Cancer-associated Fibroblast-specific Expression of the Matricellular Protein CCN1 Coordinates Neovascularization and Stroma Deposition in Melanoma Metastasis

Abstract

Melanoma is the leading cause of skin cancer-related death. As prognosis of patients with melanoma remains problematic, identification of new therapeutic targets remains essential. Matricellular proteins are nonstructural extracellular matrix proteins. They are secreted into the tumor microenvironment to coordinate behavior among different cell types, yet their contribution to melanoma is underinvestigated. Examples of matricellular proteins include those comprising the CCN family. The CCN family member, CCN1, is highly proangiogenic. Herein, we show that, in human patients with melanoma, although found in several tumor cell types, CCN1 is highly expressed by a subset of cancer-associated fibroblasts (CAF) in patients with melanoma and this expression correlates positively with expression of proangiogenic genes and progressive disease/resistance to anti-PD1 checkpoint inhibitors. Consistent with these observations, in a syngeneic C57BL6 mouse model of melanoma, loss of CCN1 expression from Col1A2-Cre-, herein identified as "universal," fibroblasts, impaired metastasis of subcutaneously injected B16F10 tumor cells to lung, concomitant with disrupted neovascularization and collagen organization. Disruption of the extracellular matrix in the loss of CCN1 was validated using a novel artificial intelligence-based image analysis platform that revealed significantly decreased phenotypic fibrosis and composite morphometric collagen scores. As drug resistance is linked to matrix deposition and neoangiogenesis, these data suggest that CCN1, due to its multifaceted role, may represent a novel therapeutic target for drug-resistant melanoma. Our data further emphasize the essential role that cancer-associated, (universal) Col1A2-Cre-fibroblasts and extracellular matrix remodeling play in coordinating behavior among different cell types within the tumor microenvironment.

Significance: In human patients, the expression of proangiogenic matricellular protein CCN1 in CAFs correlates positively with expression of stroma and angiogenic markers and progressive disease/resistance to checkpoint inhibitor therapy. In an animal model, loss of CCN1 from CAFs impaired metastasis of melanoma cells, neovascularization, and collagen deposition, emphasizing that CAFs coordinate cellular behavior in a tumor microenvironment and that CCN1 may be a novel target.

FIGURE 1

CCN1 is expressed by a subset of CAFs and correlates with expression of angiogenic markers. A,CCN1 expression correlates with bulk tumor stromal scores. Scatterplot representation of stromal score (ESTIMATE algorithm) to CCN1 expression in bulk primary melanoma tumors from TCGA (n = 82). The linear regression line is shown in blue, and the 95% confidence interval is shown as a gray area around the regression line. Spearman correlation coefficient and corresponding P value are displayed in the bottom right corner. B,CCN1 gene expression (TPM/10) in melanoma tumor single cells [malignant cells (n = 2,018), T cells (n = 3,321), B cells (n = 818), macrophages (n = 420), natural killer cells (n = 92), endothelial cells (n = 104), and CAFs (n = 106)]. Boxes represent interquartile ranges, and points represent individual sample values. C, Heat map representation of the 50 genes with the highest SD in CAF single cells from human melanoma tumors. CCN1 expression is higher in a particular subset of CAFs (GSE115978). Columns represent single CAF cells, and rows represent Z-score mRNA expression (TPM/10). The horizontal bar denotes different melanoma samples. Scatterplots show CAF-specific CCN1 scores versus gene set enrichment of angiogenesis (D), or EMT gene sets (E; GAGE t statistics) of melanoma bulk primary melanoma tumors from TCGA samples (n = 82). The linear regression line is shown in blue, and the 95% confidence interval is shown as a gray area around the regression line. Spearman correlation coefficient and associated P values are displayed in the bottom right corner.

FIGURE 2

Mice harboring Col1A2-Cre-fibroblast–specific deletion for Ccn1 show impaired metastasis of B16F10 melanoma cells to the lung. A–C, Mice were examined after 14 days of tumor growth. A, Sections of tumor stroma were stained with anti-CCN1 antibody (N = 3, representative images are shown), verifying loss of CCN1 protein expression in mice deleted for CCN1 in fibroblasts. Note CCN1 staining remains in the tumor. Scale bar = 50 µm. B, Representative images of lung sections from wild-type or mice harboring a deletion for Ccn1 in Col1A2-Cre-fibroblasts. Hematoxylin and eosin was used to reveal detect dense metastatic foci (purple). Scale Bar = 300 µm. Total area of the lung section covered by metastases was quantified. Deletion of Ccn1 caused reduced metastasis (N = 6, ***, P < 0.001). C, Reduced CCN2 expression in the stroma did not significantly alter tumor growth. Kruskal–Wallis analysis (N = 8). D, To more accurately define the cells in which CCN1 was deleted, three-week old Col1A2-cre(ER)T/); mTmG mice were injected for 5 consecutive days with tamoxifen to activate Cre recombinase. Six weeks later, skin was isolated, digested with collagenase, and cells expressing GFP were isolated by FACS, and subjected to RNA-seq analysis. Col1A2 lineage fibroblasts were identified as universal fibroblasts based on the gene expression of Pi16, Col15a1, C3, and Cd84 genes, but not papillary or reticular fibroblasts based on the gene expression of Crabp1 and Defb8 or Nexn, Trim63, Actn2, and Hspb7, respectively.

FIGURE 3

In a syngeneic model of metastasis, CCN1-deficient stroma possesses reduced collagen deposition, whereas myofibroblast formation is unaffected. Skin tissue showed a significant decrease in collagen shown trichrome and Sirius red staining (A; arrow indicates stroma). Representative images shown; Graph shows mean area of collagen, detected using trichrome stain, present in stroma images ± SD; Student t test; n  =  3). α-SMA–expressing myofibroblasts (B) and YAP-positive nuclei (C) are present in both wild-type and CCN1-deficient stroma (Representative images shown; Graphs represent mean α-SMA positive cells ± SD; one-way ANOVA; n  =  5). Mice were examined after 14 days of tumor growth of B16F10 melanoma cells.

FIGURE 4

CCN1-deficient stroma show disorganized collagen fibers. Digital pathology phenotypic quantification (PharmaNest, Inc) phenotypic fibrosis histologic heat chart (A) where each row represents a principal qFT, at the three phenotypic levels (collagen deposition, fibers morphometry, and fibrosis architecture). B, The Ph-FCS recapitulates all the qFTs for one sample, and quantifies the phenotype of fibrosis and its differences between CCN1-deficient (KO; n = 5, mean = 1.5, Std dev = 0.5) and wild-type (WT) mice (n = 6, mean = 4.6, Std dev = 0.9) groups (3.1 fold change, P < 0.001, Student t test). C, The collagen fiber morphometric scores recapitulates the differences at the morphometric level between KO (n = 5, mean = 1.6, Std dev = 0.5) and WT (n = 6, mean = 4.5, Std dev = 0.9) groups (2.9 fold change, P < 0.001, Student t test). D, Representative images with augmented digital pathology layers. “Entropy Texture” is one of the architectural phenotypes of fibrosis (39). KO, knockout.

FIGURE 5

Loss of CCN1 expression from CAFs does not affect cell proliferation or apoptosis in a syngeneic model of melanoma metastasis. A, PCNA assay in (top) tumor and (bottom) stroma. Cell proliferation was detected using an anti-PCNA antibody. Representative images shown; graphs show the mean percentage of PCNA positive cells in CAFs ± SEM; CCN1^f/fn = 7 and CCN1^−/−n = 4, Student t test. B, TUNEL assay to detect apoptosis in stroma. Representative images shown; graphs show the mean percentage of TUNEL positive cells in stroma ± SEM, CCN1^f/fn = 7 and CCN1^−/−n = 4, Student t test.

FIGURE 6

Deletion of Ccn1 in Col1A2-Cre-fibroblasts reduces tumor vasculature. A, Tumors in wild-type or mice deleted for Ccn1 in Col1A2-Cre-fibroblasts were perfused with a CT-contrast agent, as described in Materials and Methods. Volume (mL) of tumor occupied by vasculature was calculated. Mice lacking Ccn1 in their fibroblasts had significantly reduced vascular volume (t test, n = 3, P < 0.05). B, Tumors and associated stroma in wild-type or mice deleted for Ccn1 in fibroblasts were stained with anti-CD31 antibodies (t test; n = 3; *, P < 0.05).

FIGURE 7

CCN1 expression by fibroblasts is associated with increased tumor penetration of CD4⁺ T cells and resistance to anti-PD1 checkpoint inhibitors. A, Tumors were subjected to flow cytometry with anti-CD4 anti-CD8 antibodies. Percentage of positive CD45-staining cells are indicated wild type (n = 3, CCN1^f/f) and CCN1-deficient (n = 3, CCN1^−/−) tumors (Student t test). Primary flow data of the median value is presented Tissue was examined after 14 days of tumor growth. B, Boxplot shows higher levels of CAF-specific CCN1 scores in patients on checkpoint inhibitors with progressive disease (GSE78220). A six-gene set of CAF-specific genes correlating with CCN1 (PDGFRA, COL1A1, DCN, TAGLN, COL6A3, and LPAR1), as described in Materials and Methods, was analyzed. P value (unpaired t test) displayed in top left corner. n = 14 in the responders and n = 12 in the progressive disease. C and D, Loss of CCN1 expression results in reduced MMP-9 expression in CCN1-deficient mice. Real-time PCR analysis of RNAs isolated from Ccn1^+/+ and Ccn1^−/− dermal fibroblasts from 3 different mice (N = 3; C) or skin (N = 4 Ccn1^+/+, N = 8, Ccn1^−/−; D) reveal that CCN1-deficient fibroblasts have reduced expression of the anti-PD1 resistance marker/effector MMP-9 (Student unpaired t test; *, P < 0.05). Please note that D was conducted on identical skin samples to those published previously (29).