MicroRNA Profile of Tumorigenic Cells During Carcinogenesis of Lung Adenocarcinoma

To obtain microRNA (miRNA) profile and clarify their biological function in tumorigenic Sca‐1+CD34+ cells during carcinogenesis of lung adenocarcinoma. After intranasal infection with recombinant Adeno‐Cre viruses (AdV‐Cre), lung adenocarcinoma was identified pathologically in Lox‐stop‐lox Kras (LSL‐Kras) G12D mice. Sca‐1+CD34+ cells were sorted by flow cytometry and tested for tumor‐initiating ability, self‐renewal and tumorigenicity. MiRNA profiles were obtained using microarray and further confirmed by real‐time RT‐PCR (qRT‐PCR). MiRNA functions were predicted bioinformatically, and miR‐294 function was verified to explore its role in tumor migration and invasion. Lung adenocarcinoma was induced in LSL‐Kras G12D mice within 30 days. In vivo, the tumorigenicity of Sca‐1+CD34+ cells was 25 times stronger than Sca‐1−CD34− cells. During tumorigenesis of lung adenocarcinoma, the expression of 145 miRNAs in Sca‐1+CD34+ cells increased and 72 miRNAs decreased (P < 0.01). Four successively up‐regulated miRNAs (miR‐15a*, miR‐203, miR‐294 and miR‐295*) and three successively down‐regulated ones (miR‐19b, miR‐483 and miR‐615–5p) were identified. Among them, miR‐294 could constitutively bind to 3'‐UTR of matrix metalloproteinase 3 (MMP3), and down‐regulate MMP3 protein expression. MiR‐294 also significantly inhibited migration and invasion of Lewis lung cancer cells. MiRNAs are characteristically expressed in tumor‐initiating Sca‐1+CD34+ cells of lung adenocarcinoma, and may play important roles during the carcinogenesis of lung adenocarcinoma. J. Cell. Biochem. 116: 458–466, 2015. © 2014 The Authors. Journal of Cellular Biochemistry published by Wiley Periodicals, Inc.

L ung cancer is the leading cause of cancer-related deaths worldwide. About 1,000,000 people die of lung cancer every year, including small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC) [Parkin et al., 2005]. The overall five-year survival rate for the patients with lung cancer is still below 15% [Jemal et al., 2006].
Tumorigenic cells, also designated as cancer-initiating cells or cancer stem cells (CSCs), have been proposed to be involved in cancer initiation, progression, metastasis and therapy resistance [Mimeault et al., 2007a]. Similar to normal stem cells, CSCs have unlimited capability of proliferation and self-renewal, and can generate a progeny of differentiated cells. These two kinds of cells have similar cell-surface markers, and are regulated by similar signaling pathways, including Notch, Wnt, and Hh [Tsai, 2004;Mimeault et al., 2007b]. Cumulative oncogenic mutations during carcino-genesis may turn normal stem cells into cancer-initiating cells characterized by excessive, uncontrollable proliferation.
In recent years, the role of non-coding RNAs in lung cancer attracts much attention. MicroRNAs (miRNAs) are a class of small (19-25 nt) non-coding RNAs that regulate the expression of protein transcriptionally or post-transcriptionally [Hsu et al., 2006;Shcherbata et al., 2006]. MiRNAs exist in a wide variety of organisms, and play important roles in cell growth, apoptosis, stem cell differentiation and tumorigenesis [Cheng et al., 2005;He et al., 2005;Garzon et al., 2006]. Aberrant expression of miRNAs has been reported in many tumor types, such as lung cancer, breast cancer, and glioblastoma, indicating that miRNAs play an important part in initiation and growth of malignant tumors [Calin et al., 2004;Hayashita et al., 2005;Volinia et al., 2006]. Recent studies have revealed that miRNAs regulate differentiation of CSCs and malignant transformation of normal tissue stem cells into CSCs [Ibarra et al., 2007;Yu et al., 2007].
Since first being isolated and expanded from leukemia, CSCs have been reported in several human solid tumors [Ricci-Vitiani et al., 2007]. The role of bronchioalveolar stem cells (BASCs) is striking in all the studies on lung cancer stem cells [Evans et al., 1978;Fehrenbach, 2001;Kim et al., 2005]. The regenerative potential of stem cells residing in the bronchoalveolar junction of adult lungs have been further identified and characterized. Sca-1 þ CD34 þ cells in Lox-stoplox Kras (LSL-Kras) G12D mouse model of lung carcinogenesis are capable of differentiating into type II alveolar epithelial cells and Clara cells. LSL-Kras G12D mouse is a type of transgenic animal, harboring a conditional Kras knock-in allele. Kras mutation is activated only after Cre-recombinase is delivered via recombinant adenoviral vectors [Meuwissen et al., 2001]. LSL-Kras G12D mice would develop lung adenocarcinomas after intratracheal administration of AdV-Cre. To date, the mechanism by which BASCs cumulatively mutate into lung adenocarcinoma-initiating cells remains unknown. In the current research, we planned to compare miRNA profiles between BASCs and lung adenocarcinoma stem cells, and verify miRNA function during the tumorigenesis of lung adenocarcinoma.

MATERIALS AND METHODS
CELL LINES AND ANIMALS HEK293 and Murine Lewis lung cancer cell lines (LLC) (ATCC, Manassas, VA) were used. Recombinant Adeno-Cre viruses (AdV-Cre) were commercially purchased (Microbix Biosystems, Canada). LSL-Kras G12D mice were obtained from the MMHCC (Mouse Models Human Cancer Consortium) as a gift. BALB/c nude mice (4-5 weeks old, No. SCXK-2007001) were provided by the Experimental Animal Center of Third Military Medical University, China. All the animal experiment procedures were carried out with the approval of the Animal Ethics Committee of Third Military Medical University. Conditioned DMEM-F12 medium consisting of ITS (1:100), 20 ng/ml epidermal growth factor (EGF), 10 ng/ml basic fibroblast growth factor (bFGF), 5 mM 4-(2-hydroxyethy1)-1-piperazineethanesulfonic acid (HEPES), and 0.5% bovine serum albumin (BSA) was prepared.

CONSTRUCTION OF A MOUSE MODEL OF LUNG ADENOCARCINOMA
LSL-Kras G12D mice were used to construct the lung adenocarcinoma model. HEK293 cells and AdV-Cre were co-cultured until typical pathogenic cell infection occurred (Cytopathic Effect, CPE). Infected HEK 293 cells were digested and repeatedly frozen at À80°C and thawed at 37°C, and oscillated three times. Next, the virus supernatant was collected, purified with an Adeno -X TM Virus Purification Kit (BD Biosciences, Clontech), aliquoted, and stored at À80°C in cryopreservation media. Then a conditional knockout mouse model of lung adenocarcinoma was constructed by triggering Kras gene via intranasal infection with AdV-Cre as previously described [Meuwissen et al., 2001]. Lung tissues were taken on day 15 or day 30 and observed under a reverse microscope. Serial sections of 8-10 mm in thickness were stained with FITC-CD34, PE-Sca-1 (1: 100), and DAPI (1: 1,000).

CELL DIFFERENTIATION IN VITRO
Sca-1 þ CD34 þ cells were cultured in a standard culture medium and digested with 0.25% trypsin-EDTA solution into single cells on days 2, 4, 8 and 16. After incubating PE-Sca-1 and FITC-CD34 for 30 min at 4°C, single cell suspension was fixed with 1% paraformaldehyde and analyzed by flow cytometry.

TUMORIGENICITY IN VIVO
Sca-1 þ CD34 þ cells were subcutaneously inoculated in the right armpit of nude mice with 500 K, 100 K, 50 K, 20 K, and 10 K (1K ¼ 1 Â 10 3 cells), respectively (n ¼ 6). Same experiments were repeated in Sca-1 -CD34cells as a control. Tumor size was calculated by V ¼ LW 2 /2, in which L and W mean the long and short diameters, respectively.

MICROARRAY FABRICATION AND miRNA HYBRIDIZATION
On days 0, 15, and 30, lung tissues were taken from the LSL-Kras G12D mouse of intranasal infection with AdV-Cre (designated as groups A, B, and C above). Extraction of total RNAs and miRNA microarray hybridization were conducted as previously described [Liao et al., 2008]. Microarray (Kang Cheng Biological Chip Co. Ltd, China) was made by mParaflo TM microfluidic chip technology, and miRNA probe sequences were obtained from the database of Sanger miRBase version 12.0 (http://microrna.sanger.ac.uk/sequences/). The intensity of the green signal was calculated after background subtraction and replicated spots on the same chip had been averaged by median intensity. We used Median Normalization Method to obtain "Normalized Data" (Normalized Data ¼ [Foreground-Background]/median). The median was 50 percent quantile of miRNA intensity, which was larger than 50 in all samples after background subtraction. Normalized data were statistically analyzed by T test (P < 0.05 was considered statistically significant). Each probe of mmu-miRs (miR) was repeated four times.

TARGET PREDICTION
Three online software programs, TargetScan, MiRbase, and miRanda, were used to predict miRNA targets [Liao et al., 2008;Chen et al., 2009]. Comprehensive consideration of functional conservation in different organism species, binding free energy, and scores of base matching, miR-294 and its potential target matrix metalloproteinase 3 (MMP3) were focused in the following work.

WESTERN BLOT
Procedures of Western blot were described in detail in our previous work [Ma et al., 2011]. Simply, miR-294 mimics and miR negative control (RiboBio Co. Ltd., Guangzhou, China) were transfected into LLC by lipofectamine 2000, respectively, to reach a final concentration of 30 nM. After 48 h, MMP-3 mRNA and protein were detected by qRT-PCR and Western blot, respectively. MMP3 primary antibody (Abcam, USA) and secondary antibody labeled with horseradish peroxidase (HRP) (Zhongshan Biotechnology Company, Beijing, China) were used in Western blot.

MATRIGEL INVASION AND TRANSWELL MIGRATION ASSAYS
MiR-294 mimics and miR negative control were transfected into LLC as described above. Matrigel invasion and transwell migration assays were performed according to a earlier report [Batts et al., 2011]. Each experiment was repeated three times.

CONSTRUCTION OF MOUSE MODEL OF LUNG ADENOCARCINOMA
To confirm successful Lox-stop-lox modification of experimental mice, we designed primers to amplify LSL-Kras fragment from genomic DNA by PCR. The primer sequences were as follows. 5 0cctttacaagcgcacgcagactgtaga-3 0 , 5 0 -agctagccaccatggcttgagtaagtctgca-3 0 . A 550-bp PCR product was obtained, preliminarily proving that the mouse model was LSL-Kras G12D (Fig. 1a). Immunofluorescence showed that green fluorescent FITC-CD34 was distributed at the joint of the alveolar isolation (Fig. 1b), and red fluorescent PE-Sca-1 at terminal and respiratory bronchioles, localized in the cytoplasm around the nucleus (Fig. 1c). DAPI labeled cell nuclei were prominently blue (Fig. 1d), and a few Sca-1 þ CD34 þ cells with yellow fluorescence in large and round shape were distributed at BADJ between alveolar duct and respiratory bronchioles (Fig. 1e).
In Group A, the lung tissues were pink and smooth, without any abnormal change such as nodules or hyperemia (Fig. 1f), and no dysplasia appeared in pulmonary bronchial epithelium in HE staining (Fig. 1h). In Group B, atypical hyperplasia of pulmonary bronchial epithelium appeared in bronchioles and the mucosa became thicker (Fig. 1i). In Group C, dim gray nodules appeared on the surface (Fig. 1g), and pulmonary bronchial epithelium showed diffused adenocarcinoma nests (Fig. 1j).

ISOLATION AND CULTURE OF
The expression of Sca-1 and CD34 in lung tissues was analyzed by flow cytometry. We observed 0.7-1.1% Sca-1 þ CD34 þ cells and 1.7-2.0% Sca-1 -CD34cells, and sorted the cells using flow cytometry. This subpopulation was highly purified (71.5-80%). At the original position where the positive cells existed, only a few cells left.
FITC-CD34 and PE-Sca-1 immunocytochemical staining was done in spheres on day 6. In the merged picture, double-stained cells with both FITC and PE fluorescence were distributed in the cytoplasm around the nucleus, indicating that both CD34 and Sca-1 were highly expressed in spheres (Fig. 2c-f).
In colony formation assay, Sca-1 þ CD34 þ cells could produce significantly more colonies or spheres than Sca-1 -CD34cells in both standard culture medium and conditioned culture medium In Group A (day 0), lung tissues displayed smooth surface (f, h). In Group B (day 15), pulmonary bronchial epithelium atypical hyperplasia appeared in bronchioles and the mucosa became thicker (i). In Group C (day 30), dim gray nodules appeared on the lung surface (g), and atypical hyperplasia appeared on the bronchial epithelium covered by adenocarcinoma nests (j).
When being transferred into a standard culture medium with 10% FBS, Sca-1 þ CD34 þ cell spheres adhered to the culture plate in 12 h. The percentage of Sca-1 þ CD34 þ declined with the culture time. After 16 days, Sca-1 þ CD34 þ percentage decreased to a normal level of parental LLC cells, indicating that Sca-1 þ CD34 þ could differentiate into other molecular phenotypes of cells (Fig. 2h).
In xenograft experiment in nude mice, tumors could be observed in every group when 500 K cells were inoculated subcutaneously. Sca-1 þ CD34 þ cells gave rise to larger tumors than the control (P < 0.01, Table I). In gradient assay, Sca-1 þ CD34 þ cells could generate tumors when 10 K cells were inoculated, while Sca-1 -CD34cells could not grow into a tumor when 50 K cells were inoculated. To reach a similar tumorigenic efficacy, the quantity of required Sca-1 þ CD34 þ cells was 20 K while that of Sca-1 -CD34cells 500 K. Thus, the tumorigenicity of Sca-1 þ CD34 þ cells was 25 times stronger than Sca-1 -CD34cells (Fig. 2i, Table 1).
miRNA PROFILE OF Sca-1 þ CD34 þ CELLS MiRNA microarray fluorescent signals in groups A, B and C are shown in Fig. 3a, b, and c. MiRNAs were regarded as measurable transcription elements if they satisfied two conditions: signal intensity >3 Â background standard deviation and spot coefficient of variability (spot CV) <0.5. The clustering and correlation coefficient matrix was performed and mmu-miRNA profiles were obtained. As mentioned above, experimental LSL-Kras G12D mice were divided into 3 groups (groups A, B and C) according to days 0, 15 and 30 after intranasal infection with AdV-Cre. Although the lung tissues from groups A, B and C were of very similar origin, differences in miRNA profile were detectable. A total of 145 miRNAs increased successively in groups A, B, and C (P < 0.01); 72 miRNAs decreased successively (P < 0.01) (Supplemental Table. S1 & S2); four miRNAs (miR-15a*, miR-203, miR-294 and miR-295*) increased successively over 1.2 fold in groups A, B, and C; three miRNAs (miR-19b, miR-483 and miR-615-5p) decreased successively over 0.5-fold. Quantitative analysis of microarrays in group B/ A and group C/B showed significant difference (P < 0.05, Table II). All the data had been uploaded and submitted onto Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc ¼ GSE60001).
Special attention was paid to the successively increased or decreased miRNAs between groups A, B and C. To verify the accuracy of the miRNA array, seven miRNAs were chosen for further qRT-PCR assay. Amplification curve of miR-294 was demonstrated as a representative of qRT-PCR assay (Fig. 3d). The results of qRT-PCR assay were basically consistent with microarray, thus validating the authenticity of miRNA microarray (Fig. 3e-f). Quantitative analysis of qRT-PCR assay in group B/A and group C/B also showed significant difference (P < 0.05, Table 2).

TARGET VERIFICATION
According to bioinformatics analysis, most of the targets were oncogenes, anti-oncogenes or regulatory factors involved in metabolism, cell cycle, transcription, apoptosis, proliferation, and differentiation (Supplemental Table S3). For example, dozens of potential genes were targeted by miR-294, such as Wee1, Gpr34, MMP3, Cdca4, Hsp110, Atp10d, Egr3, and Map3k1. Several potential targets were chosen to verify (data not shown). Here MMP3 was shown as a target gene.
In dual luciferase reporter assay, 48 h after cotransfection of recombinant plasmids and miR-294 mimics into LLC cells, luciferase activities of firefly and Renilla were detected in all the four groups. Firefly/Renilla ratios were 6.43 AE 1.03, 13.75 AE 2.61, 13.59 AE 1.67, and 11.90 AE 1.49 in test group, control group-1, control group-2 and empty vector group, respectively. A significant decrease was observed in the test group instead of the other three groups (P < 0.05).
In qRT-PCR assay, miR-294 expression in LLC cells with miR-294 mimics transfection was 2.86 and 7.37 folds higher than that in miR negative control transfection and parental LLC cells, respectively. Then MMP3 mRNA was also detected by qRT-PCR. No significant changes were found in different transfection assays (Fig. 4a). However, in Western blot, miR-294 mimics could significantly decrease MMP3 protein expression compared with miR negative control transfection and parental LLC cells (P < 0.05) (Fig. 4b).

FUNCTIONAL ANALYSIS OF miR-294
In matrigel invasion assay, the invading tumor cells were 108.33, 165.00 and 168.67 on average in miR-294 mimics transfection, miR negative control transfection and parental LLC cells, respectively. The results showed that miR-294 mimics (Fig. 4c) significantly decreased invasion of LLC cells compared with miR negative control (Fig. 4d) and parental LLC cells (Fig. 4e, P < 0.05).
In transwell migration assay, similar results were observed. The migrating tumor cells were 109.00, 177.33 and 190.33 on average in miRNA mimics transfection, miR negative control transfection and parental LLC cells, respectively. MiR-294 mimics (Fig. 4f) significantly decreased migration of LLC cells compared with miR negative control (Fig. 4g) and parental LLC cells (Fig. 4h, P < 0.05).

DISCUSSION
It is well known that TICs are the source of non-stem cancer cells and the root of treatment resistance of cancer. A previous study has shown that stem-like cancer cells derived from normal cell transition displays tumorigenicity (Chambers and Smith, 2004). In the present study, we not only developed a mouse model of lung adenocarcinoma, but also clarified the molecular changes and biological mechanism underlying lung carcinogenesis. We found that Sca-1 þ CD34 þ cells showed strong colony-forming and proliferation capabilities in vitro and they could differentiate into Sca-1 -CD34cells in the serum. By detecting tumorigenicity, we found that Sca-1 þ CD34 þ cells displayed tumorigenic efficacy 25 times higher than Sca-1 -CD34cells. These findings were consistent with earlier research [Kim et al., 2005] and paved the way for further work. It is believed that miRNAs take part in the tumorigenesis of many solid cancers such as breast cancer [Volinia et al., 2006;Yu et al., 2007]. Several studies have also shown that aberrant expression of miRNAs is involved in generation of lung cancer, such as miR- Fig. 3. miRNA profile of Sca-1 þ CD34 þ cells and associated miRNAs in tumorigenesis. Fluorescent signals of miRNA microarray are shown of Sca-1 þ CD34 þ cells from mice with intranasal infection of AdV-Cre at day 0 (a, Group A), day 15 (b, Group B) and day 30 (c, group C). qRT-PCR amplification curve of mmu-miR-294 was exhibited as a representative (d). Seven miRNAs detected by qRT-PCR in different groups (e) and compared with miRNA chips (f). and let-7 [Hayashita et al., 2005;Volinia et al., 2006]. We previously identified fifty miRNAs in TICs from human A549 lung cancer cells [Lin et al., 2012]. Recently, it was reported that five miRNAs (let-7a, miR-221, miR-137, miR-372, and miR-182*) are closely related to the survival and tumor relapse in NSCLC patients [Yu et al., 2008]. Nevertheless, little is known about miRNA function in lung carcinogenesis. By isolating Sca-1 þ CD34 þ cells from lung tissues, we investigated the changes in miRNA profiles in carcinogenesis of lung adenocarcinoma. Our research indicates that several miRNAs, successively over-expressed or under-expressed in Sca-1 þ CD34 þ tumorigenic cells, probably play important roles in lung adenocarcinoma initiation. We further explored down-stream target genes of the miRNAs of interest. We predicted and verified that miR-294 down-regulated MMP3 target gene. Matrix metalloproteinases (MMPs) are a family of proteolytic enzymes that degrade the extracellular matrix and play important roles in tumor invasion and metastasis (Hewitt et al., 1991;Chambers and Matrisian, 1997). MMP3, a member of MMPs, has a function analogous to its homologue MMP9 that can promote cancer cell invasion, migration and metastasis in several tumors including lung cancer [Fang et al., 2005;Zhang et al., 2011]. Our study revealed that miR-294 constitutively binds to 3 0 -UTR of MMP3. Subsequent research showed that miR-294 mimics can decrease MMP3 protein expression, though no obvious change in mRNA level was observed. Matrigel invasion assay and transwell migration assay further confirmed the inhibition of the migration and invasion in LLC cells by miR-294 mimics transfection.
MiR-294 has been reported to promote the induction of pluripotency in embryonic stem cell (ESC) [Judson et al., 2009;Hanina et al., 2010]. Its expression increases during G1 phase of ESC cycle [Hanina et al., 2010;Wang et al., 2013]. Consistent with these findings, we observed that miR-294 is highly expressed in stem celllike Sca-1 þ CD34 þ cells, suggesting that miR-294 up-regulation correlates with stemness. Of interest, we found that miR-294 mimics could decrease MMP3 protein expression and suppress LLC cell invasion and migration. To our best knowledge, this is the first time, miR-294 function was tested in regulating MMP3 and cancer cell invasion and migration. miR-294 probably plays different roles in different tumor types and different pathological processes. Since MMP3 is critical for the maintenance of stemness for CSCs that exhibit potent capacities of invasion and metastasis [Jin et al., 2013], the up-regulation of miR-294 in CSCs and its regulation on MMP3 in LLC suggested a novel cell-type dependent feedback regulatory mechanism, by which miR-294 may negatively modulate the stemness. Our subsequent work will focus on the biological function of miR-294 in regulating self-renewal and metastasis in stem celllike cancer cells.
In this study, miR-294 has been used to demonstrate the biological function of miRNAs in Sca-1 þ CD34 þ tumorigenic cells. The findings are consistent with our hypothesis that miRNAs play important roles in malignant transformation of BASCs. We will further investigate biological functions of more miRNAs and their targets in the future.

CONCLUSIONS
Here, we investigated the miRNA profile of the Sca-1 þ CD34 þ population in an LSL-Kras G12D mouse model of lung adenocarcinoma. miR-294 could down-regulate MMP3 and significantly inhibit migration and invasion of LLC cells. Our identification of Sca-1 þ CD34 þ -related miRNAs can be a start to exploring the functions of these miRNAs. This study adds a new dimension to our understanding of the carcinogenesis of lung adenocarcinoma and assists oncologic biologists and clinical oncologists in designing novel therapeutic strategies.