SPP1 targeted by miR-4262 in gastric cancer functions as an enhancer of cell growth and correlates with dismal prognosis

Background Advanced gastric cancer (GC) induces diamal prognosis and high mortality. Discovery of new biomarkers or differentially expressed genes (DEGs) is serving for early diagnosis, prevention and therapautic treatmen in GC. In this study, by combining with biostatistics analysis, we aimed to verify the aberrant high expression and enhancing effects of SPP1 on GC, and to explore the probable relative post-transcriptional regulation. Methods Three datasets (GSE13911, GSE19826 and GSE27342) from NCBI GEO database were explored. SPP1 was screened out and detected in 105 real GC patients through immunohistochemistry analysis and RT-qPCR assay. The patients’ clinicopathologic features were collected and analyzed. The expression of SPP1 was examinated in three GC cell lines (MKN-45, AGS and SNU-16) . MKN-45 cell model with SPP1 depletd was constrcted through shRNA transfection. CCK8 assay, cell cycle detection and apoptosis rate calculation were conducted to evaluate the ability of cell growth. MiR-4262 was filtered out as a potential up-streaming regulator of SPP1 mRNA through bioinformatic prediction, and the dual-luciferase reporter assay was used for validation. Rescue experiment was introduced to confirm the post-transcriptional regulation. significant in both the 105 patients’ samples and the patients’ Whilst, degenerating dual-luciferase assay miR-4262. luciferase in by SPP1. The results are means of three independent experiments ±SD. (*P< 0.05). FG. Cell apoptosis was detected by flow cytometry. The representative histograms describing cell apoptosis profile in MKN-45 cells are shown. The apoptosis rate of MKN-45 cells was significantly increased from 7.92% to 14.12% via SPP1 depletion. The results are means of three independent experiments ±SD. (**P< 0.01).

3 to promote GC prevention, diagnose and therapeutic treatment.

Background
As one of the global health problem, gastric cancer (GC) results in the third leading cause of cancerrelated death (1). Even though the improvements for GC treatment have been gradully achieved, the prognosis and over-all survival rate of the patients remain dismal (2).
The application of Gene Expression Omnibus (GEO) database from National Center for Biotechnology Information (NCBI) provides us possibility for intensive mining the differential expression genes (DEGs) involve in tumorigenesis. In this study, we set up an absolute value of fold-change (FC) of gene mRNA expression with a threshold criteria of log 2 FC≥2.0 and P value 1.0E-04, by which we clustered 13 DEGs amplified and 33 ones decreased from three NCBI GEO datasets (GSE13911, GSE19826 and GSE27342) (3)(4)(5). Among the candidates of the DEGs, we screened Secreted phosphoprotein 1 (SPP1) out of the 13 remarkably amplified DEGs for further study.
SPP1 was firstly reported concering in cell epitheial transformation process (6). Further studies then revealed SPP1 as a pivotal multi-functional genes participating in different processes intracellular, such as the attachment of osteoclasts to mineralized bone matrix (7), and up-regulation of IFN-γ and IL-12 through a cytokine way (8). Deregulation of SPP1, especially amplification in expression, has been reported in various human cancers, exerting critical effects on tumorigenesis and progress. For instance, high expressed SPP1 in epithelial ovarian cancer cells specifically activate Integrinβ/FAK/AKT pathway, and induces cell migration (9); up-regulated SPP1 in prostate cancer, SPP1 mediates the activity of Smad4/PTEN pathway and leads to tumor recurrence and metastasis (10).
However, reports of the exact expression profile and the specific functions of SPP1 in GC are limited.
Thus, we detected and verified the pathological over-expression status of SPP1 in both the real GC patients' specimens and the GC cell lines. MKN-45 cell, one of the GC cell line highly expressing SPP1, was selected for the use of cell model construction. And depletion of SPP1 in the cell model further demonstrate suppressed characteristics of cell growth and mobility. Intriguingly, miR-4246 was detected as an upstream post-transcriptional regulator SPP1 through a classic miRNA modulated degeneration. All the aforementioned findings stongly suggest SPP1 as an effective enhancer in GC 4 progress, which could be negatively regulated by miR-4262.

Gene expression profile data and the identification of DEGs
The gene expression datasets GSE11399, GSE19826 and GSE27342 were downloaded from GEO database (https://www.ncbi.nlm.nih.gov/geo/). Platforms of GEO datasets is GPL570 (Affymetrix Human Genome U133 Plus 2.0 Array) (Agilent Technologies, Santa Clara, CA, USA). There totally contains 130 cases of GC tumor tissues and 126 cases of non-cancerous gastric mucosa in the datasets for comparison.
Dataset with 55 GC tissues and 35 non-cancerous tissues from the Cancer Genome Altas (TCGA) database was obtained and was intensively explored to further validate the expression profile of the candidate DEGs (https://tcga-data.nci.nih.gov/tcga/). R language, followed by normalization, was conducted by two professional bioinformatics analysts to preprocess the downloaded data, including background correction and transformation from probe level to gene symbol. DEGs differentiated between GC samples and non-cancerous tissues were defined basing on a t-test of linear models for microarray analysis package in R (Version 3.3, http://www.biocon ductor.org) (11). A threshold criteria of log 2 FC≥2.0 and P value 1.0E-04 was set for DEG selection on gene expression FC calculated in this study. Funrich Software (Version 3.0, http://funrich.org/index.html) was utilized to analyze the DEGs overlapping characteristic among the three datasets.
The online database of Cancer Cell Line Encyclopedia (CCLE: https://portals.broadinstitute.org) was introduced for determining the gene expression status of the candidates among differential tumor cell lines.
The main data collected and processed from GEO, TCGA, CCLE and dbDEMC database was listed (Suppl . Table. 1).

Cell culture and surgical specimens,
The immortalized gastric epithelium cell line (GES-1) and three GC cell lines (MKN-45, AGS and SNU-

Plasmid construction and transfection
MKN-45 cells in exponential phase were prepared and transfected with shRNA suppressing SPP1 mRNA translation through pGU6/Neo vectors (GenePharma, Shanghai, China) along with the construction of the control ones. Transfected cells were selected using medium mixed with G418 (Santa Cruz Biotechnology, Inc; 400 μg/ml).
Introduction of miR-4262 in MKN-45 cells (MKN-45/miR-4262) was carried out through mimic method similar to the description in our previous publication followed by the Dual-luciferase reporter assay (12,13). And the control ones were setup simultaneously (NigmiR).

Immunohistochemistry assay and the Western blot analysis
Antibodies against SPP1 and GAPDH (Santa Cruz, USA) were prepared, along with horseradish peroxidase-conjugated secondary antibody (Abcam, USA). Immunohistochemistry (IHC) assay was conducted on paired tissues from the patients. SPP1 antibody was utilized as the manufactory instruction described (1:50), IgG antibody was used as control. Samples treated were then exammed by two professional pathologists blindly.
RIPA buffer containing Protease Inhibitor Cocktail (Pierce, USA) was prepared lysing cells.
Concentration of the protein was measured \using BCA Protein Assay Kit (Pierce, USA). The extracted proteins were electrophoresed and electrotransfered. SPP1 antibody against (1:1000) and GAPDH (1:5000) were probed afterward. Horseradish peroxidase-conjugated secondary antibody was applied as further probe. GAPDH was regarded as the loading control.

Statistics Analysis
Results derived and generated from specimens and cells were analyzed by using SPSS 18.0. Paired t-8 test and Fisher's exact test were considered statistically significant as P values < 0.05.

Selection of the DEGs increased or decreased in GC
According to the criteria for DEGs selection of log 2 FC≥2.0 and P value 1.0E-04, 464 genes increased in GC tissues were set compared with non-cancerous specimens. And, 1136 genes extremely decreased in GC tissues were selected simultaneously.
Among these genes, 46 genes were filtered out via FunRich software, among which 13 genes were over-expressed in tumor tissues, while the other 33 ones were decreased. Respectively, the 13 high

SPP1 is highly expressed in GC tissues and cells, and is correlated with patients' survival status
According to the primary exploration of the GEO database, we screened SPP1 out of the 13 extremely highly expressed DGEs in GC tissues (P=2.58E-13 for GSE13911, P=8.32E-5 for GSE19826 and P=2.52E-10 for GSE27342) ( Fig. 2A C). In supplementary, we also analyzed the database of CCLE and observed the obvious inclination of high expression of SPP1 in different human malignancies including GC, and SPP1 presents a commonly incident up-regulation in multiple GC cell lines (Suppl. Fig. 2).
We detected the expression of SPP1 in three GC cell lines (MKN-45, AGS and SNU-16) through qRT-PCR assay and Western-blot analysis, and found SPP1was significantly increased in all the three GC cell lines compared with the control GES-1 cells (Fig. 2D, E) (P 0.05).
The expression of SPP1 in 105 paired of specimens from GC patients was detected by IHC assay. SPP1 was verified highly expressed in 74.28% tumor tissues (78/105); and only 24.76% (26/105) samples of non-cancerous tissues present a relatively higher expression. Hence, the expression of SPP1 in tumor tissues was frequently and significantly higher than adjacent non-cancerous tissues (P 0.01). (Fig. 3A,

B)
Additionally, Kaplan-Meier plot was generated to indicate the correlation between SPP1 and patents' over-all survival (OS) or progression free survival rate (PFS). We collected 876 GC cases' follow-up information, and the generated OS and PFS data indicated that the patients' are significantly impaired when SPP1 was highly expressed (OS: HR=2.23, P=2.7E-14; PFS: HR=1.98, P=1.1E-10) (Fig. 3C).

High expression of SPP1 is associated with GC clinicopathologic features
The correlation between SPP1 expression status and the clinicopathologic features of the 105 GC cases was analyzed. As Table. 1 shown, there is no significant correlation between OPN expression and the patients' age, gender or tumor location. On the contrary, high expression of SPP1 inclines toward larger tumor size (P 0.05), more frequent lymph node metastasis (P 0.05), deeper local invasion (P 0.05), and more advanced TNM stages (P 0.05) in cases with higher SPP1 expression level, which indicates a doubtless correlation with SPP1 over-expression and a part of the GC clinicopathologic features.

SPP1 depletion suppresses cell proliferation, arrests the cell cycle and promotes cell apoptosis in MKN-45 cells
MKN-45 cells was found expressing the highest level of SPP1 among the three GC cell lines, and was selected and transfected with pGU6/Neo vectors for SPP1 effect. The effect of transfection was validated through RT-qPCR and Western blot analysis (Fig. 4A, B).
CCK8 assay was applied for evaluating the effect of SPP1 on GC cell proliferation. A significant suppression of cell proliferation in SPP1 depleted MKN-45 cells compared with the control MKN-45 cells, as P value 0.05 for d1, and P value 0.01 for d2 to d4 (Fig. 4C). Likewise, the flow cytometry analysis for cell cycle demonstrated that the cell cycle of MKN-45 cells was arrested at G0/G1 phase via SPP1 depletion (Fig. 4D, E). The percentage of MKN-45 cells in G0/G1 phase was raised from 46.29% to 64.78% (P 0.01); Cell portion distributing at S phase was decreased from 32.69% to 25.98%; and for G2/M phase, was decreased from 21.02% to 12.25%. Whilst, we also noticed an increased apoptosis rate of MKN-45 treated with SPP1 depletion, from 7.92% to 14.12% (P 0.01) (Fig.   4F, G).
All these results above indicate that SPP1 depletion significant defects the tumor cell growth in GC.

SPP1 is a post-transcriptionally targeted by miR-4262 in GC cells
MiRNAs are short non-coding RNAs consist of around 22 nucleotides, inducing post-transcriptional regulation of transcribed genes. By using online bioinformatics and prediction tools described in methods parts above, we found miR-4262, a miRNA aberrantly expressed in multiple human malignancies, is potentially function as the upstreaming modulator of SPP1 (Fig. 5A). Online software of dbDEMC (Version 2.0) indicates the expression of miR-4262 in human malignancies, which presents miR-4262 as a remarkably down-regulated miRNA in GC (Fig. 5B). On this basis, we detected the expression of miR-4262 in the three GC cell lines, and found that the expression of miR-4262 was significantly decreased in GC cells compared with GES-1 cells (P 0.01) (Fig. 5C).
The direct interaction between SPP1 mRNA and miR-4262 was detected via dual-luciferase reporter assay. As Fig 5D shown (Fig. 5D). And the SPP1 expression at either mRNA status or protein status was significantly decreased in MKN-45/miR-4262 cells (Fig. 5E, F). These findings above suggest that SPP1 is one of the direct targets suppressed by miR-4262 in GC.

Re-upregulation of SPP1 in MKN-45 cells rescues the phenotype of cell growth and apoptosis induced by introducing miR-4262
In GC, miR-4262 had been verified significantly down-regulated in tumor tissues and cells, and plays a suppressive role in tumor progress, remarkably inhibiting the cell proliferation, and promotes the cell apoptosis (14).
Since we have validated the direct interaction between miR-4262 and SPP1 mRNA, we assumed that by re-introducing SPP1 into MKN-45 cells could rescue the phenotypes induced by miR-4262. In our study, we discovered that the suppressing function of miR-4262 of cell proliferation was significantly reversed by re-introducing SPP1 in MKN-45 cells (P value 0.05 for d1 d2, and P value 0.01 for d3 d4). Simultaneously, the cell cycle arrest and increase of cell apoptosis in MKN-45 cells were also partly rescued (Fig. 6).
Thus, we suggest that SPP1 is an enhancer of cell growth in GC, targeted by miR-4262.

Discussion
The tumorigenesis and progress of GC involves in multiple and complex mechanisms and molecules, such as mRNAs, noncoding RNAs and aberrant expressed proteins (15). Accumulating studies have reported a large amount of genes function in GC either promote or suppress the tumor process. As for the promotors, Cyclin Dependent Kinase 6 (CDK6) was reported significantly inducing cell invasion in GC (16); Plant homeodomain finger protein 10 (PHF10) promotes GC through enhancing cell proliferation (17); And, Solute carrier-7A5 (SLC7A5) and V-crk avian sarcoma virus CT10 oncogene homolog-like adapted protein of CRK family (CRKL) have been validated enhancing GC cell proliferation (18,19). On the contrary, other genes were found playing suppressive or inhibiting roles in GC, and these genes commonly present decreased expression profiles in tumor tissues. For instance, EPB41L3 is a suppressor of GC cell invasion and migration, and is negatively regulated by miR-223 (20); Runt Related Transcription Factor 3 (RUNX3) is involved with transcription activity to suppress GC process, and is inactivated in GC cells by protein mis-localization (21); Moreover, the miR-338 targets SSX2IP is one contributing to the suppression induction of GC tumorigenesis (22).

Data mining for individual gene information via the datasets of NCBI GEO and TCGA databases could
primarily give an integrative learning of the expression profiles and the relative functional network involved in GC process. In this study, we explored three GC patients' datasets of from the GEO database, which totally included 130 GC samples and 126 non-cancerous gastric mucosa specimens for intensive analysis. By using log 2 FC=2.0 as the cut-off value, we observed 464 genes significantly increased in GC, and 1136 ones remarkably decreased, among tens of thousands of genes. We suppose to detect teh probable promotors or suppressors of GC among these candidates. The overlapped results from three individual datasets gave out two cohorts, inw which present 13 genes abnormally amplified and 33 genes decreased in tumor tissues.We observed that SPP1 was significantly increased in within the tumor tissues, and literatures and the GO and KEGG pathway analysis indicated SPP1 as an pivotal genes linked to several critical pathways close to tumor process, including integrin-mediated signaling pathway, extracellular matrix organization, cell-matrix adhesion and ECM-receptor interaction. Thus, SPP1 may probably affect GC process in an enhancing way. SPP1 was originally reported involved in attachment of osteoclasts to the mineralized bone matrix (23). Evidence for the involvement of SPP1 in human malignancies indicates SPP1 as one of the promotors in multiple cancers. For instance, in colon cancer, SPP1 is highly expressed and significantly promotes tumor cell proliferation and angiogenesis (24); in non-small lung cancer, high expression of SPP1 induces tumor growth and enhances cell mobility (25). Recently, several studies have observed and suggested SPP1 as an indicator of the prognosis in GC (26)(27)(28). However, evidence of SPP1's promoting functions in GC is still limited.
In this study, our exploration indicated a common amplification of SPP1 expression from 105 paired real patients' tumor samples, and also three GC cell lines. Noticeably, survival analysis on either OS or PFS in GC patients demonstrated significant poor outcome of the patients with higher SPP1 expression. Meanwhile, we co-analyzed the SPP1 expression status and GC patients' clinicopathologic features. We discovered that high SPP1 expression is significantly correlated with some of the 13 parameters dismal in prognosis, especially the larger tumor size, higher local invasion degree, severer lymphnode metastasis and more advanced tumor stages. Considering the findings above, we further focused and detected the effect of SPP1 on GC tumors growth. As we supposed, the depletion of SPP1 significantly arrested the cell cycles of GC cells and impaired the ability of cell proliferation. And in accordance with these, SPP1 also induced significant cell apoptosis in MKN-45 cells. Thus we regard SPP1 as a biomarker indicating GC tumor progress and poor prognosis, enhancing the cell growth of tumor.
Specific cellular mechanisms are involved with the regulation of the signature genes or biomarkers in tumor. As acknowledged, miRNAs, which are transcripts with non-coding characteristics composed with around 22 nucleotides, are a kind of upstreaming regulators involves in tumor process suppressing or degenerating target mRNAs (29,30). We supposed post-transcriptional regulation upstreaming SPP1 plays critical role in GC process.
By using bioinformatics prediction, we found miR-4262 is a potential upstreaming probably binding to the 3'UTR of SPP1 mRNA. According to the reports and literatures, miR-4262 is a yet validated noncoding molecule participating in multiple processes either promoting or suppressing tumorigenesis.
For example, in hepatocellular carcinoma, miR-4262 negatively regulates PDCD4 and then induces activation of NF-κB pathway, causing tumor cell proliferation (31); In acute myeloid leukemia, high expression of miR-4262 is found in both bone marrow and patients' serum, which is associated with short survival period and poor prognosis (32). On the contrary, miR-4262 acts as an important suppressor in cervical cancer by directly degenerating mRNA of ZETB33, and impacts cell proliferation and EMT process (33). Recent report has pointed out that miR-4262 expression is significantly decreased in GC cells, and exerts suppressive effects cell growth by targeting proto-oncogene CD163, along with the promotion of cell apoptosis (14).
In our study, the significant down-regulation of miR-4262 in GC cells through qRT-PCR assay was

Consent for publication
The individual personal data in this study could be used for publication with the consents.

Availability of data and material
All data obtained in this study is included in this published article and its supplementary materials.

Funding
This study was kindly supported by grants from the following: National Natural Science Foundation of  DEGs selection from NCBI GEO and TCGA database A. Venn chart of the increased genes in GSE13911, GSE19826 and GSE27342 datasets. Thirteen DEGs over-expressed in the three datasets were selected. B. Venn chart of the decreased genes in the NCBI GEO datasets above. Thirty-three DEGs were selected from the three datasets. C. Heatmap generated from TCGA database for GC. The bar indicates the relative expression level of separate genes in differential samples, from blue to red, which indicates low expression of gene mRNAs to high expressed ones.