Revealing cytokine-induced changes in the extracellular matrix with secondary ion mass spectrometry

Abstract

Cell-secreted matrices (CSMs), where extracellular matrix (ECM) deposited by monolayer cell cultures are decellularized, have been increasingly used to produce surfaces that may be reseeded with cells. Such surfaces are useful to help us understand cell-ECM interactions in a microenvironment closer to the in vivo situation than synthetic substrates with adsorbed proteins. We describe the production of CSMs from mouse primary osteoblasts (mPObs) exposed to cytokine challenge during matrix secretion, mimicking in vivo inflammatory environments. Time-of-flight secondary ion mass spectrometry (ToF-SIMS) data revealed that CSMs with cytokine challenge at day 7 or day 12 of culture can be chemically distinguished from one another and from untreated CSM using multivariate analysis. Comparison of the differences with reference spectra from adsorbed protein mixtures points towards cytokine challenge resulting in a decrease in collagen content. This is supported by immunocytochemical and histological staining, demonstrating a 44% loss of collagen mass and a 32% loss in collagen I coverage. CSM surfaces demonstrate greater cell adhesion than adsorbed ECM proteins. When mPObs were reseeded onto cytokine-challenged CSMs they exhibited reduced adhesion and elongated morphology compared to untreated CSMs. Such changes may direct subsequent cell fate and function and provide insights into pathological responses at sites of inflammation.

1. Introduction

The extracellular matrix has long been investigated as a key director of the fate, behavior and function of cells and tissues. Cells may sense surface composition, topology and mechanical properties of the ECM. Through complex signaling pathways these may influence factors as diverse as cell adhesion, contractility, morphology, and gene expression. These in turn drive phenotypic characteristics and regulate the remodeling of the ECM itself. In many inflammatory disorders, changes to ECM composition and structure drives both disease progression and severity [1,2]. For example, in inflammatory environments the action of cytokines such as IL-1β, TNF-α and IFN-γ disrupt matrix homeostasis by reducing ECM deposition and enhancing ECM resorption. Simultaneously, altered ECM deposition in inflamed tissues can actively influence immune responses at such sites [3].

The influence of the ECM on cell behavior can be challenging to investigate experimentally in a biologically relevant manner. Adsorbed protein surfaces have been widely used, although they only provide a flat, adhesive surface that does not approach the complexity of native ECM. Many developments have been made in the production of decellularized tissues and organs [4–6], their characterization [5,7,8], and developing use in the clinic [9–11]. However, whilst matrices may be produced from different tissue sources, anatomical locations, disease models or species, given their in vivo source, little control over matrix composition and structure at the point of deposition is possible. Therefore, the production of in vitro cell-secreted matrices (CSMs), where cells of interest are stimulated to deposit an extracellular matrix in situ before decellularization, providing a surface that may mimic in vivo niche environments, have been increasingly reported, with the aim of more closely recapitulating the in vivo environment.

CSMs, also referred to in the literature as cell-derived matrices (CDMs) have been produced from a range of cell types including mesenchymal stem cells [12–16], hematopoietic stem/progenitor cells [17], epithelial and endothelial cells [18], fibroblasts [19] [20], hepatic cells [21], as well as osteoblasts, both primary [22,23] and from cell lines [24]. They have found a number of applications, from early studies to understand cell proliferation and migration [19,25], to later studies investigating how the ECM directs stem cell differentiation [14,16,26]. CSMs have also been used to understand a number of disease states, notably examining cancer cell,migration and metastasis [23,27,28] [29,30], and genetic bone diseases exhibiting abnormal matrix deposition [22]. Recent studies report the use of CSMs in a broad range of studies including those of vascularization [31], MSC differentiation [32], and cell migration [33] A key advantage of cell-secreted matrices is that they may be manipulated or modified at multiple points during their deposition. This may be through selection of cell types [22,26], culture conditions [14], differentiation protocols, or post-decellularization modification [22,26].

Most commonly CSM surfaces have been studied in situ by immunocytochemical and histological techniques [15,22], or ex situ, solubilizing the ECM before analysis by immunolabelling techniques such as ELISA or Western blotting. For example, Baht et al. (2011) used immunohistochemistry to reveal increased levels of ECM components in CSMs from osteoblasts from donors with the genetic bone disease cranialsynostysis compared to healthy donors [22]. Whilst these approaches may be used to identify specific components, prior knowledge is required to select and identify appropriate targets. Spectroscopic and spectrometric techniques are useful as they collect data representative of the sample as a whole, rather than targeting specific components. Recently proteomic methods employing mass spectrometry have been used to study cell secreted matrices and decellularized tissues [34] including the identification of novel matrix components from CSM secreted by a fibrotic liver cell line [21]; a comparative analysis of CSMs that support or inhibit the maintenance of pluripotency in human embryonic stem cells [35]; and proteomic analysis of decellularized human vocal fold mucosa [36]. However, such approaches require solubilizing the sample, removing one of the advantages of CSM systems—that the ECM proteins are presented at the surface in their native “as secreted” orientations and structures. Therefore in situ surface analytical methods are advantageous.

Time-of-flight secondary ion mass spectrometry (ToF-SIMS) is a surface analytical technique that has been widely used to study a range of biological samples including proteins [37–40], lipids [41], DNA microarrays [42], cultured cells [43,44], along with soft [45–47], hard [48] and decellularized tissues [7,8]. In ToF-SIMS a primary ion beam (in this study Bi₃⁺) is used to impact the surface of interest, yielding characteristic fragments from the top few nanometers of the surface. A small proportion of these fragments are charged ions and may be analyzed using a time-of-flight detector to yield mass/charge spectra characteristic of the surface. These spectra may be interrogated for information about the chemistry [49], structure [50] and molecular orientation [51] of the surface. ToF-SIMS provides certain advantages in the analysis of biological materials over both other proteomic mass spectrometry based methods and traditional biochemical techniques. Notably ToF-SIMS allows surfaces to be analyzed in situ without digestion or disruption of the surface; and does not require a priori selection of targets of interest. When proteins are analyzed by ToF-SIMS, the positive ion spectra are dominated by their amino acid building blocks and associate molecular fragments [37]. The study of amino acid homopolymers [52,53] has allowed characteristic peaks for each amino acid residue to be identified, and this information can be used to probe the nature and orientation of proteins at interfaces [37].

Multivariate analysis techniques are useful to aid the interpretation of data from complex, multicomponent surfaces [54]. These mathematical methods can be used to identify distinct groupings and subtle spectral differences in a large number of complex samples. Principal components analysis (PCA) is one multivariate technique that has been widely used to study protein rich samples [7,37,55,56]. PCA has previously been used to distinguish spectra from adsorbed protein mixtures [37], decellularized tissues [7,8], or residual ECM proteins following cell-lift off [56,57].

With carefully decellularized surfaces and powerful analytical techniques available, this study sets out (1) to use ToF-SIMS to identify differences in the composition of cell-secreted matrices stimulated by exposure to the pro-inflammatory cytokines IL-1β and TNF-α and IFN-γ; (2) to compare findings from ToF-SIMS analysis with those from complementary biochemical techniques; and (3) to relate the structure of these cell-secreted matrices to the behavior of cells reseeded on them.

2. Materials and methods

2.1. Cell culture

Mouse primary osteoblast cells (mPObs) were isolated from the calvariae of 1–3 day old CD1 mice as previously described [58]. Briefly, calvariae were dissected and digested in a solution of 0.5 mg/ml trypsin II S (Sigma) and 1.4 mg/ml collagenase IA (Sigma). Cells released during the first two digestion periods (10 minutes each) were discarded. Cells collected from the subsequent three digestion periods (20 minutes each) were pooled and cultured at a density of 6.6 × 10³ cells/cm² in a basal medium consisting of minimum essential medium-α (αMEM; Lonza) containing 10% foetal bovine serum (FBS; Sigma), 100 U/mL penicillin, 100 µg/mL streptomycin and 2 mM L-Glutamine. Cells were passaged once before cryopreservation and used experimentally before passage 4.

2.2. CSM production

mPOb cells were plated in basal medium on 12-well tissue culture treated polystyrene (TCP) multi-well plates or 13 mm diameter Thermanox coverslips (Thermo) at a density of 10,000 cells/cm². Cells were allowed to adhere overnight before culture in osteogenic medium consisting of basal medium supplemented with 50 mM beta-glycerophosphate, 50 µg/ml L-ascorbate-2-phosphate, and, when cytokines were not present, 10 µM dexamethasone (all from Sigma, Inc.).

After 14 days, cultures were decellularized, but retained an intact extracellular matrix surface on the substrate. Cultures were washed gently with phosphate buffered saline (PBS) before treatment with 20 mM ammonium hydroxide in water for 20 minutes at 37°C with intermittent shaking, followed by a further gentle wash in PBS before treatment with 50 U/ml DNase I in PBS for 1 hour at 37°C. Subsequent to decellularization, CSMs were washed gently in PBS and allowed to soak overnight in PBS. CSMs were stored in PBS at 4°C for no more than 2 weeks before use.

2.3. Cytokine challenge

Cytokine challenge medium was prepared by supplementation of osteogenic medium (without dexamethasone) with 1 ng/ml recombinant human IL-1β, 10 ng/ml human TNF-α and 100 ng/ml mouse IFN-γ (All R&D systems). IL-1β and TNF-α have cross-species reactivity between mouse and human, whilst IFN-γ does not. Control medium was not supplemented with cytokines. After 7 or 12 days of mPOb culture for CSM production, cells were exposed to cytokine challenge medium or control medium for 48 hours, before continuation of culture in osteogenic medium.

2.4. Protein coatings

As comparator surfaces, binary mixtures of collagen I and fibronectin were adsorbed onto TCP multi-well plates or Thermanox coverslips. For ToF-SIMS analysis, single protein solutions of rat tail collagen type I or bovine fibronectin (both Sigma) were prepared at 100 µg/ml in PBS. From these 25:75, 50:40 and 75:25 volume:volume ratio mixtures were prepared. 13 mm diameter Thermanox coverslips were covered with 500 µl of single protein solutions and mixtures before incubating for 1 hour at 37°C to allow for protein adsorption. Protein coated coverslips were washed six times in ultrapure water and air dried before use. For cell culture experiments, single protein solutions of rat tail collagen type I or bovine fibronectin (both Sigma) were prepared at 10 µg/ml in PBS. From these a 50:50 volume:volume ratio mixtures was also prepared. All solutions were sterile filtered. 12-well tissue culture polystyrene multi-well plates were covered with 500 µl per well of protein solution before incubating for 1 hour at 37°C to allow for protein adsorption. Protein coated wells were washed twice in PBS before air drying and storage at 4°C.

2.5. ToF-SIMS analysis

For ToF-SIMS analysis, CSMs were prepared as described previously on 13 mm diameter Thermanox coverslips chosen for their stable support under deceullularization. To remove residual salts and lipids, CSMs were immersed in 0.15 M ammonium acetate for 1 minute with gentle agitation before dehydration through 20%, 40%, 60%, 80% and 100% ethanol in water before air drying.

Samples were mounted on a flat stage with double-sided sticking tape, or in a back-mount stage and held in place by metal clips, before entry into the instrument. ToF-SIMS spectra were obtained for CSMs using an ION-ToF 5–100, and for adsorbed proteins using an ION-ToF 4–100. Both instruments were equipped with a Bi₃⁺ liquid metal ion gun with an energy of 25 keV. Total ion dose was maintained under 10¹² ions/cm² to ensure static SIMS conditions. The area of analysis was 500×500 µm, which was rastered in a random pattern across 256×256 pixels. Positive ion spectra were collected over the mass range of m/z 0–800. Spectra of negative ions were not collected as they yield less information from proteinaceous surfaces. [4–6,14,37]. At least three replicates were prepared for each sample type, with at least three spectra obtained from separate sites on each replicate sample, yeilding at least 9 spectra per treatment condition. Positive ion spectra were calibrated against peaks for CH₂⁺, CH₃⁺, C₃H₂⁺, and C₄H₃⁺ before further analysis.

2.6. Principal components analysis

Principal components analysis (PCA) performs an axis rotation to observe more clearly the variance in the data set along new axes called principal components [54]. Two resulting plots are vital for interpretation: scores and loadings. The scores are a projection of the original data points onto the new axis (principal components) and reveal the relationship between samples in a given principal component. Loadings describes the rotation between the original axis and new principal components axis. They reveal the variables that are responsible for the separation seen in the scores plot. PCA was performed using the spectragui package of the NESAC/BIO MVA Toolbox running in MATLAB R2012a (The MathWorks, Inc). Amino acid fragment peaks from the list previously used by Canavan et al. for analysis of ECM-surfaces [56] and adapted from work by Wagner and coworkers [37] were manually identified using SurfaceLab 6 (ION-ToF) and selected in overlaid spectra from each sample set, before adding to a peak list. Using the statistics function, peak intensities were compiled into a single tab-separated text file. This was imported into spectragui. Sample names were created to group repeated samples together. Spectra were normalized against the total ion intensity of the peak set and square-root mean centred before principal components analysis was performed. The calculated scores and loadings were saved to the MATLAB workspace and plotted using Graphpad Prism 6.

2.7. F-Actin and DNA visualisation

To visualize the F-actin cytoskeleton, samples were fixed in a solution of 4% w/v paraformaldehyde (PFA) in PBS for 30 minutes and permeablized with 0.1% Triton X-100 in PBS for 30 minutes, before incubation with 66 nM AlexaFluor488 phalloidin (green) or AlexaFluor546 phalloidin (red, both Invitrogen) in a solution of 1% w/v BSA in PBS for 20 minutes at room temperature whilst protected for light. Where DNA staining was also desired, 1 µg/ml Hoechst 33258 (Invitrogen) was included with the phalloidin solution. Samples were then washed three times in PBS before imaging by fluorescence microscopy.

2.8. DNA quantification

Freshly prepared CSMs, or mPOb cultures prior to decellularization, were scraped into RIPA buffer (Sigma) on ice for 10 minutes. Standards of calf thymus DNA (Sigma) were made up in RIPA (Radio-Immunoprecipitation Assay) buffer. Samples or standards were diluted as required in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 7.5) and 100 µl transferred to triplicate wells of black micro-plates. An equal volume of PicoGreen reagent (Sigma) diluted 1:200 in TE buffer was added. After incubation at room temperature for 5 minutes protected from light, fluorescence intensity was measured at 480/520 nm (ex/em) with a micro-plate reader (Infinite 200 Pro, Tecan).

2.9. Immunocytochemistry

Extracellular matrix components collagen I and fibronectin were visualized by immunocytochemical staining. Samples were fixed in 4% (w/v) PFA for 20 minutes. After washing with PBS, samples were permeablized for 30 minutes with 0.1% Triton X-100 in PBS for 30 minutes and washed again in PBS, before blocking for 30 minutes with 1% (w/v) bovine serum albumin (BSA) in PBS supplemented with 3% (v/v) donkey serum. The blocking solution was removed and the primary antibody added without further washing. Primary antibodies used were rabbit anti-mouse collagen I polyclonal (Millipore, AB765P) and sheep anti-human fibronectin polyclonal (R&D Systems, AF1918). The primary antibody was allowed to incubate overnight at 4°C. Subsequently, samples were washed with PBS (3×5 minutes) before incubation with the secondary antibody for 1 hour at room temperature. Secondary antibodies used were donkey anti-rabbit IgG AlexaFluor488 conjugate or donkey anti-sheep IgG AlexaFluor546 conjugate (Both Invitrogen). Fluorescence microscope images were taken using a Leica DM-IRB inverted microscope and Volocity Imaging Software (Improvision).

Fibronectin content was quantified by enzyme linked immunosorbent assay (ELISA). CSMs were scraped into RIPA buffer on ice for 10 minutes. Fibronectin concentration was determined using a fibronectin mouse ELISA kit (Abcam) according to the manufacturer’s instructions and standardized against known concentrations of mouse fibronectin.

Percentage area coverage of collagen I was determined by image analysis of representative images of samples stained by immunocytochemistry as described above. Representative images were thresholded manually in ImageJ to select positively stained areas only. Consistent thresholding parameters were used for all images. After conversion to a binary image, percentage area was calculated using ImageJ’s measure function, limiting the ROI to threshold of the binary image.

2.10. Histology

Collagens and glycosaminoglycans were visualized by staining with Sirius Red and Alcian Blue. Samples were fixed in 10% (v/v) neutral buffered formalin for 20 minutes and rinsed three times with tap water before staining for proteoglycans with Alcian Blue solution (0.5% w/v Alcian Blue 8GX, 1% v/v acetic acid distilled water) for 10 minutes. Samples were then rinsed in tap water for 1 minute before staining in molybdophosphonic acid (1% w/v in distilled water) for 15 minutes. Samples were rinsed again in tap water before staining for collagens with Sirius Red solution (1% w/v Sirius Red 3B, 30% v/v saturated picric acid in distilled water) for 1 hour. Samples were then rinsed and imaged immediately by brightfield microscopy.

Elution and absorbance measurement of dye from Sirius Red stained samples was used to quantify collagen content of CSMs. Formalin fixed samples were rinsed in tap water for 1 minute before staining in Sirius Red solution (1% w/v Sirius Red 3B, 30% v/v saturated picric acid in distilled water) for 1 hour. Samples were then rinsed in tap water and allowed to dry. Dye was eluted into a 50:50 v/v mixture of ethanol and 50 mM NaOH. Triplicate 100 µl samples were transferred to a 96-well multi-well plate and absorbance at 540 nm read using a micro-plate reader (Infinite 200 Pro, Tecan). Samples were calibrated against known concentrations of bovine collagen I dried onto multi-well plates before staining.

2.11. Total protein quantification

Total protein content of CSMs was determined by micro-BCA protein assay kit (Thermo). Samples were collected with a cell scraper into RIPA buffer on ice and diluted as required in PBS. The micro-BCA assay was performed according to the manufacturer’s instructions. Absorbance was measured at 526 nm using a micro-plate reader (Infinite 200 Pro, Tecan), and standardized against known concentrations of BSA in RIPA buffer.

2.12. Adhesion

mPOb cells were seeded at 10,000 cells/cm² in serum free basal medium onto cytokine-challenged CSMs; adsorbed proteins or untreated tissue culture polystyrene in 12-well plates. After 30 minutes, 2, 4 or 6 hours plates were washed three times with PBS to remove un-adhered cells. Remaining live cells were stained with 4 µm calcein AM (Invitrogen) in basal medium for 30 minutes at 37°C whilst protected for light. Cells were then washed in PBS before scraping into 200 µl of RIPA buffer on ice to lyse cells and release fluorescent calcein. Fluorescent intensity of 100 µl lysate samples was measured at 494/517 nm (ex/em) using a micro-plate reader (Infinite 200 Pro, Tecan). Results were standardized against known cell numbers prepared by serial dilution.

2.13 Morphological assessment

mPOb cells were seeded onto CSMs or adsorbed proteins at 5000 cells/cm² in basal medium and incubated at 37°C and 5% CO₂ for 24 hours before PFA fixation and staining for the F-actin cytoskeleton as described previously. Representative fluorescent microscope images of F-Actin stained samples were imported into ImageJ. Regions of interest (ROIs) following the perimeter of least 100 cells per sample image were manually selected using the wand (tracing) tool and measured using the Analyse Particles package. Cell spread area represents the area of each ROI. Aspect ratio is calculated as the ratio of major axis length to minor axis length for an ellipse fitted around the region of interest. Scatter plots of cell area versus cell aspect ratio were produced in R, displaying a density heat map to aid visualization of sub-populations.

2.14 Statistical analysis

All statistical analysis was performed in Prism 5 (Graphapad). Where only two groups were compared an unpaired t-test was used. Where multiple comparisons were made One-way or two-way ANOVA was employed followed by Sidak’s multiple comparisons test [59].

3. Results

3.1. Decellularization

CSMs were produced by seeding mPObs on tissue culture plastic multi-well plates or Thermanox coverslips and culturing for 14 days in an osteogenic medium. Subsequently monolayer cultures were decellularized in a two-step protocol using a low concentration alkaline solution and a nuclease treatment. Staining of undecellularized mPOb cultures and CSM for both DNA and the F-Actin cytoskeleton was performed, along with quantification of residual dsDNA and assessment of metabolic activity. Visible F-actin and DNA are eliminated by decellularization (Figure 1A) and dsDNA levels are significantly (Figure 1B, p<0.0001, t-test) reduced to 1.27 ± 0.154 µg/cm². Calcein fluorescence seen on CSMs treated with calcein AM are significantly reduced compared to live, confluent mPObs prior to decellularization (Figure 1C, p<0.0001, t-test.). Fluorescent intensity for CSMs is less than 1% of that seen for mPObs, suggesting almost total cessation of cellular metabolism by decellularization. As well as confirming removal of cellular components, retention of ECM components collagen I and fibronectin was assessed by immunocytochemical staining. In CSMs both collagen I and fibronectin are seen as a dense fibrous matrix that appear little disrupted compared to mPOb cultures before decellularization, with little change in spatial distribution and level of staining (Figure 1A). When observed by phase contrast microscopy CSMs appear as a dense, intact network and adhered across the surface (Figure 1A).

Figure 1.

(A) Representative phase contrast and fluorescence microscope images of CSMs before (left) and after (right) decellularization, stained for DNA (Hoechst), F-Actin (Alexa 488-Phalloidin conjugate), along with immunocytochemical staining for Fibronectin and Collagen I. Scale bars represent 200 µm. (B) Quantification of dsDNA in CSMs before (left) and after (right) decellularization by PicoGreen assay. (C) Assessment of cell metabolic activity by Calcein AM staining before (left) and after (right) decellularization. In B & C bars represent mean ± 1 SD, n=3. **** represents significant difference, p<0.0001, t-test.

3.2. ToF-SIMS analysis

Representative positive ion ToF-SIMS spectra of CSM surfaces, model adsorbed protein films and the underlying Thermanox substrate, shown in Figure 2, reveal a complex pattern of characteristic peaks. The spectra of the Thermanox substrate (Figure 2C) can be easily visually distinguished from those from CSMs (Figures 2D, 2E & 2F) and adsorbed proteins (Figures 2A & 2B). Thermanox is understood to be a custom poly(ethylene terephthalate) (PET) resin subsequently functionalized to contain nitrogen based groups to improve cell adhesion [60]. Many prominent peaks are known fragments of the PET backbone, including the PET-repeat unit (m/z 193) and the benzyol fragment C₇H₅O⁺ (m/z 104), whilst others, not seen in the spectra of pure PET previously reported [61] can be identified as nitrogen containing organic secondary ions, presumably resulting from proprietary modifications to PET, including negative ions m/z 26 and m/z 42.

Figure 2.

Positive ion ToF-SIMS spectra obtained from CSMs with (D & E) or without (C) cytokine challenge, adsorbed collagen I (A) or fibronectin (B) coatings, or as-received Thermanox (C). Prominent peaks are labeled with their mass/charge.

The spectra of the proteinaceous surfaces analysed are visibly distinct from that of the underlying Thermanox substrate. Positive ion spectra are dominated by peaks that have been identified as fragments associated with the amino acids building blocks of the proteins found at the surface such as m/z 30 (common to all amino acids, although most characteristic of glycine) and arginine (m/z 43, 59). The prominent peak at m/z 70 found in all proteinaceous spectra is not unique to a single amino acid residue and may result from arginine, leucine or proline. Whilst it is possible to identify known fragments of individual amino acids in the spectra, the complexity of ToF-SIMS spectra from proteinaceous surfaces makes it difficult to directly identify differences between samples of interest and associate differences with particular proteins. Therefore principal components analysis was used to identify differences in the intensities of known amino acid peaks in the positive ion spectra of CSMs and adsorbed proteins. Figure 3 presents scores and loadings plots for the first three principal components including all samples. The scores represent the relationship between samples, whilst the loadings reveal which peaks are responsible for the separation shown in the scores plot.

Figure 3.

Principal components analysis scores (left) and loadings (right) plots for principal component 1 (A & B), 2 (C & D) and 3 (E&F) from PCA of amino acid peaks selected from ToF-SIMS spectra obtained from as-received Thermanox coverslips (grey); adsorbed binary mixtures of fibronectin and collagen type I (blue through purple); and CSM surfaces with or without cytokine challenge (red = challenge at day 7, orange = challenge at day 12, green = no challenge). Scores plots (left) show individual sample scores along with boxes marking median, 25th & 75th percentiles and whiskers marking minimum and maximum values. Notable significant differences are highlighted (* p<0.05, **** p<0.0001, ANOVA with Sidak’s multiple comparisons test). Scores represent the relationship between samples In loadings plots (right) highly loading peaks are labelled with their mass and assigned amino acid identity. Loadings show which peaks are responsible for the separation seen in the scores plots.

Principal component 1 (PC1) separates the as-received Thermanox substrate from all CSM surfaces and model protein coatings, and describes 90% of the variance in the data set. The associated scores plot (Figure 3A) shows that PC1 gives high positive scores (0.39 ± 0.017, mean ± 1 SD) to sample points on the Thermanox substrate, whilst all other samples are scored negatively (Min: −0.087; Max: −0.023). The corresponding loadings plot (Figure 3B) shows five peaks with positive loadings greater than 0.1 and fifteen with negative loadings less than −0.1. Positively loaded peaks on PC1 include m/z 86 (C₅H₁₂N⁺) and m/z 100 (C₄H₁₀N₃⁺). Untreated Thermanox coverslips would not be expected to yield fragments of amino acid residues leucine/isoleucine (m/z 86), and arginine (m/z 100). Therefore these peaks are likely products of nitrogen functionalization of the Thermanox detected at similar mass/charge values. Most of the remaining peaks from the peak list load negatively on PC1 towards the protein containing CSMs and adsorbed proteins.

Principal component 2 separates CSM surfaces significantly (p<0.0001) from all model protein coatings and describes 6% of the variability within the data set. In the scores plot (Figure 3C), CSM surfaces are scored positively (Min: 0.030; Max: 0.053) and are significantly different from all other groups (p<0.0001). However, they are not significantly separated from one another (p>0.9). The Thermanox substrate has a low positive PC2 score (0.0081 ± 0.0038). Adsorbed binary mixtures of collagen I and fibronectin are scored in decreasing order moving from fibronectin scored neutrally (−0.0024 ± 0.0045), to collagen I with a high negative PC2 score. The corresponding loadings plot (Figure 3D) shows that the majority of peaks have positive PC2 loadings greater than 0.1 (21 of 45). Only 6 peaks have negative PC2 loadings less than −0.1 These include peaks at m/z 68 (C₄H₆N⁺ Pro) 70 (C₄H₈N⁺ Arg/Leu/Pro), 73 (C₃H₇N⁺ Arg), 115 (C₄H₇N₂O₂⁺ Gly), and 127 (C₅H₁₁N₄⁺ Arg).

Principal component 3 reveals significant differences between CSM surfaces with different cytokine challenge regimes. It describes 3% of the variability within the data set. In the scores plot for PC3 (Figure 3E), all CSM surfaces are significantly separated from one another (p<0.05). CSM surfaces produced with cytokine treatment at on day 7 score neutrally (−0.0040 ± 0.0062). CSM produced with cytokine treatment on day 12 have a more negative score (−0.019 ± 0.0077). CSM surfaces not treated with cytokines have the most negative scores of the three CSM surfaces (−0.026 ± 0.0035). The scores plot clearly and significantly separates the five collagen I / fibronectin model protein coatings in order from pure fibronectin with a high positive score (0.051 ± 0.0040), through mixtures moving from positive to neutral scores, to pure collagen I with a high negative score (−0.038 ± 0.0068). The Thermanox substrate scores neutrally (−0.0031 ± 0.0070). The loadings plot for PC3 (Figure 3F) highlights ten peaks that exhibit positive loadings greater than 0.1. These include peaks that can be identified with the amino acids threonine (m/z 68, 74), valine (m/z 72), tryptophan (m/z 130, 159, 170) and tyrosine (m/z 136). Six peaks have negative loadings less than −0.1. and may be attributed to amino acid residues glycine (m/z 30), arginine (m/z 43, 100), proline (m/z 68) and leucine/isoleucine (m/z 86). The peak at m/z 44 may be associated with alanine or cysteine.

Plots for principal component 4 (PC4) and above are not shown. PC4 describes 0.954% of the variability in the data set. No significant separation between sample groups was observed in this or higher PCs, suggesting that this and further PCs describe variability within sample groups or noise within the data set.

2.3. Histological and immunocytochemical analysis

Whilst the composition of naturally, in vitro produced extracellular matrices are too complex to identify all components using ToF-SIMS, it has been demonstrated here to be a useful tool to identify key compositional changes in CSMs. However it is important to compare conclusions drawn from ToF-SIMS spectra with conventional biochemical techniques.

Quantification of collagen mass in CSM surfaces through Sirius Red staining (Figure 4BA) reveals a significant decrease in collagens with cytokine treatment on day 7 of culture (p<0.0001, 2-way ANOVA with Sidak’s multiple comparisons test) from 9.75 ± 1.05 µg in untreated CSMs to 5.45 ± 0.582 µg in treated CSMs (mean ± 1 SD), a 44.1% reduction. No significant difference in collagen content of CSM was observed with cytokine treatment on day 12 of culture (p<0.05).

Figure 4.

(A) Quantification of collagen I surface coverage in cytokine-challenged CSMs assessed by image analysis of CSMs stained for collagen type I by immunofluorescence, representative images used are shown in (B). (C) Quantification of total protein in cytokine-challenged CSM surfaces by micro-BCA assay. (D) Quantification of collagen content in cytokine-challenged CSMs by elution of dye from Sirius Red stained samples. (E) Representative bright field microscope images of CSMs stained with Alcian Blue (GAGs, blue) and Sirius Red (Collagens, pink). (F) Quantification of fibronectin content of cytokine-challenged CSMs by ELISA. In all plots bars represent mean +/− 1 SD (n=6). Significant difference between challenged and unchallenged CSMs are highlighted (* p<0.05, **** p<0.0001, ns p>0.05, 2-way ANOVA with Sidak’s multiple comparisons test). In all images scale bars represent 200 µm.

Representative bright field microscopy images of CSM surfaces, stained with Sirius Red and Alcian Blue for collagens and GAGs respectively, show a dense web-like network of collagens in untreated matrices (Figure 4E). In matrices subjected to cytokine challenge at day 7 a notable disruption of the collagen matrix was observed, with areas of little or no staining. This does not appear to be seen with cytokine challenge at day 12. All CSM surfaces show positive staining for GAGs in a diffuse network of fibres. GAG deposition or structure does not appear to be affected by treatment with cytokines at either time point.

Representative fluorescence microscope images of CSMs stained for collagen I (Figure 4B) show that it is deposited in an interwoven, fibrous matrix, intact across the substrate. In CSMs not treated with cytokines more staining is seen, and appears almost continuous across the surface. By comparison, in CSM surfaces treated with cytokines there appears to be less staining. This is particularly notable in CSM treated at day 7 of culture, where areas without collagen I staining can be seen. Additionally, stained areas appear to be less organized than those seen in their untreated counterparts. Using image analysis to quantify area percentage positively stained using consistent thresholding (Figure 4A), a significant decrease in collagen I coverage is seen in CSM treated at day 7 compared to its untreated control (p<0.05, 2-way ANOVA with Sidak’s multiple comparisons test), with coverage reduced from 76.9% ± 13.1%, to 52.0% ± 10.70% (mean ± 1SD, n=3); a 32.4% reduction. No significant reduction in coverage is observed with cytokine treatment at day 12 (p>0.05).

Total protein content of CSM lysates were determined by micro-BCA (Figure 4C) and fibronectin was assayed by ELISA (Figure 4F). No significant differences in total protein concentrations or fibronectin levels between CSM surfaces were found (p>0.05 2-way ANOVA with Sidak's multiple comparisons test).

3.4. Cell adhesion

The adhesion of mPOb cells to CSMs and adsorbed proteins was assessed under serum free conditions (Figure 5). At all time points adhesion to pure collagen I was significantly lower than for either pure fibronectin or a 50:50 v:v mixture of collagen I and fibronectin (p<0.0001, 2-way ANOVA with Sidak’s multiple comparisons tests). Cell adhesion was also significantly lower than on untreated tissue culture polystyrene at 30 minutes and 2 hours. Adhesion to CSMs reached higher levels than for adsorbed proteins. The proportion of cells adhered to all CSM surfaces was significantly higher than all protein surfaces at 2 and 6 hour time points (p<0.001). However after 30 minutes, adhesion to all CSM surfaces was significantly lower than to pure fibronectin (p<0.001), although the differences were not always significant for other CSM/protein comparisons at this time point.

Figure 5.

Bar chart showing adhesion of mPOb cells to CSMs with or without cytokine challenge (left) and adsorbed ECM proteins (right). mPOb cells were seeded in serum free conditions at 10,000 cells/cm². After 30 mins, 2 and 6 hours non adhered cells were removed by washing in PBS and adhered cells quantified by staining with Calcein AM. Fluorescence of Calcein was normalized vs known cell numbers. Bars represent mean ± 1 SD, n=6. Significant differences in adhesion between treated and control matrices are highlighted (* p<0.05, ****, p<0.0001, ANOVA with Sidak’s multiple comparisons tests).

Comparing the influence of cytokine challenge during CSM production on subsequent mPOb adhesion, significant decreases were observed at 2 hour (p<0.05) and 6 hour (p<0.0001) time points between CSM surfaces exposed to cytokines at day 7 and their associated untreated comparator. Whilst significant, these differences remain small, with cytokine challenge at day 7 decreasing percentage adhesion at 6 hours from 73.48% ± 5.4 to 63.80% ± 4.9. No significant differences in adhesion were noted when cytokine challenge occurs at day 12 of CSM production. Furthermore, no significant differences were observed between CSMs not treated with cytokines at day 7 or day 12, indicating that these control groups are comparable (p<0.05).

3.5. Cell morphology

Representative fluorescence microscope images of F-actin stained mPOb cells seeded at sub-confluence onto CSMs or model proteins are shown in inserts of Figure 6. Clear morphological differences can be noted. Cells seeded onto CSMs appear more irregular in shape, more elongated and less highly spread than those seeded on proteins. The morphological characteristics of mPObs seeded at subconfluence onto CSMs or model proteins were assessed by image analysis to quantify cell spread area and aspect ratio. Figure 6 shows these as scatter plots. Cells on CSMs (all groups) can be seen to have significantly lower cell spread area than those on collagen or a 50:50 v:v mix of collagen I and fibronectin (p<0.001). Few cells on CSMs are seen with spread area >4000 µm². Little difference in the distribution of cell spread area is seen between CSMs exposed to different cytokine challenge regimes. However, some differences in the distribution of cell aspect ratios observed are noted. Whilst the majority of cells on all surfaces exhibit aspect ratios between 1 and 4, a notable number of cells on CSMs appear with elongated morphologies with aspect ratios between 4 and 8. Few cells are seen with such elongated morphologies on adsorbed proteins or TCP. Scatter plots also reveal a number of very highly elongated cells with aspect ratio >9 observed only on CSMs exposed to cytokine challenge (either at day 7 or day 12) during production. Cells with these morphologies are clearly separated from the bulk population, with very few cells observed with aspect ratios between 8–9. On CSMs not exposed to cytokines, few mPObs are seen with aspect ratios >8.

Figure 6.

Scatter plots showing cell area vs cell aspect ratio (long axis / short axis) for mPOb cells seeded onto cytokine challenged CSMs (A) or adsorbed proteins (B). Cells were seeded at 10,000 cells/cm2 for 24 hours before fixing in PFA and staining the F-Actin cytoskeleton with Alexa 488-Phalloidin conjugate. Cells were imaged by fluorescence microscopy before manually bounding >100 cells per sample in ImageJ to calculate cell area and aspect ratio. Heat map represents probability density (low = white, high = red). Insets show representative images used for image analysis. Scale bars represent 200 µm.

4. Discussion

4.1. Decellularization

First we validated the effectiveness of the decellularization protocol. A key characteristic of decellularized matrices is that cellular components have been removed, and extracellular components and structure are retained [62]. For cell-secreted matrixes to be useful surfaces they must also remain intact and adhered to the underlying substrate through decellularization, storage and subsequent use; as well as support the culture of relevant cell types [14,18].

In this study a decellularization protocol of low concentration ammonium hydroxide treatment to lyse cells followed by a nuclease treatment to remove residual nuclear material was used. Similar protocols have been previously reported to produce CSMs from endothelial [63] and osteoblast cells types [23,24]. One of the most widely used decellularization protocols for the production of CSMs also utilizes ammonium hydroxide treatment together with a detergent Triton X-100 [13,14,18,22]. Detergent treatment was not used in this study as decellularization appeared effective without its use. It has been suggested that detergent use may disrupt some ECM structures and components such as GAGs [62], whilst decellularization with ammonium hydroxide alone has been demonstrated to retain these components [24].

Staining for, and quantification of, DNA has been the most widely used and recommended assessment of decellularization effectiveness [62]. CSMs produced in this study showed no visible DNA with Hoechst staining (Figure 1A), and dsDNA levels quantified to be under 1.5 µg dsDNA/cm² (Figure 1B) The removal of other cellular architecture, such as the F-actin cytoskeleton [18,64], and cessation of metabolic activity [14] have also been considered by earlier studies, and provide further validation of decellularization. The protocol used in this study was shown to effectively obliterate cytoskeleton, as assessed by F-actin staining (Figure 1A). Additionally, elimination of intracellular esterase activity demonstrated effective cessation of cellular metabolism (Figure 1A).

Successful decellularization should not only remove cellular products, but also retain ECM components in a structural and conformational presentation close to that found prior to decellularization. Collagen I and fibronectin are key components of the ECM produced by osteoblasts, both in vivo in the osteoid matrix that precedes mineralization, and in in vitro cultures. Immunofluorescence imaging shows that both are retained in CSMs in similar levels and structures seen prior to decellularization (Figure 1A).

The matrix surface appears intact and continuous. Where damage or loss of matrix surfaces was observed, this tended to be seen at lower edge of the matrix, where contact during aspiration of medium or decellularization solutions was possible. As such, care was taken when preparing CSMs to avoid contacting the ECM surface. Others have reported similar challenges with matrix fragility [65]. Prewitz and coworkers recently showed that covalent binding of fibornectin to a glass substrate significantly increased matrix robustness [18].

To assess the effects of pro-inflammatory cytokines on mPOb cultures being prepared for CSM surfaces, two cytokine challenge regimens were used. The combination of 1 ng/ml IL-1β, 10 ng/ml TNF-α and 100 ng/ml IFN-γ, was selected on the basis of their reported synergistic effect, and prior studies demonstrating their effectiveness in promoting an inflammatory response in mPObs [58]. This is the first study of CSMs produced following cytokine exposure, although Mountziaris et al. (2008) had previously described the behavior of MSCs exposed to TNF-α and cultured on cell-derived matrices grown on acellular PCL/ECM scaffolds [66].

4.2. ToF-SIMS analysis

Having prepared and validated mPOb deposited CSMs from cytokine-challenged environments, ToF-SIMS analysis of both cytokine-challenged CSMs was performed. To test the hypothesis that collagen content is reduced by cytokine challenge, particularly at day 7, where there is time for matrix remodeling, CSM surfaces were compared to model proteins: specifically mixtures of collagen I and fibronectin adsorbed onto Thermanox coverslips. Fibronectin was selected as a non-collagenous protein, secreted by osteoblasts, critical for their differentiation [67] and survival [68], and demonstrated above to be present in the CSM surfaces. Fibronectin is readily identifiable by ToF-SIMS analysis due to its distinctive content of cyclic amino acids tyrosine and tryptophan [39].

From examination of the raw ToF-SIMS spectra it is difficult to distinguish between CSM groups and adsorbed proteins, being formed predominantly of the same 21 amino acids, whilst the as-received Thermanox substrate is easily distinguishable (Figure 2).

As with previous surface analytical studies of extracellular matrix samples [7,8,56,57], multivariate analysis is required to aid the identification of spectral differences between sample groups. To further simplify the number of variables (mass peaks) used for multivariate analysis, and eliminate peaks from other components such as the substrate, or residual cellular material, these studies examined the intensities of specific peaks known to be characteristic of amino acid fragments.

Using an amino acid fragment peak list previously used to study extracellular matrix surfaces [57], principal components analysis of cytokine treated CSMs revealed spectral differences between surfaces from different treatment regimes, with CSMs treated at day 12 being more similar to untreated CSMs than those treated at day 7 (Figure 3). Identification of highly loading peaks in PCA comparing both CSMs alone and to model protein coatings, using a subset of known amino acid related peaks, suggest that these differences are due to the reduction in relative collagen content with cytokine treatment.

The first principal component separates the as-received Thermanox substrate from the other samples (Figures 3A & 3B). As its spectral differences from the other samples are easily identifiable in the raw spectra it is unsurprising that the first principal component dominates the variance in the data set. With the peak list restricted to those thought to be characteristic of amino acids (i.e., supervised PCA), it is somewhat surprising that Thermanox yields secondary ions at these mass/charge ratios. However, proprietary nitrogen functionalisation of the PET backbone, for example by gas plasma modification, could result in nitrogen-containing positive ion fragments with masses, structures and charges similar to several amino acid fragments.

Further discussing PC1, consideration of the substrate is important for ToF-SIMS analysis of thin samples or those of incomplete coverage. Here these are considered by including the bare substrate in the PCA. Some previous studies of ECM components on a pNiPAM substrate have included known substrate peaks [57,69], whilst in later studies from the same group these were not considered [56].

In principal component 2, which separates CSMs from adsorbed proteins (Figure 3C), the large number of peaks loading positively, contributing to the high score of CSM surfaces compared to the few peaks loading negatively (Figure 3D), illustrates the complexity of CSM surfaces compared to single proteins or binary protein mixtures. CSM surfaces contain many components that have been secreted and remodeled by a population of cells in response to their environment. These may include structural ECM proteins such as elastin, or ECM proteins such as osteopontin or osteocalcin involved in osteogenesis and maturation from a collagen rich osteoid matrix to mineralized bone-like nodules. In comparison, the defined components and controlled adsorption conditions of the model protein coatings make them simpler surfaces than CSM, with known composition, and homogenous spatial distribution and orientation—all surface characteristics that may influence resulting spectra. Whilst binary mixtures of collagen I and fibronectin can be separated from one another, PC2 does not allow CSM surfaces to be distinguished according to cytokine exposure.

Comparing the scores in PC3 with those for model protein mixtures, it can be seen that CSM surfaces are collagen rich, displaying scores between those for 100% collagen I and a 75:25 Coll I:Fn ratio (Figure 3E). Untreated CSM surfaces score more towards the pure collagen sample than those treated with cytokines, which score towards fibronectin-containing protein mixtures. This suggests that untreated CSM shares more spectral similarities with pure collagen than those exposed to cytokine challenge. The neutral scoring of the Thermanox substrate indicates that it does not contribute to the peaks seen in the loadings plot for PC3 (Figure 3F).

In summary, principal components analysis of cytokine treated CSMs allows for the identification of distinct spectral differences between surfaces from different treatment regimes, with CSMs treated at day 12 being more similar to untreated CSMs than those treated at day 7. Identification of highly loading peaks in PCA comparing both CSMs alone and to model protein coatings using a list of known amino acid-related peaks suggest that these differences are due to the reduction in relative collagen content with cytokine treatment.

Assessment of the composition of complex protein-rich surfaces is challenging. Whilst proteomics based techniques allow detailed and accurate compositional analysis, and have been employed to study a range of decellularised systems including CSMs [21,34–36], they typically require sample digestion. ToF-SIMS, on the other hand, allows for samples to be examined as deposited – retaining their natural structure and conformations. Although ToF-SIMS can be used to readily identify components in single [37] and binary [39] adsorbed protein systems, it is difficult to quantitatively assess the composition of multicomponent protein surfaces. However it is possible to describe qualitative trends through comparison to reference protein spectra [38]. This approach has been taken by the few studies that have reported ToF-SIMS analysis of ECM-based systems. Two sequential studies performed ToF-SIMS analysis of decellularized tissues including oesophagus, bladder, small intestine, and liver. PCA using an amino acid peak list was able to distinguish samples both by anatomical location and method of preparation [8]. Separate PCAs of decellularized tissues and model proteins were qualitatively compared and associated with tissue origin, although conclusions on the identification of ECM components were not drawn [7].

In another series of studies, multivariate analysis of ToF-SIMS data allowed for poly (N-isopropyl acrylamide) (pNIPAM) coated tissue culture polystyrene after thermal cell-sheet lift-off to be distinguished from both the underlying substrate and surface modification [57]. By deriving principal components from model proteins alone to form a model into which data from surfaces of interest were projected, the residual ECM could be associated with laminin as opposed to fibronectin or collagen [56]. Whilst such surfaces may not be considered as true decellularized matrices as their purpose is to release intact and viable cell sheets from the surface leaving some residual extracellular components, rather than retain ECM structures for subsequent experiments, they nevertheless provide a useful comparator to the CSMs presented here.

Whilst the compositions of naturally, in vitro produced extracellular matrices are too complex to identify all components using ToF-SIMS, it has been demonstrated as a useful tool to identify key compositional changes. In order to confirm these conclusions it is important to compare them with standard biochemical techniques.

4.3. Validation of ToF-SIMS conclusions with biochemical techniques

Inspection of CSMs treated with pro-inflammatory cytokines at day 7 of culture when stained with Sirius Red and Alcian Blue (Figure 4E) or by immunofluorescence for collagen I (Figure 4B) showed that cytokine treatment disrupts the level and distribution of collagen content. However, no significant difference is seen with cytokine challenge at day 12. These results validate the conclusions reached from ToF-SIMS data that cytokine challenge at day 7 of matrix production leads to a reduction in collagen content compared to untreated matrices, whilst when cytokine challenge takes place at day 12 there is less change in collagen content.

No significant change in total protein content of CSM surfaces is seen with cytokine challenge at either time point (Figure 4C). A reduction of collagen content, yet maintenance of overall protein content, would suggest that levels of another non-collagenous ECM protein may be elevated. However, no significant increase in fibronectin content is observed by ELISA (Figure 4F).

Previous studies in primary calvarial cells [70] and MC3T3-E1 cells [71] showed that exposure to TNF-α alone could reduce collagen production by 30–35%. Stimulation with IL-1β alone decreases collagen deposition in both MC3T3-E1 cells [72] and the osteosarcoma cell line MG-63 [71]. These studies support the drop in collagen identified from ToF-SIMS analysis and the subsequent quantification of a 44.1% reduction in collagen content in CSM after exposure to a cocktail of pro-inflammatory cytokines at day 7 in our model.

Whilst several studies have suggested that IL-1β [72] and IFN-γ [73] induce a slight decrease in non-collagenous protein expression by bone cells, no significant decrease in fibronectin or total protein content was seen in CSMs produced after exposure to a combined cocktail of cytokines at early or late time points.

The mechanisms of action of IL-1β, TNF-α and IFN-γ have been well described. These pro-inflammatory cytokines are understood to influence the extracellular matrix through multiple pathways including the down-regulation of collagen I gene expression [74,75]. TNF-α in particular has also been shown to disrupt collagen fibril assembly [76]. Additionally cytokines including IL-1β and TNF-α are known to promote expression of a range of matrix metalloproteases (MMPs), many of which exhibit collagenolytic activity [3,77]. Whilst IFN-γ has been shown to down-regulate several MMPs, the synergistic effect these cytokines in bone cells has been demonstrated to result in a loss of collagen [78]. Reduced percentage area and increased heterogeneity of collagen I coverage may potentially be explained by the action of MMPs to actively degrade previously deposited ECM, although their presence or activity was not investigated here. Not all ECM components are down-regulated by inflammatory cytokines. The deposition of proteoglycans, includign decorin, biglycan and versican, which act to both structurally organise the ECM and are part of complex transduction networks for inflammatory signals, are increased in fibroblasts treated with IL-1β and TNF-α [79]. Decorin expression is increased in smooth muscle cells treated with IL-1β. Whilst not explored in this study, it may be hypothesised that cytokine induced proteoglycan synthesis is a factor in the maintenance of total protein levels observed in cytokine challenged-CSMs.

4.4. Adhesion of reseeded mPOb cells

As the initial interaction between seeded cells and a surface, adhesion plays an important role in subsequent cell behaviour. On proteinaceous surfaces this is mediated by the interactions of cell binding ligands with specific amino acid binding motifs found in the surface’s constituent proteins. The serum-free adhesion of mPOb cells was higher on CSM surfaces than on any single ECM protein or binary mixture. Whilst the adsorbed protein coats will present appropriate binding motifs, their homogeneous nature may mean that these sites are sub-optimally oriented or available for integrin interactions. CSMs, however, are deposited in situ in a complex structure of fibers and bundles determined by the secreting cell population, and therefore may present binding sites with better accessibility and availability. Additionally, CSMs contain a wide mixture of components, some of which may complement and assist in cell binding. Differences in the adhesion of mPOb cells to different adsorbed protein mixtures were observed (Figure 5). Consistently, adhesion was lowest on pure collagen I and highest on pure fibronectin, with 50:50 v:v binary mixture showing intermediate adhesion, corresponding to previously reported results [80]. This is likely due to differences in the rate of integrin binding. Binding to collagen I is predominantly through the α1β1 and α2β1 integrins, whilst fibronectin binding is mediated through the interaction of a range of integrins with the Arg-Gly-Asp (RGD) motif [81]. Mixed formulations may not perform as expected for a simple binary mixtures as collagen I contains a specific fibronectin binding site that regulates fibronectin fibril formulation, possibly leading to more complex interactions between the two components [82].

A small but significant decrease in cell adhesion is seen between CSMs exposed to cytokine challenge at day 7 and their control. A similar difference is not observed with cytokine challenge at day 12. This difference in adhesion may be due to several changes in the CSM induced by exposure to cytokines during deposition. The drop in collagen content and increase in the heterogeneity of distribution caused by cytokine-challenge demonstrated earlier, may result in decreased density of binding motifs in collagen I available for integrin binding; or the distortion of ECM structure, and potentially other related parameters including mechanical and topological properties, leading to concealed or obscured binding sites. It has recently been shown that cell adhesion is dependent on matrix fibre stiffness, orientation and available adhesive area [83]. Alternatively cytokine treatment may have damaged other ECM components that support cell binding such as fibronectin with its extensive RGD motifs. Although beyond the scope of the present investigation, these hypotheses may be tested in future studies through the use of integrin blocking or knockdown experiments.

4.5. Morphology of reseeded mPOb cells

When seeded at sub-confluence to minimize cell-cell interactions, mPOb cells were shown to present a smaller spread area and higher aspect ratio on CSM surfaces (Figure 6A) than on adsorbed proteins (Figure 6B), although many cells remained rounded (aspect ratio <3). A large number of cells were seen with more elongated aspect ratios. This corresponds with previously reported studies of cell morphology on CSMs [84].

Cell spreading and aspect ratio may be controlled by a range of surface parameters including but not limited to ECM composition, stiffness, and ligand dentistry; topographic roughness and alignment, ECM restriction, porosity, elastic behaviour and crosslinking. Regulated by cell-cell signaling, cell contractility, cell adhesions cell motility [85]. In CSMs, which present a fibrous surface rather than the homogeneous relatively flat surface of adsorbed proteins. Here alignment, both through preferential movement and spatial restriction, can be seen to result in elongation of cell shape. Additionally the presumably more heterogeneous distribution of binding ligands in CSMs may lead to contortion of cell shape as focal adhesions are formed.

On CSMs not exposed to cytokines, mPObs reach a maximum aspect ratio of ~8. However, on CMSs that did see cytokine challenge, both at day 7 or day 12, a notable cluster of small, highly elongated cells is noted. This cluster, representing <7% of cells, is clearly separated from the bulk population. Importantly, biochemical analysis could not identify a significant difference between CSMs with or without cytokine treatment at day 12. However, ToF-SIMS analysis was able to separate these matrices, indicating both the advantages of sensitive surface analysis techniques, and that that the mechanisms causing this difference in cell behavior may be more complex than a simple alteration of ECM composition and collagen distribution.

Considering the effects of cytokine treatment on osteoblast CSM previously discussed, it is possible to hypothesis as to the cause of this morphological cluster. Firstly this may be due to areas of decreased availability of specific binding ligands, either through decreased density, or deactivation through cryptic or unfolding mechanisms [86]. Equally, cytokine challenge may alter levels of ECM crosslinking, which may control folding and stretching in new ECM deposited by reseeded cells as they interact with the matrix [87]. Previous studies have demonstrated that decreased collagen density may yield increased cell aspect ratio and decreased surface area, increased number of projections of fibroblasts [88]. Decreased collagen coverage prompted by cytokine challenge may increase the size of lacunar area (areas without matrix), and hence between available collagen I binding sites. Supporting this hypothesis, Dumas et al. (2008) showed that mechanical stimulation during CSM production increased matrix lacunar areas, leading to a subsequent decrease of cell circularity [89]. In synthetic studies, altering spacing and size of protein islands may manipulate cell morphology by controlling integrin clustering and focal adhesion maturation [90,91]. Secondly, possible changes to ECM topology may lead to localized spatial restriction of cells as they align to features of the matrix. A variety of synthetic systems have shown how topography, both random and aligned, can direct cell morphology [92]. In a CSM system, Cukierman et al. showed that flattening matrices with a weight increased the cell spread area and reduced the aspect ratio of reseeded fibroblasts [93]. Thirdly, changes to ECM stiffness or elasticity may also come into play. The fibrillar nature of ECM can result in complex non-elastic, viscous, behaviour [94]. Prewitz and coworkers (2013) showed that CSMs from different culture environments exhibit significantly different Young’s Modulus [18]. It is likely that all of these mechanisms, and more, contribute to the effects observed. Although not investigated in this study, these factors could be considered through the comparison of cell behavior on CSMS to that on other quasi-vivo substrates such as matrigel, or synthetic substrates with controlled stiffness and topography. The possibility that changes in both morphology and adhesion are the effects of an inflammatory response to residual cytokines, entrapped in the matrix from prior cytokine challenge and retained post-decellularization, must be considered as a potential confounding factor. Previous studies have shown that growth factors including BMPs and VEGF may be retained in CSMs [23,24]. However, preliminary studies to measure the inflammatory marker nitric oxide on mPOb cells reseeded on cytokine challenged showed this to be not significantly different from baseline (data not shown), suggesting that this is of not a confounding effect on the conclusions drawn here.

5. Conclusions

This study has demonstrated the production of cell secreted matrices from mouse primary osteoblasts exposed to cytokine-challenge during matrix deposition. CSMs were found to be cell-free and retain relevant ECM components and structure. Analysis of ToF-SIMS spectra obtained from CSMs allows the different cytokine-challenge regimes to be differentiated. Comparison with adsorbed protein mixtures of collagen I and fibronectin suggest that the spectral differences between matrix produced with and without cytokines are associated with a loss of collagenous proteins. This is confirmed by histology and immunocytochemistry, demonstrating a drop in collagen coverage and mass in CSMs exposed to cytokine challenge mid-way through matrix deposition. These differences in ECM composition and structure were demonstrated to influence the adhesion and morphology of cells reseeded on cytokine-challenged matrices, which may direct subsequent cell fate and function.

This is the first study in which CSMs have been produced from a simulated inflammatory environment, and further demonstrates the utility of CSM systems to investigate ECM-based disease states. This work provides insights into the effect of cytokines involved in inflammation on ECM composition and structure; as well as subsequent cellular behavior. This study advances on previous surface analytical studies of ECM materials and represents the first investigation of CSMs by ToF-SIMS. Looking forward, it provides a basis for further application of surface analytical techniques to examine a range of decellularized systems.