CAF-Associated Genes in Breast Cancer for Novel Therapeutic Strategies

Abstract

Breast cancer (BC) is the most common cancer in women, and therapeutic strategies for it are based on the molecular subtypes of luminal BC, HER2 BC, and triple-negative BC (TNBC) because each subtype harbors different unique genetic aberrations. Recently, features of the tumor microenvironment (TME), especially cancer-associated fibroblasts (CAFs), have been demonstrated to play a critical role in BC progression, and we would like to understand the molecular features of BC CAFs for novel therapeutic strategies. In a recent study, 115 CAF-associated genes (CAFGs) were identified in a public database of microdissection and microarray data (GSE35602) from 13 colorectal cancer (CRC) tumors. Using a public database (GSE10797) of 28 BC tumors, a similar analysis was performed. In BC, 59 genes from the 115 CAFGs identified in CRC (CRC CAFGs) were also closely associated with a CAFs marker, SPARC (R = 0.6 or beyond), and POSTN was of particular interest as one of the BC CAFGs with the highest expression levels and a close association with SPARC expression (R = 0.94) in the cancer stroma of BC tumors. In BC stroma, POSTN was followed in expression levels by DKK3, MMP2, PDPN, and ACTA2. Unexpectedly, FAP and VIM were not as highly associated with SPARC expression in the cancer stroma of BC tumors and exhibited low expression. These findings suggested that ACTA2 might be the most relevant conventional CAFs marker in BC, and ACTA2 was actually correlated in expression with many CRC CAFGs, such as SPARC. Surprisingly, the SE ratio values of the BC CAFGs were much lower (average SE = 3.8) than those of the CRC CAFGs (SE = 10 or beyond). We summarized the current understanding of BC CAFs from the literature. Finally, in triple-negative BC (TNBC) (n = 5), SPARC expression uniquely showed a close association with COL11A1 and TAGLN expression, representing a myofibroblast (myCAFs) marker in the cancer stroma of the BC tumors, suggesting that myCAFs may be molecularly characterized by TNBC in contrast to other BC phenotypes. In summary, CAFs could have unique molecular characteristics in BC, and such TME uniqueness could be therapeutically targeted in BC.

Keywords: SPARC; breast cancer; cancer-associated fibroblasts (CAFs); colorectal cancer; triple-negative breast cancer; tumor microenvironment (TME).

Figure 1

Correlation with SPARC in cancer stroma and S/E ratio of 28 breast cancer cases in GSE10797. (a) Upper panel, the correlation coefficient with SPARC was calculated in the stroma of 28 breast cancer cases, and 58 genes with R ≥ 0.6 (59 genes including SPARC) were finally selected. Gold bars represent R ≥ 0.9, purple bars represent R ≥ 0.8, and reddish brown bars represent R ≥ 0.6, and black bars is CAFs markers (ACTA2 in the figures). Lower panel shows SE ratio of the above 59 genes. The red line is drawn at 10 because an SE ratio of 10 or beyond is considered a CAFGs. (b) All 5 genes associated with SPARC expression in cancer stroma of the BC tumors (R ≥ 0.9). The highest expression was reproducibly confirmed in GSM272720, while the blue arrow shows the BC case with the highest expression among Triple negative (TN) type BC (TN max). (c) Representative SPARC target genes identified by SPARC knockdown in YS1 cells (CAFs cell line), in which their expression was associated with SPARC expression in cancer stroma. The figure includes another SPARC probe (SPARC(2)), which had inconsistent data.

Figure 2

Molecular features of representative CAFGs in BC (BC CAFGs) among CRC CAFGs. (a) Upper panel, expression amounts of CAFGs in GEM272720. Middle panel, correlation coefficient between individual CAFGs and strSPARC in BC. Lower panel, SE ratio (all below 10). (b) Correlation of BC CAGs (DKK3 and MMP2) and well-known CRC CAFGs (ACTA2, FAP, and VIM) to strSPARC in BC is shown, in which FAP and VIM are not strongly associated with strSPARC (R = 0.47 and R = 0.18, respectively).

Figure 3

BC CAFGs and current understanding its functional relevance. (a) Relationship between CAFs expressing POSTN and/or ACTA2 (myCAFs) and CAFs expressing HIF1A. (b) scRNA analysis was performed on mesenchymal cells from mouse breast cancer and showed that they could be classified into 4 CAFs categories (vascular CAFs-vCAFs, matrix CAFs-mCAFs, cycling-cCAFs, and developmental CAFs-dCAFs). Well known CAFs markers (FAP, SPARC, ACTA2, and VIM) were expressed in all 4 CAFs, while PDGFRA was expressed only in mCAFs (this figures was made from reference [3]). (c) Correlation coefficient (R = 0.6 or beyond) between CRC CAFGs and strACTA2 in BC. (d) Representative correlation of CRC CAFGs associated with ACTA2 expression in cancer stroma.

Figure 4

CAFGs collagens in BC and novel therapeutic potential. Adipocyte-derived fibroblasts adjacent to breast tumor cells increase secretion of fibronectin and collagen I, promoting migration and invasive potentials, which can be dependent on activation of the Wnt/β-catenin pathway. Endo180 promotes cell migration and ECM remodeling by collagen uptake. PYCR1 is a key enzyme in proline synthesis and is involved in the production of ECM that promotes tumorigenesis and may be a therapeutic target.

Figure 5

Molecular characteristics of Semi-CAFGs in BC that may be involved in CAFs activation and have novel therapeutic potential. (a) Upper panel, correlation coefficients (R-index) between genes defined as Semi-CAFs in CRC and strSPARC. R-index = 0.6 or beyond represents colored bars. Lower panel, expression amounts of each Semi-CAFs in GSM272720. Outstandingly, TIMP3 expression was robust in cancer stroma of the BC tumors. (b) Semi-CAFGs (TIMP3, WISP1, TGFB1, and NOTCH4) expressions associated with SPARC expression in cancer stroma. (c) Upper panel, correlation coefficients (R-index = 0.6 or beyond with strSPARC) in BC among CRC CAFGs. Lower panel, expression amounts of each L-CAFs in GSM272720. (d) Correlation of CAFGs with strSPARC in BC (R ≥ 0.9) commonly in whole BC and TNBC was individually shown. (e) Correlation of CAFGs (COL11A1 and TAGLN) with strSPARC (R ≥ 0.9) only in TNBC was individually shown.TIMP (tissue inhibitor of metalloproteinase) is an inhibitor of the matrix metalloproteinase (MMP) family. Quadruple TIMP knockout (TIMPless, TIMP1, bSE = 1.7; TIMP2, bSE = 2.6; TIMP3, bSE = 3.1; TIMP4, bSE = 2.0) fibroblasts to unleash MMP activity and show that complete Timp loss is sufficient for the acquisition of hallmark CAFs functions in BC [40]. These findings suggested that MMP activity is critical for CAFs activity in BC.