The TGF-β1-Induced Expression of Matrix Metalloproteinases in Mesenchymal Stromal Cells is Influenced by Type of Substrate

Transforming growth factor (TGF)-β1 activates the expression of matrix metalloproteinases (MMPs) in fibroblasts. Attachment of these cells to laminin-111 further raises the TGF-β1-induced expression of MMP-3 and MMP-10. Mesenchymal stromal cells (MSC) attach to a variety of extracellular matrix proteins during development and wound healing. We therefore investigated the TGF-β1-regulated expression of MMPs in MSC upon attachment to laminin-111 and type I collagen. The expression of MMPs was determined by quantitative reverse transcription polymerase chain reaction and enzyme-linked immunosorbent assay. The TGF-β1 signalling pathways were investigated by immunoblot and by pharmacological blocking of Smad2, MEK/ERK and p38MAPK activities. Overall, TGF-β1 significantly activated the expression of mRNA encoding MMP-3 (p≤0.05), MMP-13 (p≤0.05) and TIMP-1 (p≤0.01) in MSC. Induction of MMP-10 was not significant. In contrast to our observation on fibroblasts, the attachment of MSC to laminin-111 did not affect the TGF-β1-induced expression of MMP-3 and MMP-10. Attachment to type I collagen reduced the TGF-β1-induced secretion of MMP-3 and MMP-10 compared to cells grown on laminin-111 or tissue culture plastic dishes. The expression of MMP-3 was induced by TGF-β1 via Smad2, ERK1/2 and p38MAPK. The expression of MMP-10 was regulated by Smad2 and ERK/1/2, whereas the expression of MMP-13 was shown to be p38 MAPkinase dependent. We conclude that the regulation of MMP-3, MMP-10, and MMP-13 by TGF-β1 proceeds via distinct signalling routes. In contrast to the regulatory pathways in fibroblasts, we could not prove a co-signalling of TGF-β1and integrin-dependent pathways for the regulation of MMP-3 and MMP-10 in MSC upon attachment to laminin-111. Therefore, MSC respond differently to TGF-β1 and extracellular matrix molecules compared to fibroblasts. Journal of Tissue Science & Engineering J o u r n a l o f T iss ue S cience &ngine e r i n g

domain common to all MMPs contains the conserved zinc-binding domain and a unique methionine-turn [22] which play a role in matrix degradation during physiological or pathological processes [22][23][24][25][26]. The stromelysins MMP-3 and MMP-10 degrade a variety of proteins including pro-collagen-I and thus contribute to matrix degradation in bone [27] and cartilage [28]. The collagenase MMP-13 cleaves type II collagen more efficiently than type I or III collagens and is considered to play an important role in physiological and pathological degradation of articular cartilage. The activity of many MMPs is regulated by tissue inhibitors of metalloproteases (TIMPs) [29].
We recently showed that TGF-β1 cooperates with integrin signalling upon attachment of synovial fibroblasts to LM-111, significantly boosting the expression of MMP-3 and MMP-10 [1,2]. Crosstalk between TGF-β and the integrin signalling pathway has been reported in epithelial cells as well [30]. We therefore investigated in this study whether attachment to LM-111 or attachment to another major component of the extracellular matrix, Col-1, affects the expression of MMPs in human TGF-β1-activated MSC. In addition, the TGF-β signalling pathways were studied by immunoblots and by specific inhibitors blocking the phosphorylation of Smad2, the p38 mitogenactivated protein kinase (MAPK) and extracellular signal-regulated kinases (ERK)1/2.

Isolation and characterization of MSC
After written consent, the MSC were isolated by density gradient centrifugation from femoral bone marrow aspirates of patients undergoing endoprosthetic surgery (n=15). The study was approved by the local ethics committee. MSC were grown in MSC Growth Medium (Lonza, Basel, Switzerland) and characterized as described recently [4,31]. Briefly, flow cytometry was performed to confirm the surface expression of CD73 (BD Pharmingen, San Diego, CA, USA), CD90, CD146 (R&D Systems, Minneapolis, MN, USA) and CD105 (Serotec, Raleigh, NC, USA), as well as lack of CD11b, CD14 (BD Pharmingen), CD34 (Biolegend, San Diego, CA, USA), and CD45 (R&D Systems) on the cells obtained [31,32]. Further, we proved their ability to differentiate in adipogenic, chondrogenic and osteogenic lineages by established protocols [31,33]. Differentiation of MSC to adipogenic cells was detected by staining the intracellular lipid vesicles with Oil Red-O. The proteoglycans in chondrogenically differentiated pellets were detected by Alcian blue staining. Detection of the mineralized extracellular matrix by von Kossa staining was used to detect osteoblasts after in vitro differentiation of MSC.
The cells were harvested at the same time and RNA was isolated (RNeasy Mini Kit, Qiagen, Hilden, Germany). Reverse transcription of RNA was carried out employing the RT-for-PCR kit (Clontech, Mountain View, CA, USA). Transcript levels of MMP-1, -2, -3, -9, -13 and TIMP-1, -2 and -3 were measured after reverse transcription by quantitative polymerase chain reaction (qRT-PCR) using commercially available primers (Search LC, Heidelberg, Germany). For MMP-10 the following reagents were used: forward primer: 5'-GGCTCTTTCACTCAGCCAAC-3', reverse primer: 5'-TCCCGAAGGAACAGATTTTG-3', product size: 175 bp, access number NM_002425.1). The amplification of cDNA was performed in 35 cycles (LightCycler 1.5, Roche, Mannheim, Germany). Following one initial cycle (95°C 10 sec, 68°C 10 sec, 72°C 16 sec, temperature transition rate 20°C sec -1 ) the annealing temperature was dropped to 58°C with a step size of 0.5°C. PCR amplification was evaluated by the "fit-points" method [34]. The data represent the mean mRNA expression levels of MMP and TIMP normalized to the expression levels of the housekeeping gene GAPDH. Known amounts of a recombinant standard DNA were used to calibrate each run. The quality of amplification was investigated by melting point analysis (95°C, 58°C 10s, 95°C). The PCR products were then separated by electrophoresis on 1.5% agarose gels and visualized by UV-activated ethidium-bromide fluorescence to confirm the expected size.
To investigate the contribution of the signalling pathways on regulation of the MMPs, MSC were grown in TCPS flasks over night and the signalling pathways were blocked by addition of the following inhibitors: 100 nM of TGF-β RI kinase inhibitor (ALK5i) which blocks the phosphorylation of Smad2/3, or 20 µM of the MEK inhibitor which blocks the phosphorylation of ERK1 and ERK2 (PD 98059), or 20µM of the p38MAPK inhibitor (SB 203580) (all from Calbiochem). After 30 min pre-incubation with these inhibitors, cells were stimulated by addition of 10 ng/mL TGF-β1. Cells activated by TGF-β without inhibitors and cells left untreated served as controls. After 24 h, supernatants were collected to perform ELISA assays as described above; and RNA was isolated to perform qRT-PCR to quantify transcript levels of MMP-3, MMP-10 and MMP-13.

Statistical analysis
The mean values, corresponding standard deviations and statistical significance between groups of data were assessed with a two-sided paired Student's t-tests using the GraphPad Prism® software. Probability values (p) equal to or less than 0.05 (*) or 0.01 (**) were considered to be statistically significant and marked in the figures accordingly.

Characterization of isolated MSC
MSC were characterized by flow cytometry and differentiated as  [4]. The MSC displayed strong staining for CD73, CD90, CD105 and CD146, and absence or low expression of CD11b, CD14, CD34 and CD45 (see supplement, Figure S1). The MSC showed a fibroblast-like morphology and could be differentiated in vitro into adipogenic, chondrogenic and osteogenic lineages (see supplement, Figure S2).
To investigate the expression of the MMPs on the protein level, supernatants of MSC were collected and tested by ELISA. The secretion of MMP-3 was enhanced by TGF-β1 approximately fivefold above controls (n.s., n=5, Figure 2B), and that of MMP-10 fifteen-fold above controls (p<0.01, n=5, Figure 2D). The basal MMP-13 protein levels were below the detection limit. However, TGF-β1 stimulated the secretion of MMP-13 in MSC ( Figure 2F).
Attachment of MSC to Col-1 in absence of TGF-β1 did not   significantly alter the MMP-3 secretion in comparison to that in TCPS adherent MSC (11.5 pg/mL ± 6). But when exposed to TGF-β1, MSC growing on Col-1 produced less than half of MMP-3 (34.4 pg/mL ± 17, n=5) compared to TGF-β1 activated MSC on TCPS (82.5 pg/mL ± 66, n=5). The difference in production of MMP-3 by TGF-β1-activated MCS on TCPS versus Col-1 did not reach significance. The attachment of MSC to LM-111 did not affect their TGF-β1-induced expression of MMP-3 (77.8 pg/mL ± 58, n=5) compared to cells on TCPS ( Figure  2B). Similar results were obtained for MMP-10. TGF-β1 elevated the secretion of MMP-10 in MSC on TCPS (183.2 pg/mL ± 79, p<0.01, n=5), as well as on LM-111 (183.2 pg/mL ± 51, n=5, Figure 2D). Note that the MMP-10 secretion from TGF-ß1 activated MSC grown on Col-1 (76.3 pg/mL ± 32, p<0.01, n=5) remained 2.4 fold below the production measured in supernatants of MSC incubated on plastic or LM-111. Again, the difference in production of MMP-10 by TGF-β1-activated MCS on TCPS versus Col-1 did not reach significance. Spontaneous secretion of MMP-13 was at or below detection levels. TGF-β1 stimulated the production of MMP-13 in MSC attached to TCPS, Col-1 and LM-111. The attachment on Col-1 had no effect on the secretion of MMP-13, but cellular contact to LM-111 augmented the release of MMP-13 to some extent ( Figure 2F). The expression of MMP-1, MMP-2 and MMP-9 and of TIMP-1, -2 and -3 was not affected by the attachment of the cells to Col-1 or LM-111 (data not shown).

TGF-β1 signalling pathway
In order to better understand the TGFβ1-regulated expression of the MMPs in MSC, we investigated signal transduction by Smad2, ERK-and p38MAPK, (see supplement, Figure S3), and tracked the cell signalling pathways (Figure 3).
Activation of Smad2 as visualized by phosphorylation of the protein was recorded 30 to 120 minutes after addition of TGF-β1, peaking one hour after induction. Prior to addition of TGF-β1, phosphorylated Smad2 was not detected in MSC (see supplement, Figure S3). Phosphorylation of the 43 kDa ERK1 was slightly raised, again peaking after 60 minutes of activation, whereas phopshorylation of the 41 kDa ERK2 was more prominent. A strong increase in signals indicating phosphorylation of p38 MAPK were not observed in TGF-β1-induced MSC compared to control cells. Detection of total p38 MAPK protein served as loading control ( Figure S3).
TGF-β1 elevated the MMP-10 transcript levels more than 5-fold. They were reduced slightly by the AKL5/Smad2 inhibitor and by the MEK/ERK inhibitor PD98059, respectively ( Figure 3C). In contrast to regulation of MMP-3, the p38 MAPK inhibitor had no significant effects on the mRNA expression of MMP-10 ( Figure 3C). The MMP-10 protein level was elevated from 17.4 pg/mL to 196 pg/mL by TGF-β1 and was diminished to 28 pg/mL by addition of the AKL5/Smad2 inhibitor, to 61 pg/mL by the MEK/ERK inhibitor, and to 109 pg/mL by the p38 MAPkinase inhibitor ( Figure 3D). Thus, in MSC, TGF-β1 regulated the expression of MMP-10 mainly via activation of Smad2 and ERK1/2.
In contrast, the TGF-β1-induced MMP-13 mRNA expression could be diminished by addition of the p38 MAPkinase blocker SB 203580. TGF-β1 induced the MMP-13 mRNA expression significantly (8-fold, p<0.05). The addition of the SB203580 reduced the TGF-β1-induced expression of MMP-13 significantly to (p<0.01), whereas the TGF-βRI and MEK/ERK inhibitors had only minor and non-significant effects ( Figure 3E). Spontaneous release of MMP-13 from MSC was very low in vitro but was elevated more than 10-fold above the detection limit of the ELISA by TGF-β1 ( Figure 3F). In the presence of each of the inhibitors interfering with signalling of ALK5/Smad2, MEK/ERK or p38MAPK, the production of MMP-13 was reduced to the detection levels of the assay ( Figure 3F). Thus, on the transcript level, TGF-β1 appears to regulate the expression of MMP-13 mainly via activation of p38 MAPK, and to a lesser extent through Smad2 and ERK1/2.
In MSC, an activation of MMP-2 and MT1-MMP was reported after 72 h of treatment with TGF-β1, which caused increased migratory ability of the cells. But no change in the expression of MMP-2, MT1-MMP and TIMPs was observed after 24 h of TGF-β1 treatment [42]. With respect to the TIMPs this corroborates our results, and suggests a time-dependent induction of MMP expression by TGF-β1. Therefore, MMP-3, MMP-10 and MMP-13 could be among the first proteases induced by TGF-β1, leading to matrix degradation in the MSC niche. Later on, other MMPs, such as MMP-2 and MT1-MMP, could promote migration of the MSC to their target tissue during wound healing.
TGF-β1 signalling employs the TGF-β receptors TGF-βRI and -RII, and the intracellular pathways via Smad2/3, ERK1/2 and p38MAPK. By aid of activation-specific antibodies and by use of specific inhibitors, we could prove that in MSC the expression of MMP-3 is elevated by TGF-β1 via all three pathways, Smad2, ERK1/2 and p38MAPK. This has also been seen in synovial fibroblasts [1,2]. Thus, the expression of MMP-10 is regulated mainly by Smad2 and ERK1/2 in MSC, and expression of MMP-13 is regulated in TGF-β1 activated MSC foremost via p38MAPK. In chondrocytes IL-1β activates the expression of MMP-13 in the same manner [43].
Furthermore, the regulation of MMPs also depends on the substratum to which the cells attach [44][45][46][47]. The expression of MMP-3 and MMP-10 is not affected by the attachment of MSC to LM-111. But release of MMP-13 is elevated when MSC adhere to LM-111. In contrast, when MSC are grown on Col-1-coated dishes, the TGF-β1induced protein expression of MMP-3 and MMP-10 decreases. The attachment of MSC to Col-1 and LM-111 is amongst others mediated by integrins [13]. Integrin α2β1 plays a key role in the attachment of MSC to Col-1. The activation of MSC by TGF-β1 enhances the binding of MSC to this protein [4]. We assume that integrins not only provide contact points for the attachment of MSC to Col-1 or LM-111, but also influence the TGF-β1-induced signalling pathways in these cells.
In contrast to MSC, attachment of synovial fibroblasts to LM-111 modulated the expression of MMP-3 and MMP-10, and the TGF-β1 induced expression of MMP-3 and MMP-10 was even enhanced in fibroblasts [1,2]. Activation of osteoblasts by TGF-β1 also activated the expression of MMP-3 and MMP-10. As reported in this study for MSC, in osteoblasts the TGF-β-facilitated expression of MMPs was not enhanced by attachment to LM-111, and was also not ameliorated by attachment to Col-1 (unpublished observation). Differences in expression of integrins or other extracellular matrix binding cell surface proteins between MSC, fibroblasts and osteoblasts may account for the distinct sensitivities of these cells in response to cytokine plus substratum-dependent cell activation.
Crosstalk between cell-matrix attachment and cytokine-or growth factor-induced signalling is known to occur in many cell types, e.g. in epithelial cells [30], where the "crossing point" between the two signalling pathways is the p38MAPK. A similar crosstalk could exist in MSC, enhancing the expression of MMP-13 upon attachment to LM-111 and exposure to TGF-β1. Moreover, a paracrine crosstalk may exist between the MSC themselves, and the individual setup of the experiment may influence the outcome. Therefore the induction index computed for the MMPs in a given sets of experiments (e.g. Figure 1, optimized for mRNA yield) will differ from the results observed in another set of investigations (e.g. Figure 2, optimized for exploring key transcripts and cell supernatants from the same culture).
The co-regulation of MMPs by TGF-β1 and integrin receptors may play a role during developmental processes such as osteogenesis, and in the chondrogenic differentiation of MSC. During these processes, a permanent reconstruction of the ECM takes place in the developing tissues. Furthermore, the regulated expression of MMPs in response to a combination of growth factor signalling and integrin-mediated ECM recognition may allow directed migration of MSC from their niches into the target tissue during development or wound healing [42].
Moreover, modulation of cytokine responses by the composition of biomaterials and scaffolds may also be a useful tool to further improve the outcome of tissue engineering. The cells applied together with such an implant will degrade the scaffold during wound repair. Scaffolds generated from a substratum promoting for instance a mild expression of MMPs at the rim of the implant could facilitate its integration. But controlling the expression of MMPs in a load-bearing zone in the center of the scaffold by e.g. Col-1 could indirectly promote the deposition of an extracellular matrix by lowering the expression of the stromelysins MMP3 and MMP10. This may then accelerate the regeneration of the defect. This, however, must be addressed specifically in a separate study in the future.