Cell Motility and the Cytoskeleton 45:1–9 (2000) Molecular Responses of Human Dermal Fibroblasts to Dual Cues: Contact Guidance and Mechanical Load V.C. Mudera,1* R. Pleass,2 M. Eastwood,3 R. Tarnuzzer,4 G. Schultz,4 P. Khaw,2 D.A. McGrouther,1 and R.A. Brown1 1University College London, Tissue Repair Unit, Division of Plastic and Reconstructive Surgery, London, United Kingdom 2University College London, Wound Healing Group, Department of Pathology, Institute of Ophthalmology, London, United Kingdom 3University of Westminster, Centre for Tissue Engineering Research, Department of Technology and Design, London, United Kingdom 4University of Florida, Department of Obstretrics and Gynecology, Gainesville, Florida, USA Fibroblast contraction in wound healing involves the interaction of several cell types, cytokines, and extracellular matrix molecules. We have previously developed fibroblast alignment models using precise uniaxial mechanical loads in 3D culture and using contact guidance on fibronectin strands. Our aim here was to use contact guidance to place fibroblasts in their potentially most sensitive configuration, i.e., perpendicular to the axis of loading, to present cells with conflicting guidance cues. Gene expression at the mRNA level of cells recovered from different zones of the 3D collagen gel (with distinct orientation) was determined by quantitative RT-PCR for the matrix proteases MMP1, 2, and 3, and inhibitors TIMP1 and 2.Our results show a 2-, 4-, and 3-fold increase in MMP1, 2, and 3, respectively, in the non-aligned strain zone, relative to the aligned strain zone. These results suggest that cells unable to align to applied loads remodel their matrix far more rapidly than orientated cells. Where fibroblasts were held in an alignment perpendicular to the applied load by contact guidance, the fall in MMP mRNA expression was largely abolished, indicating that these cells remained in a mechano-activated state. The protease inhibitors TIMP1 and 2 were poorly mechano-responsive, further suggesting that changes in MMP expression result in functional matrix remodelling. These results indicate how mechanical loading in tissues may influence matrix remodelling, particularly under conflicting guidance cues. Cell Motil. Cytoskeleton 45:1–9, 2000. r 2000 Wiley-Liss, Inc. Key words: quantitative RT-PCR; contact guidance; mechanical load; MMP; TIMP; fibroblasts; collagen gel INTRODUCTION Fibroblast populated collagen lattices (FPCLs) have been widely used as an in vitro model for wound contraction. Studies have tested the effect of mechanical loads on fibroblasts in collagen lattices. These have ranged from the effect of stress on cyclic AMP [He and Grinnell, 1994] and PDGF receptors [Lin et al., 1998], Tenascin C [Chiquet-Ehrismann et al., 1994], alpha smooth muscle actin [Arora et al., 1994], increased cell proliferation [Butt et al., 1995], and morphological alter- r 2000 Wiley-Liss, Inc. This work was carried out jointly at the Tissue Repair Unit and Wound Healing Group, University College London. Templates for Quantitative RT-PCR were synthesised and purified at the University of Florida. Contract grant sponsor: Engineering and Physical Sciences Research Council; Contract grant sponsor: Guide Dogs for the Blind. *Correspondence to: Dr. Vivek Mudera, University College London, Tissue Repair Unit, Division of Plastic and Reconstructive Surgery, 67–73 Riding House Street, London W1P,7LD, UK. E-mail: email@example.com Received 17 March 1999; accepted 21 September 1999 2 Mudera et al. ations as feedback reactions [Chamson et al., 1997]. Previous work from our group examining the effects of precise uniaxial mechanical loading on FPCLs has demonstrated an ability to predictably align fibroblasts along lines of principal strain [Eastwood et al., 1998] (see Fig. 5) and mechanical responses of fibroblasts to external loading, which maintains a tensional homeostasis [Eastwood et al., 1996]. In our FPCL model using the Tensioning culture force monitor, we have also demonstrated using Finite Element Analysis, whereby in the regions where mechanical loading was non-directional in the FPCL, called the ␦ zone (because of the stiffness of the material used to anchor the FPCLs), the cells did not become bipolar and elongate but remained stellate in shape even though they were subjected to the same mechanical loading (see Fig. 6) [Eastwood et al., 1998]. This has led to the hypothesis that the morphological and motile behavior of fibroblasts tends to ‘‘stress shield’’ them from the perceived mechanical loads wherever they are able to. (Note: ‘‘percieved load’’ depends on the applied load and the material properties, i.e., stiffness of the matrix.) This shielding corresponds to a bipolar, elongate morphology orientated parallel with the principal strain. An alternative means to control fibroblast alignment uses topographic guidance, for example micromachined grooves of different dimensions [Oakely et al., 1997; Curtis and Wilkinson, 1997]. Contact guidance of cellular migration and orientation has been described on orientated collagen fibrils [Guido and Tranquillo, 1993] and on fibronectin fibres [Ejim et al., 1993]. Contact guidance to fibronectin fibres acts at two levels. Firstly, the cells on the fibres become bipolar and align parallel with the fibre (i.e., direct cell-fibre interaction). Subsequently, multiple layers of cells ‘‘dock’’ parallel to this first layer of cells apparently by cell-cell attachment [Ejim et al., 1993], representing a second, indirect form of contact alignment. This ability to control cell orientation experimentally (using contact guidance or tensional models) has great potential in understanding the basic biological mechanisms regulating development, growth, and repair. It is also critical for the development of advanced forms of tissue repair and cell engineering therapies for example in peripheral nerve repair, production of tendon and ligament substitutes in vitro, and control of microvascular repair. The aim in all cases is to orientate repair cells and new tissue formation, not least the plane in which the extracellular matrix is laid down. Whilst these models provide detailed information on cellular responses to single cues, cells in vivo are subjected to complex and interdependent combinations of mechanical and contact guidance cues. The aim of the present study was to test the concept that cells subjected to conflicting orientating cues would behave in a manner that is a predictable compromise between each cue. Dermal fibroblasts were in the first stage aligned by contact guidance on fibronectin strands embedded in an FPCL. They were then subjected to uniaxial mechanical loads applied at 90° to the fibronectin fibres. Cell orientation on fibronectin fibres gives a principal orientation without external tension. The second external loading cue, perpendicular to the contact guidance cue, would be predicted to cause some realignment of contact guided cells. This hypothesis predicts that cell movement (and eventual matrix deposition) will lead to optimal stress shielding from external loads, i.e., parallel alignment; consequently, cells that have been pre-aligned perpendicular to the applied load should be maximally stimulated. To test this hypothesis we monitored the effects of these conflicting cues on gene expression of three important enzymes in matrix remodeling MMP1, 2, and3 and their natural inhibitors TIMP1 and 2. MATERIALS AND METHODS Human dermal fibroblasts were cultured from explants of normal skin taken directly from the operating theater [Burt and McGrouther, 1992] in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% (v/v) Fetal Calf Serum (FCS, First Link, West Midlands, UK), with glutamine (2 mM, ICN, Biochemicals Ltd, Thyne, UK), and penicillin/streptomycin (1,000 U/ml/100 µg/ml) (Gibco Life Technologies, Paisley, UK). Cells were used between passage 3–6 for experiments. Preparation of Fibronectin Strands Fibronectin strands were extruded by a modified method as described by Underwood et al. . Briefly, fibronectin-rich plasma solution obtained as a by-product (Scottish National Blood Transfusion Service, Plasma Fractionation Laboratory, Edinburg) was used. The protein concentration of the solution was 4.7 mg/ml of which 3.4 mg/ml was fibronectin and 1.3 mg/ml was fibrinogen. The solution was acidified to precipitate proteins by addition of 0.1M Citric acid at a v/v ratio of 1 acid:2 protein solution to give a final pH between 4.0 to 4.5. The acidified solution was gently stirred to facilitate protein aggregation. Strands of aggregated fibronectin were drawn up using a fine glass rod and washed repeatedly in copious amounts of DW followed by isotonic saline and DMEM to neutralize the acidity. Mechanical Loading Model Fibroblast Populated Collagen Lattices (FPCL) were set up as described earlier [Eastwood et al., 1998]. Gels were prepared by mixing 4 ml of 2.28 mg/ml solution of native acid soluble Type I rat tail collagen (Advanced Protein Products, West Midlands, UK) with 0.5 ml of Molecular Responses to Conflicting Guidance Cues 3 FPCLs were treated for mRNA extraction as described below. Assembly of the FPCL for the dual cue model (Fig. 1) involved formation of a two-layered gel. The first layer was made by casting 2 ml of collagen containing proportionate fibroblasts, followed by a layer of dermal fibroblasts and three fibronectin strands perpendicular to the long edges of the gel followed by another layer of fibroblasts and a further 2 ml of collagen solution containing fibroblasts. This was to ensure even distribution of fibroblasts throughout the gel to maintain similarity with the mechanically loaded model. This was then allowed to gel at 37°C for 5 min, topped up with 15 ml of culture medium, mounted onto the tensioning-culture force monitor, and cyclically loaded as described. Extraction of mRNA Fig. 1. Overhead and side view schematic of Dual Cue Model showing fibronectin strands along which cells align, perpendicular to the applied mechanical load. 10⫻ DMEM and neutralized with 5M NaOH prior to addition of fibroblasts (106 cells/ml of gel solution) suspended in 0.5 ml of DMEM. The 5-ml collagen/cell suspension formed a gel within 5 min at 37°C, at which stage the cell chamber was topped up with a further 15 ml of culture medium. Rectangular FPCLs (75⫻ 25 mm) were restrained by their opposing short edges to give a high aspect ratio configuration as described earlier [Eastwood et al., 1998]. The entire apparatus was then transferred and fixed to the tensioning-Culture Force Monitor (t-CFM) a modification of the Culture Force Monitor described previously [Eastwood et al., 1998]. FPCLs were loaded with 120 Dyne force in a 1-hr cycle after an 8-hr pre-culture period without any loading, during which time the cells established a tensional homeostasis [Brown et al., 1998]. The loading cycle consisted of 15 min of linear loading at a rate of 480 dynes/hr, followed by a 15-min resting phase (with no change of load), then unloading for 15-min reversal of loading, and finally a further 15-min resting phase. The total cycle time was 1 hr repeated for 16 hr after which the FPCLs were recovered and the different zones (described in Fig. 1) were separated using a scalpel and immersed in modified lysis buffer for mRNA extraction as described below. Free-floating gels were set up as described above except that the gels were untethered (i.e., not attached to the t-CFM). After casting, the FPCL was detached using a fine bore needle and, after the addition of 15 ml of culture medium, allowed to float unrestrained. After 24 hr, the mRNA was extracted from gels using Qiagen kits with a minor modification. The gels were cut using a surgical scalpel into different zones of alignment and these gel pieces were immersed in the lysis buffer supplied with the kits. An extra 5 µl of ␤-mercaptoethanol/ml of lysis buffer was added to facilitate break up of disulfide bonds and the lysis buffer with the gel was frozen overnight at ⫺20°C . Upon thawing, the gel with the resident cells had lysed completely and the subsequent extraction process was followed according to the manufacturer’s protocol. Only samples with an OD 260:280 ratio between 1.8–2.0 were used for subsequent RT-PCR analysis. Preparation of Competitive MMP Template and Quantitative(QC)-RT-PCR Protocol Competitive MMP template was prepared as described by Tarnuzzer et al.  from a synthetic Super-Template Plasmid Construct. cDNA was synthesized using serial dilutions of the competitive MMP template from 6.8 ⫻ 102 to 6.8 ⫻ 109 copies/reaction along with 0.1 µg of extracted RNA in each tube for all samples along with 2.5 mM oligo(dt)16, 1.5 mM MgCl2, 200 µM dNTP (Promega), 50 U/ml Human Placenta Ribonuclease Inhibitor, Tris-HCl, pH 8.3, 50 mM KCL and 200 U/µg RNA Moloney Murine Leukemia Virus (MMLV)- RT (Life Technologies, Gaithersburg, MD). The reaction was incubated at 25°C for 10 min, 37°C for 60 min, and 92°C for 5 min. cDNA amplification was carried out in a 50 µl reaction volume containing 5 µl of the RT reaction, 200 µM dNTP (Promega USA), 50 pmol of each 38 and 58 PCR primer, 1.5 mM MgCl2, 10 mM Tris-HCL, pH 8.3, 50 mM KCl, and 0.25 µl/reaction Taq DNA polymerase (Perkin-Elmer). Amplification reactions were carried out in 40 sequential cycles of 94°C for 1.5 min, 58°C for 2 min, and 72°C for 3 min. 4 Mudera et al. Fig. 2. Photograph of 2% agarose gel stained with ethidium bromide showing decreasing band intensity for sample as template concentration increases with primers for MMP2. Template (333 bp) and sample bands (615 bp) are clearly separated. Marker used is Phi X/ Hae III. Primer Sequences for MMP Template MMP1 38: AGGTTAGCTTACTGTCACAC MMP1 58: TTGTCCTCACTGAGGGAAAC MMP2 38: GTACTTGCCATCCTTCTCAA MMP2 58: CCTGTTTGTGCTGAAGGACA MMP3 38: GTTCTGGAGGGACAGGTTCC MMP3 58: TCAGAACCTTTCCTGGCATC TIMP1 38: GACACTGTGCAGGCTTCAGT TIMP1 58: CAGACCACCTTATACCAGCG TIMP2 38: GTTGGAGGCCTGCTTATGGG TIMP2 58: TCTGGAAACGACATTTATGG Detection and Quantitation of PCR Products PCR products were separated on 2% agarose gels containing 25 ng/ml ethidium bromide and photographed. (Fig. 2). The photographs were scanned and band intensities measured using NIH image. Band intensity values were normalized based on molecular weight of the products. The log ratio of band intensities within each lane was plotted against the log of the copy number of template added per reaction. Quantity of target messages was determined where the ratio of template and targetband intensities was equal to 1. As the RNA used in the RT reaction is constant, the copy numbers per cell were extrapolated using 26 pg as the universally accepted standard of RNA per cell [Tarnuzzer et al., 1996]. RESULTS Baseline responses of cells within a 3D collagen matrix were determined for fibroblast MMP and TIMP gene expression after 24 hr in unloaded free floating collagen lattices. Since the cells themselves generate an Fig. 3. Bar diagram for unloaded collagen lattice comparing copy numbers/1,000 cells ⫾ S.D. for MMP1, 2, and 3 and TIMP1 and 2. endogenous contraction force, some localised tension between neighbouring cells is inevitable. However, it was possible to use untethered free-floating gel (i.e., with no reactive loading) as a control for nil external mechanical loading. Cells in free-floating gels did not align and were mostly stellate in shape. Since external loading was uniformly minimal these gels were analysed in total and not divided into zones as with subsequent gel models. Figure 3 shows the copy number/1,000 cells of MMP1, 2, and 3 and TIMP 1 and 2 for free-floating gels. The mean copy number among the MMPs was lowest for MMP3, and TIMP1 mean copy number was 410 ⫾ 34 as compared to 494 ⫾ 40 for TIMP2. Figure 4 compares the MMP mean copy number/ 1,000 cells from cells within the non-aligned ␦-zone and cells from the single-cue mechanically aligned zone. In each case ␦ and single-cue zone cells were analysed from the same gel. Non-aligned ␦-zone cells (Fig. 5) showed the highest total MMP expression. Cells that were aligned parallel to the applied mechanical load (Fig. 6) in the single-cue zone showed downregulation of total MMP expression relative to cells in the nonaligned ␦-zone zone. Molecular Responses to Conflicting Guidance Cues 5 Fig. 5. Picture of cells in the ␦ zone. Note stellate shaped cells with no orientation. Scale 20⫻. Fig. 4. Bar diagram for mechanically loaded gel comparing MMP1, 2, and 3 copy numbers/1,000 cells ⫾ S.D. in non-aligned delta zone and single cue mechanically aligned zone. Note: Down regulation of MMPs in single cue aligned zone. MMP1 mean copy number in the aligned zone was 50%, MMP2 was 22%, and MMP3 was 37% of that in the non-aligned ␦-zone within the same gel. This suggests that cells that have been mechanically loaded but are unable to align significantly increase their MMP expression. In contrast, cells that were able to align downregulated their MMP expression. MMP2 seemed to be the most responsive species to mechanical stimulii (over fourfold reduction in copy numbers). Figure 7 shows the comparable expression of TIMP 1 and 2 from cells within the non-aligned ␦-zone and cells from the single-cue mechanically aligned zone. In this case, aligned cells showed 53% TIMP1 and 63 % TIMP2 of mean copy number/1,000 cells when compared with non-aligned mechanically loaded cells within the same gel. There was no significant increase in TIMP expression by non-aligned loaded cells (␦ zone) relative to freefloating gels (Fig. 3) despite the upregulation of MMPs. However, TIMP expression in single cue mechanically loaded and aligned cells was down regulated when compared to cells from the ␦ zone. Although differential Fig. 6. Picture of cells aligning along lines of principal strain. Note the elongate and bipolar morphology. Scale 20⫻. responses were seen between MMP and TIMP in the ␦ zone, both species of TIMP responded in parallel throughout the experimental series. Figure 8 compares MMP expression of cells in non-aligned ␦ zone with that in the dual cue zone. MMP1 6 Mudera et al. Fig. 7. Bar diagram for mechanically loaded gel comparing TIMP1 and 2 copy numbers/1,000 cells ⫾ S.D. in non-aligned delta zone and single cue mechanically aligned zone. Note: Down regulation of TIMPs in single cue aligned zone is less than MMPs. copy number in the dual cue zone was 86% of that in the non-aligned zone, whilst MMP2 was 72% and MMP3 was 78%. These reductions in expression were modest and contrast dramatically with the fourfold down regulation of copy number in cells from the single cue aligned zone. This absence of down regulation of copy numbers occurred despite the fact that by the end of the 16 loading cycles, most of the cells had realigned to become parallel to the applied mechanical load, which was now predominant.(Fig. 9a,b). Indeed, 80% of cells were estimated to have realigned to become parallel to the applied mechanical load whilst 20% of cells were adherent to the fibronectin fibres and remained aligned along them, perpendicular to the applied load. The absense of predicted down regulation, despite 80% cell realignment, suggests that the cells that were still adherent and perpendicular to the loading contributed a disproportionately large proportion of the total signal, effectively obscuring the expected down regulation as seen in Figure 3. Figure 10 shows that TIMP1 and 2 show 83% and 86% of copy numbers in the dual cue zone when Fig. 8. Bar diagram comparing copy numbers/1,000 cells ⫾ S.D. of MMP1, 2, and 3 from non-aligned delta zone and dual cue zone. Note: Down regulation of MMPs is not as much as in single cue aligned zone. compared to the non-aligned zone. These figures again reflect a much lower down regulation in copy number than expected when compared to the down regulation by single cue mechanically aligned cells (Fig. 7). DISCUSSION Previous studies on topographic cues have shown that many cell types like fibroblasts, nuerites, oesteoblasts, and macrophage-like cells will take on alignment parallel to that of ridges or fibres of appropriate dimension in their substrate [Curtis and Wilkinson, 1997; Gray et al., 1996; Matsuzawa et al., 1994]. A specific form of contact guidance using aligned fibronectin fibres has been developed as a model here [Ejim et al., 1992; Underwood et al., 1999]. In this model, cells are aligned to discreet artificial fibres of fibronectin and migrate along those tracks in vitro [Wojciak-Stothard et al., 1997]. Another major organizing cue is known to be mechanical loading across the substrate. Mechanical loading in vitro using a Fig. 9. a: Photograph of cells in dual cue zone after 16 cycles of mechanical loading. Note: A layer of cells still aligned along fibronectin strand while the rest have realigned parallel to the direction of mechanical loading. Scale 20⫻. b: Higher magnification of cells in dual cue zone after 16 cycles of mechanical loading. Scale 40⫻. 8 Mudera et al. Fig. 10. Bar diagram comparing copy numbers/1,000 cells ⫾ S.D. of TIMP1 and 2 from non-aligned delta zone and dual cue zone. flex cell device [Butt et al., 1996] has been reported, though the complex nature of predominant mechanical forces in this model makes reliable correlation with cell behavior difficult. In fibroblasts, response to mechanical loading has been shown to result in an elongate bipolar morphology parallel with the axis of a uniaxial repetitive mechanical strain [Eastwood et al., 1998]. Using this same uniaxial strain model for mechanical loading fibroblasts, it has also proved possible to identify responsive species of matrix metalloproteases. In a previous study, gross changes in protein expression were monitored over prolonged experimental periods and total protease activity was measured by zymography [Prajapati, 1998]. This demonstrated that uniaxialy loaded fibroblasts appear to dramatically alter their rates of matrix remodeling and also that some species of MMPs are differentially modulated by loading, representing potentially good markers of cell activation. Having established that cells can be organized and aligned in a predictable manner in vitro by single cues, i.e., either contact guidance or mechanical load, the aim here has been to determine the cellular response to multiple conflicting cues. In effect, the interplay between cues has been tested in this model in an attempt to determine which form of cue is dominant in any given circumstance. Analysis of the cellular responses in this model was adapted to make use of QT-RT-PCR for the measurement of mRNA expression since direct measurement of protein measurement for this low number of cells is not presently feasible. In effect, this analysis of gene expression provides a short measure of cell activity at the time of the end of mechanical loading. The initial finding of this study was that uniaxial mechanical loading was able to reorganize cells that were not directly attached to the fibronectin fibre cues. Eighty percent of cells guided by these were aligned by cell-cell contact as opposed to direct cell-fibre contact (Fig. 9a,b). These cells in this conflicting, dual cue zone appeared to be under weaker guidance, since at the end of the experiment they were realigned parallel to the applied load. Analysis of MMP expression in the three zones— namely, non-aligned mechanical loaded, single cue uniaxial aligned load, and dual cue guided zones—indicated that the ability to align to the mechanical loading in the single cue zone led to a dramatic down regulation of message. Cells under the same mechanical load but without any predominant alignment (hence unable to become aligned) express far higher levels of MMP message than even cells in the unloaded collagen lattice, which also have no alignment. This is consistent with the idea that non-aligned, mechanically loaded cells are stimulated to move and remodel their matrix whilst the ability to align (and to minimise the perceived strain) reverses this process, potentially leading to matrix accumalation. The most differentially reactive species of the three tested was the gelatinase, MMP2. MMP2 has been implicated in other systems in the process of cell motility [Gianelli et al., 1997]. The additional differential response in the down regulation of MMPs (MMP2) is likely to be accentuated in effect since down regulation of TIMP1 and 2 was far less in the single cue zone, potentially increasing the TIMP:MMP ratio. The importance of cell alignment to applied load was supported by the minimal down regulation of MMP and TIMP in the dual cue zone. This was the zone where cells were predicted to be most sensitive to mechanical loading because of their orientation across the loading axis. Though less than 20% of the cells retained the perpendicular alignment, down regulation of signal was reduced by three- to fourfold, indicating that these cells produced disproportionately higher MMP and TIMP message, obscuring the down regulation of message of the remaining 80% of cells that did become aligned. Findings from the model provide a new insight into the Molecular Responses to Conflicting Guidance Cues differential response of cells to realistic cues, likely to occur in natural tissues or in vitro, in engineered tissues. In this instance, mechanical loading appeared to dominate over cell-cell contact guidance on previously aligned cells. However, cells directly in contact with the guiding substrate could not be realigned, and appeared to be maximally stimulated by the mechanical strain. It seems likely that cell movement and remodeling associated with this stress shielding may be mediated through expression of MMP activity. Further studies are in progress to characterize MMP responses as potential markers of mechanical activation. Practical application of guidance cues, for example, in tissue engineering, are likely to require multiple control cues, since contact guidance is effective in guiding and accelerating cell recruitment. At the same time, mechanical cues will inevitably be present as fibroblasts remodel and contract, collagenous matrices leading to complex and conflicting cues. ACKNOWLEDGMENTS We are grateful to Dr. Ann Lee and Ms. Kirsty Smith for photography. REFERENCES Arora PD, Bibby KJ, McCullogh CA. 1994. Slow oscillations of free intracellular calcium ion concentration in human fibroblasts responding to mechanical stretch. J Cell Physiol 161:187–200. Brown RA, Prajapati R, McGrouther DA, Yannas IV, Eastwood M. 1998. Tensional Homeostasis in dermal fibroblasts: Mechanical responses to mechanical loading in three-dimensional substrates. J Cellular Physiol 175:323–332. Burt AM, McGrouther DA. 1992. Production and use of skin cell cultures in theraputic situation. In: SpierRE, Griffiths JB, editors. Animal cell biotechnology. New York: Academic Press. p 150–168. Butt RP, Laurent GJ, Bishop JE. 1995. Mechanical load and polypeptide growth factors stimulate cardiac fibroblast activity. Ann NY Acad Sci 752:387–393. Chamson A, Sudre F, Le-Guen C, Le J, Rattner A, Frey J. 1997. Morphological alteration of fibroblasts mechanically stressed in collagen lattice. Arch Dermatol Res 289:596–599. Chiquet-Ehrismann R, Tannheimer M, Koch M, Brunner A, Spring J, Martin D. 1994. Tenasin C expression by fibroblasts is elevated in stressed collagen gels. J Cell Biol 127:2093–2101. 9 Curtis A, Wilkinson C. 1997. Topographical control of cells. Biomaterials 18:1573–1583. Eastwood M, Porter R, Khan U, McGrouther DA, Brown RA. 1996. Quantitative analysis of collagen gel contractile forces generated by dermal fibroblasts and the relationship to cell morphology. J Cell Physiol 166:33–42. Eastwood M, Mudera VC, McGrouther DA, Brown RA. 1998. Effect of precise mechanical loading on fibroblast populated collagen lattices: morphological changes. Cell Motil Cytoskeleton 40: 13–21. Ejim OS, Blunn GW, Brown RA. 1993. Production of artificially orientated mats and strands from plasma fibronectin: a morphological study. Biomaterials 14:743–748. Gianelli G, Falk-Marzillier J, Schiraldi O, Stetler-Stevenson WG, Quaranta V. 1997. Induction of cell migration by matrix metalloprotease-2 cleavage of laminin-5. Science 277:225–228. Gray C, Boyde A, Jones SJ. 1996. Topographically induced bone formation in vitro: implications for bone implants and bone grafts. Bone 18:115–123. Guido S, Tranquillo RT. 1993. A methodology for the systematic and quantitative study of cell contact guidance in oriented collagen gels. Correlation of fibroblast orientation and gel birefringence. J Cell Sci 105:317–331. He Y, Grinnell F. 1994. Stress relaxation of fibroblast activates a cyclic AMP signaling pathway. J Cell Biol 126:457–464. Lin YC, Ho CH, Grinnell F. 1998. Decreased PDGF receptor kinase activity in fibroblasts contracting stressed collagen matrices. Exp Cell Res 240:377–387. Matsuzawa M, Potember RS, Krauthamer V. 1994. Use of chemically patterened substrate to study directional effect of damaging electrical stimulation on cultured neuroblastoma cells. Brain Res 667:47–53. Oakley C, Jaeger NA, Brunette DM. 1997. Sensitivity of fibroblasts and their cytoskeletons to substratum topographies: topographic guidance and topographic compensation by micromachined grooves of different dimensions. Exp Cell Res 234:413–424. Prajapati R. 1998. The regulation of proteases and mechanical loading during fibroblast populated collagen lattice contraction. Ph.D. thesis. University of London. Tarnuzzer RW, Macauley SP, Farmerie WG, Caballero S, Ghassemifar MR, Anderson JT, Robinson CP, Grant MB, Humphreys-Beher MG, Franzen L, Peck AB, Schultz GS. 1996. Competitive RNA templates for detection and quantitation of growth factors, cytokines, extracellular matrix components and matrix metalloproteinases by RT-PCR. Biotechniques 20:670–674. Underwood S, Afoke A, Brown RA, MacLeod AJ, Dunnill P. 1999. The physical properties of a fibrillar fibronectin-fibrinogen material with potential use in tissue engineering. Bioprocess Engineering 20:239–248. Wojciak-Stothard B, Denyer M, Mishra M, Brown RA. 1997. A study of the adhesion, orientation and movement of cells cultured on ultrathin fibronectin fibres. In Vitro Cell Dev Biol 33:110–117.