Human mesenchymal stromal cells in adhesion to cell‐derived extracellular matrix and titanium: Comparative kinome profile analysis

The extracellular matrix (ECM) physically supports cells and influences stem cell behaviour, modulating kinase‐mediated signalling cascades. Cell‐derived ECMs have emerged in bone regeneration as they reproduce physiological tissue‐architecture and ameliorate mesenchymal stromal cell (MSC) properties. Titanium scaffolds show good mechanical properties, facilitate cell adhesion, and have been routinely used for bone tissue engineering (BTE). We analyzed the kinomic signature of human MSCs in adhesion to an osteopromotive osteoblast‐derived ECM, and compared it to MSCs on titanium. PamChip kinase‐array analysis revealed 63 phosphorylated peptides on ECM and 59 on titanium, with MSCs on ECM exhibiting significantly higher kinase activity than on titanium. MSCs on the two substrates showed overlapping kinome profiles, with activation of similar signalling pathways (FAK, ERK, and PI3K signalling). Inhibition of PI3K signalling in cells significantly reduced adhesion to ECM and increased the number of nonadherent cells on both substrates. In summary, this study comprehensively characterized the kinase activity in MSCs on cell‐derived ECM and titanium, highlighting the role of PI3K signalling in kinomic changes regulating osteoblast viability and adhesion. Kinome profile analysis represents a powerful tool to select pathways to better understand cell behaviour. Osteoblast‐derived ECM could be further investigated as titanium scaffold‐coating to improve BTE.

bioactive molecules and by transducing mechanical signalling (Discher, Mooney, & Zandstra, 2009;Guilak et al., 2009;Hynes, 2009;Reilly & Engler, 2010). Bone matrix is composed of collagen and noncollagenous proteins to maintain bone flexibility, whereas its stiffness is achieved by hydroxyapatite crystals, which makes bone a peculiar type of connective tissue (Alford & Hankenson, 2006;Gentili & Cancedda, 2009). Because of its composition, bone ECM is essential for the structure and the strength of the bone and it also actively participates in bone formation and bone metabolism, by regulating mineralization and modulating growth factor availability (Alford & Hankenson, 2006;Bonewald & Dallas, 1994). The physical cues of bone ECM proteins are mechanosensed by bone cells via integrin-mediated signalling, which converts the biomechanical properties of the ECM, eventually acting on cell adhesion, proliferation and differentiation (Grzesik & Robey, 1994;Hidalgo-Bastida & Cartmell, 2010;Marie, Haÿ, & Saidak, 2014).
The molecular details of the signalling pathways that mediate the relay of information from integrin engagement to altered cellular physiology remain, however, largely obscure.
In recent years, the interest in bone ECM for regenerative purposes has grown rapidly. Bone tissue engineering (BTE) applications have been proposed as bone graft substitutes in large bone defects when bone healing capacity is lost. BTE involves the combination of scaffolds (osteoconduction), osteogenic factors (osteoinduction), and autologous mesenchymal stromal cells (MSCs; osteogenesis) to mimic the native bone ECM structure and stimulate the osteogenic differentiation of local progenitors, driving new bone formation Meijer, de Bruijn, Koole, & van Blitterswijk, 2007). Scaffolds serve as a structural template for osteogenesis, being biocompatible and osteoconductive (Bose, Roy, & Bandyopadhyay, 2012). Among the several scaffolds that can be used for BTE, titanium shows good mechanical properties and it can be tailored in porosity to suit cell adhesion. Although not bioresorbable, titanium scaffolds are already clinically used for orthopaedic and dental implants and in load-bearing areas for their good mechanical properties (Holland & Mikos, 2006). As scaffolds do not reproduce the native structure of bone ECM, decellularized ECMs represent an alternative cell-instructive microenvironment to guide endogenous repair (Badylak, Freytes, & Gilbert, 2009;Benders et al., 2013). In this context, cell-secreted ECMs have also been proposed, as they are readily available and can be customized for the use as scaffold-coating (Decaris, Binder, Soicher, Bhat, & Leach, 2012;Fitzpatrick & McDevitt, 2015;Hoshiba, Lu, Kawazoe, & Chen, 2010).
We and others demonstrated that osteoblast-derived ECM stimulates MSC osteogenesis and promotes bone formation ( Baroncelli M., just accepted;Datta, Holtorf, Sikavitsas, Jansen, & Mikos, 2005;Mauney, Kaplan, & Volloch, 2004). Moreover, cell-secreted ECMs have already been used to coat and modify titanium surfaces, showing that ECM influences gene expression and enhances osteogenic differentiation of MSCs (Datta, et al., 2005; Q. P. Pham, et al., 2008;M. T. Pham, Reuther, & Maitz, 2003). In this context, the aim of this study was to compare the naturally secreted devitalized ECM to titanium and investigate how they differentially regulate cell adhesion.
Kinase activity lies at the core of cell signal transduction, as activation of specific kinases mediates the induction of signalling cascades resulting into cellular processes such as cell metabolism, differentiation, and cytoskeletal rearrangements during cell adhesion (Peppelenbosch, Frijns, & Fuhler, 2016;Robertson, et al., 2015;Zaidel-Bar & Geiger, 2010;Zambuzzi, Coelho, Alves, & Granjeiro, 2011). Integrin-mediated activation of kinase signalling cascades such as FAK and Src family kinase converts mechanical forces into biochemical signals and results in the efficient adhesion of the cell to the surface. At the same time, deregulation of kinase-mediated signalling pathways leads to pathological states, emphasizing that studying kinase activity is crucial to understand biological functions.
The aim of our study was to assess specific kinomic changes upon MSC adhesion to cell-derived ECM and titanium surfaces, by using tyrosine-kinase PamChip® array which to the best of our knowledge has not been used before to investigate cell adhesion of human MSCs.

| Cell culture and ECM preparations
Human bone marrow-derived MSCs were used to prepare the osteopromotive devitalized ECM as previously described (Baroncelli et al., 2017). Briefly, MSCs (5,128 viable cells/cm 2 ; PT-2501; Lonza, Walkersville, MD) from a single donor at passage seven were cultured in growth medium for 2 days (α-Mem phenol-red free MSCs were devitalized by freeze-thaw cycles, DNAse treatment (10 U/ml; Sigma-Aldrich), extensive washings with phosphate buffer saline (PBS; Gibco), and sterile air drying. Devitalized ECMs were stored at −20°C until further use.

| Tyrosine-kinase activity profiling using PamChip peptide microarray
To check the effect of the devitalized ECM and titanium on MSC behaviour, MSCs (28,300 viable cells/cm 2 ) were cultured on these surfaces in growth medium. After 4 hr, cells were scraped in M-PER mammalian protein extraction buffer (ThermoFisher Scientific, Rockford, IL) containing halt phosphatase and protease inhibitors (ThermoFisher Scientific), allowed to lyse at 4°C for 10 min and lysates were cleared by centrifugation at 14,000g for 10 min.
Supernatants were stored at −80°C until use. Cell lysates (5 μg protein for all samples) were loaded on a PamChip tyrosine-kinase microarray (PamGene International BV, 's-Hertogenbosch, The Netherlands). PamChip® is a high-throughput and cost-effective peptide array that allows the study of kinome profile changes without a priori assumptions (Peppelenbosch, 2012). In the PamChip platform, cell lysates are continuously pumped past 144 consensus BARONCELLI ET AL.

| 2985
peptide-sequences spotted on a three-dimensional porous microarray, and the phosphorylation of their specific target substrates by kinases present in the whole cell lysate is fluorescently detected, describing the entire tyrosine-kinase activity profile within a single experiment (Diks et al., 2004;Lemeer et al., 2007;Sikkema et al., 2009 Figure 1b).
Finally, a group of peptides displayed very slow initial activation followed by a rapid increase in reaction velocity toward the end of the experiment. These peptides were identified by inspection of the visual aspect of the curve and classified as late V max , whereas V max was calculated by using the tangent of apparent substrate phosphorylation between 1,440 and 1,840 s (Supporting Information Figure 1c). A detailed flowchart of kinome profile analysis is presented in Supporting Information Figure 1d. V max values below zero were artificially set to zero. Only V max values with average above zero were considered for further analysis. Markov state analysis was performed to determine "on" and "off" calls of peptide phosphorylation on ECM and titanium (Alves et al., 2015). In detail, for each substrate the 143 peptides were ranked for V max intensity, and a linear trend line was set for the lowest 60 peptides considered as background. Peptides whose average phosphorylation minus 1.95 times the standard deviation being higher than the background signal were considered as Markov-positive "on" calls and further analyzed (Supporting Information Figure 2a,b).

| Kinome array analysis
Protein and gene annotations of the kinase substrates on PamChip were searched through Uniprot Knowledgebase (www.uniprot.org).
Markov-positive peptides were analyzed through Qiagen's Ingenu-ity® Pathway Analysis (IPA ® ; Qiagen Redwood City; www.qiagen. com/ingenuity) against human genome provided by Ingenuity Knowledge Base as background. Gene IDs of the parent proteins of Markov-positive peptides on ECM and titanium were used in IPA. As the peptide ART_004_EAIYAAPFAKKKXC is phosphorylated by ABL1 kinase if artificial (Kua et al., 2012) as in our case, ABL1 was also included in the IPA analysis. Consensus phosphopeptides representing different phosphorylation sites of the same protein were considered together. The Canonical Pathway analysis tool was used for IPA analysis. Intracellular signalling pathways not restricted to a specific cellular type were selected and used to generate the heat map using R, together with the relative Gene IDs of the phosphorylated kinase substrates.
Because specific kinases activate signalling cascades, the kinase substrates that were phosphorylated in cells on ECM and on titanium were further fitted into signalling pathways and cell-related functions as previously described (Sikkema et al., 2009) to confirm IPA analysis.

| Statistical analysis
Area under curve (AUC) of the kinetic reaction was calculated for each Markov-positive peptide, and nonparametric Wilcoxon matched-pairs signed rank test used to calculate significance.
Functional attachment data were representative of three independent experiments, with one or two technical replicates per each experiment, and all values were displayed as average ± standard deviation of biological replicates otherwise indicated elsewhere.
One-way analysis of variance, followed by Bonferroni post hoc test was used to calculate significance, unless otherwise indicated.  Table 1).
On titanium, 37 peptides were phosphorylated with maximum reaction speed early in the analysis, with an additional 10 at mid stage and 12 at later stages (Supporting Information Figure 2b  (AUC of 34.1 ± 7.6 vs. 27.6 ± 6.5, P < 0.0001, data not shown). Several of the peptides that were significantly phosphorylated early in MSCs cultured on ECM, only achieved significant phosphorylation at later time points when cells were cultured on titanium, suggesting lower levels of active kinase present in these latter lysates. Comparing the kinase activity profiles, most of the phosphorylated peptides (55) were shared between the two substrates, highlighting a substantial overlap between the kinome profiles of MSCs in ECM and titanium, whereas eight kinase substrates were exclusively phosphorylated on the cell-derived ECM and four on titanium (Figure 1c).   Table 3 for titanium). We further focused on intracellular signallings, not specific for a selected cell type, resulting in a total of were activated on titanium as on the ECM (p = 7.94 × 10 −11 , 6.30 × 10 −11 , 2.29 × 10 −17 , and 5.75 × 10 −6 , respectively), as well as ERK/MAPK (p = 3.46 × 10 −7 ) and PI3K/AKT (p = 1.91 × 10 −4 ), confirming the overlap between these substrates.
IPA analysis revealed that the activated kinases were involved in multiple signalling cascades. In addition, we used the results of the peptide array to fit each kinase that phosphorylates a selected peptide into one specific signalling pathway, as previously done (Sikkema et al., 2009), in a more biased approach but more osteoblast-oriented (complete list in Supporting Information Table   4). This approach confirmed the IPA analysis, as of the 63 kinase substrates phosphorylated on ECM, four induced the activation of FAK signalling and a total of 11 phosphopeptides were involved in cytoskeletal functions (Supporting Information Figure 3a,b; Supporting Information Note. Peptides are in alphabetical order; numbers indicate the position of the first and last amino acid of the peptide in the complete human protein.
phosphorylated on titanium (Supporting Information Table 6).
Phosphorylation of four peptides induce the activation of FAK signalling. Moreover, MAPK (two peptides) and PI3K (three peptides) signalling were activated also in cells in adhesion to titanium (Supporting Information Figure 3c,d). The

| Functional consequences of reduced PI3K activation
PamChip microarray analysis revealed that the PI3K/AKT signalling pathway among others was activated in cells adhering to both substrates ( Figure 2), but with a higher activity on ECM than on titanium (Figure 3a,d). The temporal kinetics of the peptides clustered in PI3K/AKT signalling are displayed in Figure 4. PI3K signalling has been shown to be important for several cellular functions, including cell adhesion. We validated this by allowing MSCs to adhere to ECM for 4 hr in the absence or presence of the PI3K kinase inhibitors Wortmannin or LY290042. Figure 5a shows that Wormannin significantly reduced cell attachment to confirming that the PI3K inhibitors were not toxic (Figure 5d).
The phosphorylation of selected kinases such as PKB (as well as ERK) was also confirmed to be reduced in cells cultured on titanium compared to cultures on ECM, also when assessing nonadherent cells (Supporting Information Figure 5a,b). Thus, we next investigated adhesion of MSCs to titanium and showed that while the highest concentration of PI3K inhibitors (10 μM) reduced adhesion, this effect did not reach statistical significance, in accordance with the

| DISCUSSION
In this study, we comprehensively described the kinome profiles of human MSCs during adherence to a cell-derived ECM and to titanium, by successfully using PamChip array technology. MSCs on the two substrates showed a substantial overlap of kinase signatures.
Cells on ECM typically activate kinase reactions that conform classical kinetics, that is that maximum reaction speeds are seen early in the experiment, and have higher level of active kinases.
Importantly, without a priori assumptions, we used the PamChip kinase array to identify PI3K signalling and further functional experiments showed its importance in MSC viability and adhesion.
This observation may guide rational design of novel scaffolds for tissue engineering.
Cell-surface interplay has been studied to develop biomaterials to improve BTE Zambuzzi et al., 2011). Upon cell adhesion to a surface, mechanical forces are converted into biochemical signals by integrins, that induce the activation of FAK, Src family kinases, and an intricate network of signalling pathways, such as PI3K, MAPK ERK1/2, PKC, and Rho-family GTPase, that eventually modulate cell behaviour (Marie et al., 2014). Osteoblast adhesion is controlled mainly by PKA, PKC, and RhoA proteins that promote cell cycle arrest and mediate cytoskeletal rearrangements (Zambuzzi, Bruni-Cardoso, et al., 2009). In line with this, we showed that pathways such as FAK, PAK, Paxillin, ILK, and Rho GTPase family signalling were activated upon MSC adhesion to ECM and titanium. Moreover, PamChip kinase array revealed the activation of ERK/MAPK signalling. MAPKs are a central hub in controlling bone homoeostasis, as they are activated by extracellular stimuli and ECM-mediated integrin activation via Src/FAK signalling network, but also promote osteoblast survival and differentiation by controlling osteogenic transcription factors (Greenblatt, Shim, & Glimcher, 2013;Marie et al., 2014). Cell adhesion to an ECM has been studied through mass spectrometry showing the high level of tyrosine phosphorylation in adhesion complexes, thus revealing the importance of kinases in cell adhesion (Robertson et al., 2015;Zaidel-Bar & Geiger, 2010) The behaviour of calvarial osteoblasts has been analyzed through PepChip kinase-array screening technology. PamChip is a cost-effective high-throughput array that can simultaneously identify rapid changes in kinome profiles (Peppelenbosch, 2012). PamChip and other kinase-array platforms drive hypothesis formation, due to the variable number of putative upstream kinases that could phosphorylate the peptides. They represent powerful tools to select pathways that might be crucial in physiological functions, but the kinase activation needs to be validated by immunoblot analysis (Arsenault, Griebel, & Napper, 2011;Sikkema et al., 2009). We used IPA for functional clustering of the peptides into signalling pathways, as previously done (Kuijjer et al., 2014), and we confirmed it by fitting each peptide in one specific signalling pathway. However, software to fit phosphopeptides into cascade signalling networks needs to be implemented with kinomic-oriented tools.
In this study, PamChip revealed the activation of PI3K/AKT in MSCs adhering to ECM and titanium, which corroborates with previous findings on polystyrene and on hydroxyapatite Milani et al., 2010). Conversely, PI3K/AKT signalling was found to be downregulated during osteogenic differentiation in standard culture conditions and on Matrigel (Chaves Neto et al., 2011;Marumoto et al., 2017). The PI3K/AKT signalling pathway regulates many cellular functions such as proliferation, adhesion and migration, and its activation promotes cell survival (Manning & Toker, 2017 (Fujita et al., 2004), though findings are still controversial as for the influence of PI3K signalling in osteogenic differentiation (Kratchmarova, Blagoev, Haack-Sorensen, Kassem, & Mann, 2005;Viñals, López-Rovira, Rosa, & Ventura, 2002). PI3K signalling is involved in integrin-mediated signal transduction during cell adhesion (Chen, Appeddu, Isoda, & Guan, 1996;King, Mattaliano, Chan, Tsichlis, & Brugge, 1997). In this study, we showed that PI3K is involved in osteoblast adhesion to ECM and titanium, in agreement to previous studies where PI3Kmediated AKT activity was shown to be reduced when PI3K inhibitors were present in COS7 cells and MSCs in adhesion to fibronectin (Chaudhary et al., 2000;Liu et al., 2010). When PI3K inhibitors were used, the number of nonadherent cells increased, with a stronger effect on ECM than titanium. This is probably due to the fact that PI3K signalling was more active on ECM than on titanium, as shown by the temporal kinetics of the PI3K kinase substrates, thus making it easier to visualize the inhibitory effect and highlighting the importance of performing kinetic analyses.
The mechanisms of cell adhesion to titanium has been previously investigated by using FAK and Src phosphorylation as biomarkers to monitor cell or biomaterials interplay, as FAK and Src have been proven to be phosphorylated upon integrin activation in cells when adhering to different substrates (Zambuzzi et al., , 2009Zambuzzi, Milani, & Teti, 2010

CONFLICTS OF INTEREST
All authors declare that there are no conflicts of interest.