Nanofiber curvature with Rho GTPase activity increases mouse embryonic fibroblast random migration velocity

Abstract

Mechanotransduction arises from information encoded in the shape of materials such as curvature. It induces activation of small GTPase signaling affecting cell phenotypes including differentiation. We carried out a set of preliminary experiments to test the hypothesis that curvature (1/radius) would also affect cell motility due to signal pathway crosstalk. High molecular weight poly (methyl methacrylate) straight nanofibers were electrospun with curvature ranging from 41 to 1 μm⁻¹ and collected on a passivated glass substrate. The fiber curvature increased mouse mesenchymal stem cell aspect ratio (P < 0.02) and decreased cell area (P < 0.01). Despite little effect on some motility patterns such as polarity and persistence, we found selected fiber curvatures can increase normalized random fibroblastic mouse embryonic cell (MEF) migration velocity close to 2.5 times compared with a flat surface (P < 0.001). A maximum in the velocity curve occurred near 2.5 μm⁻¹ and may vary with the time since initiation of attachment to the surface (range of 0–20 h). In the middle range of fiber curvatures, the relative relationship to curvature was similar regardless of treatment with Rho-kinase inhibitor (Y27632) or cdc42 inhibitor (ML141), although it was decreased on most curvatures (P < 0.05). However, below a critical curvature threshold MEFs may not be able to distinguish shallow curvature from a flat surface, while still being affected by contact guidance. The preliminary data in this manuscript suggested the large low curvature fibers were interpreted in a manner similar to a non-curved surface. Thus, curvature is a biomaterial construct design parameter that should be considered when specific biological responses are desired.

Statement of integration, innovation, and insight 

Replacement of damaged or diseased tissues that cannot otherwise regenerate is transforming modern medicine. However, the extent to which we can rationally design materials to affect cellular outcomes remains low. Knowing the effect of material stiffness and diameter on stem cell differentiation, we investigated cell migration and signaling on fibrous scaffolds. By investigating diameters across orders of magnitude (50–2000 nm), we identified a velocity maximum of ~800 nm. Furthermore, the results suggest large fibers may not be interpreted by single cells as a curved surface. This work presents insight into the design of constructs for engineering tissues.

INTRODUCTION

Cell migration is an essential function for multicellular organism development and homeostasis, as well as regenerative engineering. The extracellular matrix (ECM) provides physical structure leading to the complex signaling cascades that give rise to cell motility through adhesion and contraction. Spatiotemporal initiation of membrane protrusion and retraction events is controlled in part by small GTPases. Rho GTPases include RhoA, known to be involved in cellular functions such as ruffling [1], focal adhesion assembly [2], and contractility [3], whereas cdc42 is associated with cell spreading [4], and filopodia [5]. We hypothesized that there would be differential Rho GTPase involvement in migration on synthetic ECM mimetic scaffolds with controlled curvatures.

Mesenchymal migration, exhibited by diverse cell types including fibroblasts, has been observed on flat substrates for many years. However, the in vivo environment is rarely flat or as stiff as a petri dish. Variation of substrate stiffness affects cellular function including differentiation [6], as demonstrated by Engler et al. using a variable stiffness hydrogel [7]. In addition to differentiation [8], the mechanosensitive machinery responds to material shape affecting migration [9], cell morphology [10], and cell type specific functions [11–14]. Both natural and synthetic materials can act as in vitro substrates to understand bioinstructive cell-matrix interactions [15, 16], with some being better at isolating single parameters than others [17].

Micro- and nano-fibers have been used to mimic the information rich ECM. These can be created with a variety of methods including phase separation [18] and electrospinning [19]. Nanofiber membranes find biomedical uses as artificial basement membranes, encapsulation membranes [20, 21], hard and soft tissue reconstruction structures [22, 23], and even a twisted suture [24]. Without additional process modifications, electrospun fibers collected on a flat substrate most closely mimic the basement membrane ECM. They are planar, have features in the range of ECM proteins, are highly porous, and scaffold cells in three dimensions (3D). However, studies of nanofiber based cell migration are few despite the fact that our understanding of scaffold design and 3D cell migration could be improved.

Although parameters in the fabrication process can be used to control aspects of the fiber network, an incomplete knowledgebase regarding how the fiber curvature affects cellular behaviors limits the ability to effectively use this capability. In the context of tissue engineering, one can imagine quantifiable outcomes (such as migration, differentiation, proliferation, etc.) that inform priorities for a nanofiber scaffold design. For instance, if a scaffold will be implanted and host cells should populate the scaffold to generate the desired tissue, stimulation of migration into the scaffold would take a higher priority than if cells will be seeded prior to implantation. In the later case, encouraging the cells to stay on the scaffold (low migration), while supporting their viability and differentiation, would likely be the prudent approach.

Some existing studies examine the effect of fiber curvature on proliferation and differentiation. Christopherson et al. studied polyethersulfone nanofibers of curvatures 7.07 ± 44.44, 2.67 ± 13.07, and 1.38 ± 6.41 μm⁻¹ finding proliferation of rat neural stem cells decreased with curvature [25]. In a study of human mesenchymal stem cells on PEOT/PBT (poly (ethylene oxide terephthalate)-poly (butylene terephthalate)) fibers, Moroni et al. found that more than curvature (< 2 μm⁻¹) of the fibers, the porosity of the individual fibers had a greater effect on proliferation [26]. This suggests that smaller physical sizes are required. Indeed Yang et al. found that 6.67 μm⁻¹ fibers promoted neural differentiation better than 1.33 μm⁻¹ counterparts [27]. Nanofibers provide a powerful controllable substrate on which to study phenomena that arise from 3D features.

In this work, control of curvature was accomplished by varying the concentration of polymer in the electrospinning solution, without changing the composition of the fiber or the molecular weight of the polymer. Here, we only examined fiber diameters in a subcellular size range, not considering multiple micron fibers, which stimulated a different regime of responses in epithelial tissue phenotypes [28]. The fiber size range we chose encompasses natural matrix fiber sizes of ~100 and 800 nm diameter fibers that have been identified for collagen and elastin based substrates as well as encompassing diameters that are used in tissue engineering studies [29–32]. Straight fibers with a predominate alignment were fabricated, which have been shown in multiple reports to increase migration velocity compared with randomly oriented fibers [33–36]. Being straight, the fibers presented the cells with positive out-of-plane curvature, as opposed to a fiber that is not straight (has a tortuous path from one point to another). Furthermore, the number of fiber intersections was minimized by increasing alignment, reducing the complexity of data interpretation [37]. Using this system, we observed motility parameters of cells migrating on fibers with curvatures across an order of magnitude during continuous extended time windows with small GTPase inhibition.

MATERIALS AND METHODS

Substrate preparation

Electrospun nanofiber scaffolds were fabricated from high molecular weight polymethylmethacrylate (PMMA; HMW–PMMA, 996 kDa) dissolved in dimethylformamide:tetrahydrofuran (DMF:THF) (60:40) and collected on a target using a 15–20 cm working distance, ~15 kV driving voltage, 0.7–2.5 ml/h flow rate, and an 18 or 30 gauge needle (adjustments used to control quality of fibers). Fibers were collected on glass coverslips that were first spin-coated with a layer of cell adhesion resistant poly (2-hydroxyethyl methacrylate; pHEMA). A prolonged electrospinning time was used for the smallest fibers to overcome embedding in the pHEMA layer. Coverslips spin-coated with HMW–PMMA were used as flat surface controls.

Collector electric field manipulation created straight fibers with a predominate orientation. In some experiments (Field Aligned Fibers), fiber alignment was created by two conductive copper bars positioned ~12-mm apart and 22 × 22-mm pHEMA coated coverslips were affixed in front of this for fiber collection (Fig. 1A). The area in the center with the greatest alignment was used for imaging. In a similar setup, a continuous copper plate was covered with paraffin wax except for two exposed areas on either side of an area that was ~22-mm wide (Fig. 1B).

Figure 1.

Nanofiber substrate fabrication by electrospinning. Two collector configurations utilized to produce Field Aligned Fibers (A, B) and Field and Kinetic Aligned Fibers (C, D). Center area shows relative scale cross sections of the fiber sets interleaved. (E) Example sequence of images showing the manually identified center of mass of each cell at each time point.

To increase production rate, reproducibility between samples, and degree of orientation along an axis; a custom designed rotating collector was designed to take advantage of electric field and kinetic alignment forces (Field and Kinetic Aligned Fibers). The collector structure was 3D printed on a SeeMeCNC ROCKSTOCK MAX V2 printer (STereoLithography file available upon request; Fig. 1C and D). Copper plates affixed at intervals on the circumference were linked together by conductive wires and then connected through a disc and brush system to the grounding lead on the high voltage power supply. The rotation was powered by a variable speed mill equipped with a chuck. Approximately 8–10 22 × 22-mm square coverslips were collected at a time and could be repeated by simply removing coverslips and affixing new, minimizing the effect of day-to-day humidity and temperature variation.

The concentration was the driving force for curvature control where 1.54, 3.1, 3.7, 6, and 8.3% HMW–PMMA by weight was used for curvatures of 41.41, 10.45, 8.36, 4.56, and 1.65 μm⁻¹, respectively, corresponding to diameters 48, 191, 239, 438, and 1212 nm, respectively (Field Aligned Fibers, Supplemental Table 1, see online supplementary material) and 3, 6.5, 7.5, 10, and 12% HMW–PMMA by weight was used to fabricate scaffolds with target curvatures of 10, 4, 2.5, 1.333, and 1 μm⁻¹, respectively, corresponding to diameters of 200, 500, 800, 1500, and 2000, respectively (Field and Kinetic Aligned Fibers, Supplemental Table 1, see online supplementary material).

A scanning electron microscope was used at the Materials Characterization Lab in the Penn State Materials Research Institute. Samples were sputter coated and imaged using the parameters displayed in the information bar at the bottom of each panel including 5 keV, 56pA, and a working distance ranging from 3.2 to 3.5 mm. Diameter measurements were made on the images using ImageJ (National Institutes of Health, USA). Linear and exponential fits to the Field Aligned Fiber curvatures were used to guide selection of concentrations for Field and Kinetic Aligned Fibers. Embedded in each figure, notations of the fiber set used have been made.

Cell culture

Mouse embryonic fibroblasts (MEFs) were cultured in DMEM (Life Technologies 11 965-092) supplemented with 15% Fetal Bovine Serum (FBS, Atlanta Biologicals, Flowery Branch, GA) and 1% Penicillin/Streptomycin (P/S, Corning, Corning, NY, USA, Part#: 30-002-CI). MEFs were used for migration studies between passage 7 and 25. GFP positive mouse mesenchymal stem cells (mMSCs) were isolated from a GFP positive C57BL/6 mouse and were a kind gift from Professor Daniel Hayes. mMSCs were also cultured in DMEM, but with 10% FBS and 1% P/S. All cells were cultured at 37°C in a humidified incubator with 5% CO2. Substrates were placed under the ultraviolet (UV) lamp in the biosafety cabinet for 20 min to deactivate microbial contamination. Cells were then seeded onto the substrates at ~5 × 10⁴ cells per construct (22 × 22-mm) for imaging.

Live-cell imaging

Cells were time lapse imaged in supplemented FluoroBrite DMEM (ThermoFisher) media on an upright microscope. Cells and scaffolds were held in P30 culture dishes custom prepared with a glass coverslip attached to the dish covering an 18-mm diameter circular hole cut using a heated metal cork borer. The plate was sealed with UV activated optically clear Norland Optical Adhesive 68. The 22 × 22-mm coverslip was mounted either in or to the bottom of the dish depending on the final use and dish size. A heated stage with a layer of white mineral oil over the media was used. Images were taken at intervals of 90 s (unless noted otherwise) with differential interference contrast (DIC) illumination using a 10X long-working-distance water dipping objective.

Live-cell imaging was done in FluoroBrite DMEM (Life Technologies A18967-01) supplemented with 15% FBS, 1% Glutamax (Life Technologies 35 050-061), and 25 ul/ml HEPES (AMRESCO J848). Cells were allowed to attach in the humidified incubator prior to the collection of time-lapse images. For each set of experiments (i.e. one of each curvature and a flat control: six conditions), cells were first divided at decreasing densities into P60 dishes, from which cells would be lifted and plated at consistent times before that sample would be imaged to control for passage and density effects on migration speed. Unless noted otherwise, ~12-h post seeding in normal growth media, media was changed to the live-cell imaging media. The plate was then placed on the heated stage warmer (BioPTECH, Butler, PA, USA), with inhibitors added at this time if applicable. Contact was formed with the water dipping objective followed by application of a layer of white light mineral oil (AMRESCO PN J217) to prevent evaporation.

Motility quantification

DIC images for cell migration velocity measurements were collected with a 10X objective at 90-s intervals for ~12 or 24 h. During post-analysis, a 6-h time window was selected with a starting point that ensured the time between plating and the quantified window was consistent for the set of acquisition. Cells were selected to reduce confounding factors including: migration out of the field of view, interaction with other cells, detachment from nanofibers, and cell division. Cells were tracked with the Manual Tracking ImageJ plugin (NIH, USA). Because automated cell tracking techniques do not work well for cells that are not labeled, manual mouse clicks were positioned to estimate the center of mass at each time point for each cell in their unmanipulated state, allowing for frequent imaging over a long period (Fig. 1E). Most migration data were presented after standardizing to the flat surface migration velocities for that experimental set. Therefore, if one batch of cells exhibited a higher migration activity than another, the effect would have been equalized in the final analysis. Graphs were presented at a log base 2 scale so that a distance above 1 is the same interpretation as the equal distance below 1 (i.e. 4-fold and 0.25-fold). Migration velocity was calculated directly from the Manual Tracking ImageJ plugin output. MSD and velocity magnitude polarization profiles were calculated by plugging the coordinates of each tracked cell to the MATLAB scripts provided by Wu et al. as a toolkit to analyze in an anisotropic persistent random walk (APRW) model [38]. Individual cell MSD plots can be found in Supplementary Figs 5 through 9, see online supplementary material for a color version of these figures. Cell track plots were constructed from the raw data using a custom MATLAB script (Supplementary Information). Cell persistence was calculated using a custom Python (3.7.6) script written with the Math (www.python.org), SciPy [39], NumPy [40], and Pandas [41] libraries. The script read in the XY coordinates of each cell from a comma-separated values (CSV) file, calculated the angle between trajectories at each time step, recording the timesteps at which the trajectory changed by more than π/2 from the previous one. Persistence time for each cell was calculated by converting timesteps to minutes.

In a post-analysis, the time-lapse videos were reviewed noting the frames where the number of fibers the cell-of-interest was touching or the number of nearby cells the cell-of-interest was contacting changed. Time weighted averages for each cell for selected conditions were then plotted against the normalized velocity to check for a correlation. Only those fiber curvatures that could be resolved in the images were quantified (< 5 μm⁻¹) for the 6-h migration windows on both fiber sets. The Pearson correlations were not significant for these data.

Cell morphology quantification

GFP positive mouse derived stromal/stem cells (Obatala, New Orleans, LA, USA) were allowed to attach and migrate on the substrate for a period of 18–22 h before being fixed and mounted to prepare for imaging. Briefly, the cells were washed twice with phosphate buffered saline, fixed in 4% paraformaldehyde in phosphate buffered saline (PBS), rinsed in PBS, and then mounted to a glass slide using anti-fade mounting media. Images were taken using a 40X water dipping objective with a GFP filter set.

Cell morphology was quantified using CellProfiler (www.cellprofiler.org) open-source software [42], which allows the same ‘pipeline’ of processing modules to be applied to a large group of images in an automated fashion. The output of CellProfiler can include (based on the images input) information about the cell, nucleus, focal adhesion, shape, size, intensity, association, and other morphological parameters.

Statistics

Raw data were organized and mathematical transformations performed in Microsoft Excel or MATLAB, whereas statistical analysis and graphing was conducted in GraphPad Prism (v8, San Diego, CA, USA). The raw velocity data (μm/h) were compared against the normal distribution using four tests (Anderson-Darling, D’Agostino & Pearson, Shapiro–Wilk, and Kolmogorov–Smirnov tests), none of which were able to confirm the null hypothesis (the distribution was not different than the normal distribution), meaning that the sampled data are not likely from a normal distribution. In most cases the null hypothesis was rejected, while in a few cases a test failed, but another test rejected the null hypothesis for that data set. However, parametric statistical tests were utilized to examine the data because tests including the analysis of variance (ANOVA) test are considered robust to data that is not highly normal [43]. Therefore, Brown-Forsythe and Welch ANOVA tests with Dunnett’s T3 multiple comparisons test, where individual variances were computed for each comparison, were applied for cases where only curvatures were compared. For experiments where different timepoints or treatment with inhibitors were compared, the ordinary two-way ANOVA was used (sphericity not assumed) with Tukey’s multiple comparisons test applied with individual variances computed for each comparison. For cell morphology measurements, single one-way Brown-Forsythe and Welch ANOVA tests with Dunnett’s T3 multiple comparisons test, where individual variances were computed for each comparison, were run for the aspect ratio, the mean cell intensity and the cell area separately. The major or minor axis was not compared since the aspect ratio was derived from the ratio of the major and minor axis. Tables of ANOVA comparison results can be found in Supplementary Tables 2–6, see online supplementary material.

RESULTS

Electrospun nanofiber curvature was varied between 1 and 41.41 μm⁻¹ (or 2000 and 48.3 nm in diameter), using two overlapping sets of nanofibers (Fig. 1, Supplemental Table 1, see online supplementary material). One set varied between 1.65 and 41.41 μm⁻¹ (1212 and 48.3 nm, Field Aligned Fibers, Fig. 2A and B), whereas the other ranged between 1 and 10 μm⁻¹ (2000 and 200 nm, Field and Kinetic Aligned Fibers, Fig. 2B, Supplemental Fig. 1, see online supplementary material for a color version of this figure). Results were normalized and presented on a logarithmic scale (log 2) in order to make experiments comparable. Log base 2 was chosen so that the distance above and below 1 (flat, zero curvature) represents the same increase or decrease fold change (double and half are the same distance away from 1).

Figure 2.

Nanofiber curvature is controlled by polymer concentration and migration velocity varies with time attached to substrate. (A) Scanning electron micrographs of five PMMA concentrations with corresponding mean diameters shown. (B) Plot showing the inverse relationship between radius and curvature. Inset: Positive relationship between concentration and diameter of electrospun fibers. (C) Relative migration velocity of MEFs migrating on Field Aligned Fiber substrates during a 6–12 and 13–19-h post-seeding to substrate window. +/− SEM shown, n = 5 cells per condition, adjusted P-values: ^*P < 0.01. (D, E) Absolute migration velocity of 6–12 h (D) and 13–19 h (E) migration windows. Note: smallest fibers were not included in this analysis because cells did not attach to the 41.41 μm⁻¹ substrate for the 6–12-h experiment.

In a preliminary experiment using Field Aligned Fibers, MEF migration velocity followed an opposite trend verses nanofiber curvature when 6–12 and 13–19-h windows were compared from separate experiments (Fig. 2C). These migration velocities were normalized to the flat surface shown on the left-hand dependent axis, with the correlating μm/h mean velocities shown by the lower curves and plotted on the right-hand dependent axis (Fig. 2C), and with individual mean values for each cell shown in Fig. 2D and E. Although many possible explanations for the differential velocity trend exist, one observation was that the degree of spreading on the fiber substrate appeared to be greater in the 13–19-h window (Video 1, 2). Since the 13–19-h window produced the higher mean velocities, we selected similar time windows so treatments would have a measurable effect.

To investigate the time dependent migration velocity phenomena further, a new set of MEF cells were monitored over a 20-h migration period on Field and Kinetic Aligned Fibers. The MSD and polarity were similar between fiber curvatures (Fig. 3A, Supplemental Fig. 2A, see online supplementary material for a color version of this figure), while the persistence time was greatest on the largest fibers (Fig. 3B). During the 20-h window as a whole, four fiber curvatures, excluding the smallest, displayed an increased mean relative velocity compared with the flat surface control (Fig. 3C and D  P < 0.01). Quantified windows of 2.5 h in duration demonstrated that the relative effect of curvature tended to vary slightly as cells were on the substrate for increasing amounts of time (Fig. 3E). The variability trended the greatest on the 2.5 μm⁻¹ fibers (Fig. 3F). Although many factors may affect the propensity of a cell to migrate, as time since attachment to a new substrate elapses, a few possibilities seem more plausible. (i) Recovery of surface proteins may not be complete for many hours following proteolytic cell lifting procedures [44]. (ii) Cell area tends to increase when attached to a relatively stiff substrate that is sparsely seeded [45, 46]. (iii) Cellular activity (measured as protrusion or bleb activity) tends to increase with time after seeding to a dish [47]. In the case of PMMA polymer nanofibers deposited on a glass substrate, sufficient rigidity exists to resist traction forces applied by the cells and therefore the cells spread.

Figure 3.

Curvature-velocity correlation varies with time. (A) MSD for MEFs migrating during a continuous 0–20-h experiment on Field and Kinetic Aligned Fibers. (B) Mean persistence shown in minutes. P-values indicate the multiple comparison corrected significance. (C) Mean cell velocities for MEFs migrating during the entire 20-h window. (D) Absolute migration velocities for each fiber size. (E) 2.5-h windows extracted from the 0–20-h block (black region in each stopwatch shaped chart covers the 2.5-h window that is averaged in the graph). (F) Mean MEF velocities grouped by fiber size to visualize any effect of time on a single size (relative scale cross-sections of fibers). +/− SEM shown, n = 5 cells per condition, adjusted P-values: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001.

Regardless of the dominate mechanism, the time window for analysis was clearly important. The time window for motility analysis in this paper was selected to increase the time in the spread state. Not including earlier time points where the cell would be initially transitioning to a spread state reduced the influence of pre-spreading attachment activity. During the first hour of introduction to a new substrate the cell settles and deforms in a predominately passive process that can be characterized by mechanical deformation of the actin cortex [48], followed by the initiation of molecular events that begin an active spreading phase [49].

The spreading process is also reset during cell division. Hirai et al. showed that NIH3T3 fibroblasts tended to increase in projected area between 4.8 and 9.5 h after trypsinization and before the first cell division [50]. Following this logic, tracking cell motility in a 12–24-h window should be after most cells have divided once and are in the spreading period once again [51]. As noted in the methods, cells that divide during the observation period (usually 6 h) were not selected for tracking.

To begin to understand why curvature affects aspects of cell motility, cell morphology was examined in fiber attached cells by image analysis of GFP labeled cytoplasm in mouse stromal/stem cells harvested from GFP mice. In our hands, mMSCs spread more on adhesive surfaces than MEFs and were thus selected for these measurements. The minor axis of nanofiber attached cells averaged lower than the flat surface controls (Fig. 4A). Because the major axis length did not vary noticeably, the cell aspect ratio tended to increase compared with the flat surface and was significantly higher on the 1.33 μm⁻¹ curvature fibers compared with the flat and 10 μm⁻¹ curvature fibers (Fig. 4A). The mean fluorescence intensity in the widefield image (an approximation of cytoplasm thickness or cell height) tended to be larger than the flat surface control at a similar proportion as the decrease in cell area (Fig. 4B and C, darker areas in images correspond to greater intensity of GFP fluorescence). All fiber sizes encouraged significantly lower cell areas compared with the flat surface, while all but the 4 μm⁻¹ fibers promoted a higher mean fluorescence intensity (Fig. 4B).

Figure 4.

Cells adopt an elongated shape on straight nanofibers. (A) The minor and major axis as well as the calculated aspect ratio of fixed mouse derived stromal / stem cells constitutively expressing GFP plotted against fiber curvature using Field and Kinetic Aligned Fibers. (B) The projected cell area and intensity of GFP from attached cells. (C) Representative widefield images of the cells quantified in panels A and B, showing different shapes on the straight fiberous substrates. Darker areas indicate greater intensity of GFP (imaged with a ×40 objective), suggesting thicker cytoplasm. +/− SD shown, n = 7, 5, 9, 10, 7, 10 for curvatures 10, 4, 2.5, 1.333, 1 and 0 (flat) μm⁻¹, respectively, adjusted P-values: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, and ^****P < 0.0001. When no line is drawn (panel B) symbols denote comparison with the flat surface control (zero curvature). Note: statistical tests were not run on the major and minor axis measurements, as these were used to calculate the aspect ratio.

Cells tend to form bridges between fibers, which has been studied by others [52] and we have observed this here as well as in live cells moving away from an acute angle fiber intersection (data not shown). A cell stretching further apart would experience increasing tension at the leading edge perhaps similar to durotaxis observed in cells migrating up a stiffness gradient [53–55]. Since cells with a larger comparative area also experience greater actinomyosin tension [56–58], we next wondered how the small GTPase pathways involving RhoA and cdc42 affect cell motility in the context of fiber curvature.

Inhibition of Rho-kinases (ROCK) with Y27632 applied to MEFs on Field Aligned Fibers decreased relative velocities on nanofiber substrates with different curvatures (Fig. 5A–C). Interestingly, the decrease in velocity under inhibitor treatment was relatively uniform across the range of nanofiber curvatures, with a noticeable difference being the significant change at 4.56 μm⁻¹ curvature. This suggested that migration velocity dependence on curvature was linked to RhoA activity, as inhibition on 4.56, 8.36, and 10.45 μm⁻¹ fibers significantly reduced relative velocity (Fig. 5A, D, and E, Supplemental Fig. 3, Video 2, 3, see online supplementary material), but did not affect persistence (Fig. 5F and G). The trends were generally consistent across the 6 h window, seldom reducing the mean velocity to less than that of the control MEFs migrating on a flat surface (Fig. 5C).

Figure 5.

RhoA contributes to cell migration on curved substrates. (A, B) Fibroblastic migration velocities verses curvature during 6-h imaging windows (hours post seeding were 13–19, 17–23 for MEF, MEF Y27632, respectively). (C) This effect is shown for 1-h windows (black region in each stopwatch shaped chart covers the 2.5-h window that was averaged in the graph). All groups (treated and untreated) were normalized to the untreated flat group that displayed a velocity of 72.13 μm/h. Absolute migration velocities (D, E) and persistence time (F, G) shown for indicated fiber curvatures. +/− SEM shown, n = 5 cells per condition, adjusted P-values: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, and ^****P < 0.0001 by two-way ANOVA.

On Field and Kinetic Aligned Fibers we investigated this further by using ML141 to inhibit cdc42 activity in comparison with Y27632 in MEFs, which did not cause noticeable differences in the MSD curve shapes (Fig. 6A–C) or polarity (Supplemental Fig. 2B–D, see online supplementary material for a color version of this figure). The absolute velocity was increased on the 4 μm⁻¹ fibers, whereas under inhibition with Y27632 no differences were observed (Fig. 6D and E) and treatment with ML141 caused a decrease in velocity on the 2.5 μm⁻¹ fibers (Fig. 6F). The persistence time was only statistically increased on the 2.5 μm⁻¹ fibers under treatment with ML141 (Fig. G, H, and I).

Figure 6.

Small GTPases contribute to cell motility on curved substrates. (A–C) MSD of MEFs on Field and Kinetic Aligned Fibers with or without inhibition of RhoA (Y27632) or cdc42 (ML141). Absolute migration velocities (D–F) and persistence time (G–I) shown for indicated fiber curvatures. +/− SEM shown, n = 5 cells per condition, adjusted P-values shown.

Similar to the Field Aligned Fibers, a trend of decreased relative velocity on fibers treated with Y27632 occurred within a middle curvature range (1.33–4 μm⁻¹, Fig. 7A, Video 4–6), but this only produced a significant difference on the same fiber size for ROCK inhibition on the 4 μm⁻¹ fibers, whereas multiple significant cross fiber size and inhibition comparisons existed (Supplemental Fig. 4A–C, see online supplementary material for a color version of this figure). This decreased velocity trend also occurred with ML141 treatment (Fig. 7A). Unexpectedly however, at curvature extremes this trend did not hold. On the smallest fibers, only Y27632 produced a decreasing trend. The 4 μm⁻¹ untreated fibers stimulated a highly significant velocity compared with untreated 1 μm⁻¹ and Y27632 inhibited 2.5 μm⁻¹ fibers (Fig. 7A and B, Supplemental Fig. 4A–C, see online supplementary material for a color version of this figure). Of note, many comparisons between the inhibitor treated groups were noted as significant including a decrease on the 4 μm⁻¹ fibers with ROCK inhibition (P < 0.01, Supplemental Fig. 4A, see online supplementary material for a color version of this figure). Also many statistical differences that crossed inhibition and fiber size existed (Supplemental Fig. 4A–C, see online supplementary material for a color version of this figure), underscoring the effects of small GTPases in curvature driven cell motility. With respect to duration of migration, flat surface migrating cells consistently displayed a higher velocity when treated with an inhibitor (Fig. 7C). However, it was not until 17 h that the smallest curvature fibers took on relative trends closer to the flat surface. In contrast, at most time points the fibers with curvatures greater than 1.33 μm⁻¹ followed the average trends. Thus, we suggest that cells lose some ability to differentiate a curved surface from a flat surface when the fiber has a sufficiently small curvature.

Figure 7.

Small GTPases contribute to cell migration velocity on curved substrates. (A, B) Trends and migration traces for untreated and inhibitor treated MEFs. (C) Effect of inhibitors throughout the course of the 6-h window. Each panel represents the indicated 1-h window (black region in each clock shaped chart covers the 1-h window that was averaged in the graph, all adjusted P values >0.5). All groups (treated and untreated) were normalized to the untreated flat group that displayed a velocity of 64 μm/hr. +/− SEM shown, n = 5 cells per condition, adjusted P-values: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, and ^****P < 0.0001 by two-way ANOVA.

DISCUSSION

Rational design of tissue engineering scaffolds requires knowledge of cell migration. Since natural material scaffolds often present difficulties in isolating single variables, synthetic materials have a role in basic biological studies as well as direct tissue engineering applications. We utilized electrospinning to control fiber curvature across orders of magnitude (1–40 μm⁻¹) and down to length scales matching key migratory protein complexes such as focal adhesions. In our model, fibril curvature increases random fibroblastic cell migration velocity. Although the relationship with curvature during consecutive 2.5-h windows did vary over the course of the 20-h migration window, these were not significant. Overall, it was in the middle range of curvatures that we found the greatest stimulatory effect of fibrous substrates on MEF migration velocity. Sharma et al. observed an increase in nanofiber based migration speed over the flat surface control [59], similar to our results, suggesting cell-attachment to a curved substrate alters the migration control mechanisms. Also in agreement with our work, Qu et al. studied MSC attachment and migration on fibers of three curvatures prepared from tussah silk fibroin compared against a flat surface poly-l-lysine control. All three fiber sizes stimulated migration faster than the flat surface, with the peak being on the 5 μm⁻¹ curvature fibers compared with 2.5 and 1.67 μm⁻¹ fibers [33].

Our increased velocity peaked at ~2.5 μm⁻¹ (800-nm diameter) and decreased as the fibers became larger or smaller. These preliminary studies agree with the 882-nm diameter maximum in p38 signaling we found previously in preosteoblast cells [60], suggesting conserved curvature sensing mechanisms. Interestingly, activation of p38 stimulates migration in diverse cell types including endothelial cells [61], neurons [62], and tumor cells [63, 64]. MEFs on fibers often migrated twice as fast as the flat surface controls, except when the fiber curvature was sufficiently small (approaching zero curvature, or large diameter) where we observed a pattern similar to a flat surface. This suggests the scale of curvature relative to the molecular sensing apparatus is important.

As part of our work to understand how this velocity trend arises, we observed significant decreases in projected cell area of cytoplasmic GFP expressing cells cultured on nanofibers of all curvatures in Field and Kinetic Aligned Fibers. A decrease in cell width, not length, appeared to drive the cell area decrease. Badami et al. noted a decreased projected cell area on fibers compared with TCPS and spin-coated surfaces [65]. Looking at a wide range of fibers (0.2, 0.5, and 4 μm⁻¹) Tian et al. also found an increase in cell length on smaller fibers [66]. These findings support our hypothesis that as fiber curvature decreases, cells are less likely to display phenotypes characteristic of fibrous substrates.

Since the purpose of our studies was to determine effects of curvature, we constructed the substrate to isolate that factor. The electrospinning method was based on work by Khanlou et al. [67] Because the effect of fiber density was noted with random fibers (Supplemental Fig. 10, see online supplementary material for a color version of this figure) as being partially related to cell residence time at fiber intersections, straight fibers with a predominate alignment were utilized to reduce these intersections. Furthermore, we did not find a correlation between normalized migration velocity and number of fibers the quantified cells were attached to during migration or the number of cells the cell-of-interest was contacting regardless of inhibition (Supplemental Fig. 11, see online supplementary material for a color version of this figure). This suggests that for the cells quantified in this paper, the fiber density was not a confounding factor.

In future research, it would be desirable to reduce beading that occurred in the highest curvature samples, perhaps by fine tuning of the applied voltage [68]. We conducted some preliminary studies on suspended nanofibers, however cell-attachment to suspended fibers introduces stiffness as a variable [69]. Therefore, we chose to deposit fibers to a consistent cell-attachment-passivated glass substrate. This prevented cell manipulation of the fibers, thereby reducing the effect of curvature on the effective stiffness of the fiber. Although adding kinetic alignment to our collector mechanism may have increased alignment, the main advantage we obtained was the ability to collect fibers on more samples in a given amount of time. Unfortunately, we did not do a direct comparison between the two sets of fibers, and so we cannot make claims about the differences cells would perceive.

The PMMA fibers utilized in our work were relatively stiff and did not have intrinsic biological binding capability, but instead acted as a substrate for protein deposition from the culture media that facilitated cell-attachment. Nevertheless, comparing to fibers made of natural matrix proteins can add context to our data [29–32]. Collagen fiber curvature fluctuates physiologically by the bundling of individual fibrils and is therefore variable. Shih et al. compared cell migration on electrospun collagen randomly oriented nanofibers of three curvature ranges: 2–4, 4–10, and 10–40 μm⁻¹. Compared with the TCPS control, cells on all three fiber sizes were slower with a minimum on the 4–10 μm⁻¹ size [70]. By using electrospinning to make these fibers, some of the concerns with covarying properties were avoided, making relevant data. Fiber alignment and different surface chemistries may have contributed to a difference with the exact trend we observed. In our studies, variable amounts or conformations of adsorbed proteins to the surfaces of different shapes and curvatures may occur, modulating the cellular response [71, 72]. Pelipenko et al. hypothesized that the soft nature of PVA fibers contributed to a keratinocyte and dermal fibroblast increasing trend in migration as fiber curvature increased [73]. These seemingly contrary results underscore the fact that cell migration is affected by a variety of mechanical properties. In any case, the fact that relationships to curvature were found is in agreement with the conclusions of this work.

Inhibition of ROCK in MEFs migrating on fibers produced a decrease in migration velocities in both sets of fibers we examined. Rho/ROCK signaling is integral to certain migration and stem cell differentiation pathways. Noriega et al. found that 6.67 μm⁻¹ fibers down regulated ROCK-1 gene expression compared with 2 μm⁻¹ fibers while RhoA was nearly the same on both fiber sizes [74]. This fits with our result that MEF migration was not as markedly affected by ROCK inhibition on the smaller fiber curvatures. Using CH310T1/2 cells, Andalib et al. compared ROCK expression and cell shape on random and aligned fibers of about 15.38 μm⁻¹. The aligned and random fibers caused an increasing trend in ROCK expression, however, only in the aligned fibers was this significant [75]. Considering these studies, a relationship between nanofiber based migration and Rho/ROCK signaling was not surprising, however it was not known if Rho/ROCK was required. Our results suggest curvature may be a regulator of Rho/ROCK signaling.

Although our study focused on the curvature of fibers, contact guidance may have also been involved [76]. For instance, 1D flat analogs of nanofibers affected cell migration through contact guidance [77]. Similarly, Kaiser et al. compared cell migration velocity on micro-machined titanium alloy surfaces with varied parallel grooves and found that all cells migrated faster on grooves than the flat surface [78]. We found that fiber-based MEF migration may have been influenced by both ROCK and cdc42, however the effect on fiber migration was similar across the middle of the curvature range assayed. Notably, as the fiber curvature decreased the effect of inhibition with Y27632 or ML141 became similar to a flat surface. This implies that below a critical threshold curvature (perhaps <1.33 μm⁻¹) certain information encoded in out-of-plane curvature was no longer detectable by an attached MEF. It is also possible multiple curvature ranges exist where curvature, small GTPase biology or contact guidance dominate. Further work will be required to separate the effects of curvature from the effect of contact guidance. A curvature threshold may also form a foundation for future studies into the molecular basis for curvature sensing.

In conclusion, it is only through studies exploring parameter spaces that tissue engineers can take advantage of the control afforded by electrospinning to study cell biology and improve tissue engineering scaffold function. We utilized an electrospun fiber substrate designed to isolate the factor of curvature. The flat surface controls were made of the same material. Random migration of MEFs was quantified via time-lapse imaging and cell shape of eGFP expressing mMSCs was qualified by semi-automated methods. We found that random fibroblastic migration velocity was increased up to 2.5X the flat surface control by fiber curvature. The shape of the cell cytoplasm was characterized by high aspect ratios compared with the flat surface. Pharmacological inhibition of both cdc42 and ROCK decreased migration velocity in the middle range of curvatures. However, when the fiber curvature approached zero, the effect of inhibition flipped and followed the pattern of the flat surface. This suggests that curvatures may be grouped based on biological effects. Although contact guidance may be a contributing cause of the observations we have made, our experiments cannot rule out the effect of curvature. The study presented here gives evidence that migration velocity can be controlled by fiber curvature through intracellular signaling pathways. Further work is likely to result in the ability to rationally design tissue engineering scaffolds, learning about cell biology in the process.

Authors’ Contributions

Daniel T. Bowers: methodology, software, formal analysis, investigation, data curation, writing – drafting & editing, project administration; Justin L. Brown: conceptualization, methodology, resources, writing – review & editing, supervision, project administration, funding acquisition.

Funding

This work was supported by the National Institute of Biomedical Imaging and Bioengineering at the National Institutes of Health (R21EB019230); and National Institute of Arthritis and Musculoskeletal and Skin Diseases at the National Institutes of Health (R03AR065192). The funding sources did not have any influence on the decision to conduct, analyze, or publish this work.

Conflict of interest statement

The authors confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.

Data Availability

Raw data and high-resolution raw images are available upon request.