In 2004, Braakhuis and colleagues stated that stem cells in oral mucosa had yet to be definitively identified, and so in a paper they presented [37] they proposed a genetic model for the progression of oral squamous cell carcinomas, based on clonal evolution. The group stated that somatic oral stem cells could be found in the basal layer of oral mucosa, whilst more differentiated cells, which were not themselves terminally differentiated, were term ed transit amplifying cells and could be found in the basal and suprabasal layers in oral mucosa [37]. Braakhuis and colleagues suggested that as stem cells had the longest lifespan, these were the cells most likely to be the cells of origin for oral cancers, and that oral cancer manifested itself as a cluster of cells (patch) in which p53 was mutated [37]. The group proposed a theory (the "patch-field carcinoma model"), whereby upon an oral stem cell receiving one or more genetic "hits" (one of which was probably alteration of p53), it forms a patch consisting of daughter cells carrying these genetic alterations [37]. Further genetic alterations then cause the stem cell to escape the bounds of normal stem cell regulations, which give it a growth advantage. This then results in lateral expansion of the patch and symmetric division would allow for expansion of stem cell numbers, thus providing more target cells for genetic "hits" [37]. M ultiple
clones develop and expand within the field and selection of these clones results in invasive
carcinoma with the potential to metastasise [37]. The authors suggest that this model could explain tum our re-growth following surgical tum our resection [37].
In the same year, Mackenzie studied the colony forming characteristics of oral squamous cell carcinomas (OSCCs), using an organotypic culture method, in which OSCC cell lines were grown on collagen IV matrices plated with normal oral fibroblasts and normal oral kératinocytes [38]. This resulted in each of the OSCC lines reforming an altered epithelium with a stratified structure, but failed to produce well-differentiated epithelium, and no stratum corneum was formed [38]. Instead, the tum our cells clustered in the basal layer creating colonies [38]. Cells from the OSCC lines were also found to be able to survive and proliferate when the cells represented 0.1% of the plating mixture [38]. However, it was noted by the author that 90% of the OSCC cells which were plated to form cultures did not survive, and the author suggested that this could mean that malignant oral cells were highly heterogeneous and only contained a small fraction of cells with clone-forming abilities [38].
Prince and colleagues [39] used head and neck squamous cell carcinomas (HNSCC) rather than specifically focusing upon oral squamous cell carcinomas, in their study. The group used tw o mouse models to examine the characteristics of tumours from HNSCC patients, the non-obese diabetic/ severe combine immunodeficient (N O D / SCID) model and the Rag2/ cytokine receptor common y- chain double knockout (Rag2yDK0) model [39]. The group found that 52% (1 3 / 25) of the samples of patient tumours implanted into the mouse models, yielded tumours and that growth rates were similar for both mouse models, leading the group to conclude that either mouse model could be used [39]. In this paper, the group also detailed that from nine patients, among the samples they collected, three were from the tongue and two were from the floor of the mouth, and histology showed that the tumours ranged from poorly differentiated to well differentiated [39]. The group used the CD44 marker, and seven markers grouped together, which they called lineage markers, to separate non-epithelial from epithelial cells (Lin), and found that CD44^Lin' cells produced tumours which contained both CD44^Lin' and CD44'Lin' cells, concluding that the CD44^Lin' population
contained the HNSCC stem cell population [39]. They group also found that BMI staining of the nuclei of cells within the CD44^Lin' population, given that BMI had been implicated in the self-renewal of cells from other cancers, verified that the CD44^Lin' population contained HNSCC stem cells [39]. A point of note was that the group noted that 20 of 31 injections of cells which stained for CD44, i.e. were CD44^Lin", formed tumours when a minimum of 5 x 10^ cells were injected, but they also found that one of 40 injections in which cells stained CD44'Lin‘ yielded tumours and th a t they failed to form
tumours when 5 x 10® cells were used [39]. This could suggest that if the CD44 Lin‘ population did contain cancer stem cells (CSCs), they were very rare rather than not present at all, and this would follow the cancer stem cell model.
Chiou and colleagues found that cells from one highly malignant OSCC cell line and cells from one less malignant OSCC line, enriched for stem-like properties using FACS, expressed greatly increased levels of three genes associated with stem and progenitor cells, Oct-4, Nanog and Nestin, compared with their parental OSCC cells [40]. These stem-like cells from the highly malignant cell line also formed tumours in two of three mice, compared with its parent line (not enriched for stem-like cells) which produced tumours in one of three mice and required more cells to do so (1 x 10® versus the 1 x 10"^ required for the stem-like cells) [40]. The group also found that increased Oct-4, Nanog and CD133 expression correlated with advanced-stage OSCC and therefore poor prognosis, with increased Nanog expression giving the worst predicted patient survival [40].
Locke and colleagues found upon examining 15 HNSCC lines (many of which were from the oral cavity) that despite small differences between the cell lines, all tended to form three distinct types of colony, holoclones, meroclones and paraclones and that these formations were maintained as the colonies grew [23]. As stated previously in the "Cancer Stem Cell Theories" section, of the three colony formations, cells constituting the holoclones were able to reproduce cells demonstrating the full range of phenotypes from the cell line from which the holoclones came [23]. The group also identified that holoclone cells adhered more quickly to their growth surfaces than cells from the other tw o colony formations, and therefore concluded that rapid adherence correlated with increased clonogenicity [23]. Holoclone cells were deemed to be the source of the rest of the cell population, and the group concluded that HNSCCs contained a sub-population of cells possessing unlimited self-renewal and the ability to produce proliferating and differentiated progeny [23]. This in turn suggested that HNSCCs, or at least those examined in this study [23], followed the cancer stem cell model. Members of the same group verified these results and found that tw o additional lines created from patient tumours also followed the same colony formation and hierarchy patterns, with holoclone cells, concluded to be the most stem-like, staining for CD44, p i integrin and E- cadherin [41].
Further analysis of oral CSC biomarkers was conducted by AbdulMaJeed and colleagues, who used the putative stem cell markers ALDHl, CD271, CD44 and CD24 to examine 385 paraffin-embedded samples of normal, dysplastic (ranging from moderate to severe) and cancerous oral tissue [42]. They found that three of the four markers, ALDHl, CD44 and CD24, stained with higher intensities for
OSCC tissue compared with normal tissue, and that the intensity of ALDHl and CD24 staining correlated with increased disease severity, with the latter being able to distinguish non-malignant and OSCC tissue [42]. The group also found that ALDHl staining was increased for severe dysplasia versus moderate dysplasia and normal mucosa [42]. The group concluded that CD24 achieved sensitivity and specificity scores of 70.9% and 75.3% respectively for OSCC versus non-malignant tissues, 68.9% and 80% respectively for dysplastic versus normal tissues, and 63.6% and 35.9% respectively for severe versus moderate dysplasia [42]. This suggested that CD24 staining of putative stem cells could be used to differentiate OSCC tissue from normal tissue, but showed limited ability to differentiate grades of dysplasia and dysplastic versus normal tissue on the basis of their putative stem cell populations.
The studies reviewed tended to use biomarkers, such as CD44, p i integrin and E-cadherin, to distinguish stem-like from non-stem like cells in oral cancer cell lines. However, as Gonzalez-Moles and colleagues point out in their review of the cancer stem cell hypothesis as it pertains to oral cancers, few of these markers have been found to reliably enrich for oral cancer stem cells, as they also enrich for non-stem cells, and normal stem cells [43]. Gonzalez-Moles and colleagues also identify excretion of Hoechst dye, and expression of ABCG2, Bmi-1 and Oct-4 as markers of cancer stem cells in HNSCCs, but state that these marker procedures are not suitable for routine application. The authors do however suggest that CD44 expression could be lost because, like E-cadherin, it could be involved in the epithelial-to-mesenchymal transition and mesenchymal -to-epithelial processes (EMT and MET, respectively) [43]. A point of note is that the authors of many papers have not mentioned the potential advantage of having a recognised panel of cell lines, for specific cancers such as OSCCs (separated from the general category of HNSCCs), to allow "like-for-like" comparisons to be made, without heterogeneity of cells within different lines contributing errors.
W hat is clear is that the number of biomarkers for cancer stem cells that have been trialled in oral cancer, is low (especially when comparing work in OSCC with that of Quintana and colleagues who used 85 biomarkers to examine CSCs within melanomas [33]) and many studies do not use the gold standard serial xenotransplantation to verify results. Therefore, more research appears to be required to find alternative biomarkers for oral cancer stem-ness, or to develop a system whereby the area of difference between oral cancer stem-like cells (OCSCs) and oral cancer non-stem like cells (be it intracellular, e.g. from the cytoplasm or genetic material, or related to the cell membrane) could be identified prior to an exhaustive test of biomarkers being conducted.