The RUNX genes have come to prominence recently because of their roles as essential regulators of cell fate in development and their paradoxical effects in cancer, in which they can function either as tumour-suppressor genes or dominant oncogenes according to context. How can this family of transcription factors have such an ambiguous role in cancer? How and where do these genes impinge on the pathways that regulate growth control and differentiation? And what is the evidence for a wider role for the RUNX genes in non-haematopoietic cancers?

The til-1 locus was identified as a common retroviral integration site in virus-accelerated lymphomas of CD2-myc transgenic mice. We now show that viral insertions at til-1 lead to transcriptional activation of PEBP2␣A (CBFA1), a transcription factor related to the Drosophila segmentation gene product, Runt. Insertions are upstream and in the opposite orientation to the gene and appear to activate a variant promoter that is normally silent in T cells. Activity of this promoter was detected in rodent osteogenic sarcoma cells and primary osteoblasts, implicating bone as the normal site of promoter activity. The isoforms encoded by the activated gene all encompass the conserved runt DNAbinding domain and share a novel N terminus different from the previously reported PEBP2␣A products. Minor products include isoforms with internal deletions due to exon skipping and a novel C-terminal domain unrelated to known runt domain factors. The major isoform expressed from the activated til-1 locus (G1) was found to account for virtually all of the core binding factor activity in nuclear extracts from its corresponding lymphoma cell line. Another member of this gene family, AML1(CBFA2), is well known for its involvement in human hemopoietic tumors. These results provide evidence of a direct oncogenic role for PEBP2␣A and indicate that the Myc and Runt family genes can cooperate in oncogenesis.

The runt related transcription factor CBFA1 (AML3/PEBP2alphaA/RUNX2) regulates expression of several bone- and cartilage-related genes and is required for bone formation in vivo. The gene regulatory mechanisms that control activation and repression of CBFA1 gene transcription during osteoblast differentiation and skeletal development are essential for proper execution of the osteogenic program. We have therefore defined functional contributions of 5' regulatory sequences conserved in rat, mouse and human CBFA1 genes to transcription. Deletion analysis reveals that 0.6 kB of the bone-related rat or mouse CBFA1 promoter (P1, MASNS protein isoform) is sufficient to confer transcriptional activation, and that there are multiple promoter domains which positively and negatively regulate transcription. Progressive deletion of promoter segments between nt -351 and -92 causes a striking 30- to 100-fold combined decrease in promoter activity. Additionally, 5' UTR sequences repress reporter gene transcription 2- to 3-fold. Our data demonstrate that CBFA1 is a principal DNA binding protein interacting with the 5' region of the CBFA1 gene in osseous cells, that there are at least three CBFA1 recognition motifs in the rat CBFA1 promoter, and that there are three tandemly repeated CBFA1 sites within the 5' UTR. We find that forced expression of CBFA1 protein downregulates CBFA1 promoter activity and that a single CBFA1 site is sufficient for transcriptional autosuppression. Thus, our data indicate that the CBFA1 gene is autoregulated in part by negative feedback on its own promoter to stringently control CBFA1 gene expression and function during bone formation.

The Runx genes present a challenge to the simple binary classification of cancer genes as oncogenes or tumor suppressors. There is evidence that loss of function of two of the three mammalian Runx genes promotes cancer, but in a highly lineage-restricted manner. In human leukemias, the RUNX1 gene is involved in various chromosomal translocation events that create oncogenic fusion proteins, at least some of which appear to function as dominant-negative inhibitors of the normal gene product. Paradoxically, evidence is mounting that structurally intact Runx genes are also oncogenic when overexpressed. All the three murine genes act as targets for transcriptional activation by retroviral insertional mutagenesis, and the oncogenic potential of Runx2 has been confirmed in transgenic mice. Moreover, the RUNX1 gene is often amplified or overexpressed in cases of acute leukemia. The state of progress in elucidating the oncogenic roles of the Runx genes is the subject of this review, and we draw together recent observations in a tentative model for the effects of Runx deregulation on hematopoietic cell differentiation. We suggest that lineage-specific factors determine the sensitivity to the oncogenic effects of loss or overexpression of Runx factors.

Members of the Runx and MYC families have been implicated as collaborating oncogenes. The mechanism of this potent collaboration is elucidated in this study of Runx2/MYC mice. As shown previously, ectopic expression of Runx2 in the thymus leads to a preneoplastic state defined by an accumulation of cells with an immature phenotype and a low proliferative rate. We now show that c-MYC overexpression is sufficient to rescue proliferation and to release the differentiation block imposed by Runx2. Analysis of Runx2-expressing lymphomas reveals a consistently low rate of apoptosis, in contrast to lymphomas of MYC mice which are often highly apoptotic. The low apoptosis phenotype is dominant in Runx2/ MYC tumors, indicating that Runx2 confers a potent survival advantage to MYC-expressing tumor cells. The role of the p53 pathway in Runx2/MYC tumors was explored on a p53 heterozygote background. Surprisingly, functional p53 was retained in vivo, even after transplantation, whereas explanted tumor cells displayed rapid allele loss in vitro. Our results show that Runx2 and MYC overcome distinct ''fail-safe'' responses and that their selection as collaborating genes is due to their ability to neutralize each other's negative growth effect. Furthermore, the Runx2/MYC combination overcomes the requirement for genetic inactivation of the p53 pathway in vivo. (Cancer Res 2006; 66(4): 2195-201)