Stiffness increases with myofibroblast content and collagen density in mesenchymal high grade serous ovarian cancer

Abstract

Women diagnosed with high-grade serous ovarian cancers (HGSOC) are still likely to exhibit a bad prognosis, particularly when suffering from HGSOC of the Mesenchymal molecular subtype (50% cases). These tumors show a desmoplastic reaction with accumulation of extracellular matrix proteins and high content of cancer-associated fibroblasts. Using patient-derived xenograft mouse models of Mesenchymal and Non-Mesenchymal HGSOC, we show here that HGSOC exhibit distinct stiffness depending on their molecular subtype. Indeed, tumor stiffness strongly correlates with tumor growth in Mesenchymal HGSOC, while Non-Mesenchymal tumors remain soft. Moreover, we observe that tumor stiffening is associated with high stromal content, collagen network remodeling, and MAPK/MEK pathway activation. Furthermore, tumor stiffness accompanies a glycolytic metabolic switch in the epithelial compartment, as expected based on Warburg's effect, but also in stromal cells. This effect is restricted to the central part of stiff Mesenchymal tumors. Indeed, stiff Mesenchymal tumors remain softer at the periphery than at the core, with stromal cells secreting high levels of collagens and showing an OXPHOS metabolism. Thus, our study suggests that tumor stiffness could be at the crossroad of three major processes, i.e. matrix remodeling, MEK activation and stromal metabolic switch that might explain at least in part Mesenchymal HGSOC aggressiveness.

Figure 1

Tumor stiffness increases with growth of mesenchymal HGSOC. (A) Representative views of HES staining from PDX mouse models (Up) and corresponding human HGSOC (Bottom). Mesenchymal (OV26, OV21) and Non-Mesenchymal (OV33) PDX are shown. Scale bar, 50 μm. (B) Hierarchical clustering using Euclidean distance and Complete agglomeration method based on Mesenchymal and Non-Mesenchymal gene signatures (defined in) from Institut Curie’s HGSOC data set. Each row represents a gene, and each column a tumor, with 107 HGSOC patients and 13 PDX. OV26, OV21 and OV33 are indicated. Blue and red squares indicate gene expression in each tumor below and above the mean, respectively. Color saturation indicates magnitude of deviation from the mean. The dendogram of samples (above the matrix) allows classification of patient and PDX in Mesenchymal (red) and Non-Mesenchymal (blue) subgroups. (C) Representative colored stiffness maps in the transverse plan of Mesenchymal and Non-Mesenchymal PDX tumor growth over time. Colored map represents the Young’s modulus value (E) for each pixel, stiffness scale ranging from 0 kPa (blue) to 120 kPa (red). Dotted lines, drawn by hand, delineate tumor border and area. t0 corresponds to the first day of tumor stiffness measurement, and the following days of measures are indicated. (A) Stands for tumor area and (E) for mean tumor stiffness per pixel at each time point. (D) Variations of stiffness values in tumor area over time. The total tumor area occupied by pixels of a specific stiffness value (pixel stiffness range: 0 to 200 kPa), inside the same representative tumor as in (C), from t0 and all along measurements (d, in days) in PDX models. Data are expressed as percentages rather than in bins in order to compensate for the increasing number of pixels obtained as tumors grow. (E–G) Correlation plots between mean tumor stiffness and mean tumor area upon growth of tumors from Mesenchymal OV26 (n = 22) (E) and OV21 (n = 30) (F), and Non-Mesenchymal OV33 (n = 18) (G) PDX models. Each dot refers to a single tumor measurement at a given time. The number of measures per tumor (m = 73 (E), 156 (F), 91 (G)), depends on the PDX follow-up duration limited by ethical concerns. Correlation coefficient σ and P value are based on Spearman’s rank correlation test. (H) Mean tumor stiffness curves over time for Mesenchymal OV26 (n = 20) and OV21 (n = 22) (F), and Non-Mesenchymal OV33 (n = 16) PDX models. P values are based on Welch's t-test. (I) Histograms of stiffness values in tumor area. The total tumor area occupied by pixels of a specific stiffness value (pixel stiffness range: 0 to 200 kPa) between soft and stiff Mesenchymal OV26 (soft: dark blue dashed line, n = 8; stiff: red line, n = 7), soft and stiff Mesenchymal OV21 (soft: purple dashed line, n = 13; stiff: light red line, n = 9) and Non-Mesenchymal OV33 (soft: light blue dashed line, n = 15) tumors. Data are expressed as percentages of tumor area rather than in bins in order to compensate for the increasing number of pixels obtained as tumors grow. (J) Correlation plot between stiffness value of each pixel and distance from the tumor barycenter in Mesenchymal OV26 (soft n = 8; stiff: n = 8) and OV21 (soft n = 13; stiff n = 9) and Non-Mesenchymal OV33 (soft n = 15) tumors. Correlation coefficients σ and P value are based on Spearman’s rank correlation test.

Figure 2

Tumor stiffness is associated with high myofibroblast content in Mesenchymal HGSOC. (A) Representative views of HES staining showing stromal (orange) and epithelial (pink) compartments in soft and stiff Mesenchymal and Non-Mesenchymal tumors. Scale bar, 200 μm. (B) Same as in (A) showing smooth muscle α-actin (SMA) immunostaining. (C) Scatter plot showing percentages of stroma in soft (dot) versus stiff (triangle) Mesenchymal (OV26: soft n = 9, stiff n = 11; OV21: soft n = 13, stiff n = 9) and Non-Mesenchymal (n = 15) HGSOC. Data are shown as mean ± S.E.M. P values from Mann Whitney test. (D) Same as in (C) showing SMA histological scores (Hscores). P values from Welch's t-test. (E–G) Same as in (A) at the center (top) and periphery (bottom) of soft (left) and stiff (right) Mesenchymal (E,F) and Non-Mesenchymal (G) HGSOC. (H–J) Scatter plot showing percentages of stroma at the center (C, plain) and periphery (P, empty) in soft (dot) versus stiff (triangle) Mesenchymal (OV26: soft n = 6, stiff n = 6) (H); (OV21: soft n = 13, stiff n = 9) (I) and Non-Mesenchymal (n = 15) (J) HGSOC. Data are shown as mean ± S.E.M. P values from Welch's t-test.

Figure 3

Tumor stiffening is correlated with collagen remodeling in Mesenchymal HGSOC. (A) Representative views of Masson’s trichrome staining showing nuclei (dark purple), cytoplasm (purple) and collagen (green/blue) in soft and stiff Mesenchymal and Non-Mesenchymal HGSOC. Scale bar, 200 μm. (B) Scatter plot showing collagen density in soft (dot) and stiff (triangle) Mesenchymal (OV26: soft n = 6, stiff n = 6; OV21: soft n = 13, stiff n = 9) and Non-Mesenchymal (n = 15) tumors. Data are shown as mean ± S.E.M. P values from Welch's t-test. (C) Correlation plots between stroma content (as percentage of total tumor section) and collagen density (evaluated using image J software, see “Methods”) in Mesenchymal (OV26: n = 12; OV21: n = 22) and Non-Mesenchymal (n = 14) HGSOC. Correlation coefficients σ and P values are based on Spearman’s rank correlation test. (D) Same as in (C) between mean tumor stiffness (kPa) and collagen density. (E) Representative projected stack images of SHG signal in soft and stiff Mesenchymal and Non-Mesenchymal tumor sections. Scale bar, 100 μm. (F,H,J) Scatter plots showing collagen fiber length (F), thickness (H) and integrated density (see “Methods”) (J) in soft (dot, n = 3) and stiff (triangle, n = 3) Mesenchymal and Non-Mesenchymal (n = 3) tumor sections. Around 100 collagen fibers were measured in at least 10 representative regions per tumor. Data are shown as mean ± S.E.M. P values from Unpaired t-test. (G,I,K) Same as in (F,H,J) showing collagen fiber length (G), thickness (I) and integrated density (i.e. product of area and mean grey value) (K) at center (C, plain) or periphery (P) of soft (n = 3) and stiff (n ≥ 2) Mesenchymal and Non-Mesenchymal (n = 3) tumor sections. Around 100 collagen fibers were measured in at least 10 representative regions per tumor. Data are shown as mean ± S.E.M. P values are based on Paired t-test when comparing center versus periphery of the same tumor and Welch’s t-test when comparing soft versus stiff tumors.

Figure 4

MEK is activated upon tumor stiffening of mesenchymal HGSOC. (A) Representative views of YAP staining in soft (left) and stiff (right) Mesenchymal OV26 tumors. Scale bar: 100 μm. (B) Bar plots showing CYR61, CTGF, ANXA3 and ANKRD1 mRNA expression levels normalized to cyclophilin A in soft (n = 6) and stiff (n = 6) tumors. Data are shown as mean ± S.E.M. P values from Student’s t-test. (C) Representative western blots showing the phosphorylated form (P-) and the total protein levels of MEK in soft (n = 7) versus stiff (n = 7) Mesenchymal (OV26) and Non-Mesenchymal (OV33) (n = 12) HGSOC. (D) Same as in (C) for P38, AKT, JNK-1 and JNK-2 in soft (n = 7) versus stiff (n = 7) Mesenchymal OV26 and Non-Mesenchymal (OV33) (n = 11) tumors. Dashed lines are used to delineate different parts of two different gels run, blotted and revealed at the same time, with the same time of exposure. (E) Scatter plots of P-MEK/MEK and MEK/Actin ratios from soft (dot) and stiff (triangle) Mesenchymal (OV26: soft n = 7, stiff n = 7; OV21: soft n = 10, stiff n = 9) and Non-Mesenchymal (OV33, n = 12) tumors, as assessed by densitometry analysis of western blots shown in (C and Supplementary Fig. 2E). Data are shown as mean ± S.E.M. P values from Welch's t-test (left panel) and Mann–Whitney test (right panel). (F) Same as in (E) for P-P38/P38, P-AKT/AKT and P-JNK/JNK ratios (but n = 11 for OV33 Non-Mesenchymal tumors). (G) Representative images of human MRC5 fibroblasts, cultured 60 h either on soft (1 kPa—left) or stiff (50 kPa—right) polyacrylamide hydrogels. (H) Representative western blots showing P-MEK and MEK protein levels from MRC5 cells cultured as described in (G). Dashed line is used to delineate different parts from the same gel. (I) Representative images of ovarian cancer cells cultured in hanging drops for 6 h to 72 h or on plastic dish for 24 h. (J) Representative western blots showing P-MEK and MEK protein levels from SKOV3 and CAOV-3 ovarian cancer cell lines cultured 6 h or 24 h either on plastic plate (stiff) or in hanging drops (soft). Actin is used as an internal control for all protein loadings.

Figure 5

A metabolic switch occurs along tumor stiffening, in Mesenchymal HGSOC. (A–F) Representative pathways up-regulated in different tumor stiffness conditions and tumor localizations, as indicated by the different schemes. A plain circle (epithelial compartment) surrounded by wavy lines (stromal compartment) represents soft tumors. A plain circle (epithelial compartment) surrounded by straight lines (stromal compartment) illustrating matrix contraction, represents the core of stiff tumors. A thick circle line (epithelium) surrounded by straight lines (stroma) represents the periphery of stiff tumors. Orange represents the stromal compartment localization and green the epithelial compartment localization when performing differential analysis in stiff vs. soft tumors or within stiff tumors. Stromal and epithelial genes expression are based on murine MTA 1.0 or human HTA 2.0 microarrays, respectively. T-test was used to compare gene expression between two conditions. For each condition, the top 3 DAVID biological pathways defined by the up-regulated gene lists are represented. We analyzed 3 soft tumors, 5 stiff tumors both at their periphery or at their center. P values are presented as – Log 10. (A) Genes up-regulated in the epithelium at the center of stiff tumors (left) compared to genes up-regulated in the epithelium of soft tumors (right). (B) Genes up-regulated in the epithelium at the periphery of stiff tumors (left) compared to genes up-regulated in the epithelium of soft tumors (right). (C) Genes up-regulated in the epithelium at the center (left) compared to the periphery (right) of stiff tumors. (D) Genes up-regulated in the stroma at the center of stiff tumors (left) compared to genes up-regulated in the stroma of soft tumors (right). (E) Genes up-regulated in the stroma at the periphery of stiff tumors (left) compared to genes up-regulated in the stroma of soft tumors (right). (F) Genes up-regulated in the stroma at the center (left) compared to the periphery (right) of stiff tumors.