Analysis of Stromal Gene Expression for the Identification of Prognostic and Predictive Molecular Markers in Cancer Therapy

In all epithelial tumors, cancer cells coexist with other cell types that together make up the tumor-associated stroma. These stromal cells include fibroblasts, inflammatory cells, and vascular cells. Cancer cells functionally interact with stromal cells and also define their phenotype through soluble factors, like cytokines and chemokines, leading to tumor-stroma interactions that involve complex feedback loops between these cells. It is thought that these interactions create an environment that principally supports tumor progression. Stromal cells are also actively recruited by cancer cells, and this recruitment of stromal cells into solid tumor is essential for tumor growth^1–3. Vascular cells are required for neo-vascularization and formation of microvessels, which allows tumors to grow while the microvessels can become highways for inflammatory cells to get into the tumors and for cancer cells to escape the tumor mass and metastasize into distant organs. Tumor-associated fibroblasts have diverse functions and are essential for tissue restoration processes in the expanding tumor, commonly referred to as the wound healing response of tumors, and they participate in tumor growth by providing the extracellular matrix structure within the tumor and by influencing cancer cell differentiation. Fibroblasts can induce what is known as the epithelial to mesenchymal transition or transformation of cancer cells, a phenotype that leads to increased mobility and an enhanced metastatic potential of the transformed cells. In addition, cancer cells actively recruit inflammatory cells and modify their phenotype to foster tumor development. Tumor-associated macrophages and the reprogramming of cytotoxic T cells into regulatory T cells are examples of a cancer cell-induced immune microenvironment where re-programmed immune cells impede the development of an anti-tumor immunity and secrete pro-inflammatory mediators, leading to a metastatic cancer phenotype, increased tumor angiogenesis, and inhibition of the anti-tumor activity of chemotherapeutics, radiation, and immune therapies^4,5.

The tumor-associated stroma frequently accounts for a large fraction of the tumor mass and can largely determine the phenotype of a tumor. Accordingly, distinct molecular features of stromal cells and stroma-related gene signatures were found to define the aggressiveness of human tumors⁶. For example, stromal gene signatures predict clinical outcome and resistance to therapy in breast cancer^7,8. Other studies revealed that fibroblast-derived transcriptional signatures are associated with cancer progression and poor outcome in human breast and lung cancer^9,10. These findings highlight the importance of stromal biology and specifically the process of fibroblast-induced wound healing in cancer biology, disease aggressiveness, and response to therapy. Thus, the study of the tumor stroma has the promise to advance discovery of prognostic and predictive molecular markers for patient management and cancer therapy.

In this issue, Edlund and coworkers describe stromal CD99 expression as a novel prognostic marker in human non-small cell lung cancer (NSCLC). CD99 is a glycoprotein encoded by the MIC2 gene and is best known as a leukemia antigen and a diagnostic marker for Ewing’s sarcoma. It is expressed by many hematopoitic cells. In NSCLC, CD99 is expressed by stromal cells including fibroblasts and tumor-infiltrating lymphocytes. The prognostic significance of CD99 in NSCLC was discovered when Edlund et al. examined the global gene expression of laser capture microdissected (LCM) tumor epithelium and compared the obtained profiles with those of bulk tumor tissue samples containing both tumor epithelium and tumor-associated stroma. This comparison yielded a short list of 200 candidate stroma-associated genes that were differently expressed between the LCM-obtained and bulk tumor samples. Further mining of Human Protein Atlas expressing data and immunohistochemical analysis of a NSCLC tissue microarray showed that 12 candidates were expressed in NSCLC stroma. Of those, one gene, CD99, was found to be independently associated with increased long-term survival of lung cancer patients, which was corroborated by the analysis of a second dataset. The approach in this study was different from those taken by others who directly analyzed LCM-obtained stroma^6,7,11. However, stroma without contamination by cancer cells can be difficult to obtain and the research approach chosen by Edlund et al. may yield findings that are comparable to those from directly analyzing stroma samples.

How does increased stromal CD99 expression extend lung cancer survival? The authors did not examine the underlying mechanism(s) by which CD99 expression may influence disease aggressiveness and cause increased long-term survival of NSCLC patients, which is a limitation of the study. Others reported that CD99 is a cell adhesion molecule and its expression inhibits proliferation and migration of various human cancer cell lines^12,13, consistent with a tumor suppressor function of CD99. These studies were initiated because decreased CD99 immunoreactivity in cancer cells of osteosarcoma, pulmonary carcinoid tumors, and gastric adenocarcinoma was found to be associated with lymph node metastasis and poor survival^12–14. We can only speculate that CD99 expression in stromal cells may have effects similar to those observed in human cancer cell lines. Alternatively, CD99 expression in stromal cells may alter tumor-stroma interactions that are based on cell-cell contacts between stromal cells and cancer cells, or may affect the architecture of the stroma and for example inhibit the process of neo-angiogenesis. Despite this uncertainty, the literature reports are consistent in their findings that CD99 is a candidate tumor suppressor in a broad range of human tumors and low tumor CD99 expression is a marker of poor outcome.

While there is a myriad of studies in the peer-reviewed literature that have examined tumor epithelial molecular markers for prognosis and guidance of cancer therapy, only few reports have focused on the analysis of the tumor-associated stroma for the identification of prognostic and predictive markers in cancer therapy. This is at least partly explained by the difficulty to obtain pure stroma samples and also by the fact that the tumor-associated stroma is made up of several different cell types whose relative abundance in the stroma can greatly change within a tumor and from one patient to another. Nevertheless, several recent publications and the study by Edlund and coworkers have shown that the analysis of the stroma is technically feasible and can excel molecular marker discovery for future clinical application. Given the importance of stromal cells in cancer progression, it is crucial that future studies more rigorously examine the contribution of stromal biology to cancer outcomes and resistance to therapy.