Prognostic Value of LGR5 in Colorectal Cancer: A Meta-Analysis

Objective Leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5) has recently been reported to be a marker of cancer stem cells (CSCs) in colorectal cancer (CRC), and the prognostic value of LGR5 in CRC has been evaluated in several studies. However, the conclusions remain controversial. In this study, we aimed to evaluate the association between the expression of LGR5 and the outcome of CRC patients by performing a meta-analysis. Methods We systematically searched for relevant studies published up to February 2014 using the PubMed, Web of Science, EMBASE and Wangfang databases. Only articles in which LGR5 expression was detected by immunohistochemistry were included. A meta-analysis was performed using STATA 12.0, and pooled hazard ratios (HRs) with 95% confidence intervals (CIs) were used to estimate the strength of the association between LGR5 expression and the prognosis of CRC patients. Results A total of 7 studies comprising 1833 CRC patients met the inclusion criteria, including 6 studies comprising 1781 patients for overall survival (OS) and 3 studies comprising 528 patients for disease-free survival (DFS). Our results showed that high LGR5 expression was significantly associated with poor prognosis in terms of OS (HR: 1.87, 95% CI: 1.23–2.84; P = 0.003) and DFS (HR: 2.44, 95% CI: 1.49–3.98; P<0.001). Further subgroup analysis revealed that many factors, including the study region, number of patients, follow-up duration and cutoff value, affected the significance of the association between LGR5 expression and a worse prognosis in patients with CRC. In addition, there was no evidence of publication bias, as suggested by Begg’s and Egger’s tests. Conclusions The present meta-analysis indicated that high LGR5 expression was associated with poor prognosis in patients with CRC and that LGR5 is an efficient prognostic factor in CRC.


Introduction
Colorectal cancer (CRC) is the most common malignancy of the gastrointestinal tract worldwide. As one of the leading causes of cancer-related mortality [1], CRC accounts for more than 600,000 deaths every year [2]. Despite advances in curative surgery and adjuvant therapy, as well as extensive CRC-focused research over the past 20 years, the 5-year survival rate is still poor [3]. Relapse, metastasis and drug resistance are the main factors contributing to the high mortality and poor survival rate of this disease [4]. Increasing evidence suggests that a population of selfrenewing tumor cells, known as cancer stem cells (CSCs), is responsible for tumor progression, relapse, metastases and therapeutic resistance [5,6]. Therefore, the identification of CSCs is crucial in the search for therapeutic targets and useful prognostic markers for CRC.
Becker et al. suggested that leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5) may be a better marker of CSCs in CRC [7]. LGR5 was initially identified as an orphan G protein-coupled receptor (GPCR) that belongs to the subfamily of glycoprotein hormone receptors [8], and it contains a large extracellular domain with 17 leucine-rich repeats and a seventransmembrane domain. Recently, elevated LGR5 expression has been observed in several types of cancers, including hepatocellular carcinoma [9], CRC [10], ovarian cancer [11], and basal cell carcinoma [12]. In particular, many studies have suggested that LGR5 plays a key role in colorectal carcinogenesis and is associated with the poor outcome of CRC patients [13][14][15][16][17][18]. Although LGR5 allelic variation can affect LGR5 protein expression in colorectal cancers, the somatic LGR5 genotype seems to be relatively stable in primary tumors. Moreover, patients with variant alleles in SNPs of the LGR5 gene showed similar prognosis as patients with wild type LGR5, no significant difference was observed [19]. Therefore, it was expected that LGR5 expression in CRC is an ideal prognostic marker that is correlated with low survival.
In fact, in recent years, many studies have shown that the expression of LGR5 is positively associated with poor prognosis in CRC [13,15,17]. However, no correlation was found between the expression of LGR5 and a poor clinical outcome in CRC in another previous study [20]. The prognostic value of LGR5 in CRC patients is controversial, and an insufficient sample size and several other factors likely resulted in the contrary results of different clinical studies. However, to date, there has been no meta-analysis of LGR5 expression and the prognosis of patients with CRC. To clarify the exact prognostic value of LGR5 in CRC, we performed a meta-analysis of eligible studies to investigate the relationship between LGR5 expression and the prognosis of CRC patients.

Selection criteria
The inclusion criteria for the eligibility of a study were as follows: (1) patients with distinctive CRC diagnosis by pathology, (2) an assessment of the relationships between LGR5 expression and the prognosis of CRC patients, (3) sufficient information provided to estimate hazard ratio (HR) for overall survival (OS) or disease-free survival (DFS), and (4) publication with English or Chinese. The following studies were excluded from the analyses: (1) letters, reviews and conference abstracts, due to the limited data, and (2) articles about animal or cell lines. Regarding multiple publications from the same population, only the most recent or the most complete study was included in the analysis.

Data extraction and quality assessment
Two investigators (Qing Chen and Xin Zhang) reviewed each eligible article independently and extracted information from all the publications meeting the inclusion criteria. The following characteristics were collected from each study: the first author's name, the year of publication, the country of origin, the number of patients, the age of the patients, the time of follow-up, the disease stage, the cutoff value, LGR5 expression levels and survival data. Disagreements were resolved by discussion until consensus was reached.
Quality assessment was conducted for each of the available studies using the Newcastle-Ottawa quality assessment scale [21]. The score assessed eight items on methodology that were LGR5 and the Prognosis of Colorectal Cancer PLOS ONE | www.plosone.org categorized into three dimensions, including selection, comparability and outcome. Interpretation of the scale is performed by awarding points, or ''stars'', for high-quality elements. The stars are then added up and used to compare study quality in a quantitative manner. We specified a priori that a score of 7 or higher indicated high quality, a score of 5 or 6 indicated moderate quality, and a score of 4 or less indicated low quality.  LGR5 and the Prognosis of Colorectal Cancer PLOS ONE | www.plosone.org

Statistical analysis
Survival outcome data were synthesized using the HR and its 95% confidence interval (CI) to analyze the impact of LGR5 expression on the survival of CRC patients. Several of the included studies provided HRs and 95% CIs, which we pooled directly. Otherwise, we calculated the HR and its 95% CI from available data or a Kaplan-Meier survival curve using Engauge Digitizer version 4.1 (free software downloaded from http://sourceforge. net). By convention, an observed HR.1 implied worse survival for a group with increased LGR5 expression. The impact of LGR5 expression on survival was considered to be statistically significant if there was no overlap of the 95% CI with 1. The heterogeneity among the studies was assessed by a chi-square-based Q statistic test [22], and the I 2 value was used to quantify the heterogeneity (I 2 = 0-50%, no or moderate heterogeneity; I 2 .50%, significant heterogeneity) [23]. If homogeneity was not significant (P.0.10 for the Q test), the fixed-effect model was used; otherwise, the random-effect model was used. Publication bias was assessed using Egger's test and Begg's test [24,25]. To adjust for multiple comparisons, we applied the stepdown Bonferroni method, which control for familywise error rate (FEW). In addition, sensitivity analysis was performed to examine the stability of the pooled results. The statistical analyses were conducted using STATA 12.0. All P values were two-sided, and P,0.05 was considered to be statistically significant.

Study selection and characteristics
As shown in Fig. 1, a total of 58 articles were initially retrieved from the above databases using the search strategy described above, and the details of search results in different databases were shown in Table S1. After reviewing the titles and abstracts of the articles, 41 articles that were irrelevant to our aim were excluded. The remaining 16 articles were then independently scrutinized by two of the authors. Of the articles, 10 articles were excluded: full text could not be found for 2 articles [13,26], 1 article was about LGR5 expression in the peripheral blood [18], 4 articles did not provide OS or DFS data [10,[27][28][29], 2 articles were about variants or polymorphisms of LGR5 [29,30], and 1 article duplicated a cohort of patients [17]. Eventually, 7 articles met the inclusion criteria and were included in the meta-analysis [14][15][16]20,[31][32][33].
All 7 eligible articles evaluated the correlation between LGR5 expression in CRC tissues and the prognosis of CRC. The major characteristics of the 7 studies are summarized in Table 1. These studies were published from 2010-2014 and conducted in 3 countries (the USA, Japan and China). The total number of enrolled patients was 1833, with individual samples ranging from 52-891 (median 180). The reported mean age of the patients ranged from 57-66.9 years across the eligible studies. The followup period ranged from 28.3-180 months. Several of the studies defined the cutoff value using complex scoring metrics combining the intensity and percentage of LGR5 expression, whereas other studies only used the percentage of LGR5 expression. Positive LGR5 expression was observed to range from 36-60%. There were 2 studies that utilized both OS and DFS to assess the prognostic value of LGR5 expression in CRC patients [16,32].  LGR5 and the Prognosis of Colorectal Cancer PLOS ONE | www.plosone.org Additionally, 4 studies used only OS as an indicator [15,20,31,33], and 1 study used only DFS [17]. HRs and 95% CIs were directly obtained from 5 studies, and for the remaining 2 studies, HRs and 95% CIs were extrapolated from Kaplan-Meier curves. Overall, the studies included in mate-analysis were of high quality, and of the 7 studies, 3 (43%) scored 8, 3 (43%) scored 7 and 1 (14%) scored 6.
LGR5 expression and prognosis of CRC 6 out of the total of 7 studies reported data on LGR5 expression and OS in CRC [15,16,20,[31][32][33], so the combined data from all 6 studies were pooled in the meta-analysis. As shown in Fig. 2, due to a significant degree of heterogeneity (I 2 = 80.1%, P,0.001), the pooled HRs and 95% CIs were calculated using the random-effect model (REM). The results showed that high LGR5 expression was associated with poor OS in CRC (HR: 1.87, 95% CI: 1.23-2.84, P = 0.003, REM). Additionally, the pooled HR for OS was 1.34 (95% CI: 1.17-1.54, P,0.001) calculated by the fixed-effect model (FEM). 3 of 7 studies reported data on LGR5 expression and DFS, and as shown in Fig. 3, high LGR5 expression was significantly correlated with DFS, with a pooled HR estimate of 2.44 (95% CI: 1.49-3.98, P,0.001, FEM). No significant heterogeneity was observed (I 2 = 0%, P = 0.649).
We conducted a subgroup analysis of the association of LGR5 expression with OS by study region, patient number, age, followup, tumor stage and cutoff value, and the main results are shown in Table 2. The results showed that LGR5 expression was significantly correlated with OS (HR: 2.52, 95% CI: 1.75-2.89) in patients from Asian countries, but not in patients from non-Asian countries (HR: 1.09, 95% CI: 0.93-1.28), here, we should note the fact that only study from non-Asian countries (USA) was included. The subgroup analysis showed a significant correlation between LGR5 expression and OS in studies in which the number of patients was less than 200 (HR: 2.22, 95% CI: 1.66-2.98). The subgroup meta-analysis of the studies with a follow-up period of less than 60 months indicated that high LGR5 expression was a predictor of poor OS (HR: 2.38, 95% CI: 1.79-3.15). When grouped according to the stage of CRC, a significant relationship between LGR5 expression and OS was observed in both the early stage (HR: 2.27, 95% CI: 1.52-3.38) and the later stage (HR: 1.70, 95% CI: 1.01-2.87) of CRC. We also observed a statistically significant effect of LGR5 expression on OS based on the studies using a cutoff .6, with an HR of 2.09 (95% CI: 1.36-3.21). Additionally, the results also showed that the high LGR5 expression was independent predictor of poor OS in CRC patients with the mean age #60 (HR: 2.27, 95% CI: 1.52-3.38).
Because of the low number of studies, the subgroup analysis of the association of LGR5 expression with DFS was not further pursued.

Publication bias analysis
In this meta-analysis, publication bias was tested using Begg's funnel plot and Egger's test. As shown in Figure 4, the funnel plot presented no proof of obvious publication bias in any of the included studies for OS or DFS, and Egger's test also showed no obvious publication bias in the studies for either of the two outcomes (OS, P = 0.471; DFS, P = 0.749; Fig. 5).

Sensitivity analysis
To gauge the stability of the result, a sensitivity analysis was conducted, in which one study was deleted at a time. As shown in Table 3, the corresponding pooled HR for OS did not significantly change, regardless of which study was deleted, suggesting that the result was robust. However, the pooled HR for DFS appeared to be significantly altered, which may have been due to the insufficient number of studies.

Discussion
CRC is a malignant disease with high mortality worldwide, and its prognosis is still poor, although tremendous progress has been achieved [4]. Recently, alteration of molecular biological markers in tumor tissues has become an important part of predicting the prognosis of patients with malignant tumors. LGR5, known as a target of the Wnt signaling pathway, has been reported to be a potential marker of stem cells in the small intestine and colon [34,35]. Accumulated studies have shown that LGR5 is closely associated with tumorigenesis and tumor invasion in CRC and is likely to be a relevant marker of CSCs in CRC [12,36]. CSCs are thought to be responsible for the local invasion, metastasis and recurrence of malignant tumors because of their self-renewal and multi-differentiation potential and thus are also a major obstacle to improving overall cancer survival.
In recent years, the correlation between LGR5 expression and the survival of patients has been explored in many studies due to the key role of LGR5 in tumorigenesis. High LGR5 expression has been extensively reported to be an unfavorable prognostic indicator in various human cancers [13,15,33,37]. However, certain reports have also suggested contrary results for the correlation between LGR5 expression and the prognosis of CRC. For example, Ziskin et al. reported that LGR5 expression was not significantly associated with the outcome of tumors [20]  LGR5 and the Prognosis of Colorectal Cancer PLOS ONE | www.plosone.org and was not a prognostic marker of CRC (HR: 1.09, 95% CI: 0.93-1.28). To validate the exact relationship between LGR5 expression and the prognosis of patients with CRC, a metaanalysis was performed, including recent related studies and generally using a comprehensive search strategy. To the best of our knowledge, this is the first meta-analysis that has documented the prognostic value of LGR5 in CRC. In the present metaanalysis, through combining the outcomes of 6 published studies, comprising 1781 patients with CRC, we draw the conclusion that high LGR5 expression was significantly associated with poor OS (HR: 1.87, 95% CI: 1.23-2.84, P = 0.003), moreover, the high LGR5 expression was also significantly associated with worse DFS (HR: 2.44, 95% CI: 1.49-3.98, P,0.001) by combining the outcomes of 3 studies comprising 528 patients with CRC. Additionally, the sensitivity analysis showed that the significant association between high LGR5 expression and poor OS was not altered, regardless of whether one of these studies was omitted, suggesting the robustness of this result. However, the significant association between high LGR5 expression and poor DFS was affected by omitting study, which may be due to the small number of studies. In this meta-analysis, the quality assessment was performed independently and reproducibly by two authors according to Newcastle-Ottawa guidelines. By evaluating the articles comprehensively and scientifically, we ensured that the included studies were of high-quality. Warrants caution is because the baseline characteristics of patients might have affected the conclusion of each included report, including the sample size, follow-up period, clinical stage and cutoff scores for the definition of positive staining, among other aspects. Therefore, further subgroup analysis was performed by study region, patient number, age, follow-up duration, stage and cutoff value. We found that high LGR5 expression was significantly associated with poor OS in the studies performed in Asia (HR: 2.52, 95% CI: 1.75-2.89), but not in one study done outside Asia. Therefore, further investigation is needed to verify whether the prognostic value of LGR5 in CRC is associated with the region inhabited by patients. A significant association was observed between LGR5 expression and poor OS only when the follow-up period was #60 months; this finding should be interpreted cautiously due to increased mortality with the prolongation of follow-up time. In addition, the subgroup analysis by clinical stage revealed that high LGR5 expression was significantly associated with poor OS in patients, regardless of the stage (early or late) of CRC, indicating that LGR5 was an independent predictor of prognosis during the progression of CRC. This observation was consistent with the role of LGR5 in carcinogenesis, including enriching CSCs and promoting tumor formation and progression. As there is no uniform standard to define positive LGR5 expression, the cutoff greatly differed among the studies, and the results may also have changed due to the cutoff value. The subgroup analysis showed that LGR5 expression was significantly associated with an unfavorable prognosis in patients with CRC when studies set the cutoff score at .6 (HR: 2.09, 95% CI: 1.36-3.21), suggesting that a higher cutoff score was more likely to lead to a differential conclusion. Previous studies have demonstrated that elevated LGR5 expression significantly correlates with lymphatic invasion, vascular invasion, tumor depth, lymph node metastasis, and tumor recurrence [38,39].
Publication bias is a major concern in all forms of meta-analysis. In the present study, neither Egger's test nor Begg's funnel plot showed evidence of publication bias. However, there are several limitations to the current meta-analysis, as we could not prevent all potential bias among the studies. First, the studies included in our meta-analysis were restricted to only articles published in English or Chinese, and the number was relatively small. Second, heterogeneity among the studies were found in the main analysis, maybe due to the diversity of the techniques used to identify the expression of LGR5. For example, all of the included studies were required to detect LGR5 expression by IHC, but the results of IHC greatly depended on methodological factors such as primary antibody type and antibody concentration, leading to betweenstudy heterogeneity. For this reason, random effect model and subgroup analysis were adopted to adjust for this shortcomings. Third, the method of HR and 95% CI extrapolation should be mentioned, as an HR and a 95% CI calculated from data or extracted from survival curves might be less reliable than those obtained by direct analysis of variance. Additionally, we need to consider the fact that studies with positive results are easily accepted, whereas studies with negative results are often rejected.
In summary, despite the limitations listed above, we found that high LGR5 expression may be an independent risk factor for patients with CRC. Based on currently published articles, high LGR5 expression is an unfavorable prognostic factor in CRC patients and could be helpful to optimize therapeutic schemes. Further, larger prospective studies are needed to validate our results.

Author Contributions
Conceived and designed the experiments: QC XZ PZ. Performed the experiments: QC XZ. Analyzed the data: WL YJ. Contributed reagents/ materials/analysis tools: WL HC. Contributed to the writing of the manuscript: QC. Screened the full text of selected studies to confirm eligibility, assess quality, and extract data: QC XZ.