Keratin 17 Is a Prognostic Biomarker in Endocervical Glandular Neoplasia

Daniel Mockler, MD; Luisa F. Escobar-Hoyos, PhD; Ali Akalin, MD, PhD; Jamie Romeiser, MPH; A. Laurie Shroyer, PhD; Kenneth R. Shroyer, MD, PhD


Am J Clin Pathol. 2017;148(3):264-273. 

In This Article


K17 was identified in normal cervical tissue and cervical adenocarcinoma (n = 1) in the mid-1980s while profiling intermediate filaments for diagnostic purposes.[21] Early studies of K17 in tissue were greatly limited by lack of access to antibodies that were reactive in FFPE.[22] In 1989, the monoclonal antibody (E3) directed against K17 was identified,[23] and subsequent studies in the 1990s looking at a limited number of cases of glandular neoplasia found weak- to moderate-intensity staining in most cells of adenocarcinoma cases as well as identifying a reserve cell (RC)/stem cell phenotype (CK 8, 14, 16, 17, 18, and 19 and bcl-2) of the normal endocervix.[24,25] It was hypothesized that cervical lesions that maintain an RC phenotype may develop into malignant lesions while loss of an RC phenotype may predict lesions that are regressive in nature and that cervical adenocarcinomas do not arise from the columnar cells of the endocervix but rather the RC compartment, where K17 expression was always found.[24] This bipotential nature/keratin expression profile of RC could suggest that a cervical intraepithelial neoplasia lesion could give rise to adenocarcinoma with persistence of all of the cytokeratins except K5, K6, and K16.[22] RCs also have been studied in microglandular hyperplasia, a common endocervical glandular proliferation that can either produce a stable population of RCs or terminate in mature squamous metaplasia. It has been postulated that the plastic nature of RC cells and the molecular mechanisms governing these phenotype switches in microglandular hyperplasia might explain the multiplicity of neoplastic phenotypes (squamous and glandular) observed in cervical neoplasia.[26]

Since the first microscopic investigations of the endocervical canal, the presence of RCs has been noted. On light microscopy, these cells are small cuboidal cells with relatively large nuclei that are located subjacent to the columnar epithelium and between the basement membrane of endocervical glands. Anatomically, these cells are present in the proximal endocervix and have a discontinuous distribution pattern, with the greatest concentration being present at the squamo-columnar junction.[27] It is believed that these RCs regenerate the columnar epithelium as well as respond to trauma and hormonal cues by enabling columnar to squamous metaplasia.[26] RCs have been shown to have a complex yet distinct keratin expression pattern, which includes K5, K8, K14, K17, K18, and K19,[28] with K17 being the most specific marker for RCs and RC hyperplasia.[29] These cells are noncycling (Ki-67 negative) and unrelated to human papillomavirus (HPV) status (p16 negative), and it is believed that these RCs function as the stem cell population of the cervix, which is supported by studies showing ubiquitous expression of Bcl-2 and p63 in these cells.[25,29,30]

A columnar epithelial hypothesis of cervical carcinogenesis has been proposed. In this hypothesis, the HPV target cell is not the basal layer of the squamous epithelium but the subcolumnar RCs of the endocervical mucosa.[31] This HPV target cell can be identified by the combined expression of K17 and p63, and once this target cell is infected by HPV, it has two potential fates: the cells can remain in their original form or become benign squamous epithelium by normal metaplasia; however, if the cells undergo HPV-mediated transformation, they can produce AIS or further undergo atypical metaplasia to produce a squamous intraepithelial lesion, which is a logical explanation for why endocervical glandular and squamous lesions often occur together.[29] This theory also correlates with the anatomic location of RCs, which are located near the squamocolumnar junction and could explain why recurrence rates are lower following excisional or ablative treatments of the proximal endocervical canal.[32]

K17 has been studied previously in small numbers of adenocarcinoma cases secondarily in the context of cytokeratin profiling, and to our knowledge, no studies have looked specifically at K17 expression in human FFPE tissue as a clinical prognostic indicator in cervical adenocarcinoma and adenocarcinoma in situ.[22,24,25] Prior studies have shown that K17 expression can be correlated with poor prognosis since CD44+/K17+ cells derived from primary cervical carcinoma have been shown to have stem-like properties, including the capacity for cell renewal, chemoresistance, and in vivo tumorigenicity.[33] In bladder cancer, K17 has been shown to mark a basal-like population of highly tumorigenic cancer cells that are located at the invasive tumor front.[17] K17 has also been shown to be a poor prognostic indicator in other epithelial malignancies, including gastric adenocarcinoma, bladder carcinoma, ovarian carcinoma, and triple-negative breast carcinoma.[16–19] Since K17 is overexpressed in many different types of cancer, it is unlikely to be a specific marker for primary cervical cancer vs carcinomas that have spread to the cervix from other anatomic sites. Mechanistic insight into its role as a prognostic biomarker, however, is based on recent work from our laboratory that has shown that K17 functions as an oncoprotein by mediating cancer cell-cycle progression by promoting p27KIP1 nuclear export and degradation in cervical SCC, and similar mechanisms are likely to be important in endocervical adenocarcinoma.[15]

It should be noted that we did not detect p16 staining in 18 (20%) of 90 cases. Although most cervical adenocarcinomas are associated with high-risk HPV, it should be noted that in recent years, it has been reported that the more uncommon morphologic subtypes of endocervical adenocarcinoma, such as clear cell and adenoma malignum (which were included in our study), are often unassociated with HPV infection.[34]

In summary, we have shown that K17 expression in normal endocervix is limited to subcolumnar RCs near the squamocolumnar junction, and expression in the columnar endocervical epithelium was highly specific for glandular neoplasia of the cervix. Most important, we further determined that high K17 expression is a powerful, negative prognostic biomarker that can be used to identify patients who have the shortest survival probability following the diagnosis of endocervical adenocarcinoma. Thus, K17 immunohistochemical test results could provide important additional data to help inform clinical decisions related to optimal therapeutic intervention.