Tissue stiffening promotes keratinocyte proliferation through activation of epidermal growth factor signaling

In the epidermis of the skin, the balance between keratinocyte proliferation in the basal layer and terminal differentiation and shedding in the upper layers maintains normal tissue homeostasis (Blanpain and Fuchs, 2009). These processes depend on a variety of extracellular cues and signals, such as soluble growth factors (Reiss and Sartorelli, 1987; Rheinwald and Green, 1977; Zhu and Watt, 1999), cell-cell adhesion (Green and Simpson, 2007; Niessen, 2007), and cell-extracellular matrix (ECM) interactions (Adams and Watt, 1989; Jones and Watt, 1993). Dysregulation of key extrinsic signaling pathways can lead to an imbalance in growth and differentiation and often contributes to the pathogenesis of skin diseases including chronic wounds (Herrick et al., 1992; Stojadinovic et al., 2005; Wysocki et al., 1993), blistering (Bruckner-Tuderman et al., 1989), and cancer progression (Gat et al., 1998; Martins et al., 2009; Reiss and Sartorelli, 1987; Uribe and Gonzalez, 2011).

Epidermal growth factor (EGF) signaling is one of the major regulatory axes controlling keratinocyte proliferation and survival. The EGF receptor (EGFR) is a receptor tyrosine kinase that is highly expressed in the basal layer of the epidermis and, upon binding of EGF ligands, such as EGF, amphiregulin and transforming growth factor α (TGF-α), the receptor dimerizes and becomes activated by autophosphorylation at multiple tyrosine (Y) residues (Jost et al., 2000). Under homeostatic conditions, EGF signaling promotes growth and survival of basal keratinocytes through downstream activation of mitogen activated protein kinase (MAPK) and phosphotidylinositide 3-kinase (PI3K) signaling pathways (Assefa et al., 1997; Wan et al., 2001). However, overexpression of EGFR or its ligands is associated with a variety of hyperproliferative conditions, such as psoriasis (Piepkorn, 1996) and cancer (Reiss and Sartorelli, 1987; Uribe and Gonzalez, 2011).

While the roles of many biochemical factors in the regulation of keratinocyte function have been described in detail, little is known about the contribution of mechanical or biophysical cues. In our previous studies, we used micro-patterned substrates and established that simple changes in keratinocyte shape and adhesion are potent regulators of terminal differentiation (Connelly et al., 2010). Similarly, reduced tethering of ECM molecules to cell culture supports can induce terminal differentiation (Trappmann et al., 2011). While bulk material stiffness appears to have little effect on keratinocyte differentiation, the impact on additional cell functions or fate over longer time scales has yet to be determined. As tissue stiffness regulates the proliferation and self-renewal of multiple cell types, including mammary epithelia (Klein et al., 2009; Paszek et al., 2005), muscle-derived stem cells (Gilbert et al., 2010), hematopoietic stem cells (Lee-Thedieck et al., 2012) and mesenchymal stem cells (Chowdhury et al., 2010), it may also be an important mediator of epidermal keratinocyte growth.

In the present study we investigated the effects of altered matrix stiffness on keratinocyte behavior using model silicone substrates. We show that increased matrix stiffness promotes epidermal proliferation independently of changes in cell fate, and that EGF signaling mediates this response. We also demonstrate that EGF signaling is elevated within keloid scar tissue, which is ∼30-fold stiffer than normal skin. These findings provide significant insights into the mechanisms of mechanosensing within the epidermis, and their impact on tissue homeostasis and scar formation.

To investigate the influence of matrix stiffness on long-term keratinocyte growth and differentiation, we generated cell culture substrates with defined elastic moduli using polydimethylsiloxane (PDMS). PDMS substrates were crosslinked with 2% or 20% (w/w) curing agent to produce non-porous substrates with elastic moduli of 180 kPa or 2 MPa, respectively (Fig. S1). Our previous atomic force microscopy (AFM) analysis of normal skin measured the elastic modulus of the basement membrane to be ∼140 kPa (Kao et al., 2016). Thus, the PDMS substrates with 2% curing agent were most similar to normal skin, while the substrates crosslinked with 20% curing agent represented a ten-fold increase in matrix stiffness.

Primary human keratinocytes were seeded onto 2% or 20% PDMS substrates at clonal density and cultured for 10 days in low-Ca²⁺, serum-free medium with 0.1 ng/ml EGF. Cells formed a similar number of colonies on both substrates but the colonies on the stiff 20% substrates were significantly larger (Fig. 1A,C). Tracking of individual colonies over the first seven days revealed a more rapid, exponential increase in the number of cells per clone on the 20% substrates compared to the 2% substrates (Fig. 1D), and keratinocytes on the stiff substrates had a higher proliferative rate at day 7 (Fig. 1E,F). There were no detectable differences in initial adhesion or viability of keratinocyte cultured on the soft and stiff substrates (Fig. S2).

To assess whether substrate stiffness influenced epidermal cell fate, keratinocytes were first expanded clonally on 2% or 20% PDMS for 10 days, then dissociated and expanded a second time on 2% or 20% PDMS. Keratinocytes formed similar numbers of colonies under all conditions (Fig. 1G,H), indicating that previous exposure to a soft or stiff environment did not affect the proportion of colony-initiating cells within the culture, a common read-out of epidermal stem cell function in vitro (Jones and Watt, 1993). In addition, expression of the terminal differentiation marker involucrin, was similar for cells cultured on 2% or 20% PDMS for 5 days followed by stimulation with Ca²⁺ (1.8 mM CaCl₂) for 2 days to induce terminal differentiation (Fig. 1I). Likewise, there were no striking differences in Ca²⁺-induced assembly of adherens junctions over this range of substrate moduli (Fig. S4). Taken together, these findings indicate that increased substrate stiffness specifically stimulates keratinocyte proliferation but does not affect adhesion, survival or terminal differentiation. We conclude that matrix stiffness regulates keratinocyte growth independently of changes in cell fate.

To gain insight into the mechanism by which matrix stiffness regulates keratinocyte proliferation we first examined the effects on EGF signaling, a key regulator of proliferation and survival (Jost et al., 2000). Cells were cultured on 2% or 20% PDMS substrates with or without collagen coating for 24 h in EGF-free medium, then stimulated with 10 ng/ml EGF for 15 min. High levels of EGFR phosphorylation at Y1068 were observed on all substrates following EGF stimulation (Fig. 2A), but there was a significantly higher level of basal EGFR phosphorylation only on the stiff, collagen-coated substrates (Fig. 2A,B). Increased EGFR phosphorylation on stiff substrates prior to stimulation was also detected by immunofluorescence staining, while receptor internalization following treatment with EGF appeared to be unaffected by substrate stiffness (Fig. 2C). A similar increase in the basal level of phosphorylated Y845, Y1086 and Y1173 was also observed on the stiff substrates (Fig. 2D, Fig. S3A).

The effects of matrix stiffness on EGF signaling occurred in a dose-dependent manner. EGFR phosphorylation progressively increased with increasing substrate stiffness for PDMS crosslinked with 2%, 5%, 10% or 20% curing agent (Fig. 2D). There was also elevated EGFR phosphorylation in keratinocytes cultured on stiff substrates coated with fibronectin; however, the response on fibronectin was associated with increased total EGFR, suggesting ECM-specific effects as well (Fig. S3B). Finally, western blot analysis revealed higher levels of phosphorylated ERK1/2 (pERK) and Akt (pAkt) – downstream targets of EGFR – on stiff PDMS substrates (Fig. 2F,G). We conclude that increased matrix stiffness promotes activation of the EGF signaling pathway in keratinocytes.

To determine the functional role of EGF signaling in the growth response of keratinocytes to increased matrix stiffness, we performed colony-forming assays on 2% or 20% PDMS substrates with a range of different EGF concentrations. In the absence of exogenous EGF, keratinocytes formed colonies only on the stiff PDMS substrates (Fig. 3A). At low EGF concentrations (0.1 and 1 ng/ml), there were no differences in the number of colonies formed, but colony size was significantly greater on the stiff, 20% PDMS (Fig. 3B,C). Colony size increased with increasing EGF concentrations on the soft 2% substrates and, at the highest dose of 10 ng/ml, the difference in colony size between soft and stiff substrates was not statistically significant (Fig. 3B,C). Clonal growth was completely blocked by treatment with the EGF inhibitor AG1478 (Fig. 3A,C). We conclude that EGF signaling mediates the effects of matrix stiffness on keratinocyte proliferation. Moreover, a minimal level of EGF signaling is required by keratinocytes to initiate colony formation but, at higher concentrations, EGF primarily regulates colony size.

We tested whether autocrine growth factor signaling was responsible for the increased growth on stiff substrates. Expansion of cells on PDMS substrates in the presence of conditioned medium from keratinocytes on 2% or 20% PDMS enhanced overall clonal growth, but there were no observable differences between the effects of medium from cells on either substrate (Fig. S3C). Similarly, there were no measurable differences in the level of amphiregulin, the primary EGF-family ligand produced by keratinocytes (Piepkorn et al., 1994), released into the medium by cells on 2% or 20% substrates (Fig. S3D). These findings suggest that matrix stiffness regulates EGFR activation through an intrinsic signaling mechanism rather than altered production of EGF ligands.

We next examined how stiffness-induced changes in EGF signaling depended on mechanical linkage with the ECM. Consistent with previous findings (Trappmann et al., 2011), there were no measurable differences in cell spreading or organization of the F-actin cytoskeleton between keratinocytes cultured on 2% or 20% PDMS substrates (Fig. 4A,B). However, cells on the stiff PDMS surfaces displayed significantly more focal adhesions detected by paxillin immunofluorescence (Fig. 4C,D). A similar response was observed for vinculin, whereas there were no differences in overall expression of β1 integrin (Fig. S4). Moreover, paxillin expression, as well as focal adhesion kinase (FAK) phosphorylation at Y397, increased in a dose-dependent manner with increasing substrate stiffness (Fig. 4E,F). Together, these results indicate that over this range of elastic moduli, substrate stiffening increases focal adhesion number and total FAK activation.

To investigate the crosstalk between focal adhesion and EGFR signaling, we first assessed the direct interaction (within 30–40 nm) of EGFR with focal adhesions (i.e. paxillin) by using a proximity ligation assay. There was a significantly higher interaction signal between EGFR and paxillin on 20% PDMS compared to 2% PDMS, and this relationship was reversed when acto-myosin contractility was inhibited with Blebbistatin (Figs 5A,B and 4S). Increased levels of phosphorylated EGFR (pEGFR) (Y1086) on 20% PDMS also colocalized with paxillin (Fig. S3A). Moreover, the disruption of cytoskeletal tension reversed the effects of substrate stiffness on EGFR phosphorylation, with higher levels on 2% PDMS compared to 20% PDMS when treated with Blebbistatin (Fig. 5C), while latrunculin treatment blocked the effects of substrate stiffness on EGFR phosphorylation and reduced overall receptor expression (Fig. 5C). Additionally, we used siRNA to partially knock down vinculin and FAK (Fig. S4). Reduction in total FAK led to higher EGFR phosphorylation on 2% PDMS compared to 20% PDMS, while reduced vinculin levels caused total EGFR to be downregulated, with no differences in phosphorylation between soft and stiff substrates (Fig. 5D). Together, these results indicate that the balance between ECM stiffness and F-actin cytoskeletal tension regulates the recruitment of EGFR to focal adhesions and that activation depends on FAK.

Functionally, cytoskeletal tension and FAK activity were important for keratinocyte growth in colony formation assays. Treatment with Blebbistatin reversed the effects of substrate stiffness on colony size, consistent with the effects on EGFR phosphorylation, while FAK inhibition with FAK inhibitor 14 completely blocked all colony formation (Fig. 5E). These results demonstrate that focal adhesion assembly and signaling, as well as tension within the F-actin cytoskeleton, are required for stiffness-dependent changes in EGFR phosphorylation and clonal growth in keratinocytes.

Finally, to explore the physiologic significance of mechanically-regulated EGF signaling within the skin we compared the biomechanics of normal skin with stiff keloid scar tissue. Keloids are severe, injury-induced scars that expand beyond the initial wound margins and are characterized by excessive ECM production and hyperproliferation (Andrews et al., 2016). They are believed to be stiffer than normal skin (Huang et al., 2016), but the mechanical properties have not been formally established yet. We performed micro-indentation testing of the dermis of normal adult skin and the expanding margins of keloid scars (Fig. 6A,B), and the tangent modulus of the stress-strain curves at 30% strain was calculated as previously described (Delaine-Smith et al., 2016). The average modulus of normal skin samples was ∼6.1±2.9 kPa (mean±s.d.), and the moduli of the keloid samples were significantly greater (10×–100×), ranging from 50–650 kPa (Fig. 6B,C). Compared to our previous AFM analysis, the lower absolute values of moduli measured by micro-indentation were most likely to be due to a larger test area, which included the softer dermal tissue than the basement membrane alone (Kao et al., 2016).

In conjunction with mechanical testing, we analyzed EGFR phosphorylation within the epidermis of keloid scars by immunofluorescence staining and compared the levels of pEGFR to the extra-lesional (uninvolved) epidermis adjacent to the scar. Increased levels of pEGFR (Y0168) within the keloid scar could be observed in two out of three frozen sections examined, and levels of pEGFR (Y845) were ∼50% higher within the scarred skin across all three patient samples (Fig. 6D,E). Consistent with our in vitro studies, these findings demonstrate that stiffening of the underlying dermis in keloids scars corresponds with EGFR activation in the epidermis and suggest that tissue mechanics also regulates EGF signaling in vivo.

In the present study we employed silicone-based biomaterials with tunable mechanical properties to investigate the effects of matrix stiffening on epidermal growth and differentiation. Our findings provide clear evidence that elevated matrix stiffness beyond normal physiologic levels stimulates keratinocyte proliferation but does not affect the ability to self-renew or terminally differentiate. Thus, matrix stiffness specifically regulates keratinocyte proliferation independently of effects on stem cell fate. Recent studies have shown that cultured human keratinocytes switch between expanding and balanced modes of growth, which depend on cell-cell contact and EGF signaling (Roshan et al., 2016). Our findings support this model and identify matrix stiffness as a key upstream regulator. In addition, another recent study demonstrated that increased matrix stiffness promotes directional migration of HaCaT keratinocytes (Wickert et al., 2016). Together, these results and our own establish a mechanism for how keratinocytes sense and respond to changes in bulk material elasticity.

Mechanistically, we show that EGF signaling mediates the growth response of human keratinocytes to altered matrix stiffness, and phosphorylation of EGFR depends on focal adhesion assembly and acto-myosin contractility. Moreover, this response involves direct interaction between EGFR and focal adhesions, suggesting that focal adhesion signaling molecules, such as FAK, may regulate EGFR activity. However, it is also possible that EGFR activation regulates focal adhesion assembly, and potential bi-directional signaling and feedback mechanisms will be an important area of future investigation. Our findings are consistent with previous studies in normal and cancerous mammary epithelia (Klein et al., 2009; Paszek et al., 2005; Wang et al., 1998), as well as recent findings, which link stiffness-dependent EGFR activation to Src-family kinases in fibroblasts (Saxena et al., 2017). Thus, bio-mechanical regulation of focal adhesion assembly may play a key role in modulating EGF signaling across diverse cell types.

In addition to cell-matrix adhesions, cell-cell adhesions also contribute to the biomechanical regulation of EGFRs. In simple epithelia, increased substrate stiffness inhibits adherens junction assembly, which in turn enhances EGF sensitivity and proliferation (Kim and Asthagiri, 2011). Recent studies also suggest that, in the epidermis, elevated acto-myosin tension within adherens junctions in the granular layer negatively regulates EGFR activity (Rübsam et al., 2017). While our studies here aimed to establish the direct signaling between the ECM and EGFR by using sparse, low-Ca²⁺ culture conditions, the potential crosstalk with cell-cell adhesion mechanics cannot be completely excluded. It will be interesting in future work to explore how biomechanical cues from both cell-ECM and cell-cell adhesions are integrated and regulate keratinocyte function under more confluent, in vivo-like conditions. Direct measurement of forces at cell-cell adhesions (Borghi et al., 2012) and localization of activated EGFR within different mechanical environments will be of particular interest. Likewise, the role of additional cell adhesion receptors, such as desmosomes (Broussard et al., 2017), and other growth factor receptors (Conway et al., 2013) in cellular mechanosensing will be important areas of future investigation. Finally, it will also be necessary to consider the dynamics of focal adhesion turnover and EGFR recycling, as EGFR is recycled together with β1 integrins (Caswell et al., 2008).

Although many studies have investigated the effects of substrate mechanics on cell function using controlled in vitro models, only a few have linked in vivo biomechanics to normal or pathological functions (Gilbert et al., 2010; Levental et al., 2009). Our results indicate that EGFR phosphorylation correlates with increased tissue stiffness in human skin and may play a role in keloid scar pathogenesis. While genome-wide association studies have only linked a handful of genes to keloid susceptibility (Nakashima et al., 2010), the underlying causes of keloid scar formation remain almost completely unknown. It is interesting to note that keloids often develop in areas of skin with high tension (Andrews et al., 2016; Huang et al., 2016; Ogawa et al., 2012), which combined with our findings, further supports a role for biomechanics in scar formation. Future studies examining the inter- and intra-keloid heterogeneity in mechanics, as well as the crosstalk with growth factor signaling and epidermal cell mechanics, will be important and will hopefully shed new light on these disfiguring and often painful conditions.

PDMS substrates (Sylgard 184, Dow Corning) were prepared by mixing the PDMS base with crosslinker at ratios varying from 50:1 (2%) to 5:1 (20%). PDMS mixtures were de-gassed under a vacuum, spread onto 13 mm diameter glass coverslips or 6-well plates, and cured overnight at 70°C. To functionalize the PDMS substrates with ECM, the surfaces were covered with a solution of 50 mg/ml sulfo-SANPAH (Thermo Scientific) in water and exposed to 365 nm UV light for 10 min. This process was repeated twice, followed by incubation with either 50 μg/ml rat type I collagen (BD Biosciences) or 50 μg/ml human plasma fibronectin. Samples were rinsed three times with PBS, and sterilized with UVB and 70% ethanol prior to cell seeding. All chemicals were from Sigma-Aldrich unless otherwise noted.

Primary human keratinocytes were isolated from neonatal foreskin and maintained on a layer of J2 3T3 fibroblasts in FAD medium as previously described (Rheinwald and Green, 1977). For studies on PDMS substrates, fibroblasts were removed using Versene (Invitrogen), and keratinocytes (passage 2–6) were trypsinized and seeded onto PDMS substrates in keratinocyte serum-free medium (KSFM, Invitrogen) supplemented with bovine pituitary extract, penicillin/streptomycin, and 0.1 ng/ml EGF. To induce terminal differentiation, keratinocytes were cultured in KSFM further supplemented with 1.8 mM CaCl₂.

Skin samples were obtained from keloid patients and healthy volunteers from the plastic surgery department at Barts Health NHS Trust. All tissue samples were from dark skinned (South Asian or Afro-Caribbean) adult donors (male and female, under the age of 50); body sites included the back, shoulder, chest and stomach. All subjects gave informed consent and the study was conducted under local ethical committee approval (East London Research Ethics Committee, study no 2011-000626-29).

Primary keratinocytes were seeded onto non-ECM coated PDMS surfaces within a 6-well plate at a density of 1000/well. Cells were cultured for 10 days in KSFM supplemented with 0–10 ng/ml EGF, 10 µM AG1478, 50 µM Blebbistatin, or 416 nM FAK inhibitor 14 (Tocris Bioscience, Bristol, UK). Cells were fixed with 4% paraformaldehyde (PFA), stained for 30 min with 0.06% Crystal Violet, and rinsed copiously with water. Scanned images of the stained wells were analyzed with ImageJ to quantify colony size and number.

To measure DNA synthesis, keratinocytes were cultured on PDMS substrates for 7 days in KSFM then incubated with 10 µM 5-ethynyl-2′-deoxyuridine (EdU) for 1 h at 37°C and rinsed twice with Edu-free KSFM. Cells were fixed with 4% PFA, and EdU incorporated into the DNA was tagged with AlexaFluor-568 by using the ‘Click-It’ kit (Thermo Scientific) according to the manufacturer's instructions. Samples were co-stained with DAPI and imaged using a Leica DM5000B microscope.

Cells were washed in PBS and incubated in radioimmunoprecipitation assay (RIPA) buffer plus protease and phosphatase inhibitors for 10 min on ice. Cells were scraped off the dish, briefly sonicated, and centrifuged (300 g for 5 min) to remove insoluble material. Protein concentration was determined by the BCA assay (Thermo Scientific). Lysates were combined with loading buffer (Thermo Scientific) and 1% β2-mercaptoethanol, and equal amounts of total protein were resolved on a 4% or 10% polyacrylamide gel (Bio-Rad) and transferred onto nitrocellulose membranes (GE Lifesciences). Membranes were blocked for 1 h in either 5% non-fat dry milk or 3% BSA, before being incubated either 1 h at room temperature or overnight at 4°C with primary antibodies against Involucrin [SY7, CRUK (1:1000)], pEGFR Y1068 [cat. no. 3777, Cell Signaling Technology, Tyr1068 (D7A5) Rb XP (1:1000)], EGFR [cat. no. 4267, Cell Signaling Technology (D38B1) Rb XP (1:1000)], pERK1/2 [cat. no. 9106S, Cell Signaling Technology (1:1000)], ERK1/2 [cat. no. 4695S, Cell Signaling Technology (1:1000)], pAkt [cat. no. 9271, Cell Signaling Technology Ser473 Rb (1:1000)], Akt [cat. no. 9272, Cell Signaling Technology Rb (1:1000)], Paxillin [cat. no. 610569, BD Biosciences (1:1000)], pFAK [cat. no. 611722, BD Biosciences (1:500)], FAK [cat. no. 0537, Millipore (1:500)] or GAPDH [cat. no. ab9485, Abcam Rb (1:2000)]. Secondary detection was performed with HRP-conjugated anti-rabbit or anti-mouse antibodies (1:5000, Dako). Proteins were visualized using the enhanced chemiluminescence detection system (Millipore, Watford, UK).

The enzyme-linked immunosorbent assay (ELISA) for amphiregulin was performed on conditioned medium from cells cultured on PDMS substrates for 24 h according to the manufacturer's protocol for the DuoSet ELISA development kit (R&D Systems, Abingdon, UK). Briefly, ELISA MaxiSorp 96-well plates (Thermo Scientific) were coated overnight at room temperature with the capture antibody diluted in a 1% BSA/PBS solution. Plates were washed three times with 0.05% Tween 20 in PBS and blocked with 1% BSA for 1 h. Following a second wash step, 100 µl of the conditioned medium samples (pre-diluted 1:10) were incubated at room temperature for 2 h. The plate was washed three times, incubated with a biotinylated anti-human amphiregulin antibody (from the kit) for 2 h at room temperature, and washed three more times. Detection was performed by incubation with 100 µl horseradish peroxidase (HRP)-tagged streptavidin for 20 min, followed by washing and incubation with the substrate solution for 20 min. The reaction was stopped with 50 µl Stop Solution and read at 450 nm using a Synergy HT plate reader.

For immunofluorescence staining, keratinocytes on PDMS substrates were fixed with 4% PFA and permeabilized with 0.1% Triton X-100 for 5 min. Samples were blocked with 10% FBS plus 0.25% gelatin in PBS for 1 h and incubated overnight at 4°C with primary antibodies against pEGFR Y1068 (as above, 1:500), pEGFR Y1086 [cat. no. ab32086, Abcam (1:500)], paxillin (as above, 1:500), vinculin [hVIN-1; cat. no. V9131, Sigma-Aldrich (1:1000)], E-cadherin [HECD1; cat. no. ab1416, Abcam (1:100)] or total EGFR (as above, 1:100). Secondary staining of pEGFR antibodies was performed with anti-rabbit AlexaFluor-488 [cat. no A11008, Thermo Scientific (1:1000)], and anti-mouse AlexaFluor-568 or-488 [cat. nos A10037 and A11001, Thermo Scientific (1:1000)]. F-actin was labeled with phalloidin-AlexaFluor-568 or -488 (1:500, Sigma-Aldrich) included in the secondary solution. For the proximity ligation assay, fixed samples were co-stained with antibodies against paxillin and EGFR, and detected using the Duolink Mouse-Rabbit Red kit (Sigma-Aldrich) according to the manufacturer’s instructions. Samples were mounted on glass microscope slides with Mowiol and imaged using a Zeiss 710 confocal microscope (Carl Zeiss) and 20× or 63× objectives.

Normal skin and keloid tissue samples were either snap frozen in optimal cutting temperature (OCT) compound (BD Biosciences) or fixed with 4% paraformaldehyde and embedded in paraffin. Freshly cut frozen sections were fixed for 10 min in 4% paraformaldehyde, blocked with 10% FBS and 0.25% gelatin, and stained with anti-pEGFR Y1068 (as above, 1:300). Paraffin sections were de-waxed, and heat-mediated antigen retrieval was performed with 10 mM sodium citrate. Sections were blocked as before and stained with anti-pEGFR Y845 (1:100, Cell Signaling, 6963S). Slides were imaged with a Leica DM5000B microscope, and mean fluorescence intensity of the basal layer was quantified with ImageJ.

Keratinocytes were seeded onto collagen-coated PDMS surfaces in a 6-well plate and cultured overnight. Cells were transfected with 4 pmole siRNA and 4 µl of Jet Prime reagent (Polyplus Transfection) per well according to the manufacturer's instructions. Cells were cultured for 72 h and harvested for western blot analysis. Silencer Select (Thermo Fisher) validated small interfering RNAs (siRNAs) were used for PTK2 (s11486) and VCL (s14764) knockdown.

Apoptosis and viability were analyzed by flow cytometry. Keratinocytes were cultured on PDMS substrates for 7 days, trypsinized, resuspended in 200 µl Annexin V buffer (50 mM HEPES, 700 mM NaCl, 12.5 CaCl₂) plus 5 µl Annexin V-FITC (Invitrogen) and DAPI, and incubated at room temperature for 15 min. Annexin V-positive (apoptotic) and DAPI-positive (dead) cells were analyzed by flow cytometry using the Becton Dickinson LSRII. As a positive control for apoptosis, cells were treated with 10 mJ/cm² UVB light 20 h prior to analysis.

For the mechanical testing of the PDMS, samples were cast into Petri dishes and, after curing were cut to an approximate size of 5 mm wide×25 mm long×1.5 mm thick. Tensile testing was conducted using an Instron universal testing system. The PDMS samples were held in place using pneumatic grips and tested to failure at a strain rate of 20 mm/min. The elastic modulus was calculated from the slope of the stress versus strain curve. Six samples per setting (2% and 20%) were tested.

For scanning electron microscopy (SEM) analysis of surface topography, cover slips were coated with PDMS and gold-coated using an Agar auto sputter coater. SEM imaging was conducted using an FEI Inspect-F with an accelerating voltage of 5 kV and a working distance of 10 mm. The roughness measurements were performed using AFM (nTegra, NT-MDT). The samples were imaged over a 10×10 μm area in semi-contact mode with a silicon nitride cantilever (MLCT, Bruker, spring constant k=0.6 N/m). The sample height at each point within the image and average roughness were measured using the AFM software (Nova, NT-MDT).

The tissue moduli of keloid and normal skin samples were measured by micro-indentation. Frozen specimens were fully thawed at room temperature in PBS for 1 h before testing. Mechanical indentation was performed using an Instron ElectroPuls E1000 (Instron) equipped with a 10 N load cell (resolution=0.1 mN). Specimens were indented using a stainless steel plane-ended cylindrical punch with a diameter (Ø_i) of 2 or 1 mm. Specimen thickness (T_s) was measured as the distance between the base of the test dish and top of the sample, each detected by applying a pre-load of 0.3 mN. Specimen diameter (Ø_s) was measured using electronic callipers. Indentation was performed at room temperature with specimens fully submerged in PBS throughout testing. Tests were performed using a ramped displacement-control regime whereby each specimen was displaced to 30% of their measured thickness at a rate of 1% s⁻¹. The resulting load detected from the sample was recorded at 10 Hz. To minimise errors in calculations of tissue moduli, specimen to indenter ratios were kept to Ø_s:Ø_i ≥4:1 and T_s:Ø_i ≤2:1.

The geometric correction factor к accounts for large-deformation, non-linear behavior and values for strains >15% can be determined from linear interpolation (Zhang et al., 1997). The second geometrical correction factor, G_к, is applied from Delaine-Smith et al. (2016). Poisson's ratio (ν), was assumed to be 0.499 for all specimens.

All data were analyzed by ANOVA and Tukey's test for post-hoc analysis with sample size of independent experiments or patients indicated in the figure captions.

We thank Dr Gary Warnes for assistance with annexin V analysis and Oscar Pundel for maintenance of primary keratinocyte cultures.