In the previous issue of Breast Cancer Research Xiao-Jun Ma and colleagues of the Dennis C Sgroi laboratory report on a comparative analysis of global gene expression in the microenvironment of ductal carcinoma as it relates to the acquisition of invasiveness [1]. The authors utilize an approach that significantly reduces confounding signals in the data due to inter-patient variability by performing laser capture on breast tissue from 14 breast cancer patients with matched normal and tumor samples. The majority of the cohort was pre-menopausal (mean age 41 years), with estrogen receptor positive tumors (78%) and ipsilateral nodal involvement (78%). Thus, this study focuses on young women's breast cancer, an important but understudied subset that is associated with poor prognosis [2, 3].

Previous work from this group showed that gene expression changes in epithelium occur prior to local invasion, as early as atypical ductal hyperplasia [2]. Surprisingly, no major changes in epithelial gene expression were identified between in situ and invasive cancers. The authors now confirm these earlier results and demonstrate that, like the epithelium, tumor stroma undergoes dramatic gene expression alterations early, at the transition from normal to ductal carcinoma in situ (DCIS). Over 3,500 stromal genes were identified to be differentially regulated in DCIS compared to normal, whereas only 305 stromal genes were differentially regulated with the transition to invasion. These observations highlight early events that occur prior to histological evidence of local invasion, and the co-dependency between epithelial cells and the stroma, in regulating tumor progression.

Another important contribution of this work is the identification of a stromal gene signature that marks the transition to local invasion. While only three epithelial genes were identified as differentially regulated between DCIS and locally invasive disease, 76 stromal genes were upregulated and 229 downregulated in invasive cancers compared to DCIS. These new data further challenge the epithelial cell-centric, multi-hit model of metastasis by demonstrating that it is a stromal gene signature, not an epithelial signature, that distinguishes pre-invasive from invasive cancer. In 2002 van 't Veer and colleagues [3] demonstrated the power of global gene profiling by showing for the first time that expression signatures of the primary tumor could predict risk of metastasis with high accuracy. The van 't Veer study fundamentally undermined the prevailing hypothesis that metastasis is solely mediated by the acquisition of rare and rate limiting mutations within a minority of tumor cells [4].

Subsequent studies have demonstrated that having a poor prognosis gene signature is not sufficient for metastasis, as many cancers with such signatures do not metastasize [5]. These studies and others helped re-awaken interest in the role of the tumor microenvironment in orchestrating the metastatic switch [6].

The paper by Xiao-Jun Ma and colleagues adds to an impressive list of recent gene expression studies designed to investigate the role of the tumor microenvironment in breast cancer [7–10]. Prior to the reporting of these gene profiling studies, a histologically defined fibrotic response referred to as tumor desmoplasia had been correlated with poor prognosis [11, 12]. In the study by Xiao-Jun Ma and colleagues, the short chain collagen types X and VIII and fibronectin were upregulated four- to six-fold in stroma surrounding DCIS compared to adjacent normal stroma, implicating these extracellular matrix (ECM) proteins in cancer progression. Pre-clinical studies strongly support the hypothesis that collagens and fibronectin directly promote metastatic phenotypes in cancer cells. For example, fibronectin signaling through β 1 integrin was recently found to cause loss of tumor cell quiescence via phosphorylation of myosin light chain and cytoskeletal reorganization [13]. Thus, it was somewhat surprising when two independent studies, utilizing gene profiles identified from very disparate tissue samples, reported that ECM proteins associated with the fibrotic response, including fibronectin, collagens I, III, and VI, biglycan, fibulin 2, SPARC, and connective tissue growth factor were associated with improved survival [10, 14]. Resolution of these apparent contradictory observations will require significant advancement in the ECM field, including additional functional characterization of the ECM constituents, such as the 26 genetically distinct collagen types identified to date. Improvements in our understanding of matrix assembly, cross-linking, signaling and turnover will be required as well.

Insight into why a tumor fibrotic ECM signature could have dual effects on survival prognosis may be provided by the list of stromal genes upregulated at the DCIS to invasion transition, as reported by Xiao-Jun Ma and colleagues. This transition was accompanied by increased stromal expression of the matrix metalloproteinases MMP-2, MMP-11 and MMP-14. Further, these authors found that stroma of invasive tumors was associated with a strong immune response signature, consistent with a large body of literature revealing multiple mechanisms by which immune cells positively regulate tumor progression [15]. Given that fragments of ECM proteins are key attractants and activators of immune cells under pathological conditions, differential proteolysis of tumor ECM and subsequent recruitment of immune cells may discriminate between tumor fibrosis associated with good prognosis from poor prognosis. Global gene expression profiling studies have helped formulate these exciting hypotheses and will no doubt continue to challenge our assumptions about breast cancer initiation, invasion and metastasis, permitting us to make inroads into this disease in ways previously unimagined.