Cell state plasticity, stem cells, EMT, and the generation of intra-tumoral heterogeneity

Abstract

Cellular heterogeneity in cancer represents a significant challenge. In order to develop effective and lasting therapies, it is essential to understand the source of this heterogeneity, and its role in tumor progression and therapy resistance. Here, we consider not only genetic and epigenetic mechanisms, but also inflammation and cell state reprogramming in creating tumor heterogeneity. We discuss similarities between normal mammary epithelial developmental states and various breast cancer molecular sub-types, and the cells that are thought to propagate them. We emphasize that while stem cell phenotypes and mesenchymal character have often been conflated, existing data suggest that the combination of intrinsic genetic and epigenetic changes, and microenvironmental influences generate multiple types of tumor propagating cells distinguishable by their positions along a continuum of epithelial to mesenchymal, stem to differentiated and embryonic to mature cell states. Consequently, in addition to the prospect of stem cell-directed tumor therapies, there is a need to understand interrelationships between stem cell, epithelial-mesenchymal, and tumor-associated reprogramming events to develop new therapies that mitigate cell state plasticity and minimize the evolution of tumor heterogeneity.

Fig. 1

Changes in mammary stem cell activity throughout development. a Shown is one view of the mouse mammary cell hierarchy, starting from fetal development. Beginning at embryonic day 16 (E16), mouse cells from fetal mammary rudiments exhibit the multipotency and self-renewal functions expected for bipotential mammary stem cells. Mammary stem cell frequency increases until birth and then decreases dramatically. Currently, it is not possible to say with confidence what the frequency of adult mammary stem cells is, or whether the adult mammary gland is maintained by lineage restricted myoepithelial and luminal stem cells. By analogy with other organs, tissue disruption, inflammation and oncogene activation may enable adult mammary cells to reacquire fetal-like plasticity. This may explain the differences in stem cell frequency measured in the adult using lineage tracing, which preserves tissue structure, and transplantation, which disrupts it. b Probabilistic representation of mammary development depicting likelihood of cells entering a multipotent (stem) state. Changes in stem cell activity in the mammary epithelium are accompanied by cellular changes in proliferation, epigenetics, signaling, and microenvironment as well as stage specific gene expression patterns. The peak of fetal mammary stem cell activity correlates with an increased number of cells that co-express myoepithelial and luminal differentiation markers (e.g., keratins 14 and 8, respectively) and transcriptional regulators (e.g. Gata3, Elf5, Sox9, Sox10, p63). Stem cell capacity and marker co-expression are lost as development progresses. c Genetic, epigenetic and microenvironmental factors may expand the cellular heterogeneity of the epithelium conferring adaptability and plasticity to the population. Mesenchymal character can be conferred to epithelial cells through transcriptional regulators such as Slug and Sox 10, and may depend on both the level and duration of their expression. Stem cell activity may also be increased by appropriate balance of mesenchymal and epithelial factors (such as Slug with Sox9), or Sox10 with other factors yet to be identified, or other balanced combinations of luminal and myoepithelial specifiers. Characteristics such as EMT, multi-lineage potential and self-renewal may make certain cells better able to adapt to stresses such as those encountered in transplantation, tumor progression, migration and metastasis, or drug challenge. The position of the normal basal (B) and Luminal (L) compartments along the mesenchymal (M) to epithelial (E) axis is shown

Fig. 2

Breast cancer cell lines show differential enrichment for fetal epithelial signatures (fMaSC) and fetal non-epithelial signatures (fStromal). Basal A Cell Lines, which are most similar to ‘Basal-like’ breast cancers, show enrichment for fMaSC signatures. Basal B cell lines, which have a more mesenchymal phenotype, exhibit classic BCSC markers and correspond to ‘Claudin Low’ tumors. Claudin low tumors are enriched for non-epithelial signatures and are anti-correlated to the fMaSC state. The SUM149PT cell line shows enrichment for non-epithelial signatures similar to other cell lines classified as Basal B in the Neve et al. data set, but shows less repression of the epithelial fMASC signature. This observation, and this cell line’s subsequent characterization as Basal-like as opposed to Claudin Low, may be attributable to an intermediate or heterogeneous phenotype among its constituent cells. Claudin low cell lines also show significant enrichment for the aMaSC signature (not shown). Assessment of signature enrichment was based on signatures from Spike et al. and methods described in Segal et al.

Fig. 3

Diverse states of the normal and neoplastic mammary epithelium relative to development, stem cell differentiation, and epithelial/mesenchymal phenotypes. a Changes in potential for cells to enter the stem cell state throughout development can be conceptualized in a coordinate space that relates stem cell abundance and differentiation (y-axis) with developmental progression (z-axis) and epithelial/mesenchymal phenotypes (x-axis). The various cell types within the mammary epithelium occupy distinct regions in this space over the course of development (as also proposed by Granit et al.). b The different intrinsic subtypes of breast cancer occupy some of the same space as the normal mammary epithelium leading to partially shared gene expression profiles and operative molecular mechanisms. We suggest using tumor propagating cell (TPC) as a general term to describe cells able to generate and propagate tumor xenografts. These cells are expected to exhibit cell state plasticity to generate intra-tumoral heterogeneity in either immune compromised mice when assaying human tumor cells, or immune intact mice when using isogenic mouse cells for analysis. It remains to be determined whether TPC initiate or propagate tumors in humans. TPC may either be mesenchymally oriented as are classic “cancer stem cells” such as BCSC, or they may resemble the more SLCC that share similarities with fMaSCs. CL Claudin Low, BL basal like, H2 Her2-like, LB luminal B, LA Luminal A. Inter-tumoral and intra-tumoral heterogeneity may further rest on local differences in expression of key regulators (e.g., Sox10, Cripto) – the expression and activity of which can be affected by autocrine or paracrine factors produced in the epithelium or microenvironment. Furthermore, genetic background and microenvironmental influences including immune responses and local and systemic stresses may impact aggressiveness by shifting the thresholds that distinguish stem cells from non-stem cells. c a model involving the generation of a tumor from a luminal progenitor. The Luminal progenitor already harbors some stem cell characteristics, such as cell state plasticity and a higher proliferative index that may facilitate epigenetic reprogramming to bipotentiality. Luminal progenitors appear to dedifferentiate into a bipotent-like state by activation of diverse oncogenic pathways and loss of tumor suppressors previously associated with BLBCs. d More mature cells may also have tumorigenic potential following tissue disruption, inflammation or other proliferative stimuli that induce facultative stem-like phenotypes. If these influences promote plasticity and reprogramming in situ, the resulting stem cell-like tumors could in fact derive from a variety of precursor cells. e The cell or cells of origin of Claudin low tumors remain to be defined, and this figure presents mechanisms for their genesis involving intrinsic or extrinsic EMT promoting signals. The genesis of these tumors may also depend upon induction of phenotypic plasticity and tumor cell-associated reprogramming