Soil Primes the Seed: Epigenetic Landscape Drives Tumor Behavior

Abstract

Whether invasive tumor phenotypes like EMT arise from oncogenic drivers or from priming of the pre-tumor cell of origin remains unknown. In this issue of Cell Stem Cell, Latil et al. (2017) show that the pre-tumor niche establishes a chromatin state predisposing squamous cell carcinomas to undergo EMT and metastasis, suggesting that the pre-tumor epigenome has prognostic value.

In cancer patients, distant sites of the body have been termed “the soil” to describe where secondary metastatic tumor cells (“the seeds”) can grow (Paget, 1989). While a large body of work supports this soil-and-seed hypothesis, a similar question has not been addressed: whether the pre-tumor niche, or soil, in which the tumor initiates, can also predict tumor behavior. On the one hand, many studies show that chromatin and transcriptional states change dramatically as cells move between tissue environments, suggesting a strong but distinct influence of separate environmental soils in regulating tumor phenotype (Adam et al., 2015). However, a large body of work details hard-wired chromatin states that act as barcodes to pinpoint the site of origin and fail to change even when the cells are passaged in different culture conditions, suggesting dominating intrinsic mechanisms persist within the seed (Chang, 2009). Epithelial-to-mesenchymal transition (EMT) is a process involving loss of epithelial cell-cell contacts and promotion of migratory cell behavior and represents one of the most lethal, invasive, and metastatic cancer phenotypes (Chen et al., 2017). Mammalian skin provides a useful setting for answering whether EMT arises from a cell-intrinsic oncogenic insult (the seed) or is influenced by the pre-tumor niche (the soil) because of its multiple well-defined and genetically characterized cell niches.

In this issue of Cell Stem Cell, Latil and colleagues (Latil et al., 2017) describe a role for the pre-tumor epigenetic landscape and its influence on subsequent skin tumor behavior. The authors sought to determine whether the same oncogenic insult that causes squamous cell carcinoma (SCC) of the skin, applied to cells within distinct skin niches, would influence the resulting tumor’s propensity to undergo EMT. To that end they utilized two well-established oncogenic insults using the keratin 14 (K14) and LGR5 Cre recombinase alleles to target tumor growth to the interfollicular epidermis (IFE) or hair follicle (HF) compartments, respectively. The authors found that the IFE compartment produced less invasive and well-differentiated SCCs, while tumors from the HF formed significant populations of mesenchymal-like tumor cells that have undergone EMT (Figure 1). Single-cell transplantation experiments elegantly revealed that individual LGR5+, but not K14+, derived tumor cells possess the capability to produce both well-differentiated epithelial and invasive mesenchymal tumors in ectopic environments, demonstrating that the HF cancer cell of origin determines tumor aggressiveness.

Figure 1. Local Niches and Niche Factors Prime the Epigenetic Landscape of Normal and Cancer Stem Cells to Respond to Morphogens and Oncogenic Insults.

The same embryonic stem cell responds differently to diverse combinations of morphogens (BMP/FGF/Wnt), while application of local factors can alter the phenotypic outcome in disparate cell types, even with the same factor combination. Keratinocyte-derived tumors from the hair follicle, but not interfollicular epidermal, niche respond to one oncogenic insult (HRas^G12D), but not another (loss of Ptch), by undergoing epithelial-to-mesenchymal transition (EMT). These data suggest that the pre-tumor epigenome plays an important prognostic role in predicting the behavior of human cancers.

To better understand gene regulatory signatures between the normal cell of origin and tumors, Latil et al. then investigated the chromatin state of the HF-derived tumors that led to EMT. An important contribution came from a comparative genomic analysis of the LGR5 and K14-derived tumors with and without oncogenic insult. Interestingly, the authors highlight a significant degree of similarity in transcriptional and chromatin accessibility profiles for normal HF cells and HF-derived tumor cells, suggesting that the chromatin state of cells residing within the HF niche is primed to promote EMT upon oncogenic transformation. Latil et al. define a core set of open chromatin regions containing enrichment for transcription factor (TF) binding motifs common to both K14-derived and LGR5-derived SCCs, suggesting that they play a role in SCC formation, but not EMT. By contrast, the authors observe that chromatin accessibility sites for the pro-differentiation TF p63 are reduced in SCCs from LGR5+ cells and that these HF-derived cells lack the ability to repress transforming growth factor beta (TGF-β), a known driver of EMT gene expression (Moustakas and Heldin, 2016). Subsequent expression of p63 in LGR5+ cells reduced EMT and metastasis, illustrating how this TF opposes EMT in the particular IFE SCC cell of origin.

The present study brings into focus the interaction of the pre-tumor epigenome with the oncogenic stimulus and generates several new questions. The first is how EMT, an inherently transient and reversible process during development, compares with oncogenic EMT, in which the pre-tumor chromatin landscape assumes a stable and oncogenically permissive conformation throughout tumorigenesis (Biggs and Mikkola, 2014). Transient EMT during normal hair follicle morphogenesis occurs when local niche factors like noggin and Wnt/β-catenin downregulate E-cadherin and cell adhesion to allow inward growth of the hair follicle (Jamora et al., 2003). The observation that HF-derived tumors in the Latil study maintained their EMT behavior demonstrates the cell-intrinsic and irreversible nature of EMT, indicating that the seed eliminates its dependence on the local environment. The epigenetic mechanism driving irreversible EMT following an oncogenic stimulus (in this case ras activation with p53 loss), and how oncogenic EMT compares to the normal homeostatic process, remains an open question.

Another issue is whether or not the pre-tumor chromatin state confers the same phenotype if transformed with different oncogenic insults. SCCs and another common tumor type, basal cell carcinoma (BCC), can be produced from the same cell of origin, but the latter is driven by overexpression of a different signaling pathway, the hedgehog pathway. In contrast to the Latil study, loss of the hedgehog pathway tumor suppressor Ptch1 in the Lgr5+ HF compartment produced BCCs with qualities similar to those from cells derived from K14+ IFE cells, and they lacked significant EMT (Kasper et al., 2011). This finding suggests that the EMT phenotype may be unique to the initial tumor type or oncogenic insult provided. It is also reminiscent of previous studies in the stem cell literature wherein distinct morphogenic stimuli applied to identical embryonic stem cells resulted in diverse embryonic lineages and tissue outcomes (Murry and Keller, 2008). This finding additionally suggests that understanding the phenotypic output from an oncogenic stimulus will require a complex knowledge of its interaction with a particular pre-tumor epigenome.

Ultimately, to understand tumor behavior and to provide an accurate prognosis for patients, additional studies are required that define the degree to which the pre-tumor epigenome can be altered by its associated oncogenic stimuli. An important extension of this work will be to identify which factor or factors within the pre-tumor niche determine the final chromatin state. Suppression of such factors could change the pre-tumor epigenome to prevent invasive properties like EMT and would be a useful preventative agent in cancer. An equally provocative question is how a particular oncogenic stimulus interacts with the pre-tumor epigenome and its associated TFs to drive proliferation, invasion, metastasis, and drug resistance. The available data suggest that the same pre-tumor niche may prime separate seeds, transformed by different oncogenic insults, in diverse ways. Additional studies are needed to precisely define the contributions of both the seed and the pre-tumor soil in cancer prognosis.