IntroductionA growing body of evidence supports a role for members of the mammalian Hox homeobox gene family of transcription factors in the regulation of hematopoiesis. 1 In a subgroup of human myeloid leukemias, the HOXA9 gene is involved in a recurrent translocation between chromosomes 7 (NUP98) and 11 (HOXA9), resulting in a fusion transcript that encodes the NUP98-HOXA9 oncoprotein. 2,3 Recently, the expression of HOXA9 was also shown to be the single most highly correlating factor (of 6817 genes tested) for poor prognosis in human acute myeloid leukemia (AML), 4 possibly indicating a broader role for this gene in human leukemia beyond that caused by its chromosomal translocations.The leukemogenic potential of Hoxa9 was directly demonstrated by the development of AML in mouse bone marrow transplantation chimeras that received a graft of primitive hematopoietic cells engineered by retroviral gene transfer to overexpress Hoxa9. 5,6 Although the latency for the development of AML in these chimeras is much shorter than reported for other similar Hox gene chimeras (ie, Hoxa10, Hoxb3, Hoxb8), overexpression of Hoxa9 alone is not sufficient to induce leukemia, and additional somatic mutation(s) are needed. [6][7][8][9] In leukemic transformation, genetic interactions have been demonstrated between Hoxa9 and molecules such as Meis1 and E2A-PBX1, 5,6 both of which can bind DNA cooperatively with Hoxa9. 10,11 This suggests the presence of a leukemogenic complex, which includes Hoxa9 and potentially some of its DNA-binding cofactors.In normal human CD34 ϩ bone marrow cells, HOXA9 is preferentially expressed in subfractions that are highly enriched for primitive long-term culture-initiating cells or myeloid progenitor cells, and its expression is significantly reduced in the erythroid lineage. 12,13 This pattern of Hoxa9 expression is conserved in murine hematopoietic cells, with high levels detected in the Sca-1 ϩ lineage-negative (LinϪ) bone marrow subpopulation that is enriched for primitive hematopoietic cells, whereas much lower levels are present in the progenitor-depleted fraction (eg, Sca-1 Ϫ Lin ϩ ). 14,15 Hoxa9 is also expressed in purified subpopulations of B and T lymphoid cells, although at levels much lower than those found in primitive bone marrow cells. 14 Analysis of the Hoxa9 null mutant mice showed that lack of Hoxa9 caused a significant decrease in bone marrow myeloid and pre-B-cell progenitors, and Hoxa9 Ϫ/Ϫ myeloid progenitors have a blunted response to granulocyte colony-stimulating factor, a growth factor known to enhance the production and release of granulocytes from the bone marrow. 14 Although adult thymocyte development was only mildly abnormal in Hoxa9 Ϫ/Ϫ mice, the fetal thymus cellularity was reduced 5-to 10-fold, due in part to a delay in thymocyte progression from immature single-positive (T-cell receptor-negative) to double-positive (ie, CD4 ϩ CD8 ϩ ) T cells. 14,16 In contrast to lymphoid and myeloid progenitor cells in the Hoxa9 Ϫ/Ϫ mice, no significant decrease was detected in more...

Tumor-suppressor genes (TSGs) have been classically defined as genes whose loss of function in tumor cells contributes to the formation and/or maintenance of the tumor phenotype. TSGs containing nonsense mutations may not be expressed because of nonsense-mediated RNA decay (NMD). We combined inhibition of the NMD process, which clears transcripts that contain nonsense mutations, with the application of high-density single-nucleotide polymorphism arrays analysis to discriminate allelic content in order to identify candidate TSGs in five breast cancer cell lines. We identified ARID1A as a target of NMD in the T47D breast cancer cell line, likely as a consequence of a mutation in exon-9, which introduces a premature stop codon at position Q944. ARID1A encodes a human homolog of yeast SWI1, which is an integral member of the hSWI/SNF complex, an ATPdependent, chromatin-remodeling, multiple-subunit enzyme. Although we did not find any somatic mutations in 11 breast tumors, which show DNA copy-number loss at the 1p36 locus adjacent to ARID1A, we show that low ARID1A RNA or nuclear protein expression is associated with more aggressive breast cancer phenotypes, such as high tumor grade, in two independent cohorts of over 200 human breast cancer cases each. We also found that low ARID1A nuclear expression becomes more prevalent during the later stages of breast tumor progression. Finally, we found that ARID1A reexpression in the T47D cell line results in significant inhibition of colony formation in soft agar. These results suggest that ARID1A may be a candidate TSG in breast cancer.