), the equivalent loop diameters were determined to be between 20 and 40 µm in our study. Finally, for grid patterns, the inter-fibre distances were varied between 0 and 120 µm in orthogonal directions.
Imaging of cell migration experiments were processed with Z-stack projection of cells' GFP fluorescence (SI. figure 1). Results confirmed that each individual fibre had the capacity to guide cell migration (figure 2(a), supplementary movies 1-4 available online at stacks.iop.org/BF/14/024102/mmedia). Aiming at cell on-fibre migration, only cases of single cells on fibril topographies were identified and analysed for the duration. Depending on the fibre pattern, cells adhering to individual fibres, or two parallel fibre tracks, performed guided movement along fibre axes. Cells on a network of orthogonal crossed fibres migrated along either horizontal or vertical directions. Figure 2(b) compares the average step speed (calculated from the distance travelled within 10 min of the imaging interval) for single cell migration on different fibre patterns. The average step speed ranges between 0.2 and 0.5 µm min −1 on the four typical global fibre patterns, with no significant difference found between migration on straight aligned fibres and on other complex fibre patterns, or the off-fibre control group. With the finding above, it is suggested that the global fibre pattern affect the direction of cell movement, but not the individual cell step speed. Cell migration speed could be controlled by the intrinsic cell mobility, and the fibre local topographies such as fibre dimension/diameter and surface textures [61][62][63][64], which were kept constant in our experiments. Next, analysing single cell morphology in figure 2(c) revealed that as fibre patterns change from open linear to enclosed patterns, the cell aspect ratio increased to a maximum of 0.8 (where aspect ratio = 1 describes a perfect circular shape). Thus, the average shape of a cell is strongly influenced by the fibre pattern, which might have further implications for fibre-guided cell migration. The  Aspect ratio = minor axis length/major axis length; as each single cell was fitted with an ellipse that optimally approached the cell's boundary. Number of single cells studied: n = 62 cells on straight fibres; n = 90 cells on wavy fibres; n = 98 cells on grid fibres; n = 52 cells on loop fibres; n = 64 cells on no fibre substrates. All data presented are means ± SDs with superimposed data points. Two sample t-tests were performed, ns denotes no significant difference between groups, * * p < 0.01, * * * p < 0.001.
following sections analyse cell migration in each of the fibre pattern case in detail, and evaluate how the associated migration behaviour is correlated to the size and shape of the cells.

Migration dynamics along straight, parallel fibre patterns
For the straight fibres, cells' major axes aligned with the fibre axis, as seen in figure 3(a). Probability maps are used to illustrate the spatial distribution and activity of the cells over time. Live-cell images were captured at intervals of 10 min and overlapped into a 2D projection image, such that each image pixel numerically represents cell occupancy probability, with a value between 0 and 1. During single cell migration on straight fibres, their major axis was aligned with the fibre axis, and topography guidance provided by the polystyrene microfibres efficiently drove the cells to move along the fibre direction, as shown by figure 3(b). In some cases, a cell could stretch its body sideways and attempt to connect to multiple fibres during migration (bridging between parallel fibres); while in other cases, a cell remained solely on its original fibre track throughout the imaging period (see figures 3(c) and (d), supplementary videos 5 and 6). To evaluate the migration statistics for the single-fibre migration case, cell displacement along the fibre direction was calculated denotes the position vector (in 1D system only motion in one axis is considered) of the jth cell at time t, and t 0 represents the corresponding original starting time. The displacement vs. time plot for all the cells tracked for the straight fibre patterns, is shown in figure 3(e). The ensemble-averaged displacement trajectories of cells were in symmetry with respect to the zero horizontal line, supporting the equal likeliness for the cells to migrate towards either end of the fibre. Further categorisation of single-fibre and parallel fibre migration behaviours in figure 3(f) suggests that the spacing between parallel fibres (i.e. fibre interval) is a determining parameter. When the inter-fibre distance was smaller than 30 µm, there was no preferred choice for the cell behaviour (either remain on a single fibre or bridging to neighbouring fibres). However, when the fibre interval was increased to over 35 µm, cell bridging was no longer observed, and cells remained on their original linear fibre. Figure 3(f) shows that cell minor axis length all had values below ca. 18 µm. Thus, it is suggested that the distance of adjacent fibres is the characteristic dimension of the parallel fibres, which determines whether a cell is able to be diverted from its current fibre migration path. The length of the cell minor axis could be used as a measure of their ability to stretch their body orthogonal to the fibre direction, and reach an adjacent fibre for migration.

Migration dynamics along grid patterns
Looking over the 12 h live-cell imaging period, an interesting observation is that most cells are restricted to their local grid/boxed meshes, which is shown in the fluorescent microscopic image as well as the probability map in figure 4(a). The possible reason is that within the time of measurement, cells would inevitably come across orthogonal fibre joints, which appear as a local protuberant landscape to the cells. When a cell comes across an orthogonal fibre joint, its protrusions are able to sense orthogonal fibres, and the cell reacts accordingly to adapt to different fibril topography. The ultimate displacement of cells during the entire experimental session was calculated and displays in figure 4(b). Compared to cells on straight aligned fibres, this distribution shows that grid joints provide an extra degree of freedom to cell migration.
Next, the single cell migration was studied. Figure 4(c) illustrates three types of single cell behaviour that were observed when interacting with orthogonal fibre junctions: (i) bridged at joints, where a single cell would bridge to the adjacent parallel fibre track and then continue migrating along the two parallel tracks simultaneously; (ii) anchored at joints, where a single cell would remain anchored at these joints; and (iii) steered at joints, where a single cell would be redirected along one of the orthogonal fibre tracks. Figure 4(d) further demonstrates typical cases of the above scenarios in sequential fluorescent microscopic images, indicating general time durations (additional detailed examples are shown in supplementary videos 7-9). To characterise the migration behaviour, 'cell adhesion time' is defined as the total time that a cell is adhered at a fibre junction, from the initial point the cell body first touches the joint, until the moment that all parts of the cell body have lost contact with the joint. As shown in figure 4(e), when the normalisation of this parameter (cell adhesion time/total imaging time) is plotted against the adjacent fibre distance, the different cell behaviour falls into distinct categories. When the adjacent fibre distance is less than 30-35 µm, multiple cell behaviours was observed, though cell bridging dominates at narrow fibre spacing. As adjacent fibre distance becomes greater than ∼30-35 µm, cell bridging can no longer take place. For adjacent fibre distance above 70 µm, cells would then be redirected at these junctions to an orthogonal fibre track for continued migration.

Migration dynamics along wavy and loop fibre patterns
ECM fibrils often have curly geometries [58,60], with features similar to cell sizes, and such geometries can be categorised into two major groups: wavy and closed-loop patterns. For wavy fibre patterns, the curve can be described by a sinusoidal function of y = A · sin (2πx/λ), with A being the amplitude and λ the wavelength. Focusing on single cell-fibre interaction, the migration along a wavy fibre segment is illustrated in figure 5(a), where cells were observed to maintain contact with the fibre throughout the migration. A cell entering a wavy curve is defined to be the moment when the cell body firstly touch a trough/crest (where dy/dx = 0); and a cell leaving the wavy curve is defined as the moment when the cell body has fully passed over the curve trough/crest. Cell residence time represents the total   [10][11][12]. Cell residence times are also displayed alongside the probability maps, highlighting that as the waveform amplitude increased, coupled with increased curliness, cell residence seemed to be prolonged.
To quantify the curliness of wavy fibres, 'curve steepness' was characterised as the maximum derivative (|dy/dx| max ) for each of the sinusoidal segment. Thus figure 5(c) shows the effect of wavy fibre contour, characterised by wave height (2 × A) and steepness (|dy/dx| max ) on cell residence. When the wave height increased beyond ∼30-35 µm, along with the increased curve steepness, the normalised cell residence times increased rapidly. In some cases, the 'cell residence time/total time duration' ratio approached 1, indicating that these cells were trapped inside the curve throughout the entire imaging period. It should be noted that as the curve steepness is increased, there are still cases of cells having lower residence time. This is because cells were observed to enter fibre curves at different times during the imaging session, thus low residence values were recorded for cells entering towards the end of the imaging session.
The close-loop patterns demonstrated an efficient 'cell trapping' ability when the loop diameters are close to the cell size. Figure 6 [13][14][15]. For small loops with diameters of around 20 µm, cells tended to occupy the entire loop space, and having some of its cell body extending beyond the loop. For medium size loops with diameters of approximately 25-30 µm, cells would still be able to accommodate the majority of the loop space. In contrast, for large loops, with diameters greater than 35 µm, cells could only adhere to a portion of the loop boundary, and would accordingly only occupy smaller portions of the loop. Figure 6(c) presents the correlation between the normalized cell residence and the loop diameter. As the loop diameter increases from 20 to 35 µm, the cell residence gradually shifted up to around 70% of the time.

Discussion and conclusion
For all the global microfibre patterns studied in this work, MDA-MB-231 cells were able to execute fibreguided migrations, while adapting their morphologies to fit in local topographies as well as global fibre patterns. As shown in figure 7(a), our results point to the threshold feature length for aligned, wavy, loop Our live cell imaging results indicated highly dynamical nature of visible cellular protrusions, which were formed to probe surrounding topography, subsequently guide the displacement of the cell body (SI. figures 2 and 3). Imaging data showed that visible cell protrusion lengths could go far beyond the range of 30-35 µm, which were generated and afflicted to the fibre structures (e.g. SI. figure 3). In addition, our prior work showed that long protrusions could break from the main cell body to release the cell to travel away from the protrusion breakage fragments [65]. The above experimental observations suggest that the upper lengths of protrusions themselves is not a factor for crossing migration paths.
Therefore, this work examined the minor axis length and the major axis length of the cell bodies in all experiments, as plotted in figures 7(b) and (c). It is shown that the upper limit of the cell body's minor axis length, and the lower limit of cell body's major axis length, are both in the range of 30-35 µm. The 'coincidence' of these two values led us to postulate the following: A cell can have different movements to accommodate fibre patterns during migration; different fibre patterns will determine the motion as well as the shape of the cell body. However, chances for a cell to alter its pre-existing fibre migration path depend on its body's ability to stretch sideways and reach the adjacent fibre, and thus could be determined by the upper limit of its minor axis. For MDA-MB-231 cells which attached to an existing fibre, the maximum distance that they can reach and form stabilised adhesion on another fibre, over the preference of existing fibre alone, is in the range of 30-35 µm. Although experiments in this work were performed on fibres patterned on a planar (2D) substrate, it is postulated that the similar outcomes would be manifested in a 3D fibrous network. A plane is defined by two lines that intersect or are parallel; hence, the fibres on a 2D geometry represent a subsect that reflects the upper limit for track diversion in a 3D fibrous network.
To conclude, this work reports the systematic analysis of MDA-MB-231 cell migration dynamics along polystyrene fibres fabricated using low-voltage continuous electrospinning. The fabrication of fibre networks with straight, curly and grid patterns provided geometric environments to study how cells sense and interact with complex morphologies. Cells were able to dynamically adapt their shapes in response to the adhered fibre, as well as to the neighbouring patterns; such adaptation reflects their ability to traverse between fibre tracks. Key dynamical and morphological features such as the variation of cell step speed and minor axis were identified for understanding cell migration in fibril matrices, as well as designing future ECM-mimic fibril models for cell migration studies.

Coating glass slides
The glass slides (diameter: 22 mm) used for depositing polystyrene fibres were coated using a spincoater (Model WS-650MZ-23MPPB). For each glass slide, 70 µl 2 wt% polystyrene (Mw ∼ 280.000 kDA, Sigma-Aldrich) in toluene (anhydrous 99.8%, Sigma-Aldrich) solution was initially added at the centre. The glass slides were spun for 10 s at 500 rpm, followed by 50 s at 1000 rpm. Coated glass slides were then dried at reduced pressure for 24 h, following which they could be stored at room temperature for several weeks.

Fibre printing
Low-voltage continuous electrospinning patterning technology (LEP) was used to print polystyrene fibres on polystyrene coated glass slides. The setup consists of a supporting aluminium plate and silicon plate under the glass slide, aluminium initiators at both sides of the polystyrene coated glass slide, and a power source and syringe pump. Coated glass slides were placed between two aluminium initiators on a supporting aluminium plate for fibre deposition. The printing stage was heated up to 33.5 • C to ensure continuous jetting during printing. After the stage had settled at this temperature, a 1 ml syringe with a 25 G needle was filled with a 25 wt% polystyrene in dimethylformamide (DMF) (Sigma-Aldrich) solution for fibre printing, and a flow rate of 160 µl h −1 was set. During the printing process, the stage was initially lifted slowly until it had reached the needle of the syringe. The positive and negative poles of the voltage output device were connected to the needle and the stage respectively. Polystyrene fibres were directly written onto the coated substrate under an applied voltage ranging from 0.3 to 0.7 kV. The global fibre pattern shape can also be affected by the voltage level applied. A voltage of 0.5 kV marked the transition between straight and curly fibres for 25 wt% polystyrene/DMF mixture. To fabricate straight aligned fibre patterns, the working voltage was set at 0.3 kV. To fabricate all curly fibres, working voltage was further increased to 0.5-0.7 k V. To fabricate grid fibre patterns, a first layer of straight aligned fibres were initially deposited, followed by rotation of the substrate by 90 degrees to deposit a second layer of aligned fibres.

Atomic force microscopy (AFM) characterisation
In situ AFM nano-indentation was performed using a Nanowizard AFM (JPK Instruments, Germany). Measurements were taken in distilled water at room temperatures. The preparation procedure involved fixing the sample onto a piece of mica using araldite and inserting it into a liquid cell. The samples were immersed in distilled water for over 10 min prior to measurement. A Hertizian model [66] was employed to extrapolate the apparent contact stiffness and a silicon tip (NSC36, Mikromasch, frequency 100-200 kHz) was used for the measurement. For the in situ AFM, measurements were performed on polystyrene fibres deposited on a thin spin-coated PDMS (polymethyldioxane, modulus ∼500 kPa) film. This allows to show contrasting stiffness values between the fibre and the background in the stiffness mapping.

Fibre pre-treatment for cell migration
A series of pre-treatments were followed to make fibres on polystyrene-coated glass slides ready for cell adhesion and migration experiments. First, silicon rubber compound (RS 555-588) was used to fix the fibre-printed glass slide on a petri dish (diameter: 35 mm). This petri dish had a hole cut in the centre, using a machine in advance, and the fibre-printed glass slide was fixed right above this hole. After fixing, plasma treatment was applied with a flow rate of 25 m 3 s −1 for 45 s to provide polystyrene fibres with hydrophilic surface. Afterwards, the petri dish was sterilised under UV light for 30 min, following which the fibre-printed glass slide were immersed in 4 ml PBS for 24 h in the hood. Finally, PBS was replaced with 3 ml cell culture medium and the fibres were immersed for 48 h in a standard cell incubator; the immersing culture medium was changed every 24 h.

Cell culture
GFP-labeled MDA-MB-231 human breast cancer cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) supplemented with 1% L-glutamine (Invitrogen), 10% foetal bovine serum (FBS), 1% penicillin and 100 µg ml −1 streptomycin (Sigma Aldrich). To reduce fluorescence quenching during culture, aluminium foil was used to wrap the culture flasks. The culture medium was changed every 2 d, and cell passage was performed when 80% of the flask was covered by cells.

Live cell imaging
Live cell imaging was performed on 35 mm petri dishes with fibre-deposited, PS-coated glass slides. 30 000-40 000 cells were suspended in 2 ml cell culture medium, resulting in a cell density of 1.5-2 × 10 4 cells ml −1 to be deposited into the petridish for the single cell migration imaging. Higher cell seeding densities would potentially make more cells attach to fibre paths and eventually affect the motions of their neighbouring cells, while we tried to stay with a 'single-track' concept. An estimated period of 2-3 h was given in advance before the imaging initiated, in order for most of the cells to attach to fibres; and their migration dynamics were captured every 10.5 min in the time-lapse microscopy for a continuous 12 h time period. Z-stack imaging of the cells' GFP fluorescence were applied, since a single plane of confocal imaging does not provide sufficient imaging quality for cell shape segmentation. The cell Z-stack imaging and time-lapse experiments were performed using an environmentally controlled chamber joined to a Leica confocal microscope (Leica TCS SP5). The heater was set to 37 • C to heat up the stage top incubator, and sterilised water and CO 2 were used to provide a suitable moisture and pH inside the box-type incubator. During imaging, excitation was set at 480 nm and detection was set at around 500 nm to observe the GFP signal, and the bright field signal was also collected. For cell migrations along straight aligned fibres, experiments were conducted with N = 7 replicates; 4-16 single cells were observed in each experimental replica, resulting a total of 62 single cells analysed. For cell migrations along grid fibres, experiments were conducted with N = 9 replicates; 3-23 singles cells were observed in each experimental replica, resulting in a total of 98 single cells analysed. For cell migrations along wavy fibres, experiments were conducted with N = 4 replicates; 15-34 single cells were observed in each experimental replica, resulting in a total of 90 single cells analysed. For cell migrations along loop fibres, experiments were conducted with N = 4 replicates; 7-23 single cells were observed in each experimental replica, resulting in a total of 52 single cells analysed. For the control groups of cell migrations along no-fibre substrates, experiments were conducted with N = 3 replicates; 5-50 single cells were observed in each experimental replica, resulting in a total of 64 single cells analysed.

Post-processing and statistical analysis
For the image analysis, an image analysis program written in-house was used to pre-process the livecell images and to create a database to classify the microenvironment-based interaction automatically. The program incorporates a code to analyse the z-stack images, and a number of open-source applications, including ImageJ plug-ins and CellProfiler pipelines, in an adapted procedure following our prior study [67]. The GFP intensity of the z-stack imaging acquired during the time-lapse experiments was projected into one x-y plane at each time point; thus the projected cell area, that is the 2D projection of the 3D cell shape, was obtained at each time point. The Manual Tracking plugin of ImageJ, applied to the projected GFP sequence, was used to track the position of the cancer cell nuclei over time. Single cells that were long-lasting in the acquisition view, kept moving along single a fibre and were distanced from other cells, were generally chosen for tracking. Homewritten Matlab codes were used for most analysis of cell migration. The image analysis software CellProfiler was applied to perform the segmentation of the cancer cell projected area and to extract the cell shape features in the different microenvironments at each time point. Cells were labelled with unique ID, and cell shapes were recognised and segmented from the microscopic images. In terms of single cell migration studies, only single cells with good shape recognitions were collected; cells with defective shape recognitions were excluded. The cancer cell coordinates were corrected to take into account the fluctuations of the microscope stage that occurred during experiments.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.