To EMT or not to EMT: Ablation of mesenchymal tumor cell lineages reveals the essential role of EMT in pancreatic cancer initiation and evolution

Abstract

Epithelial-to-mesenchymal transition (EMT), a complex biological pathway that facilitates cellular plasticity, is used by tumor cells to enable metastasis and drug resistance. Our functional understanding of the impact of EMT on cancer has been limited by the lack of effective tools to ablate tumor cells as they become mesenchymal. In a recent study published in Nature, Perelli et. al used elegant genetically engineered lineage tracing and ablation strategies to track and eliminate tumor cells as they undergo EMT in pancreatic cancer¹. In a two-pronged approach, they queried the functional consequences of ablating EMT tumor cells before pancreatic ductal adenocarcinoma (PDAC) formation or in advanced PDAC tumors. These experiments collectively revealed that epithelial tumor cells only progress to low-grade lesions with minimal proliferative potential, while mesenchymal tumor cells undergo EMT early on to become malignant and metastasize. Profiling of mesenchymal tumor cell lineages revealed an altered chromatin landscape that leads to chromosomal instability (CIN) and disease progression. CIN is facilitated through complex structural rearrangements and chromothripsis, ultimately driving increased tumor heterogeneity and enhanced proliferation in EMT cells. This work reveals that EMT is an important driver of tumor heterogeneity and progression as a downstream consequence of CIN and provides mechanistic insight into how cellular plasticity can lead to genomic changes that drive disease progression.

Epithelial-to-mesenchymal transition (EMT; also referred to as epithelial-to-mesenchymal plasticity or EMP) is a developmental pathway that is co-opted by tumor cells to facilitate cellular plasticity, which is their inherent ability to adapt and alter their cellular identity during tumor progression². Canonically, EMT is marked by tumors of epithelial origin, termed carcinomas, losing their epithelial identity and acquiring mesenchymal features. This process has been shown to be essential for metastasis, but that paradigm has been challenged in recent years³. In addition to EMT favoring a pro-metastatic phenotype, EMT and associated cellular plasticity have been implicated in driving tumor heterogeneity, clonal evolution, and drug resistance. In a study recently published in Nature, Perelli and colleagues have developed cutting-edge somatic mosaic genome engineering technologies to trace and ablate mesenchymal lineages in pancreatic cancer to demonstrate that EMT is essential for pancreatic cancer evolution by favoring genome instability¹. Remarkably, tracing experiments revealed that advanced tumors and metastatic lesions were universally established from mesenchymal cells, with epithelial lineages only contributing to neoplastic lesion formation that are uniformly counterselected during progression (Figure 1). Thus, mesenchymal tumor cells with high fitness and metastatic potential are selected for early in pancreatic cancer mouse models, in agreement with prior work showing that EMT and dissemination occur early on in Pancreatic Ductal Adenocarcinoma (PDAC) mouse models⁴.

Figure 1. EMT tracing and ablation in pancreatic cancer.

Schematic representation of EMT tracing and ablation techniques used in Perelli et. al.

A significant challenge in modeling EMT in autochthonous genetically engineered mouse models (GEMMs) of cancer has been that any mesenchymal marker that marks EMT cells will also mark stromal cells in the tumor microenvironment. To get around this limitation, the authors built a PDAC GEMM that uses a flip-excision (FLEx) system to ablate Vimentin (Vim)-positive cells that have arisen from Pdx1-expressing tumor cells. This innovative approach is a significant technological advance because it ablates tumor cell mesenchymal lineages in vivo while sparing Vim⁺ stromal cells, such as cancer-associated fibroblasts. Using this new ablation system, the authors demonstrated that EMT cell ablation prior to tumor formation almost completely abrogated malignant progression, resulting in low-grade cystic lesion formation with limited proliferative capacity and no metastatic ability (Figure 1). EMT-intact tumors, on the other hand, progressed to metastatic PDAC. This demonstrated EMT is dispensable for cystic tumor initiation, but indispensable for progression to carcinoma, where highly proliferative and metastatic subclones expand.

To test the role of EMT in established PDACs, the authors built a second GEMM that utilizes a ganciclovir (GCV)-inducible ablation approach to selectively remove mesenchymal lineages in advanced tumors, which similarly reduced tumor growth and metastasis (Figure 1). Intriguingly, the withdrawal of GCV, and thus the re-population of the mesenchymal compartment, completely rescued the phenotype, resulting in highly proliferative and metastatic tumors. This suggests that the epithelial cell compartment can putatively serve as a reservoir for mesenchymal cells or that a remnant pool of Vimentin-low mesenchymal cells that escaped ablation can re-populate post-ablation. Collectively, these state-of-the-art mouse models of EMT have revealed previously unappreciated roles for mesenchymal plasticity in driving progression and tumor cell evolution.

To determine the mechanism of enhanced tumor growth, progression, and metastasis, the authors performed a series of experiments to demonstrate that mesenchymal tumor cells have a higher proliferation index and a significant increase in aberrant mitoses, which results in enhanced chromosomal instability (CIN). The ablation of mesenchymal lineages restrained the emergence of CIN, suggesting that EMT is a driver of genomic instability. As such, EMT-intact tumors were enriched for complex structural rearrangements and chromothriptic (extensive chromosomal rearrangements to one or more chromosomes) events, while EMT-depleted tumors showed very little evidence of chromothripsis. In PDAC patients and PDX models, the authors similarly found that tumors enriched for EMT signatures were more prone to accumulating chromothripsis and other complex structural variants.

To understand the functional role of these genomic alterations on tumor heterogeneity, the authors performed single-cell RNA-sequencing (scRNA-seq) and single-cell genome and epigenome by transposases (GET) sequencing (scGET-seq) on cultured lineage-labeled epithelial and mesenchymal populations. The activation of an EMT program resulted in enhanced chromosome accessibility in centromeric and pericentromeric chromosomal regions that putatively delay mitosis and resulted in lagging chromosomes and extranuclear chromatic sequestration that are permissive for chromothripsis, allowing for the emergence of clonal diversity. Finally, lineage trajectory analysis of primary and metastatic tumors revealed that EMT-competent tumors are enriched for complex transcriptomic branching with high levels of transcriptomic entropy that fuels their increased tumor heterogeneity. Collectively, these data suggest a paradigm-shifting model, whereby EMT and cellular plasticity are required for tumor growth and clonal expansion.

Canonically, as tumor cells enter the metastatic cascade and begin to disseminate, they turn on EMT programs that enhance their migratory and invasive capabilities. They maintain this mesenchymal state until they reach the metastatic site where they undergo the reverse mesenchymal-to-epithelial transition (MET) to re-establish their epithelial cell identity and proliferate². The major paradigm in the field posits that EMT cells favor motility over proliferation, with a clear dichotomy between migration and growth phenotypes. Perelli and colleagues demonstrate through their in vivo tracing and ablation systems that tumor cells can acquire mesenchymal states and simultaneously enhance their proliferative potential, thus reshaping the EMT paradigm. This, in part, explains a paradox in pancreatic cancer patients, where quasimesenchymal/squamous/basal-like subtypes that are enriched for cells undergoing EMT are more aggressive than their classical subtype counterparts, which are primarily epithelial in nature.

It has become increasingly accepted that the transition from an epithelial to a mesenchymal state occurs across a continuum, with tumor cells often residing in intermediate states that retain both epithelial and mesenchymal markers⁵. Notably, the approach taken in this work marks mesenchymal cells based on Vimentin positivity, which will trace and deplete both complete EMT cells, which have completely lost their epithelial identity and express various mesenchymal markers, as well as hybrid/partial EMT cells that co-express both epithelial and mesenchymal genes. It is possible that hybrid EMT cells, which have been proposed by others to have enhanced proliferative capacity⁶, are a subset of Vimentin-positive cells that are being depleted by this approach and may drive the reduced proliferation phenotype. Future experiments should look to selectively trace and deplete hybrid or complete EMT cells to determine the contribution of the different cell states to these phenotypes.

Surprisingly, the selective ablation of mesenchymal lineages resulted in tumors with relatively flat genomes, despite having a combination of mutant Kras^G12D and deletion of Trp53, Cdkn2a, and Cdkn2b. Epithelial lineages are thus less likely to harbor genome alterations. One possible complication in interpreting these experiments is that GCV-based depletion is known to ablate surrounding cells (in this context Vimentin-negative cells), known as the bystander effect. Thus, GCV ablation of EMT cells in this system may also ablate neighboring cells that are not undergoing EMT, which will alter the interpretation of these results. This is particularly relevant in PDAC, which exhibits a high degree of EMT, and the cumulative bystander ablation of non-EMT cells is likely to be significant in number, and thus an important caveat to consider.

The authors’ discovery that EMT facilitates CIN is exciting, as it offers an opportunity to selectively kill EMT cells using previously validated CIN-exploiting therapies that rely upon the idea that unstable chromosomes are often present outside of the nucleus and activate DNA sensing pathways such as cGAS-STING. Therefore, the persistent activation of the cGAS-STING pathway in CIN tumor cells might be a selective vulnerability that is enriched as cells undergo EMT. It is tempting to speculate that EMT cells have higher CIN and thus will be more susceptible to STING inhibition, resulting in tumors comprised primarily of epithelial cells with relatively flat genomes and less aggressive characteristics. This has been indirectly demonstrated by others, where STING inhibition of CIN tumors prolonged survival by reducing metastases and downregulating EMT genes⁷. Interestingly, Perelli et al. show that the interferon (IFN) gene cluster on chromosome 4C4 (syntenic to human 9p21) is lost in EMT-intact tumors, and others have shown that the same genomic region is often deleted in cancer and correlates with metastasis and immune evasion⁸. The loss of IFN genes in cells undergoing EMT might contribute to immunoediting at the metastatic site, which results in immune-evasive metastatic lesions comprised of immune-privileged clones⁹. In addition to the interferon genes, this genomic region contains CDKN2A and CDKN2B tumor suppressors that encode p16^INK4a/p14^ARF and p15^INK4b. It is provocative to think that EMT may enrich for a subpopulation of tumor cells with 9p21 loss that are highly proliferative due to the loss of CDKN2A/B and associated p53 and RB tumor-suppressive pathways. In addition to cGAS-STING and associated IFN signaling, KIF18A has emerged as a mitotic vulnerability enriched in chromosomally unstable cancers¹⁰ that may be an Achilles heel of EMT cells with CIN. Inhibitors of KIF18A activate the mitotic checkpoint and cull CIN cancer cells¹⁰, and future work should test if KIF18A inhibition selectively kills mesenchymal lineages that Perelli and colleagues have shown are enriched for aberrant mitoses. More generally, this phenomenon may be indicative of EMT driving a DNA damage response that is vulnerable to inhibitors of the pathway. Collectively, Perelli and colleagues’ discovery that EMT cells are enriched for CIN opens the door to many new therapeutic avenues to specifically target mesenchymal lineages and block metastasis and disease progression.