Square prism micropillars on poly(methyl methacrylate) surfaces modulate the morphology and differentiation of human dental pulp mesenchymal stem cells

Abstract

Use of soluble factors is the most common strategy to induce osteogenic differentiation of mesenchymal stem cells (MSCs) in vitro, but it may raise potential side effects in vivo. The topographies of the substrate surfaces affect cell behavior, and this could be a promising approach to guide stem cell differentiation. Micropillars have been reported to modulate cellular and subcellular shape, and it is particularly interesting to investigate whether these changes in cell morphology can modulate gene expression and lineage commitment without chemical induction. In this study, poly(methyl methacrylate) (PMMA) films were decorated with square prism micropillars with different lateral dimensions (4, 8 and 16 μm), and the surface wettability of the substrates was altered by oxygen plasma treatment. Both pattern dimensions and hydrophilicity were found to affect the attachment, proliferation, and most importantly, gene expression of human dental pulp mesenchymal stem cells (DPSCs). Decreasing the pillar width and interpillar spacing enhanced cell attachment, cell elongation, and nuclear deformation of nuclei, but reduced early proliferation rate. Surfaces with 4 or 8 μm wide pillars/gaps upregulated the expression of early bone-marker genes and mineralization over 28 days of culture. Exposure to oxygen plasma increased wettability and promoted cell attachment and proliferation but delayed osteogenesis. Our findings showed that surface topography and chemistry are very useful tools in controlling cell behavior on substrates and they can even help create better implants. The most important finding is that hydrophobic micropillars on polymeric substrate surfaces can be exploited in inducing osteogenic differentiation of MSCs without any differentiation supplements.

Graphical Abstract

Introduction

Tissue engineering strategies have emerged as alternatives to auto- or allografts and implants in the regeneration of defected tissues in cases of massive bone losses or fractions^[1–3]. MSCs have been of great interest in bone tissue engineering because of their high self-renewal and osteogenic capacity^[4]. Human dental pulp has been identified as an easily accessible source of MSCs for clinical applications and dental pulp stem cells (DPSCs) have been shown to exhibit a high osteogenic/odontogenic capacity^{[5, 6]}. Effective and safe induction of osteogenesis, however, has been an issue for the researchers. It is difficult to deliver soluble differentiation supplements such as bone morphogenetic proteins (BMPs) and dexamethasone in a spatially and temporally controlled manner to obtain satisfactory effects^[7]. Moreover, their use raises safety concerns including the potential side effects and the risk of tumorigenicity in supraphysiological doses^[8]. Besides biochemical factors, the fate of stem cells is also modulated by the topography of basement membranes, which regulate adhesion, distribution and organization of cytoskeletal elements, migration, survival, growth, and differentiation of adherent cells^[9–11]. For example, changes in the morphology of BMSCs upon spreading on micro- and nanoscale topography of the collagen fibrils and hydroxyapatite crystals of the bone matrix contribute to their differentiation into osteoblasts^{[12, 13]}. Then further differentiation into mechanosensitive osteocytes is triggered upon entrapment in bone lacunae of 10–30 μm diameters distorts the cuboid shape of osteoblasts into a highly-branched morphology with numerous cytoplasmic extensions running along the narrow channels called canaliculi^[14]. The mechanism by which these physical features modulate the cell fate has been explained with the mechanotransduction of local cytoskeletal tensions created by the changes in size and shape of the cells^{[11, 15]}. These tensions or signals are transmitted from focal adhesion complexes to cell nucleus either indirectly, chemically, through the signaling cascades that modulate the activity of transcription factors^[16], or directly through the linker proteins between the cytoskeletal filaments and nuclear lamina, resulting in a repositioning of chromatin and alteration of gene expression^[17–20]. The modulation of cell morphology through surface topography may therefore be a safer alternative to soluble factors for guiding lineage commitment of stem cells.

The advances in lithographic micro- and nanostructure fabrication methods in silicon microelectronics industry have enabled researchers in biology and biomedicine to design surface structures to study the responses of cells to the surface topography^{[21, 22]}. A variety of micro- and nanoscale 3D surface features including ridges and grooves, randomly or evenly distributed pits, pillars/posts, or wells^{[23, 24]} have been shown to influence the attachment^{[25, 26]}, migration^{[27, 28]}, morphology^{[29, 30]}, proliferation^[31–33], and differentiation of the cells^[34–38]. Cells were reported to align themselves along the ridges of surface patterns through a phenomenon called contact guidance^[39] and exhibit morphological changes in cell body^{[25, 40–44]}. In addition to causing alterations of cell shape, micropillar-decorated surfaces with low micron sized cues were shown to trigger severe deformations in cell nuclei^{[33, 38, 45–50]}. We also reported that DPSCs cultured on PMMA surfaces decorated with square prism micropillars displayed distinctive changes in cellular and nuclear shapes, the type and extent of which depended on the lateral dimensions and hydrophilicity of the pillars^[51].

In a more recent study, we showed that micropillars improve osteogenic activity of human osteoblast-like cells on PMMA surfaces by promoting cell attachment, alkaline phosphatase (ALP) expression and mineralization^[52]. Our findings address a fundamental question: besides improving osteogenic activity, could these micropillar decorated surfaces also induce osteogenic differentiation of mesenchymal stem cells without biochemical supplements? Changes in cellular^{[37, 53, 54]} or nuclear^[38] morphologies on patterned surfaces have been demonstrated to improve osteogenic differentiation of MSCs induced by differentiation media. Studies on the induction of stem cell differentiation on 3D topographical features without soluble factors, however, are limited. Examples of induction of osteogenesis by surface topography without chemical supplements include the culture of mesenchymal stem cells on artificial ECM^[55], silk fibroin^[56] and polyurethane^[57] substrates but these substrates were decorated with submicron ridges and grooves, on PMMA surfaces decorated with randomly distributed nanopillars^[34] and on tantalum coated microarrays with shallow (2.4 μm high) pillars^[26]. In none of these studies, however, cytoskeletal and nuclear deformations were extensively induced, emphasized or correlated with osteogenesis.

Here we hypothesize that the osteogenic differentiation of DPSCs can be induced by micropillars through the modulation of cytoskeletal and nuclear morphologies and no differentiation supplements. We planned to assess the influence of induced morphological changes on the fate of DPSCs by performing proliferation, osteogenic gene expression, bone matrix deposition and mineralization tests. For this purpose, a series of PMMA micropillar arrays with pillar widths and spacings in the size range of bone lacunae (4, 8 and 16 μm) were designed. The hydrophilicity of the substrates was improved by oxygen plasma treatment, which we had previously shown to induce morphological changes in a shorter time and to a higher extent^[51]. Findings of this present study will lead to more defined topographical parameters for use in the design of tissue engineering scaffolds and bone implants with better osteoinduction and osseointegration capacities.

Materials and Methods

Fabrication and characterization of PMMA micropillar arrays

Poly(methyl methacrylate) (PMMA; $\overline{M}\text{w}=996\text{kDa}$) was purchased from Sigma-Aldrich (USA). The production of silicon microarrays by typical photolithography, and the preparation of polymeric micropillar arrays by solvent casting on silicon array templates or wafers have been previously reported^{[45, 51, 52]}. Three types of silicon wafers were decorated with 8 μm tall square prism pillars: P4G4, P8G8 and P16G16, where P is the lateral dimension of square prism pillars and G is the interpillar gap in μm. Poly(dimethyl siloxane) (PDMS) (Sylgard 184, Dow Corning Company, UK) negatives of the wafers were produced and PMMA replicates of the original wafers were fabricated by solvent casting of 20% (w/v) PMMA solution in chloroform (Avantor J.T. Baker, USA) on the PDMS molds. All films had a surface area of 0.64 cm². PMMA replicas were Au-Pd coated under vacuum and visualized using SEM (400 F Field Emission SEM, USA). Surface wettability of the substrates was improved by oxygen plasma treatment using a Femto 40 kHz Plasma System (Diener Electronic, Germany). The pressure in the plasma chamber was maintained at 20 mbar and the substrates were exposed to oxygen plasma at 100 W for 10 min. Samples were then immediately used in contact angle measurements and in vitro tests.

Cell isolation, culture and seeding protocol

Human dental pulp stem cells were isolated from a third molar teeth via an enzymatic degradation method as reported before^{[51, 58]}. Human dental pulp extracts from impacted third molars were obtained from 3 patients (2 female, 1 male), aged 18–22 years, with informed written consent at the Middle East Technical University Medical Center with the approval of the METU Human Subjects Ethics Committee of, Ankara, Turkey (No: 28620816/505–69). Briefly, pulp tissue fragments were minced into small pieces and digested with collagenase type I and dispase II (Sigma-Aldrich, USA). Cells from 3 patients were filtered, pooled and expanded in DMEM:F12 1:1 (Lonza, Switzerland) supplemented with fetal bovine serum (FBS, 10%, v/v) (Biowest, France), penicillin and streptomycin (Pen-Strep, 100 U.mL⁻¹) and L-glutamine (5 mM) (Lonza, Switzerland). Passages 2–5 were used in the in vitro studies. Cells in the chemical induction group (control) were cultured on TCPS (n=3) in an expansion media supplemented with 100 nM dexamethasone, 10 mM β-glycerophosphate and 50 μM L-ascorbic acid (Sigma-Aldrich, USA).

Cells were seeded onto the substrates as described previously^{[45, 51]}. Briefly, PMMA films were sterilized by exposing both sides to UV for 15 min and placed into 12-well plates. Cells were detached from the tissue culture flasks, counted using Nucleocounter (Chemometec, Denmark), and seeded onto the patterned surfaces at a density of 10⁴ cells.cm⁻². The number of seeded cells were found to be too high for a morphology analysis of individual cells even at early time points. It was, however, too low for the determination of cell numbers on Day 1 by Alamar Blue assay or for the extraction of enough mRNA on Day 7 for gene expression analysis. Cell seeding density was, therefore, optimized as 3×10³ cells.cm⁻² for microscopy, 10⁴ cells.cm⁻² for proliferation and immunofluorescence, and 3×10⁴ cells.cm⁻² for cell attachment and gene expression analyses. After 2 h of seeding, 2 mL of expansion medium was gently added to each well. Plates were incubated at 37°C and 5% CO₂ incubator, and the medium was changed every 2 days.

Flow cytometry

Isolated cells were characterized for the MSC negative and positive surface antigens by flow cytometry (BD Accuri C6, USA). Trypsinized cells were washed with FACS buffer (1:1000 sodium azide (Sigma-Aldrich) and 1:100 BSA (Sigma-Aldrich) in PBS) and fixed in 4% paraformaldehyde (Sigma-Aldrich) solution for 15 min. Fixed cells were stained with mouse IgG1 anti-human monoclonal antibodies (Biolegend, USA) against CD31 (#303110), CD45 (#304017), CD90 (#328116) and CD105 (#323212) and 10⁵ cells were counted per sample.

Cell attachment and proliferation

Numbers of cells on the substrates were determined using the Alamar Blue® cell viability assay (Thermo Fisher Scientific, USA) 16 h after seeding and on Days 3, 7, 14 and 21. Substrates were washed twice with PBS and incubated in 500 μL Alamar blue solution (10% in DMEM High Glucose colorless (Lonza) supplemented with 100 U.mL⁻¹ Pen-Strep) for 2 h at 37°C and 5% CO₂. Absorbance was measured twice for each sample (n=3) at 570 nm (reduced) and 595 nm (oxidized) using a plate reader (Multiscan Spectrum, Thermo Scientific). Dye reduction (%) was calculated as described in the assay guide, and it was converted to cell numbers using a calibration curve (Fig. S1). The specific growth rate of the DPSCs during the exponential growth phase were calculated by using the cell numbers and the Equation 2 which was derived from Equation 1^[59]:

$$N_t=N_0{e}^{\mu t}$$ (1)

$$\mu =\text{ln}\left(\frac{N_t}{N_0}\right)/t$$ (2)

where N_t is the cell numbers at a specific time point within the exponential growth phase, N₀ the initial cell numbers on Day 3, t the time (d) and μ the specific growth rate (d⁻¹).

Scanning electron microscopy (SEM) imaging

Samples were rinsed with PBS, fixed in 4% paraformaldehyde, washed with PIPES (piperazine-N,N’-bis(ethanesulfonic acid)) buffer (Sigma-Aldrich, USA) and incubated in 1% osmium tetroxide (OsO₄) (Polysciences, USA) in PIPES buffer at RT for 1 h. Then, the samples were washed again with PIPES buffer and dehydrated by incubating in a series of 50, 70 and 100% (v/v) ethanol solutions at RT for 5 min each. Samples were Au-Pd coated under vacuum and examined by SEM.

Confocal scanning laser microscopy (CSLM) imaging

Samples were washed with PBS and fixed by immersing the samples in 4% (w/v) paraformaldehyde solution for 15 min at RT. Cells were permeabilized with Triton X-100 (PanReac Applichem, Germany) solution (0.1%, v/v, in PBS) for 5 min at RT. After washing with PBS, samples were incubated in BSA blocking solution (1%, w/v, in PBS) at 37°C for 30 min. Samples were incubated in Alexa Fluor 488 conjugated phalloidin (Thermo Fisher Scientific, USA) solution (1:50 in 0.1% BSA) at 37°C for 1 h and in DRAQ5 (ab108410, Abcam, UK) solution (1:1000 in 0.1% BSA) at RT for 1 h to label actin cytoskeleton and nuclei, respectively. Then, the samples were washed twice with PBS and analyzed by using Leica SPE confocal laser scanning microscope (Germany).

qRT-PCR

The total RNA was collected from 12-well culture plates (n=4) for the chemical induction group or n=4 samples (each repeat consisted of a pool of 5 separate PMMA replicas) on Days 7, 14, and 28 using Masterpure RNA Purification Kit (Epicentre, USA) according to the manufacturer’s instructions. First-strand cDNAs were reverse transcribed using RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) from each sample using a thermal cycler (iCycler, BIORAD, USA). qRT-PCR reactions were performed using GoTaq® qPCR Mastermix (Promega, USA) with Rotor-Gene Q real-time PCR cycler (Qiagen, Germany) using forward and reverse primers specific for glyceraldehyde 3-phosphate dehydrogenase (GAPDH), Osterix (OSX), alkaline phosphatase (ALP) and osteocalcin (OC) listed in Supplementary Table 1. PCR conditions were: an initial denaturation at 95°C (2 min) and 35 cycles of denaturation at 95°C (15 s), and annealing/extension at 60°C (60 s). The C_t values for OSX, ALP and OC were normalized to that of the housekeeping gene GAPDH. Relative expression of the genes was calculated by the ΔΔC_t method as described elsewhere^[60]. Briefly, the fold changes in gene expression (2^−ΔΔCt) by chemically induced cells on TCPS and by the cells cultured on micropatterned surfaces were calculated by normalization to expression levels in growth medium without osteogenic supplements and to expression levels on unpatterned control surfaces, respectively.

Immunofluorescence

Cells attached on the samples were fixed with paraformaldehyde solution. They were then incubated at 37°C for 1 h in blocking solution (5% v/v goat serum, 1% v/v Tween 20, 0.1% w/v BSA, 0.1% w/v sodium azide in 0.01 M PBS). For the immunostaining of collagen type I and osteopontin, samples were incubated within primary antibody (Rabbit anti-osteopontin, ab91655; mouse anti-collagen type I, ab23446, Abca, UK) for 1 h and in secondary antibody (Goat anti-Rabbit IgG Alexa Fluor 488, A-11008; Goat anti-Mouse IgG Alexa Fluor 532, A-11002, Invitrogen, USA) for 30 min. Cell nuclei were stained with DRAQ5 for 15 min at RT. After preparation, specimens were kept in the dark, humidified containers until microscopic examination. CLSM images were collected using Leica SPE and the laser power was kept constant for each channel. Signal intensity for each marker was normalized to the signal intensity of nuclei to normalize matrix deposition to cell numbers.

Mineralization

Mineralization (calcium deposition) to evaluate the degree of osteogenesis on the samples was determined by Alizarin red staining. Samples were fixed in 4% paraformaldehyde solution for 15 min, rinsed with 0.01M PBS, stained with Alizarin Red S (Cyagen, USA) for 5 min and rinsed with 0.01M PBS before imaging. Images were collected with Olympus IX70 inverted light microscope (Japan) with a 20x objective.

Statistical Analysis

All quantitative data in this study (n≥3) are expressed as mean ± standard deviation. One-way analysis of variance (One-way ANOVA) with Tukey’s pairwise comparison test was performed to analyze the results between the groups. p-values less than 0.05 were considered statistically significant. Pearson product moment correlation test was performed to calculate the correlation coefficients (r) between two variables. Variables with |r| values larger than 0.5 and 0.85 were considered correlated and strongly correlated, respectively.

Results and Discussion

Characterization of micropillar decorated PMMA surfaces

The topographical features of the ECM are known to regulate the responses of MSCs. Therefore, engineering of substrate topography has been widely investigated in the last decade with the aim of guiding lineage commitment of stem cells^{[9, 23]}. We prepared micropillar arrays on PMMA surfaces that we previously reported to generate distinct morphological changes in DPSCs^[51]. Three types of PMMA surfaces with micropillars of 8 μm height and varying lateral dimensions (P4G4, P8G8 and P16G16 in Fig.1A, P and G represent the width of square pillars and interpillar gaps in μm, respectively) were prepared (Fig.1B). Unpatterned PMMA substrates were used as negative control surfaces. We previously showed that the Young’s moduli (E: 1300 MPa) and ultimate tensile strength (UTS: 41 MPa) (Fig.S2) of solvent cast PMMA films were in the range of natural bone tissue (E: 100–30000 MPa, UTS: 5–150 MPa)^[52], indicating that the mechanical properties of our substrates were suitable for osteogenesis. Wettability of polymeric surfaces modulates the type, amount, and conformation of the adsorbed proteins, which in turn influences adhesion, proliferation, and differentiation of mammalian cells^[61]. O₂ plasma treatment introduces oxygen rich polar groups on substrate surfaces and modulate surface wettability,^{[62, 63]}, which is known to affect behavior of adherent cells^[64]. A set of substrates were therefore exposed to oxygen plasma prior to cell seeding (Fig.1C) to investigate the influence of improved wettability of micropillar arrays and earlier and more robust alterations in cell morphology^[51] on the fate of DPSCs. AFM analysis (Fig.S3) showed that oxygen plasma treatment slightly increased the nanoscale roughness of unpatterned surfaces, bringing RMS deviation from 6.4 to 15.9 nm and average height from 33.1 to 54.6 nm (Supplementary table 2).

Figure 1.

Design and surface modification of PMMA films. (A) Designed dimensions of three PMMA surfaces decorated with square prism micropillars. (B) SEM micrographs of the micropatterned PMMA films. Scale bars: 10 μm. (C) Representative SEM micrographs of DPSCs on untreated (UT) and plasma treated (PT) P8G8 surfaces on Day 7.

Characterization of isolated cells

Cells isolated from dental pulps had MSC-like well spread, broad, and flattened shapes on TCPS (Fig.S4A) which is consistent with literature^{[58, 65]}. Cells at passage 2 were found to be positive (>50%) for MSC surface markers CD90 (100%) and CD105 (73.6%) and negative for hematopoietic stem cell surface markers CD31 (0.3%) and CD45 (0.4%) (Fig.4A). The relatively lower expression of CD105 at low passages was consistent with other studies^{[66, 67]}, and it was increased significantly at passage 5 (99.6%, Fig.S4B) probably because of a gradual decrease in the proportion of CD105 negative ectomesenchymal stem cells that share a common origin with neural crest cells^{[67, 68]}.

Figure 4.

Nuclear and cytoskeletal morphology of DPSCs on the substrates. (A) Confocal micrographs of the DPSCs on PMMA films on Day 7 (Scale bars: 50 μm). (B) Z-stack images of the cells on P4G4 surface on Days 1, 7 and 14 (Scale bars: 30 μm). Red: Actin cytoskeleton (Alexa Fluor 488-conjugated Phalloidin), green: nucleus (DRAQ5).

DPSCs cultured on TCPS in osteogenic induction media displayed a significant increase in the expressions of OSX, ALP, and OC over 28 days of culture (Fig.2B), confirming the osteogenic capacity of the isolated cells. OSX expression was slightly upregulated on Day 7, followed by a 10-fold increase on Day 14, and a sharp decrease back to the baseline on Day 28. A similar observation was reported by Igarashi et al., who observed an initial increase of OSX expression with differentiating calvarial cells for the first 15 days and a gradual decrease afterwards^[69]. Peak OSX levels on Day 14 was accompanied by significant upregulation of ALP and OC expressions on Days 14 and 28. This was expected, because OSX is an early osteogenic transcription factor that lies downstream of runt-related transcription factor 2 (Runx2) and stimulates bone matrix deposition and mineralization through elevated levels of collagen type I, ALP and OC^[70].

Figure 2.

Characterization of the cells isolated from dental pulp tissues. (A) Flow cytometry histograms of passage 2 cells gated on the forward and side scatter of analyzed particles for the expression of MSC positive CD90 and CD105 (red), and negative CD31 and CD45 (green) surface markers. Black curves: Unstained control. (B) Relative expression of the osteogenic marker genes osterix (OSX), alkaline phosphatase (ALP), and osteocalcin (OC) by the cells chemically induced with dexamethasone. Expression levels in the chemical induction group were normalized to the corresponding levels in the maintenance media without osteogenic supplements.

Attachment of DPSCs on the substrates

Cell numbers on the substrates 16 h after seeding given in Fig.3A showed a lower number of cells attached on the substrates than the TCPS control. Meanwhile, a significantly higher number of cells attached on the patterned surfaces compared to the unpatterned control in the untreated group indicated a positive effect of micropillars on cell attachment. Enhanced cell attachment by micropillars was reported earlier for astrocytes^[71], cardiomyocytes^[72], and MSCs^{[26, 73]}, which can be attributed to larger contact area^[73] and better formation and maturation of focal adhesions^[74] assisted by 3D patterns. Despite the decrease in wettability with a decrease in pillar and gap length, cell attachment was higher on P4G4 than on P8G8 and P16G16. This observation cannot be explained with increased contact as all patterned surfaces used in this study had geometrically the same total surface area. Higher cell attachment on P4G4 was probably a result of the highest number of pillars per unit area that a cell can interact, considering that the edges of pillar tops were shown to be where focal contacts were mainly concentrated on^[75].

Figure 3.

Cell attachment on the substrates. (A) The number of DPSC on PMMA films 16 h after cell seeding. (Seeding density: 3×10⁴ cells.cm⁻²). Hatched columns represent oxygen plasma treated substrates. Data represent mean ± SD of three replicates, *p < 0.05, **p < 0.01 and ***p < 0.001. Asterisks above the bars represent significance compared to the unpatterned control. n.s.: not significant. (B) SEM micrographs of DPSCs on (i-iv) untreated and (v-viii) oxygen plasma treated substrates on Day 1 (Scale bars: 20 μm, UC: unpatterned control).

Oxygen plasma treatment dramatically increased cell attachment on all surfaces but P4G4. Cell attachment on the plasma treated surfaces increased with increasing pillar and gap width, from P4G4 towards P16G16, possibly because of the larger interpillar gaps being able to accommodate more cells. The increase in cell attachment upon plasma treatment can be attributed to the change of profile and conformation of adsorbed proteins with an increase in hydrophilicity as well as the nano-roughening of the surfaces due to plasma etching^[64]. A positive effect of oxygen plasma treatment on cell attachment was also reported before by others^{[76, 77]}. In those studies, WCA were in the range 40°−70°, which is generally accepted to be optimum for mammalian cell adhesion on polymeric surfaces^[78]. The range of WCA of the plasma treated surfaces used in this study, however, was much lower (between 10–20°)^[51] and cell attachment was still very high compared to untreated counterparts. Similar observations of significantly improved cell attachment on extremely wettable surfaces were also reported with nerve cells and osteoblasts on oxygen plasma treated poly(l-lactic acid) (PLLA)^[79] and hexamethyldisiloxane surfaces^[80], respectively. Our findings and those of others indicate that the optimal wettability for cell attachment is not a fixed value but rather vary with the material and cell type, and treatment conditions.

Positioning and morphology of DPSCs on the substrates

Lateral dimensions and hydrophilicity of the micropillars determined the relative positions and morphology of DPSCs as can be seen in SEM micrographs given in Fig.3B (lower magnification in Fig.S5). Cells on the untreated smooth surfaces (Fig.3B-i) did not spread well while the cells on the O₂ plasma treated counterparts (Fig.3B-v) did with numerous extensions. On the untreated P4G4 and P8G8 surfaces, most of the cell bulk was located on pillar tops (Fig.3B-ii, iii), forming bridges over the interpillar spaces and extending their filopodia downward to the substrate floor. On the plasma treated counterparts (Fig.3B-vi, vii), on the other hand, cells were localized on the substrate floor with their filopodia extended towards and stretched over the pillar tops. Cells on both the untreated and the plasma treated P16G16 surfaces (Fig.3B-iv, viii) were spread in the interpillar gaps.

Confocal microscopy images of the cells presented in Fig.4A shows the distinct changes in cell and nucleus morphology. Cells were squeezed between the pillars and conformed their cytoplasm to the interpillar spaces, constraining their typical spread morphology seen on the unpatterned surface. Cells were extremely elongated on P4G4 but highly branched on P8G8 and P16G16 surfaces. Same trends in cell morphology were also observed on the plasma treated surfaces. Cells were localized more in the interpillar spaces and surround the pillars on plasma treated P8G8 and P16G16 substrates. Together with cell bodies, nuclei were also inserted between the pillars and distorted on patterned surfaces, particularly on P4G4 (Fig. 4B). Interestingly, cells that were located in between the pillars of untreated P4G4 and P8G8 (Fig.S6) surfaces on Day 7 crawled back on top of the pillars on Day 14 and showed no significant nuclear deformation. On the oxygen plasma treated counterparts, however, majority of the cells were still squeezed in between the pillars and had deformed nuclei on Day 14.

Cell proliferation on the substrates

Both micropatterns and plasma modification had a profound effect on cell proliferation on the substrates (Fig.5A). Cell numbers (Fig.S7) on untreated P8G8 and P16G16 surfaces were comparable with that on the unpatterned control, while there were significantly lower number of cells on P4G4 surface at all time points. Cell numbers on oxygen plasma treated substrates were significantly higher than those on untreated counterparts throughout 21 days of culture, but P4G4 surface still had less cells than unpatterned or P8G8 and P16G16 surfaces on Days 3 and 7. Oxygen plasma treatment and larger pillar/gap widths were found to increase proliferation rate on the substrates between days 3 and 7. The positive effect of oxygen plasma modification on cell division was also reported by others^{[81, 82]} and it can be attributed to the introduction of oxygen-rich groups such as hydroxyl and carboxylic acid, which were shown to modulate protein adsorption and subsequent transmission of proliferative signals mediated by integrin-ligand interactions^[83]. After Day 7, specific growth rates on plasma treated surfaces decreased compared to the earlier time intervals and became lower than those on the untreated substrates (Supplementary Table 3), probably because of contact inhibition^[84] after reaching confluency.

Figure 5.

Proliferation of DPSCs on PMMA substrates. (A) Cell numbers on the substrates determined by Alamar blue assay on days 3, 7, 14, and 21 (Seeding density: 10⁴ cells.cm⁻²). Dashed lines represent proliferation on the oxygen plasma treated substrates. (B) Analysis of correlation between average nuclear deformations and specific growth rates (μ, day⁻¹) on untreated and oxygen plasma treated substrates between Days 3 and 7.

Considering the increase in nuclear deformations and decrease in proliferation rates with a decrease in pillar and gap size from P16G16 towards P4G4, a negative effect of nuclear deformations on cell proliferation can be suggested, particularly at early time points. Mitosis is a highly organized process that involves precise and predetermined rearrangements of genetic material in the nucleus; therefore, severe deformations in the nuclei might interrupt cell division by obstructing the access of the replication machinery to DNA^[85]. Indeed, a negative correlation was found between the specific growth rates and the average nuclear deformation values on the substrates during Days 3–7 (Fig.5B). Nagayama et al. also reported inhibition of the proliferation of vascular smooth muscle cells on micropillar decorated surfaces, which they explained by the condensation of the chromatin in the deformed nuclei^[33]. Specific growth rates between days 7 and 14 on untreated P4G4 and P8G8 surfaces leveled with those on the unpatterned and P16G16 surfaces (Supplementary Table 3), probably because of the gradual disappearance of nuclear deformations on P4G4 and P8G8 as the cells crawled onto the pillar tops at later time points.

Expression of osteogenic marker genes on the substrates

Micropillar arrays on which distinct changes were observed in cytoskeletal and nuclear morphologies led to alterations in gene expression (Fig. 6A–C) as expected. On Day 7, OSX expression was 2-fold higher on the untreated P4G4 and P8G8 but not on P16G16. However, the relative expression of OSX on P4G4 and P8G8 surfaces levelled down to the baseline on Days 14 and 28. The early upregulation of OSX on P4G4 and P8G8 surfaces was followed by approximately 4-fold and 6-fold increases in ALP expression on Days 14 and 28, respectively. The ALP levels on untreated P4G4 and P8G8 on Day 14 were higher than that was obtained with chemical induction (Fig.2B), suggesting that the commitment to osteogenic lineage was triggered by surface patterns earlier than with soluble factors. Unlike chemical induction, however, the expression of OC on untreated micropillar arrays did not change significantly over 28 days of culture. OC is a relatively late osteogenic marker^[86] and was shown before to be expressed later than ALP during osteogenesis^{[87, 88]}, indicating that micropillar arrays might have induced osteogenesis earlier than by chemicals but did not lead to a mature osteoblastic phenotype over the 4 weeks of culture. Gene expression at later time points should be analyzed to investigate whether a fully mature osteoblast phenotype can be achieved on the patterned surfaces. In this case, the time that cells spend in between the micropillar features could be increased by delaying confluency by decreasing the initial number of cells seeded.

Figure 6.

Relative expression of the osteogenic marker genes (A) Osterix (OSX), (B) Alkaline phosphatase (ALP), and (C) Osteocalcin (OC) The gene expressions on micropillar decorated films were normalized to the mean values on the unpatterned control films. Hatched columns represent oxygen plasma treated substrates. Data represents mean ± SD for four samples, *p < 0.05, **p < 0.01 and ***p < 0.001. *, # and + above the bars represent significance compared to smooth, P4G4 and P8G8 surfaces, respectively. n.s: not significant.

Interestingly, no significant change was observed in the expression of osteogenic marker genes on the oxygen plasma treated substrates during the first 2 weeks of culture (Fig.6A–C, hatched columns). The only significant change was the upregulation of OSX on P4G4 and P8G8 surfaces on Day 28, suggesting a much later commitment to osteogenesis on the plasma treated substrates. Our hypothesis was that cytoskeletal tensions due to the micropillar mediated morphological changes and nuclear deformations could trigger osteogenesis. Considering earlier and more robust morphological changes we reported before on the oxygen plasma treated micropillar arrays^[51], we were expecting the cells on such surfaces to differentiate earlier. An explanation to this situation could be the initial high proliferation rates on the plasma treated surfaces as summarized in Supplementary Table 4. Cell cycle withdrawal and differentiation are coupled processes; differentiating cells need a longer cell cycle to respond to external signals and to accumulate differentiation-inducing transcription factors, resulting in a decrease in proliferation rate^{[89, 90]}. For example, pre-osteoblastic cells differentiated by forced upregulation of Runx2 and OSX were shown to suspend cell cycle, resulting in a transition from active cell growth to quiescence^[91]. Significantly higher initial proliferation rates we recorded on plasma treated surfaces point out to shorter cell cycles, which might have prevented cells from getting into a differentiation state. The influence of cellular shape on the expression of osteogenic genes was investigated by correlation test. Surface topography-induced elongation of MSCs was reported by others to enhance osteogenesis by chemical induction^{[37, 92, 93]}. This was later explained with the upregulation of OSX via reorganization of cytoskeletal elements and following signaling cascades on patterned surfaces^[95–97]. Here, on the untreated surfaces, we found no correlation between cell elongation (observed on P4G4) or branching (observed on P8G8 and P16G16) and OSX expression (Supplementary Table 5, Fig.S8A). This finding suggests that cytoskeletal tension resulting from changes in cell morphology in the forms of elongation or branching modulates osteogenesis on patterned substrates.

It is noteworthy that OSX and ALP were upregulated on P4G4 and P8G8 but not on P16G16, implying an effect of nuclear deformation on osteogenesis. Deviations from the elliptical shape of cell nucleus was proposed as one of the factors altering structure and organization of chromatin and therefore modulating gene expression^[98]. Osteogenic genes including OSX, ALP and OC were shown to be located within the telomeric regions of large chromosomes, which were predisposed to repositioning upon intranuclear reorganization^[18]. Indeed, we found a positive correlation between OSX and ALP expressions and the extent of nuclear deformations (Supplementary Table 6, Fig.S8B). Liu et al. reported earlier that nuclear deformations on micropillars significantly enhanced the expressions of Runx2 and ALP by MSCs cultured in a cocktail of osteogenic and adipogenic supplements^[38]. Here we report a similar observation but without chemical induction.

Bone extracellular matrix deposition and mineralization

Immunofluorescence was used to detect deposition of bone ECM proteins, collagen type I (COL I) and osteopontin (OP), in response to physical cues (Fig. 7A). Cells on untreated substrates, particularly those on micropillar decorated surfaces, were found to aggregate into osteogenic nodule-like structures rich in COL I and OP. Cells on the plasma treated surfaces, however, were well spread and did not form aggregates. A few large aggregates were observed on P4G4 while there were multiple, denser aggregates on P8G8 and P16G16. Human MSCs undergoing osteogenic differentiation are known to form multilayered nodular structures in vitro^{[99, 100]}, and hydrophobic substrates were shown before to force stem cells to aggregate and enhance osteogenesis^[101]. Contrary to a previous report of aggregate formation on nano but not on micropillars^[102], our substrates triggered formation of bone nodule-like aggregations.

Figure 7.

Bone matrix deposition by DPSCs on PMMA films. (A) Immunohistochemical staining of collagen type I (COLI) and osteopontin (OP) on Day 28. Green: COLI, purple: OP, cyan: nucleus (DRAQ5). Scale bars: 200 μm. Semiquantitative analysis of the relative fluorescence intensities of (A) COLI and (B) OP on the substrates normalized to DNA signal. Relative intensities on patterned surfaces were normalized to unpatterned controls. Hatched columns represent oxygen plasma treated surfaces. Data represents mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, and ****p<0.0001.

The deposition of COL I and OP were semi-quantified in Fig.7B and Fig.7C, respectively, by normalization of fluorescence intensities to nuclear signal. Plasma modification was found to increase COL I and OP deposition on unpatterned surfaces. Moreover, normalized intensities of COL I and OP were lower on both untreated and plasma treated micropillar arrays compared to the unpatterned control surfaces. These observations suggest that bone matrix deposition increase with increasing cell spreading. We previously showed that oxygen plasma modification increases cell spread area on unpatterned surfaces. It can also clearly be seen that cell spreading is significantly restricted on micropillar decorated surfaces, particularly on untreated hydrophobic substrates. Indeed, synthesis and deposition of COL I were shown to increase with increasing cell spreading even if the levels of pro-α1(I) collagen mRNA were the same^[103]. COL I deposition was higher on the untreated micropillar arrays except P8G8 than their plasma treated counterparts, while OP deposition was higher on plasma treated substrates (except P16G16). Surface hydrophilicity was shown to enhance deposition of OP both in vitro^[104] and in vivo^[105] because of higher BMP signaling on hydrophilic surfaces.

The nodular aggregates on untreated P4G4 and P8G8 surfaces were found to be positive for calcium deposits by Alizarin red staining (Fig. 8, shown with yellow arrows), indicating mineralization. Only a low basal staining was observed on the plasma treated counterparts, indicating no significant mineralization over 28 days of culture. Higher deposition of COL-I and OP on plasma treated surfaces but lack of mineralization may look contradictory considering that bone matrix proteins promote mineralization. However, it should be noted that mineralization was observed on untreated P4G4 and P8G8 surfaces, where cell bone nodule-like aggregates were positive for both COL-I and OP, and the expression of ALP was significantly upregulated on days 14 and 28 following the early upregulation of OSX on Day 7. On plasma treated P4G4 and P8G8 surfaces, on the other hand, no significant change was observed in ALP levels compared to the unpatterned control over 28 days of culture. Upregulation of ALP is essential for bone mineralization since it promotes calcium phosphate deposition by increasing the local concentrations of inorganic phosphate and decreasing the levels of extracellular pyrophosphate, which is known to inhibit mineralization^[107].

Figure 8.

Alizarin red staining of DPSCs on PMMA films on Day 28. Yellow arrows point dark red accumulations depicting calcific deposition by the cells. Scale bars: 200 μm.

Overall, our findings indicate that square prism micropillars cause significant alterations in cytoskeletal and nuclear morphology and induce upregulation of early osteogenic genes and mineralization on hydrophobic PMMA surfaces. We propose that sharp edges of the square prism pillars and narrow interpillar gaps on P4G4 and P8G8 substrates generated high degrees of cytoskeletal tension and induced differentiation of multipotent DPSCs towards bone lineage rather than soft tissue cells such as adipocytes. It was previously shown that MSCs grown on adhesive microislands with sharp corners and curvature in the shape of rectangles or stars induced osteogenesis while the cells on smoother patterns like round islands or flowers underwent adipogenic differentiation^[92]. Besides the nature of morphological changes, high stiffness of the solvent-cast PMMA films probably favored differentiation towards osteogenesis rather than adipogenesis or myogenesis. Considering the influence of matrix stiffness on lineage commitment of MSCs^[108], the elastic modulus of our solvent cast PMMA films (~1.3 GPa^[52]) was far too high for adipogenic (1–5 kPa) or myogenic (10–20 kPa) differentiation but favors osteogenesis (>25 kPa)^[109].

Conclusions

In this study, PMMA micropillar arrays with varying lateral dimensions were fabricated and their wettability was improved by treating with oxygen plasma. These micropillar structures were shown to affect morphologies of cell body and nucleus, and modulate attachment, proliferation and osteogenic differentiation of DPSCs. Hydrophobic micropillar arrays with lateral dimensions of 4 (P4G4) and 8 μm (P8G8) induced the expression of early osteogenic genes and mineralization in bone nodule-like cell aggregates without any differentiation supplements in the growth media. In conclusion we propose that hydrophobic square prism micropillars bring osteoinduction to PMMA surfaces without biochemical supplements. Our approach is more advantageous over other methods employing nanoscale patterns that require more complicated and costly methods for fabrication^[9] or using materials such as genetically engineered proteins^[55] or decellularized bone matrices^[106] that are harder to process. Hence, P4G4 and P8G8 surfaces have significant potential for use of biomaterial topographies as a safer alternative to soluble factors in inducing the osteogenic differentiation of mesenchymal stem cells and as a result for more effective bone implants. Future studies will focus on the expression of bone marker genes at later time points to determine whether osteogenesis is fully induced also on the plasma treated surfaces and whether a fully mature osteoblast phenotype can be achieved on the untreated hydrophobic substrates without chemical induction. Besides potential applications on the surface of bone biomaterials, our strategy to induce cell differentiation by modulation of cellular morphology on micropillar decorated surfaces can also be employed in tissue engineering applications with biodegradable polymers such as poly(L-lactic acid) (PLLA) or poly(lactic-co-glycolic acid) (PLGA) using the same approach [45]. Multilayered 3D scaffolds can also be constructed by stacking individual micropillar decorated films on top of each other^[110]. Moreover, the strategy can be expanded to investigate myogenic differentiation due to the extensive cell elongation observed on P4G4 surfaces but by using softer materials such as collagen^[46].

Highlights.

• Micropillars increased cell attachment but reduced proliferation rate at early time points.

• Oxygen plasma treatment enhanced both cell attachment and proliferation rate on all substrates.

• Untreated hydrophobic arrays induced osteogenic differentiation of DPSCs without chemical induction over 28 days of culture.

• The expression levels of osteogenic genes were positively correlated with the degree of nuclear deformations.