Ras–dependent Paracrine Cascades
, by Kenneth P. Olive
Dr. Kenneth Olive is Assistant Professor at Columbia University and a member of the Herbert Irving Comprehensive Cancer Center. He trained with Tyler Jacks at MIT and David Tuveson at UPenn and the University of Cambridge. His lab uses mouse models and human clinical samples to understand, image, and treat pancreatic cancer.
One of the prevailing mysteries of Ras is the disconnect between the known cell autonomous effectors of mutant Ras and the cascade of non-cell autonomous changes it induces in the local microenvironment. Despite outstanding progress by the Ras community in delineating the proximal pathways of Ras signaling, there are few secreted or cell-surface proteins represented on the resulting maps (including Ras 2.0). Yet alterations in the tumor microenvironment may be one of the most profound consequences of Ras activation. The cancers with a high frequency of Ras mutations (e.g. pancreatic, lung, and colon) are characterized by extensive desmoplasia and stromal cell infiltration, and many studies now find associations between Ras mutation and the activation of specific paracrine pathways that effect biological responses in stromal cells. Genetically engineered mouse models have been particularly useful in characterizing the paracrine effects of mutant Ras expression. A pancreatic cancer model based on tetracycline-inducible mutant K-ras demonstrated rapid stromal regression following withdrawal of mutant K-ras expression.1 However, detailed molecular mechanisms linking direct Ras effector proteins to secreted/cell surface ligands remain elusive.
Some of the most direct evidence of Ras–dependent paracrine signaling comes from studies of tumor development. Many inflammatory conditions predispose to related cancers (eg. chronic pancreatitis is a strong risk factor for pancreatic ductal adenocarcinoma), yet acute inflammation is part of the normal, healthy response to tissue injury. An overly festive night out might injure some acinar cells in the pancreas, resulting in acinar-to-ductal metaplasia (ADM) and inflammation. Normally, after the tissue is repaired, ADM cells return to an acinar state and the acute inflammation resolves. However, in the context of a K-ras mutation, this cycle is broken and the tissue is unable to appropriately resolve the inflammatory response, leading to chronic inflammation and increasing cancer risk.2
Normal epithelial (glandular) tissues have a limited complement of stromal cells. For example, fibroblasts make up just a few percent of cells in the normal pancreas. However, in the setting of chronic inflammation, and later in cancer desmoplasia, fibroblasts and other stromal cell types can dominate the tissue parenchyma. A host of cytokines, chemokines, and other factors are secreted, leading to changes in cellular composition, remodeling of the extracellular matrix, increased mechanical stiffness of the tissue parenchyma, and modulation of the local immune microenvironment, all of which can promote malignancy. Another consequence of desmoplasia can be changes in tissue perfusion that limit the delivery of drugs to tumor tissues, reducing drug efficacy.3,4
Despite the obvious importance of these processes, we actually have a shockingly incomplete understanding of the cellular and extracellular composition of desmoplastic tumors. Fibroblasts, immunocytes, and endothelial cells are clearly major contributors, but these are all broad categories. How many different types of fibroblasts exist in a tumor? How do we visualize them? What are their functional distinctions? The question is even more complicated for immunocytes, a huge category comprising dozens of specific cell types with diverse and often opposing functional roles. Endothelial and lymphendothelial cells can vary in type and form, yielding diverse structures that control perfusion, oxygen saturation, nutrient levels, and other biophysical properties. There are also “immature” cell types such as mesenchymal stem cells, immature myeloid progenitor cells, and endothelial progenitor cells. The word “immature” is a diminutive and often equated to “non-functional”, but in truth such cells are associated with critical functions, as in the case of myeloid derived suppressor cells, which help mediate local immunosuppression. Moreover, there are cell types that exist outside the tumor mass that probably affect biology within the tumor. Tertiary lymph nodes can form in proximity to some tumors, perhaps reflecting a failed or coopted immune response against the tumor. Nerve cells project axonal processes into GI and other cancers, prostate, gastric, and pancreatic tumors, and make direct connections to specific subsets of malignant epithelial cells and potently influencing tumor biology.5 Even within the malignant compartment, there is significant heterogeneity of differentiation states or progenitor identities. What is clear is that existing cell-specific markers fail to fully capture the heterogeneity of cellular composition within tumors and that even highly multiplexed flow cytometry or immunofluorescent technologies still struggle to provide an unbiased and complete representation of tissue composition. Advances in single-cell sequencing and CyTOF technologies6 may provide tractable means to properly quantify the diversity of cell types that contribute to even the most complex of solid tumors.
The stroma regression observed in the inducible K-ras model6 indicates that mutant Ras plays a major role in orchestrating this stromal response, but the mechanisms are limited to vignettes at this point. For example, K-ras mutation in the pancreas is associated with an increase in the expression of the Sonic Hedgehog ligand, which induces downstream pathway activity in certain fibroblasts to increase cell proliferation and collagen deposition.7,8 Similarly, K-ras mutation is also associated with the overexpression of GM-CSF, which promotes the recruitment of immunosuppressive myeloid lineages.9 In the latter case, inhibitors were used to ascertain that K-ras mediated induction of GM-CSF is dependent on both MAPK and PI3K, but the full mechanism leading to its elevated expression is cryptic. It is unclear whether Ras-dependent paracrine signals from the malignant cell can directly control all of the different cell types in the stroma. However, it is apparent that secondary signals propagate between different stromal cell types, indirectly responding to cues from Ras. One example of such a “paracrine cascade” is the indirect regulation of CD8+ T-cells by Gr-1+, CD11b+ myeloid cells responding to the K-ras dependent induction of GM-CSF.10
I expect that multicellular paracrine cascades are going to be the rule, rather than the exception, in the tumor microenvironment. This brings up a critical issue: the paucity of model systems for studying complex, multicellular interactions. Reconstitution/mixing models are getting more sophisticated, but they still are largely limited to two cell types, perhaps three. How can we extend this to 5, 10, or 20 different cell types? What are the sources of cells used for such assays and do they truly reflect the cognate cells derived from a human tumor? Moreover, I am convinced that spatial geometry is critical. Anyone who stares at cell-specific immunohistochemical stains of tumor sections will notice that cells are not randomly scattered throughout the tissue. Different cell types exist in specific locations relative to malignant cells, blood vessels, etc. A recent example illustrates how different subtypes of fibroblasts are located at different distances from malignant pancreatic tumor cells.11 Certainly authochthonous mouse models are a great tools, but manipulating individual cell types is a protracted and tedious process. The use of mouse-derived organoids may offer another alternative, particularly when implanted into recipients that are genetically modified to modulate stromal cell populations. Another possibility is the use of tumor explant cultures, which harbor all of the cell types present in a tumor and may possibly be cultured for long enough to enable genetic or pharmacological manipulation of cell types. Finally, exciting new bioengineered models are being developed that take into account some of the biophysical parameters of native tumors.12 Innovations like these will be crucial for deconstructing the cascades of paracrine signaling that propagate through the stroma of tumors in response to K-ras mutations.