Expression Of Intracellular Components of the NF-κB Alternative Pathway (NF-κB2, RelB, NIK and Bcl3) is Associated With Clinical Outcome of NSCLC Patients

A growing number of studies has shed light on the role of the NF-κΒ in non-small-cell lung cancer (NSCLC). To address the significance of major effectors of the NF-κΒ alternative pathway, we investigated the relationship between NF-κΒ2, RelB, NIK and Bcl3 expression (mRNA and protein) and the clinical outcome of NSCLC patients. NF-κΒ2, RelB, NIK and Bcl3 protein expression levels were assessed by immunohistochemistry in tissue samples from 151 NSCLC patients who had curative resection. mRNA levels were also evaluated in 69 patients using quantitative real-time PCR. Although all studied proteins were overexpressed in NSCLC (P < 0.001 for all), only RelB mRNA levels were strongly increased in cancerous specimens compared to tumor-adjacent non-neoplastic tissues (P = 0.009). Moreover, NF-κB2, RelB and Bcl3 expression was associated with overall survival (OS). In particular, cytoplasmic and mRNA expression of RelB was related to 5-year OS (P = 0.014 and P = 0.006, respectively). Multivariate analysis also showed that Bcl3 expression (nuclear and cytoplasmic) was associated with increased 5-year OS (P = 0.002 and P = 0.036, respectively). In addition, higher Bcl3 mRNA levels were associated with inferior OS in stages I & II and improved OS in stages III and IV after 5-year follow-up (P = 0.004 and P = 0.001, respectively). Furthermore, stage I patients with lower NF-κB2 mRNA levels had better 5-year survival in univariate and multivariate analysis (P = 0.031 and P = 0.028, respectively). Interestingly, RelB expression (cytoplasmic and mRNA) was inversely associated with relapse rates (P = 0.027 and P = 0.015, respectively), while low NIK cytoplasmic expression was associated with lower relapse rates (P = 0.019). Cytoplasmic NIK expression as well as NF-κB2/ Bcl3 detection was associated with lymph node infiltration (P = 0.039 and P = 0.014, respectively). The present study confirms the deregulation of the NF-κB alternative pathway in NSCLC and also demonstrates the importance of this pathway in prognosis, recurrence and infiltration of regional lymph nodes.

nf-κB2, RelB, NIK, Bcl3 are overexpressed in NSCLC. All studied molecules were expressed more frequently in tumor tissues compared to tumor-adjacent, non-neoplastic tissues. In particular, cytoplasmic NF-κB2 and RelB were detected in 97.4% and 67.1% of NSCLC specimens, while nuclear immunostaining for NF-κB2 and RelB was observed in 18.7% and 47.2%, respectively. On the contrary, NF-κB2 and RelB were observed in both compartments in only 10% of tumor-adjacent non-neoplastic tissues (Fig. 2). Moreover, expression levels of NF-κB2 and RelB were significantly higher in tumors compared to non-neoplastic tissues (Fig. 2, P < 0.001 for both). Bcl3 and NIK were also detected in most NSCLC cases (100% and 92.5%, respectively), while no signal was detected in adjacent non-neoplastic specimens (Fig. 2, P < 0.001 for both of them). With regard to mRNA expression, RelB mRNA levels were strongly increased in cancerous specimens compared to tumor-adjacent non-neoplastic tissues (Fig. 3b, P = 0.009), while mRNA levels of NF-κB2, Bcl3 and NIK did not differ between neoplastic and non-neoplastic tissues (Fig. 3).
Expression levels of RelB, Bcl3, ΝΙΚ and NF-κB2 were associated with overall survival. By univariate analysis, patients with low or intermediate cytoplasmic expression of RelB had improved 2-year survival compared to patients with higher expression levels (P = 0.031). However, this difference was lost when assessing 3-and 5-year survival (Fig. 4a, P = 0.183 and P = 0.128, respectively). Survival was also associated with RelB mRNA levels ( Fig. 4b, P = 0.023 for 5-year follow-up). In addition, the prognostic significance of the cytoplasmic and mRNA expression of RelB for 5-year OS was also observed using multivariate Cox proportional hazards models adjusted for age, grade, primary location, smoking, stage, histological subtype and maximum diameter (P = 0.014; HR, 0.288; 95% CI, 0.107-0.776 and P = 0.006; HR, 1.242; 95% CI, 1.065-1.449, respectively).
With regard to Bcl3, both protein and mRNA expression levels were associated with OS. Low or high nuclear or cytoplasmic protein levels were associated with worse 5-year survival outcome compared to intermediate expression in univariate analysis (Fig. 4c,d, P = 0.011 and P = 0.096, respectively). Furthermore, these observations were statistically significant in Cox regression models, using the age, grade, gender, histological subtype and primary location as coefficients (P = 0.002; HR, 2.588; 95% CI, 1.409-4.751 and P = 0.036; HR, 1.901; 95% CI, 1.044-3.460, respectively). Although mRNA levels appeared to have a poor prognostic value in the cohort as a whole ( Fig. 5a, P = 0.316), stratification according to stage and using a cut-off point of 2.09, showed an inferior OS in stage I and II patients with high Bcl3 mRNA levels ( Fig. 5b) not only in univariate (P = 0.004), but also in multivariate analysis (age, grade, histological subtype, primary location, maximum diameter, smoking as coefficients, P = 0.030). Higher Bcl3 mRNA levels in stages III and IV were associated with improved clinical outcome after a period of 5-year observation, which was statistically significant using a cut-off point of 0.43 as determined by the X-tile tool in this subpopulation (Fig. 5c, P = 0.001). In addition, this association remained after multivariate analysis (P = 0.013; HR, 0.493; 95% CI, 0.283-0.859).
Interesting was also the observation that although NIK cytoplasmic and mRNA levels weren't associated with OS in the whole cohort (P = 0.125 and P = 0.760, respectively), further stratification based on disease stage showed a significant association of NIK cytoplasmic signal with OS. Particularly, patients of stages II and III with low or intermediate levels had better clinical outcome after 5-year observation (P = 0.006, Fig. 4e), but not patients of stage I (P = 0.540, Fig. 4f). This correlation continued to be statistically significant in Cox regression models, using age, grade, gender, histological subtype and primary location as coefficients (P = 0.035; HR, 0.475; 95% CI, 0.237-0.951).
Another interesting finding was the association of the NF-κB2 mRNA expression with OS after stratification with pathological stage. In particular, patients of stage I with lower NF-κB2 mRNA levels had better 5-year survival outcome compared to patients with higher expression in univariate (Fig. 5d, P = 0.031), as well as in multivariate analysis using the age, grade, gender, histological subtype and primary location as coefficients (P = 0.028; HR, 0.043; 95% CI, 0.003-0.714).
In addition, patients without regional LN infiltration and lower NF-κB2 mRNA levels had statistically significant improved survival rates after 5 years observation in comparison to thοse with higher expression (Fig. 5e, P = 0.004). Interestingly, the same correlation was also observed in multivariate analysis adjusted for age, grade, gender, histological subtype and primary location (P < 0.001; HR, 0.043; 95% CI, 0.007-0.245). On the contrary, patients with infiltrated N1 or N2 LNs and increased NF-κB2 mRNA expression had better survival than patients with decreased NF-κB2 gene expression. This association was statistically significant using Cox proportional hazards models adjusted for age, grade and primary location (P = 0.007; HR, 4.673; 95% CI, 1.519-14.372), but not in univariate analysis where it approached but did not quite achieve statistical significance ( Fig. 5f, P = 0.059). www.nature.com/scientificreports www.nature.com/scientificreports/ Expression levels of NIK and RelB were correlated with relapse rate. Expression levels of RelB and NIK were associated with relapse rate (Fig. 6). In particular, RelB cytoplasmic signal was inversely associated with relapse rate, with higher cytoplasmic signal being related to lower relapse rate in univariate analysis (Fig. 6a, Association of relapse rate with NIK cytoplasmic expression was also observed, with lower expression being related to no relapse ( Fig. 6c, P = 0.019). This association was also observed in logistic regression analysis using age, stage, maximum diameter, histological subtype and gender as coefficients (P = 0.028; HR, 2.87; 95% CI, 1.119-7.364). On the contrary, NIK mRNA levels weren't correlated with relapse status (Fig. 6d) and, similarly, no significant association was found between Bcl3 and NF-κΒ2 expression, either in protein or in mRNA expression levels, with respect to relapse rate. NIK expression is related to regional lymph node infiltration. Cytoplasmic NIK expression in primary lesions was associated with infiltration of lymph nodes ( Supplementary Fig. 1a,b, P = 0.039), while this association was not observed for mRNA expression (P = 0.743). Patients with infiltrated lymph nodes at baseline had lower cytoplasmic NIK expression compared to patients with absence of lymph node infiltration. On the contrary, no association was found between cytoplasmic NF-κΒ2 expression, RelB, Bcl3 and NF-κΒ2/ RelB "co-expression" and regional lymph node infiltration. Interestingly, NF-κΒ2/ Bcl3 "co-expression" was associated with lymph node infiltration, with concurrent signal for both molecules being higher in patients without lymph node disease (P = 0.014).
Associations of studied molecules with pathological parameters. Interestingly, patients of stages I and IV were found to have lower NF-κΒ2 cytoplasmic expression compared to patients of stages II-III (P = 0.018). Similar differences were not detected in mRNA levels, as gene expression of NF-κΒ2 was stable across stages (P = 0.356). www.nature.com/scientificreports www.nature.com/scientificreports/ Furthermore, protein and/or mRNA levels were correlated with tumor size. In particular, higher levels of NF-κΒ2 mRNA were related to larger primary lesions (P = 0.020). On the contrary, lower nuclear expression of Bcl3 was associated with larger maximum diameter of primary lesions (P = 0.039). In addition, using a two-tier grading system, tumors with high grade displayed higher expression of cytoplasmic NF-κΒ2 and NF-κΒ2 mRNA levels ( Supplementary Fig. 2a, P = 0.046 and Supplementary Fig. 2b).
With regard to histological subtype, expression of RelB and NIK but not NF-κΒ2 and Bcl3 was statistically significant. RelB cytoplasmic expression was higher in squamous-cell carcinomas compared to adenocarcinomas  Fig. 3a, P = 0.003). In addition, RelB nuclear expression was detected in higher levels in squamous carcinomas than adenocarcinomas ( Supplementary Fig. 3b, P = 0.039). No difference was found in mRNA levels between adenocarcinomas and squamous cell carcinomas, while both had lower RelB mRNA levels compared to large cell carcinomas. In addition, cytoplasmic signal for NIK in squamous cell carcinomas was higher compared to adenocarcinomas ( Supplementary Fig. 3c, P = 0.022), while no difference was found in mRNA levels.

Discussion
The involvement of the alternative pathway of NF-κΒ has been increasingly recognized in lung cancer initiation, progression and clinical outcome as well as in response to treatment 17,20 . We have demonstrated previously that NF-κΒ2 and RelB are overexpressed in non-small-cell carcinomas 17 . Prompted by the previously reported preliminary data, we sought to clarify further the role of this pathway in NSCLC by assessing protein and gene expression of NF-κΒ2, RelB, Bcl3 and NIK in a bigger cohort. We showed that the expression of the major intracellular components of the NF-κΒ alternative pathway i.e. NF-κΒ2, RelB, NIK and Bcl3 are particularly deregulated in NSCLC. All four proteins as well as RelB mRNA levels were increased in neoplastic tissues compared to tumor-adjacent, non-cancerous tissues. In addition, expression levels of RelB, Bcl3 and NF-κΒ2 were associated with OS and those of NIK and RelB with relapse rate. Furthermore, NIK and NF-κΒ2 were related to regional lymph node infiltration. These findings further reinforce the evidence of the activation of this pathway in lung cancer, with findings to be consistent with previous research 17,18,21 .
Overexpression of NF-κΒ2 has been documented also in hematological cancers (cutaneous T lymphomas, NK/T lymphomas), as well as in a plethora of solid tumors such as myeloid thyroid cancer, pancreatic adenocarcinoma, breast, prostate, esophageal, NSCLC and colon cancers 17,[22][23][24][25][26][27][28][29][30] . Moreover, the observed overexpression of NF-κΒ2 in NSCLC is consistent with the overexpression of the TNF receptor family member, LTβR (lymphotoxin-β receptor), that occurs in 87 to 96% of a wide range of solid tumors, including lung cancert 21,31 . LTβR is upstream of NF-κΒ2 and its activation results in the activation of the NF-κΒ alternative pathway 21 . In addition, another activator of the alternative NF-κΒ pathway, CD40 and its ligand, CD154, have been found to be overexpressed in 51.9 and 58.9% of NSCLC patients, respectively 32 . This further reinforces the fact that this pathway is undergoing deregulation. Notably, one of the major findings of this study is the correlation between high cytoplasmic expression of RelB and poor OS in NSCLC patients. Our findings are corroborated by a number of other studies [33][34][35] . OS rates from the KM estimate, based on mRNA levels from publicly available databases, also suggest a significant difference for adenocarcinomas but not for squamous carcinomas (Supplementary Fig. 4). In addition, low RelB activity has been related to a favorable survival of patients with chronic lymphocytic leukemia (CLL) 33 . Furthermore, an association of high RelB protein levels with shorter OS of NSCLC patients has also been reported 34 . Recently, the prognostic significance of RelB levels has also been shown for patients with grade III and IV gliomas, whereby low RelB levels were associated with longer OS 35 .
A possible explanation for the translational value of RelB may lie on the functional connection of RelB with cancer cell growth, migration, and invasion. In DU145 prostate cancer cells, RelB seems to function as an oncogene 36 . Also, RelB in combination with RelA activity sustains the basal survival of CLL cells and renders them sensitive to proteasome inhibition 33 . In addition, RelB in advanced ovarian cancers, supports tumor-initiating cells through the cancer stem-like associated enzyme aldehyde dehydrogenase (ALDH), and the loss of RelB leads to reversion of chemoresistance and inhibition of tumorigenesis in mouse xenograft models 37 . RelB has also been implicated in fostering the stemness of osteosarcoma cells through the paracrine action of cancer-associated mesenchymal stromal cells 38 . Furthermore, in multiple myeloma, it is well known that RelB is able to exert a crucial anti-apoptotic role in malignant cells 39 . In addition, we have to note that RelB regulates gene expression by alternative mechanisms (e.g. epigenetic modifications, dimerization with dimers with the aryl hydrocarbon receptor) and it is not only part of the alternative signaling pathway of NF-kB 40,41 .
Similarly to RelB, elevated expression of Bcl3 is related to poor OS in stages I-II and to improved OS in stages III-IV patients. Elevated expression of Bcl3 at diagnosis has also been associated with poor prognosis of patients with multiple myeloma 42 . Similarly, an inverse correlation of Bcl3 expression with survival has been documented for patients with colorectal adenocarcinomas 43 . Importantly, this finding is consistent with our data of poor survival and in concordance with KM analysis especially with increased gene expression of Bcl3 in patients with stages I-II ( Supplementary Fig. 5) 44 . On the contrary, increased Bcl3 expression was associated with improved OS in stage III and IV patients, a finding which is also consistent with the trend observed in survival analysis from KMplotter, although it didn't reach statistical significance (Supplementary Fig. 5c). Notably, this stage-dependent association between mRNA levels and OS may reflect the different treatment regimens administered in the different stages of disease.
In addition to associations with OS, expression of members of the alternative NF-κΒ2 pathway was also related to local cancer relapse. Notably, our study is the first to our knowledge to report that RelB expression (in protein and mRNA level) was inversely associated with local relapse. Again, this observation is in agreement with Moreover, it appears that this association may be tumor type specific, as no association was noted between expression of RelB and recurrence in ER-positive breast cancer patients 45 .
A linear association of relapse rate and regional lymph node infiltration with NIK cytoplasmic expression was also detected, with lower expression related to the absence of relapse. These findings may reflect NIK's involvement in NSCLC progression through the activation of the NF-κΒ alternative pathway. Although, here, we document for the first time that NIK is overexpressed in NSCLC, it is well documented that NIK is overexpressed in culture and in in vivo models of NSCLC, representing a molecular switch of the activation of the NF-κΒ alternative pathway 46 . Furthermore, possible mechanisms through which NIK influences metastatic potential in lung cancer cells have been described. For example, NIK depletion induced apoptosis in A549 cells and in H1299 cells reduced their colony-forming efficiency, contributing to the oncogenic phenotypes of NSCLC cells 46 . In addition, NIK has been implicated in the regulation of breast cancer stem cells, while its inhibition can impair clonogenicity and tumorigenesis, not only through the activation of NF-κΒ, but also through the activation of the ERK1/2 pathway 47 .
Despite the promising results, we must acknowledge some limitations of our study. Tissue samples were all surgical specimens and therefore represent mainly early and locally advanced NSCLC patients, while stage IV is underrepresented. Additionally, a larger cohort could lead to even more robust associations especially in assessment of gene expression and in stratification analyses.
In conclusion, we report for the first time a comprehensive expression analysis of the major intracellular components of the alternative pathway of NF-κΒ and its prognostic significance in NSCLC. Our data provide strong evidence that this pathway is particularly deregulated in NSCLC, influencing concurrently the prognosis of patients.

Methods
Study design, population, tissue specimens and data collection. In this study, ethical guidelines of the Helsinki Declaration were followed (2013) 48 . Prior to study initiation, approval was granted by the Scientific Committee and the Committee on Research and Ethics of the University Hospital of Patras (Greece). Informed consent was obtained from all the participants.
Protein and mRNA expression of four members of the alternative pathway (NF-κB2, RelB, NIK and Bcl3) were studied in patients, who had undergone a curative resection of a lung tumor at the University Hospital of Patras between 2005 and 2010. This retrospective analysis was performed blindly using an archival database from the Pathology Department of the University Hospital of Patras which allowed access to formalin-fixed paraffin-embedded (FFPE) tissue specimens of both invasive NSCLC as well as adjacent non-neoplastic lung parenchyma. Clinical information was collected from medical records or through direct communication with the patients. Overall survival (OS) was defined up to a follow-up period of 60 months. The workflow of the study is described in Fig. 7. immunohistochemical analysis. Immunohistochemistry was performed as described previously 17,18 . The primary antibodies used against NF-κΒ2, NIK and Bcl3 were mouse monoclonal, while the antibody against RelB was rabbit polyclonal. Conditions for each primary antibody (clonality, clone, dilution, antigen retrieval and incubation time) are presented in Supplementary Table S2. The Envision detection kit (DAKO) was used for detection and diaminobenzidine (DAB) was used as the chromogen for visualization according to the manufacturer's instructions. Dehydrated Harris' hematoxylin solution was used for counterstaining the sections. The specificity of the method was tested using protein blocking solution instead of the primary antibodies in consecutive sections. Inflammatory cells of the tumor microenvironment were used as internal positive controls. evaluation of immunohistochemistry. All slides were assessed independently and blinded to each case by one pathologist (H.P.) and one investigator (F.D.). The histological type and tumor grade were confirmed based on the 2004 WHO classification 49 . Evaluation of the immunohistochemical signal was performed as described www.nature.com/scientificreports www.nature.com/scientificreports/ previously 17,18 . Cases were considered positive when staining was noted in >10% of cells. The distribution and intensity of the NF-κB2, RelB, NIK, and Bcl3 signals were used to estimate NF-κB2, RelB, NIK, Bcl3 expression. Staining was graded on a scale of 0-3 according to the intensity and the percentage of immunopositive cells as follows: 0: no staining or <10% positive cells; 1: weak staining in >10% of cells or moderate staining in 10-70% of cells; 2: moderate staining in >70% of cells or strong staining in 10-70% of cells; 3: strong staining in >70% of cells. NF-κB2, RelB, NIK, and Bcl3 protein expression in cancer cells was categorized in three groups (high vs medium vs low) using as a cut-off the 33 rd and 66 th percentiles or cut-offs derived from X-tile software 50,51 . For each slide, a total score was calculated as the sum of the intensity and the distribution (values ranging between 0 and 6). In order to obtain microphotographs, a Nikon DXM 1200 C digital camera with ACT-1C software mounted on a Nikon Eclipse 80i microscope (Nikon Instruments Inc., Melville, NY, USA) was used.
Gene expression analysis by quantitative real-time PCR analysis (qRT-PCR). RNA preparation. Four 10 μm slides of neoplastic, and when available, paired, non-malignant, adjacent FFPE tissue specimens from 69 NSCLC patients were used to extract RNA samples using the commercially available kit, NucleoSpin ® totalRNA FFPE Kit (MACHEREY-NAGEL, GmbH & Co., Düren Germany), according to the manufacturer's instructions. Isolated RNA samples were then treated with DNase (Ambion, Austin, TX, USA), and total RNA was quantified using a Nanodrop-1000 spectrophotometer (NanoDrop, Fisher Thermo, Wilmington, DE, USA) and stored at −80 °C. cDNA synthesis. A total of 3 μg of RNA was reverse transcribed into cDNA using 100U of Superscript III Reverse Transcriptase (Life Technologies), 300 ng of random nonamer primers (Foundation for Research and Technology-Hellas, Crete, Greece) and 100 nM dNTPs (Stratagene) in a total volume of 50 μl. To control the RNA samples for DNA contamination, a no enzyme control was used. Additionally, a commercially available RNA sample (Stratagene) was used as a calibrator in every batch of reverse transcription reactions to account for run to run variablility. The mixture was incubated in a C1000 Touch thermal cycler (Bio-Rad) at 25 °C for 5 minutes, 50 °C for 60 minutes, and 70 °C for 15 minutes. CDNA was diluted to 15 ng/μl and stored at −20 °C.
Quantification of gene expression. Expression levels of the NF-κB2, RelB, NIK and Bcl3 genes were quantified by real-time PCR (qPCR) assays. Specific primers and probes for NF-κB2, RelB, NIK, Bcl3 and IPO8 genes (Importin 8 was used as a reference gene) were designed to bind to all isoforms using OligoAnalyzer 3.1 (Integrated DNA Technologies, Inc.) 52 , according to the sequences provided in NCBI (http://www.ncbi.nlm.nih.gov/). Primers and probes were synthesized by IDT. Primer sequences and reaction conditions can be provided upon request. The qPCR reactions were carried out in triplicates, in a total volume of 20 μl, containing 5 μl of cDNA in 1× Kapa Probe Fast Master Mix (KAPA BIOSYSTEMS, Woburn, MA, USA) in an MX3000p cycler (Stratagene, La Jolla, CA, USA). Relative expression levels were calculated using the LinReg Program 53 and were normalized to levels obtained for the calibrator sample and to IPO8 levels.
Statistical analysis. The Statistical Package for Social Sciences Version 17 (SPSS, Chicago, IL, USA) was used for statistical analysis. Associations between protein expression and clinicopathological parameters of patients were assessed by using the χ 2 test for nominal variables and the Kruskal-Wallis or the Mann-Whitney tests for ordinal variables. T test was used for continuous variables such as gene expression. Spearman's correlations were used to assess associations between variables. The association of expression levels with relapse rates was evaluated using logistic regression models, adjusted for coefficients. The Kaplan-Meier method and the log rank test were used for the estimation of survival rates. The prognostic significance of the studied molecules was evaluated by Cox regression analysis. The X-tile software was used in order to provide the best cut-off points 51 . For all comparisons, statistical significance was defined as P < 0.05.