Tumor-stroma relationships: who's the driver?

Cancer is characterized by somatic mutations that provide a growth and survival advantage. The mutations are a function of chance because of the random nature of mutagenesis. However, the biology in the tumor cells drives the selection of mutations that are advantageous to the cancer. In addition, the tumor environment exerts a selective pressure on the tumor. For example, the hypoxic environment of a tumor induces HIF, stimulating VEGF production, which signals to the non-tumor endothelial cells to stimulate angiogenesis. Witkiewicz et al. have assessed the association of another non-tumor marker, MCT4, in triple-negative breast cancer (TNBC).

Triple-negative breast cancers are estrogen and progesterone receptor-negative and HER-2-negative and account for 10–20% of all breast cancers. Standard treatment is surgery with radiotherapy and adjuvant chemotherapy, sometimes with biologic agents. Although TNBCs are generally very susceptible to chemotherapy, they are often associated with a shorter median time to relapse and early death. One important goal is to identify prognostic biomarkers to reliably select high- and low-risk subsets of TNBC.

Prognostic biomarkers do not have to be limited to the tumor cell. The ratio of tumor to stroma in TNBC is a predictor of outcome; tumors with < 50% stroma have a 5‑year progression-free survival and overall survival of 85% and 89% compared with 45% and 65% in tumors with > 50% stroma. Does the biology of the stroma predict response in TNBC? Loss of stromal caveolin-1 (Cav-1) is predictive in TNBCs; 75.5% of patients are alive at 5 y with high stromal Cav-1 compared with 9.4% of patients with low stromal Cav-1. There is some controversy; not all data supports the prognostic association, with arguments that Cav-1 is a tumor suppressor in some situations and an oncogene in other settings. Loss of Cav-1 increases the levels of enzymes in glycolysis and lactate dehydrogenase. Hence, the intracellular levels of lactate could be elevated in Cav-1-null stromal cells, and the level of the lactate transporter MCT4 could be increased to maintain intracellular pH. An inverse correlation between Cav-1 and MCT4 expression is a plausible hypothesis. Witkiewicz et al. showed high stromal MCT4 predicts for poor outcome in TNBC. Combining the data for both Cav-1 and MCT4 improved the prognostic power of the data, certainly for the intermediate risk groups.

MCT4 is a plausible therapeutic target whether considering excessive production of lactate by stromal tissue or tumor cells. In either scenario, the metabolism of the tumor would be disturbed, causing intracellular acidosis if MCT4 is on the tumor cells and starving the tumor if MCT4 is on the stromal cells. Would patients with TNBC with Cav-1 low/MCT4 high stromal tissue elicit a clinical benefit from decreasing the availability of lactate? Possibly. However, the tumor metabolism might adapt to use alternative sources of carbon. On the upside, the acidosis in the MCT4-positive stromal cells might kill the stromal cells, lowering the stroma/tumor ratio and subsequently extending relapse-free survival.

The characterization of the cancer-stromal relationship is gaining momentum, from direct interactions, e.g., through integrins, to signaling molecules such as VEGF.. Metabolically, chronic lymphocytic leukemia cells can modulate their redox status through the synthesis of GSH from cysteine that is released by bone marrow stromal cells. The bone marrow cells are not bystanders but import cystine and convert it to cysteine that is exported to the microenvironment.

In conclusion, there is crosstalk at multiple levels between tumor cells and the stroma, providing a rationale to target the stroma. This is unlikely to be a panacea for treating cancer as shown by the inter- and intra-tumor-type heterogeneity that has been shown for tumor-stroma interactions. What will predict which tumors will respond to stroma directed therapies; the biology of the stroma, the biology and genetics of the tumor or both? As the volume of cancer genome data grows, we would be wise to consider how the genetics of tumors is molded by the microenvironment and vice versa, and how we might target the microenvironment for clinical benefit.