Biglycan expression in the melanoma microenvironment promotes invasiveness via increased tissue stiffness inducing integrin-β1 expression

Novel targeted and immunotherapeutic approaches have revolutionized the treatment of metastatic melanoma. A better understanding of the melanoma-microenvironment, in particular the interaction of cells with extracellular matrix molecules, may help to further improve these new therapeutic strategies. We observed that the extracellular matrix molecule biglycan (Bgn) was expressed in certain human melanoma cells and primary fibroblasts when evaluated by microarray-based gene expression analysis. Bgn expression in the melanoma tissues correlated with low overall-survival and low progression-free-survival in patients. To understand the functional role of Bgn we used gene-targeted mice lacking functional Bgn. Here we observed that melanoma growth, metastasis-formation and tumor-related death were reduced in Bgn−/− mice compared to Bgn+/+ mice. In vitro invasion of melanoma cells into organotypic-matrices derived from Bgn−/− fibroblasts was reduced compared to melanoma invasion into Bgn-proficient matrices. Tissue stiffness as determined by atomic-force-microscopy was reduced in Bgn−/− matrices. Isolation of melanoma cells and fibroblasts from the stiffer Bgn+/+ matrices revealed an increase in integrin-β1 expression compared to the Bgn−/− fibroblast matrices. Overexpression of integrin-β1 in B16-melanoma cells abolished the survival benefit seen in Bgn−/− mice. Consistent with the studies performed in mice, the abundance of Bgn-expression in human melanoma samples positively correlated with the expression of integrin-β1, which is in agreement with results from the organotypic invasion-assay and the in vivo mouse studies. This study describes a novel role for Bgn-related tissue stiffness in the melanoma-microenvironment via regulation of integrin-β1 expression by melanoma cells in both mice and humans.


Organotypic invasion assay
Collagen I was isolated from rat tails using acid extraction. Mouse embryonic fibroblasts (immortalized) directed the polymerization of collagen thus synthesizing the membranes for cell invasion. Fibroblasts were provided collagen 1 and placed in 35 mm Petri dishes, the consequent polymerization of the collagen formed membranes of a size smaller than 15 mm. The membranes were then incubated with a suspension of 1x10 4 melanoma cells for three days, initially, to be later placed on metal grid platforms. An air/ liquid gradient drove invasive cell migration in direction of the media. After a 7 day migration period, membranes were fixed with paraformaldehyde and embedded in paraffin blocks. Thereafter the paraffin-embedded membranes were sectioned and stained with hematoxylin/eosin. Cell migration was quantified as the median of cells invading the membrane. This was calculated by taking the median of three microscopic fields at 10x magnification. For protein expression analysis, cells were isolated from the membranes after 7 days of migration using collagenase 1. The method was done as previously described (1).

Generation of MEFs on C3H genetic background
For a further series of organotypic invasion assay experiments, Bgn +/+ MEFs on C3H genetic background were obtained from Invivogen. These cells were isolated from 13.5 day old C3H WT mouse embryos and immortalized by stable transfection of an SV40 large antigen-expressing plasmid. To be able to compare the two cell lines, C3H derived Bgn -/-MEFs were immortalized using the same SV40 large antigen-expressing plasmid that was used for the other cells.

Picrosirius red staining
Paraffin-embedded organotypic cultures after 7 days of cell migration were sectioned and stained with Weigert's Hematoxylin for 8 minutes and counterstained with 0.1% Picrosirius Red (Direct Red 80; Sigma Aldrich) for 90 minutes as described previously (2). Images were taken using Axioplan2 fluorescence microscope (Zeiss) equipped with an analyzer and polarizer and Axiocam camera. All the images were taken using identical settings and exposure time. Collagen fibril density was quantified on images taken with orthogonally oriented polarized light using image J software. After setting a minimal threshold, the area (in pixel) of the brightness of the threshold light was calculated from a minimum of 9 images per condition.

Indentation-type atomic force microscopy
Indentation-type atomic force microscopy (AFM) is well suited for measuring the elasticity of soft collagencontaining matrices (3). Both WT-and Bgn-KO-MEFs contracted collagen matrices were measured on the same day after 13 days of contraction. Each sample was prepared directly before the experiment. Since the matrices were very soft and also fragile, it was crucial not to apply any excess force on the regions of interest during sample preparation. The individual matrices were gently transferred out of the 35mm Petri dish by completely filling the dish with buffer solution, closing the lid and flipping it upside down. As a consequence, the matrix floated into the lid. The 35mm dish was lifted away from the lid and replaced by a 100mm dish (without lid). After flipping this assembly again and removing the lid the collagen matrix was inside the larger dish and the buffer solution was dispersed. Next, a circular piece was punched out of the center using a 8mm tissue punch. A coverslip (22mm x 22mm) was glued on top of this piece using cyanoacrylate. The glass was used to lift this assembly with tweezers and to glue it into a 60mm dish which was filled immediately with PBS and used for the elasticity measurements.
Indentation experiments were performed on a JPK NanoWizard 3 BioScience AFM system equipped with a CellHesion 200 module, motorized x-y-precisionstage and TopView optics for sample alignment (JPK Instruments). Spherical indenters were prepared by gluing glass beads (PGB-007, Kisker Biotech) to soft tipless AFM cantilevers (MikroMasch CSC38/tipless/No Al, nominal spring constant 0.03 Nm −1 , NanoandMore) using ultraviolet curing glue (Norland Optical Adhesive 61, Norland Products). For the gluing procedure the AFM was mounted on an inverted optical microscope (Nikon Eclipse Ti-E with Intensilight as UV source for glue curing, Nikon Instruments). The bead radius was determined using the optical microscope. All measurements were performed with the same cantilever (r indenter =32μm). For force calibration, the nominal spring constant was used and the sensitivity was determined before the experiment series on bare glass in PBS buffer.
The indentation experiments were carried out in 60mm Petri dishes filled with PBS. The length of the force-distance curves was 25μm, the force setpoint was 1.5nN and the extend/retract speed was 5μm/s. On a central region of each sample between 100 and 150 force-distance curves were recorded on different points that were arranged like a grid and separated by at least 50μm. For both phenotypes three different matrices were measured and the respective curves were pooled for data analysis. The Young's moduli were determined by fitting the individual curves to a Hertz model using the JPK data processing software. Curves were rejected if they clearly deviated from a simple Hertzian-type indentation or if the fit residuals were significantly higher than for a typical indentation curve from the respective condition. The data are from the following number of curves: WT-MEFs (n=3) 349 curves were measured, 340 accepted, 9 discarded (2.6%). Bgn-KO-MEFs, (n=3) 415 curves measured, 391 accepted, 24 discarded (5.8%). The Young's moduli of the accepted curves are shown as scatter plots and notched boxplots in Figure 3. A two-sample Kolmogorov-Smirnov test indicates a significant difference between the two distributions with a p-value < 0.0001. OriginPro (Originlab) was used for data plotting and statistical testing.

Fibronectin fiber orientation
Bgn +/+ and Bgn -/-MEFs were plated on 15 μ-ibidi slides. After 5 days the samples were fixed, permeabilized and stained with anti-Fibronectin antibody (Sigma) and Hoechst33342. 15 μm Z-stacks were imaged and a maximal projection was analysed using MetaMorph. Tresholded fibers with less than 5 pixels were excluded. Percentages of distributed angles of fibers from -20° to 20° were combined as a measure for parallel/organised fibers as previously described (4). For comparison of the two groups the Mann-Whitney-Test was performed.