Background
Lung cancer is the leading cause of cancer-related death in developed countries [
1,
2]. In Europe 410,000 new cases of lung cancer and 353,000 related deaths were estimated to occur in 2012 [
3]. In the most frequent type, non-small-cell lung cancer (NSCLC), more than 70% of patients debut with locally advanced or metastatic disease [
4]. In patients that present with early-stage disease, surgery alone or surgery followed by cisplatin-based adjuvant treatment is the first line of treatment. Although surgery is considered a curative treatment, 30–55% of patients will relapse during the first two years [
5]. These data highlight the need to further investigate this disease and consolidate useful prognostic markers.
NK2 homeobox 1 (NKX2–1), also known as thyroid transcription factor-1 (TTF-1), is a key transcription factor that orchestrates the development of the lung, thyroid and forebrain in the embryonic period [
6]. In adult lung tissue, NKX2–1 is expressed in conducting airways type II alveolar epithelial cells and in Clara cells and uniformly in the terminal respiratory unit [
7]. NKX2–1 regulates surfactant protein transcription by directly binding to the promoter of SP-A [
8], SP-B [
9], SP-C [
10], clara cell secretory protein (CCSP) [
11], and T1〈 [
12]. NKX2–1 is commonly used in clinical practice for the differential diagnosis of the adenocarcinoma (ADK) subtype of NSCLC [
13,
14]. More than 70% of ADK are positive for NKX2–1 by immunohistochemistry, independent of disease stage [
15].
Most of the studies that have examined the role of NKX2–1 in oncogenesis have highlighted its role as tumor suppressor. NKX2–1 inhibits proliferation by inhibiting the embryonal proto-oncogene High Mobility Group At-Hook 2 (HMGA2) [
16]. In addition, NKX2–1 inhibits cell motility and metastatic capacity through modification of intercellular junctions and cytoskeleton elements. It promotes the expression of E-Cadherin, Ocludin (OCLN) and CLN 1/18 [
17]. It also inhibits epithelial-to-mesenchymal transition (EMT) by repressing transforming growth factor β (TGF β), which produces an increase of E-Cadherin levels [
18]. NKX2–1 also activates MYBPH synthesis, which inhibits actomyosin filaments, crucial elements in cell migration [
19]. In addition, NKX2–1 regulates transcription of P53, a gene frequently lost or mutated in NSCLC patients [
20,
21]. Reduced expression of NKX2–1 has been associated with initiation and progression of invasive mucinous lung ADK in patients harboring KRAS mutations [
22].
MicroRNAs (miRNAs) are small non-coding RNAs (22–24 nucleotides in length) that regulate post-transcriptional processes via sequence-specific interactions with the 3′ untranslated regions (UTR) of mRNA [
23]. Two miRNAs have been associated with NKX2–1: miR-365 and miR-33a. Using an in silico method, Mu et al. identified and validated miR-365 as a miRNA that directly regulates NKX2–1 by binding to its 3’UTR and inhibiting its translation [
24]. Interestingly, a feed-back signaling loop between miR-365 and TGF β was described, illustrating that miR-365 participates in the NKX2–1 repression of TGF β [
24]. miR-33a, located downstream of NKX2–1, was previously shown to have cholesterol homeostasis regulatory activity and to bind to the 3’UTR of HMGA2. These data indicate that NKX2–1 upregulates miR-33a, which represses HMGA2 [
25] and could inhibit EMT, thus controlling lung cancer metastasis [
26].
The prognostic value of NKX2–1 expression is controversial. Although most studies report that lower expression of NKX2–1 is associated with shorter overall survival (OS) in NSCLC [
27‐
30], others report an inverse prognostic correlation [
31]. We hypothesized that miR-365a, NKX2–1 and miR-33a could work together to influence growth and differentiation of lung cancer cells and that the study of this axis could thus help to understand the discrepancies observed in the prognostic impact of NKX2–1. We have studied the expression of NKX2–1 and its associated miRNAs, miR-365 and miR-33a, in a cohort of 110 early-stage NSCLC patients and correlated our findings with overall survival (OS).
Methods
Patient samples
From June 2007 to November 2013, tumor tissue samples were prospectively collected from 110 adult patients diagnosed with stage I-II NSCLC who underwent complete surgical resection in our institution. All patients had Spanish ethnicity. Tissue samples were immediately immersed in RNALater® (Ambion) and stored at −80 °C until processing. Clinical data were recorded on admission: age, gender, smoking history, preoperative pulmonary function tests, chronic obstructive pulmonary disease (COPD), Eastern Cooperative Oncology Group (ECOG) performance status (PS), clinical and postoperative staging according to TNM 7th edition [
32], type of surgical resection and pathological findings (histological subtype, and the presence of emphysema). Information regarding adjuvant treatment, relapse and clinical outcomes was also recorded. The mutational status of TP53 and K-RAS was assessed in all patients, and EGFR mutational status was assessed in ADK patients.
RNA extraction and gene expression analysis
Total RNA was isolated from frozen tissue using TriZol® Reagent (Life Technologies) according to the manufacturer’s protocol. RNA from samples was quantified using a NanoDrop ND-1000 Spectrophotometer (Fisher Scientific, Madrid, Spain).
cDNA was obtained from 500 ng of total RNA using the High Capacity cDNA Reverse Transcription Kit® (Life Technologies) as per manufacturer’s protocol. NKX2–1 mRNA expression levels were quantified using a TaqMan Gene Expression assay (Hs00968940_m1) in a 7500 Real-Time PCR System (Life Technologies). Relative expression levels were calculated by 2-ΔΔCt method using 18S as endogenous control.
miRNA quantification
miR-365 and miR-33a expression was analyzed using TaqMan MicroRNA Assay (Applied Biosystems) as previously described [
33]. Relative quantification was calculated using 2
-ΔΔCt. Normalization was performed with miR-191 [
34]. All experiments were performed in triplicate.
TP53, K-RAS and EGFR mutation analysis
PCR to identify TP53, K-RAS and EGFR mutations was performed on 50-ng DNA samples and the Sanger sequencing process was performed by STAB Vida (Caparica, Portugal). The mutation analysis for TP53 included the exons 5–8 and used the following primers: exon 5 forward(F) 5′- GTTTCTTTGCTGCCGTCTTC-3′, 5 reverse (R) 5’-GAGCAATCAGTGAGGAATCAGA-3′; exon 6F 5’-AGAGACGACAGGGCTGGTT-3′, 6R 5’-CTTAACCCCTCCTCCCAGAG-3′; exon 7F 5’-TTGCCACAGGTCTCCCCAA-3′, 7R 5’-AGGGGTCAGAGGCAAGCAGA-3′; exon 8F 5’-GGGACAGGTAGGACCTGATTT-3′, 8R 5’-TAACTGCACCCTTGGTCTCC-3′.
The mutation analysis for K-RAS included the codons 12 and 13 and used the following primers: F 5’-TTAACCTTATGTGTGACATGTTCTAA-3′, R 5’-AGAATGGTCCTGCACCAGTAA-3′.
The mutation analysis for EGFR included the exons 18, 19, 20 and 21 and used the following primers: exon 18 F 5’-GCATGGTGAGGGCTGAGGT-3′, 18R 5’-TGCAAGGACTCTGGGCTCC-3′; exon 19 F 5’-TGCATCGCTGGTAACATCCA-3′, 19R 5’-GAAAAGGTGGGCCTGAGGTT-3′; exon 20F 5’-TCCTTCTGGCCACCATGC-3′, 20R 5’-TGGCTCCTTATCTCCCCTCC-3′; exon 21F 5’-ATGCAGAGCTTCTTCCCATGA-3′, 21R 5’-CAGGAAAATGCTGGCTGACC-3′.
TTF-1 immunohistochemistry staining
IHC was performed on formalin-fixed, paraffin-embedded tissue sections of 16 lung carcinomas and 3 normal lung controls from the Pathology Service of the Hospital Clinic of Barcelona after review by a thoracic pathologist. 4-μm-thick transverse sections of formalin-fixed, paraffin-embedded tissue were serially cut and mounted onto Dako Silanized Slides (S3003; Dako, Glostrup, Denmark). For antigen retrieval, the sections were manually immersed in Target Retrieval solution, high pH (Dako) and heated in a water bath at 95–99uC for 20 min. Endogenous peroxidase activity was quenched by immersion in Dako Real Peroxidase-Blocking solution for 10 min. The tissue sections were incubated with primary antibody against TTF1 (1:100, 8g7g3/1, DAKO, glostrop, Denmark). The slides were then washed in Tris–HCl and detected with horseradish peroxidase-conjugated anti-rabbit EnVision + kit (DAKO). Finally, sections were stained with hematoxylin. All slides were blindly scored by the same two pathologists. Nuclear staining of tumor cells was considered TTF1+. Tumors with completely no TTF1 expression in nuclei were de ned as TTF1 − .
Statistical analyses
The primary endpoint of the study was OS, defined as the time between surgery and death from any cause. Kaplan-Meier curves for OS were drawn and compared by means of a log-rank test. All factors with
P ≤ 0.1 in the univariate analysis were included in the Cox multivariate regression analyses for OS. Optimal cut-offs of NKX2–1 expression data for OS were assessed by means of maximally selected log-rank statistics using the Maxstat package (R statistical package, v. 2.8.1, Vienna, Austria) [
35] and confirmed by the Kaplan-Meier test. Student T-Test or Mann-Whitney U test, as appropriate, were used for comparisons between two groups or ANOVA when more than two groups were compared. Pearson correlation was used to compare the NKX2–1 expression with its associated miRNAs. All statistical analyses were performed using PASW Statistics 21 (SPSS Inc.) and R v2.8.1. The level of significance was set at
P ≤ 0.05.
Discussion
The potential prognostic impact of NKX2–1 in NSCLC is unclear. Several studies have found an association between low NKX2–1 expression and good prognosis [
27‐
30,
37,
38] while others have reported an association with poor prognosis [
39,
40] and still others have found no association at all [
41‐
43]. In the present study, we assessed NKX2–1 expression in a cohort of 110 patients with stage I-II NSCLC who had undergone surgical resection as their first therapeutic approach. Our findings indicate that high NKX2–1 expression is associated with longer OS both in the entire cohort (
P = 0.035) and in the subgroup of stage I patients (
P = 0.031). Moreover, among patients with neither TP53 nor KRAS mutations, NKX2–1 expression emerged as an independent prognostic factor for OS (OR 5.335;
P = 0.018) and DFS (OR 4.333,
P = 0.008).
The conflicting findings of previous studies may be due to several factors, including the techniques used for NKX2–1 analysis and the heterogeneity of patient cohorts.
Most of the studies evaluated NKX2–1 expression by Immunohistochemistry [
38], a technique that only differentiates between the presence or absence of a protein but cannot quantify the levels of expression, although in some studies [
30] the authors use an automated quantitative analysis of protein concentration within subcellular compartments to establish a range of expression. We have evaluated NKX2–1 levels using RT-PCR, which is a highly sensitive technique that can better classify patients with intermediate levels of a gene.
Arguably, the most important difference between studies of the prognostic impact of NKX2–1 lies in the clinical characteristics of the patients included, especially disease stage. A meta-analysis of 2235 patients included in 17 studies of NKX2–1 found that 937 patients (80%) had stage IIIb-IV disease [
38]. In fact, the few studies reporting a negative prognostic impact for high NKX2–1 expression were performed in cohorts enriched in stage IIIb patients. Since the inclusion of these patients with locally advanced disease may well confound the identification of prognostic markers, we focused our study on a well-characterized cohort of early-stage patients. Our results are in line with other previous studies [
27] where high NKX2–1 expression was related to longer survival. Interestingly, this association was maintained in the subgroup of stage I NSCLC patients, who had not received adjuvant treatment after surgery, suggesting a clear prognostic role for NKX2–1 mRNA expression.
NKX2–1 expression was upregulated in our patients with wild-type TP53. This can be explained by previous findings that NKX2–1 has been linked to regulation of TP53 transcription via LKB1 loss and IKKβ/NF-κB activation [
20,
44].
.
NKX2–1 gene amplification displayed a positive correlation with the presence of KRAS mutations; however, its prognostic impact remains controversial. In a Japanese ADK cohort,
NKX2–1 amplification was an independent predictor of poor prognosis [
45,
46] NKX2–1 has also been linked to KRAS in that NKX2–1 gene haploinsufficiency in patients harboring KRAS mutations, with a consequent loss of function of NKX2–1, may promote tumorigenesis in mucinous ADK [
47]. In our patients, however, we found no significant differences in NKX2–1 expression according to KRAS status. Nevertheless, when we analyzed the effect of NKX2–1 expression on OS in patient subgroups classified according to TP53 and KRAS mutation status, we found a remarkable impact among patients with both wild-type TP53 and wild-type KRAS. OS in these patients was much longer for patients with high NKX2–1 levels – both in the entire cohort (
P = 0.017) and in stage I patients (
P = 0.005). Furthermore, in an exploratory analysis in this subgroup of patients, we found a strong association between high NKX2–1 expression and longer disease-free survival – again both in the entire cohort (
P = 0.006) and in patients with stage I disease (
P = 0.003). These findings suggest that NKX2–1 functions are linked to the normal functioning of TP53 and KRAS and that mutations in these genes influence the prognostic impact of NKX2–1 expression.
As expected, NKX2–1 expression in our cohort was higher in ADK than in SCC patients, although high NKX2–1 expression has previously been described in SCC as well [
48]. Moreover, we observed an inverse correlation between smoking habit and NKX2–1 levels that is in line with previous reports [
49,
50], suggesting that NKX2–1 downregulation could be related to smoking habits.
The underlying mechanism of NKX2–1 as a tumor suppressor is not fully understood but we speculated that it might be linked to its associated miRNAs, miR-365 and miR-33a. We found a negative correlation between miR-365 and NKX2–1 expression, indicating that miR-365 directly regulates NKX2–1. We also observed an inverse relation between miR-365 and NKX2–1 expression in ADK vs SCC, where miR-365 expression was lower in ADK patients while NKX2–1 expression was higher. Moreover, patients with high levels of miR-365 showed a trend towards shorter OS (P = 0.073). Taken together, these findings provide further evidence for a regulatory role of miR-365 over NKX2–1. In contrast, no significant correlation between NKX2–1 and miR-33a was observed, suggesting that factors other than NKX2–1 are involved in the regulation of miR-33a.