Microarray technology reveals potentially novel genes and pathways involved in non-functioning pituitary adenomas

As a kind of benign adenomas in the pituitary gland, clinically non-functioning pituitary adenomas (NFPAs) are the most common type of pituitary macroadenomas in adults. The NFPAs account for about 34.0% [1] of all pituitary adenomas (PAs) that occur at a prevalence rate of 75-94 per 100,000 [1,2]. Patients with NFPAs generally suffer from headaches, hypopituitarism, hypogonadism and visual field defects. Late diagnosis due to inconspicuous signs and symptoms, extension to the cavernous sinus and sellar floor, resistance to pharmacological therapy and high recurrence rate, make their treatment disappointing and challenging [3]. Approximately 80.0% of NFPAs originate from gonadotroph cells (gonadotroph pituitary adenoma, GnPA) [4], and other NFPAs are mainly associated with null cells (null cell pituitary adenoma, ncPA). The identification of novel therapeutic targets for human NFPAs depend on a good understanding of the molecular mechanism of NFPAs [5].

Progression in understanding the mechanism of PAs, especially NFPAs, has been achieved over the last several years. According to the reports, germline mutations in AIP or MEN1 genes are associated with young age-onset PAs [6,7]. The HGF and c-MET genes are frequently expressed in PAs, and their expressions are correlated with phos-phorylated Akt expression [8]. Durán-Prado et al. [9] identified that sst5TMD4, a truncated variant of somatostatin receptor 5, appeared in 85.0% PAs rather than normal pituitary, and it may play an inhibitory role in PAs that possess poor response to somatostatin analogs. Raf/ MEK/ ERK and PI3K/Akt/mTOR signaling pathways are perturbed in NFPAs [10]. As a target of the SF1 gene in gonadotroph cells, CYP11A1 is up-regulated in human GnPA, and Cyp11a1 promotes survival and proliferation of primary cells and cell lines of rat PAs [5]. Rotondi et al. [11] suggested that the gonadotroph phenotype was strongly associated with AIP expression in NFPAs. The AIP level is higher in GnPA than that in ncPA, and both AIP and cyclinD1 levels are high in most NFPAs. The AIP level correlates with follicle-stimulating hormone β (FSHβ) and cyclinD1 levels in GnPA. However, AIP is not involved in the aggressiveness of NFPAs [11]. Recently, CCNB1 was found to mediate the proliferation-inhibiting role of miR-410, a small non-coding RNA, in GnPA [12]. Additionally, Chesnokova et al. [13] have identified that human pituitary tumors originated from gonadotroph cells express abundant FOXL2, and both FOXL2 and PTTG promote cluster- ing expression and secretion from gonadotroph cells, thus restraining the proliferation of pituitary cells.

Along with the development of microarray, transcriptome analysis has been widely utilized in understanding tumor mechanism. Based on the gene expression microarray dataset GSE26966, Michaelis et al. [14] identified that GADD45β, a downstream effector of p53, is a tumor suppressor in gonadotroph tumor. Its overexpression in mouse gonadotroph cells blocks cell proliferation and promotes apoptosis [14]. Based on the same dataset, Cai et al. [15] identified the coexpressed and altered genes involved in gonadotroph tumors and suggest that ITGA4, MPP2, DLK1, CDKN2A and ASAP2 might be biomarkers. However, pathways or functions of the altered genes were not studied by Michaelis et al. [14], and the protein-protein interactions (PPIs) between genes were not investigated in the two aforementioned studies [14,15]. In particular, Zhao et al. [16] performed an integrated analysis of five available microarray datasets of various PAs, to detect 3994 differentially expressed genes (DEGs) (including 2043 up- and 1951 down-regulated genes), and conducted a PPI network analysis. However, PPIs of more DEGs are needed to be analyzed, and more potential novel PAs-related genes are still unknown. Moreover, molecular mechanisms underlying the pathogenesis of PAs, particular NFPAs, remain unclear, and it is still essential to comprehensively investigate and annotate the alterations in gene expression profiles. In the present study, NFPAs-related microarray data uploaded by Michaelis et al. [14] were analyzed to identify significant DEGs, study NFPAs-related functions and pathways, construct interaction network, and identify potential novel NFPAs-related genes.

Non-functioning pituitary adenomas comprise about 34.0% of pituitary tumors, while their molecular mechanism is still incompletely understood [5]. In the current study, we comprehensively analyzed the gene expression profile of NFPAs and healthy pituitary glands. As a result, 604 DEGs were identified between NFPAs and controls, including 177 up- and 427 down-regulated genes, which were much less than those identified by Michaelis et al. [14]. However, in the current study, we analyzed the same microarray data using different software, algorithms, and analysis criteria (corrected p value <0.05 and [log₂ FC] >2) in order to focus on the DEGs that were more significant.

In the current study, mean FC of the up-regulated genes was 6.6, and mean FC of the down-regulated genes was –19.2, which were different from those in the previous study by Michaelis et al. [14] (4.5 and –32.2, respectively). The differences of mean FC values might be caused by the different DEG sets in the two studies [14]. The major DEGs found by Michaelis et al. [14] had similar expression change patterns in the current study, e.g. for the PLAGL1, CDKN1A, RPRM, PMAIP1, MDM2, GADD45A, GAD-D45B and GADD45G genes.

Of the top DEGs, DLK1, GH1, CDKN2A and MEG3 were significantly down-regulated in NFPAs in comparison with normal pituitary glands in this study. According to the report, the MEG3 and DLK1-MEG3 locus are silenced in human NFPAs of gonadotroph origin, and DLK1-MEG3 locus plays a tumor suppressor role in NFPAs [30]. Based on proteome data and microarray data or reverse transcription quantitative real-time polymerase chain reaction analysis, Moreno et al. [31] found that DLK1, GH1 and PRL are down-regulated in NFPAs when compared with normal pituitary glands, whereas IDH1 is significantly up-regulated. The CDKN2A and DLK1 are considered as biomarkers of gonadotroph tumors by Cai et al. [15], and gene silencing mediated by hypermethylation of the CpG island within exon 1 in CDKN2A is associated with NFPAs [32]. As clearly shown in Figure 7, the expression change patterns of known disease genes DLK1, GH1, PRL, CDKN2A and IDH1, were consistent with the aforementioned studies [30-32], demonstrating the high accuracy of our results.

Expressions of EGFR in NFPAs varied in different studies [10,26-28]. In the current study, EGFR showed low expression in NFPAs (Figure 7), and it interacted with known disease gene CDKN2A, indicating that low expression of EGFR might be associated with NFPAs. We also found that CDKN2A was a top DEG, and it interacted with 22 DEGs in the whole PPI network and most DEGs in the PPI sub-network, suggesting that CDKN2A might play a crucial role in the progression of NFPAs.

Furthermore, potential novel genes were identified (Figure 6), especially COL4A5, LHX3, MSN and GHSR. The role of these genes in NFPAs has not been investigated by previous studies. According to the report, mRNA level of GHSR in NFPAs is lower than that in growth hormone-producing PAs [33]. In the present study, COL4A5, LHX3, MSN and GHSR were significantly down-regulated in NF-PAs in comparison with normal controls, and they interacted with known NFPA-related genes such as EGFR, PRL, and GH1. These results indicated that COL4A5, LHX3, MSN and GHSR might participate in the initiation and progression of NFPAs via interaction with EGFR, PRL and GH1, respectively.

We found DEGs were significantly enriched in the p53 (Figure 3) and Jak-STAT signaling pathways (Figure 5), which had been reported to take part in PAs pathogenesis [8,24]. The p53 signaling pathway is involved in biological processes such as cell cycle arrest, apoptosis, senescence, DNA repair and changes in metabolism. Expression level of p53 correlates with the proliferative state of PAs [24]. The Jak-STAT pathway is an important downstream pathway for growth factor receptors and cytokine receptors, and it is involved in the regulation of cell proliferation and survival [34,35]. As all of the DEGs mapped on these pathways were remarkably down-regulated in NFPAs, p53 and Jak-STAT signaling pathways might play roles in the progression of NFPAs.

In addition, DEGs were significantly enriched in GO terms mainly about cell communication, signaling, ECM, plasma membrane, collagen, transcription factor activity and receptor binding (Table 2). The ECM, plasma membrane, and receptor binding are the basis of cell communication and signaling between pituitary cells, which play crucial roles in the development and invasion of PAs [36, 37]. As DEGs mapped on these GO terms were remarkably dysregulated in NFPAs, cell communication and signaling might contribute to the progression of NFPAs.

In conclusion, a number of genes (e.g. COL4A5, LHX3, MSN and GHSR) identified in this study, might be potential novel NFPA-related genes. Furthermore, cell communication and signaling pathways (e.g. p53 and JakSTAT) might be implicated in the pathogenesis of NFPAs. Currently, no effective medical therapies are available for NFPAs, due to their unclear mechanism. Although further validation is required, our findings might provide information to guide future researchers and even benefit the development of medical therapy for NFPAs.