Expression of the GLI family genes is associated with tumor progression in advanced lung adenocarcinoma

Background The hedgehog (Hh) signaling pathway is aberrantly activated in various cancers. Expression of the GLI family of genes, which encode for transcriptional factors of the Hh pathway, has not been fully assessed in clinical samples of advanced lung adenocarcinoma. In this study, we retrospectively evaluated the expression of the GLI family of genes in advanced stage lung adenocarcinoma samples and determined their relation to patient survival. Methods The levels of GLI1, GLI2, and GLI3 mRNA expression were measured by quantitative real-time polymerase chain reaction in surgically obtained tissue samples from stage II-IV lung adenocarcinoma patients (n = 102). Pairwise comparisons between all three GLI mRNA expression were performed, and after dichotomizing the patients into low and high expression groups according to each GLI mRNA expression level, survival curves were calculated and multivariate analyses were conducted. Results Significant positive correlation was found between GLI1 and GLI3 mRNA expression (P <0.001). Tumors with higher expression (upper 15%) of GLI1 or GLI3 mRNA were associated with poor survival in stage II-IV (5-year overall survival rates: GLI1 mRNA low, 41.7% vs. high, 20.0%, P = 0.0074; GLI3 mRNA low, 43.1% vs. high, 13.3%, P = 0.0062) and stage III-IV (5-year overall survival rates: GLI1 mRNA low, 34.0% vs. high, 0%, P = 0.0012; GLI3 mRNA low, 33.4% vs. high, 7.7%, P = 0.057) lung adenocarcinoma patients. GLI2 mRNA expression did not appear to have great clinical significance. Multivariate analysis revealed higher GLI1 mRNA expression as an independent factor for unfavorable patient survival (P = 0.0030, hazard ratio = 3.1, 95% confidence interval = 1.5-6.2), as well as tumor differentiation and stage. Conclusions Expression of GLI1 and GLI3 mRNA was strongly correlated, and their overexpression, especially that of GLI1, was found to be predictive of aggressive tumor behavior. This study indicates that the Hh pathway may be a key oncogenic signaling network in tumor pathogenesis and, thus, a potential therapeutic target in advanced lung adenocarcinoma.

Background Non-small cell lung cancer (NSCLC) has been the leading cause of cancer-related deaths world-wide, and adenocarcinoma accounts for the majority of all NSCLC cases [1,2]. Although any possibility of a cure rests almost entirely with surgical resection with or without chemo-or radiotherapy, postoperative recurrence rate is still high and survival rate remains low compared with other types of cancers. In particular, advanced NSCLC can be exceptionally lethal even with aggressive anticancer therapies, despite the dramatic treatment shift of some targeted therapeutic agents for lung adenocarcinoma [3,4]. Developing and deploying new treatment approaches are vital to improve patient prognoses.
The hedgehog (Hh) signaling pathway, which consists of three distinct homologues (sonic, indian, and desert Hh) is constitutively activated in a variety of human tumor entities [5]. Normally, Hh signaling plays an important role during vertebral embryonic development and tissue maintenance [6]. The Hh proteins are secreted molecules, and their receptor, Patched, usually functions as a downstream inhibitor in the absence of Hh. Hh signaling is activated through binding to the Patched receptor, which unleashes the transmembrane protein, Smoothened, and ultimately upregulates GLI [5]. GLI proteins are zinc finger transcription factors which exert their functions by nuclear localization and promoting target gene transcription, and in humans, at least three distinct GLI genes, GLI1, GLI2, and GLI3, have been identified [7]. As the GLI family of proteins are the final mediators of Hh signaling, their expression, especially that of GLI1, is deemed to be the best reflection of the net effect of Hh activation, thus, survival studies on GLI may indicate the potential involvement of Hh in tumor pathogenesis [8].
In addition to the tumorigenesis based on genetic alterations of the Hh pathway predominantly seen in basal cell carcinoma and medulloblastoma, extensive involvement of ligand-dependent Hh activation has been suggested in various cancer entities [9,10]. With regard to lung cancer, while the involvement of Hh signaling in small cell lung cancer has been shown to be evident [11][12][13], it remains poorly understood in lung adenocarcinoma, especially in the advanced stages.
In this study, we evaluated the expression levels of the three GLI family genes (GLI1, GLI2, and GLI3) in surgical samples from advanced lung adenocarcinoma patients (stage II-IV), and revealed the possible involvement of the Hh signaling pathway in tumor advancement by analysis for correlations between GLI expression and patient prognosis. As we found considerable impact of GLI1 and GLI3 gene expression on patient survival, this study demonstrates that inhibition of Hh signaling may be a promising drug target for the treatment of advanced lung adenocarcinoma.

Patients, tissue samples, and tumor information
Tissue samples were obtained from patients who were diagnosed as stage II-IV lung adenocarcinoma after surgical resection at the Kyoto University Hospital between July 2001 and December 2007. Continuous surgical cases with sufficient tissue material for evaluations were included in this study (n = 102). All of these tumors were histologically confirmed as lung adenocarcinoma, and their tumor staging and degree of differentiation were all assessed by board-certified pathologists in the Department of Pathology of the Kyoto University Hospital. Tumor stage was determined by the latest tumor-node-metastasis classification system [14]. Histological type and grade of cell differentiation were determined according to the WHO classification system [15]. Their demographics and baseline characteristics are presented in Table 1. Pre-or postoperative chemotherapy was mainly composed of orally administered tegafur-uracil and/or conventional intravenous chemotherapeutic agents (carboplatin, paclitaxel, docetaxel, gemcitabine, vinorelbine) for NSCLC. Only one patient (stage II) received radiotherapy prior to surgery (concurrently with chemotherapy), and postoperative radiotherapy (mediastinum or chest wall) was performed in four patients (stage III). Complete tumor resection was achieved in 77 patients (75.5%), whereas in 25 patients (24.5%) macro-or microscopic residual tumors were suspected after surgery (for example, pleural dissemination, malignant pleural effusion, positive stumps, or distant metastasis prior to surgery). Informed consent for participation in this study was obtained from all patients prior to their surgeries, and this study was reviewed and approved by the Ethics Committee of the Graduate School and Faculty of Medicine at the Kyoto University.

Preparation of tissue mRNA
For sample collection, tumor tissue samples were dissected immediately after surgical resection and soaked in RNAlater TissueProtect Tubes (Qiagen, Tokyo, Japan) for more than 48 h before storage at -80°C until use. Total RNA was isolated from tissue samples using RNeasy Plus Mini Kit (Qiagen), and reverse transcription of total RNA was conducted using the Ready-To-Go You-Prime First-Strand Beads (Amersham Biosciences, Uppsala, Sweden) to obtain cDNA.

Quantification of GLI mRNA
To quantify GLI mRNA expression levels of each sample, quantitative real-time PCR was performed using the LightCycler thermal cycler system (Roche Diagnostics Japan, Tokyo, Japan). The PCR primers used for the quantitative amplification of GLI1 mRNA were forward: 5′-CTCCCGAAGGACAGGTATGTAAC -3′ and reverse: 5′-CCCTACTCTTTAGGCACTAGAGTTG -3′, those of GLI2 mRNA were forward: 5′-AGAAGCAGCGCAAT GACGTG -3′ and reverse: 5′-GTCATCCAGTGCCGT CAGGT -3′, and those of GLI3 mRNA were forward: 5′-AAACCCCAATCATGGACTCAAC -3′ and reverse: 5′-TACGTGCTCCATCCATTTGGT-3′. The primers for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA, used as an internal control, were forward: 5′-ACAACAGCCTCAAGATCATCAG-3′ and reverse: 5′-TCTTCTGGGTGGCAGTGATG-3′. After a 20-μL reaction mixture containing 0.5 μM forward/reverse primers and 0.03 μg cDNA in QuantiTect SYBR Green PCR Master Mix (Qiagen) was prepared, PCR amplification was initiated by preincubation for 15 min at 95°C for initial activation, followed by 40 cycles of the following protocol: denaturation at 94°C for 15 s, annealing at 59°C for 15 s, and elongation at 72°C for 15 s with detection of fluorescence products. The quantitative data were analyzed with LightCycler analysis software version 5.03 (Roche Diagnostics Japan). The expression levels of GLI genes were calculated as the ratio of GLI mRNA value to GAPDH mRNA value, and are expressed as median and mean ± standard deviation.

Statistical analysis
Relationships between each pair of mRNA expression were estimated by calculating the Spearman's rank correlation coefficient (r s ). Pairwise comparisons of postoperative survival of each GLI low and high groups were compared using the Mann-Whitney U test. Survival curves were evaluated by the Kaplan-Meier method, and cumulative survival data were compared using the logrank test. Multivariate analysis of prognostic factors was performed by the Cox proportional hazard model. Differences were considered significant when P <0.05. All statistical analyses were performed using StatMate IV software version 4.01 (ATMS, Tokyo, Japan) and JMP software version 8 (SAS Institute Japan, Tokyo, Japan).
GLI1 and GLI3 mRNA expression levels were remarkably correlated By comparing GLI1, GLI2, and GLI3 mRNA expression levels, strong positive correlation was observed between GLI1 and GLI3 mRNA expression (r s = 0.46, P <0.001) (Figure 2), whereas no such relations were found between GLI2 and GLI1 or GLI3 mRNA expression levels.
Even with exclusion of the four observed outliers (n = 98), strong positive correlation was also found between GLI1 and GLI3 mRNA expression (r s = 0.40, P <0.001).
Higher GLI1 mRNA expression was correlated with significantly worse prognosis especially in stage III-IV lung adenocarcinoma patients Subset analysis of patients with stage III-IV lung adenocarcinoma (n = 71) showed significantly worse clinical outcomes in both the higher GLI1 and GLI3 mRNA expression groups. In this cohort, 5-year OAS rates in the GLI1 mRNA low and high groups were 34.0% and 0%, respectively (P = 0.0012, HR = 2.8 (1.9-13.5)), and 5-year DFS rates in these groups were 18.3% and 0%, respectively (P = 0.027, HR = 2.1 (1.1-7.2)). Although the difference in OAS between the GLI3 mRNA expression groups was marginal (low, 33.4% vs. high, 7.7%, P = 0.057), 5-year DFS rates were statistically different between GLI3 mRNA low and high groups (low, 17.6% vs. high, 7.7%, P = 0.030, HR = 1.9 (1.1-5.3)). Survival curves are presented in Figure 4. Higher GLI1 mRNA expression was an independent prognostic factor in advanced lung adenocarcinoma The results of multivariate analyses for each clinicopathological parameter are presented in Table 2. Higher GLI1 mRNA expression was demonstrated to be an independent prognostic factor for poor patient prognosis (P = 0.0030, HR = 3.1 (1.5-6.2)) as well as tumor differentiation (well vs. poor) and tumor stage. No correlations were found for GLI2 and GLI3 mRNA expression with other clinicopathological features.

Discussion
Previous reports on Hh molecules have already provided us with convincing information on its etiologic contribution to NSCLC [16]. A large percentage of primary NSCLC cell lines have been shown to express Hh target genes including GLI1, indicating constitutive activation of the Hh pathway in NSCLC cell lines [17]. It was recently shown that disrupting the Hh pathway inhibited NSCLC proliferation in vitro and abrogated tumor growth in vivo [18]. In clinical samples, immunohistochemical analysis of tissue microarray have revealed significant correlation among the Hh molecules in early-stage NSCLC and of lymph node metastasis with nuclear GLI1 immunolocalization in lung adenocarcinoma [19,20]. However, the impact of Hh signaling on patient survival has not been previously reported in NSCLC. In this study, expression of the three GLI genes was demonstrated in almost all cases in this cohort, although their expression levels varied widely. Interestingly, some lung adenocarcinoma samples showed relatively high GLI1 and GLI3 mRNA expression, and their expression patterns were considerably correlated. Although GLI1 is considered to be the major effector in Hh signaling, the role of GLI3 is still controversial. Whereas GLI3 acts mainly as a repressor in normal Hh signaling [21], some reports have suggested its activator capacity in oncogenesis [22][23]. The present study also provides evidence of the capacity of GLI3 to activate Hh signaling in concert with GLI1 in advanced lung adenocarcinoma. With regard to GLI2, very little is characterized about its expression in NSCLC, and only correlation between nuclear GLI2 with sonic Hh expression in lung squamous cell carcinoma has been reported [19]. From our research, no relation between GLI2 mRNA expression and GLI1 mRNA level, GLI3 mRNA level, or patient survival was detected, thus, the role of GLI2 in the pathogenesis of lung adenocarcinoma is still unclear. While GLI1 is mainly regulated at the transcriptional level, GLI2 is considered as a latent transcriptional regulator that can be activated by Hh signaling. Modification of the N-terminus of GLI2 is a critical step in the regulation of the transcriptional activity of the protein [8]. This as well as the difference in cancer types may explain the discrepancies in GLI2 expression and patient outcomes between our results and previous ones [24]. Our research is based on the genetic expression of GLI, whereas for GLI2, it may be necessary to assess its functional protein level instead.
The mechanisms by which Hh signaling is involved in cancer pathogenesis are not uniform, and there are several different models of Hh signal transduction. Besides the most evident pattern of ligand-independent mutational Hh activation seen in basal cell carcinoma and medulloblastoma, ligand-dependent autocrine Hh activation in tumor cells has also been shown to be common in many cancers including NSCLC. In addition, ligand-dependent paracrine Hh signaling that involves the tumor microenvironment can lead to tumor development [25]. Crosstalk with other developmental cascades such as Wnt or Notch is also frequently observed [26]. This diversity of Hh signaling in cancer development in part explains the differences across cancer types [27,28].
Previous reports on expression profiles of Hh-related molecules in NSCLC specimens had focused mainly on the estimation of protein levels by immunohistochemistry [19,20], and their gene expression has not been thoroughly investigated. This study is unique in that thorough investigation was performed for the gene expression of all three GLI family members, which enabled us to directly compare their levels in clinical samples and to reveal the unexpected correlation between GLI1 and GLI3 mRNA expression. Interestingly, some GLI1/GLI3 highly expressing tumors (in this cohort, defined as the upper 15th percentile of patients for each mRNA expression) showed significantly poorer patient prognosis in advanced lung adenocarcinoma. Although a statistically significant negative correlation between survival and GLI1 expression has been shown in breast [29], esophageal [30], colorectal [31], and bladder cancers [32], this is the first study to demonstrate a clinical impact of GLI1 on patient survival in lung adenocarcinoma. Our speculation was further supported by the multivariate analysis, which showed that GLI1 mRNA expression is an indicator for unfavorable prognosis. Although the independency of Hh molecule expression on patient survival has been previously shown in bladder and prostate cancers [32,33], we are the first to reveal that of GLI1 in lung adenocarcinoma, which is indicative of the involvement of Hh signaling in a subset of NSCLC. Investigation on GLI3 expression in tumor samples has been rare, and the possible clinical impact of GLI3 expression raised by our study needs further corroboration by additional research studies. Irrespective of these novel observations, some limitations should be addressed in the present study. First, the validity of the cutoff of mRNA values was not ensured. Unfortunately, for our data, dichotomizing patients by either median or mean value of mRNA expression provided little information on survival difference, thus, we defined the upper 15% expression level as the cutoff. Further, patient backgrounds were diverse. Due to the retrospective nature of analysis of these advanced stage cancer cases, pre-or postoperative therapies were not standardized and the completeness of resection was not fully warranted (for example, in stage IV patients, operations were mainly performed for salvage or diagnosis). We believe that these limitations can be overcome by expanding patient numbers and/or careful selection of patients with similar clinical backgrounds. Additionally, analyzing the expression profiles of GLI proteins as well as that of other Hh molecules in this cohort would further facilitate the deepening of our understanding.
With the advent of small molecule Hh inhibitors, such as vismodegib [34,35], recent interest has focused on the clinical application of Hh inhibitors in various types of tumors harboring aberrantly activated Hh signaling [25,36]. Based on our results, Hh inhibition may be a good candidate treatment for advanced lung adenocarcinoma. Because a similar study on stage I-II lung adenocarcinoma did not show any significance on patient survival [19], this signaling pathway may be involved more in tumor progression rather than tumorigenesis, and disrupting Hh may be more efficacious in the advanced stages of lung adenocarcinoma. Further studies including analysis of the expression of other Hh molecules or on other histological types of NSCLC are necessary to provide extended support of our findings.

Conclusions
We found that GLI1, GLI2, and GLI3 mRNA were expressed in almost all advanced stage lung adenocarcinoma tissue samples, and strong correlation was observed between GLI1 and GLI3 mRNA expression. While GLI2 mRNA expression level was of little clinical significance, a subset of patients showing relatively high GLI1/GLI3 expression showed poorer patient survival. In particular, higher GLI1 mRNA expression was an independent prognostic factor for unfavorable prognosis. This is the first report to describe the clinical involvement of the GLI family of genes in advanced lung adenocarcinoma, and these results suggest that Hh signal inhibition is a good candidate therapeutic target for advanced stage lung adenocarcinoma, which provide an impetus for further investigations on the Hh signaling in NSCLC.