MicroRNA and mRNA Features of Malignant Pleural Mesothelioma and Benign Asbestos-Related Pleural Effusion

Introduction. We investigated the expression of microRNAs and mRNAs in pleural tissues from patients with either malignant pleural mesothelioma or benign asbestos-related pleural effusion. Methods. Fresh frozen tissues from a total of 18 malignant pleural mesothelioma and 6 benign asbestos-related pleural effusion patients were studied. Expression profiling of mRNA and microRNA was performed using standard protocols. Results. We discovered significant upregulation of multiple microRNAs in malignant pleural mesothelioma compared to benign asbestos-related pleural effusion. Hsa-miR-484, hsa-miR-320, hsa-let-7a, and hsa-miR-125a-5p were able to discriminate malignant from benign disease. Dynamically regulated mRNAs were also identified. MET was the most highly overexpressed gene in malignant pleural mesothelioma compared to benign asbestos-related pleural effusion. Integrated analyses examining microRNA-mRNA interactions suggested multiple altered targets within the Notch signaling pathway. Conclusions. Specific microRNAs and mRNAs may have diagnostic utility in differentiating patients with malignant pleural mesothelioma from benign asbestos-related pleural effusion. These studies may be particularly helpful in patients who reside in a region with a high incidence of mesothelioma.


Introduction
Malignant pleural mesothelioma (MPM) is an aggressive tumor and remains as a significant worldwide health problem because of its poor prognosis and increasing incidence [1]. The major known risk factor is exposure to asbestos. Although MPM has a poor clinical outcome and is frequently untreatable, recent series have demonstrated that early diagnosis and aggressive treatment strategies may improve overall survival.
The detection of early stage patients with MPM and the differentiation of early stage MPM from benign asbestosrelated pleural effusion (BAPE) are critically important for improving the survival of MPM patients. The typical histologic findings of BAPE on pleural tissue are nonspecific pleuritis/fibrosis, which is a nondiagnostic entity. The diagnosis of BAPE is based on a history of asbestos exposure and exclusion of other causes, together with a period of observation to exclude malignancy [2]. Unfortunately, there is no reliable clinical, radiological, or laboratory data that can reliably differentiate early stage MPM from BAPE.
Recent advances in cancer molecular biology include the identification of microRNAs (miR) involved in regulating gene expression at the posttranscriptional level. It has been shown that approximately half of the known miR are located in cancer-associated genomic regions or in fragile sites [3]. They can act as oncogenes or tumor suppressors [4][5][6]. Studies have shown that miR profiling can differentiate tumor from normal cells, different tumor histological subtypes from each other, and may potentially predict clinical outcome [6][7][8]. miR have multitarget characteristics and can regulate groups of genes; thus, low-level modulation from individual miR could have an additive effect on multiple gene targets [9].
One can speculate that the analysis of expression signatures for miR and mRNAs in patients with MPM could differentiate benign pleural disease or metastatic cancer to the pleura from MPM, be useful for all of the histological subtypes of MPM, correlate with the extent of disease in order to monitor treatment response, and predict outcome. Only a small number of studies have been done in MPM to examine miR expression [10][11][12][13][14][15][16][17][18][19] and corresponding gene expression of these tumors [20][21][22][23][24][25].
In this study, we investigated the miR and mRNA expression levels and their integrated analysis in both BAPE and different stages and histological subtypes of MPM.

Patients and Specimens.
A total 18 MPM and 6 BAPE patients who were diagnosed and followed at the Eskisehir Osmangazi University Hospital in Turkey were enrolled in this study. None of the patients had received chemotherapy or radiotherapy prior to diagnosis. Clinical data including age, sex, asbestos exposure history, histology, stage, treatment history, and survival characteristics were collected from all mesothelioma patients, while age, sex, asbestos exposure history, and survival data were collected from the BAPE patients (Table 1). Follow-up was conducted for three years or until death. The study was approved by the Local Ethics Committee in Eskisehir Osmangazi University, and all of the patients provided written informed consent for genetic analysis on their pleural specimens.
Pleural specimens were obtained by medical thoracoscopy, CT-guided Abrams needle, or thoracotomy. Biopsy specimens were processed with a portion frozen for future analysis and another part fixed in formalin and sent to the pathology department for diagnosis. All biopsy samples underwent review in Turkey by a single pathologist and were subsequently confirmed by a lung pathologist at Mayo Clinic. Histological diagnosis of MPM was confirmed using immunohistochemistry. Diagnosis of BAPE was based on a nonmalignant pleural biopsy, a history of asbestos exposure, and the exclusion of the other causes, together with a three-year observation period to exclude malignancy. The frozen samples were transported to the Mayo Clinic for gene value from the average miR value. The standard deviation of the difference was calculated from the standard deviation of the miR and endogenous control values. The results of Δ were mean ± S.D. Low miR Δ value corresponded to high miR expression. The calculation of ΔΔ involved subtraction by the Δ calibrator value. This was subtraction of one of the Δ values as an arbitrary constant. The standard deviation of ΔΔ was the same as the standard deviation of the Δ value. The fold changes for relative miR expression were determined by 2 −Δ(Δ ) . miR which did not detect any of samples were excluded from analyses.
Student's -test was used to compare the expression of each miR between MPM and BAPE. All corrections for multiple hypothesis testing (miR and Affymetrix) were by value function in R [26]. Only upregulated miR passed criteria for multiple hypothesis testing ( -value < 0.05). miR with more than 3 mesothelioma samples with > 35 were not reported as upregulated (Table 2).
Receiver operating characteristics (ROC) curves were plotted and the area under the curve (AUC) computed to access the individual ability of each miR to differentiate BAPE from MPM and on the histology and stage of MPM.

Gene Expression Data.
The microarray signal intensity (.CEL) files were normalized and processed by the "gcrma"  [27]. SNR values greater than and less than zero potentially indicate over-and underexpression in MPM compared with BAPE, respectively. We also required that the average expression in samples overexpressing a gene had greater than 3.5 Log 2 intensities. Log 2 expression intensities for the gcrma normalized data ranged from 2 to 15. Based on our experience with quantitative RT-PCR, gene expression intensities below 3.5 were not reliable and frequently not detected. Significant figures for overexpression in MPM compared with BAPE were calculated by -test and then corrected for multiple comparison correction using the " value" package in R [26]. Downregulated probesets in MPM compared with BAPE did not pass our criteria for multiple hypothesis testing ( -value < 0.05). Upregulated probesets with best SNR and < 0.01 were reported.

miRNA and mRNA Integrated Analysis.
We used the mgsa program in R for integrated analysis to identify canonical pathways altered in MPM [28]. Upregulated probesets (2270) with SNR > 0.2 and value < 0.01 were selected. These probesets represented 1740 unique genes. Also, gene targets of 5 miR with significant downregulation in MPM ( < 0.05) were identified using the c3 data set in version 3.1 of the molecular signature database (MSDB) from the Broad Institute (http://www.broadinstitute.org/gsea/msigdb/index.jsp). These targets included 1258 unique genes. The list of upregulated genes and targets of downregulated miR were combined and used in the canonical pathway analysis using c2.cp in MSDB (version 3.1). Reported results were based on 1275 genes in the combined list that mapped to c2.cp.
Within the MPM samples, there did not appear to be any significant miR expression differences among epithelial, sarcomatous, and mixed type histology or significant differences between early stage (I-II) and late stage (III-IV) malignant disease. Microarray analyses were performed to identify differentially expressed mRNA. We identified a number of dynamically regulated mRNAs including members from the FGF (FGF9), TGFB (TGFB2), and WNT (WNT3) signaling pathways (Table 3, Figure 2).  In Study Set: number of genes in the pathway from genes selected for the analysis. In Population: number of genes in the pathway. The column "estimate" for MGSA indicates the marginal value of a term being in the "active" state; 0.4 was used as the threshold. We also identified multiple mRNA targets of miR. From integrated analysis we identified eight significant pathways, including two pathways related to NOTCH signaling (Table 4).
Interestingly, MET and CCNE2 were overexpressed in MPM and were also targets of underexpressed miR. Other than these two, there were 13 other overexpressed genes in MPM that were also targets of underexpressed miR. Importantly, MET was the most highly overexpressed gene with over 15-fold overexpression in MPM compared to BAPE (Table 5). This finding is consistent with the literature regarding the role of the c-MET oncogene in mesothelioma.

Discussion
Differentiating between MPM and BAPE on pleural biopsy can be difficult in some cases, even with the use of immunohistochemistry on biopsy samples obtained by thoracoscopy. When the histological finding of pleural tissue is called nonspecific pleuritis/fibrosis, there are no definitive clinical, radiological, or laboratory data that can help clinicians determine the next step in management. One is therefore left to consider one of two approaches: observation with prolonged follow-up or further biopsies. Both of these considerations are not ideal, especially for individuals residing in regions where  Figure 3: Selected pathways and genes by an integrated analysis using overexpressed mRNA (microarray) and targets genes of underexpressed miR. Yellow depicts targets of underexpressed miR and blue depicts overexpressed genes by microarray. Green depicts overexpressed genes that are also targets of underexpressed miR. There were 232 genes in the 8 pathways. Only 32 genes that participated in more than 1 pathway are shown. Only one of two pathways related to NOTCH signaling was included in creating this figure. MAPK1, TGFBR2, EP300, CDC42, MET, IGF1R, and SMAD2 participate in 3 or more pathways. MET and CCNE2 mRNAs are overexpressed and are targets of underexpressed miR. they are exposed to environmental asbestos and the incidence rate of mesothelioma is high. Patients with nonspecific pleuritis/fibrosis on thoracoscopic biopsy can be diagnosed as MPM within one year [29]. MPM patients who have early stage disease can be candidates for aggressive treatment and may have improved survival if they are diagnosed at the presentation and not during prolonged observation.
Guled and colleagues showed that in MPM compared with normal samples, twelve miR were highly expressed whereas nine other miR either were not expressed or had severely reduced expression levels [11]. In another study of MPM and human normal mesothelial cell cultures, 22 differentially expressed miR were identified, some of which have been linked to oncogenesis as members of the miR-17-92 cluster [15].
Benjamin and colleagues developed an assay which differentiates MPM from carcinomas using miR expression [13]. They found that the hsa-miR-200 family is strongly expressed in adenocarcinoma samples from a variety of epithelial tissues but minimally so in MPM [13]. Gee and colleagues identified a panel of miR that were specifically downregulated in MPM (irregardless of histological subtype) compared to lung adenocarcinoma and included miR-141, miR-200a * , miR-200b, miR-200c, miR-203, miR-205, and miR-429 [14]. They also evaluated the individual ability of these miR to distinguish between sample types and found that all were good discriminators of disease. The authors concluded that when the pathological tests are inconclusive, measuring a combination of miR could lead to an accurate diagnosis [14].
Several studies have evaluated miR as a new potential biomarker for the diagnosis of mesothelioma comparing the expression levels of miR in mesothelioma and asbestos exposed control groups using ROC analysis [17][18][19]. Santarelli et al. found that miR-126 could significantly differentiate high-risk individuals from mesothelioma patients with a sensitivity of 73% and specificity of 74% [17]. Another study reported that cut-off values for miR-103 could discriminate mesothelioma patients from asbestos exposed controls, revealing a sensitivity of 83% and specificity of 71% [18]. Further studies demonstrated that miR-625-3p level was significantly elevated in serum from mesothelioma patients compared with asbestosis patients with a sensitivity of 70% and a specificity of 90% [19].
Previous studies have reported different expression levels of several miR among histopathological subtypes of MPM [11,12]. Among our samples we did not find differentially expressed miR related to histopathological subtypes, likely because the majority of our samples were epithelial. In regard to the stage of MPM, results have been contradictory, possibility related to small sample sizes. For example, some authors found no correlations with miR expression and tumor stage, while other studies showed that some miR were downregulated in advanced stage MPM [12,17].
We also identified several genes with altered expression from the FGF (FGF9), TGFB (TGFB2), and WNT (WNT3) signaling pathways. Gee et al. used target prediction software to identify proteins predicted to be downregulated by miR [14]. They then queried their results with two different predictive algorithms to find targets with a higher probability of interactions. They predicted that the downregulation of unique miR would result in the loss of multiple levels of posttranscriptional gene regulation of the WNT signaling pathways [14].
Specific growth factors such as epidermal growth factor (EGF), hepatocyte growth factor (HGF), vascular endothelial growth factor (VEGF), and insulin-like growth factor (IGF) and their receptors have been shown to play a significant role in the oncogenesis, progression, and resistance to therapy in mesothelioma [30]. HGF is a multifunctional growth factor that can induce many important biological functions related to the malignant phenotype, including cell scattering, invasion, proliferation, and morphogenesis. HGF induces these biological functions through binding to its transmembrane tyrosine kinase receptor, c-mesenchymal-epithelial transition (c-MET) [31]. The majority of mesothelioma cases express c-MET [32]. We performed an integrated analysis of mRNA and miR interactions and also found overexpression of c-Met mRNA and corresponding underexpression of its targeted miR (Table 5). Our data suggest that the c-MET pathway may have promise as a therapeutic target in future clinical trials.
HGF and c-Met are highly expressed in SV40-positive mesothelioma samples [33][34][35]. It has been shown that SV40 induces HGF/Met receptor activation, telomerase activity, and Notch-1 activation in human mesothelial cells and MPM biopsy samples [33][34][35]. When SV40 infects human mesothelial cells it causes Met activation via an autocrine loop. Furthermore, SV40 replicates in human mesothelial cells and infects adjacent human mesothelial cells, inducing an HGF-dependent Met activation and cell cycle progression into S phase [33]. Notch signaling pathways were found in our study but we do not have data regarding the SV40 status in our patient population.

Conclusions
We have identified specific miR as having potential diagnostic utility in patients with MPM or BAPE. We believe that evaluation of miR would be helpful in asbestos exposed patients diagnosed with nonspecific pleuritis/fibrosis on a pleural biopsy, especially if a patient comes from a region with high mesothelioma incidence rates. Also, we have identified specific genes and signaling pathways that may have promise as a therapeutic target in patients with MPM. These results will need to be validated in a larger cohort of patients to confirm their diagnostic and therapeutic utility.