Epithelial to Mesenchymal Transition Tumors: Fallacious or Snail's Pace?

In this issue of Clinical Cancer Research, Yauch et al. (1) present evidence that a subset of non–small cell lung carcinomas that do not respond to erlotinib therapy have low levels of E-cadherin and higher levels of mesenchymal-type transcripts. The bulk of their evidence uses non–small cell lung carcinoma cell lines but has some confirmatory evidence with immunohistochemical staining on tissue sections of human tumors. The investigators suggest that an epithelial to mesenchymal transition (EMT) signature phenotype determines the clinical response. The investigators fall short of calling these tumors with the EMT marker proteins EMT tumors; however, they do bring up an area that has recently received a great deal of attention.

The lines in the sand were drawn in a recent exchange in Cancer Research (2, 3). Tarin discussed the fallacy of EMT tumors in a recent essay in response to an article by Thompson and Newgreen (2). Tarin states that “…EMT is an unfortunate misconception resulting from erroneous interpretation of pathologic data” (3). The rejoinder from Thompson and Newgreen joins the battle (3), suggesting the issue is one of semantics: “Exactly what constitutes an EMT is open to interpretation….” Thompson and Newgreen allow “…expression of mesenchymal marker proteins and altered morphology resembling mesenchymal cells” (3). Tarin does not and, insisting on rigor, takes the view that the “… premise that EMT occurs in real cancers is seriously in doubt and the interpretation of the molecular changes should be revisited.” It is clear, from a quick glance at the literature and the article under discussion, that EMT and EMT tumors have been adopted by the scientific community with a meaning and implication not intended by Tarin. The current article documents the protein markers, but not the morphologic changes.

The debate, in our view, has been productive. Yauch et al. document that some human cancers without a predominant spindle cell component lose some of the epithelial markers and gain some of the mesenchymal markers. Most beneficially, the EMT markers can identify tumors that do not respond to a specific therapy, namely erlotinib. Because the images of the tumors provided do not clearly exhibit spindle cell or metaplastic morphology, they will not fulfill Tarin's rigorous requirements and would not be considered EMT by most morphologists. This makes me wonder whether we have been missing something. Have we been overlooking important prognostic factors simply because we have no easy conceptual niche for the phenomenon? Is it right to call them EMT or is this overreaching the conceptual boundaries? Should we call these EMT tumors or merely poorly differentiated tumors?

Why even concern ourselves with the nomenclature? What's in a name? Why even worry about seemingly semantic controversy? Words represent concepts and concepts represent meanings. The concept of epithelial mesenchymal transition has opened new ways of assessing and interpreting the physical entities that have been previously observed but frequently have been either ignored or misinterpreted. The fact is that investigators are observing old phenomena with useful new concepts and a fashionable new term, EMT.

Rather than join the terminology fray, we will provide a biological view from the perspective of mouse tumor biology. The mouse, with its experimental flexibility, is frequently an informative source. Spindle cell tumors have been found in the mouse for many years. They have been routinely diagnosed as sarcomas and associated with infection by retroviruses or with chemical carcinogenesis. With the advent of genetic engineering, increasing numbers of sarcomatous lesions have been appearing in unusual tissues and under unusual circumstances (4). Some years ago, in the process of developing transgenic mammary glands, we observed that the morphology and biology of mammary cells transfected with H-ras was dose dependent (5). Limited doses of the oncogene resulted in premalignant growths, whereas high doses of replication-competent retrovirus resulted in spindle cell carcinosarcomas (5). In genetically engineered mice, sarcomas and angiomas are routinely found in tumor suppressor animals, such as Pten or p53 knockouts (6–11). Most mouse pathologists have been perplexed by the phenomenon; however, little has been written.

For my colleagues and I, personally, the conceptual breakthrough came from two separate laboratories that simultaneously observed cohorts of transgenic mice in which the oncogenic transgene was not expressed in the resulting mammary tumors (8, 12, 13). Investigation of these tumors showed that, in both cases, the tumors were predominantly spindle cell tumors that expressed vimentin, smooth muscle actin, keratin 8/18, and keratin 14 (8, 12, 13). On this basis, the tumors could have been classified as myoepithelial carcinomas. They also expressed, however, transcription factors associated with EMT, such as Slug, Snail, or Twist (13). With this observation, we returned to our archives and reexamined some of the spindle cell tumors from several p53 cohorts (Fig. 1). They also turned out to coexpress smooth muscle actin, vimentin, keratin 14, and keratin 18. Subsequently, virtually every genetically engineered mice–related mouse spindle cell tumor sent to us has been found to have a mixture of epithelial and mesenchymal markers. The highly cellular stroma frequently observed with some other mouse tumors is also composed of smooth muscle actin staining cells. Again, the spindle cell population has a mixture of stromal and epithelial markers. Are these cells myofibroblasts or EMT cells that have been overlooked in our ignorance?

Although these observations can be regarded as anecdotal, the pathologist is in a position to draw inferences from broad experience. The experimental evidence to support the transition from a pure epithelial to a mesenchymal phenotype comes from the doxycycline-inducible models developed in the laboratory of Lewis Chodosh (6, 11). In these models, the oncogenes are induced with doxycycline in the drinking water, resulting in rapid development of palpable tumors. Removal of the doxycycline results in the dramatic disappearance of the tumor phenotype. Some tumors, however, either persist or reappear. Most tumors will reappear with reintroduction of the antibiotic into the drinking water. The recurrent or persistent tumors usually retain the tumor phenotype expected for that particular transgene. Some tumors, however, recur as spindle cell tumors. The wnt, erbB2, and p53^null tumors were more frequently spindle cell than the myc tumors. The immunohistochemical staining on these tumors typically exhibits dual lineage.

These observations led to a more detailed study of cells from the erbB2 model (11). The analysis showed that recurrent doxycycline-resistant tumors develop EMT features. Because syngeneic mice could be used, the molecular and biological evolution of tumor cells could be followed by transplantation into wild-type hosts. Again, the experiments made it abundantly clear that tumor recurrence was associated with EMT. Chodosh's lab found that Snail was overexpressed in the EMT tumors (11). Transfecting a transgenic neu tumor cell line with Snail resulted in a decreased latency time and the EMT phenotype. A meta-analysis of extant data suggested that recurrent human breast cancers are associated with Snail overexpression. Perhaps, the EMT-associated transcription factors identified by these basic studies should be added to the clinical scientist's list of criteria.

With this verification of the EMT phenotype, we have used the combination of spindle cell morphology, dual staining with antivimentin and antikeratin, and staining with anti–smooth muscle actin as diagnostic criteria for EMT tumors in the mouse (Fig. 1). We generally find the anti–keratin 5 or 14 and anti–smooth muscle actin the most revealing stains, often showing very subtle gland-like relationships that are difficult to appreciate in the H&E stains. These criteria are consistent with those suggested by Thompson et al. (3).

The idea of some type of mesenchymal element in human breast cancer was described many years ago using terms such as fetal-type fibroblasts and malignant stroma to describe cases with hypercellular stroma (14–16). Some investigators found the same genetic abnormalities in the cancer and the cancer-associated fibroblasts (17). Studies by Cunha suggest that cancer associated fibroblasts can transform hyperplastic epithelium (reviewed in ref. 18). These studies went beyond the recent identification of host and stromal factors that influence mammary gland development and, more specifically, breast cancer. The idea proved difficult to verify, however, and seemed to drift out of the limelight for lack of attention.

Although acknowledged in surgical pathology nomenclature, these types of phenomena never quite made it into classification schemes. Antibodies against keratin 14 and smooth muscle actin have identified spindle cell elements that may or may not be clinically significant (19–23). A collection of 100 dual-staining spindle cell breast carcinomas in the Armed Forces Institute of Pathology archives was described in 1989 (24). One wonders what they would be called in 2005. With the rekindled interest, EMT tumors now have been described in the breast and other organs (25–32).

It has all been very confusing and the term EMT really has not been extensively used in the clinical arena. Some pathologists, such as Tarin, deny the existence of EMT tumor of the human breast. The current article reopens the debate in other terms. Our view is that the biology behind the nomenclature is valid and that the observations warrant full investigation before discarding either the term or the concept. As useful as the debate may be, a word of caution must be added before the Yauch et al. article stimulates a proliferation of EMT phenotype articles. EMT cannot be defined in terms of a single antigen such as E-cadherin. Otherwise, clearly epithelial tumors, such as the lobular carcinomas of the breast, could become EMT tumors simply because of the silencing of E-cadherin. The images used by Yauch et al. to illustrate the non–small cell lung carcinoma clearly show cohesive epithelial cells. Further, dual staining cannot be the single criterion; otherwise, entities such as leiomyomas of the uterus, synovial sarcomas, and other dual-staining tumors become EMT tumors.

Thus, EMT is here to stay. There is biology behind the application of the term to tumors. The lessons from mouse tumor biology resonate and instruct the examination and reconsideration of phenomena previously observed but not fully appreciated in human pathology. Seen with fresh eyes, reinforced by new concepts, and confirmed with new technologies, we are being led, albeit at a snail's pace, toward new twists that may serve our patients well. The article by Yauch et al. reinforces the need to completely explore the implications of the new concept. The scientific community that is defining EMT will need to get together with the clinical scientists to determine appropriate criteria.