Towards a complete picture of cancer development and progression, a comprehensive delineation of underlying sources contributing to the manifestation of malignant phenotypes and their heterogeneity, is critical. In previous studies we have uniquely identified intrinsic phenotypes reflecting histologic tumor variability but independent of contextual influences, such as stromal induction 
. By employing experimental coculture in conjunction with novel primary tumor cell lines, here our approach has identified FTExT, a programmed response to stromally secreted factors. Instead of a stromal impact that uniformly promotes or inhibits all tumors, our data demonstrate the significance of variable epithelial responsiveness in disease outcome. Together with correlative analyses of clinical tissue, such focused yet broad-scope investigative strategies to define the functional framework of cancer phenotypes could substantially improve the range and accuracy of druggable target identification.
The FTExT profile generated through experimental coculture portrays stromal regulation of major biological programs, suggesting a key role for stroma-assisted promotion of early hallmarks of cancer encompassing evasion of regulatory circuits, such as immune surveillance, oxidative stress, and metabolic deficiency 
, in addition to the hallmarks of advanced cancer 
. Consequently, the FTExT expression profile was found to be associated with a poor clinical outcome in two independent datasets, representing over 600 cases. Consistent with this correlative clinical data, we have further demonstrated, that the downstream functional effects of stromally induced transcriptional changes were dramatically variable between primary breast tumor cells of divergent histopathologic derivation. In aggressive breast cancer cells, tumor-promoting effects were evident as an increased ability for oxidative stress neutralization, apoptosis evasion, and rapid cell cycling. In contrast, non-aggressive cells displayed the consequences of negative growth regulation or tumor inhibitory effects in response to paracrine interaction. It can thus be concluded that FTExT is a functionally validated surrogate for measuring the impact of stromal signaling on malignant epithelium within cancerous breast tissue. Based on a consistent pattern of responsiveness irrespective of the source of breast fibroblasts, our data suggest that contrasting responses of malignant cells are characteristically ‘hard wired’ into the tumor cell genome. We have previously demonstrated that 2/3 novel high grade cell lines in this study cluster with other well-known luminal cell lines 
. In contrast, our low and intermediate grade cell lines cluster apart from all other established cell lines, suggesting the possibility that the unique responsiveness of such novel breast cancer cell lines to stromal signaling reflect a distinct molecular makeup. Reflecting a model of ‘effective cooperativity’, neighboring stroma within malignant tissue plays the role of a key accomplice, but only to the tumor epithelium armed with significant biological derangement. Consequently, treatment strategies could be better informed by the early detection of those responsiveness patterns wherein stromal interaction promotes tumor survival and growth at the expense of the host.
In terms of relevance to histologic grade, this is the first report of a functionally mediated, stromally induced molecular component associated with global prognostic profiles of grade 2 and 3 tumors. Grade is an important histologic determinant of patient management. More recently, in attempts to eliminate the subjectivity underlying its histologic assignment, a molecular grade gene index has been implemented 
based on genes, which are predominantly associated with tumor cell proliferation 
. The distinctive components of the FTExT signature, and their role in distinguishing those grade 2 and 3 tumors that do not confer an expected poor outcome, could further assist in improving the reliability of tumor grading and prognostic assessment. Our model maintains that in indolent breast tumors, where differentiation signaling is relatively intact, stromal fibroblasts and malignant cells are not engaged in growth promoting ‘cooperation’. On the other hand, the loss of responsiveness to differentiation cues in grade 3 and some grade 2 lesions induces a dramatic shift in the epithelial-stromal dialog towards tumor promotion. It could be argued that generally at advanced stages, grade 3 cancer displays low stromal cellularity, and therefore crosstalk between malignant cells and other components of the tumor, is marginal. In such tumor cells, repression of FTExT genes might be an inherent phenotype independent of stromal signaling. While this possibility cannot be categorically eliminated at this time, our experimental data demonstrating that cells derived from grade 3 tumors harboring phenotypes of poor prognosis 
are indeed promoted further in their aggressive attributes through paracrine interaction, are strongly supportive of a stromal role. In an alternate scenario, the molecular profile characteristic of high histologic grade is established early in tumorigenesis partially through stromal interaction, and maintained thereafter. An example closely supporting our model of interactive cooperativity is that of dermal fibroblasts, which repressed the growth of non-metastatic melanoma lesions while proliferation of metastatically competent melanoma cells was stimulated in their presence 
. In this light, it could be speculated that FTExT might play a role in the increased risks associated with mammographic breast density 
based on the rationale that overabundance of ‘cooperative’ fibroblasts associated with responsive premalignant or malignant cells leads to more aggressive lesions. This is also consistent with the finding that carcinomas that exhibit a greater stromal reaction manifested microscopically as a “desmoplastic response” are associated with higher histologic grade and poor patient outcome 
Intriguingly, considerable variability in signal induction and/or elicited response appears to occur between cancer-associated fibroblasts from different tumor types emphasizing the importance of specificity in the crosstalk between stromal and malignant cells of diverse tissue origin. With regard to 26 probe sets common between the coculture transcriptional profiles in our study and myeloma cells cocultured with bone marrow derived stromal cells 
, 25 probe sets displayed suppression in the FTExT profile, but in remarkable contrast their expression was induced in myeloma cocultures. Such inter-study comparisons are invaluable and underscore the importance of well-defined model systems for the derivation of clinically relevant data for each tumor type. As a striking example of tissue specific effects, although loss of function for PTEN, a FTExT gene, is a well-known indicator of tumor aggressiveness 
, this carcinoma-associated tumor suppressor gene appears to be overexpressed in myeloma patients with poor clinical outcome 
. Similar contrasting patterns of expression were noted underlying the role of FTExT genes: CYBA, GNAI2, GRN, JAK3, TRADD, BCL2A1, BPGM, and CD48 in carcinomas of the breast and other tissues, in comparison to myeloma. However, both in myeloma as well as in breast cancer, signatures portraying altered functionality due to paracrine exchange between stromal and tumor cells, were characteristic of poor clinical outcome, suggesting similar underlying pathways albeit transcriptionally distinctive. The fact that although stroma of different tissues regulate a closely similar set of genes, but in opposing directions in two tumor types, might reflect contrasting functions of their precursor cell types - epithelium vs. B lymphocytes, in this case. This is a critical finding in and of itself, emphasizing careful consideration prior to data extrapolation from one cellular subset or tissue type to another. Thus caution must be exercised in employing irrelevant cell types and their modulating microenvironment in highly cost-intensive drug discovery efforts. In a similar vein, it is notable that the FTExT signature includes several genes characteristic of immune response. While the induction of these genes in tumor tissue confers a favorable prognosis in breast cancer patients 
, it is consistent that their repression would have the reverse effect as demonstrated by our data and by others 
Our approach complements profiling strategies for identifying the biological basis of the clinical heterogeneity of human malignancy in general, and breast cancer in particular, but differs in the striking innovation of engaging functional cell models spanning the clinicopathologic spectrum of this disease. The need for hypothesis validation through gene-by-gene manipulation of cellular and rodent models, severely limited in the representation of the full range of human cancer pathology, is thus circumvented. Portrayed as dynamic paracrine interactions, which lead to the clinical, molecular, and functional heterogeneity of breast cancer, our data present novel opportunities for tumor targeting. Thus, therapies that enable the disruption of cancer-promoting interactions in the malignant microenvironment could serve as an important adjunct to current approaches, which only target malignant subpopulations to a variable degree. The observation that secreted stromal factors are remarkably effective in modulating tumor phenotypes brings a long desired goal in the field closer to fruition, whereby the detection of tumor associated markers in body fluids could be facilitated in the future. Concerted efforts towards the identification of stromal signals that fuel dramatic changes represented by the FTExT profiles of tumor cells should be a priority for impeding such crosstalk in high risk tumors, such as triple-negative breast cancer. In this regard, previous investigations of microdissected fibroblasts and tumor cells within archived breast tumor tissue have indeed demonstrated characteristic features of stromal and epithelial cell lineages 
, but not distinguished between inherent and induced characteristics of each subset, thus only partially addressing their respective role in the overall tumor phenotype. On the other hand, an inducible model of squamous cell carcinoma highlighted the importance of tumor-stroma interaction and identified the temporal induction of ß1 integrin as a potential nexus involved in Ras-driven tumor progression 
, but functional dissection of inter-tumor heterogeneity was not a major goal. Among those that could benefit considerably from targets manifested early in tumorigenesis are women that harbor DCIS, currently the most common diagnosis in the Western world associated with breast related symptoms 
. Since DCIS is a preinvasive lesion, and often indolent, it presents the clinician with the dilemma of distinguishing between those that will recur with invasive disease, thus requiring immediate aggressive therapeutic measures, from those that will not. As suggested by the FTExT signature, stromally induced cues might promote the release of in situ
tumor cells from the confines of the basement membrane, as well as support subsequent steps in tumor survival and expansion. Thus, detection of FTExT-related aberrations within diagnostic biopsies could improve clinical decision-making, and offer clear opportunities for personalized patient management.