A gender specific improved survival related to stromal miR-143 and miR-145 expression in non-small cell lung cancer

Micro RNAs (miRNA) are small non-coding RNAs that post-transcriptionally regulate gene expression. Dysregulation of miRNA cluster 143/145 has been reported in several malignancies, but their role in non-small cell lung cancer (NSCLC) remains elusive. This study investigates the prognostic impact of miR-143 and miR-145 in primary tumors and metastatic lymph nodes in NSCLC tissue. Tissue from 553 primary tumors and 143 matched metastatic lymph nodes were collected and tissue microarrays were constructed. In situ hybridization was used to evaluate miR-143 and miR-145 expression in tumor epithelial cells and stromal cells in the primary tumors and lymph nodes. In vivo data was supplemented with functional studies of cell lines in vitro to evaluate the role of miR-143 and miR-145 in NSCLC tumorigenesis. In our cohort, stromal miR-143 (S-miR-143) and miR-145 (S-miR-145) expression in primary tumor tissue were independent prognosticators of improved disease-specific survival (DSS) in female (S-miR-143, HR: 0.53, p = 0.019) and male patients (S-miR-145, HR: 0.58, p = 0.021), respectively. Interesting correlations between the miR cluster 143/145 and previously investigated steroid hormone receptors from the same cohort were identified, substantiating their gender dependent significance.

Tissue microarray construction. All specimens were embedded in paraffin blocks and examined by two experienced pathologists. Detailed methodology regarding TMA construction has previously been published 15 . Briefly, (1) representative areas of stromal and tumor tissue in primary tumors and tumor tissue from lymph nodes were identified and sampled with a 0.6 mm stylet, (2) transferred to the recipient TMA block and (3) cut into 4μm sections with a Micron microtome (HM355S) prior to in situ hybridization. Normal lung tissue far from the site of the tumor, and lung tissue samples from 20 emphysema patients without any history of neoplastic disease, were used as controls and for comparing biomarker expression level in malignant vs non-malignant tissue.
4 µm TMA sections were incubated overnight at 60 °C to attach tissue to Super Frost Plus slides. To ensure good distribution of reagents and protect sections from desiccation, LCS (Liquid Coverslip oil, Roche, 5264839001) was added. Deparaffinization was performed in EZ Prep buffer (Roche 5279755001) at 68 °C (3 × 12 min). Demasking was done at 95 °C with CC1 buffer (Roche, 6414575001) for 40 minutes. Subsequently, sections were rinsed with Reaction Buffer (Roche 5353955001) and RiboWash, SSPE buffer (Roche 5266262001).
All slides were denaturated for 8 min. at 90 °C. Hybridization with probes was performed for 60 min at 54 °C for miR-145, 55 °C for miR-143, 57 °C for scramble miR and 55 °C for U6. Stringent wash procedures were done at 2 × 8 min with 2.0X RiboWash, SSPE buffer with the same temperatures as used under hybridization for each probe. Blocking against unspecific bindings followed, with blocking solution (Roche, 5268869001) for 16 min. at 37 °C. Alkaline phosphatase (AP)-conjugated anti DIG (Anti-DIG-AP Multimer, Roche 07256302001) was incubated for 20 min. at 37 °C for immunologic detection. After rinsing, substrate enzymatic reactions were carried out with NBT/BCIP (CromoMap Blue kit, Roche 526661001) for 60 min at 37 °C, to give a blue precipitate to detect the microRNA. Sections were again rinsed and counterstained in 4 min with Red Stain II (Roche 5272017001). Increasing gradients of ethanol solutions was used for dehydration. Finally, all sections were mounted using the Histokitt mounting medium (Assistant-Histokitt, 1025/250 Sondheim/Rhoen Germany).

Scoring of ISH.
All tissue samples were independently and semi-quantitatively scored by an experienced pathologist (SAS) and a trained medical doctor (KS). Biomarkers were evaluated by intensity in neoplastic epithelial cells and stromal cells; 0 (no staining), 1 (weak), 2 (intermediate) and 3 (strong) and density in stromal cells; 0 = absent, 1 = 1-5%, 2 = 6-50%, 3 = >50%. Due to homogenous staining in neoplastic epithelial cells, scoring of biomarker density was not deemed necessary. For stromal biomarker expression (S-miR) the mean value of intensity and density combined, was calculated. Staining of fibroblasts, fibrocytes, lymphocytes, smooth muscle cells (SMC) and endothelial cells in blood and lymph vessels were included while scoring tumor stroma. Striking positivity was noted in endothelial cells lining the blood vessels and SMCs, including the smallest capillaries. Each variable was dichotomized for survival analyses based on a minimal p-value approach. A high score was defined as a score ≥ mean value for stromal-miR-143 (S-miR-143, mean value: 1.87) and tumor-miR-143 (T-miR-143, mean value: 1.98) and >0 for S-miR-145 and T-miR-145. The same scoring approach was used in PT, LN+, positive and negative tissue controls. For LN+ however, the stromal compartment was not scored due to large numbers of excessively stained lymphocytes. In normal lung tissue from NSCLC patients, collected far from the site of the tumor, miR-143 was prominently expressed in type 2 pneumocytes and macrophages. Collagen and SCieNTifiC RePoRts | (2018) 8:8549 | DOI:10.1038/s41598-018-26864-w endothelial cells lining the alveolar wall were mostly negative. miR-145 expression was observed in a few pneumocytes type I, while most were negative. Staining in macrophages was predominantly negative.
Total RNA isolation. Total RNA from the cells were isolated using the miRNeasy Mini Kit (cat.# 217004, Qiagen, Hilden, Germany). First, 700 μl QIAzol lysis reagent was used to lyse the cells before homogenization and a 5 minute incubation at room temperature. Second, 140 μl chloroform was added, samples shaken, and then incubated for 3 minutes at room temperature. Third, samples were centrifuged at 12000 G for 15 minutes at 4 °C before the upper aqueous phase was transferred and mixed with 100% ethanol. Finally, the samples were transferred to the RNeasy ® Mini column and washed in several steps before elution with 50 μl ddH2O. Samples were stored at −70 °C. cDNA synthesis. For the first strand cDNA synthesis, the miScript II RT Kit (cat.# 218160, Qiagen, Hilden, Germany) was used. First, 100 ng total RNA was mixed with 4 μl 5X miScript HiSpec buffer, 2 μl 10X Nucleics mix, 2 μl miScript reverse transcriptase mix, and RNase-free water to a final volume of 20 μl. Second, samples were incubated for 1 hour at 37 °C, and then incubated at 95 °C for 5 minutes. Finally, all samples were diluted to a total volume of 200 μl using RNase-free water, and stored at −70 °C.
Proliferation assay. The ability of cancer cells to proliferate was evaluated using the real-time cell analyzer xCELLigence, RTCA DP (catalog#05469759001, ACEA Biosciences, San Diego, USA) fitted with the E-plate 16 (catalog#05469830001, ACEA Biosciences, San Diego, USA). Prior to seeding, cells were trypsinized until detached, resuspended in complete growth media, and counted. In accordance with the manufacturer protocol, cells were seeded in quadruplicates into an E-plate after baseline measurements. The E-plate containing cells was positioned in the RTCA DP instrument, located in an incubator preserving the same conditions as used for routine cultivation of cell lines. The cell index was automatically measured every 30 minutes throughout the experiment duration. Growth curves were calculated with the RTCA software version 1.2.1. A minimum of three independent experiments were performed for each cell line.
Migration assay. The ability of cancer cells to migrate was assessed using ibidi TM culture inserts (ibidi GmbH, Planegg, Germany). The inserts consist of two 0.22 cm 2 silicone chambers separated by a 0.5 mm divider. The inserts were positioned into a 12-well tissue culture dish, one insert per well. Roughly 70 µl pre-transfected cell-suspension containing 4-6 × 10 5 cells/ml were added to each chamber. The cells were left to adhere for 24 hours before the insert was removed and images acquired across the cell-free zone at time points 0 hours and 20 hours. The migration potential into the 0.5 mm gap was calculated using the free online software TScratch, version 1.0 (CSElab, Computational Science and Engineering Laboratory, Switzerland). Initially, the functional experiments for this study were designed using three cell lines; the large cell carcinoma cell line H460, the squamous cell carcinoma cell line H520, and the adenocarcinoma cell line A549. In our experiments, however, the cell lines H460 and H520 did not exhibit migrational properties, leaving only the A549 cell line representing the migration experiment. To strengthen our results, we therefor included the large cell carcinoma cell line H661 in the migration study. Interobserver reliability between scorers was assessed by a two-way random effects model with absolute agreement definition. Associations between marker expression, and marker expression and clinicopathological parameters, were examined by Spearman's rank correlation and χ 2 test or Fisher's exact.
Wilcoxon non-parametrical test was used to assess the difference in biomarker expression between lung tumor tissue and non-malignant lung tissue. Statistical significance between proliferation curves was assessed by one-way ANOVA. The Kaplan-Meier method was used to visualize association between marker expression and disease-specific survival (DSS) and the statistical significance between survival curves was tested using the log-rank test. DSS was defined as the time from surgery to lung cancer death. Variables with significant p-values from the univariate analyses were entered into Cox proportional Hazard models. The final models were derived from a backward conditional method with probability for stepwise entry and removal at 0,05 and 0,10.

miR-143 and miR-145 expression in NSCLC cells. ISH expression of miR-143 and miR-145 in NSCLC
cells and metastatic lymph nodes. miR-143 was primarily observed in the cytoplasm of tumor epithelial and stromal cells, while miR-145 was mainly observed in the epithelial and stromal cell nuclei (Fig. 1). Table 2 Fig. 1).  miR-143 and miR-145 inhibit NSCLC migration. Transfection with miR-143 and miR-145 inhibited migration in both the A549 and H661 cell line when compared with cells transfected with the negative control miRNA (Fig. 2). The inhibition was strongest for miR-145 in both cell lines.

Inhibition of proliferation by miR-143 and miR-145.
Both miR-143 and miR-145 inhibited proliferation in the cell lines H460 and A549, and the inhibition was more evident for cells transfected with miR-145 (Fig. 3A,B). Transfection of miR-143 promoted proliferation in the H520 cell line, whereas miR-145 had an inhibitory effect on proliferation in the same cell line (Fig. 3C). In the cell lines A549 and H460, the inhibitory effects of co-transfection with miR-143 and miR-145 in equal concentrations, were equivalent to that of the miR-145 transfection alone. When co-transfecting the H520 cell line with equal concentrations of miR-143 and miR-145, the inhibitory effects displayed by transfecting miR-145 alone were reduced to a degree where the proliferation-rate was not significantly different to the negative control. Simultaneously, the increase in proliferation caused by the miR-143 transfection alone, was greatly reduced when the H520 cell line was co-transfected with both miR-143 and miR-145 in equal concentrations.
Correlation with clinical variables and other molecular markers. There were no significant associations between miR-143 and miR-145 expression in PT or LN+ and clinicopathological prognosticators listed in Table 1.
Between marker correlations with likely biological significance were as follows: LN+T-miR-143 was positively correlated with PT stromal AR expression (r = 0.494: p < 0.001), and inversely correlated with PT tumor epithelial PGR expression (−r = 0.453: p < 0.001). T-miR-143 in PT was correlated with cytoplasmic ERβ in PT (r = 0.215: p < 0.001) and T-miR-145 in PT was correlated with nuclear ERβ expression in tumor cells (r = 0.212: p < 0.001). Other significant correlations were also observed (Supplementary Table 1).  Table 1. The impacts of biomarkers on DSS in PT are presented in Table 2 and Fig. 4. Neither epithelial nor stromal expression of miR-143 or miR-145 showed significant impact on DSS in the overall cohort. Following gender stratification, however, high S-miR-143 was a positive prognosticator in female patients (p = 0.011), while high S-miR-145 was a positive prognosticator in male patients (p = 0.013). Further, the combination of low S-miR-143 and low S-miR-145 was associated with an unfavorable prognosis in the overall cohort (p = 0.007, Fig. 5). In LN+, neither miR-143 nor miR-145 showed impact on DSS in the overall cohort or stratified by gender.

Discussion
In this large retrospective study of 553 NSCLC patients, S-miR-143 and S-miR-145 expression in PT were positive prognosticators in female and male patients, respectively. Further, the combination of low stromal expression of both miR-143 and miR-145 predicted poor DSS in the overall cohort. Cell line studies confirm the tumor suppressive role of miR-143 and miR-145 in NSCLC, further substantiating their importance in lung cancer pathogenesis. We also observe significant correlations with our previously investigated steroid hormone receptors, suggesting that a biologic rationale may cause, or contribute to, the gender related survival impact observed.
To our knowledge, this is the first study investigating the prognostic impact of miR-143 and miR-145 in neoplastic epithelial cells, tumor associated stromal cells and matched metastatic lymph nodes in the same NSCLC cohort.
Associations between miR cluster 143/145 and cancer survival have been reported for different malignancies, results are, however, conflicting. In prostate cancer (PCa), Avgeris et al. 17 demonstrated a shorter disease-free survival in PCa patients with low miR-145 expression levels. Campayo et al. 18 reported similar results for miR-145 in NSCLC patients. These reports confirm our suggestion of high miR-145 expression as a positive prognosticator, herein in NSCLC patients. Contradicting our results, Al Feber et al. 19 and Avgeris et al. 20 , both reported associations between high levels of miR-143 and miR-145 and poor overall survival in esophageal and bladder cancer, respectively. Importantly, none of the aforementioned studies have evaluated survival impact according to cellular compartment, as was performed in our study.    12 demonstrated an inhibition of migration and proliferation of NSCLC cells following transfection with miR-143. Surprisingly, when we transfected the squamous cell carcinoma cell line H520 with miR-143, proliferation was dramatically increased (Fig. 3C). This is in contrast to most studies reporting on the effects of miR-143 on proliferation [23][24][25][26][27] . However, there are studies depicting alternative roles for the miR cluster 143/145 28 , and members of our research group have reported similar findings using breast cancer cell lines 29 . Interestingly, when co-transfecting miR-143 and miR-145 in equal concentrations, the proliferative capacity was markedly reduced (Fig. 3A,B), meaning the net effect of co-transfecting is tumor suppressive.
The use of verified laboratory techniques with meticulously prepared protocols for biomarker handling is a strength with regards to reliability and reproducibility of our results. Our patient cohort is large with an extensive follow-up time, which further substantiate our results. The retrospective study design may represent a weakness with inaccurate clinical patient data.
In line with previous reports, we detect downregulation of both miR-143 and miR-145 in four independent cancer cell lines relative to levels in a non-cancerous cell line ( Supplementary Fig. 1) 30,31 . Thus, extending the general sense of miR-143/miR-145 downregulation in cancer, including NSCLC 21 . Interestingly, this is in contrast with our ISH-results, reporting significantly increased expression of miR-143/miR-145 in tumor cells and adjoining stromal cells in comparison to non-malignant tissue. To our knowledge, this is the first large-scale miRNA in situ NSCLC tissue hybridization analysis reporting an upregulated miR-143/miR-145 expression. These findings are conflicting with the smaller study (n = 48) by Shen et al. 31 , reporting a downregulation of miR-145 expression in NSCLC, by the use of ISH technique. We present a thorough and comprehensive study of miR-143 and miR-145 expression in appropriate cell types by the use of several acknowledged techniques, giving an optimal account of miR-expression in the tumor environment. Due to contributions from stromal cells in tumor growth, it is pivotal to consider the stromal compartment when elucidating biological mechanisms in epithelial cancers 32 . We found that miR-143 was primarily observed in the cell cytoplasm, while miR-145 was mainly observed in the nuclei of epithelial and stromal cells. Further, we report an abundant positivity of miR-143 and miR-145 in fibroblasts and SMCs lining the blood and lymph vessels, consistent with previous reports 9, 33 . miRs are traditionally considered to act within the cell cytoplasm, regulating gene expression post-transcriptionally 34 . However, a number of miRNAs have been localized in the nuclei, although their nuclear functions remain elusive 35 .
In an extensive meta-analysis, Kent et al. 9 highlighted the crucial importance of cell-type localization of miR-NAs, and how a lack of consideration of specific cellular expression of miRNAs may lead to a general misconception that miRNAs are downregulated in neoplastic tissue. Chivukula et al. 36 , Dimitrova et al. 28 and Akao et al. 37 all published results indicating that neither miR-143 nor miR-145 are expressed in tumor epithelial cells, causing the latter group to conclude that these miRNAs are downregulated in malignant tissue. Similar results have been published by other groups investigating a variety of malignancies 21,38,39 . However, only one 28 of the previous studies has focused on the histological cell-type localization of miR-143/miR-145. These factors, combined with the lung cancer cell lines lack of stroma, inflammatory cells and vascularization, may contribute to the discrepancy observed between miR-143/miR-145 expression in cell lines and tissue samples.  In this study, we present interesting correlations between miR-143/miR-145 and steroid hormone receptors expressed in the NSCLC tissue. The finding of gender specific survival significance of miR-143 and miR-145, forces us to consider sex hormones as a relevant factor. Delfino et al. 40 reported a gender-specific miRNA targeting of molecules related to glioblastoma survival. Further, Duttagupta et al. 41 reported differential miRNA expression levels in a gender specific manner. Mounting evidence confirms activation of hormone receptors to be of outmost importance in lung cancer pathogenesis and several interesting cross-talk pathways between steroid hormones and miR-143/miR-145 have been found [42][43][44][45][46] . Both miRNAs play a critical role in ovarian functioning, and a recent report presents miR-143 affecting estradiol production in granulosa cells by targeting KRAS 47,48 . Further, Spizzo et al., 2011 reported that miR-145 downregulates ERα expression in breast cancer 11 . We found correlations between miR-143/miR-145 and AR, the rate limiting enzyme in estradiol production, suggesting the miRs may interact with regional estradiol production and ER signaling in the lung, as observed in breast tissue. In 2012, Paris et al. presented a study on estrogen effects in breast cancer, showing a direct regulation of miRNA expression and ERβ signaling 49 . Herein, ERβ expression correlated with miR-143/miR-145 expression, suggesting a similar link may exist in NSCLC. The aforementioned reports, assembled with our findings, provide a compelling rationale for a biological cross-talk between miR-143/miR-145 and hormone receptors. If validated in larger, confirmatory studies, this may in fact represent new possibilities for targeted therapy for NSCLC patients, using gender, miRNA and hormone receptor expression as therapy selection criteria.

Conclusion
We present high stromal expressions of miR-143 and miR-145 as positive prognosticators in a gender specific manner in early stage NSCLC patients. Our findings indicate that miR-143/miR-145 acts as tumor suppressor molecules in lung cancer, suggesting that these miRNAs may be useful in miRNA based therapy in NSCLC. Further, we highlight the complexity of miR expression, and stress the importance of cell-type specific expression profiling. By accentuating the correlation between miRNA expression and hormone receptor expression, we emphasize the importance of exploring multi-targeted therapies in the treatment of NSCLC patients, as anti-hormonal therapy is highly accessible.