Cell matrix remodeling ability shown by image spatial correlation

Abstract

Extracellular matrix (ECM) remodeling is a critical step of many biological and pathological processes. However, most of the studies to date lack a quantitative method to measure ECM remodeling at a scale comparable to cell size. Here, we applied image spatial correlation to collagen second harmonic generation (SHG) images to quantitatively evaluate the degree of collagen remodeling by cells. We propose a simple statistical method based on spatial correlation functions to determine the size of high collagen density area around cells. We applied our method to measure collagen remodeling by two breast cancer cell lines (MDA-MB-231 and MCF-7), which display different degrees of invasiveness, and a fibroblast cell line (NIH/3T3). We found distinct collagen compaction levels of these three cell lines by applying the spatial correlation method, indicating different collagen remodeling ability. Furthermore, we quantitatively measured the effect of Latrunculin B and Marimastat on MDA-MB-231 cell line collagen remodeling ability and showed that significant collagen compaction level decreases with these treatments.

Figure 1

Experimental workflow—Collagen remodeling analysis using image spatial correlation. Upper right panel: collagen SHG images were acquired by 2-photon microscopy. 10 z-stack slices across 200 μm depth from culture dish surface were collected. Each slice covers 850 μm × 850 μm. Upper left panel: an example of collagen SHG images from close to the culture dish surface (slice 1) to 200 μm above (slice 10). The SHG intensity decreases with the increasing depth of acquisition. However, since image spatial correlation is not sensitive to average intensity change, image spatial correlation (1) was applied to each slice, and the results were averaged to produce one correlation image. Lower left panel: the spatial correlation image calculated from the SHG z-stack. The relatively symmetrical autocorrelation pattern indicates that with our sample preparation, collagen fibers were not preferentially aligned toward specific direction at the scale of the whole image. Lower right panel: the x and y cross sections from the spatial correlation image were fitted with two Gaussian components. One component has the standard deviation close to pixel size (green curve); the other component has broader standard deviation (blue curve), and its width varies based on the collagen compaction level.

Figure 2

Representative SHG images of collagen matrix with and without cells. (a) Type I collagen matrix without cells after 1, 2, and 4 days of incubation. There is no recognizable collagen fiber distribution heterogeneity or collagen compaction. (b) Collagen matrix with MCF-7 cells after 1, 2, and 4 days of incubation. There is no distinguishable feature until day 4, in which collagen matrix with MCF-7 cells shows mild remodeling, although the majority of the collagen matrix was not compacted by cells. (c) Collagen matrix with NIH/3T3 cells after 1, 2, and 4 days of incubation. A higher degree of collagen compaction compared to the collagen matrix with MCF-7 cells can be seen since day 2. Higher collagen density was also observed, possibly due to collagen synthesis by the fibroblasts. (d) Collagen matrix with MDA-MB-231 cells after 1, 2, and 4 days of incubation. Collagen matrix shows significantly higher degree of remodeling compared to both MCF-7 and NIH/3T3 cell-embedded collagen matrices. (e) Actin-GFP labeled MDA-MB-231 cells cultured in type I collagen matrix for 4 days, showing that collagen matrix (magenta) has higher density close to cells (green). The higher density area around the cell is referred to as collagen compaction, and its size is quantitatively evaluated by image spatial correlation in this report.

Figure 3

The standard deviation of narrow peak is not significantly affected by collagen remodeling. For the collagen matrix with or without MDA-MB-231 cells embedded, the standard deviation of the narrow peak from spatial correlation images (the green curve in Figure 1) did not show any significant difference, and its value was close to the pixel size.

Figure 4

Comparison of collagen remodeling degree with different cell lines. The degree of collagen remodeling by MCF-7, NIH/3T3, and MDA-MB-231 cell lines over 4 days of incubation was evaluated by σ of the low spatial frequency component (the blue curve in Figure 1). Of all three cell lines, MDA-MB-231 showed the highest degree of collagen remodeling, and the collagen compaction degree increases over 4-day period. MCF-7 exhibits the lowest degree of collagen compaction and has the σ value similar to the collagen matrix without cells. Dashed line shows the average σ value of collagen matrix without cells at day 1. * indicates that the standard deviation is significantly different from the baseline control (collagen matrix without cells for each equivalent age of the sample, P  value < 0.01 by t-test).

Figure 5

MMP inhibitor and actin polymerization inhibitor affect MDA-MB-231 collagen remodeling ability. MDA-MB-231 cells were cultured in collagen matrix and treated with Marimastat (MMP inhibitor) or Latrunculin B (actin polymerization inhibitor) for 4 consecutive days. Both treatments significantly compromised the collagen compaction ability of MDA-MB-231 cells. While the effect of Marimastat is milder, Latrunculin B treatment resulted in the compaction degree similar to collagen matrix with MCF-7 cells. * indicates the standard deviation is significantly different from the collagen matrix without cells (P  value < 0.01 by t-test).