Expression of CD44 (standard or variant isoforms) has been considered a prognostic marker for the progression of prostate cancer. The mechanism by which CD44 regulates the progression of prostate cancer is largely unknown. The present study was performed to evaluate the role of CD44 in prostate cancer-induced bone metastasis. We screened three cell lines (PC3, DU145, and LNCaP) for the expression of CD44. Normal prostatic epithelial (HPR-1) and benign prostatic hyperplasic cells (BPH) were used as controls. PC3 and DU145 cells were established from the bone and brain metastatic lesions of a prostate cancer patient, respectively. Our studies are in agreement with the majority of earlier studies[53, 54] in the expression of CD44 in androgen independent PC3 and DU145 cells, but not in androgen dependent LNCaP cells, which is established from a lymph node metastasis. Stable expression of androgen receptor in PC3 cells reduces CD44 expression to a significant level (data not shown).
The present study was undertaken to determine the possible mechanisms involved in the formation of osteolytic lesions associated with metastasis of prostate cancer cells to bone and the significance of CD44 and αvβ3 signaling. Previous studies in CD44 knockout mice link CD44 receptor with RANKL expression. Our results in PC3 cells show that RANKL expression is in part mediated by CD44 signaling through RUNX2. As a result of CD44 expression, we have found expression of RANKL and MMP9 through RUNX2-dependent signaling in PC3 cells. RUNX2 SiRNA reduces MMP9 expression but not MMP2 at mRNA level. On the other hand, androgen-dependent LNCaP cells demonstrated expression and secretion of MMP2 as a major metalloproteases (Additional file1: Figure S1). MMP2 expression may occur independent of RUNX2 and CD44 signaling in LNCaP cells. Consistent with our studies, others have shown negligible Runx2 in normal prostate epithelial and non-metastatic LNCaP cells. High Runx2 levels are associated with development of large tumors, increased expression of metastasis-related genes (MMP9, MMP13, VEGF, osteopontin) and secreted bone-resorbing factors (PTHrP, IL8) promoting osteolytic disease. Moreover, it was identified in co-culture studies that PC3 cells promote osteoclastogenesis and RUNX2 has a role in it. This suggests a role for RUNX2 in the expression of RANKL.
RUNX proteins are expressed in prostate tissue and prostate cancer cells[18, 55, 56]. Breast and prostate cancers over expressing RUNX2 metastasized predominantly to bone[16, 20]. We have shown a direct relationship of CD44 expression with RUNX2 activation in androgen-independent PC3 cells. Knockdown of CD44 reduced the expression of RUNX2 at mRNA and protein levels and hence reduced RUNX2-mediated signaling. Our studies demonstrate the possible role of CD44 signaling in RUNX2-mediated expression of RANKL. One possible explanation for RUNX2-regulated RANKL expression in PC3 cells may be associated with the lack of androgen receptor signaling. Androgen receptor was shown to bind RUNX2 and abrogates its binding to DNA and possibly to other nuclear DNAs. It appears that CD44 expression in androgen-independent cells (e.g. PC3 cells) counteracts androgen receptor effects in terms of activation of RUNX2- mediated events. Therefore, knockdown of CD44 signaling in PC3 cells has the potential to reduce RUNX2 mediated signaling.
Hyaluronan (HA), the major non-protein glycosaminoglycan component of the extracellular matrix in mammalian bone marrow, functions in part through its receptor, CD44, to stimulate a series of intracellular signaling events that lead to RANKL expression. We have shown previously that osteopontin (OPN) is secreted by PC3 cells. Over-expression of OPN in PC3 cells increases the secretion of RANKL through αvβ3 signaling. Our current mechanistic evaluation studies in PC3 cells suggest a role for CD44 signaling in the phosphorylation of a RUNX2 and integrin αvβ3 signaling in the phosphorylation of Smad-5 independent of CD44 signaling. However, further studies are required to understand the precise contribution of downstream kinase(s) to the regulation of RUNX2 phosphorylation.
Runx2 nuclear localization was found to be up-regulated in prostate cancer and was suggested that this could be used as a predictor of metastasis in prostate cancer. Several studies have shown that RUNX2 regulates localization of activated Smads in the subnuclear loci[24, 58, 59]. RUNX2 cooperates with Smads to induce differentiation of osteoblasts[26, 60] and expression of collagenase in breast cancer cells. RUNX2 forms complexes with Smad proteins as a requirement for mediating BMP/TGF β responsiveness in tumor cells. These effects contribute to tumor growth in bone and the accompanying bone loss in metastatic breast cancer cells. Formation of the Runx2/Smad transcriptional complex is dependent on the phosphorylation state of these proteins. Likewise, we detected predominant localization of phosphorylated RUNX2 and Smad 5 in the nuclei of lysates made from PC3 cells, prostatic adenocarcinoma and in tissue microarray sections containing primary prostatic tumor (grade 2–4).
Distinct relationship has been shown to exist between each Smad and RUNX2,[26, 27, 58, 62, 63]. Not only Smad 5 but also Smads 2 and 3 were shown to physically interact with RUNX2 in P19 embryonic carcinoma cells. RUNX2/Smad 3 interaction stimulated collagen 3 expression in breast cancer cells. Runx2/Smad3 complex negatively regulated endogenous and TGF-beta-induced connective tissue growth factor gene expression in vascular smooth muscle cells. We have found that PC3 cells express Smad −2, -3 and −5 (Additional file2: Figure S2). Smad 5 interaction was more with RUNX2 and this interaction regulates the expression of RANKL in prostate cancer cells.
RUNX2/Smad complex was shown to regulate the expression of RANKL in osteoblasts. Although various studies have addressed the role of RUNX2 and Smad(s) in the regulation of expression of RANKL, the mechanisms underlying this process have remained largely unknown. Also the role of Smad5 in the expression of RANKL needs further elucidation. The data presented here show that Smad 5 and RUNX2 are co-immunoprecipitated in the nuclear fraction. RUNX2/Smad 5 complex regulates the expression of RANKL in PC3 cells. Interaction of RUNX2 with RANKL promoter was observed with CHIP assay. Binding of RUNX2 to the ctggaaccact ggagt motif site on the RANKL is shown by CHIP assay. Although knockdown of RUNX2 or inhibition of phosphorylation of Smad-5 by an inhibitor to αv reduces the levels of RANKL, direct binding of Smad 5 with RANKL promoter was not observed. Future studies should delineate the relevant interactions between these proteins.
Interestingly, we have also observed reduced levels of RUNX2 and RANKL expression in cells treated with an inhibitor to αv or SiRNA to Smad5. These results indicate that RUNX2 is a major target gene of CD44 and Smad 5 signaling pathway. This is consistence with observations shown by others that Smad 5 is an upstream regulator of RUNX2[26, 51, 60]. Over expression of Smad 5 increases RUNX2 levels in human MG63 osteosarcoma cells. RUNX2 expression is transiently up regulated by TGF-β and BMP-2 activated Smads in mesenchymal precursor cell differentiation[26, 60]. Smad 2 and 3 are expressed in PC3 cells; however, these proteins could not compensate the function of Smad 5. Therefore, it is possible that, a) Smad 5 which induces RUNX2 expression might also be translocated to subnuclear loci by RUNX2; b) Smad 2 or 3 interaction with RUNX2 may not occur for RANKL expression in response to integrin αvβ3 signaling. BMP2 signaling contributes to the high level of Runx2-Smad interaction which activates RANKL in osteoblasts. CD44/Smad signaling pathway has been shown to have a regulatory role in osteoblast differentiation in the absence of BMPs. The underlying molecular mechanism by which αvβ3-activated Smad 5 regulates RUNX2 expression needs further elucidation. Taken together, bone metastatic prostate cancer cells (PC3) are osteomimetic and are expressing genes and proteins as observed in osteoblasts. However, the expression of osteoblastic specific genes in metastatic cancer cells does not necessarily involve the same pathway as observed in osteoblasts.