Specific peripheral miRNA profiles for distinguishing lung cancer from COPD

Lung Cancer 74 (2011) 41–47

Contents lists available at ScienceDirect

Lung Cancer journal homepage: www.elsevier.com/locate/lungcan

Specific peripheral miRNA profiles for distinguishing lung cancer from COPD Petra Leidinger a,1 , Andreas Keller b,c,d,1 , Anne Borries c , Hanno Huwer e , Mareike Rohling f , Junko Huebers f , Hans-Peter Lenhof d , Eckart Meese a,∗ a

Institute of Human Genetics, Saarland University, Medical School, Homburg, Germany Biomarker Discovery Center, Heidelberg, Germany Febit Biomed GmbH, Heidelberg, Germany d Center for Bioinformatics, Saarland University, Saarbruecken, Germany e Department of Cardiothoracic Surgery, Voelklingen Heart Center, Germany f Department of Pneumology, Voelklingen Lung Center, Germany b c

a r t i c l e

i n f o

Article history: Received 25 October 2010 Received in revised form 15 December 2010 Accepted 7 February 2011 Keywords: Blood Chronic obstructive pulmonary disease Expression profile Lung cancer Microarray MicroRNA

a b s t r a c t Recently we reported differential miRNA signatures in blood cells of lung cancer patients and healthy controls. With the present study we wanted to investigate if miRNA blood signatures are also suited to differentiate lung cancer patients from COPD patients. We compared the expression of 863 human miRNAs in blood cells of lung cancer patients, COPD patients, and healthy controls. The miRNA pattern from patients with lung cancer and COPD were more similar to each other than to the healthy controls. However, we were able to discriminate lung cancer patients and COPD patients with 90.4% accuracy, 89.2% specificity, and 91.7% sensitivity. In total, 140 miRNAs were significant for the comparison COPD and controls, 61 miRNAs were significant for the comparison lung cancer and controls, and 14 miRNAs were significant for the comparison lung cancer and COPD. Screening target databases yielded over 400 putative targets for those 14 miRNAs. The predicted mRNA targets of three of the 14 miRNAs were significantly up-regulated in PBMCs of lung cancer patients compared to patients with non-malignant lung diseases. In conclusion, we showed that blood miRNA signatures are suitable to distinguish lung cancer from COPD. © 2011 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Worldwide more than one million people die each year from lung cancer making it the leading cause of cancer related deaths [1]. The five-year survival rate is among the lowest of all cancers. A major challenge is the prevailing lack of specific biomarkers for detection and monitoring of lung cancer. While single markers show a rather low sensitivity and specificity for the identification of lung cancer and other cancers, complex marker signatures often reach a significantly higher degree of accuracy. Most recently, microRNA (miRNA) expression profiles have been proposed as potential biomarkers for cancer diagnosis and treatment monitoring [2]. MicroRNAs are small (17–24 nt) non-coding RNA transcripts, involved in physiological and pathophysiological processes including cancer development through the regulation of gene expression [3,4].

∗ Corresponding author at: Institute of Human Genetics, Saarland University, Medical School, Building 60, 66421 Homburg, Germany. Tel.: +49 06841 1626038; fax: +49 06841 1626186. E-mail address: [email protected] (E. Meese). 1 Both authors equally contributed to this work. 0169-5002/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.lungcan.2011.02.003

Based on the differential expression of miRNAs in tumors, miRNA expression signatures allow classification of many cancers, including lung cancer [5]. First attempts were made to differentiate specific subtypes of lung cancer such as primary lung cancer from metastatic lung tumors by miRNA signatures [6]. While the overwhelming majority of published miRNA signatures has been established on tumor cells, recent studies have identified specific miRNA signatures in sera and peripheral blood cells from cancer patients [7–11]. The remarkable stability of miRNAs makes miRNA signatures in body fluids especially intriguing for future minimally invasive diagnostics [10,12]. Recently, we identified a microRNA expression signature in blood cells that allows for differentiation between lung cancer patients and healthy individuals [13]. To further evaluate the utility of blood based miRNA signatures for diagnostic purposes, it is important to compare cancer not only to healthy controls but also to other non-cancer diseases of the same organ. Here, we analyzed Chronic Obstructive Pulmonary Disease (COPD), a common pulmonary affliction encompassing chronic obstructive bronchitis and lung emphysema [14]. COPD is a global burden affecting 10–15% of adults older than 40 years [15]. COPD is not only a common co-morbidity but also precedes lung cancer in 50–90% of cases [16]. We compared the expression of 863 miRNAs in blood cells of lung cancer patients, patients suffering

42

P. Leidinger et al. / Lung Cancer 74 (2011) 41–47

from COPD, and healthy individuals to further our knowledge of miRNA signatures in blood cells of lung cancer and COPD patients. In detail, we address the following questions: How similar are miRNA signatures among the aforementioned three groups? Can lung cancer be distinguished from COPD by blood based miRNA signatures? Which miRNAs contribute most to a separation of lung cancer from COPD? What are the predicted target genes of miRNAs differentially expressed in blood cells of patients with lung cancer and COPD. Overall, the study shows that blood miRNA signatures are suitable to distinguish lung cancer from COPD. Furthermore, we identified significant deregulated miRNAs and analyzed their putative target genes by database screening. 2. Materials and methods See Online Data Supplement for additional methologic details. 2.1. Blood samples The local ethics committee approved the analysis of miRNA expression in blood from lung cancer patients, COPD patients, and healthy subjects. We obtained 5 mL peripheral blood in PAXgene Blood RNA tubes (BD, Franklin Lakes, NJ, USA) with patients’ informed consent from 28 lung cancer patients from Department of Cardiothoracic Surgery, Voelklingen Heart Center and 24 patients with COPD from Department of Pneumology, Voelklingen Lung Center. The blood was drawn prior to surgery of lung cancer. Spirometrical data for lung cancer and COPD patients were available. Control blood (5 ml) was collected from 19 volunteers without any known abnormality of the lung by the Department of Human Genetics, Medical School, Saarland University. Again, informed consent was obtained from each study subject. More detailed information on patients and controls is given in Table E1 in the Online Data Supplement.

Fig. 1. First versus second principal component (PC) for miRNA expression pattern generated from blood cells of lung cancer patients, COPD patients, and healthy controls. The colored crosses show the quantiles for each group and the median of both principal components. Outlined peaks are indicated for the control cohort in the second PC on top and for the COPD cohort in the first PC on the right.

present in the data. This is achieved by a transformation to a set of uncorrelated vectors, the principal components. These components are ordered such that the first components contain most of the information present in the complete data set. 2.5. Pathway analysis

2.2. RNA isolation and microarray screening The isolation of total RNA from whole blood was performed using the miRNeasy kit (Qiagen GmbH, Hilden) as previously described [17]. RNA was then analyzed with the Geniom Realtime Analyzer (GRTA, Febit Biomed GmbH, Heidelberg, Germany) using the Geniom Biochip miRNA Homo sapiens containing 7 replicates of at least 863 miRNAs as annotated in the Sanger miRBase version 12.0 [18,19] and microfluidic-based enzymatic on-chip labeling of miRNAs (MPEA [20], as previously described [17]). Microarray data are publicly available in the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/, GSE24709). 2.3. Statistical analysis of microarray data We carried out parametric t-test (unpaired, two-tailed) for each miRNA separately, to detect miRNAs that show different behavior in different groups of blood donors. The resulting p-values were adjusted for multiple testing by Benjamini–Hochberg [21,22] adjustment. Classification of samples using miRNA patterns was carried out using Support Vector Machines (SVM, [23]) as implemented in the R e1071 package [24] using 20 repetitions of standard 10-fold cross-validation and a subset selection technique based on t-test. 2.4. Principal component analysis The fundamental idea of principal component analysis (PCA) [25,26] is to reduce the dimensionality of data sets containing correlated variables while retaining most variation (information)

To detect pathways, whose members are significantly enriched for targets of miRNAs being deregulated, we utilized the “miRNA to pathway dictionary” [27]. Here, all significant target pathways are listed for each miRNA. These have been computed by separately applying standard Over-Representation Analysis for each set of miRNA targets. 3. Results Using the Geniom Real Time Analyzer Platform, we analyzed the expression of the 863 human miRNAs annotated in miRBase version 12.0. In total, we screened the miRNA expression in blood cells from 71 different individuals, including 28 lung cancer patients, 24 COPD patients, and 19 healthy controls. The group of lung cancer patients included both patients without COPD (n = 15) and with COPD (n = 13). Intensity values of all analyzed miRNAs and samples were subjected to quantile normalization prior to all other computations. We first computed the principle components of the complete data set containing 863 × 71 data points. As shown in Fig. 1, samples of the healthy control individuals and the lung cancer patients exhibited higher variability in the first principal component. Samples of the COPD patients were more variable in the second principal component. Lung cancer patients and the COPD patients were more similar to each other than lung cancer patients to controls or COPD patients to controls. This finding led us to address the question whether differential miRNA expression permits discrimination of lung cancer patients from COPD patients. For the separation of lung cancer from COPD, we obtained the highest area under the receiver operator character-

P. Leidinger et al. / Lung Cancer 74 (2011) 41–47 Table 1 Classification performance of the Support Vector Machine. Scenario

Biomarkers

Accuracy

Specificity

Sensitivity

Control vs. Lung cancer Control vs. COPD Lung cancer vs. COPD

70 220 250

0.874 1 0.904

0.884 1 0.892

0.863 1 0.917

istics curve (AUC) value (0.876) for hsa-miR-675. We combined the information of several biomarkers and tested different classification approaches including Linear Discriminant Analysis and several Support Vector Machines, and different subset selection techniques including stepwise forward wrapping and filtering based on t-test significance values or AUC values. The best approach was radial basis function Support Vector Machine combined with t-test filtering. In agreement with the aforementioned results, we obtained high accuracy values for the classification of COPD versus healthy controls. By using only two miRNAs (hsa-miR-200a* and hsamiR-518a) we separated COPD from healthy controls with 83.7% accuracy. By using 220 miRNAs accuracy, specificity, and sensitivity increased to near 100% (Table 1). Using the same approach and 70 miRNAs, we separated lung cancer from healthy controls with an accuracy of 87.4%, a specificity of 88.4%, and a sensitivity of 86.3%. The largest number of miRNAs was required for accurate discrimination of lung cancer from COPD. By using 250 miRNAs, we reached an accuracy, specificity, and sensitivity of 90.4%, 89.2%, and 91.7%, respectively (Fig. 2 and Table 1). However, even a small subset of only six of the 250 markers (2.4%) still separates lung cancer from COPD with an accuracy, specificity, and sensitivity of 86.25%, 83.3%, and 89.2%. Next, we carried out pairwise group comparisons to identify miRNAs that contribute significantly to the different separations, including COPD versus healthy controls, COPD versus lung cancer, and lung cancer versus healthy controls. For each comparison we computed significance values by using t-test and AUC. Adjusting the significance values for multiple testing by controlling the false discovery rate (FDR) according to Benjamini–Hochberg and setting the alpha level to 0.01 we identified miRNAs that were significant for each group comparison. We found 140 significant miRNAs for the comparison of COPD versus controls, 61 miRNAs for the comparison of lung cancer versus controls, and 14 miRNAs for the comparison of

Fig. 2. Accuracy, specificity and sensitivity as function of the classification subset size for the scenario lung cancer versus COPD. The numbers of miRNAs are indicated on the x-axis and the degree of accuracy, sensitivity, and specificity is indicated on the y-axis. With increasing number of miRNAs the accuracy increases and fluctuates near 100% due to the various combinations of miRNAs.

43

lung cancer versus COPD. These results may be slightly biased due to the different group sizes that influence the significance value. Nevertheless, the findings reinforce the result of the principal component analysis in that both lung cancer and COPD can clearly be separated from healthy controls by an extended number of significant miRNAs. We found far fewer significant miRNAs for the separation of lung cancer from COPD. The most significant up- and down-regulated miRNAs (p-value < 0.001) are shown in Fig. 3A–C. Remarkably, we did not find significant miRNAs for the separation of lung cancer patients with COPD and lung cancer patients without COPD using an adjusted t-test p-value of <0.05. In addition, we analyzed the influence of other parameters, including gender, lung cancer histology or COPD GOLD (Global initiative for chronic Obstructive Lung Disease) status by pairwise group comparison. In the group of lung cancer patients we compared males versus females, smokers versus non-smokers, and squamous cell lung carcinoma versus adenocarcinoma. In the group of the COPD patients we compared males versus females, COPD with emphysema versus COPD without emphysema, and GOLD IV versus GOLD II and III. For all six abovementioned comparisons we did not find significant miRNAs with an adjusted p-value < 0.05. The t-test p-values and the AUC values for the most significant miRNA and all pairwise comparisons are summarized in Table 2. Out of the 14 miRNAs that were significant different in the comparison lung cancer versus COPD, miRNA hsa-let-7d* and hsamiR-328 were also significant for the separation between lung cancer and healthy controls and eight miRNAs (hsa-miR-26a, hsamiR-641, hsa-miR-383, hsa-miR-940, hsa-miR-662, hsa-miR-92a, hsa-miR-369-5p, hsa-miR-636) were significant for the separation between COPD and healthy controls. The miRNAs hsa-miR-675, hsa-miR-93*, hsa-miR-513b, and hsa-miR-1224-3p were significant for the separation between lung cancer and COPD, only. Table 3 lists the median signal intensities and p-values of the aforementioned 14 miRNAs for all three comparisons. Fig. 4 provides a graphical presentation of the median signal intensities of the 14 miRNAs for all blood samples. Additionally, we computed a multiple sequence alignment of the 14 miRNAs to detect a potential redundancy of the miRNA sequences. The alignment analysis did not reveal pairs of miRNAs with significant sequence similarity. We also calculated pairwise correlation values for the 14 miRNAs. Only hsa-miR-7d* and hsa-miR-328 showed a high correlation value of 0.797. We performed database analysis to identify putative targets of the 14 miRNAs that were significant for the separation between lung cancer and COPD. Here, we identified 424 putative target genes according to the microcosm database (http://www.ebi.ac.uk/enright-srv/microcosm/). The most frequently predicted gene was IQWD1 that was targeted by 3 miRNAs, namely hsa-let-7d*, hsa-miR-92, and hsa-miR-636. Five genes were potential targets of two miRNAs each. Further, two of the 14 miRNAs identified nine experimentally validated targets according to the miR2Disease database (http://www.mir2disease.org/). Specifically, for hsa-miR-328 we found three of the validated target genes and for hsa-miR-26a six of the validated target genes as summarized in Table 4. Hsa-miR-328 was up-regulated and hsa-miR-26a was down-regulated in lung cancer patients, as compared to COPD patients. To relate the miRNA expression to the expression of the target genes, we analyzed the database Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/). The dataset GSE13255 provides the gene expression data of peripheral blood mononuclear cells (PBMCs) of patients with lung cancer and patients with non-malignant lung diseases, many of them COPD. We extracted 15,227 features from the GEO dataset GSE13255 and computed the median fold change in lung cancer and in non-malignant lung diseases for each of these features. Using the list of genes sorted by

44

P. Leidinger et al. / Lung Cancer 74 (2011) 41–47

Table 2 t-Test p-values and AUC values for the most significant miRNA and all pairwise comparisons. Comparison

Most significant miRNA

Adjusted p-value

AUC

Lung carcinoma smoker vs. lung carcinoma non-smoker Squamous cell lung carcinoma vs. adenocarcinoma Lung carcinoma male vs. lung carcinoma female COPD GOLD IV vs. COPD GOLD II, III COPD with emphysema vs. COPD without emphysema COPD male vs. COPD female Lung cancer vs. COPD Healthy vs. COPD Healthy vs. lung cancer

hsa-miR-199b-3p hsa-miR-33a* hsa-miR-635 hsa-let-7b hsa-miR-200a* hsa-miR-935 hsa-miR-662 hsa-miR-518b hsa-miR-20b

0.99639 0.68642 0.98572 0.96453 0.98509 0.88926 0.0001 0.00001 0.00006

0.183 0.094 0.209 0.126 0.252 0.829 0.884 0.947 0.926

Table 3 Significant markers for differentiation of lung cancer versus COPD (p-value 0.01). miRNA

Control

COPD

Lung cancer

Control vs. COPD

Control vs. Lung cancer

Lung cancer vs. COPD

hsa-miR-641 hsa-miR-662 hsa-miR-369-5p hsa-miR-383 hsa-miR-636 hsa-miR-940 hsa-miR-26a hsa-miR-92a hsa-miR-328 hsa-let-7d* hsa-miR-1224-3p hsa-miR-513b hsa-miR-93* hsa-miR-675

76.68 90.65 33.46 74.96 246.59 225.92 7269.84 13651.44 59.92 70.76 137.63 66.76 893.5 254.2

143.15 23.1 97.1 142.06 106.39 152.89 7975.44 9554.17 76.93 102.75 109.61 80.41 1303.7 149.11

59.58 95.46 33.25 73.83 222.87 247.83 5568.45 13651.44 208.31 250.42 233.37 39.04 2321.35 287.83

0.00013 0.0003 0.00041 0.00122 0.00186 0.00583 0.00931 0.00957 0.96379 0.05763 0.08731 0.03264 0.99299 0.04421

0.90088 0.5175 0.60298 0.87052 0.72712 0.94678 0.21746 0.80809 0.00428 0.00006 0.86406 0.12765 0.01562 0.04842

0.00075 0.0001 0.0001 0.00316 0.00016 0.00683 0.00047 0.00156 0.00126 0.00278 0.00316 0.00411 0.0068 0.00156

their median fold change [38], we carried out a gene set enrichment analysis using GeneTrail (standard parameters, p-value of targets: 0.01) to detect miRNAs whose targets are significantly upor down-regulated in lung cancer as compared to non-malignant lung pathologies. Of the 14 miRNAs, we identified three miRNAs hsa-miR-662, hsa-miR-328, and hsa-miR-675, whose predicted targets showed significant up-regulation in PBMCs of lung cancer patients as compared to PBMCs of patients with non-malignant lung pathologies in the dataset GSE13255. Notably, the three miRNAs were up-regulated in our study in blood cells of lung cancer patients compared to blood cells of COPD patients. 4. Discussion Here, we compared miRNA expression profiles of peripheral blood cells in patients with lung cancer, patients with COPD, and healthy individuals. As expected, the miRNA signatures of lung cancer patients and COPD patients were more similar to each other than to the signature of healthy controls as revealed by principle components analysis. However, the miRNA signatures of blood cells were still so different between lung cancer patients and COPD patients as to permit separation between these diseases with an accuracy of 90.4%. Previously, messenger RNA (mRNA) expression profiles in peripheral blood cells have been reported for other Table 4 Validated target genes and miRNAs extracted from miR2Disease. miRNA

Gene

hsa-miR-328

ABCG2 BACE1 CD44

hsa-miR-26a

CCND2 CCNE2 EZH2 HMGA1 HMGA2 PTEN

human cancer including renal cell carcinoma and breast cancer [28–30]. A recent study compared the mRNA expression pattern in PBMCs from patients with lung cancer and from controls with non-malignant lung diseases [31]. The authors found a 29-gene signature, which distinguished both patient groups with 91% sensitivity and 79% specificity. For future diagnostic purposes, it may be beneficial to utilize miRNA patterns of blood cells instead of mRNA expression patterns. The smaller number of miRNAs relative to the number of mRNAs is likely to yield a signature with less background noise. The nature of the blood cells that contribute to the miRNA signature is still elusive. The number of circulating tumor cells appears far too small to strongly influence the miRNA expression pattern of whole blood. More likely, the tumor influences immune cells by releasing factors such as cytokines. It has been shown that tumors can induce suppressor cells that are not only found in lymph nodes and in spleen, but also in peripheral blood [32–36]. The finding of a more pronounced expression signature in patients with advanced tumors is in agreement with the idea that malignant cells can induce a cancer-specific mRNA and miRNA expression pattern in peripheral blood cells. Comparison of mRNA expression signatures between matched lung cancer samples taken presurgery and postsurgery frequently shows a diminished mRNA signature in blood cells after surgery [31]. This finding also supports the idea of a communication between tumor and peripheral blood cells. In our study, we found 14 miRNAs expressed differentially in blood cells of lung cancer patients compared to COPD patients. The rather limited number of studies on expression signatures of peripheral blood cells complicates the interpretation of the biological meaning of these miRNAs. Furthermore, all studies to date used different technical approaches for various diseases. As mentioned previously, the study of Showe et al. which most closely resembles our analysis identified a 29-gene signature in PBMCs of patients with lung cancer as compared to patients with nonmalignant lung diseases [31]. Out of this 29-gene signature, the gene ASCC3 (activating signal cointegrator 1 complex subunit 3) was predicted as target of the miRNA hsa-miR-26a that was differ-

P. Leidinger et al. / Lung Cancer 74 (2011) 41–47

45

Fig. 4. (A and B) Bar chart of the normalized intensities of the 14 miRNAs being significant for the comparison between lung cancer and COPD. The median expression is shown for all blood donors, including lung cancer patients, COPD patients, and healthy controls. The bar charts are shown in a splitted diagram (A and B) due to different scales on the y-axis.

Fig. 3. (A–C) Scatterplots of miRNA intensity values measured in blood cells of (A) lung cancer patients and healthy controls, (B) COPD patients and healthy controls, and (C) lung cancer patients and COPD patients. The plots are shown on a logarithmic scale. Up- and down-regulated (p < 0.001) miRNAs are indicated by red and green font. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

entially expressed in lung cancer blood cells compared to COPD blood cells in this study. ASCC3 is a transcriptional coactivator involved in several pathways, including stimulation of transactivation by SRF (serum response factor), AP-1 (activating protein 1) and NF-kappaB (nuclear factor kappaB) through direct binding to SRF, c-Jun, p50, and p65. ASCC3 was 1.87 fold up-regulated in PBMC from lung cancer patients than in PBMC from patients with non-malignant lung diseases. In contrast, we found a 1.4 fold down-regulation of the miRNA hsa-miR-26a. To elucidate the function of the down-regulation of hsa-miR-26a in blood of lung cancer patients compared to COPD patients, further knowledge on the protein expression of ASCC3 in blood cells of lung cancer patients is required. The above mentioned miRNA hsa-miR26a shows six experimentally validated targets including cyclins CCND2 and CCNE2 that are required for G1/S transition, EZH2 that plays a role in the haematopoietic system, HMGA1 and HMGA2 that are transcription regulators, and the tumor suppressor PTEN. Independent analysis of the complete GEO dataset (GSE13255) of the study of Showe et al. yielded three miRNAs (hsa-miR-662, hsa-miR-328, and hsa-miR-675) with predicted targets that were up-regulated in PBMCs of lung cancer patients as compared to non-malignant lung pathologies [31]. In our study, the respective three miRNAs were also up-regulated in blood cells of lung cancer patients compared to COPD. For the miRNA hsa-miR-328 we identified experimentally validated targets including BACE1 (protease, involved in Alzheimers disease), ABCG2 (part of ABC transporter), and CD44 (cell-surface glycoprotein involved in cell–cell interac-

46

P. Leidinger et al. / Lung Cancer 74 (2011) 41–47

tion, cell adhesion, and cell motility). In addition, one needs to bear in mind that the miRNA expression and the mRNA expression signatures stem from two different studies. While our miRNA analysis compared lung cancer patients with COPD patients, the study by Showe compared lung cancer patients with various kinds of noncancer lung diseases, including a large proportion of COPD patients but also patients with granulomous inflammation, non-malignant nodules, sarcoidosis, and pneumonia [31]. COPD as common pulmonary affection is not only a common lung cancer co-morbidity but it is also associated with higher risk for the development of lung cancer [37]. With the present proofof-principle study we demonstrate that lung cancer patients and patients with COPD show a significantly different miRNA expression signature in peripheral blood. High specificity for a certain disease is a prerequisite for the development of a diagnostic test. Using 250 miRNAs we were able to differentiate lung cancer patients and COPD patients with 90.4% accuracy, 89.2% specificity, and 91.7% sensitivity. Even with only six miRNAs we still obtained an accuracy of 86%. As a large proportion of the lung cancer patients included in our study also suffered from COPD we searched for miRNAs that were differentially expressed between lung cancer patients with and without COPD. Since we did not find miRNAs significant deregulated between the two patient groups the classification yielded only about 50% accuracy. In agreement with our study, Showe et al. did not differentiate lung cancer patients with obstructive lung disease from lung cancer patients without obstructive lung disease. It is legitimate to speculate that expression pattern elicited by lung cancer may override the pattern elicited by COPD. Our samples also did not allow to test for the influence of smoking due to the lack of complete information on the smoking habits of the lung cancer patients and controls. Interestingly, there is evidence that the mRNA signature in lung cancer patients does not seem dependent on the smoking status [31]. 5. Conclusion In conclusion, we showed for the first time that lung cancer patients are distinguishable from COPD patients by their peripheral miRNA expression profile with high accuracy. Therefore, our proof-of-principle study strengthens the hypothesis that blood based miRNA signatures might be potential cancer biomarker. In combination with imaging techniques blood based miRNA signatures might contribute to an earlier lung cancer detection and thus improve the survival rates. Prospective studies will clarify whether lung cancer specific miRNA signatures can be detected very early in or even before lung cancer development and if these signatures are also suitable to distinguish lung cancer patients from COPD patients. It may even be possible to take an advantage of specific miRNAs in future lung cancer therapies. Conflict of interest Andreas Keller and Anne Borries are employees of febit biomed GmbH. Acknowledgements We acknowledge the technical assistance of Hannah Schroers and Pamela Haeberle. We thank Dr. Jack Leonard for carefully proofreading and correcting the manuscript. Funding: This work was supported by funding of the German Ministry of Research Education (BMBF) under contract 01EX0806, the Hedwig-Stalter foundation, Deutsche Forschungsgemeinschaft (DFG, LE2783/1-1), and HOMFOR 2010.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.lungcan.2011.02.003. References [1] Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin 2005;55(2):74–108. [2] Roa W, Brunet B, Guo L, Amanie J, Fairchild A, Gabos Z, Nijjar T, Scrimger R, Yee D, Xing J. Identification of a new microRNA expression profile as a potential cancer screening tool. Clin Invest Med 2010;33(2):E124. [3] Sassen S, Miska EA, Caldas C. MicroRNA: implications for cancer. Virchows Arch 2008;452(1):1–10. [4] Esquela-Kerscher A, Slack FJ. Oncomirs—microRNAs with a role in cancer. Nat Rev Cancer 2006;6(4):259–69. [5] Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D, et al. MicroRNA expression profiles classify human cancers. Nature 2005;435(7043):834–8. [6] Barshack I, Lithwick-Yanai G, Afek A, Rosenblatt K, Tabibian-Keissar H, Zepeniuk M, Cohen L, Dan H, Zion O, Strenov Y, Polak-Charcon S, Perelman M. MicroRNA expression differentiates between primary lung tumors and metastases to the lung. Pathol Res Pract 2010. [7] Wang K, Zhang S, Marzolf B, Troisch P, Brightman A, Hu Z, et al. Circulating microRNAs, potential biomarkers for drug-induced liver injury. Proc Natl Acad Sci USA 2009;106(11):4402–7. [8] Wang J, Chen J, Chang P, LeBlanc A, Li D, Abbruzzesse JL, et al. MicroRNAs in plasma of pancreatic ductal adenocarcinoma patients as novel blood-based biomarkers of disease. Cancer Prev Res (Phila Pa) 2009;2(9):807–13. [9] Cortez MA, Calin GA. MicroRNA identification in plasma and serum: a new tool to diagnose and monitor diseases. Expert Opin Biol Ther 2009;9(6):703–11. [10] Gilad S, Meiri E, Yogev Y, Benjamin S, Lebanony D, Yerushalmi N, et al. Serum microRNAs are promising novel biomarkers. PLoS One 2008;3(9):pe3148. [11] Chin LJ, Slack FJ. A truth serum for cancer—microRNAs have major potential as cancer biomarkers. Cell Res 2008;18(10):983–4. [12] Chen X, Ba Y, Ma L, Cai X, Yin Y, Wang K, et al. Characterization of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases. Cell Res 2008;18(10):997–1006. [13] Keller A, Leidinger P, Borries A, Wendschlag A, Wucherpfennig F, Scheffler M, et al. miRNAs in lung cancer—studying complex fingerprints in patient’s blood cells by microarray experiments. BMC Cancer 2009;9:353. [14] Tammemagi CM, Neslund-Dudas C, Simoff M, Kvale P. Impact of comorbidity on lung cancer survival. Int J Cancer 2003;103(6):792–802. [15] van Gestel YR, Hoeks SE, Sin DD, Huzeir V, Stam H, Mertens FW, et al. COPD and cancer mortality: the influence of statins. Thorax 2009;64(11):963–7. [16] Young RP, Hopkins RJ, Christmas T, Black PN, Metcalf P, Gamble GD. COPD prevalence is increased in lung cancer, independent of age, sex and smoking history. Eur Respir J 2009;34(2):380–6. [17] Keller A, Leidinger P, Lange J, Borries A, Schroers H, Scheffler M, et al. Multiple sclerosis: microRNA expression profiles accurately differentiate patients with relapsing-remitting disease from healthy controls. PLoS One 2009;4(10):e7440. [18] Griffiths-Jones S, Grocock RJ, van Dongen S, Bateman A, Enright AJ. miRBase: microRNA sequences, targets and gene nomenclature. Nucleic Acids Res 2006;34:D140–4 [database issue]. [19] Griffiths-Jones S. miRBase: the microRNA sequence database. Methods Mol Biol 2006;342:129–38. [20] Vorwerk S, Ganter K, Cheng Y, Hoheisel J, Stahler PF, Beier M. Microfluidic-based enzymatic on-chip labeling of miRNAs. Nat Biotechnol 2008;25(2/3):142–9. [21] Benjamini Y, Drai D, Elmer G, Kafkafi N, Golani I. Controlling the false discovery rate in behavior genetics research. Behav Brain Res 2001;125(1/2):279–84. [22] Hochberg Y. A sharper bonferroni procedure for multiple tests of significance. Biometrica 1988;75:185–93. [23] Vapnik V. The nature of statistical learning theory. 2nd ed. New York: Springer; 2000. [24] Team RR. A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2008. [25] Dunteman GH. Principal component analysis. Sage Publications; 1989. [26] Mardia K, Kent JT, Bibby JM. Multivariate analysis. New York: Academic Press; 1979. [27] Backes C, Meese E, Lenhof HP, Keller A. A dictionary on microRNAs and their putative target pathways. Nucleic Acids Res 2010. [28] Burczynski ME, Twine NC, Dukart G, Marshall B, Hidalgo M, Stadler WM, et al. Transcriptional profiles in peripheral blood mononuclear cells prognostic of clinical outcomes in patients with advanced renal cell carcinoma. Clin Cancer Res 2005;11(3):1181–9. [29] Sharma P, Sahni NS, Tibshirani R, Skaane P, Urdal P, Berghagen H, et al. Early detection of breast cancer based on gene-expression patterns in peripheral blood cells. Breast Cancer Res 2005;7(5):R634–44. [30] Twine NC, Stover JA, Marshall B, Dukart G, Hidalgo M, Stadler W, et al. Diseaseassociated expression profiles in peripheral blood mononuclear cells from patients with advanced renal cell carcinoma. Cancer Res 2003;63(18):6069–75. [31] Showe MK, Vachani A, Kossenkov AV, Yousef M, Nichols C, Nikonova EV, et al. Gene expression profiles in peripheral blood mononuclear cells can distinguish patients with non-small cell lung cancer from patients with nonmalignant lung disease. Cancer Res 2009;69(24):9202–10.

P. Leidinger et al. / Lung Cancer 74 (2011) 41–47 [32] Marigo I, Dolcetti L, Serafini P, Zanovello P, Bronte V. Tumor-induced tolerance and immune suppression by myeloid derived suppressor cells. Immunol Rev 2008;222:162–79. [33] Serafini P, Mgebroff S, Noonan K, Borrello I. Myeloid-derived suppressor cells promote cross-tolerance in B-cell lymphoma by expanding regulatory T cells. Cancer Res 2008;68(13):5439–49. [34] Sinha P, Clements VK, Bunt SK, Albelda SM, Ostrand-Rosenberg S. Cross-talk between myeloid-derived suppressor cells and macrophages subverts tumor immunity toward a type 2 response. J Immunol 2007;179(2):977–83. [35] Salvadori S, Martinelli G, Zier K. Resection of solid tumors reverses T cell defects and restores protective immunity. J Immunol 2000;164(4): 2214–20.

47

[36] Diaz-Montero CM, Salem ML, Nishimura MI, Garrett-Mayer E, Cole DJ, Montero AJ. Increased circulating myeloid-derived suppressor cells correlate with clinical cancer stage, metastatic tumor burden, and doxorubicin-cyclophosphamide chemotherapy. Cancer Immunol Immunother 2009;58(1):49–59. [37] Turner MC, Chen Y, Krewski D, Calle EE, Thun MJ. Chronic obstructive pulmonary disease is associated with lung cancer mortality in a prospective study of never smokers. Am J Respir Crit Care Med 2007;176(3):285–90. [38] Backes, C, Keller, A, Kuentzer, J, Kneissl, B, Comtesse, N, Elnakady, YA, Müller, R, Meese, E, Lenhof, HP. GeneTrail - advanced gene set enrichment analysis. Nucleic Acid Research, Web Server Issue 2007, http://genetrail.bioinf.unisb.de/.