MicroRNA expression profiling of sputum for the detection of early and locally advanced non-small-cell lung cancer : a prospective case – control study

Background Non-small-cell lung cancer (nsclc) is associated with very poor overall survival because 70% of patients present with locally advanced or metastatic disease at the time of diagnosis. Micrornas (mirnas) are a class of short, noncoding rna molecules whose presence in samples of biologic fluids such as sputum has demonstrated promise as a potential means of detecting nsclc. We investigated the stage-specific nsclc detection potential of an efficient panel of 3 mirnas (mir-21, mir-210, mir-372) using a single sputum sample. Methods A single spontaneously expectorated sputum sample was prospectively collected from 21 early nsclc (≤stage ii) patients, 22 advanced nsclc (≥stage iii) patients, and 10 control subjects. Mirna expression profiles were determined by quantitative real-time polymerase chain reaction and were analyzed by unsupervised hierarchical cluster analysis. Results Mean tumour size (±95% confidence interval) in the early and advanced nsclc patients was 3.3 cm ± 0.9 cm and 4.8 cm ± 0.7 cm respectively. Adenocarcinoma constituted 61.9% of the early and 45.5% of the advanced nsclc cases respectively. In comparing the early nsclc group with the control group, the mirna panel yielded a diagnostic sensitivity of 67% and a specificity of 90.0%. For the advanced nsclc group, the mirna panel detected nsclc with a sensitivity and specificity of 64% and 100% respectively. Conclusions A sputum mir-21, mir-210, and mir-372 expression profile might provide a sensitive and highly specific means for detecting nsclc. Sputum mirna analysis demonstrates promise as a potential complementary screening tool.


INTRODUCTION
Non-small-cell lung cancer (nsclc) represents the 2nd most commonly diagnosed malignancy in Canada 1 .It is the leading cause of cancer-related mortality in both men and women, accounting for 27% of all predicted cancer-related deaths in 2015 1 .Because more than half of all patients present with metastatic disease at the time of diagnosis, the overall current 5-year survival is an abysmal 17.4% 2 .Diagnosis at an early stage has a significant effect on the long-term survival rate.The 10-year survival of individuals with stage i nsclc can approach 90% 3 .
The efficacy of chest radiography and sputum cytology for screening purposes has not been supported in the literature 4 .The U.S. National Lung Screening Trial demonstrated that a reduction in lung cancer mortality can be achieved through the use of low-dose computed tomography (ct) imaging in high-risk populations 5 .The long-term sequelae of repeated exposure to ct radiation and the morbidity related to the invasive investigations of e87 Current Oncology, Vol.23 incidentally discovered nodules leaves room for the development of a complementary screening modality.
Since their discovery in 1993, small noncoding endogenous single-stranded rnas (approximately 22 nucleotides), called "micrornas" (mirnas), have received significant attention 6 .Currently, posttranslational regulation of mirnas is implicated in multiple biologic processes such as proliferation, apoptosis, cell differentiation, oncogenesis, and metastasis [7][8][9][10][11][12][13] .The role of mirnas in lung cancer initiation, progression, and treatment response prediction is the subject of ongoing evaluation 14,15 .In the setting of lung cancer detection, sputum provides a potential source of mirnas because of their presence in cancer cells exfoliated into the airway and ultimately expectorated as sputum.However, the extracellular milieu also contains detectable mirnas because of a variety of transport mechanisms such as exosomes, microvesicles, rna binding protein complexes, or apoptotic bodies 16,17 .Even in extremely small quantities, mirnas can be detected by sensitive techniques such as real-time quantitative polymerase chain reaction (rtqpcr).Furthermore, mirnas display remarkable stability, even under adverse experimental conditions [18][19][20] .Those features favourably support the future utilization of mirnas as a practical screening biomarker.
Our previous work used single sputum samples from 24 prospectively accrued patients of all nsclc stages and from 6 healthy control subjects to evaluate a 5-mirna panel (mir-21, mir-143, mir-155, mir-210, mir-372) for the detection of nsclc, demonstrating 83.3% sensitivity and 100% specificity 21 .A post-hoc analysis revealed that 3 mirnas (mir-21, mir-210, mir-372) were primarily responsible for the diagnostic power of the panel.Individually, those 3 mirnas have been used in nsclc blood and sputum evaluations because of their overexpression, as documented in our own experience and in the literature 21,22 .This specific 3-mirna panel has not previously been reported in the literature.The first aim of the present study was therefore to determine whether this small and cost-effective 3-mirna panel could accurately diagnose the presence or absence of nsclc.Second, we investigated whether nsclc stage influenced the ability of the mirna panel to detect the presence of malignancy.To those ends, we prospectively recruited two cohorts: n An early-stage group consisting of individuals with stages i and ii nsclc n An advanced-stage group consisting of patients with stages iii and iv nsclc Both groups had been staged according to the 7th edition of the American Joint Committee on Cancer staging system 23 .To our knowledge, the screening potential of an mirna panel has not previously been investigated in a prospective stage-dependent manner.

Patient Selection and Data Collection
Adults diagnosed with nsclc of any histologic type, which had been confirmed either by bronchial wash cytology or by transbronchial or percutaneous fine-needle aspiration, were recruited.Participants were categorized as having either advanced or early nsclc.The early nsclc cohort comprised individuals staged either i or ii.Reported staging was based on the final postoperative pathology review.The advanced nsclc cohort comprised individuals staged either iii or iv.Advanced nsclc was staged using standard clinical practices including imaging modalities such as ct and positron-emission tomography, with nodal evaluation conducted either by endobronchial ultrasound-guided biopsy or cervical mediastinoscopy.Tumour size was based on the reported maximum tumour size.
Participants 18-75 years of age, with a minimum life expectancy of 3 months or more, who were medically stable, with an Eastern Cooperative Oncology Group score of 2 or less, were eligible 24 .All nsclc-negative controls had undergone either radiography or ct imaging of the chest within 12 months of sputum collection, confirming the absence of intrathoracic pathology.The presence of pulmonary medical disorders such as asthma or chronic obstructive pulmonary disease were not an indication for exclusion, given their prevalence in the nsclc population.Individuals with a history of malignancy except for nonmelanoma skin cancer were excluded from study recruitment.
All study participants completed a detailed questionnaire compiling information about their past medical history, smoking history, potential lifetime occupational hazard exposure, functional status, and demographic information.Imaging, pathology reports, and past medical and surgical histories were also retrieved for each study participant from our health region's electronic medical records system.Control subjects and early-stage patients were prospectively recruited from a preoperative clinical evaluation by a thoracic surgeon; advanced-stage patients were recruited from our local medical oncology treatment facility (Cross Cancer Institute, Edmonton, AB) before initiation of any treatment.
The study was approved by the Human Research Ethics Board of the University of Alberta (Edmonton, AB) and the Alberta Cancer Research Ethics Committee (Alberta Health Services, Edmonton, AB).All study participants provided informed consent.

Collection and Processing of Sputum
Sputum samples were collected from a single spontaneous expectoration obtained before any nsclc-directed treatment (surgery, chemotherapy, radiation).Participants were instructed to rinse their mouth with water after waking in the morning, to inspire, to hold that breath, and subsequently, to produce a deep cough.The sputum was directly collected into sterile containers and stored at 4°C until further processing was performed within 24 hours of collection.Code identifiers were applied to each sample container, blinding the individual performing the mirna analysis, and thus preventing subsequent statistical bias.
Based on our previous experience, we did not evaluate sputum cytology as a surrogate marker for sputum adequacy 21 .Sputum cytology is associated with a high rate of false positives for the detection of nsclc 25 .Furthermore, cytology does not evaluate the extracellular mirna component present in sputum 16,17 .Xie et al. 16 demonstrated a significant difference in sensitivity when they compared sputum cytology with mirna expression.Total rna isolation, rna reverse transcription, and rt-qpcr, which were conducted according to the manufacturer's recommendations, are described in our previous study 21 .The procedure is briefly summarized in the subsections that follow.
The TaqMan MicroRNA Reverse Transcription Kit for individual mirnas (Applied Biosystems, Foster City, CA, U.S.A.) was used for reverse transcription.Reverse transcription was performed using a StepOnePlus rt-pcr instrument (Applied Biosystems).The reverse transcription products were stored at -80°C until subsequent rtqpcr analysis.

RT-qPCR
The assays for individual mirnas were performed in duplicate using the reverse transcription reaction product already described.The cycle threshold (cyth) was defined as the number of cycles required for the fluorescent signal to cross the threshold in pcr.The SDS software application (Applied Biosystems) was used to identify cyth values.Expression values of mir-21, mir-210, and mir-372 were normalized to the commonly used endogenous control U6 and to a reference sample derived from normal lung fibroblast cells (MRC-5) 26,27 .The comparative method (ΔΔCτ method) was used to quantify the rt-qpcr mirna expression data.
The relative mirna expression of a given sample was determined using where Rn represents the required quantity of mirna to be tested, and ΔΔCτ was calculated as with cyth m being the cyth for the measured mirna and cyth ec being the cyth for the endogenous control mirna (U6) for the sputum samples to be tested (sample) and for the reference sample [reference (from the MRC-5 cell line)].

Statistical Analysis
The normalized sputum expression profiles for each individual (the 3-mirna panel) were evaluated using hierarchical cluster analysis as validated in our previous work 21 .Cluster analysis is a statistical method that groups similar (or related) objects within a group ("cluster"), thereby placing dissimilar objects in other groups 28 ; it is useful for analyzing complex biologic data.The "nearest neighbour cluster" method and the squared Euclidean distance were used to measure the intervals.Two separate cluster analyses were used to compare the control samples with the samples from the early-and advanced-stage nsclc samples.Hierarchical cluster analyses were performed using the SPSS software application (version 13: SPSS, Chicago, IL, U.S.A.).
Study population data are expressed as means or medians (± 95% confidence intervals) for continuous variables and as frequencies with proportions for categorical data.Cohort population comparisons were made using the Student t-test when the data were normally distributed and the Mann-Whitney U-test when the data were non-normally distributed continuous variables.

RESULTS
The study enrolled 10 control subjects, 21 patients with early nsclc, and 22 patients with advanced nsclc.Table i presents cohort-specific characteristics.Table ii shows characteristics, lung disease status, and smoking history for the individual control participants.Tables iii and iv show the patient and tumour characteristics for the individual patients with early and advanced nsclc respectively.
Most of the patients with early and advanced nsclc (≥86%) had a current or previous history of smoking; 50% of the control cohort had a smoking history.Mean pack-years for the individuals with smoking histories in the control, early nsclc, and advanced nsclc groups were 41.7 ± 17.0, 46.1 ± 13.6, and 33.4 ± 10.2 respectively.
For the patients with early and advanced nsclc, mean tumour size in the largest measured diameter was 3.3 ± 0.9 cm and 4.8 ± 0.7 cm respectively.Adenocarcinoma and squamous cell carcinoma constituted 62.0% and 23.8% of the tumours in the early nsclc group, and 45.5% and 22.7% of the tumours in the advanced nsclc group.No study participant had a prior history of malignancy.Every participant in the early-stage nsclc group underwent curative-intent resection.
Figures 1 and 2 present dendrograms comparing the control subjects with, respectively, the early and the advanced nsclc patients, based on their individual normalized mirna expression results (Table v).In the dendrograms, a horizontal line segregates the study participants into 2 large clusters, differentiating those with and without nsclc.Individuals clustered closer to 0 demonstrate a greater degree of relatedness according to their 3-mirna expression profiles.
In the early nsclc group, the mir-21, mir-210, and mir-372 expression profile yielded a diagnostic sensitivity of 67% and a specificity of 90%.False negatives were generated for participants E07, E13, E14, E17, E19, E20, and E21.Only 1 participant with squamous cell carcinoma generated a false-negative result; of the remaining 6 participants, 5 had adenocarcinoma, and 1 had large-cell nsclc.Just 1 control subject (C02) generated a false-positive result.The positive and negative predictive values of the panel in the early nsclc group were 93% and 56% respectively.Comparing the control subjects with the patients who had advanced nsclc, a similar sensitivity and specificity (64% and 100%) was observed.False negatives were generated for participants A02, A03, A04, A11, A12, A14, A15, and A18.As in the comparison involving the early-stage group, only 1 participant with squamous cell carcinoma generated a false-negative result; of the remaining 7 participants, 4 had nsclc not otherwise specified, and 3 had adenocarcinoma.No false-positive results were observed, leading to positive and negative predictive values of 100% and 56% respectively.

DISCUSSION AND CONCLUSIONS
The burden that lung cancer places on society leads to a great demand for the development and validation of biomarkers that can easily be measured in accessible biologic fluids to accurately detect nsclc.Sputum is one such source that has been investigated to determine if it contains elements that reflect the presence or absence of malignancy.Cytology alone-or even in combination with chest radiography-is not recommended for screening purposes 4 .Aberrant over-or underexpression of mirna in lung cancer tissue has clearly been shown 15 .The mirnas play an important role in controlling cell growth, the cell cycle, and apoptosis.An understanding of specific mirnas and their roles in various malignancies has grown tremendously in recent years.The promising results from Based on the 7th edition of the American Joint Committee on Cancer staging system 23 .NA = not applicable; NSCLC = non-small-cell lung cancer.present study, which used a small 3-mirna panel to detect nsclc, have, interestingly, also been implicated in a variety of solid malignancies.Overexpression of mir-21 has been associated with malignancies such as glioma and colorectal and prostate cancers [29][30][31] .A recent meta-analysis that evaluated 598 specimens of lung cancer and 528 normal lung tissue controls found that mir-21 and mir-210 upregulation were statistically significant in the tumour tissue group 32 , and mir-21 is known to target multiple negative regulators of the Ras/mek/erk pathway, thereby promoting proliferation 33 .Furthermore, mir-21 inhibition of pro-apoptotic gene products further contributes to tumourigenesis and tumour progression 32 .
Overexpression of mir-210 has also been demonstrated in lung, esophageal, and pancreatic cancer 34,35 , and this mirna is often referred to as the "master" hypoxia mirna 36 .The response to hypoxia has been well demonstrated to result in upregulation of mir-210 37,38 .Using the nsclc cell line A549, Grosso et al. 38 demonstrated that mir-210 confers radioresistance through hypoxia inducible factor 1.
The oncogenic role of mir-372 has been demonstrated by its downregulation of the tumour suppressor gene large tumour suppressor homolog 2 39 .Using the lung adenocarcinoma cell line CL 1-0, Lai et al. 40 demonstrated that the overexpression of mir-372 corresponded with increased invasiveness.
We initially hypothesized that advanced nsclc would result in an increase of mirna within the sputum milieu, correlating with an enhanced rate of sensitivity in the advanced nsclc group.Our results demonstrate that nsclc stage does not appear to play a significant role in the results of our diagnostic test, such that reasonable sensitivity and very high specificity were achieved in patients with an early-stage nsclc diagnosis.
Our future direction with this test will focus on the interest in low-dose ct screening for high-risk individuals.The U.S. National Lung Screening Trial highlights a tremendous challenge facing widespread ct screening.The authors observed that 4.2% of participants required at least 1 diagnostic surgical procedure 3 .The cost to the health care system of subsequent investigations and their associated morbidity leaves room for the development of complementary diagnostic tests for future integration within low-dose ct screening programs.We have demonstrated that sensitivity is enhanced with mirna analysis of bronchoalveolar lavage compared with sputum (86%) 41 .The feasibility of routine bronchoscopy inhibits its usefulness in a screening setting 41 .Noninvasive biomarker-based techniques such as mirna analysis of sputum could further aid in cancer-risk stratification for pulmonary nodules, thereby minimizing unnecessary invasive testing.Further prospective evaluation in individuals with pulmonary nodules will have to be undertaken if sputum mirna analysis is to be validated for clinical application.
Interestingly, mir-372 expression demonstrated the greatest variability in our advanced-stage cohort.The exact reason for that variability is unclear, because mir-372 overexpression is correlated with lung cancer invasiveness.Squamous cell carcinoma constituted only 1 of the observed false negatives in each group.That observation could be related to the greater sensitivity for the detection of squamous cell carcinoma also seen in sputum cytology, because squamous cell carcinomas are frequently more centrally located 42 .
Similar results with a small, focused mirna panel have been demonstrated in the literature.Xing et al. 43 used a 3-mirna panel (mir-21, mir-31, mir-210) to evaluate sputum, yielding a sensitivity of 80.52% and a specificity of 86.08%.The mounting body of literature highlights the exciting potential of a noninvasive tool of this kind in nsclc diagnosis.Because of sample size limitations and variability in expression values, we were unable to demonstrate meaningful associations between individual mirnas and tumour subtype or patient factors.Large population-based studies will be required to answer such questions.
To summarize, hierarchical cluster analysis of sputum mirna expression yields a highly specific test for the detection of early and advanced nsclc.Future studies using internal and external validation to reconfirm the results of the present study are warranted.We propose that a test such as the one described here could have future clinical applicability as a complementary diagnostic investigation in low-dose ct screening programs.Based on the 7th edition of the American Joint Committee on Cancer staging system 23 . b Poorly differentiated tumour reported as NSCLC not otherwise specified.
Could not be accurately measured on computed tomography imaging because of lobar consolidation.NA = not available; COPD = chronic obstructive pulmonary disease.

TABLE I
Characteristics of study participants overall

TABLE II Characteristics of control participants Specimen ID Sex Age Lung disease Smoking history
e90 Current Oncology, Vol. 23, No. 2, April 2016 © 2016 Multimed Inc.

TABLE IV
Characteristics of participants with advanced-stage non-small-cell lung cancer (NSCLC)