Prognostic Value of Cancer Stem Cell Markers in Potentially Malignant Disorders of Oral Mucosa: A Meta-analysis

In last few years, cancer stem cell (CSC) hypothesis has attracted much attention with respect to tumor initiation, progression and in understanding the fundamental biology to identify molecular signatures responsible for aggressive behavior of tumor. The hypothesis emphasizes that a small subset of cancer cells with stem cell–like characteristics (known as CSCs), possess unlimited proliferative potential and are responsible for tumor formation with phenotypically heterogeneous cell population. This stem cell–like cancer cell population is distinguished from other tumor cells by the expression pattern of stemness related markers such as CD133, CD44, CD24, CD271, Octamer-binding transcription factor 4 (Oct-4), Sex determining region Y-box 2 (Sox-2), Nanog homeobox (NANOG), Aldehyde dehydrogenase-1 (ALDH1), ATP-binding cassette sub-family G member-2 (ABCG2), and Polycomb Group Ring Finger Protein (Bmi-1; refs. 1–5). It has been shown that tumor cells positive for CSC markers have property of self-renewal, tumorigenicity and aberrant differentiation potential (6–8). Existence of CSCs has been associated with tumor aggressiveness and increased resistance to radio-chemotherapeutic intervention in several malignancies (9–12), including oral squamous cell carcinoma (OSCC; ref. 3). Although the association between expression of CSC markers and survival of OSCC cases has been investigated (4, 5, 13), reports evaluating prognostic value of CSCs in oral potentially malignant disorders (OPMD) are still scarce. Certain studies have highlighted the significance of these markers in OPMDs such as oral lichen planus (LP), oral leukoplakia (OL) and oral erythroplakia (OE) to determine oral cancer–free survival (14–16). Therefore, assessing the expression of CSC markers in oral precancer appears to be the most rational approach for distinguishing indolent from high-risk cases.

The aim of present study was to review and perform a meta-analysis of the prognostic value of CSC markers in OPMDs. We also emphasize on the relevance of evaluating CSC marker expression in oral epithelial dysplasia and the limitation of single-marker investigation.

The PubMed database was thoroughly searched until December 2017, using combination of keywords: oral premalignant/precancerous lesion, oral premalignant/precancerous condition, oral potentially malignant disorders, oral leukoplakia, oral erythroplakia, oral lichen planus, oral submucous fibrosis; and cancer stem cell, CD133, CD44, CD24, CD271, podoplanin, ALDH1, ABCG2, Oct-4, Sox-2, Bmi-1, and NANOG. The search result was merged in reference manager software. The retrieved record was reviewed systematically and any discrepancy was resolved by mutual consensus (T.S. Saluja and S.K. Singh; Fig. 1). To minimize the exclusion of any relevant record, the references of suitable articles was analyzed manually. The inclusion was restricted to articles in English language. While conducting this analysis, the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) was followed (17).

Quality assessment of the retrieved records was done on the basis of reporting recommendations for tumor marker prognostic studies (REMARK) guidelines (18). Studies with score above 6 were included in this meta-analysis. The principal guidelines are summarized in Table 1.

This meta-analysis was conducted by Review Manager (Version 5.3.5; Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Relative risk (risk ratio, RR) and odds ratio (OR) with 95% confidence intervals (CI) were used to evaluate the association between CSC marker expression in OPMDs and malignant transformation. The statistical heterogeneity between studies was assessed using Cochran Q and I² statistic. RR/OR was calculated using fixed effects model when the P value was more than 0.1 else random effects model was used. Funnel plots were used to analyze the publication bias. Initially, we conducted meta-analysis for the end point “malignant transformation of OPMD” even if CSC marker or OPMD varied between the studies. Furthermore, we conducted separate meta-analysis specifically for each CSC marker and OPMD. The influence of each study was examined by excluding one study at a time (sensitivity analysis) for each meta-analysis.

The step-wise search protocol is summarized in Fig. 1. Out of 118 articles retrieved form database, 34 full-text articles were evaluated for eligibility based on inclusion/exclusion criteria. Of these, 23 articles (refs. 14–16, 19–38; Table 2; Supplementary Table S1) were assessed for quality as per REMARK criteria (Table 1) and finally 12 studies (14–16, 19–27) were selected (Table 2). Out of 12 studies, 6 evaluated the expression of two CSC markers resulting in a total of 18 investigations. Of these, podoplanin was evaluated in six, ALDH1 in four, ABCG2 and Bmi-1 in three each and CD133 in two investigations.

The total study period ranged from 2007 to 2017. Majority of the studies were retrospective. Seventy five percent of studies were from Asia, 8% from Europe, 4% from Australia and 13% from North America. Out of 18 investigations, 3 (17%) were carried out in LP, 11 (61%) in OL and 4 (22%) in OE cases. CSC marker podoplanin was analyzed in 6 (33%) investigations, whereas ALDH1, Bmi-1, ABCG2 and CD133 was analyzed in 4 (22%), 3 (17%), 3 (17%), and 2 (11%) investigations, respectively. Immunohistochemistry was used to quantify CSC expression in all the investigations. Furthermore, 6 investigations provided data for association between grades of epithelial dysplasia and CSC marker expression.

Meta-analysis of the data of 18 investigations from 12 studies is summarized in Fig. 2. Of note, we did not observe significant heterogeneity (P > 0.1) in our analyses. The pooled analysis using the fixed effect model was conducted. An increased risk [RR, 3.31; 95% confidence interval (CI), 2.72–4.02] of developing OSCC in CSC marker–positive OPMDs was seen. After restricting the results to studies that evaluated similar CSC marker, expression of CD133 (RR, 4.20; 95% CI, 2.55–6.92) was associated with the highest risk followed by ABCG2 (RR, 3.64; 95% CI, 2.12–6.25) and podoplanin (RR, 3.45; 95% CI, 2.37–5.04; Fig. 3). The analysis for specific OPMDs showed that CSC marker positive cases of LP (RR, 6.14; 95% CI, 2.70–13.95; Fig. 4A) followed by OL (RR, 3.19; 95% CI, 2.55–3.98; Fig. 4C) and OE (RR, 2.96; 95% CI, 1.84–4.76; Fig. 4B) carried highest risk of transformation to OSCC. Funnel plots for assessing publication bias of the included studies are shown in Supplementary Fig. S1. Sensitivity analysis showed that the results did not vary substantially in both the pooled and subgroup analysis.

For analysis, we considered mild dysplasia as low-grade dysplasia (LGD), and moderate and severe dysplasia as high-grade dysplasia (HGD). Random effect model was used as significant high heterogeneity was observed (I² = 71%, P = 0.004). The expression of CSC marker correlated with HGD (OR, 3.53; 95% CI, 1.13–11.07; P = 0.03; Fig. 5A). However, sensitivity analysis showed a more robust association (OR, 10.15; 95% CI, 4.31–23.89; P < 0.00001) on inclusion of investigations in which only CSC marker podoplanin was evaluated (Fig. 5B). Furthermore, percentage values of cases positive and negative for CSC markers in LGD and HGD were compared. Interestingly, we found that more than one-third CSC marker–positive LGD cases progressed to OSCC (Fig. 5C).

Cancer is the leading cause of morbidity and mortality worldwide. Despite considerable efforts to limit cancer-related deaths, the prognosis of various malignancies is far from satisfactory. According to GLOBOCAN report (2012) more than 145,000 deaths were attributed to oral cancer (39). OSCC is the most frequently occurring malignancy of oral cavity with 5 year survival rate of less than 50% (40, 41). The transformation of normal oral mucosa into OSCC usually occurs in a stepwise manner. A gamut of cellular and tissue architecture changes preceding development of OSCC manifests as clinically visible changes on oral mucosa. These changes are categorized as OPMDs (42), which were earlier [World Health Organization (WHO), 1978] classified as precancerous lesions and conditions (43). These include OL, OE, LP, proliferative verrucous leukoplakia (PVL), tobacco associated keratosis, verrucous hyperplasia, and oral submucous fibrosis (OSMF), to name a few. PVL followed by OE, OL, and OSMF carries the highest risk of malignant transformation (44, 45–52). These OPMDs have potential for regression as well as progression depending on cellular stress and tumorigenic environmental cues. As oral carcinogenesis is an intricate multistep process requiring genetic and epigenetic alterations in several genes and genomic instability (53, 54), it usually provides a reasonable time frame to intercept cancer at precancer stage.

In last decade, identification of CSCs has been an area of research interest as they are responsible for apoptosis resistance, self-renewal and differentiation, thus contributing in development and progression of cancer (55). Studies of premalignant adenomatous polyps of colon and precancerous gastric lesions (56, 57) have shown that the presence of a CSC population in precancer stage is an early indicator of malignant progression. We, therefore, conducted this meta-analysis to correlate the expression of CSC markers in OPMDs with OSCC development. A total of 1,659 cases from 18 investigations were included in this analysis. Our results demonstrated that CSC-positive OPMDs carries significantly high risk of developing into OSCC. Most of the studies included in this meta-analysis (14, 16, 19, 22–25, 27), showed superiority of CSC markers to other factors such as age, smoking habit and histology in predicting progression of OPMD to OSCC. Multivariate analysis of most of these studies suggests that presence of CSC population in OPMDs does not indicate a favorable clinical outcome.

Though several reports have demonstrated prognostic utility of examining CSCs in OPMDs, it has not been widely adopted for oral cancer risk assessment. Over the years different techniques have been introduced for determining malignant potential of oral premalignancy (58), but none is foolproof in identifying individuals at a higher risk for OSCC. Currently, the risk of OSCC development is based on histologic severity of epithelial dysplasia. However, histologic grading is inherently subjective and several authors (59–61) have not observed any significant association between grades of dysplasia and risk of malignant transformation. The data from included studies in current analysis highlight similar observation. Kawaguchi and colleagues (14) reported that more than 50% of OL patients without dysplasia developed oral cancer. de Vicente and colleagues (27) also found that 5% of OL lesions with mild dysplasia progressed whereas 25% of severe dysplasia did not progress to cancer. Habiba and colleagues (16) reported superior prognostic utility of CSC markers over histologic grading. The data extracted from these studies (Fig. 5) shows adequate number of LGD cases positive for CSC markers. It is therefore reasonable to look for CSC marker expression in different grades of dysplasia in bringing precision to oral cancer risk profiling of individual cases.

An intriguing aspect of our observation was malignant transformation of OPMDs that were negative for CSC markers. Though such number of cases was small in all the included studies it raises two crucial questions: How do these OPMDs progress to oral cancer? Is there any other group of cancer cell population responsible for progression to OSCC? In consideration of CSC hypothesis the plausible explanation is that each CSC marker labels only a distinct subset of stem cell–like cancer cell, leaving other unique subpopulation of cancer cells with stem cell characteristics unidentified. In a recent study on breast cancer cell lines and primary tumors little consistency was found in expression of CSC markers CD24, CD44, and ALDH1; and each CSC marker identified a unique population of tumor cells (62). A similar observation was made in various cell lines of ovarian cancer signifying heterogeneity between tumors (63). This implies that CSC marker–negative OPMD cases that progressed to OSCC could have been positive for other CSC markers. Therefore, a critical aspect in predicting clinical behavior of OPMDs is recognizing heterogeneous population of stem cell–like cancer cells driving tumor formation. Given the variability in CSC population and probability of recognizing the presence of more than one subset of stem cell–like cancer cell, it is difficult to ascertain malignant potential of OPMDs based on single CSC marker investigation. Thus, a multi-marker panel approach is certainly relevant to distinguish between indolent and high-risk OPMDs.

To the best of our knowledge, this meta-analysis is the first to systematically appraise the efficacy of CSC markers in predicting malignant transformation of OPMD cases. The strengths of this meta-analysis are inclusion of different studies from dissimilar populations and evaluating the ability of several CSC markers in predicting biological behavior of OPMDs. The main limitation of present meta-analysis is small number of original studies. Another limitation is inclusion of OPMDs from different sites of oral cavity. Because studies on CSCs in OPMDs are very few, we propose separate analysis and reporting for each specific site in future studies. This meta-analysis addressed the end point in OPMDs by doing multiple analyses and every time the results validated the existence of CSCs in OPMDs as a poor prognostic indicator.

In summary, it is certainly acceptable that no CSC marker can be used universally to identify stem cell–like cancer cells. Proper screening and profiling of each case with respect to several CSC markers is essential. As cancer cells are highly heterogeneous at genetic, epigenetic and phenotypic levels, the immunophenotypic characterization of OPMDs with a comprehensive CSC antibody panel could be a decisive factor for early detection of oral cancer. Future research on CSCs in OPMDs should evaluate and compare the expression of various putative CSC markers with the aim of assessing variability in their expression.

No potential conflicts of interest were disclosed.

Conception and design: T.S. Saluja, S.K. Singh

Development of methodology: T.S. Saluja, P. Mishra, S.K. Singh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.S. Saluja, M. Ali

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Mishra

Writing, review, and/or revision of the manuscript: T.S. Saluja, M. Ali, V. Kumar, S.K. Singh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.S. Saluja, M. Ali, V. Kumar, S.K. Singh

Study supervision: V. Kumar, S.K. Singh

This work was supported financially by the Department of Health Research, New Delhi India (grant no. R.12014/08/2017-HR; to T.S. Saluja).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.