Selecting the correct cellular model for assessing of the biological response of collagen-based biomaterials

Graphical abstract


a b s t r a c t
Accurate evaluation of the biological performance of biomaterials requires the correct assessment of their native-like cell ligation properties. However, cell attachment studies often overlook the details of the substrate-cell binding mechanisms, be they integrin-mediated or non-specific, and ignore the classand species-specificities of the cell adhesion receptor involved. In this work we have used different collagen (Col) substrates (fibrillar collagens I, II and III and network-forming Col IV), containing different affinity cell-recognition motifs, to establish the influence of the receptor identity and speciesspecificity on collagen-cell interactive properties. Receptor expression was varied by using cells of different origin, or transfecting collagen-binding integrins into integrin-null cells. These include mouse C2C12 myoblasts transfected with human a1, a2, a10 or a11; human fibrosarcoma HT1080 cells which constitutively express only human a2b1, and rat glioma Rugli cells, with only rat a1b1. Using these lines, the nature of integrin binding sites was studied in order to delineate the bioactivity of different collagen substrates. Integrin ligation was studied on collagen coatings alongside synthetic (GFOGER/GLOGEN) and Toolkit (Col II-28/Col III-7) triple-helical peptides to evaluate (1) their affinity towards different integrins and (2) to confirm the activity of the inserted integrin in the transfected cells. Thin films of dermal and tendon Col I were used to evaluate the influence of the carbodiimide (EDC)-based treatment on the cellular response on Col of different origin. The results showed that the binding properties of transfected C2C12 cells to collagens depend on the identity of inserted integrin. Similar ligation characteristics were observed using a1+ and a10+ cells, but these were distinct from the similar binding features of a2+ and a11+ cells. Recombinant human and rat-a1 I domain binding to collagens and peptides correlated with the cell adhesion results, showing receptor class-and species-specificities. The understanding of the physiologically relevant cell anchorage characteristics of bio-constructs may assist in the selection of (1) the optimum collagen source for cellular supports and (2) the correct cellular model for their biological assessment. This, in turn, may allow reliable prediction of the biological performance of bio-scaffolds in vivo for specific TE applications.

Statement of Significance
Integrins play a vital role in cellular responses to environmental cues during early-stage cell-substrate interaction. We describe physiologically relevant cell anchorage to collagen substrates that present different affinity cell-recognition motifs, to provide experimental tools to assist in understanding integrin binding. Using different cell types and recombinant integrin a1-I-domains, we found that cellular response was highly dependent on collagen type, origin and EDC-crosslinking status, as well as on the integrin class and species of origin. This comprehensive study establishes selectivity amongst the four collagen-binding integrins and species-specific properties that together may influence choice of cell type and receptor in different experimental settings. This work offers key guidance in selecting of the correct cellular model for the biological testing of collagen-based biomaterials.

Introduction
Natural extracellular matrix (ECM) contains a mixture of proteins and polysaccharides that display biochemical cues which influence cell behaviour. This composition determines the cellbinding affinity through specific interaction with integrins presented on the cell surface [1,2]. ECM components possess different adhesive motifs with diverse affinities towards a variety of cell recognition receptors. Despite this complex tissue composition, for many years, collagen (in forms including gels, scaffolds and membranes) has been a commonly used biomaterial due to its biocompatibility, biodegradability and low immunogenicity, together with its ability to form fibres with high tensile strength [2][3][4][5]. Collagen (Col), being the principal structural protein in all vertebrates, comprises a family of genetically distinct molecules with a common triple helix configuration of three polypeptide subunits known as a-chains [4,6]. These triple helices comprise a molecule of tropocollagen, the basic building block of collagen fibres. Tropocollagen molecules associate in a staggered fashion to produce collagen fibrils, which are strengthened and stabilized mainly by enzymatic and non-enzymatically catalysed covalent cross-links. The extent of these crosslinks is age-dependent and tissue-specific. The human genome contains 28 collagens and the corresponding proteins are made up of about 40 gene products, identified and described in varying detail [6]. Variations in collagen types are due to differences in the primary sequence and assembly of the polypeptide subunits, the lengths of the helix and the interruptions and terminations of the collagenous helical domains. The best known and the most abundant collagens are fibrillar collagens I, II and III, each containing different affinity cell-recognition motifs that support cellular activity mainly through their interaction with cell-associated integrins a1b1, a2b1, a10b1 and a11b1 [6,7]. Col I is a major ECM component and accomplishes both structural and cell adhesive roles in many vital organs and tissues [3,8]. Col II is the chief element in articular cartilage (approximately 60% of the dry weight of this tissue) [2,9] while Col III is an important component of reticular fibres, where it is commonly found alongside Col I [10], for example in skin and blood vessel walls. These collagens have been used, alone or in combination, for the design of bio scaffolds [2,3,5]. Col I is the most widely-explored option, owing to its physical and biological attributes, the ability to isolate it to high purity and its reasonable cost. Despite this, the addition of other collagens may be highly beneficial. For example, the introduction of Col III seems advantageous when engineering cellular supports for cardiac tissue replacement as this collagen, in native tissue, plays an important role by linking contractile elements of adjacent myocytes [10]. The structural diversity observed in different Col types affects their adhesive motifs which may in turn have impact on their cell-substrate interactions via integrins [6,7].
Integrins are transmembrane glycoproteins that represent a family of 24 heterodimeric signalling receptors each composed of a single aand b-subunit. These play a central role in mediating dynamic cell-cell and cell-extracellular matrix/substrate interactions. Integrins recognise a large number of similar motifs presented in the different types of collagens. They are unique, among adhesion molecules, as their adhesiveness is dynamically regulated through ''inside-out signalling," which in turn leads to ligand binding and signal transduction in the classical ''outsidein" direction [11][12][13][14]. The strength of cellular adhesiveness of an integrin is largely governed by the intrinsic affinity of the individual receptor-ligand interface, which is dynamically modulated by conformational changes. Of the four collagen-binding integrins, a1b1 and a2b1 have been studied for almost three decades whilst the properties of both a10b1 and a11b1 are still not fully explored [7].
All collagen-binding integrins are distinguished by the presence, within the a-subunit, of an inserted A-domain, termed an I domain. The I domain co-ordinates a divalent cation, Mg 2+ , in its metal ion dependent adhesion site, which is the principal site of interaction with collagens [7,13,15]. The crystal structure of integrin a2 I domain when interacting with Col triple-helical GFOGER motif has been resolved [15]. Fig. 1 schematically represents the position of this I domain in the a-subunit of a2b1 integrin and illustrates the complex formation between a collagen triple helical motif (GFOGER) and an integrin I domain via coordination with Mg 2+ . The mechanism of this interaction dictates that the carboxylate sidechain of the glutamic acid (E) residues in GxOGEx' and the presence in the culture medium of the divalent cation (Mg 2+ ) are critical for integrin-mediated binding. The removal of Mg 2+ cations by chelation with ethylenediaminetetraacetic acid, EDTA, effectively blocks this mode of interaction and is frequently used as a convenient method to deconvolute this native-like adhesion from non-integrin-based cell attachment [16,17]. Fig. 1. The crystal structures of integrin-a2 I domain and its Mg 2+dependent binding to GFOGER collagen motif was produced from pdb:1DZI. Schematic representation of a2b1 integrin with inserted I domain in the a subunit. Adapted from [13] As integrins a 1 b 1 and a 2 b 1 bind to collagen via these inserted I domains within the a-subunit, recombinant I domains offer the possibility to test for direct integrin-collagen engagement in a cell-free system, that is without interference/input of other cellsurface receptors and independently on the receptor density on the cell surface. As such this analysis can shed light on possible class-and species-specificities of integrin-substrate interactions.
Only the correct stimulation of cell-attachment mechanisms, through an intimate, biologically relevant cell-substrate ligation, leads to the further cell activity such as spreading and proliferation. Despite this, many cell attachment studies overlook the detail of the binding mechanisms, be they integrin-mediated or nonspecific. Moreover, the class-and species-specificities of the cell adhesion receptor involved in the direct substrate-integrin association are frequently ignored. In this study, the binding abilities of all four of the collagen-specific integrins have been probed by means of cells with diverse species-specificity (murine, human and rat) and which express different collagen-binding receptors: (1) C2C12 mouse myoblasts stably transfected with human a1+, a2+, a10+ or a11+ b1 integrins; (2) HT1080 human fibrosarcoma, expressing human a2b1, and (3) Rugli cells derived from a rat glioma that express rat a1b1. The original C2C12, nontransfected mother cells, which do not express collagen-binding receptors, were used as a negative control. For this study, adhesion assays were carried out on monolayer coatings thereby removing any possible impact of the substrate physical material properties [18].
To verify the activity of the transfected cells, studies of collagenbinding were carried out alongside investigation of binding to synthetic peptides containing cell-affinity motifs GFOGER and GLO-GEN, and on their Toolkit equivalents II-28 and III-7, respectively, which were obtained from libraries of overlapping triple-helical peptides. Both Toolkit samples are composed of an active guest sequence, encompassing the natural sequence of collagens II and III (Toolkit-II and III respectively) [7,[19][20][21][22][23]. A triple-helical peptide containing the sequence GFOGER (present in Col I, Col II and Col IV) is the highest affinity ligand for integrin a2b1, whilst GLO-GEN (an important ligand in Col III) is the best cell-recognition motif for integrin a1b1. The Toolkit peptides II-28 and III-7 specifically possess GFOGER and GLOGEN motifs, respectively, in their 29-residue section of primary collagen sequence, and as such, were selected for cell attachment tests, taking into account that screening of Toolkit II and III on collagen-binding integrins has previously been effectively used in our lab for the identification of cellbinding sites within collagen [7,23].
The influence of Col I source on its cell adhesive properties was studied on both monolayers and 2D films produced from insoluble bovine dermal and tendon-derived samples. Thin films were crosslinked (XL) with different dose of EDC (1-ethyl-3-(3-dimethylami nopropyl-carbodiimide hydrochloride) in the presence of NHS (Nhydroxy-succinimide), to evaluate the influence of the treatment conditions on the cellular response of different collagen substrates. Carbodiimide-based crosslinking constitutes one of the most successful chemical methods to modulate the mechanical stability and the dissolution resistance of collagenous materials. It is highly efficient, nontoxic, and by-products can be easily eluted [17,[24][25][26][27][28]. However, this procedure results in the consumption of carboxylate sidechains on glutamate (E) or aspartate (D) residues through their reactions with an adjacent lysine (K) e-amino group [24,29,30]. Both glutamate and aspartate are important amino acid components of numerous cell-recognition ligands in collagen molecules and their depletion, as shown in [7,17,31,32], may lead to the loss of integrin-substrate ligation.
Understanding the effect of the source and the type of collagen, before and after crosslinking treatment, on the biological behavior of collagenous materials may assist in the selection of the optimum components for the design of specific ECM-like devices to more closely match the properties of the tissues they aim to repair or replace. This study used cell lines that express single collagenbinding integrins as a tool to probe collagen ligation properties. The results show the importance of the receptor class-and species-specificities for cell-substrate interaction. Moreover, the work highlights the differential cell-recognition of different collagen types which is modified with ranging crosslinking densities and the importance of selecting the correct cellular model to assess the biological properties of protein-based biomaterials. subunits, were produced as described in [33]. The C2C12 cells stably transfected with integrin a10 were a kind gift from Dr. E Lungren-Akerland, Xintela AB, Lund, Sweden. Rugli (derived from a rat glioma) cells were a kind gift from Dr. J. Gavrilovic, University of East Anglia, Norwich, UK.
The human I domains from T141 to E335 for a1 and from S141 to E336 for a2 were expressed as recombinant N-terminal GST fusion proteins as detailed in [20,34]. In the same way, rat a1 I domain has been cloned from Rugli cells fused to GST, started from S226 to E421.

Synthesis of GFOGER and GLOGEN peptides
Peptides were synthesized using method described in [22,35]. Briefly, we used an Fmoc/tert-butyl solid phase strategy on a Liberty Blue TM microwave peptide synthesizer (CEM) on a 0.1 mmol scale, using HCTU as coupling reagent and DIEA as base. Peptides were assembled on Fmoc-Rink amide aminomethyl Tantagel resin (0.526 g, loading 0.19 mmol.g À1 , RAPP Polymere) to yield Cterminal amides. Fmoc removal was performed with piperidine in DMF (20% v/v). Successive amino acid couplings and deprotection steps were carried out in DMF under microwave radiation. After peptide assembly, resin beads were washed with dichloromethane (DCM) twice, methanol (MeOH) twice and DCM. Cleavage from the resin was performed over 2 h in 10 ml of a mixture of trifluoroacetic acid (TFA)/triisopropylsilane (TIS)/H 2 O 95/2.5/2.5 v/ v/v with 250 mg of dithiothreitol (DTT). The cleavage solution was concentrated and precipitated in 20 ml of cold diethyl ether. The white precipitate was filtered, washed with 10 ml of cold diethyl ether twice and redissolved in a H 2 O/acetonitrile (ACN) 95/5 v/v (0.1% TFA) mixture. The crude product was recovered after freeze-drying and purified by preparative reverse-phase high performance liquid chromatography (RP-HPLC) on a Perkin Elmer LC200 system equipped with a 10 lm Eurospher II 100-10 C18 H (Knauer, Berlin, Germany) with a linear gradient of ACN 0.1% TFA in water 0.1% TFA. Purified compounds were characterized by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry at the Protein and Nucleic Chemistry Facility (University of Cambridge, Cambridge, UK). Full sequence of GFOGER and GLOGEN peptides are GPC(GPP) 5 -GFOGER(GPP) 5 GPC and GPC(GPP) 5 GLOGEN(GPP) 5 GPC, respectively. The control triple-helical collagen-like peptide GPP10 (complete sequence, GPC(GPP) 10 GPC) was synthesized as described previously [19,22].

Monolayer coated surfaces
Collagens and peptides coatings were produced on the surface of Immulon 2HB 96-well plates (Thermo Scientific) by incubating 100 ml/well of 10 mg/ml solution or suspension (for insoluble collagen type I) in 10 mM acetic acid containing the appropriate proteins/peptides over night at 4°C. Bovine serum albumin (BSA, Sigma) and triple-helical-like sequences GPP10 were plated in triplicate to act as nonspecific background adhesion controls.

Films
First, slurries of insoluble collagen type I (derived from bovine tendon Col (S) or skin Col(D)) were produced by swelling at 0.5% (w/v) in 50 mM acetic acid at 4°C overnight and then homogenising on ice for 30 min at 13500 rpm using an Ultra-Turrax VD125 (VWR International Ltd., UK). Air bubbles were removed from slurries by centrifuging at 2500 rpm for 5 min (Hermle Z300, Labortechnik, Germany). Collagen films of $8 mm thickness were prepared by pipetting 100 lL/well of these slurries directly in Immulon 2HB plates (Thermo Scientific) and drying for 48 h in a laminar flow cabinet.
Films were cross-linked (XL) with carbodiimide (EDC) in combination with succinimide (NHS). An EDC concentration of 11.5 mg/ ml, with a molar ratio of EDC/NHS/COO -(Col) = 5/2/1 in 95% (v/v) ethanol, was taken as standard (100%) and was varied from 1 to 200%. These crosslinking conditions were selected on the basis of our previous work where the effect of the reducing of standard crosslinking concentration down to very low levels (up to 1% EDC) on different relevant material and some cell-interactive properties was elucidated [24,32]. After reaction for 2 h at room temperature, the films were washed thoroughly in deionised water (15 min Â 5) and dried in a laminar flow cabinet.

Cell adhesion and spreading
Cell adhesion in the presence of Mg 2+ (integrin-mediated) and EDTA (non-divalent cation specific) was assessed by colorimetric assay using a cytotoxicity detection kit (LDH), Roche, Cat. No 11 644 793001. This assay is based on the measurement of lactate dehydrogenase (LDH) activity release from lysed cells into the media.

Cells adhesion on surfaces and films
All cell lines were maintained in a humidified incubator with 5% CO 2 at 37°C in DMEM, containing 10% fetal bovine serum and 1% streptavidin/penicillin. Prior to cell adhesion experiments, cells were detached from the cell culture flasks with 0.05% trypsin/0.02% EDTA (GE Healthcare), washed and re-suspended in serum free DMEM.
Before cell addition, non-specific adsorption to the coatings/- Background adhesion was determined on BSA and GPP10 coated plates. Cell adhesion assays were performed in triplicate and values are reported as means ± standard deviations. For quantitative analysis calibration curves, OD (optical density) vs known cell concentration, were obtained for each experiment. These were constructed by taking 500 ml of cell suspension at a known cell density and subsequently diluting 32 to 64 times. Cells were separated by centrifuging, lysed by adding of 250 ll of buffer containing 2% v/v Triton X-100 in distilled water for 90 min at room temperature, vortexed and then 50 ml aliquots of each solution were pipetted by triplicate in the same well plate in which the corresponding cell attachment test was conducted. After that 50 ll of LDH detection substrate were added to the calibration series at the same time as to the coated wells and left until color had developed (from 10 to 30 min). The absorbance of this series was read under the same conditions/time as on coated wells. A linear regression was fitted to the known cell values, which was used to calculate the cell number in the experimental wells. The selection of the time point for absorbance reading (color development) has no impact on the percentage of adhesion calculated using calibration curves due to the same level of color development on calibration solutions and on tested wells.

Cell spreading tests
For spreading analysis, 100 ll of cell suspension at 1 Â 10 5 cells/ml containing either 5 mM Mg 2+ or 5 mM EDTA in serum free DMEM was added to BSA blocked surfaces for 90 min at 37°C/5% CO 2 . The cells were fixed by the addition of 9 ll of 37% (w/v) formaldehyde (final concentration 3.7%) directly to the cell media for 20 min at room temperature. The samples were washed with 3 Â 200 ll PBS then viewed using a LEICA DMI6000CS phase contrast microscope fitted with a LEICA DFC340FX camera. Cell spreading (percentage of spread cells versus total number of cells) was determined by analyzing 12 images per condition. A cell was scored as spread if it was phase-dark with cellular projections and a flattened morphology. Cells were scored as non-spread if rounded and phase-bright with no cellular projections as detailed in [18,36]. The percentage cell spreading was calculated by divid-ing the number of spread cells by the total number of cells present. Values are means of triplicate or quadruplicate measurements ± s tandard deviation.

Integrin I domain binding analysis
Prior to I domain binding, coated plates were blocked with 5% (w/v) BSA in binding buffer tris-buffered saline (TBS) (50 mM TRIS, 140 mM NaCl, pH 7.4, containing 1 mg/mL BSA) for 1 h at room temperature. Following BSA blocking, the samples were washed with 3 Â 200 lL of binding buffer then incubated in 10 lg/mL of the corresponding recombinant integrin a1 (human or rat) or a2

Statistical analysis
Unless otherwise stated all error bars indicate standard deviations. Statistical significance was determined with a student ttest with unequal variance where N/S indicates none statistically significant (p > .05), ⁄ indicates p .05, ⁄⁄ indicates p 0.01, ⁄⁄⁄ indicates p .001 and ⁄⁄⁄⁄ indicates p 0.0001. All statistical annotations in films indicate the statistical difference between the data point annotated and the 0% crosslinking values.

Cell adhesion and spreading on collagen coatings and peptides
Cell adhesion on coatings was carried out alongside synthetic (GFOGER and GLOGEN) and Toolkit, II-28 and III-7, peptides (1) to evaluate their affinity towards different integrins and (2) to confirm the activity of transfected cells. Short synthetic peptides possess one cell recognition ligand (GFOGER or GLOGEN) flanked by five GPP sequences and one GPC triplet on both the N and C terminal sides, which promotes a triple-helical configuration and ensures their attachment to the plastic well surface. The selection of these peptides was based on literature reports showing that Toolkit II-28 peptide and its derivative GFOGER are high affinity ligands for integrins a2b1 and a11b1 [19,37] while peptides III-7 and the corresponding GLOGEN were good ligands for a1b1 and a10b1 [20,22]. An inactive peptide containing only the flanking sequences GPP10 (GPC(GPP) 10 GPC) was used as a negative control in this study.
Magnesium-dependent and non-specific, cation-independent (EDTA) adhesion profiles for C2C12 cell line transfected with human a1+, -a2+, -a10+ or a11+ integrins on different Col coatings and peptides are presented in Fig. 2. It can be observed that all adhesion is Mg 2+ -dependent, and so integrin-mediated for all integrin transfected C2C12 cells. Original non-transfected C2C12 parent cells, which do not endogenously express collagenrecognition receptors, do not adhere to any substrate (data not shown). The ligation properties to Col I (S, D) for a2 and a11-positive C2C12 follow the affinity Col I > Col II > Col III (Fig. 2B and D).
For C2C12-a1+ cells ( Fig. 2A) there is a significant increase in adhesion on Col III in comparison with Col I and especially Col II. The level of adhesion of a10+ cells on Col II and Col III was similar but lower than on Col I (Fig. 2C).
The adhesion profiles on synthetic triple-helical (GFOGER and GLOGEN) and Toolkit peptides followed the published trends [7] thereby confirming that all transfected C2C12 cells correctly express the cloned integrin. Furthermore, very similar adhesion levels were observed on the short GFOGER and GLOGEN and their longer corresponding Toolkit peptides for all transfected cells. The binding properties of C2C12 with a1+ and -a10+ integrins appear most similar ( Fig. 2A and C). This was distinct from the comparable ligation characteristics of a2+ and -a11+cells ( Fig. 2B and D). The adhesion values on GFOGER and II-28 peptides were similar to those observed on full length Col I molecules for C2C12-a1+, -a2 + and a11+ transfected cells. C2C12-a1+ and -a10+ positive cells showed the highest attachment properties to GLOGEN and III-7 peptides which were comparable to the level of adhesion found on Col III, especially in the case of a1+ cells.
To test the activity of endogenously expressed rather than transfected integrins, cell adhesion assays were also carried out using model cell lines HT1080 (human fibrosarcoma) and Rugli (rat glioma). These each express a single, specific collagenbinding integrin; as a2b1 and a1b1, respectively. Fig. 3 shows typical adhesion profiles for Rugli and HT1080 cells ( Fig. 3B and D) together with those found for C2C12-a1+ and C2C12-a2+ cell lines at the same cell count ( Fig. 3A and C). Col IV was included as it is a known specific marker for cells expressing a1b1 integrin.
Comparison of adhesion results for C2C12-a1+ and Rugli, a1b1, ( Fig. 3A and B)  Cell adhesion studies were carried out for different cell seeding densities to assess the possible influence of this parameter on adhesion profiles. It was found that for all transfected C2C12 cells (with all Col-binding integrins) the level of cell binding was greater on Col I(S) than on Col I(D) at lower cell counts (below 1.5 Â 10 5 cells/ml). This effect can be seen in Fig. 3A and C for C2C12-a1 and a2 positive cells where adhesion values on Col I(S) are higher than on Col I(D) for cell count 1 Â 10 5 cells/ml. At cell densities higher than 1.5-2 Â 10 5 cells/ml, similar adhesion values were observed on both dermal and tendon collagens. HT1080 and Rugli cells were not sensitive to the collagen I source as the level of attachment was the same on Col I(S) and Col I(D) at any cell density studied (examples are given in Fig. 3B and D).
Adhesion dependence on the cell seeding density showed that Mg 2+ -dependent adhesion linearly increased with the initial cell concentration in a range from 0.5 to 2 Â 10 5 cells/ml but at higher values this linearity was lost reaching saturation at cell densities above 3-4 Â 10 5 cells/ml (Fig. 4). Adhesion percentages calculated from calibration curves using a cell density within the detection range are presented in Table 1. These results show that the adhesion percentage on Col I (S and D) coatings is higher for cells expressing integrins a2 (HT1080 and C2C2-a2+) than for cells with integrin a1. Col III and Col IV samples were more adhesive for cells with a1 integrin in comparison with a2 receptor. This data also highlight the differences in ligation properties between human a1 and rat a1 integrins: for example the adhesion percentages of C2C12-a1+cells on Col I (S and D) were noticeably higher than for Rugli cells. On the other hand, the absolute values of adhesion percentages (showed in Table 1) seem quite low which may be a consequence of the experimental conditions chosen for the cell adhesion assays: at room temperature, for relatively short durations, without CO 2 supply, and using a stringent washing regime to eliminate all loosely-bound cells. The level of adhesion will also reflect the level of receptor expression, not quantified here. As a result, these adhesion levels represent physiologically relevant cation-dependent integrin-mediated anchorage of the selected cell lines to the collagen substrates. This level of adhesion agrees well with our previous reports, e.g. [18].
Cellular spreading assays were performed on fibrillar collagen coatings using HT1080, Rugli and C2C12-a2 positive cells. C2C12 original non-transfected mother cells were employed as a negative control. The aim of this study was to determine whether the cell adhesion noted earlier leads to cell spreading as a result of the correct stimulation of certain signaling pathways after attachment. These assays were conducted in the presence of Mg 2+ as no cell spreading was observed in the presence of EDTA (data not shown). It should be mentioned that these tests were carried out in the absence of serum in the cell media to prevent cell adhesion to serum containing proteins such as vitronectin and fibronectin that can bridge between Col and cells. Images showing the cell shape in the presence of Mg 2+ for HT1080, Rugli, transfected C2C12-a2+ and ''blank" C2C12 are displayed in Fig. 5A. It can be observed that all cell lines with collagen-binding integrins (HT1080, Rugli and C2C12-a2+) are similarly spread on all Col coatings: almost all attached cells look flattened with a phase dark appearance. Spreading quantification (Fig. 5B) showed a high degree of cell spreading (between 95 and 85%) for all Col-based surfaces. Non-transfected C2C12 cells were not spread on any collagen samples; in the case of Col I coatings, where some attachment was observed, all cells were phase bright and rounded.

Cell adhesion on Col I films
Fibrillar collagen I samples used in films and in coatings were analyzed for their amino acid content (as described in our previous works [17,29]). Equivalent total protein content was observed in the as-received collagen preparations of insoluble Sigma and Devro collagen with 733 ± 13.1 lg/ml (S.D. n = 3) and 737.2 ± 35.5 lg/ml (S.D. n = 4), respectively. Films produced from these collagens were EDC cross-linked to establish the effect of collagen origin and the influence of variations in crosslinking conditions on cell behaviour on 2D samples. The same method for film preparation was used for both Col I samples. It was developed, validated (in term of collagen alignment, the surface structure, covering quality, thickness, etc.) and reported in our previous works [29]. For example, it was showed that the employed method provided films with the aligned fibre bundles of collagen within the plane of the films, homogeneously covering the underlying surface with similar topographic characteristics and thickness (in the range of 7-10 mm) for both non-treated and crosslinked samples. It was also demonstrated that EDC crosslinking did not induce any non-uniform changes in film morphology or protein distribution. Fig. 6 shows the adhesion profiles of C2C12 non-transfected mother cells (Fig. 6A) and C2C12 myoblasts transfected with all collagen-recognition integrins (Fig. 6B-E) on Col(S) and Col(D) films crosslinked with different concentration of EDC (from 0 to 200%). No adhesion of C2C12 ''blank" cells was detected on any film. This was expected as this cell line does not express any collagen-binding integrin (Fig. 6A). All adhesion of transfected C2C12 cells was Mg 2+ -dependent ( Fig. 6B-E) and the profiles share similarity in their dependence on EDC crosslinking dose: adhesion decreases with increasing EDC crosslinking. It can be observed that collagen origin influences the degree of adhesion, which, in addition, was sensitive to the integrin class involved in cell-substrate ligation. For example C2C12-a2 cells adhered similarly to both dermal and tendon Col samples (Fig. 6B) at all crosslinking concentrations. Conversely, for the other transfected cells (C2C12-a1, -a10 and -a11+)  Mg 2+ -dependent adhesion was higher on dermal, Col (D), in comparison with tendon, Col(S), at all XL conditions except 100 and 200% where no cell binding was observed (Fig. 6C-E). Adhesion patterns were similar for -a2+ and a11+positive cells: collagen ligation was not affected with up to 10% EDC crosslinking ( Fig. 6B and E). By contrast a1+ and a10+ C2C12 cell attachment  showed dose-dependent inhibition even with very low EDC concentrations (1% and 3% for Col (D) and Col (S), respectively, Fig. 6C and D). The adhesion of human fibroblasts HT1080 (a2b1) and rat glioma Rugli cells (a1b1) to Col I films with different crosslinking states was also tested (Fig. 7). For each cell line, the adhesion profiles were not dependent on Col origin as adhesion values on Col(S) and Col((D) were very similar for all crosslinking conditions used. However, the sensitivity of HT1080 and Rugli attachment to EDC crosslinking was very different. HT1080 (a2b1) attachment started to decrease from 10% EDC concentration onwards, showing the same profiles to those found for C2C12-a2+ (Fig. 6B). However,  for Rugli cells (rat a1b1) a dramatic drop in adhesion values was observed on both dermal and tendon Col films after crosslinking treatment with only 1% of EDC (Fig. 7B). This pattern is significantly different to that observed for C2C12-a1+ cells (Fig. 6C) suggesting different response of cells transfected with human compared to endogenous rat a1-integrins.  (Fig. 8A, B). Human a1 I domain ligation to GFOGER peptide and GFOGER-containing II-28 was very high and comparable with the level of adhesion found for the human a2 I domain. Human a2-GFOGER interactions are considered high affinity [7], therefore this human a2 control was included to allow direct comparison showing that human a1, but not rat a1, has a similar affinity to GFOGER as human a2 I domain ( Fig. 8A and C). In contrast, almost no adhesion was detected for rat a1 I domain on GFOGER and on II-28 where the absorbance signal was near to the background level and to that in the presence of EDTA, Fig. 8A, B and C. Integrin a1 human I domain and rat a1 I domain show different affinities of ligation to different collagen types (Fig. 8D). The binding of human a1 I domain on Col I, II and III was considerably greater than that of the rat a1 I domain to the same collagen samples (Fig. 8D). However, the level of adhesion on Col IV was similar for both human and rat a1 I domains.

Adhesion and spreading on coatings
Monolayer coatings allow the detection of integrin recognition sequences in collagen substrates by evaluating the affinity and availability of their biological cues towards variety of cellular receptors. These coatings were prepared from diluted (well below their saturation limits), highly homogenised solutions/suspensions of the corresponding collagen substrates. By using an appropriate coating concentration, this assessment is independent of the Fig. 8. Attachment of a1 I domain (human and rat) on different collagens and selected synthetic and Toolkit peptides: A -rat and human a1 I-domains binding to selected Toolkit and synthetic peptides. EDTA controls block all binding to peptides (data not shown); B -a 1 I-domain binding to GFOGER, GLOGEN and II-28 in presence of Mg 2+ and EDTA; C -rat a1 vs human a1 and a2 I-domains binding to GFOGER; D -human vs rat a 1 I-domains binding to different collagens. material bulk properties since only single molecule layers are deposited onto the solid cell culture plastic, so creating a surfacebound assembly of integrin recognition sequences typical to each collagen type. Three fibrillar collagens (types I, II and III) and the network Col IV, known as a selective ligand for a1b1 integrin [7], were chosen for this study to examine the cell-recognition sequences that support integrin ligation [6,7]. As these collagens possess different solubility, the effect of this parameter on their surface immobilization was investigated. Enzyme linked immunosorbent assays (see supplementary data on collagen detection) show identical detection of collagen on coatings derived from insoluble Col I suspensions or a soluble Col I solution. This demonstrates that when incubated at excess, the solubility of the starting materials does not play an important role in the total surface coverage, which, in turn, provides the basis for comparing the response of cells expressing defined receptors to the different collagens. Through this analysis, the importance of the structural and sequence diversity on their interaction with integrin receptors was determined. In our previous work, we have shown that macromolecular triple-helical structures of Col I may display differences in their fiber architecture (fiber lengths, width, entanglement), when derived from different tissues (dermal and tendon). In turn, these structural modifications induce changes in the material properties of 3D scaffolds produced from these precursors [24]. Here, we extend upon this to determine whether these changes also impact on the biological function of this protein. C2C12 cells transfected independently with each of the four collagen-binding integrins showed that at low cell counts (within the linear concentration dependent zone) tendon collagen seems to be more celladhesive than its dermal equivalent. This may be a consequence of differences in the exposure of cell-adhesion sequences in these samples. Furthermore, as these collagen preparations possess different physical properties such as fibre diameter and stiffness, the differential cell adhesion could be due to surface-induced conformational and/or orientation changes, as observed previously for gelatin-coated surfaces [18].
The differential cell binding to collagen type I, II, III and IV may be accredited to the differences in the specific GxOGEx' motifs and their corresponding affinities towards different Col-binding integrins. The distribution of GxOGEx' sequences across fibrillar collagens have been previously identified (Table 2) [7] by using homotrimeric collagen-derived triple-helical peptide libraries called Toolkits. These assays also established that the affinity order in GxOGER ligands towards a2b1 integrin ligation showed the trend where was x = F > L ! R > M > A [7], indicating that GFOGER (found in Col I and Col II) is the highest affinity ligand for this receptor. Our results agree with relative affinities. We observed the highest adhesion values of C2C12-a2+ and HT1080 cells, both expressing a2b1 integrin, on Col I/Col II coatings and on GFO-GER/Toolkit II-28 peptides (Fig. 2 and 3C and D and Table 1). C2C12-a11+ myocytes showed very similar binding properties to a2-positive cells on collagen monolayers, which followed the trend Col I > Col II > Col III, and on collagen Toolkit peptides (Figs. 2 and 3C and D). This suggests that GFOGER is a high affinity ligand for integrin a11. A high level of C2C12-a1+ and Rugli (a1b1) cell adhesion on Col III/IV and GLOGEN/Toolkit III-7 peptides is consistent with the presence of the GLOGEN motif within all of these substrates, as this motif has been identified as a high affinity sequence for integrin a1b1 [7,22]. Col IV promoted high levels of adhesion for cells containing integrin a1 (Fig. 3A, B) which agrees with the definition of this network collagen as a selective ligand for a1b1 integrins [7]. It has also been reported that the affinity of GLOGEN for integrin a1b1 is comparable with that of GFOGER [7] which is in accordance with the high attachment levels to Col I and GFOGER/II-28 peptides we observed on C2C12 transfected with human integrin a1 (Figs. 2 and 3A). Conversely Rugli cells, expressing rat integrin a1, showed very low adhesion levels on GFOGER which may explain the low attachment to Col I and Col II as these contain GFOGER as their primary cell-binding motif (Fig. 3B). Differences in the cellular response of Rugli and C2C12-a1+cells cannot be attributed to differences in the receptor density on the cell surface as both cell lines show very similar integrindependent adhesion to Col IV and on Toolkit III-7 peptides. Instead this points to different affinities between rat and human integrin a1 towards certain cell recognition sequences such as GFOGER.
Therefore this suggests that not only integrin class, but also the species-specificity, may affect collagen-receptor interactions. Thus we highlight the importance of using humanised test systems for cell adhesion analysis.
The binding properties of C2C12 a1+ and a10+ were very similar (Fig. 2) except for on Col III where C2C12-a1+ cells showed greater attachment than a10+ cells. This may suggest that either Col III contains auxiliary motifs alongside GLOGEN, which possess different affinities towards integrins a1 and a10, or alternatively that there are differences in the display of GLOGEN sequences on Col III towards a1 and a10 receptors.
The similar adhesion levels observed on short synthetic peptides and their longer Toolkit equivalents seem in contradiction with the predicted density of cell recognition sequences due to the different masses of these peptides. Each peptide contained a single active collagen-recognition motif (GFOFER or GLOGEN); however the shorter peptides contain 42 amino acids (AA) whilst the larger Toolkit counterparts are considerably longer (63AA). As a consistent mass of peptide was applied to the tissue culture plate surface, this implies that the active ligand density of the short peptide might be 50% higher that of the longer Toolkit peptide, assuming uniform monolayer coating. However, this apparent difference in density does not invoke a corresponding increase in cell adhesion. This may be due to a number of reasons. Firstly, it is possible that cell binding is saturating on the longer peptide and so the higher density of the shorter peptide cannot elicit a further effect. Secondly, it could be that the larger, native-derived Toolkit peptides have a higher affinity for cell integrins in comparison to the shorter sequences due to differences in flexibility and in conformation of their triple-helical structures. There are differences in transition temperatures (T m ) between short peptides and Toolkit peptides. The T m of GFOGER is almost 10°C higher than that of Toolkit peptides: Tm of GFOGER = 56°C [35] and T m of II-28 and III-7 were 46°C and 45°C, respectively (data not published). It is possible that this reflects a tighter, more compact triple-helical configuration of GFOGER than the longer Toolkit peptides. Native collagen approximates to 10-fold axial symmetry, whereas the GPP-containing flanking sequences adopt 7-fold symmetry. The transition between the two is not abrupt [13,15], so that the helical twist of the GFOGER motif differs slightly depending on its local setting, and may influence the position and exposure of its amino acid residues (especially E) to integrins. It is also possible that Table 2 Integrin recognition sequences in fibrillar collagen (data from [7]).

Type of fibrillar collagen
Integrin recognition sequences Single-letter amino acid code Collagen I GFOGER, GLOGER, GROGER, GMOGER Collagen II GFOGER, GLOGER, GMOGER Collagen III GROGER, GMOGER, GAOGER, GLOGEN the longer Toolkit peptides retain their native-like triple-helical conformation upon surface immobilisation resulting in greater activity towards cell receptors. Finally, the Toolkit peptides include primary sequence on each side of the active motif which may contribute to better recognition by native integrins. In accord with this theory, in the case of Rugli cells, with rat integrin a1, lower adhesion levels on shorter in comparison to the corresponding longer Toolkit peptides was noted. It is possible that the rat integrin a1 is more sensitive to the collagen conformation, indicating that a native-like structure of adhesion motifs is critical for their activity. This is consistent with numerous reports showing that cell adhesion to GxOGEx' motifs is dependent upon a triple-helical conformation. This theory may explain, to some extent, the similar level of adhesion found on complete collagen molecules compared with the corresponding peptides which, due to their differential mass, would possess different surface densities (almost 15-25 times higher on peptides than on full collagen molecules). It is also likely that adsorption efficiency to the well surface of large collagen molecules differs from the comparatively short synthetic and Toolkit peptides which may affect the exposure of their celladhesive motifs. To finish, it is plausible that changes in the topology between the intact collagen molecules and peptides coatings, for example large fibrils on fibrillar collagens vs small structures on both types of peptides, may result in altered integrin ligation. Cellular spreading analysis was performed to determine whether cation-dependent integrin-mediated cell adhesion could elicit flattened native-like cell morphology. Spreading of HT1080, Rugli and C2C12-a2+ cells, all expressing collagen-binding receptors, was high on all Col-based surfaces. This indicates that these substrates are capable of stimulating cell-signalling mechanisms that lead to spreading. As such it appears that integrin-mediated cell engagement with collagen can induce cell spreading via both integrin a2 and a1. Moreover this is not sensitive to the human or rat cell origin, showing that they interact with collagen substrates in a physiologically relevant manner leading to a very similar level of spreading that is independent on the collagen type and tissue source.

Adhesion on films
Adhesion tests were carried out on 2-dimensional films obtained from insoluble tendon and skin derived Col I and crosslinked with different doses of EDC. As Col I continue to be a predominant choice for the design of protein-based biomaterials the understanding of its adhesive characteristics, when derived from different tissues, and analysed with a diversity of cellular models, as presented here, may assist in the selection of optimum structural components for collagen-based cell supports.
EDC crosslinking is one of the most effective treatments for improved structural stability, resistance to dissolution and tailored mechanics of collagenous matrices with implications for their clinical use as tissue engineering biomaterials. As EDC-mediated bonding impacts on cell-substrate interactive properties it is important to balance the need for structural integrity and mechanics against bioactivity in each specific clinical case. In our previous works we investigated the effect of reducing the EDC concentration down to very low levels (up to 100 times dilution, 1% EDC, with respect to standard condition) on the physical and cell-interactive properties. We demonstrated, for example, that the EDC/NHS crosslinking dose influences both the level and the mode of cellular engagement to tendon-derived Col [32]. Cell attachment changed from predominantly integrin-mediated in the absence of EDC treatment, to divalent cation-independent with high EDC crosslinking concentrations (between 100 and 500%). In this work, we further assess the influence of EDC crosslinking on the ligation of cells with single, defined, integrin populations to collagen substrates. Therefore, we were interested in the effect that the collagen source induces on cell adhesion and, additionally, if cell adhesion sites on dermal and tendon-derived samples were equally altered by carbodiimide-mediated bonding. In contrast with the earlier study [32], here, through the use of specific experimental conditions, we focused on only physiologically relevant cation-dependent integrin-mediated anchorage of the four different collagenbinding integrins to collagen substrates. In particular, we introduced modifications to the cell adhesion methodology compared with that employed in the cited work [32]. This involved limiting the crosslinking concentration to 200% EDC (maintaining consistent the dilution levels up to 1%) and conducting the cell adhesion assay on the bench, at room temperature and without CO 2 . In addition, a more stringent washing regime was used to ensure the complete removal of loosely-bound or non-Mg 2+ -dependently attached cells from material surfaces. Through these modifications the noncation dependent adhesion was limited, allowing us to examine only intimate, strong cell-substrate interactions. As such, this minimised lower affinity non-native-like collagen binding. Using this experimental approach, our results showed that crosslinking strongly decreases integrin-mediated cell binding of all cell lines via all four collagen-binding integrins (a1b1,a2b1, a10b1 and a11b1) to all films. The shape of the adhesion curves (A 490 nm vs EDC %) for HT1080 and C2C12 transfected cells resemble that found for the amine group content on Col (S) and Col (D) as a function of EDC concentration [24]. According to the well-established crosslinking mechanism, the number of amine groups involved in the EDC-promoted bonding is proportional to the quantity of carboxylic moieties on the AA residues (such as glutamate and/or aspartate) participating in the linkage. These same carboxylic groups are critical for cation-coordination between I domains of integrins and collagen and so EDC crosslinking, by consuming important chemical elements of the cell recognition sequences in collagen, depletes cell attachment to highly crosslinked collagen. These results are in agreement with our previous reports [18,24,29] where we hypothesise that carbodiimide treatment of collagenous materials significantly decreased the content of important carboxylic groups on glutamate and aspartate amino acid residues.
The adhesion of transfected C2C12 cells on films showed that the collagen origin substantially influenced the adhesion profiles. Comparison between the transfected lines showed that this adhesive profile was sensitive to the class of integrin involved in cellsubstrate ligation. Contrary to the ligation levels observed on coatings, the adhesion values for almost all receptors, with the exception of a2, were higher on dermal collagen films than on their tendon-derived equivalents. This lower adhesion observed on dermal collagen monolayers could be attributed to a number of factors including the surface protein density, differences in collagen preparation from each source, or possible conformational changes due to surface interaction. Our films were $8 mm thick, meaning that the surface cell-recognition ligands are separated from the underlying plastic by a large number of collagen molecules. As such, it is highly unlikely that any surface-induced alteration of the collagen structure would be propagated to the surface and so influence the binding of cells to Col films. By contrast, on monolayers, the plastic could influence the structure of collagen molecules exposed to cells. Interestingly this higher adhesive level on Col I(D) films was noted for a1+, a10+ and a11+ but not a2+ C2C12 cells. As integrin a2 has a different affinity for cell-binding motifs in collagen, this may point to differences in the number or display of certain triple-helical motifs on dermal versus tendon collagen. It is also possible that bulk material properties may modulate, to some extent, integrin engagement to the films but not the monolayers. For example, we have found that the resistance to compression of matrices obtained from tendon collagen is higher than that of dermal (unpublished results). Therefore the cells may be responding differently to the mechanical environment in conjunction with the specific integrin binding sequences involved. This is in accordance with reports suggesting that integrins play a significant role in propagating mechanical cues to cells even at an early stage of cell-substrate interaction [1].
Differences were noted in the adhesion profiles of cells transfected with different integrins as a function of EDC-dose. This may point to differential sensitivity to EDC crosslinking of each of the glutamate-containing GxOGEx' cell binding motifs. Each motif has a different integrin binding affinity, and so an unequal alteration of these cell recognition motifs by EDC may, in turn, alter the specific integrin binding profile of collagen I. EDC crosslinking up to 10% EDC concentration did not lead to an appreciable decrease in the adhesion of a2+ and a11+ positive C2C12 cells.
These lines bind predominantly to high affinity GFOGER motifs, suggesting that the amount of unmodified GFOGER in low crosslinked samples was still sufficient to maintain cell adhesion, despite consumption of some glutamate (E) residues. Conversely, the EDC-induced decrease in a1 and a10 mediated cell binding to other GxOGER sequences, for example GLOGER, seems to be more sensitive to crosslinking. This may explain the observed drop in adhesion values for a1+ and a10+ cells even at low, 3% EDC concentrations. This effect is even more evident in Rugli cells with rat integrin a1 than for human integrin a1 containing C2C12 cells suggesting species-dependent affinity modulation of this receptor.

Attachment of -a1 human I-domain and -a1-rat I domain
Integrins a1b1 and a2b1 bind to collagen via inserted Idomains within the a-subunit (Fig. 1, [7]). Therefore, isolated recombinant human and rat integrin a1 I domains were used to assess integrin-collagen engagement in a cell-free system. The results agreed with the differences found in the adhesion of cells expressing human and rat integrin a1b1. They show that the difference in C2C12-a1+ and Rugli cell adhesion (with human and rat a1 receptors, respectively) in response to the dose of EDC used for collagen crosslinking was due to differential I domain binding.
The observed lack of rat integrin a1 I domain engagement with GFOGER and with II-28 peptides ( Fig. 8A and B) is consistent with low adhesion of Rugli cells on Col I/II coatings and these peptides.
By contrast, C2C12 cells transfected with human a1 showed high adhesion levels on the same coatings which agree with the high affinity of human integrin a1 I-domain to GFOGER collagen sequences. In general, we have showed that while human integrin a1 I domain ligates to a wide variety of collagen derived triplehelical ligands, rat I domain shows high specificity almost exclusively to GLOGEN-containing sequences, that is synthetic GLOGEN and its Toolkit equivalent III-7 (Fig. 7A). These motifs are not present in Col I (see Table 2) and so rat integrin a1 binding to Col I is dependent upon lower affinity motifs. Therefore even a small decrease in the availability of these lower affinity sites, for example through EDC crosslinking, may lead to the significant loss of collagen binding capacity of cells expressing rat integrin a1. At the same time, cells which contain the human a1 I domain bind to a wider variety of similar triple-helical sequences including the GFO-GER motif in Col I (Fig. 8A) and were much less sensitive to crosslinking. Presumably this is due to the higher affinity of GFO-GER for human but not rat integrin a1. Additionally the different ligation behaviour of human and rat integrin a1 I domains may also explain the considerably elevated adhesion of cells with human a1 on Col I and II as this can occur through high affinity GFOGER motifs. Together these results show that both the integrin class and species-specificity dictate the affinity of these receptors to cell recognition motifs in different members of collagen family.

Conclusions
Integrins play a vital role in determining the cell response to environmental cues during the early stage of cell-substrate interaction. Therefore a detailed understanding of integrin binding to collagens, as widely used in biomaterials fabrication, is pivotal for endowing these materials with appropriate bioactivity. Here, we found that the cellular response to both the collagen type and origin is highly dependent on the specific integrin type and species being studied. This is due to integrin I domain interactions with collagen as shown using recombinant human and rat integrin a1-I domain binding assays. As for whole cell assays, these exhibited receptor class and species specificities of collagen interactions. Combining the results of cell attachment on differently crosslinked collagen films with those obtained in integrin a1-I-domain bindings studies, the biological response of human and rat cells (with the same a 1 integrin) to changes in the collagen type and origin and the EDC crosslinking status were elucidated. We have shown that EDC treatment decreases the overall amount of cell recognition sequences, whilst simultaneously altering the adhesive ratio among the different motifs on collagen molecules which further affected the specific integrin class that ligates to these collagen substrates after crosslinking. This comprehensive study of all four of the collagen-binding integrins gives key guidance in selection of the correct cellular model for the biological testing of biomaterials. For example we clearly show that rat models, which are used extensively for biological studies, may give different results from human-based systems. In summary, our data show that the collagen precursor should be carefully chosen dependent on the cell type to be incorporated and on the tissue to be replaced. This will assist in the selection of optimal materials for the design of cellular supports for different TE applications.