PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Int J Cancer. Author manuscript; available in PMC Dec 15, 2011.
Published in final edited form as:
PMCID: PMC2911512
NIHMSID: NIHMS189147
Downregulated MicroRNAs in the Differential Diagnosis of Malignant Pleural Mesothelioma
Gretchen V. Gee,a Devin C. Koestler,b Brock C. Christensen,ac David J. Sugarbaker,d Donatella Ugolini,e Giovanni P. Ivaldi,f Murray B. Resnick,a E. Andres Houseman,c Karl T. Kelsey,ac and Carmen J. Marsita*
a Department of Pathology and Laboratory Medicine, and Community Health Section, Brown University, Providence, RI
b Department of Biostatistics, Brown University, Providence, RI
c Center for Environmental Health and Technology, Brown University, Providence, RI
d Division of Thoracic Surgery, Brigham and Women’s Hospital, Boston, MA
e Departments of Oncology, Biology, and Genetics, University of Genoa, Italy and Unit of Epidemiology and Biostatistics, National Cancer Research Institute, Genoa, Italy
f Department of Pneumology, A.O. Scassi, Genoa, Italy
* Corresponding author: Carmen J. Marsit, Department of Pathology and Laboratory Medicine, Brown University, Box G-E537, Providence, RI 02912, Carmen_Marsit/at/brown.edu, Phone: (401) 863-6508, Fax: (401) 863-9008
Malignant pleural mesothelioma is a rapidly fatal disease whose diagnosis, particularly through less invasive techniques such as analysis of pleural effusion, can be challenging. Currently, a commercially available diagnostic test based on microRNA expression patterns is purported to distinguish between mesothelioma and lung adenocarcinoma. Yet, the biological basis of this technology has not been reported in the literature, and little research has been aimed at determining how differential miRNA expression contributes to the differences in pathogenesis between these diseases, both of which can be caused by asbestos exposure. We sought to illuminate the molecular differences between mesothelioma and lung adenocarcinoma by using microRNA microarrays to identify patterns in the most differentially expressed microRNAs. From this, we identified a panel of microRNAs, including members of the miR-200 gene family, that were all downregulated in malignant pleural mesothelioma compared to lung adenocarcinoma. Using the more sensitive detection method of quantitative RT-PCR on an independent series of tumors, we validated the specificity of these alterations in 100 malignant pleural mesotheliomas and 32 lung adenocarcinomas. Statistical analysis reveals that these microRNAs exceed the current recommendations for biomarkers and could greatly aid in the differential diagnosis. Further examination led us to predict that they act as redundant regulators of wnt signaling, and suggests a role for this pathway in tumor progression. This research points to novel approaches utilizing microRNAs whose decreased expression is unique to mesothelioma as potentially suitable for rapid diagnosis, and reveals prospective new targets for the treatment of this deadly disease.
Keywords: Mesothelioma, microRNA, wnt-signaling, Lung Adenocarcinoma
Malignant pleural mesothelioma (MPM) is an aggressive tumor that develops in the parietal pleura and is generally attributable to asbestos exposure. Although there is a long latency period between exposure to the carcinogen and disease, once detected it is rapidly fatal, with the median survival time being less than a year after diagnosis 1. MPM can be difficult to distinguish from adenocarcinoma arising in the periphery of the lung, and epidemiological evidence suggests asbestos and smoking are shared risk factors for these diseases 24. Accurate diagnosis can require multiple, time-consuming tests and, the “golden standard” in identification of MPM relies on light microscopy, in conjunction with a panel of immunohistochemical stains 1, 57. Investigation into the molecular differences between these two diseases should lead to a better understanding of the mechanistic differences, as well as provide biomarkers useful in diagnosis.
Increasingly, microRNAs (miRNAs) are being recognized as key players in cancer incidence and progression, and a number of reports have suggested that they may hold significant clinical utility in cancer diagnosis and prognosis 8, 9. MiRNAs are small non-coding regulatory RNA elements that can bind to messenger RNA (mRNA) transcripts and suppress protein expression. They are formed in the nucleus from a stem-loop pre-miRNA precursor, which is then transported to the cytoplasm and processed into the 19–22 nucleotide double-stranded mature miRNA. Generally, the functional mature miRNA is loaded into the RNA-induced silencing complex while the minor, complementary strand, referred to as the miRNA* sequence, is often degraded 10. The miRNA can then pair with the 3′UTR of a target gene; leading to either degradation of the mature mRNA if there is complete complementarity, or with incomplete complementarity, reversible sequestering of the transcript and inhibition of protein translation 11, 12. It is clear that miRNAs play an essential regulatory role in healthy cells and miRNA gene clustering and families ensure functional redundancy and safeguards against single mutations resulting in a dysregulation of key pathways 1315. By examining the relationship between miRNA expression and their targets, we should be able to predict the outcome of loss or gain of expression and be able to design novel treatments accordingly.
Currently, there is a commercially available miRNA-based diagnostic test, marketed as ProOnc MesotheliomaDx by Rosetta Genomics and distributed by Prometheus Laboratories, Inc., that claims to be a highly accurate test 16. Yet, little work has described the basis of this test, nor have these miRNA differences been examined in regards to disease pathogenesis. While miRNA profiles of lung adenocarcinoma have been investigated, there has not been a comprehensive analysis of how these changes distinguish adenocarcinoma of the lung from mesothelioma, and it is not clear how these miRNAs influence the progression of disease 17.
We used a relatively large number of histologically confirmed, primary human clinical tumor samples to gain preliminary insight into the key differences in the pattern of miRNA expression between MPM and lung adenocarcinoma. We found these miRNAs can be used as highly sensitive and specific biomarkers of MPM which suggested that their loss may act in concert to lead to disease. Using in silico approaches, we sought to better understand how the biology of these diseases may be differentially impacted by the expression patterns of key miRNA.
Tissue ascertainment
For preliminary miRNA analysis by microarray, tissues were obtained from the CREST Biorepository of the National Cancer Research Institute, Genova Italy. The samples consisted of pleural biopsies from lung adenocarcinoma (n = 10) patients and non-histologically defined mesothelioma patients (n = 15).
The validation series consisted of a total of 100 mesothelioma tumor samples obtained from two sources: mesothelioma tumors obtained from surgicalresection at Brigham and Women’s Hospital from an incident caseseries beginning in 2005 (n = 23) and tumor samples obtained from the National Mesothelioma Virtual Bank (n = 77). These samples consisted of 32 uncharacterized mesotheliomas, 39 epithelioid mesotheliomas, 19 biphasic mesotheliomas and 10 sarcomatoid mesotheliomas. For comparison, we studied 32 lung adenocarcinoma tumor samples from the tumor bank at Rhode Island Hospital (n = 32). Non-diseased lung tissue samples were obtained from the National Disease Research Interchange (n=4). Small RNA from non-diseased tissue was extracted, pooled and used as a relative baseline for comparison in qRT-PCR validation experiments.
RNA Isolation
RNA extractions were performed using the Ambion mirVana microRNA Isolation Kit (cat # AM1561) according to the manufacturer’s instructions for total RNA and small RNA enriched fractions. RNA quality was assessed using an Agilent 2100 bioanalyzer. RNA was used immediately or aliquoted and stored at −80 °C until use.
Microarray analysis
The enriched small RNA fraction obtained from the isolation was labeled using the mirVana miRNA Labeling kit (Ambion cat # AM1562) and biotin-labeled UTP (Ambion cat # AM8452). Labeling controls were RNA oligonucleotides homologous to the C. elegans miRNAs, cel-mir-2 and cel-lin-4. Labeled RNA was immediately processed for microarray work. Experiments were performed using custom Affymetrix gene chip microRNA microarrays (TGmirV1b520432f) provided by Tony Godfrey (University of Rochester Medical School). These arrays contain 3231 individual sequences made up of 2564 human, 357 mouse and 238 rat and 72 non-endogenous control miRNAs. The microarray work was carried out in the Core Facility of the Center for Genomics and Proteomics at Brown University following standard procedures used in the core for the Affymetrix platform. The resulting array data was assessed for quality using Expression Console and processed using Partek Genomics Suite software with normalization performed using the RMA procedure. Unsupervised hierarchical clustering analysis using the Euclidean distance metric and Ward linkage as parameters was carried out on unadjusted data using the statistical software program “R” (http://www.r-project.org/).
Quantitative RT-PCR
TaqMan miRNA Assays (Applied Biosystems) were used to quantify mature miRNA expression for the chosen candidate miRNAs: hsa-miR-200a* (part # 4373273) hsa-miR-200c (part # 4395411), hsa-miR-200b (part # 4395362), hsa-miR-203 (part # 4373095), hsa-miR-141 (part # 4373137), hsa-miR-429-(part # 4373203) and hsa-miR-205 (part # 4373093) and the reference non-coding RNAs RNU44 (part # 4373384), RNU48 (part # 4373383). 5 ng of RNA enriched for small RNAs was used as the template. All cDNA synthesis and qRT-PCR were performed according the manufacturer’s recommended conditions using the TaqMan® MicroRNA Reverse Transcription Kit, and TaqMan® Universal PCR Master Mix, No AmpErase® UNG (Applied Biosystems). QRT-PCR reactions were run on an Applied Biosystems 7900 Real Time PCR System using the 384-well plate configuration. We used a non-parametric Wilcoxon signed rank test to test for differential expression between the MPM and lung adenocarcinoma samples among the seven biologically relevant miRNAs in both qRT-PCR data sets 18. Furthermore, we used the Benjamini and Hochberg False Discovery Rate to correct for multiple testing 19.
ROC/AUC Analysis
To access the individual ability of each miRNA to differentiate between tumor types, we used Receiver Operating Characteristic (ROC) curves and computed the Area Under the Curve (AUC) 20. The ROC curve is a graphical plot of the sensitivity by 1-specificty as the discrimination threshold is varied whereas AUC is a measure of the similarity of the two classes, where a class in the context of our problem represents the tumor type. The closer the AUC is to 1 for a particular miRNA, the better the individual ability that miRNA has for differentiating between tumor types. Asymptotic normal approximation was used to compute 95% confidence intervals for each of the AUC’s to determine the range of plausible AUC’s for each miRNA. Reported data is based on optimal cutoff points.
Random Forest Procedure
To determine how well, collectively, the seven candidate differentially expressed miRNAs classify tumor type, we used the Random Forest (RF) 21, a supervised method of classification, which we implemented using the RandomForest package in R, version 4.5–25 by Liaw and Wiener. The RF classifier is essentially a collection of classification trees, where each classification tree is developed using a bootstrap sample (sampling with replacement) of the data and on average, about one third of the data is not sampled [out-of-bag (OOB)]. The observations which were sampled are used as training data to develop the classification trees, where at each node of the tree, m out of the total M predictors are chosen and the best split among these m predictors is found. The default value of m in the RandomForest R-package is equation M1, which we have found, in many cases gives the lowest prediction error. The OOB data for each tree are used as test data to obtain an estimate of the OOB error rate, the percentage of times the RF prediction is incorrect.
Here, we used the independent qRT-PCR data as training data to develop the RF classifier and use the qRT-PCR data from the original samples as test data. Since both data sets had some missing values, we used a Random Forest method of imputing the missing data 21. This method imputes missing data based on the proximity matrix from the Random Forest.
Target Prediction
TargetScan Human 5.1 (http://www.targetscan.org/) and TargetScan Custom 4.2 (http://www.targetscan.org/vert_42/seedmatch.html) were used to predict miRNA targets based on a 6 or 7 nucleotide sequence, and the presence of an “anchoring” adenosine, in the 3′ UTRs of genes that bind to the seed sequence (nt 2–7) of the miRNAs 22. Targets are analyzed for evolutionary conservation across vertebrates (broadly conserved), across placental mammals (conserved) or without regard to conservation (not conserved). TargetScan Human 5.1 was used to search for miRNA families that targeted known genes involved in wnt signaling and TargetScan Custom 4.2 was used to search the seed sequence of miR-200a* for predicted human gene targets. MiRNA-target interactions were then queried using the PicTar algorithm (http://pictar.mdc-berlin.de/) and MicroCosm Targets version 5 (http://www.ebi.ac.uk/enright-srv/microcosm/htdocs/targets/v5/). PicTar uses its own algorithm 23 and MicroCosm uses the miRanda algorithm to identify potential miRNA binding sites in genomic sequences 24.
Unsupervised hierarchical clustering using Euclidean distance and Ward linkage among the top 250 most variable miRNAs demonstrated imperfect clustering of mesothelioma from lung adenocarcinoma, as evidenced by the heatmap (Figure 1A). This is reflective of the fact that these two diseases have similar etiologies and presentations, and suggests a much more subtle molecular distinction. Careful analysis of the microarray results revealed the mesothelial-specific downregulation of four out of five members of the miR-200 miRNA family (miR-200c, miR-200b, miR-141 and miR-429), and the downregulation of miR-200a*, the minor miRNA sequence of another miR-200 family member. We also noted reduced expression of two other cancer-related miRNAs, miR-203 and miR-205, in MPM compared to lung adenocarcinoma (Figure 1B).
Figure 1
Figure 1
A. Heatmap of the top 250 most variable microRNAs between 15 Mesothelioma (green) and 10 Lung Adenocarcinoma tumors (red). Unsupervised hierarchical clustering using Euclidean distances and Ward linkage did not distinguish between the two tumor types (more ...)
QRT-PCR measurement of miRNA isolated from the original tumor biopsies confirmed array-based measured downregulation of each of the seven miRNAs in MPM compared to lung adenocarcinoma. We calculated average relative expression, plus or minus the standard error and found the reduced expression to be statistically significant (p ≤ .05) in four of the miR-200 family members (miR-200c, miR-200b, miR-141 and miR-429), as well as in miR-203 and miR-205 (Table 1).
Table 1
Table 1
Average relative expression levels of original tumors measured by qRT-PCR
In order to validate our findings and increase our statistical power, we investigated the expression of these seven miRNAs in a much larger, independent data set using quantitative PCR. Consistent with array results, all tested miRNAs had significantly lower expression in MPM relative to lung adenocarcinoma (p ≤ .004, Figure 2). The average expression of miRNAs in MPM compared to lung adenocarcinoma ranged from 6-fold reduction (miR-203) to 42-fold reduction (miR-205). MiR-141 is reduced 11-fold and miR-200a* is down 9-fold. MiR-200c/200b and 429, which, based on seed sequence are all predicted to target the same genes, were reduced 30-fold, 8-fold and 7-fold respectively in MPM compared to lung adenocarcinoma. The levels of each of these miRNAs were compared among the three histological types of mesothelioma, epitheliod, sarcomatoid and biphasic, and found not to be statistically significantly different.
Figure 2
Figure 2
Quantitative PCR was used to measure the expression of miR-200c, miR-141, miR-200b, miR-200a*, miR-429, miR-203 and miR-205 in an independent sample set. We obtained and isolated miRNA from an independent sample set which consisted of 100 mesothelioma (more ...)
Next, we wanted to evaluate the individual ability of these miRNAs to distinguish between sample types within these datasets. To achieve this, we generated ROC curves using the qRT-PCR data from the independent data set, and computed the AUC with asymptotic normal 95% confidence intervals, for each of the seven miRNAs (Figure 3). All seven miRNAs prove to be good discriminators of disease, but four of the miR-200 family member miRNAs (miR-200c, miR-141, miR-200b, and miR-429), are excellent at discrimination, with AUC values greater than 0.9 (Table 2). In fact, all seven of these miRNAs exceed the recommendation of The International Mesothelioma Interest Group for identifying markers, that either the sensitivity or specificity be greater than 80% 5.
Figure 3
Figure 3
ROC curves were created using the quantitative PCR data from the independent data set. Asymptotic normal approximation was used to compute 95% confidence intervals. A; miR-200c, B: miR-141, C: miR-200b, D: miR-200a*, E: miR-429, F: miR-203, G: miR-205. (more ...)
Table 2
Table 2
Characteristics of ROC Curves: AUC (95% CI)
To examine how these miRNA may interact to distinguish between MPM and lung adenocarcinoma, we developed a Random Forest classifier using the qRT-PCR data on the independent tumor samples with all seven miRNAs (miR-200c, miR-141, miR-200b, miR-200a* miR-429, miR-203 and miR-205). This worked extremely well, resulting in a 10% misclassification error when used to classify the qRT-PCR data from the original samples with one mesothelioma misclassifying as lung adenocarcinoma and one lung adenocarcinoma misclassifying as mesothelioma. The null value represents the expected misclassification error by randomly assigning a subject to one of the two classes (Table 3).
Table 3
Table 3
Random Forest Classifier
Random Forest classification demonstrated that the combination of miRNAs we investigated are predictive of tumor type. This reflects a convergent biological quality of this group of miRNAs, and suggests that they may target a common pathway. We hypothesized that their phenotypic action would delineate critical pathways in mesothelioma tumorigenesis. To investigate this, we used target prediction software to identify proteins predicted to be downregulated by these miRNAs. We then queried our results with two different predictive algorithms in order to find targets with higher probability of interactions (Table 4). As the miR-200 miRNA family has been shown to be negative regulators of wnt signaling in invertebrates 25, we searched the 3′ UTRs of multiple proteins involved in a this pathway, for predicted miRNA targets. This approach, summarized in Table 4, revealed that the downregulated miRNAs are predicted to redundantly target multiple downstream modulators of the canonical wnt signaling pathway, such as MYC and JUN, as well as effectors of the non-canonical wnt pathways: the planar cell polarity pathway (VANGL1, RHOA, ROCK2) and the wnt/Ca2+ pathway: (WNT5A and PLCB1) 26.
Table 4
Table 4
Predicted miRNA target seed sequences in the 3′ UTRs of genes involved in wnt signaling
We investigated the nature of the potential miRNA-target duplexes and found the strengths of these interactions occur in varying degrees. Stronger pairing, indicated by either an additional match or an adenosine after the seed sequence at nucleotide 8, has been shown to increase the likelihood of an interaction. Notably, increasing degrees of complementarity in these duplexes are linked to stronger degrees of protein repression 12, 27. We also found that a number of the 3′ UTR target recognition sites are conserved in vertebrates, suggesting that they are functionally active. Most strikingly, in silico approaches predict that a number of wnt signaling proteins could be repressed by more than one of our miRNAs. From the predictions of three separate algorithms, we can reason that the loss of multiple, redundant miRNA regulators would result in aberrant wnt signaling at the level of post-transcriptional gene silencing.
Malignant Pleural Mesothelioma is a rapidly fatal disease that is often difficult to diagnose pathologically and can be confused with other cancers, often, lung adenocarcinoma. Treatment of MPM relies on a rapid diagnosis and it would be advantageous to have specific biomarkers to differentiate it from other tumors 6. MiRNA are being used clinically to aid in the diagnosis of this disease, although there is little experimental literature detailing the identification of differentially altered miRNA, or the biological pathways predicted to be affected by miRNA expression alterations. We identified a panel of miRNA that are specifically downregulated in MPM compared to lung adenocarcinoma. Expression of these miRNAs was not significantly different in the histological types of mesothelioma, suggesting that their downregulation is in fact, a general characteristic of mesothelioma, and distinguishes this disease from its main differential diagnosis, adenocarcinoma. Analysis of these miRNAs reveals that they can act as biomarkers which exceed the current guidelines for the diagnosis of mesothelioma (Table 2). In fact, miR-200c has a 100% Negative Predictive Value, indicating that detection of this miRNA is enough to rule out a diagnosis of mesothelioma. When pathological tests are inconclusive, measuring a combination of these miRNAs could lead to an accurate diagnosis which can greatly influence the treatment and outcome of the patient.
Understanding the normal, collective role of these miRNA can inform on the consequences of their loss in mesothelioma, and suggest novel targets for therapeutic intervention and disease screening, both of which are crucial to patient survival. Previous studies relied on cultured tumor cell lines to identify a panel of miRNAs which may be associated with disease 28. Although most research on miRNAs done in cell lines can provide clues to human cancers, due to the inherent artificiality of in vitro systems, conclusions drawn from experimental systems alone must be corroborated with data from human tumors. Experiments in our lab, as well as published reports have shown that cultured cell lines are inappropriate for miRNA biomarker identification andoften have no relation to the expression patterns in primary tumors 29. Our approach using human tumors allowed us to identify a group of miRNAs whose downregulation in MPM are predictive of disease. Analysis of the overlapping targets of these miRNAs reveals the biological mechanisms through which they lead to mesothelial carcinogenesis. The gene clustering and familial pattern of these miRNAs indicated that they affect common pathways 13 and led us to predict aberrant wnt signaling as a mechanism for disease, providing potential new targets for treatment.
Experiments in simple organisms that show most miRNAs are not essential, and individually often have subtle effects on their downstream targets, provide evidence for the theory of miRNA redundancy 14, 15. Five out of seven of our miRNAs are members of the same gene family (miR-200). This family is encoded in two clusters on chromosomes 1 (miR-200a/b/429) and 12 (miR-200c/141), and is evolutionarily conserved. Forty-two percent of human miRNAs are encoded in clusters less than 3000 nucleotides apart, suggesting that they are co-regulated and that their expression levels can be indicative of certain diseases and syndromes 13. These miRNAs are grouped into a family due to homology in their seed sequences, and which messenger RNAs they will target. In fact, miR-200b/c/429 have identical seed sequences, indicating that they are redundant regulators of the same genes. The loss of expression of these closely related miRNAs by at least two separate events suggests they have redundant functions that are absent in the diseased state. By comparing overlapping targets of all of our classifying miRNAs, we were predict that absence of these miRNAs would result in the loss of multiple levels of post-transcriptional gene regulation of the wnt signaling pathways.
Our observation of the downregulation of seven potential miRNA wnt antagonists in mesothelioma suggests that dysregulation of this pathway, through the loss of miRNA expression, is an essential process in disease progression. The mechanism for this is not known but may occur in various ways. Two of the miRNAs are encoded individually; miR-203 on chromosome 14 and miR-205 on chromosome 1, and loss of their expression seems to be unrelated. Downregulation of the miR-200 gene family may be coordinate, however, as the genes for the miRNAs miR-200c/141 are clustered on chromosome 12, while the genes for miR-200a/b/429 are clustered on chromosome 1. This clustering, combined with evidence that distinct methylation alterations predict MPM, indicates epigenetic silencing as a possible mechanism of repression 30. Indeed, the expression of miR-203 from a fragile region of chromosome 14 has been shown to be modulated by loss of heterozygocity and through hypermethylation in some acute lymphoblastic leukemias 31. It is of interest to investigate the mechanisms for downregulation of these specific miRNAs in MPM in order to gain further insight into the genesis of this disease, as well as to develop novel modes of treatment.
Out of the seven miRNAs that we describe as under-expressed in MPM compared to lung adenocarcinoma, two (miR-203 and miR-205), have previously been shown to be upregulated in lung adenocarcinoma compared to matched, noncancerous tissue 17. Even higher expression of miR-205 has been suggested as a biomarker for squamous cell carcinoma, as its upregulation can distinguish this tumor from adenocarcinoma 32. The miRNAs miR-203 and miR-429 have recently been shown to be downregulated in MPM 33. We found a panel of miRNAs (miR-200c/141/200b/200a*/429/203/205) downregulated in MPM that we propose as biomarkers for disease and predict that these miRNAs play overlapping redundant roles in modulation of proteins associated with wnt signaling. This is significant, as it has been shown that activation of the wnt pathway, specifically the nuclear translocation of β-catenin, occurs in mesothelioma cell lines 34. In fact, we identified multiple wnt-pathway gene products that have been shown to be upregulated in MPM, including Jun, Myc, EGR1 and Wnt5B, which are predicted to modulated but our panel of microRNAs 35. (Table 4).
The importance of wnt signaling has long been studied in development. In the canonical wnt pathway, wnt binding to its receptor leads to the stabilization of β-catenin. β-catenin can then translocate to the nucleus and interact with the transcription factors LEF/TCF (lymphoid enhancer-binding factor/T cell factor) to activate target genes 36. In Drosophila, the members of the miR-8 gene family have been shown to antagonize this signaling pathway in multiple ways, including through the inhibition of TCF translation, a positive regulator of wnt signaling, as well as by targeting effectors of the pathway 25. In human cancers, reduction in expression of the orthologous miR-200 gene family and its target E-cadherin, as well as lower miR-205 expression, is associated with metastases 37, 38. Specifically, it has been demonstrated that miR-200c can block the translation of the zinc finger transcription factor ZEB1 (TCF8/dEF1) and result in an increase in E-cadherin 39. Other evidence indicates that decreased expression of the miR-200 gene family results in the up-regulation of the E-cadherin repressors ZEB1 and ZEB2, promoting the epithelial-to-mesenchymal transition (EMT) that is indicative of malignant tumor progression 40. This pathway was revealed to be part of a regulatory loop when it was shown that ZEB1 can directly repress miR-200c and miR-141 41. The effect of the loss of cadherins on the wnt signaling pathway is still being investigated but it is known that cadherin can tether β-catenin to the cellular membrane. In conjunction with a disruption in the β-catenin degradation pathway, the loss of this adhesion molecule may elevate wnt signaling by allowing translocation of β-catenin to the nucleus 42.
Interestingly, dysregulation of non-miRNA regulators of the wnt signaling pathway has been reported in the metastases lung adenocarcinoma. In this case, disruption of the WNT/TCF signaling occurs through the transcription factors LEF1 and HOXB9 and promotes invasion and colonization of multiple organ types, and is not associated with primary lung adenocarcinoma 43. Target prediction software indicates that the miRNAs downregulated in MPM have conserved and redundant roles in the regulation of the wnt signaling pathway. Our data, combined with previously published observations provide evidence that wnt signaling dysregulation in MPM occurs through the miRNAs, but aberrant wnt signaling in lung adenocarcinoma occurs through other mechanisms.
In summary, our data shown that MPM can be distinguished from lung adenocarcinoma through the downregulation of miRNAs: miR-141, miR-200a*, miR-200b, miR-200c, miR-203, miR-205 and miR-429. The downregulation of these miRNAs are characteristic of mesothelioma regardless of histological subtype. We propose that these miRNAs can be used as biomarkers to develop an assay for detection of MPM and aid in distinguishing it from lung adenocarcinoma. In addition, analysis of the cumulative loss of these miRNAs identified aberrant regulation of the wnt signaling pathway, effected through the loss of its miRNA antagonists, as a key component of MPM. Further study of this phenomenon should lead to a better understanding of the mechanism of MPM, as well as provide new strategies for the treatment of this disease.
Acknowledgments
Mesothelioma Applied Research Foundation Grant, and NIH grants from the NCI (R01CA126939) and NIEHS (T32ES007272, G.V.G). We thank C. Schorl of the Brown University Core Facility of the Center for Genomics and Proteomics, supported by the COBRE Center for Cancer Signaling Networks (P50RR015578) for assistance in microarray analysis. We also thank the Mesothelioma Research Bank [CDC NIOSH 1-U19-OH009077-01 Mesothelioma Virtual Bank for Translational Research] for providing samples used in this study.
Abbreviations
AUCArea Under the Curve
FDRFalse Discovery Rate
MPMMalignant pleural mesothelioma
mRNAmessenger RNA
miRNA/miRmicroRNA
OOBout-of-bag
RFRandom Forest
ROCReceiver Operating Characteristic
QRT-PCRquantitative RT-PCR
UTRuntranslated region

Footnotes
Novelty & Impact: We identified a panel of downregulated miRNAs which distinguish mesothelioma from lung adenocarcinoma and analysis of the overlapping targets of these miRNAs indicates that their cumulative loss of expression is predicted to induce aberrant wnt signaling. This work should lead to the development of a clinically applicable test to identify mesothelioma, as well as provide key insight into alternative approaches for treating newly diagnosed patients.
1. Robinson BW, Lake RA. Advances in malignant mesothelioma. N Engl J Med. 2005;353:1591–603. [PubMed]
2. Blot W, JFF Cancer Epidemiology and Prevention. New York: Oxford University Press; 1996.
3. Selikoff IJ, Hammond EC, Churg J. Asbestos exposure, smoking, and neoplasia. JAMA. 1968;204:106–12. [PubMed]
4. Nelson HH, Kelsey KT. The molecular epidemiology of asbestos and tobacco in lung cancer. Oncogene. 2002;21:7284–8. [PubMed]
5. Husain AN, Colby TV, Ordonez NG, Krausz T, Borczuk A, Cagle PT, Chirieac LR, Churg A, Galateau-Salle F, Gibbs AR, Gown AM, Hammar SP, et al. Guidelines for pathologic diagnosis of malignant mesothelioma: a consensus statement from the International Mesothelioma Interest Group. Arch Pathol Lab Med. 2009;133:1317–31. [PubMed]
6. Addis B, Roche H. Problems in mesothelioma diagnosis. Histopathology. 2009;54:55–68. [PubMed]
7. Gordon GJ, Jensen RV, Hsiao LL, Gullans SR, Blumenstock JE, Ramaswamy S, Richards WG, Sugarbaker DJ, Bueno R. Translation of microarray data into clinically relevant cancer diagnostic tests using gene expression ratios in lung cancer and mesothelioma. Cancer Res. 2002;62:4963–7. [PubMed]
8. Sotiropoulou G, Pampalakis G, Lianidou E, Mourelatos Z. Emerging roles of microRNAs as molecular switches in the integrated circuit of the cancer cell. RNA. 2009;15:1443–61. [PubMed]
9. Garzon R, Calin GA, Croce CM. MicroRNAs in Cancer. Annu Rev Med. 2009;60:167–79. [PubMed]
10. Okamura K, Phillips MD, Tyler DM, Duan H, Chou YT, Lai EC. The regulatory activity of microRNA* species has substantial influence on microRNA and 3′ UTR evolution. Nat Struct Mol Biol. 2008;15:354–63. [PMC free article] [PubMed]
11. Fabbri M, Croce CM, Calin GA. MicroRNAs. Cancer J. 2008;14:1–6. [PubMed]
12. Grimson A, Farh KK, Johnston WK, Garrett-Engele P, Lim LP, Bartel DP. MicroRNA targeting specificity in mammals: determinants beyond seed pairing. Mol Cell. 2007;27:91–105. [PMC free article] [PubMed]
13. Altuvia Y, Landgraf P, Lithwick G, Elefant N, Pfeffer S, Aravin A, Brownstein MJ, Tuschl T, Margalit H. Clustering and conservation patterns of human microRNAs. Nucleic Acids Res. 2005;33:2697–706. [PMC free article] [PubMed]
14. Allen RS, Li J, Stahle MI, Dubroue A, Gubler F, Millar AA. Genetic analysis reveals functional redundancy and the major target genes of the Arabidopsis miR159 family. Proc Natl Acad Sci U S A. 2007;104:16371–6. [PubMed]
15. Miska EA, Alvarez-Saavedra E, Abbott AL, Lau NC, Hellman AB, McGonagle SM, Bartel DP, Ambros VR, Horvitz HR. Most Caenorhabditis elegans microRNAs are individually not essential for development or viability. PLoS Genet. 2007;3:e215. [PubMed]
16. ProOnc MesotheliomaDX. Prometheus Therapeutics & Diagnostics. 2009.
17. Yanaihara N, Caplen N, Bowman E, Seike M, Kumamoto K, Yi M, Stephens RM, Okamoto A, Yokota J, Tanaka T, Calin GA, Liu CG, et al. Unique microRNA molecular profiles in lung cancer diagnosis and prognosis. Cancer Cell. 2006;9:189–98. [PubMed]
18. Wilcoxon F. Individual comparisons by ranking methods. Biometrics. 1945;1:80–3.
19. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, Series B (Methodological) 1995;57:289–300.
20. Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem. 1993;39:561–77. [PubMed]
21. Breiman L. Random Forests. Machine Learning. 2001;45:5–32.
22. Lewis BP, Burge CB, Bartel DP. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell. 2005;120:15–20. [PubMed]
23. Krek A, Grun D, Poy MN, Wolf R, Rosenberg L, Epstein EJ, MacMenamin P, da Piedade I, Gunsalus KC, Stoffel M, Rajewsky N. Combinatorial microRNA target predictions. Nat Genet. 2005;37:495–500. [PubMed]
24. Griffiths-Jones S, Saini HK, van Dongen S, Enright AJ. miRBase: tools for microRNA genomics. Nucleic Acids Res. 2008;36:D154–8. [PMC free article] [PubMed]
25. Kennell JA, Gerin I, MacDougald OA, Cadigan KM. The microRNA miR-8 is a conserved negative regulator of Wnt signaling. Proc Natl Acad Sci U S A. 2008;105:15417–22. [PubMed]
26. Chien AJ, Conrad WH, Moon RT. A Wnt survival guide: from flies to human disease. J Invest Dermatol. 2009;129:1614–27. [PMC free article] [PubMed]
27. Brennecke J, Stark A, Russell RB, Cohen SM. Principles of microRNA-target recognition. PLoS Biol. 2005;3:e85. [PMC free article] [PubMed]
28. Busacca S, Germano S, De Cecco L, Rinaldi M, Comoglio F, Favero F, Murer B, Mutti L, Pierotti M, Gaudino G. MicroRNA Signature of Malignant Mesothelioma with Potential Diagnostic and Prognostic Implications. Am J Respir Cell Mol Biol. 2009 [PubMed]
29. Avissar M, Christensen BC, Kelsey KT, Marsit CJ. MicroRNA expression ratio is predictive of head and neck squamous cell carcinoma. Clin Cancer Res. 2009;15:2850–5. [PMC free article] [PubMed]
30. Christensen BC, Houseman EA, Godleski JJ, Marsit CJ, Longacker JL, Roelofs CR, Karagas MR, Wrensch MR, Yeh RF, Nelson HH, Wiemels JL, Zheng S, et al. Epigenetic profiles distinguish pleural mesothelioma from normal pleura and predict lung asbestos burden and clinical outcome. Cancer Res. 2009;69:227–34. [PMC free article] [PubMed]
31. Bueno MJ, Perez de Castro I, Gomez de Cedron M, Santos J, Calin GA, Cigudosa JC, Croce CM, Fernandez-Piqueras J, Malumbres M. Genetic and epigenetic silencing of microRNA-203 enhances ABL1 and BCR-ABL1 oncogene expression. Cancer Cell. 2008;13:496–506. [PubMed]
32. Lebanony D, Benjamin H, Gilad S, Ezagouri M, Dov A, Ashkenazi K, Gefen N, Izraeli S, Rechavi G, Pass H, Nonaka D, Li J, et al. Diagnostic assay based on hsa-miR-205 expression distinguishes squamous from nonsquamous non-small-cell lung carcinoma. J Clin Oncol. 2009;27:2030–7. [PubMed]
33. Guled M, Lahti L, Lindholm PM, Salmenkivi K, Bagwan I, Nicholson AG, Knuutila S. CDKN2A, NF2, and JUN are dysregulated among other genes by miRNAs in malignant mesothelioma -A miRNA microarray analysis. Genes Chromosomes Cancer. 2009;48:615–23. [PubMed]
34. Uematsu K, Kanazawa S, You L, He B, Xu Z, Li K, Peterlin BM, McCormick F, Jablons DM. Wnt pathway activation in mesothelioma: evidence of Dishevelled overexpression and transcriptional activity of beta-catenin. Cancer Res. 2003;63:4547–51. [PubMed]
35. Mazieres J, You L, He B, Xu Z, Twogood S, Lee AY, Reguart N, Batra S, Mikami I, Jablons DM. Wnt2 as a new therapeutic target in malignant pleural mesothelioma. Int J Cancer. 2005;117:326–32. [PubMed]
36. Lai SL, Chien AJ, Moon RT. Wnt/Fz signaling and the cytoskeleton: potential roles in tumorigenesis. Cell Res. 2009;19:532–45. [PubMed]
37. Baffa R, Fassan M, Volinia S, O’Hara B, Liu CG, Palazzo JP, Gardiman M, Rugge M, Gomella LG, Croce CM, Rosenberg A. MicroRNA expression profiling of human metastatic cancers identifies cancer gene targets. J Pathol. 2009 [PubMed]
38. Gregory PA, Bert AG, Paterson EL, Barry SC, Tsykin A, Farshid G, Vadas MA, Khew-Goodall Y, Goodall GJ. The miR-200 family and miR-205 regulate epithelial to mesenchymal transition by targeting ZEB1 and SIP1. Nat Cell Biol. 2008;10:593–601. [PubMed]
39. Hurteau GJ, Carlson JA, Spivack SD, Brock GJ. Overexpression of the microRNA hsa-miR-200c leads to reduced expression of transcription factor 8 and increased expression of E-cadherin. Cancer Res. 2007;67:7972–6. [PubMed]
40. Park SM, Gaur AB, Lengyel E, Peter ME. The miR-200 family determines the epithelial phenotype of cancer cells by targeting the E-cadherin repressors ZEB1 and ZEB2. Genes Dev. 2008;22:894–907. [PubMed]
41. Burk U, Schubert J, Wellner U, Schmalhofer O, Vincan E, Spaderna S, Brabletz T. A reciprocal repression between ZEB1 and members of the miR-200 family promotes EMT and invasion in cancer cells. EMBO Rep. 2008;9:582–9. [PMC free article] [PubMed]
42. Jeanes A, Gottardi CJ, Yap AS. Cadherins and cancer: how does cadherin dysfunction promote tumor progression? Oncogene. 2008;27:6920–9. [PMC free article] [PubMed]
43. Nguyen DX, Chiang AC, Zhang XH, Kim JY, Kris MG, Ladanyi M, Gerald WL, Massague J. WNT/TCF signaling through LEF1 and HOXB9 mediates lung adenocarcinoma metastasis. Cell. 2009;138:51–62. [PMC free article] [PubMed]