Palladin mediates stiffness-induced fibroblast activation in the tumor microenvironment.

Biophysical Journal Volume 109 July 2015 249–264

249

Article Palladin Mediates Stiffness-Induced Fibroblast Activation in the Tumor Microenvironment Joshua S. McLane1,2 and Lee A. Ligon1,2,* 1

Center for Biotechnology and Interdisciplinary Studies and 2Department of Biological Sciences, Rensselaer Polytechnic Institute, Troy, New York

ABSTRACT Mechanical properties of the tumor microenvironment have emerged as key factors in tumor progression. It has been proposed that increased tissue stiffness can transform stromal fibroblasts into carcinoma-associated fibroblasts. However, it is unclear whether the three to five times increase in stiffness seen in tumor-adjacent stroma is sufficient for fibroblast activation. In this study we developed a three-dimensional (3D) hydrogel model with precisely tunable stiffness and show that a physiologically relevant increase in stiffness is sufficient to lead to fibroblast activation. We found that soluble factors including CC-motif chemokine ligand (CCL) chemokines and fibronectin are necessary for this activation, and the combination of C-C chemokine receptor type 4 (CCR4) chemokine receptors and b1 and b3 integrins are necessary to transduce these chemomechanical signals. We then show that these chemomechanical signals lead to the gene expression changes associated with fibroblast activation via a network of intracellular signaling pathways that include focal adhesion kinase (FAK) and phosphoinositide 3-kinase (PI3K). Finally, we identify the actin-associated protein palladin as a key node in these signaling pathways that result in fibroblast activation.

INTRODUCTION Tumors are generally much stiffer than normal tissue. Breast cancer tumors have been shown to be 10 to 20 times stiffer than normal breast tissue. The stromal tissue surrounding the tumor, known as the tumor microenvironment (TME), is also typically stiffer than normal tissue, although usually only three to five times stiffer than normal stroma (1,2). This increase in stromal stiffness has been implicated in changes in both tumor cells and in tumor associated stromal cells such as fibroblasts, which are sensitive to mechanical properties of the extracellular matrix (ECM) (1–8). Fibroblasts are the workhorses of stromal tissue. They are responsible for building and remodeling the ECM and also have critical roles in wound healing, the regulation of inflammation, and control of epithelial differentiation (4,9). In the TME, fibroblasts are activated to a myofibroblast-like phenotype, commonly referred to as carcinomaassociated fibroblasts (CAFs), and frequently undergo genetic mutations such as tumor cells (9–11). Myofibroblasts are characterized by their elongated spindle-like morphology and contractile phenotype, as well as by changes in patterns of protein expression, including increases in a-smooth muscle actin (aSMA), fibroblast activation protein (FAPa), the EDA splice variant of fibronectin (FN-EDA), and palladin, which together can be used to identify differentiated myofibroblasts (9–12). These activated fibroblasts are thought to play key roles in promoting

Submitted May 11, 2015, and accepted for publication June 17, 2015. *Correspondence: [email protected]

both tumor growth and metastasis by paracrine and juxtacrine interactions with both tumor and other tumor associated cells (9–14). Both soluble factors such as TGFb and increases in matrix stiffness have been shown to contribute to fibroblast activation (5,6,15,16). Similarly, three-dimensional (3D) matrices derived from CAFs can induce fibroblast activation and the maintenance of this phenotype is dependent on matrix interactions (17,18). Some studies have shown that increased stiffness alone can lead to an activated fibroblast phenotype, but these studies were done in two-dimensional (2D) culture and examined the effects of very large changes in stiffness, rather than the more subtle changes seen in the TME (4,15), so it remains unclear if the physiological stiffness associated with the TME is sufficient to cause fibroblast activation. Collagen-I is the major component of the breast stroma. To develop a biomimetic 3D culture environment, we developed a collagen-I hydrogel system in which we can tune mechanical stiffness (19,20). The stiffness of a collagen-I hydrogel can be altered by varying parameters involved in the formation of the hydrogel such as pH, temperature, and collagen concentration (19,21,22), but these methods alter the structure and/or ligand availability of the collagen-I, potentially inducing changes in cell phenotype independent of the changes in stiffness. To avoid this complication, we developed a system in which we can control the hydrogel stiffness independently of collagen concentration, gelation temperature, or pH by cross-linking the collagen with short (12-atom) bio-inert polyethylene

Editor: Alissa Weaver. Ó 2015 by the Biophysical Society 0006-3495/15/07/0249/16 $2.00

http://dx.doi.org/10.1016/j.bpj.2015.06.033

250

glycol polymers that have been functionalized by the addition of N-hydroxysuccinimide groups—poly(ethylene glycol)-di(succinic acid N-hydroxysuccinimide ester (PEG-diNHS). PEG-diNHS can cross-link collagen-I fibrils by forming amide bonds to tether collagen molecules together, mimicking the physiological cross-links formed in vivo (23). These collagen-I PEG-diNHS hydrogels have been previously used in studies of tumor spheroid formation and in tissue engineering, and show good biocompatibility (24,25). Palladin is an actin associated Ig-domain protein that is overexpressed in many cancers and particularly in CAFs (26–32). Palladin has been shown to regulate actin dynamics and adhesion formation (30,32). It can cross-link actin filaments directly (33), as well as bind to multiple actin associated proteins such as VASP (34), profilin (35), and a-actinin (36). Palladin is associated with mechanosensitive structures such as focal adhesions (37) and stress fibers (32), suggesting that it may play a role in cellular mechanosensation or the transduction of mechanical signals. More recently, palladin has also been shown to act as a transcriptional regulator, controlling the expression of smooth muscle cell genes such as aSMA (12,38,39). Palladin has been shown to localize to the nucleus in some cell types, which may be consistent with its role in transcription (40,41) but the regulation and balance of these two distinct roles in the actin cytoskeleton and in transcription remains unclear (42). However, these two roles of palladin in both transcriptional regulation and mechanosensation make it a strong candidate to play a key role in the regulation of fibroblast activation by mechanical signals. In this study we sought to test whether the modest three to five times increase in stiffness seen in the TME is sufficient to activate fibroblasts to a myofibroblast-like state. We observed changes in morphology and gene expression consistent with activation when cells were embedded in stiffened 3D hydrogels, but that the presence of soluble factors including CC-motif chemokine ligand (CCL) chemokines and fibronectin were necessary for activation. We found that these chemomechanical signals were transduced via the C-C chemokine receptor type 4 (CCR4) chemokine receptor and b1 and b3 integrins, which then signal to intracellular signaling pathways that include focal adhesion kinase (FAK) and phosphoinositide 3-kinase (PI3K). Finally, we show that the actin-associated protein palladin is a key node in this network of signaling that results in fibroblast activation. MATERIALS AND METHODS

McLane and Ligon sodium pyruvate (Corning), and 1% penicillin/streptomycin (Corning) and maintained at 5% CO2. Media was replaced two to three times a week during standard cell culture. Cells were cultured from a minimum of 10% to a maximum of 75% confluence and only cells from passages 2 to 8 were used.

Experimental conditions For reduced serum experiments, cells were cultured in reduced serum (1%) media (RSM) for 3 days before hydrogel encapsulation. After encapsulation, fresh RSM was added to hydrogels. All conditions received fresh media at the beginning of the 3 day experimental time course, and media was not replaced during experimental time course. For 2D experimental samples, 100,000 cells were seeded into 10 cm culture dishes (product 172958, Thermo Scientific, Waltham, MA). Collagen-coated 2D dishes were coated with collagen-I solution (product C9791, Sigma-Aldrich, St. Louis, MO) at 10 mg/mL overnight at 37 C. Solution was aspirated, and dishes were allowed to dry under ultraviolet illumination before use.

Preparation of collagen-I hydrogels Hydrogels were composed of calf skin derived collagen-I (product 150026, MP Biomedicals, Solon, OH) resuspended in 0.02 N acetic acid at 3 mg/mL. Collagen-I solution was combined with neutralizing solution (0.52 M sodium bicarbonate, 0.4 M HEPES, and 0.08 N sodium hydroxide), cellular suspension media (DMEM þ 5% BSA for blocking) with or without cells, and PEG-diNHS dissolved in dimethyl sulfoxide (DMSO; 100 mg/mL, product E3257, Sigma-Aldrich) or DMSO control. The ratio for gel formation was 615: 308: 77: 4 for collagen-I: suspension media: neutralizing solution: PEG-diNHS/DMSO. Experimental cell culture hydrogels were formed as 100 mL drops in 10 cm cell culture dishes in groups of nine and maintained in 10 mL media. Hydrogels were released from dishes using a spatula and dishes were maintained on an orbital shaker to ensure gels did not settle and reattach to dish. Hydrogels were formed with an initial cell concentration of 100,000 cells/mL of hydrogel.

Rheology Rheological measurements were performed on an AR-G2 rheometer (TA Instruments, New Castle, DE). Hydrogels were formed in situ using 1 mL of hydrogel solution added between a 20 mm plate geometry and the rheometer base at 15 C, warmed to 37 C, and allowed to gel for 15 min. For experimental cell culture hydrogels that cannot be formed in situ on the rheometer and compacted during the 3 day time course, hydrogel dimensions were measured with the hydrogel constrained between parallel plates, using an 8 mm plate geometry, and plate diameter was set in the software (Rheology Advantage, TA Instruments) to that of the constrained hydrogel diameter. Consultation with the manufacturer (TA Instruments) validated this method. Three 1 ml hydrogels each from a minimum of two independent batches for cell independent, three 1 ml hydrogels each from two independent batches for cell laden, and three 100 ml hydrogels each from three independent batches for day 3 samples were characterized. Hydrogel shear modulus was measured at 1 Hz, 0.1% strain at 37 C. Young’s modulus (E) was calculated from the software recorded shear moduli (G0 ) using the formula [E ¼ 2 G0 (1 þ v)] with Poisson’s ratio (v) assumed at 0.1 as reported in the literature (43). Cell containing hydrogel moduli were obtained from gels with an initial seeding of ~200,000 cells/mL hydrogel.

Cells and cell maintenance Human fibroblast cells CCD-1065Sk (product ATCC CRL-2077, American Type Culture Collection) were cultured in T-175 flasks in MEM (Corning/ Mediatech, Manassas, VA) supplemented with 10% fetal bovine serum (Atlanta Biologicals, Norcross, GA), 1% L-glutamine (Corning), 1 mM Biophysical Journal 109(2) 249–264

Microscopy Fluorescent and phase contrast images of live or fixed cells were taken with an inverted microscope (DMI 4000B Inverted Microscope, LEICA

Fibroblast Activation in the TME Microsystems, Buffalo Grove, IL), with a 10 objective (N PLAN 10X/0.25 PH 1, N/-/B, LEICA Microsystems) or 40 objective (HCX FL FLUOTAR 40X/0.60 CORR PH 2, N/0-2/C, LEICA Microsystems) at room temperature utilizing an ORCA-ER digital camera (model C4742-95, Hamamatsu Photonics, Bridgewater, NJ) and Volocity imaging software (Improvision/ PerkinElmer, Wlatham, MA). Images of live cells were acquired through culture dishes with hydrogels in PBS whereas images of fixed cells were acquired with hydrogels on cover slips in no medium.

Viability assessment Hydrogels were prepared as above. For live/dead assessment, hydrogels were treated with propidium iodide at 10 mg/mL and Hoechst 33342 at 1 mg/mL for 5 min and imaged as described above. Cells were manually counted from images. Cells with propidium iodide staining were counted as dead and live cells were calculated from the total Hoechst stained population minus the propidium iodide population. For apoptosis and proliferation assessment, hydrogels were grown for 3 days and probed with antibodies to Annexin V and Ki-67 as described below.

Gel compaction assay Hydrogels were prepared as above and imaged daily with an 8-megapixel digital camera (XT894, Motorola, Schaumburg, IL). Gel images were measured using Fiji (ImageJ) software (44), normalizing to a known standard cell culture dish size. Hydrogels were imaged on day 0 without media and on day 3 in the presence of media. Media causes a slight distortion resulting in a 15% decrease in measured size, therefore data was normalized to reflect this.

Quantitative PCR Hydrogels were prepared as above and cultured for 3 days. To collect cells, 9 hydrogels (one experimental replicate) were treated with collagenase (product 02195109, MP Biomedicals) at 10 mg/mL until gels were digested (30 to 60 min) at 37 C on an orbital shaker. Cells were pelleted at 1000 xg for 5 min, collagenase/media supernatant was aspirated, and cell pellets were frozen at -80 C for later analysis. mRNA was purified from cell pellets according to manufacturer’s methods with QIAshredder and RNeasy kits (products 79654 and 74104, QIAGEN, Valencia, CA) with the addition of DNase (product AM2222, Invitrogen/Life Technologies, Grand Island, NY) to ensure RNA purity. qRT-PCR was carried out for 55 cycles (95 C/15s, 58 C/15s, 72 C/30s) using QuantiFast SYBR Green RT-PCR Kit (product 204154, QIAGEN) on a Roche LightCycler 480 (Roche Applied Science, Indianpolis, IN). qPCR primers are listed in Table S2 in the Supporting Material, Palladin primers were designed to detect isoforms 1, 3, and 4; base pair No. 2347–2435. Expression fold change values were calculated using the 2-DDCt method as described by Livak and Schmittgen (45). The density of fibroblasts per milliliter of hydrogel had no impact on gene expression, assessed in PEG hydrogels (Fig. S1 C). Reported values for Figs. 3 and 4 are normalized to RSM day 0 condition, all other figures are normalized to their respective 3D collagen-I cell embedded hydrogel (COL) day 3 values. For all drug treatment conditions (see Figs. 5, 6, and S3), three experimental replicates were combined before RNA purification because of low RNA yields and six technical replicates were completed.

Immunocytochemistry Hydrogels were prepared as above and cultured for 3 days. Gels were fixed with 4% paraformaldehyde solution for 45 min at 37 C, permeabilized with 0.25% Triton-X solution for 45 min at room temperature except in the case of Annexin V, washed with PBS þ 0.05% sodium azide (PBS-NaN3), and incubated for 2 h at room temperature or overnight at 4 C in blocking solution (5% goat serum, 1% BSA, 0.05% NaN3 in PBS). Samples were then

251 incubated for 24 h with primary antibodies as indicated (a-SMA @ 1:200, product MA1-37027, Thermo Scientific; FN-EDA @ 1:500, product F6140, Sigma-Aldrich; TGFb @ 1:500, product MA5-15065, Thermo Scientific; Palladin @ 1:200 (polyclonal antibody detects isoforms 1, 3, & 4; amino acids No. 725–740), product A3986, Sigma-Aldrich; a-tubulin @ 1:400, product ATN02, Cytoskeleton (Denver, CO); FAK @ 1:100, MAB-2156/ 05-1139, EMD Millipore, Ki-67 @ 1:100, product 550609, BD Biosciences (San jose, CA); Annexin V @ 1:500, product PA5-27872, Thermo Scientific) followed by three 60 min washes with PBS-NaN3. Samples were then incubated for 24 h with secondary antibodies (@ 1:300, Alexa Fluor, Life Technologies, Grand Island, NY), and/or rhodamine phalloidin for f-actin (product P1951, Sigma-Aldrich) and counterstained with 4’,6-diamidino2-phenylindole, followed by three 60 min washes with PBS-NaN3. Immunostained hydrogels were moved to eight-chamber cell culture slides (product 154534, Thermo Scientific) with 100 mL PBS-NaN3 for storage. Excess PBS-NaN3 was aspirated before imaging of hydrogel to reduce light scattering. All hydrogels for multiple trials were stained and imaged concurrently with single dilutions of antibodies and exposure times were held constant for each antibody.

Semi-quantitative immunocytochemistry calculations ImageJ was utilized for protein quantification (see Figs. 3 C and 4 D) (44). Cells were identified utilizing actin staining by threshold (triangle, auto) to define cell boundaries and measure cell area. Expression levels of marker proteins were measured from the integrated intensity of fluorescent signal from within, or outside of (FN-EDA), the identified fibroblast areas and normalized for the cell, or noncell, area. Intensity was further corrected for background fluorescence by subtracting the average intensity from hydrogels without cells. The intensity values were averaged and all values were normalized to COL RSM values. Hydrogel types displayed no difference in background fluorescence intensity (data not shown). CellProfiler (46,47) was utilized to quantify palladin localization. Nuclei (primary objects) were identified utilizing 4’,6-diamidino-2-phenylindole staining to define nuclear boundaries (adaptive threshold, robust background threshold method) and cells (secondary objects) were identified by f-actin staining (propagation identification, adaptive threshold, background threshold method). Cytoplasm was identified as cell area minus nuclear area (tertiary objects). The integrated intensity for each nucleus identified was set as a ratio to the cytoplasmic intensity for the cell (nucleus/cytoplasm).

Cellular morphology assessment Hydrogels were prepared as above and immunostained with rhodamine phalloidin for f-actin (product P1951, Sigma-Aldrich). The number of cellular projections was counted manually; cell length and width was measured manually using Volocity software (Improvision/PerkinElmer) to calculate cell aspect ratio.

Cytokine identification Cytokines present in fibroblast conditioned media were identified using a semi-quantitative, sandwich-based, human cytokine antibody array (product AAH-CYT-1000, RayBiotech, Norcross, GA) and quantified using Fiji (ImageJ) software and normalized to included controls (44). Only one experimental replicate was completed.

Signaling manipulations Function blocking antibodies to b1 integrin (product 555002, BD Biosciences) were used at 10 mg/ml and to b3 integrin (product 554951, BD Biophysical Journal 109(2) 249–264

252

McLane and Ligon

Biosciences) were used at 5 mg/ml. Inhibitors of FAK (product 3414, FAK Inhibitor 14, Tocris Bioscience, Bristol, UK), Phosphoinositide 3-Kinase (PI3K) (product 1130, LY 294002, Tocris) and CCR4 (product 227013, CAS 864289-85-0, EMD Millipore, Temecula, CA) were used at 4 uM, 5 uM, and 400 nM, respectively. Recombinant human chemokines CCL2/ MCP-1, CCL5/RANTES, CCL17/TARC, CCL22/MDC, and CCL3L1/ MIP-1a were purchased from R&D Systems (Minneapolis, MN) (products 279-MC-010, 278-RN-010, 364-DN-025, 336-MD-025, 509-MI-025) and were used at the manufacturer’s recommended concentrations as a mixed cocktail at 10 ng/ml each CCL. Human plasma fibronectin (product CB40008, Fisher Scientific, Logan, UT) was used at 50 mg/ml. TGFb (product PHG9204, Invitrogen) was used at 25 ng/mL.

Statistical analysis Statistical analysis was performed using Microsoft Excel with the Real Statistics Resource Pack (Release 2.14.1, http://www.real-statistics.com/).

RESULTS Biomimetic hydrogels to model the tumor microenvironment Although tumors are often 10- to 20-fold stiffer than normal tissue, the stroma surrounding solid tumors is only three- to fivefold stiffer than normal stroma (1,2). We sought to model this more subtle change in TME stiffness with 3D collagen-I hydrogels stiffened with a short bio-inert polyethylene glycol polymer cross-linker (stiffened versus unstiffened, PEG and COL respectively, Fig. 1 A). This PEG-diNHS cross-linking mimics collagen-I cross-linking in vivo, creating a more rigid ECM mesh without changing the concentration of collagen ligand (23). Our goal was to generate soft hydrogels (~200 Pa) and stiffened hydrogels (~800 Pa) that recapitulate the mechanical environments found in normal tissue and in the tumor microenvironment (1,2). To determine the bulk stiffness of our hydrogels, we performed rheology on hydrogels without cells and on hydrogels with cells at day 0 and day 3. The stiffness of hydrogels without embedded cells was 209.6 Pa 5 4.8 (COL) and 770.5 Pa 5 22.3 (PEG) (Fig. 1 B). The addition of cells led to a slight increase in stiffness of the soft hydrogels (289.2 Pa 5 12.7) but did not change the stiffness of the cross-linked hydrogels (769.0 Pa 5 35.7) (Fig. 1 C). In addition, the stiffness of the gels remained constant over the experimental time course (Fig. 1 D). These results indicate that we can create a controlled environment in which stiffness is precisely tunable and can be maintained during the experimental time course. Fibroblasts are activated to myofibroblasts by tumor-associated stiffness To determine if the increased stiffness found in the TME is sufficient to lead to fibroblast activation, we grew human fibroblasts in soft collagen-I hydrogels (COL) and in stiffened PEG-diNHS cross-linked collagen-I hydrogels (PEG) for Biophysical Journal 109(2) 249–264

FIGURE 1 Collagen-I hydrogel stiffness can be tuned with PEG-diNHS cross-linking. (A) Collagen-I gels form by physical cross-linking between fibers. Hydrogel stiffness can be increased by introducing more rigid cross-links between collagen-I and PEG-diNHS. (B) Stiffness of collagen-I (COL) and collagen-I PEG-diNHS hydrogels (PEG) without cells was measured by rheometry. Young’s modulus was 209.6 Pa 5 4.8 for soft COL gels (n ¼ 6) and 770.5 Pa 5 22.3 for stiff PEG gels (n ¼ 12), corresponding to values reported for normal and tumor adjacent stroma (1,2). Diamonds in the box and whisker plots represent the mean, and the midline is the 50th percentile. The top and bottom bar represent the 90th and 10th percentile, respectively, and the top and bottom boxes represent the 75th and 25th percentile, respectively. (C) Stiffness of hydrogels with encapsulated cells. The elastic modulus of COL gels with cells was 289.2 Pa 5 12.7 (n ¼ 6) and of PEG gels with cells was 769.0 Pa 5 35.7 (n ¼ 6). (D) Stiffness of hydrogels over time. Hydrogels with encapsulated cells were cultured for 3 days. Elastic modulus of COL gels at day 3 was 262.0 Pa 5 20.0 (n ¼ 9) and of PEG gels at day 3 was 716.8 Pa 5 44.6 (n ¼ 9). All Young’s modulus values were obtained by converting measured bulk modulus with the formula E ¼ 2G(1 þ v), with v as 0.1 (43). Asterisk indicates significant difference. Student’s t-test p values < 0.05. Error bars show standard error of the mean. To see this figure in color, go online.

3 days. Hydrogel cross-linking had no effect on cell viability, proliferation, or apoptosis (Fig. S1, A and B). Second-harmonic generation microscopy showed that the cells can organize the collagen matrix in both COL and PEG hydrogels (Fig. 2 A). At day 0, COL and PEG matrices are similar with short, randomly oriented collagen-I fibers, but by day 3 the matrix collagen has been rearranged into long, directional bundles in both conditions. We quantified

Fibroblast Activation in the TME

FIGURE 2 Hydrogel structure and compaction. (A) Representative second harmonic generation images of collagen-I organization in soft (COL) or stiff (PEG) hydrogels. Scale bar: 50 mm. (B) Quantification of skewness from second-harmonic generation images, normalized with collagen day 0 and 50% percentile at 0. Forty-two images from each gel were quantified, with six gels per condition. Skewness decreases as fibril organization increases (48). Box and whisker plots as in Fig. 1 B. Asterisk indicates significant difference from day 0. Student’s t-test p values < 0.0001. (C) Hydrogel compaction over 3 days, CM (COL n ¼ 41; PEG n ¼ 39). Asterisk indicates significant difference from day 0. Student’s t-test p values < 0.005. Error bars show standard error of the mean.

253

this shift from disorganized to organized using image skewness. Skewness is a measure of the imbalance of the distribution of a variable about its mean, in this case pixels that are darker or brighter than the image mean. Images with greater organization, such as those with large fibrils, exhibit a left-tailed shift in the distribution that is calculated as a negative shift in skewness (48). Both hydrogel types show a significant negative shift in skewness from day 0 to day 3, suggesting that the matrices are being organized by the embedded cells (Fig. 2 B). Compaction of a hydrogel is an additional measure of the reorganization and contraction of the matrix by embedded cells. We observed significant decreases in hydrogel area between day 0 and day 3 in both COL and PEG gels, but these were more pronounced in PEG hydrogels (Fig. 2 C), indicating that cells can compact both gels, but do so more effectively in the stiffened matrix. Together, these results suggest that fibroblasts embedded in both soft and stiffened hydrogels can reorganize and compact the matrix, but that they do so more efficiently in the stiffer matrix representative of the TME. Next we assessed fibroblast activation in soft and stiffened matrices by looking at morphology and patterns of gene expression. Fibroblasts cultured in both soft and stiff 3D matrices adopt an elongated spindle-like morphology with prominent stress fibers (Fig. 3 A). Cells grown in the stiffened gel had a significantly higher aspect ratio and fewer projections than those grown in the soft gel (Fig. 3 B). An exaggerated spindle-like morphology is indicative of myofibroblast-like activation, and these data indicate that cells in stiffer gels show a more activated phenotype than those in the soft gels. We then measured the relative expression of markers of fibroblast activation by semi-quantitative immunocytochemistry, as it was difficult to obtain sufficient material for Western blots from these small gels. Cells grown in stiff gels had higher protein levels of a-smooth muscle actin (a-SMA), palladin, and FN-EDA than cells grown in soft gels (Fig. 3 C). Extracellular FN-EDA was also increased, indicating an increase in fibronectin deposition as would be expected by activated fibroblasts. TGFb expression, on the other hand, was unchanged, suggesting that the TGFb autocrine loop seen with long-term tumor:stroma interaction was not activated in 3 days by growth in stiffened gels (49). We also assessed mRNA expression of aSMA, palladin, FN-EDA, fibroblast activation protein (FAPa), and TGFb by qRT-PCR. We extracted RNA from cells grown in both COL and PEG hydrogels, as well as from cells grown in 2D on collagen-I coated plates to determine the effect of growth in 3D versus 2D. The expression of all markers was increased when cells were grown in 3D hydrogels in comparison with growth in 2D (Fig. S2 A), and the levels of all markers except TGFb were further increased when cells were grown in stiff gels versus soft gels, consistent with the observed protein expression (Fig. 3, C and D). The expression pattern seen in the stiff PEG hydrogels is Biophysical Journal 109(2) 249–264

254

McLane and Ligon

FIGURE 3 ECM stiffness representative of the tumor microenvironment can lead to fibroblast activation. (A) Representative images of fibroblasts in soft (collagen-I) or stiff (collagen-I PEG-diNHS) hydrogels (stained with phalloidin to label f-actin). Scale bar: 20 mm. (B) Quantification of the number of projections from the cell body and cell aspect ratio (COL n ¼ 12; PEG n ¼ 11). (C) Semi-quantitative immunocytochemistry of aSMA, palladin, intracellular FN-EDA, TGFb, and extracellular FN-EDA. All values were normalized to cells in RSM COL (n ¼ 3). Asterisk indicates significant difference between soft and stiff gels. (D) qRT-PCR of markers of fibroblast activation. All levels were normalized to expression levels in cells in collagen-I hydrogels in complete media (n ¼ 3). Asterisk indicates significant difference between cells in soft COL gels and in stiff PEG gels, Student’s t-test values < 0.05. Error bars show standard error of the mean. To see this figure in color, go online.

also similar to that seen when fibroblasts are directly activated by TGFb treatment (Fig. S2 B). Together, these results suggest that growth in a stiffened collagen-I environment (800 Pa) with mechanical properties similar to the TME is sufficient to convert fibroblasts to a myofibroblast-like state. Further, they indicate that we can use this pattern of gene expression (increases in aSMA, palladin, FAPa, and FNEDA (Fig. 3 D)) as a measure of this stiffness-induced fibroblast activation. Serum factors can alter activation response Chemical signals as well as mechanical signals are known to play a role in the activation of fibroblasts. To determine if an increase in stiffness is sufficient to activate fibroblasts in the absence of chemical cues, we investigated the response of fibroblasts to stiffness under serum depletion conditions. Fibroblasts were cultured in RSM (1% serum) and then embedded in either soft or stiff gels with RSM. Growth in RSM did not affect cell viability (Fig. S1). Fibroblasts grown in RSM had a more stellate morphology than those grown in complete media (CM). Cells grown in both soft and stiff gels had multiple cellular protrusions and few stress Biophysical Journal 109(2) 249–264

fibers (Fig. 4 A). The number of projections was higher than in cells grown in CM, but there was no difference between cells grown in soft or stiff gels in RSM (Fig. 4 B). The aspect ratio was also decreased in both soft and stiff gels to the baseline level seen in soft gels in CM (Figs. 3 B and 4 B). In addition, cells grown in RSM were significantly less able to compact the hydrogel than were those grown in CM, although cells in stiff gels in RSM did show a slight increase in compaction over those in soft gels (Fig. 4 C). These data suggest that although cells grown in RSM in stiff gels are slightly more activated than those grown in soft gels in RSM, and, in both cases, they are much less activated than those grown in CM. The expression of proteins associated with fibroblast activation was also reduced in cells grown in RSM in comparison with CM (Fig. 4 D). aSMA and palladin showed a slight stiffness dependent increase in expression, although it was much smaller than that seen in CM, and FN-EDA expression was unaffected by the increase in stiffness. TGFb expression was not affected by stiffness or serum reduction. These changes in protein expression were mirrored in mRNA levels. All transcripts were greatly decreased in comparison with CM except TGFb, which


255

FIGURE 4 Depletion of soluble growth factors blocks fibroblast activation. (A) Representative images of fibroblasts grown in soft (collagen-I) or stiff (collagen-I PEG-diNHS) hydrogels with reduced serum media (RSM) (stained with phalloidin to label f-actin). Scale bar: 20 mm. (B) Quantification of the number of projections from the cell body and cell aspect ratio (COL n ¼ 8; PEG n ¼ 7). (C) Hydrogel compaction in reduced serum media (COL n ¼ 69; PEG n ¼ 61). Asterisk indicates significant difference. Student’s t-test p values < 0.005. (D) Semi-quantitative immunocytochemistry for aSMA, palladin, FN-EDA, TGFb, and extracellular FN-EDA comparing cells grown in CM with those grown in RSM in soft (COL) or stiff (PEG) gels (n ¼ 3). (E) qRT-PCR for markers of fibroblast activation comparing cells grown in CM with those grown in RSM (n ¼ 3). Asterisk indicates significant difference between COL and PEG conditions. Hash represents significant difference between RSM and CM conditions. Student’s t-test values < 0.05. Error bars show standard error of the mean. To see this figure in color, go online.

remained constant (Fig. 4 E). aSMA, palladin, and FAPa all showed small stiffness dependent increases in transcript level, as did FN-EDA, even though it did not show an increase in protein level (Fig. 4 E). Together these data suggest that mechanical stiffness alone is not sufficient to result in strong fibroblast activation, but that soluble factors found in serum are also required. Both CCL growth factors and serum fibronectin are necessary for focal adhesion clustering To explore the link between soluble factors found in serum and mechanical signaling, we first tested whether focal adhesion clustering was affected by serum concentration. Growth factor receptors have been shown to directly interact with integrins and both integrin and growth factor receptor signaling are modified by these interactions (50). Furthermore, integrins and growth factor receptors have been shown to cocluster at focal adhesions and recruit FAK to the adhesion sites (50). To determine if growth factors present in serum may account for the differences we observed in fibroblast activation between RSM and CM, we looked at FAK localization and clustering. When cells were grown in CM, clusters of FAK labeling were seen, but when cells were grown in RSM, FAK labeling was dispersed (Fig. 5 A). To quantify this

loss of clustering, we calculated the mean standard deviation of FAK intensity for each condition and saw a decrease in variation from CM to RSM in both COL and PEG, suggesting that FAK clustering occurs in the presence of serum, but not in low serum conditions (Fig. 5 G). There are many growth factors present in serum. We screened our culture media for 120 chemokines and found that prominent among them were the CCLs (Table S1). We then screened the fibroblasts for all 10 known CCL receptors and found that only CCR4 is expressed in these cells (Table S2). Pharmacological inhibition of CCR4 in cells grown in CM resulted in the same dispersed FAK signal (Fig. 5 B) and lack of clustering (Fig. 5 G) as seen in RSM conditions. In addition, mRNA expression of genes associated with fibroblast activation was also significantly decreased with CCR4 inhibition (Fig. 5 D). However, introduction of a cocktail of recombinant CCR4 ligands into RSM (CCLs 2, 4, 5, 17, and 22) (51,52), did not rescue FAK clustering (Fig. 5, B and G) or changes in mRNA expression (Fig. 5 E), suggesting that CCL signaling alone is not sufficient to induce receptor clustering and allow the cell to respond to mechanical inputs. Fibronectin has also previously been shown to play a role in fibroblast activation (53,54), so we tested whether addition of fibronectin alone to RSM could rescue FAK Biophysical Journal 109(2) 249–264

256

McLane and Ligon

FIGURE 5 Both serum CCL growth factors and fibronectin are necessary for FAK clustering and fibroblast activation. (A) Cells grown in complete media (10% serum) show punctate clusters of FAK, whereas FAK labeling is diffuse in cells grown in reduced serum media (1% serum). (B) Inhibition of CCR4 receptor with a CCR4 antagonist drug in CM blocks FAK clustering. Reintroduction of a cocktail of recombinant CCL growth factors does not rescue FAK clustering in RSM conditions. (C) Fibronectin addition to RSM does not rescue FAK clustering, but addition of fibronectin and CCLs together in RSM condition does recover punctate FAK staining. Scale bar: 20 mm. (D) qRT-PCR of markers of fibroblast activation in cells treated with CCR4 receptor antagonist in CM (n ¼ 3). Expression level of each marker is seen in RSM conditions shown with diamond markers for comparison. (E) qRT-PCR of markers of fibroblast activation in cells in RSM hydrogels supplemented with recombinant CCL growth factors (n ¼ 3). (F) qRT-PCR of markers of fibroblast activation in cells in RSM hydrogels supplemented with fibronectin (n ¼ 3). Asterisk indicates significant difference from control. Student’s t-test values < 0.05. All qPCR data normalized to COL condition. (G) Mean standard deviation of FAK staining intensity in COL and PEG hydrogels (n ¼ 5). Images with bright and dark spots (clusters and unlabeled background) have higher standard deviation than those with more uniform labeling. Asterisk indicates significant difference from CM. ANOVA p-values < 0.05. Error bars are standard error of the mean. (H) qRT-PCR of markers of fibroblast activation in cells grown in RSM supplemented with both human serum fibronectin and CCL cocktail (n ¼ 3). Asterisk indicates significant difference from RSM control. Student’s t-test values < 0.05. Error bars show standard error of the mean. To see this figure in color, go online.

clustering and fibroblast activation, but we did not see a rescue of FAK clustering (Fig. 5 C), or fibroblast activation (Fig. 5 F) with the addition of fibronectin alone. We then added a combination of fibronectin and the CCL cocktail, and we observed a significant rescue of FAK clustering (Fig. 5, C and G). Analysis of the pattern of gene expression, however, suggests that the combination of fibronectin and CCLs at the concentrations supplied only partially rescues fibroblast activation, as FN-EDA and palladin showed little or no response to this combination (Fig. 5 H). In addition, hydrogel compaction was also not completely rescued by the addition of these serum factors (data not shown). Taken together, these data indicate that both CCL growth factors and their cognate CCR4 receptors, as well as serum fibroBiophysical Journal 109(2) 249–264

nectin, are key for FAK clustering, which is necessary for cells to respond to mechanical cues and for myofibroblast activation. However, other serum factors may play a role in more complete activation or a different concentration of specific CCLs or fibronectin may be important, as some markers of activation showed little or no response and hydrogel compaction was not completely rescued, despite the recovery of FAK clustering. Fibroblast activation is regulated by b1 and b3 integrins, FAK, and PI3K Next we probed the outside-in signaling pathways to determine how the mechanical signaling from increased stiffness


leads to fibroblast activation. Cells typically adhere to the ECM via integrins, so we first used blocking antibodies that interfere with the interaction between integrins and their ligands. We treated cells with antibodies to either integrin b1 to block collagen-I binding integrins (which all include the b1 subunit), or integrin b3 that binds fibronectin and other RGD containing ECM proteins (55,56). The effects of these blocking antibodies on fibroblast activation were complex. Blocking b1 integrin in stiff gels led to increases in the expression of FAPa, and FN-EDA and a decrease in palladin expression (Fig. 6 A), and blocking b3 integrin led to small decreases in aSMA and FAPa expression and an increase in palladin expression (Fig. 6 B). Both antibodies had overall smaller effects when added to cells in soft gels except that blocking b3 integrin led to an enhanced increase in palladin expression (Fig S3), suggesting that the effects observed in stiff gels were attributable primarily to interfering with mechanical sensing. Blocking b1 integrin led to a modest increase in TGFb expression in both soft and stiff gels and blocking b3 integrin had no effect on TGFb expression in either gel, suggesting that TGFb is not sensitive to mechanical inputs. To interpret these data, we calculated the effect of these treatments on stiffnessinduced activation by comparing the expression pattern in stiff gels with and without inhibitors (Table 1). Blocking b1 integrin had a paradoxical effect—it did not block stiffness-induced activation as expected, but rather partially enhanced the activation (Fig. 6 A and Table 1). Blocking b3 integrin, on the other hand, did partially block stiffness-induced activation (Fig. 6 B and Table 1). Interestingly,

257

the effect on palladin expression was opposite that of other indicators of fibroblast activation—it was decreased by b1 inhibition and increased by b3 inhibition. Together, these data suggest that integrins have complex roles in activation and perhaps have regulatory functions. We then tested whether inhibition of FAK could block stiffness-induced fibroblast activation. FAK is found at adhesions made by both b1 and b3 integrins and we also saw that FAK clustering is correlated with fibroblast activation. Pharmacological inhibition of FAK in stiff gels significantly blocked stiffness-induced activation, resulting in lower expression of all markers of activation except FNEDA (Fig. 6 C and Table 1). The effect of FAK inhibition in soft gels was minimal and did not alter aSMA or palladin expression (Fig S3), suggesting that FAK is central in the modulation of stiffness sensing in stiffness-induced fibroblast activation. Phosphatidylinositide 3-kinase (PI3K) is downstream of FAK and known to play a role in transducing mechanical signals (57). Therefore we tested whether inhibition of PI3K could block stiffness-induced fibroblast activation. Pharmacological inhibition of PI3K in stiff gels partially blocked stiffness-induced activation, resulting in significantly reduced expression of aSMA and FAPa, but no change in the expression of FN-EDA or palladin (Fig. 6 D and Table 1). The effect of PI3K inhibition in soft gels was minimal (Fig S3), suggesting that the effects seen in stiff gels were because of interfering with mechanosignal transduction. We then wanted to test whether all FAK signaling in stiffness-induced fibroblast activation was being transmitted FIGURE 6 b1 and b3 integrins, FAK, and PI3K all play roles in control of fibroblast activation in stiff hydrogels. (A) b1 integrin activity was inhibited with a blocking antibody and the expression of markers of fibroblast activation was measured by qRT-PCR (n ¼ 3). (B) b3 integrin activity was inhibited with a blocking antibody and the expression of markers of fibroblast activation was measured by qRT-PCR (n ¼ 3). (C) Focal adhesion kinase (FAK) activity was blocked with FAK Inhibitor 14 drug and the expression of markers of fibroblast activation was measured by qRT-PCR (n ¼ 3). (D) Phosphoinositide 3-kinase (PI3K) activity was inhibited with LY 294002 and the expression of markers of fibroblast activation was measured by qRT-PCR (n ¼ 3). (E) Cells were treated with a combination of FAK and PI3K inhibitors and the expression of markers of fibroblast activation was measured by qRT-PCR (n ¼ 3). (F) Cells were treated with a combination of b1 integrin blocking antibody and PI3K inhibitor and the expression of markers of fibroblast activation was measured by qRT-PCR (n ¼ 3). Asterisk indicates significantly different increase. Hash indicates significantly different decrease. All data is normalized to COL control. Student’s t-test values < 0.05. All error bars signify standard error of the mean.

Biophysical Journal 109(2) 249–264

258

McLane and Ligon

TABLE 1 Quantification of changes in transcription of markers of fibroblast activation in stiff hydrogels with inhibitors Stiff b1 integrin antibody b3 integrin antibody FAK inhibitor PI3K inhibitor FAK and PI3K inhibitors b1 antibody and PI3K inhibitor

aSMA FAPa FN-EDA Palladin TGFb1 (%) (%) (%) (%) (%) 23 75 103 48 61 62

396 64 128 78 81 72

186 3 4 -38 20 193

79 149 75 5 417 17

218 2 40 5 32 192

probed for the localization of palladin while manipulating integrin-matrix interactions. Blocking b1 integrins in stiff hydrogels led to an accumulation of palladin in the nucleus (30% increase over control) whereas blocking b3 integrins led to an exclusion of palladin from the nucleus (18% decrease over control) (Fig. 7), with a similar effect in soft hydrogels (Fig. S4). These data suggest that palladin localization is controlled by an integrin-mediated pathway, which may then control the balance between the function of palladin as a transcriptional regulator and actin-associated protein.

Values are reported as percent change from level of untreated stiff hydrogel. Statistically significant changes are in bold. n ¼ 3.

DISCUSSION through PI3K, so we added pharmacological inhibitors of both FAK and PI3K to cells in stiff gels. If all FAK signaling goes through PI3K, then the addition of both inhibitors should mimic that of either alone. For aSMA, FAPa, and FN-EDA the effect was the same with both inhibitors together or either alone (Fig. 6 E). But the effect on palladin was strikingly different. FAK inhibition caused a decrease in palladin expression, whereas PI3K inhibition had no effect on palladin. But when both FAK and PI3K were inhibited, palladin expression was significantly higher (Fig. 6, C–E). Similarly, we sought to determine if all effects of blocking b1 integrin signaling are mediated by PI3K, so we treated cells with a combination of PI3K inhibition and b1 blocking antibodies. We observed decreases in aSMA and FAPa expression, similar to that seen with PI3K inhibition alone, and increases in FN-EDA and TGFb expression similar to that seen with blocking b1 integrin alone (Fig. 6 F). Taken together this data suggests that FAK signaling is transmitted through PI3K for aSMA and FAPa transcription, that FAK signaling for aSMA and FAPa is dependent on interactions between b1 and b3 integrins, that FN-EDA and TGFb are regulated by b1 integrin–related signaling, and that FAK and PI3K have a complex regulatory relationship involving both integrin b1 and b3 in palladin transcription.

It has been hypothesized that the elevated stiffness of the stroma surrounding tumors contributes to the transformation of stromal cells into the tumor associated phenotypes seen

Palladin localization is controlled by matrix adhesions We observed a curious pattern of palladin gene expression when we blocked b1 or b3 integrins that was opposite that of other myofibroblast associated genes (Fig. 6, A and B). Palladin has been shown to play a role in the regulation of myofibroblast associated gene expression (12,38,39), in addition to its more established actin-associated roles (33–37). Palladin has also been shown to localize to the nucleus in some cell types (40,41), which is consistent with a role in transcriptional regulation. We hypothesized that the balance of palladin localization between cytoplasmic/ actin-associated and nuclear may regulate palladin function in these cells, and further that this balance may be regulated by the mechanotransduction pathways. Therefore, we Biophysical Journal 109(2) 249–264

FIGURE 7 Palladin localization is controlled by integrins. (A) Palladin localized to the nucleus when b1 integrin was blocked and localized to the cytoplasm when b3 integrin was blocked (n ¼ 3). (B) Quantification of the change in localization relative to the amount present in controls. Asterisk indicates significant difference from control. Student’s t-test values < 0.005. All error bars signify standard error of the mean. (C) Representative images of palladin nuclear localization.


in vivo (58–60). In this study, we demonstrate that we can make a collagen-I based, stroma-mimetic, 3D hydrogel system in which we can precisely tune mechanical stiffness within the physiologically relevant range. We then used this system to show that fibroblasts are activated to a myofibroblast-like state with a physiologically relevant increase in stiffness alone. We then investigated the interconnected roles of stiffness, soluble factors such as fibronectin and chemokines, and signaling through integrin receptors upon this activation. We found that fibroblasts acquire a myofibroblast-like phenotype when cultured in collagen-I 3D hydrogels with a stiffness that mimics the tumor-associated stroma. Although others have shown that increased stiffness plays a role in fibroblast activation, these experiments were performed with models in which the stiffness was unknown or increased 10-fold or more greater than normal stroma (4,15,61). In this study we have shown that a fourfold increase in stiffness, which is more physiologically appropriate, is sufficient to lead to fibroblast activation. The activation induced by increased stiffness was similar to that seen with TGFb treatment in the soft COL hydrogels (Fig. S2 B). The transition from 2D culture to 3D also had a significant effect. Cells grown in 2D culture on collagen-I–coated surfaces had a more senescent phenotype (Fig. S2 A); this could be attributable either to the change in geometry or the fact that 2D substrate (tissue culture plastic) is orders of magnitude stiffer than the 3D hydrogels. We also found that serum factors are necessary for the stiffness-induced activation we observed, and that CCLs and fibronectin, in particular, are important in this process (Fig. 4). Other soluble factors may play a role as well, as the CCLs and fibronectin do not completely rescue the effect. Other stromal ECM components such as FN-EDA that are deposited by the fibroblasts themselves may also play a role in promoting fibroblast activation, as fibroblasts do remodel the matrix over the experimental time course. We found that FAK clustering is associated with myofibroblast activation and is controlled in part by the availability of fibronectin. Fibronectin can be classified into two subgroups, plasma and matrix, both of which have multiple isoforms (62). The stiffness-induced activation we observed was blocked in reduced serum conditions and partially rescued with the addition of plasma fibronectin (Fig. 5). Others have suggested that the matrix FN-EDA can lead to fibroblast activation, but plasma fibronectin has also been shown to alter fibroblast activity in a wound healing model (53,63). Perhaps some fibronectin domains such as FNIII10, which is conserved between plasma and matrix FN and interacts with integrin avb3 (62), are necessary for proper integrin engagement and clustering to allow for activation, whereas other domains such as EDA actually stimulate activation. We also found that the CCR4 and its ligands, the CCLs, are necessary for FAK clustering and fibroblast activation as well. CCR4 is expressed by fibroblasts, and is upregu-

259

lated in wounded tissue and correlated with wound healing, suggesting that it may be associated with fibroblast activation (64). It has been suggested that the fibroblasts in the tumor stroma behave like a wound that never heals (65,66), and CCR4 signaling may play a part in this process. Integrins and other growth factor receptors have also been shown to cocluster, which is necessary for their downstream signaling (50,67), and perhaps CCR4 and integrins cocluster and coordinate downstream signaling through focal adhesions as well. Fibroblast activation is not a simple off/on switch. There is a continuum between the two states, and many other related states as well. We examined a range of indicators of fibroblast activation, including morphology and the expression of several genes at both the protein and mRNA level. We found that there was an overall pattern of activation to a myofibroblast-like state when cells were grown in the stiffened gels, but there were subtle differences in gene expression between the various markers that may elucidate their specific regulation. a-smooth muscle actin (aSMA) is one of the most commonly employed markers of the myofibroblast phenotype, and it is known to be responsible for the generation of contractile force (68). Expression of aSMA was completely dependent on mechanical stiffness, as no drug or antibody manipulation showed an effect in soft COL (Fig. S3). aSMA expression is inhibited when FAK-PI3K signaling is blocked, suggesting that it’s activated by FAK-PI3K signaling, consistent with previous reports (69,70). We also found that blocking b3 integrin led to a decrease in aSMA expression, suggesting that b3 integrins participate in the regulation of aSMA. FAPa is a transmembrane serine protease that can degrade collagen-I and has been cited as a marker of activated fibroblasts (71). We found that similar to aSMA, FAPa is induced in a stiffness dependent manner, stimulated through FAK-PI3K signaling, in agreement with previous work (72), and regulated by signaling through b3 integrins. Inhibition of FAPa has been shown to lead to a decrease in aSMA expression as well, so it is possible that the aSMA decrease we observed is also attributable in part to the inhibition of FAPa (Fig. 6) (71). Somewhat surprisingly, blocking b1 integrins led to an increase in FAPa expression, in agreement with a previous report also showed that blocking b1 integrin-induced FAPa expression (73). These data suggest that a balance of b1 and b3 integrin interactions control fibroblast activation. Although there were slight differences in the details, overall, the expression patterns of aSMA, FAPa, and FN-EDA were similar. In contrast, the expression of palladin was strikingly different. Our data show that palladin expression is induced by stiffness like the other markers, but the regulation is very different. Blocking b1 integrins leads to a decrease in palladin expression and blocking b3 integrins leads to an increase. Further, inhibiting FAK leads to a decrease in palladin expression, but blocking PI3K has no Biophysical Journal 109(2) 249–264

260

effect, and blocking both FAK and PI3K leads to a significant increase in expression. Together, these results suggest a complex control mechanism involving b1 integrins, b3 integrins, FAK, and PI3K signaling. To explain the complex regulation of palladin and other markers of activation suggested by our data, we propose a model (Fig. 8) in which palladin expression and the expression of aSMA and FAPa can be explained by three points: 1) palladin expression is stimulated by stiffness through the FAK-PI3K-Akt pathway; 2) palladin interactions with focal adhesions and the actin cytoskeleton alter its ability to regulate the expression of other activation associated proteins; and 3) the distribution of phosphoinositide 3-kinase (PI3K) activity between FAK and other receptor tyrosine kinases (RTK) alters palladin expression. We also propose that FN-EDA and TGFb expression can be explained via a mechanism requiring both FAK and palladin transcriptional signaling. Palladin expression has been shown to be stimulated through the PI3K-Akt pathway in 2D (74), as well as by FAK (75). aSMA and FAPa have also been shown to be dependent on FAK-PI3K signaling (69,70,72). Palladin has also been demonstrated to interact with transcription factors, and is known to be required for the expression of aSMA (12,29,38,39), whose expression is also stimulated by FAPa (68,71). We suggest that the active FAK-PI3K signaling necessary for aSMA and FAPa expression is

FIGURE 8 Model of palladin, aSMA and FAPa regulation. Palladin is regulated through integrin interactions in focal adhesions as well as growth factor and nonreceptor tyrosine kinase signaling. Palladin has both f-actin and non-f-actin interactions and plays a role in the regulation of aSMA and FAPa gene expression. To see this figure in color, go online. Biophysical Journal 109(2) 249–264

McLane and Ligon

also required for palladin expression. We also suggest that palladin interactions with transcription factors, along with FAK-PI3K signaling, are necessary for the expression of FAPa and aSMA. Further, increasing aSMA expression is associated with a decrease in palladin expression (38), suggesting that aSMA protein inhibits palladin transcription, creating an inhibitory feedback loop. Palladin has two known roles in the cell—one in regulating the actin cytoskeleton and adhesions and the other in regulating gene expression. It also has two main localization patterns—cytoplasmic (i.e., associated with the actin cytoskeleton and/or at focal adhesions) (33,37,76,77) and nuclear (40,41,77). We hypothesize that the localization of palladin plays a role in the balance between these two different activities. We suggest that in fibroblasts grown in 3D hydrogels, palladin is concentrated in the cytoplasmic compartment at focal adhesions, which in our system contain both b1 and b3 integrins. This sequesters palladin and decreases the nuclear fraction, thus restricting its role in regulating gene expression. b3 integrins have been shown to play a role in the turnover of focal adhesions (78,79) and when we block b3 integrins, palladin is redistributed from the nucleus to the cytoplasmic compartment in part attributable to sequestration at stabilized focal adhesions (Fig. 7). On the other hand, blocking b1 integrins which should lead to a reduction in focal adhesions, results in the nuclear accumulation of palladin (Fig. 7). We suggest that by interfering with b integrins, we alter the formation and/or turnover of focal adhesions, which leads to changes in the distribution of palladin, which then leads to changes in palladin’s transcriptional regulation (74,80–82). Although the exact mechanism of control of localization of palladin is still unknown, phosphorylation of Ser507 by Akt1 is critical for palladin to associate with actin (83) and Ser507 is in a C-terminal domain of palladin that localizes to the nucleus when expressed independently of the N-terminus of the protein (42). This suggests that Akt1 phosphorylation may control the localization of palladin, whereas the expression of palladin is controlled by Akt2 signaling, suggesting distinct roles of Akt signaling in the control of palladin transcription and regulation. Phosphoinositide 3-kinase is a signal transduction enzyme that is known to interact both with integrin signaling via FAK and with other RTK such as growth factor receptors, and it can also signal through protein kinase B (Akt) (84–86). We hypothesize that if we inhibit the PI3K binding partner FAK, or block FAK activation by blocking b1 integrins, we shift the interaction of PI3K to its other available partners, RTKs. We suggest that this shift in PI3K association leads to a decrease in palladin transcription by both reducing FAK-PI3K-Atk2 activity, which has been shown to stimulate palladin expression, and by increasing an unknown RTK pathway, perhaps by differential phosphorylation of Akt2, which has at least five different phosphorylation sites (74,87). Concurrently, we


261

suggest that a stimulation of palladin expression by an unknown source of Akt2 activation—such as known PI3K independent Akt2 activators TNK2, Src, and PTK6—arises from the shift toward growth factor related signaling, as TNK2, Src, and PTK6 activation are linked to growth factor signaling (88). This suggests that a shift from FAK-PI3K toward RTK-PI3K activity inhibits palladin expression by both inhibiting FAK-PI3K stimulatory activity and stimulating RTK-PI3K inhibitory activity and stimulates palladin expression by stimulating an unknown Akt2 activator. Our complete model for palladin, aSMA and FAPa expression is depicted in Fig. 8 (and Fig. S5 B) with the effect of manipulations depicted in Fig. S5 A and predicted magnitude of effect of manipulations upon palladin in Table 2. Focal adhesions stimulate palladin expression through FAK-PI3K. Palladin stimulates FAPa expression, which in turn stimulates aSMA expression. aSMA in turn inhibits palladin transcription. The association of palladin with focal adhesions and the actin cytoskeleton regulates its ability to stimulate FAPa and aSMA. All the while, RTK signaling inhibits palladin expression while it has the ability to also stimulate palladin expression through an unknown PI3K independent mechanism. The regulation of TGFb and FN-EDA expression appears to be more straightforward. TGFb is a paradoxical growth factor; in normal tissues, it acts as a tumor suppressor, but in cancer, it plays a role in promoting metastasis (3). Interestingly, TGFb expression does not change in our activated fibroblasts. This is consistent, however, with reports suggesting that TGFb induced myofibroblasts take more than 42 days to double their TGFb expression (49). Our results do indicate that TGFb expression is regulated by FAK, which is consistent with existing TGFb literature (89). Our results also suggest that palladin plays a major role in TGFb expression through both expression level and cytoplasmic/nuclear localization. The fibronectin splice variant EDA is a variant of fibronectin that contains an extra alternatively encoded type III domain repeat. It is found only in insoluble secreted ECM and its expression is usually limited to embryogenesis (63). FN-EDA is induced by stiffness, but our data does not directly indicate a role for FAK-PI3K signaling in its TABLE 2

expression, despite previous reports suggesting that signaling via PI3K-Akt-mTOR can stimulate FN-EDA expression (90). There does, however, appear to be a relationship between FN-EDA and TGFb expression, similar to a recent report in which increased TGFb expression leads to increased FN-EDA, independent of PI3K signaling (91). This may be because of FN-EDA stimulation by the MAPK pathway (63) that is known to work synergistically with the TGFb pathway (92). Our model for TGFb and FN-EDA expression is depicted in Fig. S5 B with the effect of manipulations depicted in Fig. S5 C and predicted magnitude of effect of manipulations upon palladin in Table S3. Simply, TGFb expression is stimulated by FAK and palladin, and TGFb in turn stimulates FN-EDA expression (91). For instance, when we block b1 integrin, we significantly reduce the downstream FAK-PI3K signaling, which stimulates palladin transcription. This also leads to a dramatic shift in the interactions of PI3K from FAK toward RTKs by eliminating the ability of FAK to associate with b1 integrins (by reducing the formation of focal adhesions). This causes a drastic increase in RTK-associated palladin inhibition. This shift in PI3K signaling, however, also increases PI3K independent palladin stimulation. Also, palladin localization shifts to the nucleus, which allows for regulation of transcription, stimulating FAPa and aSMA transcription. This stimulation maintains aSMA protein levels, which in turn maintains an inhibitory feedback on palladin transcription. Together, this results in a net reduction in palladin transcription, overexpression of FAPa, and maintenance of aSMA transcription (Figs. 6 and 8, and Table 2). In the case of TGFb and FN-EDA, blocking b1 integrin reduces the FAK signaling necessary for TGFb transcription by both reducing focal adhesions and further by inhibiting FAK. There is also less palladin being produced to stimulate production of TGFb, but because of the nuclear localization of palladin, there is an intense stimulation of TGFb production by palladin transcriptional regulation. Together, this results in a dramatic increase in TGFb transcription, which also leads to increased FN-EDA expression (Figs. 6 and S5, and Table S3). These results offer, to our knowledge, new insights into the regulation of fibroblast activation by the ECM. We

Model predictions for the change in palladin gene expression due to inhibitory manipulations

Palladin b1 integrin antibody b3 integrin antibody FAK inhibitor PI3K inhibitor FAK and PI3K inhibitors b1 antibody and PI3K inhibitor

FAK-PI3K Loss

FAK: RTK

RTK-PI3K Inhibition

Unknown Stimulation

Actin Association

aSMA

NET

YYYY 4 YYYY YYYY YYYY YYYY

YYYY 4 Y 4 Y YYYY

4 4 4 [[ [[ [[

[[ 4 [[ 4 [[ [[

[[ 4 4 4 4 [[

4 [[ [[ [[ [[ [[

YYYY [[ Y 4 [ 4

FAK-PI3K: loss is the change because of the loss of signaling through the FAK-PI3K pathway. FAK: RTK is the change because of any shift in PI3K association with FAK versus other RTKs. RTK-PI3K inhibition is the change because of PI3K inhibition upon the RTK-PI3K inhibition of palladin expression. TNK2 stimulation is the change because of stimulation of TNK2 by a shift in PI3K to favor growth factor signaling. Actin association is the change because of the disassociation of palladin from focal adhesions (positive for disassociation). aSMA is the change because of alteration of the level of aSMA inhibition. Biophysical Journal 109(2) 249–264

262

observe that an increase in matrix stiffness of ~fourfold from ~200 Pa to ~800 Pa induces a change in fibroblast phenotype to that of a myofibroblast that does not require exogenous TGFb, is fibronectin and CCL dependent, and converges chemical and mechanical signaling pathways through CCR4 chemokine receptors, b1 and b3 integrins, FAK, and PI3K, with palladin as a central node in these pathways, to jointly control myofibroblast gene expression. SUPPORTING MATERIAL Five figures and three tables are available at http://www.biophysj.org/ biophysj/supplemental/S0006-3495(15)00616-5.

AUTHOR CONTRIBUTIONS L.A.L. and J.S.M. designed the research, J.S.M. performed the research, L.A.L. and J.S.M. analyzed data, and L.A.L. and J.S.M. wrote the article.

McLane and Ligon 12. Brentnall, T. A., L. A. Lai, ., R. Chen. 2012. Arousal of cancer-associated stroma: overexpression of palladin activates fibroblasts to promote tumor invasion. PLoS One. 7:e30219. 13. Marsh, T., K. Pietras, and S. S. McAllister. 2013. Fibroblasts as architects of cancer pathogenesis. Biochim. Biophys. Acta. 1832:1070–1078. 14. Apostolopoulou, M., and L. Ligon. 2012. Cadherin-23 mediates heterotypic cell-cell adhesion between breast cancer epithelial cells and fibroblasts. PLoS One. 7:e33289. 15. Li, Z., J. A. Dranoff, ., R. G. Wells. 2007. Transforming growth factor-beta and substrate stiffness regulate portal fibroblast activation in culture. Hepatology. 46:1246–1256. 16. Desmoulie`re, A., A. Geinoz, ., G. Gabbiani. 1993. Transforming growth factor-beta 1 induces alpha-smooth muscle actin expression in granulation tissue myofibroblasts and in quiescent and growing cultured fibroblasts. J. Cell Biol. 122:103–111. 17. Amatangelo, M. D., D. E. Bassi, ., E. Cukierman. 2005. Stromaderived three-dimensional matrices are necessary and sufficient to promote desmoplastic differentiation of normal fibroblasts. Am. J. Pathol. 167:475–488. 18. Quiros, R. M., M. Valianou, ., E. Cukierman. 2008. Ovarian normal and tumor-associated fibroblasts retain in vivo stromal characteristics in a 3-D matrix-dependent manner. Gynecol. Oncol. 110:99–109.

ACKNOWLEDGMENTS

19. Plant, A. L., K. Bhadriraju, ., J. T. Elliott. 2009. Cell response to matrix mechanics: focus on collagen. Biochim. Biophys. Acta. 1793: 893–902.

We thank the labs of M. Hahn and R. Gilbert. We also thank Edna Cukierman for helpful discussions and sharing of preliminary data.

20. Meredith, Jr., J. E., B. Fazeli, and M. A. Schwartz. 1993. The extracellular matrix as a cell survival factor. Mol. Biol. Cell. 4:953–961.

This work has been supported by Research Scholar Grant RSG-10-245-01CSM from the American Cancer Society.

21. Li, Y., A. Asadi, ., E. P. Douglas. 2009. pH effects on collagen fibrillogenesis in vitro: electrostatic interactions and phosphate binding. Mater. Sci. Eng. C. 29:1643–1649.

REFERENCES

22. Yang, Y.-L., L. M. Leone, and L. J. Kaufman. 2009. Elastic moduli of collagen gels can be predicted from two-dimensional confocal microscopy. Biophys. J. 97:2051–2060.

1. Paszek, M. J., N. Zahir, ., V. M. Weaver. 2005. Tensional homeostasis and the malignant phenotype. Cancer Cell. 8:241–254.

23. Wallace, D. G., and J. Rosenblatt. 2003. Collagen gel systems for sustained delivery and tissue engineering. Adv. Drug Deliv. Rev. 55:1631– 1649.

2. Levental, K. R., H. Yu, ., V. M. Weaver. 2009. Matrix crosslinking forces tumor progression by enhancing integrin signaling. Cell. 139:891–906.

24. Jeong, J. H., Y. Liang, ., H. Kong. 2013. Stiffness-modulated water retention and neovascularization of dermal fibroblast-encapsulating collagen gel. Tissue Eng. Part A. 19:1275–1284.

3. Leight, J. L., M. A. Wozniak, ., C. S. Chen. 2012. Matrix rigidity regulates a switch between TGF-b1-induced apoptosis and epithelialmesenchymal transition. Mol. Biol. Cell. 23:781–791.

25. Liang, Y., J. Jeong, ., H. Kong. 2011. A cell-instructive hydrogel to regulate malignancy of 3D tumor spheroids with matrix rigidity. Biomaterials. 32:9308–9315.

4. Chia, H. N., M. Vigen, and A. M. Kasko. 2012. Effect of substrate stiffness on pulmonary fibroblast activation by TGF-b. Acta Biomater. 8:2602–2611.

26. Gupta, V., D. E. Bassi, ., E. Cukierman. 2011. Elevated expression of stromal palladin predicts poor clinical outcome in renal cell carcinoma. PLoS One. 6:e21494.

5. Hinz, B., D. Mastrangelo, ., G. Gabbiani. 2001. Mechanical tension controls granulation tissue contractile activity and myofibroblast differentiation. Am. J. Pathol. 159:1009–1020.

27. Goicoechea, S. M., R. Garcıá-Mata, ., C. A. Otey. 2014. Palladin promotes invasion of pancreatic cancer cells by enhancing invadopodia formation in cancer-associated fibroblasts. Oncogene. 33:1265–1273.

6. Tomasek, J. J., C. J. Haaksma, ., M. B. Vaughan. 1992. Fibroblast contraction occurs on release of tension in attached collagen lattices: dependency on an organized actin cytoskeleton and serum. Anat. Rec. 232:359–368.

28. Pogue-Geile, K. L., R. Chen, ., T. A. Brentnall. 2006. Palladin mutation causes familial pancreatic cancer and suggests a new cancer mechanism. PLoS Med. 3:e516.

7. Tomasek, J. J., G. Gabbiani, ., R. A. Brown. 2002. Myofibroblasts and mechano-regulation of connective tissue remodelling. Nat. Rev. Mol. Cell Biol. 3:349–363. 8. Wang, H., M. W. Tibbitt, ., K. S. Anseth. 2013. Hydrogels preserve native phenotypes of valvular fibroblasts through an elasticity-regulated PI3K/AKT pathway. Proc. Natl. Acad. Sci. USA. 110:19336– 19341. 9. Kalluri, R., and M. Zeisberg. 2006. Fibroblasts in cancer. Nat. Rev. Cancer. 6:392–401. 10. Xing, F., J. Saidou, and K. Watabe. 2010. Cancer associated fibroblasts (CAFs) in tumor microenvironment. Front. Biosci. 15:166–179. 11. Cirri, P., and P. Chiarugi. 2011. Cancer associated fibroblasts: the dark side of the coin. Am. J. Cancer Res. 1:482–497. Biophysical Journal 109(2) 249–264

29. Goicoechea, S. M., B. Bednarski, ., C. A. Otey. 2010. Isoform-specific upregulation of palladin in human and murine pancreas tumors. PLoS One. 5:e10347. 30. Najm, P., and M. El-Sibai. 2014. Palladin regulation of the actin structures needed for cancer invasion. Cell Adhes. Migr. 8:29–35. 31. von Nandelstadh, P., E. Gucciardo, ., K. Lehti. 2014. Actin-associated protein palladin promotes tumor cell invasion by linking extracellular matrix degradation to cell cytoskeleton. Mol. Biol. Cell. 25:2556–2570. 32. Parast, M. M., and C. A. Otey. 2000. Characterization of palladin, a novel protein localized to stress fibers and cell adhesions. J. Cell Biol. 150:643–656. 33. Dixon, R. D. S., D. K. Arneman, ., C. A. Otey. 2008. Palladin is an actin cross-linking protein that uses immunoglobulin-like domains to bind filamentous actin. J. Biol. Chem. 283:6222–6231.

Fibroblast Activation in the TME 34. Boukhelifa, M., M. M. Parast, ., C. A. Otey. 2004. Palladin is a novel binding partner for Ena/VASP family members. Cell Motil. Cytoskeleton. 58:17–29. 35. Boukhelifa, M., M. Moza, ., C. A. Otey. 2006. The proline-rich protein palladin is a binding partner for profilin. FEBS J. 273:26–33. 36. Ro¨nty, M., A. Taivainen, ., O. Carpeń. 2004. Molecular analysis of the interaction between palladin and alpha-actinin. FEBS Lett. 566:30–34. 37. Liu, X. S., H. J. Luo, ., Z. G. Wang. 2007. Palladin regulates cell and extracellular matrix interaction through maintaining normal actin cytoskeleton architecture and stabilizing beta1-integrin. J. Cell. Biochem. 100:1288–1300. 38. Jin, L., Q. Gan, ., A. V. Somlyo. 2010. The actin associated protein palladin is important for the early smooth muscle cell differentiation. PLoS One. 5:e12823. 39. Jin, L. 2011. The actin associated protein palladin in smooth muscle and in the development of diseases of the cardiovasculature and in cancer. J. Muscle Res. Cell Motil. 32:7–17. 40. Niedenberger, B. A., V. K. Chappell, ., C. B. Geyer. 2013. Nuclear localization of the actin regulatory protein Palladin in sertoli cells. Mol. Reprod. Dev. 80:403–413. 41. Niedenberger, B. A., V. A. Chappell, ., C. B. Geyer. 2014. Actin dynamics regulate subcellular localization of the F-actin-binding protein PALLD in mouse Sertoli cells. Reproduction. 148:333–341. 42. Endlich, N., E. Schordan, ., K. Endlich; European Renal cDNA Bank Consortium. 2009. Palladin is a dynamic actin-associated protein in podocytes. Kidney Int. 75:214–226. 43. Tomei, A. A., F. Boschetti, ., M. A. Swartz. 2009. 3D collagen cultures under well-defined dynamic strain: a novel strain device with a porous elastomeric support. Biotechnol. Bioeng. 103:217–225. 44. Schindelin, J., I. Arganda-Carreras, ., A. Cardona. 2012. Fiji: an open-source platform for biological-image analysis. Nat. Methods. 9:676–682. 45. Livak, K. J., and T. D. Schmittgen. 2001. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods. 25:402–408. 46. Carpenter, A. E., T. R. Jones, ., D. M. Sabatini. 2006. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 7:R100. 47. Kamentsky, L., T. R. Jones, ., A. E. Carpenter. 2011. Improved structure, function and compatibility for CellProfiler: modular highthroughput image analysis software. Bioinformatics. 27:1179–1180. 48. Mostaço-Guidolin, L. B., A. C.-T. Ko, ., M. G. Sowa. 2013. Collagen morphology and texture analysis: from statistics to classification. Sci. Rep. 3:2190. 49. Kojima, Y., A. Acar, ., A. Orimo. 2010. Autocrine TGF-beta and stromal cell-derived factor-1 (SDF-1) signaling drives the evolution of tumor-promoting mammary stromal myofibroblasts. Proc. Natl. Acad. Sci. USA. 107:20009–20014. 50. Eliceiri, B. P. 2001. Integrin and growth factor receptor crosstalk. Circ. Res. 89:1104–1110. 51. Zhang, T., R. Somasundaram, ., D. Herlyn. 2006. Migration of cytotoxic T lymphocytes toward melanoma cells in three-dimensional organotypic culture is dependent on CCL2 and CCR4. Eur. J. Immunol. 36:457–467. 52. Chvatchko, Y., A. J. Hoogewerf, ., C. A. Power. 2000. A key role for Cc chemokine receptor 4 in lipopolysaccharide-induced endotoxic shock. J. Exp. Med. 191:1755–1764. 53. Bielefeld, K. A., S. Amini-Nik, ., B. A. Alman. 2011. Fibronectin and b-catenin act in a regulatory loop in dermal fibroblasts to modulate cutaneous healing. J. Biol. Chem. 286:27687–27697. 54. Greenberg, R. S., A. M. Bernstein, ., S. K. Masur. 2006. FAK-dependent regulation of myofibroblast differentiation. FASEB J. 20:1006– 1008. 55. Humphries, J. D., A. Byron, and M. J. Humphries. 2006. Integrin ligands at a glance. J. Cell Sci. 119:3901–3903.

263 56. Barczyk, M., S. Carracedo, and D. Gullberg. 2010. Integrins. Cell Tissue Res. 339:269–280. 57. Chaudhuri, O., S. T. Koshy, ., D. J. Mooney. 2014. Extracellular matrix stiffness and composition jointly regulate the induction of malignant phenotypes in mammary epithelium. Nat. Mater. 13:970–978. 58. Carracedo, S., N. Lu, ., D. Gullberg. 2010. The fibroblast integrin alpha11beta1 is induced in a mechanosensitive manner involving activin A and regulates myofibroblast differentiation. J. Biol. Chem. 285:10434–10445. 59. Gonda, T. A., S. Tu, and T. C. Wang. 2009. Chronic inflammation, the tumor microenvironment and carcinogenesis. Cell Cycle. 8:2005–2013. 60. Otranto, M., V. Sarrazy, ., A. Desmoulie`re. 2012. The role of the myofibroblast in tumor stroma remodeling. Cell Adhes. Migr. 6:203–219. 61. Jones, C., and H. P. Ehrlich. 2011. Fibroblast expression of a-smooth muscle actin, a2b1 integrin and avb3 integrin: influence of surface rigidity. Exp. Mol. Pathol. 91:394–399. 62. To, W. S., and K. S. Midwood. 2011. Plasma and cellular fibronectin: distinct and independent functions during tissue repair. Fibrogenesis Tissue Repair. 4:21. 63. Kohan, M., A. F. Muro, ., N. Berkman. 2010. EDA-containing cellular fibronectin induces fibroblast differentiation through binding to alpha4beta7 integrin receptor and MAPK/Erk 1/2-dependent signaling. FASEB J. 24:4503–4512. 64. Kato, T., H. Saeki, ., S. Sato. 2011. Thymus and activation-regulated chemokine (TARC)/CC chemokine ligand (CCL) 17 accelerates wound healing by enhancing fibroblast migration. Exp. Dermatol. 20:669–674. 65. Dvorak, H. F. 1986. Tumors: wounds that do not heal. Similarities between tumor stroma generation and wound healing. N. Engl. J. Med. 315:1650–1659. 66. Rybinski, B., J. Franco-Barraza, and E. Cukierman. 2014. The wound healing, chronic fibrosis, and cancer progression triad. Physiol. Genomics. 46:223–244. 67. Wang, S. E., B. Xiang, ., C. L. Arteaga. 2009. Transforming growth factor b induces clustering of HER2 and integrins by activating Srcfocal adhesion kinase and receptor association to the cytoskeleton. Cancer Res. 69:475–482. 68. Chen, J., H. Li, ., J. H. C. Wang. 2007. Alpha-smooth muscle actin expression enhances cell traction force. Cell Motil. Cytoskeleton. 64:248–257. 69. Chan, M. W. C., P. D. Arora, ., C. A. McCulloch. 2009. FAK, PIP5KIgamma and gelsolin cooperatively mediate force-induced expression of alpha-smooth muscle actin. J. Cell Sci. 122:2769–2781. 70. Wang, C. C. L., I. Gurevich, and B. Draznin. 2003. Insulin affects vascular smooth muscle cell phenotype and migration via distinct signaling pathways. Diabetes. 52:2562–2569. 71. Lai, D., L. Ma, and F. Wang. 2012. Fibroblast activation protein regulates tumor-associated fibroblasts and epithelial ovarian cancer cells. Int. J. Oncol. 41:541–550. 72. Jia, J., T. A. Martin, ., W. G. Jiang. 2014. FAP-a (Fibroblast activation protein-a) is involved in the control of human breast cancer cell line growth and motility via the FAK pathway. BMC Cell Biol. 15:16. 73. Kennedy, A., H. Dong, ., W. T. Chen. 2009. Elevation of seprase expression and promotion of an invasive phenotype by collagenous matrices in ovarian tumor cells. Int. J. Cancer. 124:27–35. 74. Chin, Y. R., and A. Toker. 2010. Akt2 regulates expression of the actinbundling protein palladin. FEBS Lett. 584:4769–4774. 75. Jin, L., M. J. Kern, ., A. V. Somlyo. 2007. Angiotensin II, focal adhesion kinase, and PRX1 enhance smooth muscle expression of lipoma preferred partner and its newly identified binding partner palladin to promote cell migration. Circ. Res. 100:817–825. 76. Hirata, H., H. Tatsumi, and M. Sokabe. 2008. Mechanical forces facilitate actin polymerization at focal adhesions in a zyxin-dependent manner. J. Cell Sci. 121:2795–2804. Biophysical Journal 109(2) 249–264

264 77. Chang, E. H., A. H. Gasim, ., C. A. Otey. 2015. Palladin is upregulated in kidney disease and contributes to epithelial cell migration after injury. Sci. Rep. 5:7695. 78. Hinz, B., V. Dugina, ., C. Chaponnier. 2003. Alpha-smooth muscle actin is crucial for focal adhesion maturation in myofibroblasts. Mol. Biol. Cell. 14:2508–2519. 79. Puklin-Faucher, E., and M. P. Sheetz. 2009. The mechanical integrin cycle. J. Cell Sci. 122:179–186. 80. Chen, L. M., D. Bailey, and C. Fernandez-Valle. 2000. Association of beta 1 integrin with focal adhesion kinase and paxillin in differentiating Schwann cells. J. Neurosci. 20:3776–3784. 81. Crowe, D. L., and A. Ohannessian. 2004. Recruitment of focal adhesion kinase and paxillin to beta1 integrin promotes cancer cell migration via mitogen activated protein kinase activation. BMC Cancer. 4:18. 82. Wang, R., J. Bi, ., X. Zeng. 2013. Lipid rafts control human melanoma cell migration by regulating focal adhesion disassembly. Biochim. Biophys. Acta. 1833:3195–3205. 83. Chin, Y. R., and A. Toker. 2010. The actin-bundling protein palladin is an Akt1-specific substrate that regulates breast cancer cell migration. Mol. Cell. 38:333–344.

McLane and Ligon gene expression in the chicken primitive streak without affecting E-cadherin expression. BMC Dev. Biol. 11:20. 86. Matsumura, A., T. Kubota, ., E. Otsuji. 2013. HGF regulates VEGF expression via the c-Met receptor downstream pathways, PI3K/Akt, MAPK and STAT3, in CT26 murine cells. Int. J. Oncol. 42:535–542. 87. Guo, H., M. Gao, ., Z. Ding. 2014. Coordinate phosphorylation of multiple residues on single AKT1 and AKT2 molecules. Oncogene. 33:3463–3472. 88. Mahajan, K., and N. P. Mahajan. 2012. PI3K-independent AKT activation in cancers: a treasure trove for novel therapeutics. J. Cell. Physiol. 227:3178–3184. 89. Birchenall-Roberts, M. C., F. W. Ruscetti, ., S. J. Kim. 1990. Transcriptional regulation of the transforming growth factor beta 1 promoter by v-src gene products is mediated through the AP-1 complex. Mol. Cell. Biol. 10:4978–4983. 90. White, E. S., R. L. Sagana, ., A. F. Muro. 2010. Control of fibroblast fibronectin expression and alternative splicing via the PI3K/Akt/mTOR pathway. Exp. Cell Res. 316:2644–2653.

84. Abid, M. R., S. Guo, ., W. C. Aird. 2004. Vascular endothelial growth factor activates PI3K/Akt/forkhead signaling in endothelial cells. Arterioscler. Thromb. Vasc. Biol. 24:294–300.

91. Phanish, M. K., F. Heidebrecht, ., M. E. C. Dockrell. 2015. The regulation of TGFb1 induced fibronectin EDA exon alternative splicing in human renal proximal tubule epithelial cells. J. Cell. Physiol. 230:286–295.

85. Hardy, K. M., T. A. Yatskievych, ., P. B. Antin. 2011. FGF signalling through RAS/MAPK and PI3K pathways regulates cell movement and

92. Chapnick, D. A., L. Warner, ., X. Liu. 2011. Partners in crime: the TGFb and MAPK pathways in cancer progression. Cell Biosci. 1:42.

Biophysical Journal 109(2) 249–264

Spread of tumor microenvironment contributes to colonic obstruction through subperitoneal fibroblast activation in colon cancer.

Fibroblast activation protein α in tumor microenvironment: recent progression and implications (review).

Contribution of microRNAs in understanding the pancreatic tumor microenvironment involving cancer associated stellate and fibroblast cells.

Neutrophils in the Tumor Microenvironment.

Extracellular Hsp90 mediates an NF-κB dependent inflammatory stromal program: implications for the prostate tumor microenvironment.

Tumor microenvironment in the brain.

Human subperitoneal fibroblast and cancer cell interaction creates microenvironment that enhances tumor progression and metastasis.

The tumor microenvironment and cancer.

HSPG2 expression in the desmoplastic prostate tumor microenvironment.

Digitally deconvolving the tumor microenvironment.

Fibroblast ERα promotes bladder cancer invasion via increasing the CCL1 and IL-6 signals in the tumor microenvironment.

The role of CXCL12 in tumor microenvironment.

The microenvironment reprograms circuits in tumor cells.

Monitoring immune responses in the tumor microenvironment.

Metabolic pattern formation in the tumor microenvironment.

DNA methylation in the tumor microenvironment.

Decreased decorin expression in the tumor microenvironment.

NK cells in the tumor microenvironment.

Therapeutic Lymphoid Organogenesis in the Tumor Microenvironment.

Immune suppressive mechanisms in the tumor microenvironment.

Network dynamics in the tumor microenvironment.

interleukin-6 signaling in the tumor microenvironment.

Tumor-associated macrophages as major players in the tumor microenvironment.

mTORC2 signaling mediates TGFβ1-induced fibroblast activation and kidney fibrosis.