Clinicopathological significance of Sox2 expression in patients with breast cancer: a meta-analysis

Abstract

Sex-determining region Y-box protein 2 (Sox2), an embryonic transcription factor located at chromosome 3q26.33, has been frequently demonstrated to be an important prognostic marker for various tumors, including breast cancer. However, its clinicopathological role in breast cancer has not been fully elucidated. To derive a more precise evaluation, we here performed a meta-analysis focusing on the association between Sox2 expression and various clinicopathological characteristics of breast cancer. Relevant publications were identified and retrieved using PubMed, Embase, Cochrane Library, Web of Science, Chinese National Knowledge Infrastructure, and Chinese Biomedical databases. Ten studies with a total of 1713 patients with breast cancer were included in our meta-analysis. Reported odds ratios (OR) and the corresponding 95% confidence intervals (95% CI) were pooled to assess the strengths of the analyzed associations. Our results revealed significant positive associations between Sox2 expression and increased tumor size (pooled OR=2.61, 95% CI=1.91-3.58), histological grade (pooled OR=2.28, 95% CI=1.72-3.03), lymph node metastasis (pooled OR=4.17, 95% CI=1.20-14.45), and the highly aggressive triple-negative phenotype (pooled OR=2.64, 95% CI=1.11-6.29). However, no associations were observed for TNM stage and estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 statuses. Overall, the results of this meta-analysis indicate that Sox2 may be considered as a prognostic marker for breast cancer. More well-designed studies with larger sample sizes are warranted to clarify the prognostic significance of Sox2 in breast cancer.

Introduction

Despite great advances in the prognostic methods and treatments, breast cancer remains the most common malignancy in women, indicating a major public health challenge [1]. In recent years, the cancer stem cell (CSC) hypothesis has shed lights on the development of breast cancer, proposing that a pool of malignant cells with stem/progenitor cell properties and an enhanced capacity to progress compared to their more differentiated non-tumorigenic counterparts exists, and may be responsible for the poor clinical outcomes associated with breast cancer [2]. Recently, pluripotency roots of cancer stem cells have been found to correlate to the embryonic signature in breast cancers, suggesting a vital role for embryonic stem cell-like cancer cells in tumor invasion, metastasis, and/or recurrence [3-5]. Hence, identification of these cells through associated biomarkers may be a reasonable approach to improve the pathological diagnosis or for predicting the clinical outcome of breast cancer.

Sox2, a High Mobility Group (HMG) domain transcription factor located at chromosome 3q26.33 is reported to be involved in the regulation of self-renewal and pluripotency in embryonic stem cells [6]. For instance, Sox2 has been found to be overexpressed in mouse neural stem cells, resulting in repression of differentiation, whereas inhibition of Sox2 conversely has been demonstrated to lead to premature exit from the cell cycle and differentiation into neurons [7]. Recently, a growing body of evidence has revealed the contribution of Sox2 to tumorigenesis, as well as suggested a number of links between Sox2 and the clinical progression of various types of tumors, including human breast [8], pancreatic [9], and lung cancers [10]. Although the association between Sox2 expression and breast cancer patients has been extensively reported in the last 3 or 4 years, the results of these studies are not always in agreement. For example, Huang et al. [11] found that Sox2 was expressed more frequently in tumors of larger size and higher histological grade, whereas Piva et al. [12] failed to show an association between Sox2 expression and tumor size. Moreover, in a study by Rodriguez et al. [8], no significant differences in Sox2 expression between breast cancer patients with different histological grades were observed. These inconsistent results may be due to insufficient numbers of samples, as well as other factors. Accordingly, we here carried out a meta-analysis of relevant studies published on the topic to quantitatively evaluate the clinicopathological significance of Sox2 in breast cancer.

Materials and methods

Search strategy

Relevant papers included in this meta-analysis were systematically searched in the PubMed, Embase, Cochrane Library, Web of Science, Chinese National Knowledge Infrastructure, and Chinese Biomedical databases (up until April 2014), using the following limits: Human, article in English or Chinese. Combinations of the following search string were used to screen for potentially related studies: (“Sox2” OR “sex-determining region Y-box protein 2”) AND (“breast cancer” OR “breast carcinoma”). The reference lists of major textbooks, review articles, and all of the articles identified by the search were then individually and manually searched to find other potentially eligible publications. The study published more recently or that contained more information was selected if duplicated data were published.

Selection criteria

For inclusion into the meta-analysis, the identified studies had to meet the following criteria: (1) randomized controlled studies or observational studies (case-control or cohort) focused on the clinicopathological associations of Sox2 expression in breast cancer; (2) all patients with a diagnosis of breast cancer had to be confirmed by pathological or histological examination; (3) the studies had to provide sufficient information about the expression levels of Sox2. Studies with no clinicopathological data and articles with insufficient published data for determining an estimate of the odds ratio (OR) and 95% confidence interval (CI) were excluded from the meta-analysis.

Data extraction

All data from the included studies were extracted independently by two reviewers using a standardized form. The form used for data extraction documented the most relevant items, including the first author’s surname, year of publication, geographical location, age distributions of the participants, protein detection methods, sample size, staining patterns of Sox2, cutoff scores for the definition of positive staining, and clinicopathological parameters (tumor size, lymph node status, histological grade, TNM stage, and estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) status (molecular subtype). Any disparities between the two reviewers were resolved by discussion and consultation with a third reviewer.

Statistical analysis

The meta-analysis was performed using Review Manager version 5.2 software (provided by The Cochrane Collaboration, Oxford, England). Given the effect value, the ORs and their 95% CIs were combined and calculated for the quantitative aggregation of the results. Heterogeneity across studies was evaluated by Cochran’s Q-statistic test (P<0.1 was considered as statistically significant heterogeneity) [13]. For the existence of significant heterogeneity, a more conservative random-effects model was used to calculate the pooled estimates; otherwise a fixed-effects model was applied [14,15]. Using Stata 12.0 software (Stata Corporation, College Station, TX, USA), the potential for publication bias was assessed by Begg’s test and Egger’s linear regression test (P<0.05 was considered significant) [16]. The quality of studies in meta-analysis was assessed based on the Newcastle-Ottawa Scale (NOS) (http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp). Each study was assessed by eight questions with a maximum score of 9, and studies with scores equal or above 6 were considered as high-quality, otherwise they were defined as low-quality. To guarantee the effectiveness, the subgroup analyses based on potentially significant variables were performed when more than six studies were included in the overall analysis. Sensitivity analyses by sequential exclusion of each study were also conducted. All of the generated p values were two-tailed.

Results

Search results

Initially, our search strategy identified a total of 81 studies assessing the association between Sox2 expression and clinicopathological parameters of breast cancer. After the titles and abstracts of these identified studies were screened, 51 articles were excluded. Subsequently, we systematically reviewed the remaining full texts and excluded another 20 articles. As a result, 10 clinical cohort studies involving 1713 patients with breast carcinoma were included in the pooled analyses [6,8,11,12,17-22]. All ten studies utilized the immunohistochemistry method for protein detection. The publication years of the eligible studies ranged from 2007 to 2014. Of these, 8 studies were reported in English and 2 in Chinese. The main characteristics of the included studies are displayed in Table 1.

Table 1.

Characteristics of the included studies

First author	Year	Study type	Country	Ethnicity	Age (year)	Histological type	No. of patients	Method	Staining patterns	Cut-off scores	Quality assessment
Huang	2014	cohort	China	Asian	NR	Mixed	552	IHC	Nuclear	>1%	8
Nagata	2014	cohort	Japan	Asian	55.2 (29-87)	Mixed	100	IHC	Nuclear/cytoplasm	scores>2	7
Piva	2013	cohort	Spain	Caucasian	NR	Mixed	81	IHC	Nuclear	scores>2	6
Abd	2014	cohort	Egypt	African	53.37 (35-69)	Mixed	126	IHC	Nuclear	≥0%	7
Lengerke	2011	cohort	Germany	Caucasian	50-69	Mixed	86	IHC	Nuclear	≥0%	7
Rodriguez	2007	cohort	Spain	Caucasian	NR	Mixed	198	IHC	Nuclear	≥0%	7
Li	2012	cohort	China	Asian	NR	NR	269	IHC	Nuclear	>10%	8
Qi	2012	case-control	Chian	Asian	55 (30-80)	Mixed	65	IHC	Nuclear	>10%	5
Xu	2013	case-control	China	Asian	53 (34-78)	Mixed	78	IHC	Nuclear/cytoplasm	scores≥3	5
Leis	2012	cohort	Spain	Caucasian	NR	Mixed	158	IHC	Nuclear	≥0%	7

NR, not reported; IHC, immunohistochemistry.

Meta-analysis

To identify the value of Sox2 expression, 10 studies were utilized to address the associations between Sox2 protein expression and clinicopathological parameters of patients with breast carcinoma. As illustrated in Figure 1 and Table 2, the pooled OR indicated a significant association between positive Sox2 expression and larger tumor size (pooled OR=2.61, 95% CI=1.91-3.58). Similarly, the overall estimates suggested that positive Sox2 expression was significantly associated with an increased risk of lymph node metastasis (pooled OR=4.17, 95% CI=1.20-14.45), and high tumor histological grade (pooled OR=2.28, 95% CI=1.72-3.03). Moreover, a borderline significance between positive Sox2 expression and high tumor stage was also revealed (pooled OR=2.94, 95% CI=0.90-9.65). However, no significant associations between Sox2 expression and ER, PR, and HER2 status were found in breast cancer patients (pooled OR=0.99, 95% CI=0.38-2.60; OR=0.89, 95% CI=0.34-2.36; OR=1.36, 95% CI=0.94-1.97, respectively). When considering the subtype of breast cancer, positive expression of Sox2 was found to be significantly associated with the biologically aggressive triple-negative phenotype (pooled OR=2.64, 95% CI=1.11-6.29). Analysis of clinicopathological parameters revealed no obvious evidence of asymmetry in the funnel plots for publication bias (Figure 2), which was also supported by the Egger’s test (tumor size, P=0.24; histological grade, P=0.185; TNM stage, P=0.789; lymph node metastasis, P=0.053; ER status, P=0.515; PR status, P=0.468; HER2 status, P=0.829; subtype, P=0.121).

Figure 1.

Meta-analysis of Sox2-positive expression and the clinicopathological characteristics of patients with breast cancer.

Table 2.

Meta-analysis results of the association between Sox2 expression and breast cancer

Variables	Number of studies	Overall	P-value	Heterogeneity		OR (95% CI) of subgroup analysis
		OR (95% CI)		Ph	I2	Ethnicity (Asian/Caucasian)	Sample size (≥100/<100)	Quality (≥6/<6)
Tumor size	7	2.61 (1.91-3.58)	<0.001	0.76	0%	2.94 (1.98-4.37)/2.08 (1.22-3.54)	3.10 (2.08-4.62)/1.93 (1.15-3.23)	2.71 (1.92-3.83)/2.17 (1.02-4.64)
Histological grade	8	2.28 (1.72, -3.03)	<0.001	0.28	19%	2.16 (1.50-3.10)/2.42 (1.37-4.27)	1.98 (1.43-2.74)/3.46 (1.34-8.97)	2.05 (1.51-2.78)/3.81 (0.79-18.41)
Tumor stage	4	2.94 (0.90-9.65)	0.07	0.06	60%	-	-	-
LNM	6	4.17 (1.20-14.45)	0.02	<0.001	91%	7.49 (0.99-56.89)/0.86 (0.29-2.54)	6.35 (0.61-66.34)/2.99 (0.81-11.02)	3.72 (0.68-20.39)/5.84 (2.63-12.95)
ER status	5	0.99 (0.38-2.60)	0.98	<0.001	85%	-	-	-
PR status	4	0.89 (0.34-2.36)	0.81	<0.001	86%	-	-	-
HER-2 status	6	1.36 (0.94-1.97)	0.1	0.88	0%	1.27 (0.76-2.13)/1.16 (0.54-2.48)	1.35 (0.90-2.02)/1.42 (0.58-3.52)	-
Subtype	5	2.64 (1.11-6.30)	0.03	0.01	69%	-	-	-

LNM, lymph node metastasis; P _h, P-value of heterogeneity.

Figure 2.

Funnel plot of publication biases on the relationships between Sox2-positive expression and the clinicopathological characteristics of patients with breast cancer.

As shown in Table 2, the subgroup analyses based on ethnicity, sample size and quality were also conducted when more than six studies were included in the overall analyses regarding the concerned variables. As a result, the association between Sox2 expression and tumor size remained significant regardless of the ethnicity, sample size and quality variations (all P<0.05). Similar association was also observed regarding tumor histological grade in all subgroup analyses (all P<0.05) except for the lack of the relationship in studies with low quality (P>0.05). Unfortunately, no significant association regarding lymph node metastasis was revealed in any subgroup analyses except for that in studies with low quality (P<0.05), however, the results was presented with obvious heterogeneity. Moreover, the stratification analysis by ethnicity and sample size also failed to show any significant association of Sox2 expression and HER-2 status (all P<0.05).

Sensitivity analysis

To examine the stability of overall estimates, sensitivity analysis was performed by sequentially eliminating studies one by one. The removal of one study each turn showed no significant variation in overall estimates regarding the tumor size, tumor histological grade, ER and HER-2 status, suggesting the robustness of these results. Furthermore, we found a slight fluctuation of the p-value around 0.05 in the sensitivity analysis. Nevertheless, a significant association was suggested by most results of sequential omission. However, the overall estimate for the association of Sox2 expression and lymph node metastasis was significantly altered by excluding the studies by Li et al. [19] and Xu et al. [21], respectively. Similarly, the significance of pooled OR was excessively affected regarding the tumor stage by omitting the study by Nagata et al. [17] and regarding PR status by omitting the study by Abd et al. [6], suggesting the necessity to validate these associations with more well-designed studies.

Discussion

To the best of our knowledge, the current meta-analysis is the first study to systematically estimate the associations between Sox2 expression and the clinicopathological characteristics of breast cancer. Our pooled results of 10 studies involving 1713 cases provide compelling evidence of a significant positive correlation between Sox2 protein expression and increased tumor size, higher histological grade, lymph node metastasis, and the highly aggressive triple-negative phenotype of breast cancer.

As transcription factors, all Sox proteins, including Sox2, do not possess sufficient affinities for DNA binding to be able to activate or repress target gene expression [23]. Instead, it is currently generally accepted that the regulatory effects of the Sox2 are mediated through interactions with partner proteins such as cytokeratin (CK) 5/6, epidermal growth factor receptor (EGFR), cyclin D1, the epithelial mesenchymal transition-related marker vimentin, and several other proteins and pathways [8,19,24,25]. Although the molecular basis of Sox2 involvement in the development of breast cancer remains largely unknown, accumulative evidence indicates that it may play roles in the regulation of cell fate determination, differentiation, proliferation, and neoplastic transformation [4]. Chen et al. found that the expression of Sox2 was significantly higher in 56 breast cancer samples compared to in 19 normal tissues, and a similar result was observed in an in vitro study, suggesting a role of Sox2 in tumorigenesis [23]. Based on our results in the present meta-analysis, we conclude that patients with high levels of Sox2 protein expression in their tumors are more likely to develop larger tumors. This may be explained by the previously suggested mechanism that overexpression of Sox2 in MCF-7 breast cancer cells resulted in enhanced cell proliferation and tumorigenesis through upregulation of cyclin D1, and by subsequently facilitating G1/S transition of the cell cycle [23]. Moreover, in agreement with most previous studies demonstrating that positive Sox2 expression contributes to an increased risk of lymph node metastasis [6,19-21], this association was confirmed in our pooled analyses, although we failed to represent this association in the subgroup analyses, which may be due to the considerable heterogeneity. In this regard, Li et al. demonstrated that Sox2 results in increased metastasis of breast and prostate cancer cells by promoting epithelial-to-mesenchymal transition through Wnt/β-catenin signaling [19], indicating a potential underlying mechanism for the role of Sox2 in breast cancer metastasis. Furthermore, as the Sox-induced generant repression of differentiation into neurons [7], a previous study revealed that the expression of embryonic transcriptional factors, including Sox2, may be highly associated with less differentiated breast cancer [26]. Similarly, Rodriguez-Pinilla et al. reported that higher Sox2 expression was more commonly observed in basal cell-like breast carcinomas compared to other subtypes, and that it strongly correlated with CK5/6, EGFR, and vimentin immunoreactivity, suggesting that Sox2 may play a role in conferring a less differentiated and a more aggressive phenotype of breast cancer [8]. In keeping with their findings, our combined results revealed that positive Sox2 expression was significantly associated with a high tumor grade and the triple-negative subtype, altogether indicating a potential role of Sox2 in the pathological diagnosis, prognosis, and clinical procedures of breast cancer.

The statistical procedures used in the present meta-analysis may account for the varying degrees of reliability across individual studies by weighting the effect size from any one study by its sample size [27]. Although our systematic review was robust in identifying the significance of Sox2 in breast cancer, several limitations still need to be addressed. First, the present meta-analysis was restricted to studies published in English and Chinese only, which may cause selection bias, however, studies reported in other languages are not usually available for investigators. Second, the immunostaining cutoff points and methodological designs were arbitrarily selected and varied between the analyzed studies, and variability in protein expression assessment may be a cause of potential bias, which would limit our confidence in drawing conclusions. Third, the histological type of breast cancer may contribute to the studied associations [18], as it was not unified in the included studies of the meta-analysis, which may also potentially influence the results. Lastly, Sox2 usually exerts its transcription regulation function through interactions with partner proteins [4,28], and analyses of any potential interactions between these other proteins and Sox2 in breast cancer were neglected due to the absence of original data. Additional well-designed studies with a unified definition in terms of the protein expression assessment, uniform cases, and larger sample sizes are needed to present more reliable results.

In summary, the present meta-analysis provides a relative outline of Sox2 and a number of common clinicopathological parameters, such as tumor size, differentiation, lymph node status, and molecular subtypes, in patients with breast cancer. Based on our results, we believe that the detection of Sox2 expression may be of great value in determining the diagnosis and prognosis of breast cancer patients. However, given the limitations of our meta-analysis, the findings presented herein warrant future investigations to validate the diagnostic and prognostic value of Sox2 using well-designed studies and larger sample sizes.

Disclosure of conflict of interest

None.