Introduction
Prolactinoma is a benign tumor that secretes too much prolactin. It is also one of the most adult pituitary tumor and appears in pituitary gland. It often happens in young women with 20-30 years old. The tumor often leads to amenorrhea, galactorrhea, loss of axillary and pubic hair, hypogonadism, gynecomastia and erectile dysfunction. Finally the tumor can result in the cessation of growth. Till now, the main way for the treatment of prolactinoma is dopamine, and it can decrease prolactin secretion (Asa and Ezzat, 1998).

Although there are many researches for the analysis of prolactinoma, the pathogenesis is still unclear. The microarray technique can analyze the differential genes at expression level more accurately (Elston et al., 2008; Evans et al., 2003). It can identify many tumor-related genes for the further analysis of cancers. In this study, we recognize many DEGs and biological process which are correlated with prolactinoma and can provide some information about the mechanism for tumors. Furthermore we can find new biomarkers for the treatment of prolactinoma (Ramasamy et al., 2008).

1. Materials and Methods
1.1 Data
The microarray data we use are downloaded from gene expression omnibus (GEO) with the number GSE36314. There are four tumor samples and three normal samples for our analysis. The tumor samples are available during trans-sphenoidal surgery, and the normal samples are available from dead individuals. All samples are analyzed through the platform Affymetrix Human Genome U95 Version 2 Array (GPL8300) [HG_U95Av2].

1.2 Microarray analysis
The format of data we download from GEO is CEL. We apply the method “RMA” in the package “Affy” to preprocess the data (Guo et al., 2014). The preprocessing includes background adjustment, quantile normalization, finally summarization, log 2 and so on. Next we use perl to delete the arrays with “NA”. Coefficient of variation (CV) is used to assess the discrete degree of all data. The genes with CV less than 0.5 in more than 80% samples are remained. CV=SD/AVE. CV is the coefficient of variation of arrays belonging to the same gene in one sample, SD is the variance of arrays, AVE is the average of sites. Average gene expression level of multiple arrays on the same gene is defined as the gene expression level of gene. Then DEGs are identified by package “samr” with FDR<=0.05 (Tusher et al., 2001; Li and Tibshirani, 2013). The FDR is calculated based on Benjamini & Hochberg. At last we distinguish the DEGs into up-regulated genes and down-regulated genes (Wu et al., 2014; Zhang et al., 2015).

1.3 Hierarchical clustering analysis
Bidirectional hierarchical clustering analysis is a computational method that is always used to explore the relationship of samples. We use Genepattern (http://www.broadinstitute.org/cancer/software/genepattern/) and calculate Euclidean distance to explore whether the similar samples exist strong relationship. We use DEGs to make the bidirectional hierarchical clustering analysis and exhibit the result through Genepattern(Varley et al., 2013).

1.4 Functional enrichment analysis
In order to identify the function of DEGs, we use DAVID (http://david.abcc.ncif crf .gov/) to explore whether DEGs are enriched in Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) with the FDR 0.05(2008; Okuda et al., 2008).

1.5 The construction of protein-protein interaction network
The protein-protein interaction network (PPIN) not only shows the relationship among different factors, but also show the function of these factors. In order to identify the function of DEGs, PPI sub-network is built by published PPI network. There are a lot of databases storing the interactions of genes, including the Biomolecular Interaction Network Database (BIND), the Biological General Repository for Interaction Data sets (BioGRID), the Database of Interacting Proteins (DIP), the Human Protein Reference Database (HPRD), IntAct, the Molecular IN Teraction database (MINT), the mammalian PPI database of the Munich Information Center on Protein Sequences (MIPS), PDZBase (a PPI database for PDZ-domains) and Reactome. The background network includes 80,980 edges and 13,361 nodes. We regard red nodes as up-regulated genes, yellow nodes as down-regulated genes and grey nodes as other nodes(Isserlin et al., 2011; Stark et al., 2011; Salwinski et al., 2004; Peri et al., 2003; Aranda et al., 2010; Licata et al., 2012; Pagel et al., 2005; Beuming et al., 2005; Yu et al., 2012; Canturk et al., 2014; Wang et al., 2014).

2 Results
2.1 The identification of differentially expressed genes
Four tumor samples and three normal samples are used to compare in this study. After preprocessing, 12625 arrays are obtained. Then we calculate the average of arrays and obtain 8305 genes for the further analysis. Then we use the package “samr” to calculate the DEGs and obtain 86 DEGs with the FDR less than 0.05. We distinguish DEGs into 35 up-regulated genes and 51 down-regulated genes.

2.2 The hierarchical clustering analysis
To measure the similarity of samples, we use hierarchical clustering analysis to identify the relationship of tumor samples and normal samples (Figure 1). Four tumor samples are clustered together and three normal samples are clustered together. The result shows DEGs can distinguish the tumor samples and normal samples significantly. This means the pattern in tumor samples or normal samples exists the similarity and the robustness. This result provides a possibility for our further analysis.

2.3 The enrichment analysis of DEGs
It is very easy to identify the potential functions and kegg pathways which DEGs are enriched in by DAVID (Figure 2). Firstly, we analyze the Biological Process (BP) which DEGs are enriched in. We find that DEGs are enriched in 33 GO terms. The GO terms are mainly involved in response to hormone stimulus, regulation of hormone levels, response to endogenous stimulus, male gonad development gonad development, development of primary male sexual characteristics and so on. The GO terms show the relationship between prolactinoma and hormone. Prolactinoma can influence the sexual development of male and female and result in maldevelopment.

Next, we analyze Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation and DEGs are enriched in 2 KEGG pathways (Table 1). We recognize that DEGs are mainly enriched in Neuroactive ligand -receptor interaction and Pathogenic Escherichia coli infection. Neuroactive ligand-receptor interaction is potentially correlated with nervous system. The neuroactive ligand-receptor interaction pathway, which is a collection of neuroactive receptors in the plasma membranes, is involved in the stability of the neuroendocrine system. This shows that neuroactive ligand-receptor interaction pathway may be associated with the development of prolactinoma (Figure 3).

2.4 The construction of PPI network
In order to analyze the function of DEGs, we input them into PPI network and construct a sub-PPI network. Then we use Cytoscape (http://www.cytoscape.org/) for the visualization of sub-PPI network. The network includes 1151 nodes and 1131 edges, and there are 70 DEGs which exists in sub-PPI network (Figure 4). In order to analyze the relationship of genes, we use MCODE to extract significant modules. We identify 16 motifs and choose the most significant one to analyze the functions. The motif includes 5 genes such as GAL, GALR1, GALR2, GALR3 and GPR151 (Yang et al., 2014).

Hub gene always play an important role in network, and we regard genes with degree more than 80 as hub genes, including MDM2, LYN, CDH1, GH1, ACTG1 and FUS. We find these genes are all DEGs, which proves the important role of DEGs in network.

Moreover, we input DEGs into miRNA-RNA network and construct a sub network, which includes 5 genes, 5 miRNAs and 6 edges. Mir124 and Mir21 are also associated with prolactinoma.

3 Discussion
In this study, we use GEO data to identify DEGs in prolactinoma by hierarchical clustering analysis, enrichment analysis and PPI network, and analyze the relationship between DEGs and prolactinoma.

The result of hierarchical clustering demonstrates that the gene expression level between tumor and normal samples has stable gene expession signatures, which is coincident with previous studies (2012). And the DEGs are mainly enriched in hormone-related biological process and many development of sex. Prolactinoma is caused by the hypersecretion of prolactin, which always happens in hypophysis and hypothalamus. The clinical manifestation is amenorrhea, galactorrhea and hyperprolactinemia. For young female, it can lead to delayed puberty and developmental retardation. For young male, it can lead to developmental retardation and sexual dysfunction. Besides it also can result in vision disorder, headache and osteoporosis. The enrichment of DEGs indicates that most DEGs are correlated with hormone, which are related to development of sex. The result means prolactinoma plays an important role in development of sex.

DEGs are also enriched in Neuroactive ligand- receptor interaction, which are mainly involved in osteoporosis, sexual precocity and ovarian hy -perstimulation syndrome. It means that this kegg pathway has strong correlation with prolactinoma (Decker et al., 2008; Frimberger and Gearhart, 2005; Bertelloni and Mul, 2008; Kasum, 2010).

In PPI network, MDM2, LYN, CDH1, GH1, ACTG1 and FUS have high degrees and are regarded as hub genes. It demonstrates that these genes are important in interaction and biological process. MDM2 can encode an ubiquitinligating enzyme which can regulate tumor suppressor gene and promote the progress of tumors(Huang et al., 2014). Besides, GH1 is a member of prolactin family, and the abnormity of this gene can influence the secretion of prolactin, which can inhibit the growth and development of sex(Vakili et al., 2014). In conclusion, these genes play important roles in prolactinoma.

We recognize many motifs and choose a most significant motif for the analysis. This motif includes GAL, GALR1, GALR2, GALR3 and GPR151. GALR1 and GALR2 are the members of galanin family. Galanin can influence nervous system and endocrine system and the abnormity of these genes can destroy the growth and development of mammals(Wynick et al., 1998). Besides they can also effect the secretion of prolactinoma. Therefore galanin can be regarded as the biomarker of curing prolactinoma. We also identify mir124 and mir21 which have strong correlation with nervous system. The differential expression of these miRNAs can regulate many nervous system diseases and prolactinoma is one of nervous system disease (Guo et al., 2009; Chen et al., 2004). Furthermore they are also associated with cell differentiation, propagation and apoptosis.

At present, the therapeutic method is mainly focused on medicine and the medicine is dopamine agonist. However it may lead to many side-effects. And the operation can influence the function of visual system and hypothalamo-hypophyseal system. Therefore it is very important to identify tumor-related biomarkers for the treatment of prolactinoma.

Acknowledgments
This work was supported by the Science Innovation Project (grants 2015003) and the Innovation and Technology special Fund for excellent academic leader of Harbin (grant number 2015RAXYJ051).

References
Aranda B., Achuthan P., Alam-Faruque Y., Armean I., Bridge A., Derow C., Feuermann M., Ghanbarian A.T., Kerrien S., Khadake J., Kerssemakers J., Leroy C., Menden M., Michaut M., Montecchi-Palazzi L., Neuhauser S.N., Orchard S., Perreau V., Roechert B., Van Eijk K., and Hermjakob H., 2010, The IntAct molecular interaction database in 2010, Nucleic Acids Res, 38: D525-531
http://dx.doi.org/10.1093/nar/gkp878

Asa S.L., and Ezzat S., 1998, The cytogenesis and pathogenesis of pituitary adenomas, Endocr Rev, 19: 798-827
http://dx.doi.org/10.1210/er.19.6.798

Bertelloni S., and Mul D., 2008, Treatment of central precocious puberty by GnRH analogs: long-term outcome in men, Asian J Androl, 10: 525-534
http://dx.doi.org/10.1111/j.1745-7262.2008.00409.x

Beuming T., Skrabanek L., Niv M.Y., Mukherjee P., and Weinstein H., 2005, PDZBase: a protein-protein interaction database for PDZ-domains, Bioinformatics, 21: 827-828
http://dx.doi.org/10.1093/bioinformatics/bti098

Canturk K.M., Ozdemir M., Can C., Oner S., Emre R., Aslan H., Cilingir O., Ciftci E., Celayir F.M., Aldemir O., Ozen M., and Artan S., 2014, Investigation of key miRNAs and target genes in bladder cancer using miRNA profiling and bioinformatic tools, Mol Biol Rep,

Chen C.Z., Li L., Lodish H.F., and Bartel D.P., 2004, MicroRNAs modulate hematopoietic lineage differentiation, Science, 303: 83-86

Comprehensive molecular characterization of human colon and rectal cancer, 2012, Nature, 487: 330-337

Decker E., Stellzig-Eisenhauer A., Fiebig B.S., Rau C., Kress W., Saar K., Ruschendorf F., Hubner N., Grimm T., and Weber B.H., 2008, PTHR1 loss-of-function mutations in familial, nonsyndromic primary failure of tooth eruption, Am J Hum Genet, 83: 781-786
http://dx.doi.org/10.1016/j.ajhg.2008.11.006

Elston M.S., Gill A.J., Conaglen J.V., Clarkson A., Shaw J.M., Law A.J., Cook R.J., Little N.S., Clifton-Bligh R.J., Robinson B.G., and Mcdonald K.L., 2008, Wnt pathway inhibitors are strongly down-regulated in pituitary tumors, Endocrinology, 149: 1235-1242
http://dx.doi.org/10.1210/en.2007-0542

Evans C.O., Reddy P., Brat D.J., O'neill E.B., Craige B., Stevens V.L., and Oyesiku N.M., 2003, Differential expression of folate receptor in pituitary adenomas, Cancer Res, 63: 4218-4224

Frimberger D., and Gearhart J.P., 2005, Ambiguous genitalia and intersex, Urol Int, 75: 291-297
http://dx.doi.org/10.1159/000089160

Guo L., Sun B., Sang F., Wang W., and Lu Z., 2009, Haplotype distribution and evolutionary pattern of miR-17 and miR-124 families based on population analysis, PLoS One, 4: e7944
http://dx.doi.org/10.1371/journal.pone.0007944

Guo Z., Zhao C., and Wang Z., 2014, Gene expression profiles analysis identifies key genes for acute lung injury in patients with sepsis, Diagn Pathol, 9: 176
http://dx.doi.org/10.1186/s13000-014-0176-x

Huang L., Wong C.C., Cheng K.W., and Rigas B., 2014, Phospho-Aspirin-2 (MDC-22) Inhibits Estrogen Receptor Positive Breast Cancer Growth Both In Vitro and In Vivo by a Redox-Dependent Effect, PLoS One, 9: e111720
http://dx.doi.org/10.1371/journal.pone.0111720

Isserlin R., El-Badrawi R.A., and Bader G.D., 2011, The Biomolecular Interaction Network Database in PSI-MI 2.5, Database (Oxford), 2011: baq037
http://dx.doi.org/10.1093/database/baq037

Kasum M., 2010, New insights in mechanisms for development of ovarian hyperstimulation syndrome, Coll Antropol, 34: 1139-1143

Li J., and Tibshirani R., 2013, Finding consistent patterns: a nonparametric approach for identifying differential expression in RNA-Seq data, Stat Methods Med Res, 22: 519-536
http://dx.doi.org/10.1177/0962280211428386

Licata L., Briganti L., Peluso D., Perfetto L., Iannuccelli M., Galeota E., Sacco F., Palma A., Nardozza A.P., Santonico E., Castagnoli L., and Cesareni G., 2012, MINT, the molecular interaction database: 2012 update, Nucleic Acids Res, 40: D857-861
http://dx.doi.org/10.1093/nar/gkr930

Okuda S., Yamada T., Hamajima M., Itoh M., Katayama T., Bork P., Goto S., and Kanehisa M., 2008, KEGG Atlas mapping for global analysis of metabolic pathways, Nucleic Acids Res, 36: W423-426
http://dx.doi.org/10.1093/nar/gkn282

Pagel P., Kovac S., Oesterheld M., Brauner B., Dunger-Kaltenbach I., Frishman G., Montrone C., Mark P., Stumpflen V., Mewes H.W., Ruepp A., and Frishman D., 2005, The MIPS mammalian protein-protein interaction database, Bioinformatics, 21: 832-834
http://dx.doi.org/10.1093/bioinformatics/bti115

Peri S., Navarro J.D., Amanchy R., Kristiansen T.Z., Jonnalagadda C.K., Surendranath V., Niranjan V., Muthusamy B., Gandhi T.K., Gronborg M., Ibarrola N., Deshpande N., Shanker K., Shivashankar H.N., Rashmi B.P., Ramya M.A., Zhao Z., Chandrika K.N., Padma N., Harsha H.C., Yatish A.J., Kavitha M.P., Menezes M., Choudhury D.R., Suresh S., Ghosh N., Saravana R., Chandran S., Krishna S., Joy M., Anand S.K., Madavan V., Joseph A., Wong G.W., Schiemann W.P., Constantinescu S.N., Huang L., Khosravi-Far R., Steen H., Tewari M., Ghaffari S., Blobe G.C., Dang C.V., Garcia J.G., Pevsner J., Jensen O.N., Roepstorff P., Deshpande K.S., Chinnaiyan A.M., Hamosh A., Chakravarti A., and Pandey A., 2003, Development of human protein reference database as an initial platform for approaching systems biology in humans, Genome Res, 13: 2363-2371
http://dx.doi.org/10.1101/gr.1680803

Ramasamy A., Mondry A., Holmes C.C., and Altman D.G., 2008, Key issues in conducting a meta-analysis of gene expression microarray datasets, PLoS Med, 5: e184
http://dx.doi.org/10.1371/journal.pmed.0050184

Salwinski L., Miller C.S., Smith A.J., Pettit F.K., Bowie J.U., and Eisenberg D., 2004, The Database of Interacting Proteins: 2004 update, Nucleic Acids Res, 32: D449-451
http://dx.doi.org/10.1093/nar/gkh086

Stark C., Breitkreutz B.J., Chatr-Aryamontri A., Boucher L., Oughtred R., Livstone M.S., Nixon J., Van Auken K., Wang X., Shi X., Reguly T., Rust J.M., Winter A., Dolinski K., and Tyers M., 2011, The BioGRID Interaction Database: 2011 update, Nucleic Acids Res, 39: D698-704
http://dx.doi.org/10.1093/nar/gkq1116

The Gene Ontology project in 2008, 2008, Nucleic Acids Res, 36: D440-444

Tusher V.G., Tibshirani R., and Chu G., 2001, Significance analysis of microarrays applied to the ionizing radiation response, Proc Natl Acad Sci U S A, 98: 5116-5121
http://dx.doi.org/10.1073/pnas.091062498

Vakili H., Jin Y., and Cattini P.A., 2014, Energy homeostasis targets chromosomal reconfiguration of the human GH1 locus, J Clin Invest, 124: 5002-5012
http://dx.doi.org/10.1172/JCI77126

Varley K.E., Gertz J., Bowling K.M., Parker S.L., Reddy T.E., Pauli-Behn F., Cross M.K., Williams B.A., Stamatoyannopoulos J.A., Crawford G.E., Absher D.M., Wold B.J., and Myers R.M., 2013, Dynamic DNA methylation across diverse human cell lines and tissues, Genome Res, 23: 555-567
http://dx.doi.org/10.1101/gr.147942.112

Wang L., Che X.J., Wang N., Li J., and Zhu M.H., 2014, Regulatory Network Analysis of MicroRNAs and Genes in Neuroblastoma, Asian Pac J Cancer Prev, 15: 7645-7652
http://dx.doi.org/10.7314/APJCP.2014.15.18.7645

Wu M.C., Joubert B.R., Kuan P.F., Haberg S.E., Nystad W., Peddada S.D., and London S.J., 2014, A systematic assessment of normalization approaches for the Infinium 450K methylation platform, Epigenetics, 9: 318-329
http://dx.doi.org/10.4161/epi.27119

Wynick D., Small C.J., Bloom S.R., and Pachnis V., 1998, Targeted disruption of the murine galanin gene, Ann N Y Acad Sci, 863: 22-47
http://dx.doi.org/10.1111/j.1749-6632.1998.tb10681.x

Yang Z., Yu F., Lin H., and Wang J., 2014, Integrating PPI datasets with the PPI data from biomedical literature for protein complex detection, BMC Med Genomics, 7 Suppl 2: S3
http://dx.doi.org/10.1186/1755-8794-7-S2-S3

Yu X., Wallqvist A., and Reifman J., 2012, Inferring high-confidence human protein-protein interactions, BMC Bioinformatics, 13: 79
http://dx.doi.org/10.1186/1471-2105-13-79

Zhang C., Zhao H., Li J., Liu H., Wang F., Wei Y., Su J., Zhang D., Liu T., and Zhang Y., 2015, The identification of specific methylation patterns across different cancers, PLoS One, 10: e0120361
http://dx.doi.org/10.1371/journal.pone.0120361