The roles of MASPIN expression and subcellular localization in non-small cell lung cancer

Abstract Accumulating studies have confirmed that mammary serine protease inhibitor (MASPIN) plays an essential role in non-small cell lung cancer (NSCLC). However, results are still controversial or inconsistent. In the present study, we attempted to identify the clinical significance of MASPIN and its potential molecular roles in NSCLC. The correlation of MASPIN with prognosis and clinicopathological characteristics was assessed by meta-analysis. Additionally, the potential molecular mechanisms of MASPIN in NSCLC was also investigated through several online databases. A total of 2220 NSCLC patients from 12 high quality studies were included and the results indicated that up-regulated MASPIN nucleus and cytoplasm expression was associated with poor overall survival (OS) (hazard ratio (HR) = 1.43, 95% confidence interval (CI) = 1.01–2.04, P<0.05), elevated MASPIN cytoplasm expression was associated with poor OS (HR = 1.45, 95% CI = 1.01–2.07, P<0.05), disease-free survival (DFS) (HR = 1.95, 95% CI = 1.31–2.88, P=0.001), and disease-specific survival (DSS) (HR = 2.17, 95% CI = 1.18–3.99, P=0.013). MASPIN both nucleus and cytoplasm location were associated with clinicopathological characteristics. Bioinformatics analysis validated the above results and suggested that human serpin family B member 5 (SERPINB5) hypomethylated levels were negatively correlated with its mRNA expression. Bioinformatics analysis also revealed the 85 most frequently altered neighboring genes of SERPINB5, and gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed 20 GO terms and 3 KEGG pathways with statistical significance. MASPIN had a statistically negative correlation with NSCLC prognosis, functioning as an oncoprotein by hypomethylation and influencing specific pathways involving the 85 genes identified herein. MASPIN might be a promising prognostic signature in NSCLC.


Introduction
The mammary serine protease inhibitor (MASPIN), encoded by the human serpin family B member 5 (SERPINB5) gene, was first reported in 1994 as a tumor suppressor [1]. MASPIN is highly expressed in normal mammary epithelial cells but is often down-regulated or not expressed in the progression of breast malignant tumors [1,2]. Advanced research found that MASPIN inhibits cell invasion, promotes apoptosis, and inhibits angiogenesis, so it is considered to have class II tumor-suppressive properties [3]. MASPIN localizes in not only primarily cytoplasm (in the mammary epithelial cells), but also nucleus (in the myoepithelial cells), secretory vesicles, and cell surface [4]. MASPIN subcellular localization appears linked to its biological functions [5], and its nuclear localization indicates a more favorable prognosis in human malignancies, such as breast cancer [6], pancreatic cancer [4], and ovarian carcinoma [7]. This phenomenon also appears in non-small cell lung cancer (NSCLC) [8], which is one of the most familiar malignant tumors [9,10] and can be mainly divided into lung squamous cell carcinoma (LUSC) and lung adenocarcinoma (LUAD), and lung large cell carcinoma, pathologically [11]. One study shows emaciated cytoplasmic MASPIN expression could consider as an independent positive predictor in primary NSCLC [12], whereas another study shows that dominant nuclear expression was related to positive prognosis in resectable NSCLC [13]. This paradoxical localization and expression of MASPIN might have cell-specific characteristics [14] and play different roles in NSCLC. Moreover, MASPIN may have an ability to limit malignancy cell phenotypic plasticity and affect the cancer cell response to drugs [15], so it might be a valuable marker and potential therapeutic agent [5]. Thus, we conducted present study to clarify the roles of MASPIN expression and subcellular localization in NSCLC.

Studies search and selection
We performed a systematic search on Medline, Embase, and Web of Science (WOS) for relevant published studies from inception to 31 January 2019, with language restricted to English. The following keywords were used in different combinations: 'Carcinoma' , 'Non-Small-Cell Lung' , 'MASPIN' , 'SERPIN-B5' , 'Prognosis' . Search strategies in Medline were available in Supplementary Table S1.
Inclusion criteria for studies: (1) participants were pathologically diagnosed with NSCLC; (2) the expression and subcellular localization were detected by immunohistochemistry; (3) patients were divided into high and low expression groups according to MASPIN expression and the relationship between MASPIN and NSCLC prognosis (such as overall survival (OS), disease-free survival (DFS), progression-free survival (PFS), and disease-specific survival (DSS)) and clinicopathological characteristics were described; (iv) relevant data were sufficient to obtain hazard ratio (HR) and corresponding 95% confidence interval (CI); (v) cohort studies. Exclusion criteria included reviews, case reports, comments, expert opinions, meta-analysis, and conference summaries. When studies or data duplication were found, only the latest or most complete studies were included.

Data extraction and management
Citations were screened at the title and abstract level by two independent reviewers, and the full text was retrieved for those potentially eligible. For studies that did not provide HRs and 95% CIs directly, we estimated them through indirect methods [16]. If only Kaplan-Meier curves were available, HRs were extracted from graphics by Engauge Digitizer [17]. One reviewer extracted data of interest into a predesigned excel spreadsheet. Then, another two reviewers checked received data and contradictory cases were resolved by the final author.

Quality assessment
Two reviewers critically and independently assessed included studies using the Newcastle-Ottawa Scale (NOS) [18,19]. The assessments include selection, comparability, and outcome. The studies were divided into three grades: low, medium, and high, with NOS scores ranging from 0 to 3, 4 to 6, and 7 to 9, respectively. All data were finally shown in the predesigned excel spreadsheet and discrepancies, contradictories were resolved by the third author.

Data synthesis and analysis
The STATA, statistical software for data science, was adopted to synthesize and analyze data. Similar to the method described previously [20], heterogeneity was assessed by I 2 tests. If the data were substantially homogeneous, a synthesis was directly conducted with the fixed-effect model. Otherwise, a synthesis was directly conducted with the random-effect model, successively subgroup analysis was performed. Sensitivity analysis was additionally conducted to investigate the robustness of our results. At last, the funnel plot and Egger's tests were applied to assess for publication bias. Later, the different subcellular location was also analyzed.

Bioinformatics analysis
The SERPINB5 mRNA and protein expression between NSCLC and normal tissue were studied using UALCAN [21] and Human Pathology Atlas [22], respectively. The association between SERPINB5 and OS or PFS and the clinicopathological significances of SERPINB5 in NSCLC were all conducted in the Kaplan-Meier plotter [23]. The patients were grouped by median SERPINB5 expression and HR with 95% CI were calculated. In addition, MethHC [24] was utilized to compare the SERPINB5 promoter region methylation between NSCLC and normal tissue as well as the relationship between mRNA expression and methylation in NSCLC patients. One-way ANOVA was used to analyze differential expression. Moreover, the genes co-expressed with SERPINB in TCGA-LUAD were identified using the  [25]. Only those genes with Spearman's correlation coefficients of more than 0.4 were identified as SER-PINB-related genes. Finally, SERPINB-related genes were loaded into DAVID [26,27] for gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, and the illustration of the biological significance was performed using the clusterProfiler package in R [28].
Twelve studies comprising 2220 patients contained information regarding OS, PFS, DFS, and DSS. All included studies were published in 2005 and later, and more than half of the studies were conducted in Japan [12,[31][32][33][34][35]39], only three studies involved LUSC or LUAD [31,33,34]. Only one study did not report gender composition [37]. Two studies did not report the median age [31,37], and the median age of the included individuals was approximately 65 years old or so. Only two studies had no staging information for NSCLC [36,37]. As for subcellular localization, three studies [12,30,34] only involved cytoplasm, while others included both cytoplasm and nucleus. Some HRs were extracted directly from corresponding studies [29][30][31][32][33][34][35], and some were extracted from corresponding curve [12,32,[35][36][37][38][39], and some were not reported [29,32,38]. The included studies also had different definitions of the cut-off value. NOS scores of all eligible studies were greater than 7. The main characteristics of the included studies are shown in Table 1.

Sensitivity analysis and publication bias
A sensitivity analysis of OS was performed to verify the robustness of consequences. The results disclosed there had no significant heterogeneity ( Figure 2D). The sensitivity analysis and assessment of publication bias with PFS, DFS, and DSS were not conducted because there were few numbers of included studies. The funnel plot indicated no publication bias analysis of OS ( Figure 2E). Moreover, Egger's test revealed no obvious publication bias (P=0.795, Figure 2F).

DNA methylation of SERPINB5 and its correlation with mRNA expression in NSCLC patients
To recognize the role of methylation in regulating SERPINB5 expression in NSCLC patients, MethHC was applied to explore the level of DNA methylation in the SERPINB5 gene promoter region and its correlation with mRNA expression. The difference of the SERPINB5 methylation level between LUAD and normal samples was statistically significant (P<0.005, Figure 8A). Additionally, SERPINB5 methylation was negatively correlated with its mRNA expression in LUAD (r = −0.128, P<0.001, Figure 8B). The statistically significant results were also indicated in LUSC patients ( Figure 8C,D).

Function and pathway enrichment analysis of SERPINB5-related genes in LUAD
To comprehend the functions and potential mechanisms of SERPINB5 in NSCLC, cBioPortal [25] was used to identify SERPINB5-related genes in LUAD. There were 85 genes with Spearman's correlation coefficients more than 0.4 and were identified as SERPINB-related genes in TCGA-LUAD. Besides, 85 genes were put into the DAVID [26,27] for GO and KEGG pathway enrichment analysis, revealing 20 GO terms (Supplementary Table S3) and 3 KEGG pathways (Supplementary Table S4) with statistical significance. These corresponding GO terms were considered as the most specific and useful for describing the concrete function of SERPINB5 in LUAD, the interaction between different GO terms were visualized as three GO maps ( Figure 9A-C). KEGG pathway enrichment analysis results subjected that the genes were enriched in amoebiasis, metabolic pathways, and small cell lung cancer ( Figure 9D).

Discussion
Up to date, more and more studies are emerging to investigate MASPIN expression and its roles in malignant tumors, and carry out in-depth researches on different mechanisms of MASPIN about inhibiting tumor cell plasticity and affecting the context-dependent response to drugs [15]. Nevertheless, only a minority of MASPIN expression and subcellular localization were evaluated in NSCLC and no unanimous agreement was reached. One study announced that MASPIN cytoplasm expression alone could be a useful negative prognostic effect in p-stage IA LUAD [33].
Other studies indicated strong MASPIN nuclear expression was an independent favorable prognostic factor in NSCLC [29,40], and it improved differentiated epithelial phenotypes, decreased tumor angiogenesis, and increased tumor sensitivity to drug-induced apoptosis [41]. In contrast, Bircan et al. indicated MASPIN had no prognostic significance although it was higher in both LUSC and LUAD [42]. There is no doubt that larger studies are urgently needed to reveal the exact role of MASPIN in the occurrence, development, and prognosis of NSCLC. Hence, we summarized 12 studies with 2220 patients, which met preordained inclusion criteria and had high quality according to NOS scores, to obtain more accurate results and further explored the mechanism of MASPIN in NSCLC through different online databases. Clinical data conflict on the prognostic utility of MASPIN [29,36,43], and we found that MASPIN expression was up-regulated in NSCLC and was closely related to prognosis. Importantly, we not only analyzed the relationship between MASPIN and NSCLC prognosis but also applied different subcellular localization. Our results suggested that MASPIN was statistically correlated with poor prognosis (OS, DFS, and DSS) regarding different subcellular localization (nucleus and cytoplasm or cytoplasm alone), which indicated MASPIN might be a promising prognostic signature in NSCLC, although the nuclear-cytoplasmic expression pattern might have stronger influences on metastasis and proliferation than nuclear expression alone [36]. Our subgroup analysis showed that rising MASPIN protein was associated with poor OS in LUAD patients with no statistical difference (P=0.626) while the results from the Kaplan-Meier plotter uncovered that negative correlation between SERPINB5 mRNA and OS and PFS was statistically significant in LUAD patients (P<0.001). In addition, the results of LUAD and LUSC were not identical. The discrepancies might be largely attributable to background heterogeneity, which suggests that larger sample size of clinically relevant studies may be needed to verify this phenomenon. But although LUSC and LUAD account for a large proportion of NSCLC [44], they still do not fully reflect all NSCLC.
In addition, we analyzed the relationship between MASPIN and clinicopathological characteristics from the perspective of meta-analysis and Kaplan-Meier plotter. The results were complementary, as one focused on subcellular localization and the other on prognosis; one is the protein condition and the other is the RNA condition. However, it is found that the expression of both SERPINB5 protein and mRNA is significantly correlated with various clinicopathological characteristics and survival outcomes. Many studies have revealed the relevant mechanism between MASPIN and the occurrence, development, treatment, and prognosis of lung cancer. In LUAD cells, MASPIN can not only down-regulate the expression of matrix metalloproteinase-2 and integrin β1 to significantly inhibit cell migration and invasion but also affect cell growth mainly by targeting AKT signaling molecules. The different functions of MASPIN may be due to its different molecular characteristics [45]. One study suggested that cytoplasmic MASPIN may play an important role in LUAD by regulating apoptosis and, thus, is a favorable prognostic marker for LUAD patients, whereas the nuclear location may be linked to the promotion of angiogenesis [46]. Others hold MASPIN inhibits cell motility by suppressing Rac1 and PAK1 activity [47], promotes cell adhesion via the PI3K/ERK pathway [47,48], and suppresses survival of lung cancer cells through modulation of AKT pathway [49,50]. It was suggested that the loss-expression of MASPIN may participate in the invasion and metastasis of NSCLC and it has a positive relationship to vasculogenic mimicry in NSCLC [51]. About the treatment, resistance to osimertinib partially arises through MASPIN [52] and a novel statistical methodology for detecting splicing changes in exon array data presented that SERPINB5 showed alternative splicing in NSCLC patients treated with bevacizumab/erlotinib [53].
To uncover additional mechanisms, we performed methylation analysis. Online databases from MethHC showed that the underlying mechanism of MASPIN affecting NSCLC survival outcomes may be related to hypomethylation and this was following a previous study [54], and this difference may be critical for epigenetic regulation of radiosensitivity [55]. And then, in this study, we demonstrated that SERPINB5 in both LUAD and LUSC were lowly methylated in the promoter region and that there was an inverse correlation between DNA methylation and mRNA expression. Hypomethylated status means activation of gene expression, therefore, it is not difficult to infer that SER-PINB5 methylation partially contributes to its increased expression and plays an important role in the occurrence and development in NSCLC. Given the statistical difference between SERPINB5 mRNA and prognosis in LUAD and not in LUAC, TCGA-LUAD was selected for the subsequent analysis of co-expressed genes and enrichment to confirmed that SERPINB5 influenced related pathways involving the neighboring genes, as is mentioned in a previous study [53,56], and future investigations and validation are needed. Sensitivity analysis demonstrated that the correlation between MASPIN and NSCLC prognosis was stable and unchanged after removing any one study. Funnel plot and Egger's test were used to detect publication bias and events indicated no obvious publication bias.
However, the present study was not without limitations. Foremost, all the data analyzed in our study were obtained from other published studies or different online databases, which might cause background heterogeneity, as mentioned earlier. Additionally, eight studies were conducted in Asia (China or Japan) which may result in an excursion in race. Later multinational and multiracial studies for this specific population are needed. Third, the included studies had different definitions of MASPIN cut-off value. Therefore, before carrying out relevant researches, it may be necessary to identify the cut-off value in a unified way. Fourth, according to Kaplan-Meier plotter, higher SER-PINB5 mRNA expression was positively correlated with OS and PFS of surgery success (surgical margins negative) (all P<0.0001). Due to the limitations of the data obtained, we could only divide the surgery success into one group, which was bound to cause the loss of information, and the later data with richer dimensions is warranted. Finally, a series of silico-bioinformatic analysis was performed to uncover the promising molecular mechanisms of MASPIN in NSCLC, and more elaborate studies are required to be designed and conducted to validate these results in the future.
In conclusion, our study demonstrated that MASPIN had a statistically negative correlation with NSCLC prognosis, functioning as an oncoprotein by hypomethylation and influencing specific pathways, which lays a foundation for MASPIN to develop into an independent prognostic factor for NSCLC patients.

Competing Interests
X.-F.W. and B.L. were members of Morning Star Academic Cooperation, Helixlife (Shanghai, China) when they did the work described in the article. The authors declare that there are no competing interests associated with the manuscript.