Global gene expression analysis of canine osteosarcoma stem cells reveals a novel role for COX-2 in tumour initiation

Abstract

Osteosarcoma is the most common primary bone tumour of both children and dogs. It is an aggressive tumour in both species with a rapid clinical course leading ultimately to metastasis. In dogs and children distant metastasis occurs in >80% of individuals treated by surgery alone. Both canine and human osteosarcoma has been shown to contain a sub-population of cancer stem cells (CSCs), which may drive tumour growth, recurrence and metastasis, suggesting that naturally occurring canine osteosarcoma could act as a preclinical model for the human disease. Here we report the successful isolation of CSCs from primary canine osteosarcoma, as well as established cell lines. We show that these cells can form tumourspheres, and demonstrate relative resistance to chemotherapy. We demonstrate similar results for the human osteosarcma cell lines, U2OS and SAOS2. Utilizing the Affymetrix canine microarray, we are able to definitively show that there are significant differences in global gene expression profiles of isolated osteosarcoma stem cells and the daughter adherent cells. We identified 13,221 significant differences (p = 0.05), and significantly, COX-2 was expressed 141-fold more in CSC spheres than daughter adherent cells. To study the role of COX-2 expression in CSCs we utilized the COX-2 inhibitors meloxicam and mavacoxib. We found that COX-2 inhibition had no effect on CSC growth, or resistance to chemotherapy. However inhibition of COX-2 in daughter cells prevented sphere formation, indicating a potential significant role for COX-2 in tumour initiation.

Figure 1. Characterisation of osteosarcoma stem cells.

Spheres can be isolated from canine KTOSA5 cells (A), the human U2OS cell line (B), human SAOS2 cell line (C). A small population of CD34+ cells can be isolated from the KTOSA5 cell line by magnetic cell sorting. Only CD34+ cells express CD34 protein (D) and only CD34+ cells could form spheres (E). Data are representative of three independent experiments (*p<0.001). Reverse transcriptase (RT)-PCR analysis of Nanog, Oct4, STAT3, and β-actin gene expression levels (F). Western blots analysis of Fibronectin, β-catenin, Zeb1, and Vimentin, with β-actin as a loading control (G).

Figure 2. Cancer stem cells are resistant to the cytotoxic effects of doxorubicin.

Spheres were isolated from the canine osteosarcoma cell lines; KTOSA5 (A) and CSKOS (B), and the human osteosarcoma cell lines U2OS (C) and SAOS2 (D). Spheres and adherent cells were treated with the indicated doses of doxorubicin and cell viability was assayed 48 hr after treatment (* p<0.005).

Figure 3. Spheres are resistant to replicative cell death after doxorubicin treatment.

Colony forming ability after doxorubicin treatment was determined in KTOSA5 cells (* p = 0.008; ** p<0.001) (A) and U2OS cells (♮ p = 0.008; ♮♮ p<0.001) (B).

Figure 4. Cancer stem cells show an increased invasive potential in vitro.

Invasive ability of KTOSA5 (A, B) and U2OS (C, D) spheres and adherent cells was analysed using a collagen based invasion assay. Invading cells were quantified by measuring the optical density at 560 nm. * p<0.005.

Figure 5. Osteosarcoma spheres are enriched for higher tumourigenicity in vivo.

Disassociated spheres and adherent, of KTOSA5 cells (A) and CSKOS cells (B), were inoculated directly onto the chorioallantoic membrane of a chicken embryo at day 7 of development. All cells were fluorescently labelled and imaged 5 days after inoculation.

Figure 6. Gene expression analysis of canine osteosarcoma stem cells.

A three-dimensional representation of a principle component analysis of expression microarray data derived from KTOSA5 adherent cells, spheres and mesenchymal stem cells (MSC) (A). Heirarchial clustering analysis of the expression data (cut off p-value of 0.005). Expression values are represented by colours: blue squares represent low-expressed genes, red squares represent high-expressed genes (B). Biological process analysis of differentially expressed genes in KTOSA5 spheres compared to adherent cells (FDR = 0.005) (C).

Figure 7. Cancer stem cells express a higher level of COX-2.

Validation of microarray with qRT-PCR (A). Expression of COX-2 protein (B).

Figure 8. COX-2 inhibition has no effect on cell viablilty or chemo-resistance of cancer stem cells.

Dissociated KTOSA5 spheres and adherent cells (A), and CD34 sorted KTOSA5 cells (B) were treated with the indicated doses of meloxicam and cell viability was assayed 72 hr after treatment. KTOSA5 CD34- and CD34+ cells were treated with both indicated doses of meloxicam and 0.05 µM doxorubicin, cell viability was assayed 72 hr after treatment.

Figure 9. COX-2 inhibition suppresses sphere forming ability.

KTOSA cells were pre-treated for 24 hr with the indicated doses of meloxicam prior to assaying for sphere forming ability (* p<0.001) (A). KTOSA5 (B), CSKOS (C), U2OS (D) and SAOS2 (E) cells were pre-treated for 24 hr with the long-acting COX-2 inhibitor mavacoxib prior to assaying for sphere forming ability (* p<0.001).