Background:

Because of limited stability and sensitivity, circulating miRNAs as noninvasive biomarkers have not so far been used for early diagnosis and prognosis of non–small cell lung cancer (NSCLC) in clinic. Therefore, it is imperative to find more reliable biomarker(s).

Methods:

We performed one of most sensitive qRT-PCR assays, S-Poly(T) Plus, to select differently expressed miRNAs from genome-wide miRNA profiling. miRNA candidates were validated through a three-phase selection and two validation processes with 437 NSCLC cases and 415 controls.

Results:

A unique set of 7 and 9 miRNAs differed significantly in adenocarcinoma (ADC) and squamous cell carcinoma (SCC) samples compared with those in controls, of which, there were 5 universal biomarkers for NSCLC (ADC or SCC). Ten of 11 miRNAs could discriminate early stage (stage I) of NSCLC from healthy individuals. Risk score was obtained from the validation set-1 and was tested using the ROC curves with a high area under ROC curve of 0.89 in ADC and 0.96 in SCC. Ultimately, potential biomarkers and the risk score were verified by the validation set-2 with a sensitivity of 94% and a specificity of 91.6% in ADC, and a sensitivity of 98.5% and a specificity of 51.5% in SCC, respectively.

Conclusions:

Taken together, 7 miRNAs and 9 miRNAs may provide noninvasive biomarkers for diagnosis and prognosis in ADC and SCC, respectively.

Impact:

On the basis of our sensitive and accurate method, we hope that these candidate miRNAs may have strong impact on the early lung cancer diagnosis.

Lung cancer is the leading cause of cancer-related deaths in the world and 80% of the patients have non–small cell lung cancer (NSCLC), which is often diagnosed at later stages with around 15% of 5-year survival rates. If discovered at stage I, 60%–80% of patients can survive (1). Unfortunately, current approaches to cancer screening are invasive and it is difficult to detect cancer at very early stage. Subsequently, the discovery and application of molecular biomarkers has opened new sights for diagnosis and prediction of cancer. Recently, the advances in miRNA biomarkers have generated some candidate markers of lung cancer with potential clinical values (1, 2). Moreover, it was reported that miRNA expression analyses in plasma samples collected 1–2 years before the onset of disease, at the time of CT detection and in disease-free smokers, resulted in the generation of miRNA signatures with strong predictive, diagnostic, and prognostic potential (3). However, many previously published studies tested either single miRNA as a diagnostic marker or validated in rather small patient cohorts. In addition, most of the studies lacked validation in different cohorts. Until today, these published results have not been applied for clinic purpose. Therefore, it is still imperative to find reliable biomarker(s) with stability and reproducibility for clinical applications, thus improving the diagnosis and prognosis of this disease.

The miRNAs are small noncoding RNAs (∼22 nucleotides), regulating the gene expression at the posttranscriptional level by binding the 3′-untranslated region (3′-UTR) of target mRNAs (4). Physiologically, miRNAs have relevance for cell differentiation, proliferation, apoptosis, and many diseases including cancer (5–7). One miRNA may regulate hundreds of mRNAs, as a result, miRNA expression patterns can be surprisingly informative (8), and to an extent, may reflect the differentiation state of the tumors (9). The miRNAs were originally studied in tissue, but multiple studies have demonstrated that miRNAs are detectable and highly stable in the circulation (10). Given the above findings, circulating miRNAs show great potential as novel noninvasive biomarkers for cancers (11–13).

To determine the expression pattern of miRNAs, an accurate and inexpensive profiling approach is needed. Microarrays or RNA sequences have been used for miRNA high-throughput profiling (14), but the sensitivity and dynamic range have been problematic and are, therefore, best suited as discovery tools in quantitative assay platforms (8). The qRT-PCR–based miRNA expression assay might be one of the most sensitive techniques for miRNA detection (15), of which, the unique design of S-Poly(T) Plus method showed an increased sensitivity when compared with others (16, 17). In addition, the S-Poly(T) Plus also has the attractive properties of increased simplicity, high efficiency, and low cost, allowing the detection of expression pattern of more miRNAs across a large panel of samples.

In this article, we aim to use genome-wide expression profiles from large cohort of patients with NSCLC to conduct an unbiased approach for identification of potentially prognostic circulating miRNAs. First, extensive literatures' search was performed. The keyword “microRNA/miRNA” was combined with the search string “cancer”. A total of 486 miRNAs were collected from the literatures based on the searched results. Afterwards, these miRNAs were assessed with large-scale datasets from patients with NSCLC and healthy volunteers; altered miRNA subsets were subsequently quantified in samples from 288 patients with NSCLC and 266 healthy donors, and then potential biomarkers and risk score were revalidated in different set of plasma samples from 149 patients with NSCLC and 149 healthy controls. Considering the accessibility and stability of circulating miRNAs, 7 miRNAs and 9 miRNAs provide noninvasive biomarkers for diagnosis and prognosis in ADC and SCC, respectively.

Plasma collection

To control the disease heterogeneity, we used patients with two common histologic subtypes of NSCLC, adenocarcinoma (ADC) and squamous cell carcinoma (SCC). Demographic and etiologic characteristics of ADC and SCC have been described previously (18, 19). All the patients were first diagnosed and the blood was collected ahead of any clinic treatments. Whole blood samples were put into EDTA-containing tubes individually and then centrifuged at 3,000 × g for 10 minutes at 4°C. The supernatants were transferred into RNase-free tubes and stored at −80°C. Given that miRNAs also expressed in red blood cells and could have influence on the assessment, hemolysis of the blood samples was excluded from consideration. Plasma samples were collected and examined under blinded testing. Qualified plasma samples were prospectively collected form 266 healthy donors and 288 patients with NSCLC in Shenzhen People's Hospital (Shenzhen, China), as well as from 149 patients with NSCLC and 149 healthy donors in Cancer Center of Guangzhou Medical University (Guangzhou, China). This study had been approved by the Institutional Ethics Committee at Shenzhen People's Hospital and Cancer Center of Guangzhou Medical University, and informed written consents were obtained from all subjects.

RNA extraction

Total RNA was isolated using our developed serum/plasma (S/P) miRsol method as described previously (16). We added 0.1 pmol/L spiked-in Caenorhabditis elegans cel-miR-54-5p to 100 μL of plasma for normalization.

Polyadenylation, reverse transcription, and RT-PCR

The polyadenylation, reverse transcription, and RT-PCR procedures were conducted exactly according to S-Poly(T) Plus protocol (16). To profile all miRNAs effectively, every 7 of 486 miRNAs were grouped together, and were polyadenylated and reverse transcripted simultaneously in one reaction. The miRNAs with identical forward primers or with more than five base-pairings between the forward primer and reverse transcripted primers should be avoided in a group. All sequences, primers, and probes were listed in the Supplementary Table S1.

RT-PCR was performed in 96-well plates by using ABI StepOnePlus thermal cycler as described previously (16). Each PCR reaction was carried out in duplicate. The miRNA expression level was normalized to spiked-in cel-miR-54. The miRNAs with cycle threshold (Ct) value less than 35 in the panel were included in data analysis.

miRNA profiling

Excellent biomarkers should be qualified in the mixed and individual samples, and in different sets of samples. Therefore, a five-step test was designed to identify prognostic miRNAs for NSCLC, and the whole study flow chart was depicted as in Fig. 1. First, 554 plasma samples from Shenzhen People's Hospital (Shenzhen, China) were enrolled into three cohorts, which were healthy control (N = 266), NSCLC stage I (N = 130), and stage II–IV (N = 158), and afterwards pooled together respectively. The miRNA profiling assay was performed using the S-Poly(T) Plus approach with every single miRNA being detected from each pool. Comparing those in healthy cohort, miRNAs in group stage I or group stage II–IV differ by more than 4-fold changes on outcome were selected and secondly confirmed. In this step, all PCR products were detected by electrophoresis in 3.5% agarose gel, and the miRNAs with nonspecific amplification were excluded; the second screening (further screening) consisted of five cohorts: healthy control (N = 266), ADC stage I (N = 96), ADC stage II–IV (N = 113), SCC stage I (N = 34), and SCC stage II–IV (N = 45). Comparing with those in healthy group, miRNAs from stage I or stage II–IV with 2-fold changes were further studied in ADC and SCC, respectively. Third, candidate miRNAs obtained from the further screening were validated using small number of individual specimens randomly selected from Shenzhen People's Hospital. The selection criterion of the training set was the same as that in further screening. Ultimately, selected miRNAs were validated using each individual of 266 healthy donors and 288 patients with NSCLC from Shenzhen People's Hospital, and then further validated in 149 patients with NSCLC and 149 healthy donors from the Cancer Center of Guangzhou Medical University.

Figure 1.

Overview of the experiment design. Flow diagram illustrates the five steps of miRNAs' selection as potential biomarkers from 486 miRNAs in NSCLC.

Figure 1.

Overview of the experiment design. Flow diagram illustrates the five steps of miRNAs' selection as potential biomarkers from 486 miRNAs in NSCLC.

Close modal

Statistical analysis

Statistical analysis was submitted to the GraphPad Prism 5. It was the unpaired two-tailed Student t test that was used to compare the miRNA expressions in plasma, and the paired two-tailed Student t test in tissue. Data were present as mean ± SE, and variables with P values less than 0.05 were retained. Bivariate logistic regression analysis was performed using SPSS software, version19.0 (SPSS Inc.) and the method was “enter.”

Studied participants and miRNAs

In this study, 486 miRNAs related to cancer were selected as candidate miRNAs (Supplementary Table S1). A total of 852 plasma samples were enrolled, including samples from 266 healthy donors and 288 patients with NSCLC in Shenzhen People's Hospital (Shenzhen, China), as well as from 149 healthy donors and 149 patients with NSCLC in Cancer Center of Guangzhou Medical University (Guangzhou, China). The information of participators was detailed in Supplementary Table S2.

Screening candidate miRNA biomarkers in NSCLC

Early screening set.

On the basis of the S-Poly(T) Plus miRNA assay (16), qRT-PCR was first utilized to detect 486 miRNAs among three pools, which consisted of 266 healthy plasma samples, 130 patient plasma samples with NSCLC stage I, and 158 samples with NSCLC stage II–IV from Shenzhen People's Hospital. These 486 miRNAs were depicted in the volcano plot as in Fig. 2A and B. Among them, fold change between the healthy pool and stage I pool were >2 and >4 for 188 and 91 miRNAs, respectively; between stage II–IV pool were >2 and >4 for 187 and 102 miRNAs, respectively (all P < 0.05; Fig. 2C). When the two results were combined, 125 significantly differentiated miRNAs (fold changes >4 and P < 0.05; Figs. 2D and 3A and B) were selected and subjected to second screening.

Figure 2.

Profiling of 486 miRNAs in plasma of NSCLC. Volcano plot illustrates miRNAs differently expressed in healthy controls compared with stage I patients (A) and stage II–IV patients (B); fold change analysis in number of altered miRNAs (C); 125 miRNAs dysregualted with at least 4-fold changes in control individuals compared with stage I or stage II–IV patients (P < 0.05; D). miRNAs with Ct value >35 were eliminated. Data are presented as fold change (2−ΔΔCt), and expression was normalized to a spiked-in (cel-miR-54). Samples were collected from Shenzhen People's Hospital.

Figure 2.

Profiling of 486 miRNAs in plasma of NSCLC. Volcano plot illustrates miRNAs differently expressed in healthy controls compared with stage I patients (A) and stage II–IV patients (B); fold change analysis in number of altered miRNAs (C); 125 miRNAs dysregualted with at least 4-fold changes in control individuals compared with stage I or stage II–IV patients (P < 0.05; D). miRNAs with Ct value >35 were eliminated. Data are presented as fold change (2−ΔΔCt), and expression was normalized to a spiked-in (cel-miR-54). Samples were collected from Shenzhen People's Hospital.

Close modal
Figure 3.

Hierarchical clustering of 125 miRNAs. Heatmap depicted miRNAs differentially expressed between healthy controls and stage I or stage II–IV patients in ADC (A) and SCC (B). Samples were collected from Shenzhen People's Hospital.

Figure 3.

Hierarchical clustering of 125 miRNAs. Heatmap depicted miRNAs differentially expressed between healthy controls and stage I or stage II–IV patients in ADC (A) and SCC (B). Samples were collected from Shenzhen People's Hospital.

Close modal

Further screening set.

We then performed individual qRT-PCR assay to quantify each of previous identified miRNAs in five pools: Healthy (N = 266), ADC stage I (N = 96), ADC stage II–IV (N = 113), SCC stage I (N = 34), and SCC stage II–IV (N = 45). The miRNA targets that met one of the two criteria would be chosen for the next test: stage I vs. healthy: fold-change>2, or stage II–IV vs. healthy: fold change>2. As a result, 30 miRNAs in the ADC and 38 miRNAs in SCC were extracted and further validated in the training step (Supplementary Fig. S1).

Training set.

Candidate miRNAs obtained from the previous step were validated using each individual randomly selected from Shenzhen People's Hospital. The number of samples was as follows: Healthy (N = 20), ADC stage I (N = 20), ADC stage II–IV (N = 20), SCC stage I (N = 10), and SCC stage II–IV (N = 10). The screening criteria were identical with that in further screening set. Consequently, 10 miRNAs in ADC and 9 miRNAs in SCC were extracted (Supplementary Fig. S2A and S2B). However, two miRNAs (hsa-miR-183-5p and hsa-miR-144-5p) in ADC and one miRNA in SCC (hsa-miR-144-3p) were excluded because the Ct values in some individual samples were larger than 35 (marker with the square).

Validation set-1.

Seven miRNAs in ADC and 9 miRNAs in SCC were validated in each individual sample from Shenzhen People's Hospital. The expression of all miRNAs significantly differed between healthy control and patients with NSCLC stage I or stage II, indicating that these miRNAs could differentiate patients with NSCLC from healthy people. Therefore, 7 miRNAs were identified as potential biomarkers in ADC, and they were hsa-miR-26a-5p, hsa-miR-126-5p, hsa-miR-139-5p, hsa-miR-152-3p, hsa-miR-451a, hsa-miR-200c-3p, and hsa-miR-3135b (Fig. 4A); and 9 miRNAs were identified as potential biomarkers in SCC, including hsa-miR-26a-5p, hsa-miR-126-5p, hsa-miR-139-5p, hsa-miR-151a-3p, hsa-miR-151a-5p, hsa-miR-151b, hsa-miR-152-3p, hsa-miR-550a-3p, and hsa-miR-3135b (Fig. 4B), of which, there were 5 universal biomarkers in ADC and SCC, and they were hsa-miR-26a-5p, hsa-miR-126-5p, hsa-miR-139-5p, hsa-miR-152-3p, as well as hsa-miR-3135b.

Figure 4.

Validation of candidate miRNAs in each individual sample from Shenzhen People's Hospital. A, 209 healthy donors and 209 patients with ADC; 79 healthy donors and 79 patients with SCC (B). Data are shown as means ± SE; ns, not significant; *, P < 0.05; ***, P < 0.001. Samples were collected from Shenzhen People's Hospital.

Figure 4.

Validation of candidate miRNAs in each individual sample from Shenzhen People's Hospital. A, 209 healthy donors and 209 patients with ADC; 79 healthy donors and 79 patients with SCC (B). Data are shown as means ± SE; ns, not significant; *, P < 0.05; ***, P < 0.001. Samples were collected from Shenzhen People's Hospital.

Close modal

Sensitivity and specificity of screening measures

To evaluate the screening measure of potential biomarkers in NSCLC, the sensitivity and specificity were illustrated in Fig. 5. In ROC curve analysis, area under the ROC curve ranged from 0.64 to 0.76 in ADC (P < 0.001; Fig. 5A), 0.64 to 0.91 in SCC (P < 0.001; Fig. 5B), and 0.67 to 0.80 for universal biomarkers (P < 0.001; Fig. 5C).

Figure 5.

ROC curves. A, ROC of 7 individual miRNAs and 5 miRNAs as one combined biomarker in ADC; B, ROC of 9 individual miRNAs and 6 miRNAs as one combined biomarker in SCC; C, ROC of 5 individual miRNAs as potential universal biomarker in all samples, as well as 2 miRNAs as one combined biomarker. Samples were obtained from Shenzhen People's Hospital.

Figure 5.

ROC curves. A, ROC of 7 individual miRNAs and 5 miRNAs as one combined biomarker in ADC; B, ROC of 9 individual miRNAs and 6 miRNAs as one combined biomarker in SCC; C, ROC of 5 individual miRNAs as potential universal biomarker in all samples, as well as 2 miRNAs as one combined biomarker. Samples were obtained from Shenzhen People's Hospital.

Close modal

When the sensitivity of one single biomarker will not be sufficient for early detection, it has been suggested that a combination of multiple biomarkers will potentially produce an improved ROC curve in clinical performance (20). On multivariate logistic regression analysis (the entered variables were 7 miRNAs in ADC, 9 miRNAs in SCC, and 5 miRNAs in universal group), the statistical significance threshold for variable entry was P < 0.05, as results, 5/6 of the miRNAs as significant predictors in ADC/SCC were remained and there were only 2 miRNAs left in universal group. In this study, the combined miRNAs as a single biomarker promised satisfactory discrimination over diagnostic tests in ADC (5 miRNAs, AUC = 0.89, P < 0.001; Fig. 5A), SCC (6 miRNAs, AUC = 0.97, P < 0.001; Fig. 5B), and universal group (2 miRNAs, AUC = 0.80, P < 0.001; Fig. 5C). The maximum Youden index (sensitivity+specificity-1) yielded a sensitivity of 98.0% and a specificity of 78.6% in ADC [cutoff = 0.0787, risk score = 0.0676 ± 0.1224 (SD; min = −1.00 and max = 0.187), P < 0.001]; a sensitivity of 100% and a specificity of 75.7% [cutoff = 0.0075, risk score = −0.0012 ± 0.10117 (SD; min = −1.00 and max = 0.022), P < 0.001]; and in universal group, the sensitivity and specificity was 87.5% and 94.6% respectively [cutoff = 0.2128, risk score = 0.1987 ± 0.1129 (SD; min = −0.90 and max = 0.23), P < 0.001; Supplementary Table S3].

Meanwhile, a formula was estimated to determine the predicted probability of being NSCLC (risk score). The developed logistic regression equation also presents the degree of contribution of each variable to the prediction. More specifically, the relationship between the risk score of ADC (P) and the relative expression of predictors is detailed as:

formula

In squamous carcinoma, the equation is as follows:

formula

In universal group, the equation is as follows:

formula

Validation of the biomarkers and risk score in NSCLC (Validation set-2)

To test the diagnostic and prognostic capability of the potential biomarkers identified in this study, 7 miRNAs in ADC and 9 miRNAs in SCC were revalidated using another set of plasma samples from the Cancer Center of Guangzhou Medical University, and all the miRNAs expressed significantly different between the healthy controls and patients with stage I and stage II–IV (Supplementary Fig. S3A and S3B). Therefore, 7 miRNAs in ADC and 9 miRNAs in SCC were identified as potential biomarkers for prognosis of NSCLC. In ROC curve analysis, miRNAs had a much better AUC for differentiating patients and healthy individuals than those in the validation set-1, more specifically, the number ranging from 0.75 to 0.96 (P < 0.001) in ADC (Supplementary Fig. S4A), from 0.80 to 0.99 in SCC (P < 0.001; Supplementary Fig. S4B), and from 0.83 to 0.98 in universal group (P < 0.001; Supplementary Fig. S4C).

In validation set-2, we tested prediction performance of risk score derived from validation set-1. The finial prediction results revealed a sensitivity of 94.0%, a specificity of 91.6%, and an overall accuracy of 92.7% in ADC; and a sensitivity of 98.5%, a specificity of 51.5%, and an overall accuracy of 75.0% in SCC. It was noticed that a high sensitivity produced a much lower specificity (51.5%) in SCC, more specifically, the 26/66 healthy control samples with continuous levels of risk score were classified into SCC group. Therefore, we considered a manually adjusted cutoff (0.14) and it might be better for prognosis with a sensitivity of 98.4%, a specificity of 90.9%, and an overall accuracy of 94.5%. For universal biomarkers, the sensitivity and specificity were 98.7% and 17.4%, respectively; and the overall accuracy only reached 58.1% (Supplementary Table S3). We, therefore, highlighted the separate prediction function of risk score in ADC or SCC for greater overall accuracy.

Validation of the potential biomarker in tissue

To further determine whether our approach could identify a plasma miRNA profile that mirrors the disease progression, we analyzed the expression pattern of potential biomarkers in NSCLC tissues. In general, miRNAs were highly expressed in the adjacent tissues, with lower concentrations in the cancer tissues (Fig. 6). Except for hsa-miR-451a, these results were inversely correlated with results generated from plasma profiles. Tissue samples used in this test were collected from Shenzhen People's Hospital (Shenzhen, China). To verify these results, different tissue samples of NSCLC from Peking University Shenzhen Hospital (Shenzhen, China) were tested. The higher level of miRNAs in adjacent tissues in new samples had similar trends as in the former samples (Supplementary Fig. S5).

Figure 6.

Expression levels of 11 miRNAs in adjacent tissues and caner tissues. Samples were collected from Shenzhen People's Hospital. It was paired two-tailed Student t test that was used in comparing the miRNA expressions between the tumor tissues and adjacent tissues. ns, not significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Figure 6.

Expression levels of 11 miRNAs in adjacent tissues and caner tissues. Samples were collected from Shenzhen People's Hospital. It was paired two-tailed Student t test that was used in comparing the miRNA expressions between the tumor tissues and adjacent tissues. ns, not significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Close modal

The investigation of biological and molecular features of lung cancer is critical for identifying specific risk markers for lung cancer development, thus achieving the earliest prediction and intervention of lung cancer (21). The recently available data from NSCLC offered the opportunity to explore miRNA-based biomarker using miRNA sequence or microarray. In our study, genome-wide miRNA expression profiling was obtained via a different platform in RT-qPCR with higher accuracy. Moreover, our identification of aberrant circulating miRNA in the data from samples has a number of advantages compared with previous studies. Our study also revealed that potential biomarkers for NSCLC diagnosis could be validated in the different subsets of patients. Also, we found that the area under ROC curve was higher in the validation set-2 than that in validation set-1, and the possible reason could be that the batch of samples collected from Cancer Center of Guangzhou Medical University has better quality, with perfect-matched age between patient and healthy group (Supplementary Table S2). Therefore, more restrictive approach and usage of different cohorts of samples could reduce the potential of misjudgment because of technical unreliability and sample heterogeneity.

Our results demonstrated these validated miRNAs can potentially serve as novel noninvasive biomarkers for early diagnosis of ADC or SCC. In this study, we found that 11 miRNAs were dysregulated in the plasma of patients with lung cancer in comparison with cancer-free control subjects, including hsa-miR-26a-5p, hsa-miR-126-5p, hsa-miR-139-5p, hsa-miR-152-3p, hsa-miR-200c-3p, miR-451a, hsa-miR-3135b, hsa-miR-151a-3p, hsa-miR-151a-5p, hsa-miR-151b, and hsa-miR-550a-3p. As for the discrimination between ADC/SCC and healthy control, our screening strategy yielded a highest AUC of 0.89 in ADC and 0.97 in SCC, respectively. Importantly, the 10 miRNAs (except for hsa-miR-451a) could discriminate stage I ADC/SCC from healthy individuals in validation set-1 as well as set-2, which generated exciting prospects for early diagnosis and prognosis. However, the accuracy of 5 universal potential biomarkers was barely satisfactory and it could be better to treat ADC and SCC separately in prognosis. According to published data, 6 of 11 circulating miRNAs have been reported being related with NSCLC and could be potential biomarkers, which were hsa-miR-126-5p (22), hsa-miR-139-5p (23), hsa-miR-152-3p (24), hsa-miR-200c-3p (25), hsa-miR-451a (26), and hsa-miR-550a-3p (27). However, no matter in plasma/serum, tissue, or cells, there are few reports analyzing the relationship between the expression levels of these four miRNAs (hsa-miR-151a-3p, hsa-miR-151a-5p, hsa-miR-151b, and hsa-miR-3135b) and the occurrence of NSCLC. Multiple studies reported that hsa-miR-26a functions in cells and tissues and contributes to lung cancer development (28–30), and yet, no study observed a high concordance between the circulating hsa-miR-26a and NSCLC.

It is well known that the miRNA expressions are specific for tissues, developmental stages, and various diseases (31, 32). In our study, the available data, herein, indicated that expression patterns of miRNAs in plasma were different from those in tissue (Figs. 4 and 6). Boeri and colleagues also reported that circulating miRNAs had the predictive role independent from tissue specimens, because plasma miRNAs in the earlier phases of disease may be different from those required for the maintenance and the progression of the tumor (21). Teng and colleagues reported that miRNAs were secreted actively through small vesicles called “exosomes” that prevent them from degradation by RNases (33), and tumor exosomal miRNA might be sorted on the basis of the miRNA biological function via a binding protein, and more specifically, miRNAs with tumor-suppressive effects involving cell growth suppression were encapsulated into exosomes that were subsequently released into circulation, and then miRNA levels in tumor tissue are reduced compared with that in the adjacent tissue (34). According to this explanation, we speculated that 10 miRNAs in our study, including hsa-miR-26a-5p, hsa-miR-126-5p, hsa-miR-139-5p, hsa-miR-152-3p, hsa-miR-200c-3p, hsa-miR-3135b, hsa-miR-151a-3p, hsa-miR-151a-5p, hsa-miR-151b, and hsa-miR-550a-3p could be suppressive genes in NSCLC. There have been published researches on suppressive effects of the first five miRNAs (hsa-miR-26a-5p, hsa-miR-126-5p, hsa-miR-139-5p, hsa-miR-152-3p, and hsa-miR-200c-3p) in NSCLC (35–39); Ho and colleagues also reported that miR-550a-3p inhibited breast cancer initiation, growth, and metastasis (40); and we need more evidences to shed light on the possibility of cancer suppressor for hsa-miR-3135b, hsa-miR-151a-3p, hsa-miR-151a-5p, and hsa-miR-151b. As for some exception like hsa-miR-200c-3p, which was upregulated in four tissue samples and downregulated in the rest of samples (Fig. 6), the possible reason could be that the downregualtion of miRNA in tumor tissue was stage-dependent, and this needs further testing using clinical samples with more details.

Currently, this study provided that the unique patterns of circulating miRNAs might serve as noninvasive biomarkers for early diagnosis of NSCLC; however, we were not able to investigate the pathogenesis of NSCLC when the expression of miRNAs was changed. On the other hand, as a novel tool for diagnosis and prognosis, the clinical utility remains to be improved. A more standardized execution of miRNA measurements needs to be established and simpler and more robust protocols will allow the potential circulating biomarkers to enter the clinical application very soon.

No potential conflicts of interest were disclosed.

Conception and design: Y. Niu, M. Su, Y. Wu, F. Li, D. Gou

Development of methodology: Y. Niu, M. Su, Y. Wu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Niu, M. Su, Y. Wu, L. Fu, K. Kang, G. Hui, D. Gou

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Niu, M. Su, Y. Wu, F. Li

Writing, review, and/or revision of the manuscript: Y. Niu, M. Su, L. Li, F. Li, D. Gou

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Niu, M. Su, Q. Li

Study supervision: Y. Niu, M. Su, L. Li, D. Gou

This work was supported by National Natural Science Foundation of China (grant nos. 81600039, 91739109, 81700054, 81570046, and 31571199), Shenzhen Municipal Basic Research Program (JCYJ20150729104027220, JCYJ20170818144127727, and CXZZ20130515092016300), Shenzhen University start-up funding (2018043), and Shenzhen University Interdisciplinary Innovation Team Project (000003).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bianchi
F
,
Nicassio
F
,
Marzi
M
,
Belloni
E
,
Dall'Olio
V
,
Bernard
L
, et al
A serum circulating miRNA diagnostic test to identify asymptomatic high-risk individuals with early stage lung cancer
.
EMBO Mol Med
2011
;
3
:
495
503
.
2.
Rabinowits
G
,
Gerçel-Taylor
C
,
Day
JM
,
Taylor
DD
,
Kloecker
GH
. 
Exosomal microRNA: a diagnostic marker for lung cancer
.
Clin Lung Cancer
2009
;
10
:
42
6
.
3.
Mattia
B
,
Carla
V
,
Davide
C
,
Luca
R
,
Piergiorgio
M
,
Federica
F
, et al
MicroRNA signatures in tissues and plasma predict development and prognosis of computed tomography detected lung cancer
.
Proc Natl Acad Sci USA
2011
;
108
:
3713
3718
.
4.
He
L
,
Hannon
GJ
. 
MicroRNAs: small RNAs with a big role in gene regulation
.
Nat Rev Genet
2004
;
5
:
522
31
.
5.
Bushati
N
,
Cohen
SM
. 
MicroRNA functions
.
Annu Rev Cell Dev Biol
2007
;
23
:
175
205
.
6.
Chen
J-F
,
Mandel
EM
,
Thomson
JM
,
Wu
Q
,
Callis
TE
,
Hammond
SM
, et al
The role of microRNA-1 and microrna-133 in skeletal muscle proliferation and differentiation
.
Nat Genet
2005
;
38
:
228
33
.
7.
Croce
CM
,
Calin
GA
. 
miRNAs, cancer, and stem cell division
.
Cell
2005
;
122
:
6
7
.
8.
Pritchard
CC
,
Cheng
HH
,
Tewari
M
. 
MicroRNA profiling: approaches and considerations
.
Nat Rev Genet
2012
;
13
:
358
69
.
9.
Lu
J
,
Getz
G
,
Miska
EA
,
Alvarez-Saavedra
E
,
Lamb
J
,
Peck
D
, et al
MicroRNA expression profiles classify human cancers
.
Nature
2005
;
435
:
834
8
.
10.
Mitchell
PS
,
Parkin
RK
,
Kroh
EM
,
Fritz
BR
,
Wyman
SK
,
Pogosova-Agadjanyan
EL
, et al
Circulating microRNAs as stable blood-based markers for cancer detection
.
Proc Natl Acad Sci U S A
2008
;
105
:
10513
8
.
11.
Gupta
SK
,
Bang
C
,
Thum
T
. 
Circulating microRNAs as biomarkers and potential paracrine mediators of cardiovascular disease
.
Circulation: Cardiovasc Gene
2010
;
3
:
484
8
.
12.
Roth
C
,
Rack
B
,
Müller
V
,
Janni
W
,
Pantel
K
,
Schwarzenbach
H
. 
Circulating microRNAs as blood-based markers for patients with primary and metastatic breast cancer
.
Breast Cancer Res
2010
;
12
:
R90
.
13.
Toiyama
Y
,
Hur
K
,
Tanaka
K
,
Inoue
Y
,
Kusunoki
M
,
Boland
CR
, et al
Serum mir-200c is a novel prognostic and metastasis-predictive biomarker in patients with colorectal cancer
.
Ann Surg
2014
;
259
:
735
43
.
14.
Nelson
PT
,
Baldwin
DA
,
Scearce
LM
,
Oberholtzer
JC
,
Tobias
JW
,
Mourelatos
Z
. 
Microarray-based, high-throughput gene expression profiling of microRNAs
.
Nat Methods
2004
;
1
:
155
61
.
15.
Chen
C
,
Ridzon
DA
,
Broomer
AJ
,
Zhou
Z
,
Lee
DH
,
Nguyen
JT
, et al
Real-time quantification of microRNAs by stem-loop RT-PCR
.
Nucleic Acids Res
2005
;
33
:
e179
.
16.
Niu
Y
,
Zhang
L
,
Qiu
H
,
Wu
Y
,
Wang
Z
,
Zai
Y
, et al
An improved method for detecting circulating microRNAs with S-Poly (T) plus real-time PCR
.
Sci Rep
2015
;
5
:
15100
.
17.
Kang
K
,
Zhang
X
,
Liu
H
,
Wang
Z
,
Zhong
J
,
Huang
Z
, et al
A novel real-time PCR assay of microRNAs using S-Poly(T), a specific oligo (dT) reverse transcription primer with excellent sensitivity and specificity
.
PLoS One
2012
;
7
:
e 0048536
.
18.
Gabrielson
E
. 
Worldwide trends in lung cancer pathology
.
Respirology
2006
;
11
:
533
8
.
19.
Sun
S
,
Schiller
JH
,
Gazdar
AF
. 
Lung cancer in never smokers-a different disease
.
Nat Rev Cancer
2007
;
7
:
778
90
.
20.
Yuan
Z
,
Ghosh
D
. 
Combining multiple biomarker models in logistic regression
.
Biometrics
2008
;
64
:
431
9
.
21.
Boeri
M
,
Verri
C
,
Conte
D
,
Roz
L
,
Modena
P
,
Facchinetti
F
, et al
MicroRNA signatures in tissues and plasma predict development and prognosis of computed tomography detected lung cancer
.
Proc Natl Acad Sci USA
2011
;
108
:
3713
8
.
22.
Shen
J
,
Todd
NW
,
Zhang
H
,
Yu
L
,
Lingxiao
X
,
Mei
Y
, et al
Plasma microRNAs as potential biomarkers for non-small-cell lung cancer
.
Lab Invest
2011
;
91
:
579
87
.
23.
Cazzoli
R
,
Buttitta
F
,
Di Nicola
M
,
Malatesta
S
,
Marchetti
A
,
Rom
WN
, et al
MicroRNAs derived from circulating exosomes as noninvasive biomarkers for screening and diagnosing lung cancer
.
J Thorac Oncol
2013
;
8
:
1156
62
.
24.
Chen
X
,
Hu
Z
,
Wang
W
,
Ba
Y
,
Ma
L
,
Zhang
C
, et al
Identification of ten serum microRNAs from a genome-wide serum microRNA expression profile as novel noninvasive biomarkers for nonsmall cell lung cancer diagnosis
.
Int J Cancer
2012
;
130
:
1620
8
.
25.
Liu
X-G
,
Zhu
W-Y
,
Huang
Y-Y
,
Ma
L-N
,
Zhou
S-Q
,
Wang
Y-K
, et al
High expression of serum mir-21 and tumor mir-200c associated with poor prognosis in patients with lung cancer
.
Med Oncol
2012
;
29
:
618
26
.
26.
Pu
Q
,
Huang
Y
,
Lu
Y
,
Peng
Y
,
Zhang
J
,
Feng
G
, et al
Tissue-specific and plasma microRNA profiles could be promising biomarkers of histological classification and TNM stage in non-small cell lung cancer
.
Thorac Cancer
2016
;
7
:
348
54
.
27.
Rani
S
,
Gately
K
,
Crown
J
,
O'Byrne
K
,
O'Driscoll
L
. 
Global analysis of serum microRNAs as potential biomarkers for lung adenocarcinoma
.
Cancer Biol Ther
2013
;
14
:
1104
12
.
28.
Liu
B
,
Wu
X
,
Liu
B
,
Wang
C
,
Liu
Y
,
Zhou
Q
, et al
Mir-26a enhances metastasis potential of lung cancer cells via AKT pathway by targeting PTEN
.
Biochim Biophys Acta
2012
;
1822
:
1692
704
.
29.
Dang
X
,
Ma
A
,
Yang
L
,
Hu
H
,
Zhu
B
,
Shang
D
, et al
MicroRNA-26a regulates tumorigenic properties of EZH2 in human lung carcinoma cells
.
Cancer Genet
2012
;
205
:
113
23
.
30.
Gao
W
,
Shen
H
,
Liu
L
,
Xu
J
,
Xu
J
,
Shu
Y
. 
Mir-21 overexpression in human primary squamous cell lung carcinoma is associated with poor patient prognosis
.
J Cancer Res Clin
2011
;
137
:
557
66
.
31.
Keller
A
,
Meese
E
. 
Can circulating miRNAs live up to the promise of being minimal invasive biomarkers in clinical settings?
Wiley Interdisciplinary Rev
2016
;
7
:
148
56
.
32.
Esteller
M
. 
Non-coding RNAs in human disease
.
Nat Rev Genet
2011
;
12
:
861
74
.
33.
Pegtel
DM
,
Cosmopoulos
K
,
Thorley-Lawson
DA
,
van Eijndhoven
MA
,
Hopmans
ES
,
Lindenberg
JL
, et al
Functional delivery of viral miRNAs via exosomes
.
Proc Natl Acad Sci USA
2010
;
107
:
6328
33
.
34.
Teng
Y
,
Ren
Y
,
Hu
X
,
Mu
J
,
Samykutty
A
,
Zhuang
X
, et al
Mvp-mediated exosomal sorting of mir-193a promotes colon cancer progression
.
Nat Commun
2017
;
8
:
14448
.
35.
Lu
YY
,
Lin
Y
,
Ding
DX
,
Su
S
,
Chi
QQ
,
Zhang
YC
, et al
MiR-26a functions as a tumor suppressor in ambient particulate matter-bound metal-triggered lung cancer cell metastasis by targeting LIN28B-IL6-STAT3 axis
.
Arch Toxicol
2017
;
92
:
1023
35
.
36.
Song
L
,
Li
D
,
Gu
Y
,
Wen
ZM
,
Jie
J
,
Zhao
D
, et al
MicroRNA-126 targeting PIK3R2 inhibits NSCLC A549 cell proliferation, migration, and invasion by regulation of PTEN/PI3K/AKT pathway
.
Clin Lung Cancer
2016
;
17
:
e65
75
.
37.
Xu
W
,
Hang
M
,
Yuan
CY
,
Wu
FL
,
Chen
SB
,
Xue
K
. 
MicroRNA-139–5p inhibits cell proliferation and invasion by targeting insulin-like growth factor 1 receptor in human non-small cell lung cancer
.
Int J Clin Exp Pathol
2015
;
8
:
3864
70
.
38.
Cheng
Z
,
Ma
R
,
Tan
W
,
Zhang
L
. 
MiR-152 suppresses the proliferation and invasion of NSCLC cells by inhibiting FGF2
.
Exp Mol Med
2014
;
46
:
e112
.
39.
Liu
PL
,
Liu
WL
,
Chang
JM
,
Chen
YH
,
Liu
YP
,
Kuo
HF
, et al
MicroRNA-200c inhibits epithelial-mesenchymal transition, invasion, and migration of lung cancer by targeting HMGB1
.
PLoS One
2017
;
12
:
e0180844
.
40.
Ho
JY
,
Hsu
RJ
,
Wu
CH
,
Liao
GS
,
Gao
HW
,
Wang
TH
, et al
Reduced miR-550a-3p leads to breast cancer initiation, growth, and metastasis by increasing levels of ERK1 and 2
.
Oncotarget
2016
;
7
:
53853
68
.

Supplementary data