Regulation of Aryl Hydrocarbon Receptor Interacting Protein (AIP) Protein Expression by MiR-34a in Sporadic Somatotropinomas

Introduction Patients with germline AIP mutations or low AIP protein expression have large, invasive somatotroph adenomas and poor response to somatostatin analogues (SSA). Methods To study the mechanism of low AIP protein expression 31 sporadic somatotropinomas with low (n = 13) or high (n = 18) AIP protein expression were analyzed for expression of AIP messenger RNA (mRNA) and 11 microRNAs (miRNAs) predicted to bind the 3’UTR of AIP. Luciferase reporter assays of wild-type and deletion constructs of AIP-3’UTR were used to study the effect of the selected miRNAs in GH3 cells. Endogenous AIP protein and mRNA levels were measured after miRNA over- and underexpression in HEK293 and GH3 cells. Results No significant difference was observed in AIP mRNA expression between tumors with low or high AIP protein expression suggesting post-transcriptional regulation. miR-34a was highly expressed in low AIP protein samples compared high AIP protein adenomas and miR-34a levels were inversely correlated with response to SSA therapy. miR-34a inhibited the luciferase-AIP-3’UTR construct, suggesting that miR-34a binds to AIP-3’UTR. Deletion mutants of the 3 different predicted binding sites in AIP-3’UTR identified the c.*6–30 site to be involved in miR-34a’s activity. miR-34a overexpression in HEK293 and GH3 cells resulted in inhibition of endogenous AIP protein expression. Conclusion Low AIP protein expression is associated with high miR-34a expression. miR-34a can down-regulate AIP-protein but not RNA expression in vitro. miR-34a is a negative regulator of AIP-protein expression and could be responsible for the low AIP expression observed in somatotropinomas with an invasive phenotype and resistance to SSA.


Introduction
Patients with germline AIP mutations or low AIP protein expression have large, invasive somatotroph adenomas and poor response to somatostatin analogues (SSA).

Methods
To study the mechanism of low AIP protein expression 31 sporadic somatotropinomas with low (n = 13) or high (n = 18) AIP protein expression were analyzed for expression of AIP messenger RNA (mRNA) and 11 microRNAs (miRNAs) predicted to bind the 3'UTR of AIP. Luciferase reporter assays of wild-type and deletion constructs of AIP-3'UTR were used to study the effect of the selected miRNAs in GH3 cells. Endogenous AIP protein and mRNA levels were measured after miRNA over-and underexpression in HEK293 and GH3 cells.

Results
No significant difference was observed in AIP mRNA expression between tumors with low or high AIP protein expression suggesting post-transcriptional regulation. miR-34a was highly expressed in low AIP protein samples compared high AIP protein adenomas and miR-34a levels were inversely correlated with response to SSA therapy. miR-34a inhibited the luciferase-AIP-3'UTR construct, suggesting that miR-34a binds to AIP-3'UTR. Deletion mutants of the 3 different predicted binding sites in AIP-3'UTR identified the c.*6-30 site to be involved in miR-34a's activity. miR-34a overexpression in HEK293 and GH3 cells resulted in inhibition of endogenous AIP protein expression.

Introduction
Germline mutations have been described in the aryl hydrocarbon receptor interacting protein (AIP) gene in the setting of either familial isolated pituitary adenoma (FIPA) or in simplex, young-onset, predominantly growth hormone (GH)-and/or prolactin-secreting pituitary adenomas [1,2]. AIP is a tumor suppressor gene and patients with acromegaly harboring germline mutations in this gene have more invasive tumors showing sparsely-granulated pattern and are less likely to respond to treatment with somatostatin analogues (SSA) [3][4][5].
Although no somatic AIP mutation has been described in pituitary adenomas to date [4,6], approximately half of the sporadic somatotropinomas present with low intrinsic AIP expression, and, similar to the mutated tumors, most of them are invasive and respond poorly to SSA therapy [7][8][9]. Therefore, low AIP expression seems to be important in determining the pathological characteristics of somatotropinomas.
MicroRNAs (miRNAs) are small non-coding RNAs with an important role in post-transcriptional regulation of mRNA and/or protein expression. miRNAs were found to be involved in the pathogenesis of many human diseases, including pituitary adenoma [10][11][12][13][14]. We hypothesized that miRNA regulation of AIP protein expression could be responsible for the low AIP levels found in approximately half of the sporadic somatotropinomas [8,9].

Subjects and Methods Patients
Thirty-four consecutive patients with acromegaly who had previously had pituitary surgery and had tissue available (paraffin block and fresh frozen tumor sample) were included in the study. This study was approved by the Ethics Committees of the Clementino Fraga Filho University Hospital/Medical School, Federal University of Rio de Janeiro and the Clinics Hospital, Ribeirão Preto Medical School, São Paulo University. All subjects gave written informed consent before study entry. Patients underwent pituitary surgery between 2006 and 2011. Biochemical diagnosis of acromegaly was based on international criteria [15,16]. Exclusion criteria included previous known AIP mutations, a family history of pituitary adenoma, presence of features or family history of Carney complex or multiple endocrine neoplasia type 1 or 4 and preoperative therapy with SSA [as treatment may increase AIP expression [17]]. Tumor invasiveness was determined according to Knosp-Steiner criteria [18]. GH-secreting pituitary tumor samples were obtained during transsphenoidal surgery: part of the sample was processed for routine histopathological and immunohistochemical studies (including anterior pituitary hormones), and part was snap-frozen and stored at -70°C for molecular biology studies. All samples were micro-dissected by an experienced pathologist in order to separate any non-tumoral tissue and homogenized using a Polytron homogenizer. In addition, five normal human pituitaries were obtained within 10 hours from the time of death at autopsies of subjects who had died from natural causes without previous evidence of any endocrine disease or pituitary abnormality.

Postsurgical evaluation
Biochemical assessment was performed 12 weeks after surgery by evaluation of oral glucose tolerance test (OGTT) and serum insulin-like growth factor I (IGF-I) levels in all subjects. Pituitary magnetic resonance imaging (MRI) was performed 3 months after the surgical procedure. Patients were considered as non-cured on the basis of the clinical picture, nadir GH levels after OGTT higher than 0.4 ng/mL, and plasma IGF-I levels higher than age-matched normal subjects. Medical therapy with long-acting octreotide (OCT-LAR) was started at a dose of 20 mg every 4 weeks, and the dose was increased to 30 mg every 4 weeks in uncontrolled patients after 3 months of therapy. Efficacy of medical therapy was evaluated at the last patient visit, and patients were considered uncontrolled if they had a basal GH value higher than 1.0 ng/mL and/or a plasma IGF-I level higher than age-matched normal subjects with at least 6 months of treatment with OCT-LAR at a dosage of 30 mg. Postsurgical follow-up ranged from 12 to 60 months (median 32 months).
Tumor volume was not considered as an endpoint in this series because the study included only postsurgical patients, which could lead to mistakes in the volume measurements due to confounding variables such as postsurgical changes.

Cytokeratin pattern analysis
The cytokeratin expression pattern was analyzed as previously published with a mouse monoclonal antibody CAM 5.2 (1:100, BD Biosciences, San Jose, CA, USA, cat. number 349205) [19]. Tumors were classified according to the cytokeratin expression as densely granulated, sparsely granulated or mixed forms according to a previously reported classification [20]. Mixed tumors were considered as densely granulated for analysis, as previously suggested [20].

AIP mutation analysis in somatotropinomas
Deoxyribonucleic acid (DNA) was extracted using the QIAamp DNA Mini Kit (Qiagen, Valencia, CA, USA) from the pituitary adenoma tissue according to the manufacturer's protocol. The entire coding sequence of AIP (NM_003977.2), the conserved splice sites (from the conserved A of the upstream branch site to +10 downstream of each exon) and 1200 base pairs of the promoter region were direct sequenced, as previously published [4]. For those tumors whose DNA was not available, the complementary DNA (cDNA) was sequenced with previously published primers [4]. Sequencing was performed with ABI 3130 Genetic Analyzer (ABI PRISM/PE Biosystems, Foster City, CA, USA).

AIP mRNA and protein expression analysis in somatotropinomas
Immunohistochemistry. AIP expression was analyzed by immunohistochemistry, using a monoclonal antibody (1:500, NB100-127, Novus, Littleton, CO, USA) in paraffin-embedded tissue sections as previously described [7,8]. For semi-quantitative estimate of cytoplasmic AIP immunostaining, slides were scored for pattern [diffuse (score 2) or patchy (score 1)] and for intensity [strong (score 3), moderate (score 2) and weak (score 1)], and the final score was calculated by multiplying the two scores (pattern and intensity), as previously described [7,8]. Final scores of 0 (no expression), 1 and 2 were considered as low AIP expression, while scores 3, 4, or 6 were considered as high expression. The adenoma scoring was performed by a single independent observer (L.K.) blinded for the clinical data of the patients.
Reverse Transcription-qPCR. The AIP mRNA expression was analyzed by real-time qPCR in somatotropinomas and normal pituitaries. Tumoral ribonucleic acid (RNA) was extracted using the RNeasy Mini kit (Qiagen) according to the manufacturer's protocol. The amount and quality of the extracted RNA were evaluated using NanoDrop 2000 (Thermo Fischer, Wilmington, DE, USA). Approximately one microgram of total RNA was used in a reverse transcription reaction of 10 μL using 2.5 μM Oligo D(T), 5.5 mM MgCl 2 , 2.0 mM dNTPs, 20 U/μL RNase Inhibitor, 50 U/μL MultiScribe TaqMan and 10x RT Buffer, in a first strand cDNA synthesis kit (Taq-Man RT reagents, Applied Biosystems, Branchburg, New Jersey, USA). The reverse transcription cycle sequence was 25°C for 10 min, 48°C for 30 min and 95°C for 5 min. The cDNA of AIP and of glucuronidase β (GUSB), TATA box binding protein (TBP) and phosphoglycerate kinase 1 (PGK1) genes, used as endogenous controls, were separately amplified in duplicates, in a total volume of 12 μl, in real-time qPCR assays, using the Applied Biosystems 7500 Real-Time PCR System (Foster City, CA, USA). Reactions were incubated in a 96-well optical plate at 95°C for 10 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 1 min. The TaqMan assays for AIP and the endogenous controls are shown in Table 1. The cycle threshold (Ct) was defined as the cycle number at which the fluorescence surpasses the fixed threshold. The Ct data were performed using default threshold settings. Expression analysis was performed with the QPCR software [21]. Efficiency of each reaction was calculated by linear regression with the LingRegPCR software. The normalization of each sample results was performed by subtracting the Ct (geometric mean) for the target gene (AIP) by the endogenous control Ct (TBP, PGK1 and GUS), generating the ΔCt [Ct sample (target gene)-Ct sample (endogenous control)]. The

Identification and expression analysis of miRNAs targeting AIP 3'untranslated region (3'-UTR) in somatotropinomas
Target site prediction. To identify AIP mRNA-miRNA interaction we initially used algorithms described in the miRNAmap prediction program [22]. This bioinformatics tool uses data from TargetScan 6.0 (http://www.targetscan.org), MiRanda (http://www.microrna.org/ microrna/home.do) and RNAhybrid (http://bibiserv.techfak.uni-bielefeld.de/rnahybrid/). The mRNA-miRNA interaction was evaluated by three different criteria: (i) target site predicted by at least two prediction programs, (ii) the target gene contains multiple target sites for the miRNA and (iii) the target sites are located in accessible regions of the RNA as determined by a pre-specified algorithm. In addition, miRNAmap used miRNA and target mRNA expression profiles from a repository to calculate the Pearson correlation coefficients for each miRNA and the target gene [22,23]. We selected to evaluate by qPCR miRNAs that reach all three miRNAmap criteria or at least two miRNAmap criteria and a negative Pearson coefficient at least -0.30. Moreover, in order to confirm and to complete our search for miRNAs interacting with AIP and to determine the exact miRNA target binding sites in both human and rat AIP we utilized TargetScan version 6. Real-time qPCR quantification of miRNAs identified by in silico target prediction. The selected miRNA expressions were analyzed by real-time qPCR in somatotropinomas and in normal pituitaries. The reverse transcription cycle for miRNAs was 16°C for 30 min, 42°C for 30 min and 85°C for 5 min. The cDNA of the selected miRNAs and of RNU38B and RNU49, used as endogenous controls, were amplified in duplicates in three different reactions in a total volume of 12 μl, in real-time qPCR assays, using the Applied Biosystems 7500 Real-Time PCR System. Reactions were incubated in a 96-well optical plate with 40 cycles of 95°C for 15 sec and 60°C for 1 min. The TaqMan assays for miRNAs and for the endogenous controls are shown in Table 1. The analysis of the results was performed with the same methodology previously described for AIP.

Generation of mutant AIP-3'-UTR Reporter Plasmids
A pGL3-vector containing the human AIP-3'-UTR was used to perform the experiments [27]. A 931-bp segment of human AIP-3'-UTR is located immediately downstream from the coding sequence of the Firefly luciferase reporter gene. To examine whether the effect on the luciferase activity of the studied miRNAs was specifically due to binding to the predicted binding sites in the AIP-3'-UTR fragment, we disrupted these sites by site-directed mutagenesis. For interrupting the perfect "seed" pairing, four nucleotides (miR-34a site A and B) or three nucleotides (site C) of the miR-34a seed sequences were deleted using the QuikChange XL-site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA, USA) and the following primers: site A: forward 5'-ggccctgccttaccaagcccactgct-3' and reverse 5'-agcagtgggcttggtaaggcagggcc-3', site B: forward 5'-cctgccaagcccctgcagctgcca-3' and reverse 5'-tggcagctgcaggggcttggcagg-3', site C: forward 5'-gcccactgctgcccagcccccctg-3' and reverse 5'-caggggggctgggcagcagtgggc-3'. Three mutant plasmids were generated with deletions at site A (MUT_A), B (MUT_B) and C (MUT_C), and a further mutant was generated with both site A and C mutations. All mutant inserts were confirmed by direct sequencing.
Luciferase gene reporter assay GH3 cells were seeded in the inner wells of 24-well plates [28] at a density of 1x10 5 cells/well. After 24h, cells were co-transfected using Lipofectamine 2000 (Invitrogen, Paisley, UK) with 0.5 μg of the pGL3-vector and 25 ng of the Renilla vector [pRL-cyto megalovirus (CMV)] as previously described [27]. For each plate, the pre-miR-34a (PM11030, Life Technologies) or pre-miR-22 (PM11752, Life Technologies) or the scrambled pre-miR (AM17111, Life Technologies) was co-transfected at a final concentration of 50 nM. Firefly and Renilla luciferase activities were measured consecutively 24h post-transfection using the Dual-Luciferase Reporter Assay System (Promega, Southampton, UK) as previously described [27]. Ratios of Firefly vs. Renilla luminescence signals served as a measure for reporter activity normalized for transfection efficiency.

Endogenous miR-34a expression in different cell lines and tissues
In order to estimate the level of expression of miR-34a in the GH3 and HEK293 cells we extracted RNA from these cells using the RNeasy Mini Kit (Qiagen). We also included in the analysis RNA from human tissues (AM6000, Ambion) previously described to express miR-34a at high (ovary, prostate and testes) or low (adipose, heart and liver) levels (http:// mirnamap.mbc.nctu.edu.tw). Real-time qPCR amplifications were run using the hsa-miR-34a TaqMan MicroRNA Assay, (4427975, Life Technologies). RNU6B was used as an endogenous control for human samples and beta-actin was chosen as a control for the rat sample. The analysis of the results was performed with the same methodology previously described for the AIP qPCR.
Endogenous AIP mRNA and protein expression after miR-34a overexpression and inhibition HEK293 and GH3 cells were seeded in 24-well plates at a density of 0.6x10 5 cells/well and 1x10 5 cells/well, respectively. After 1-24 h cells were transfected with the pre-miR-34a precursor, the anti-miR-34a inhibitor (AM11030, Life Technologies), scrambled-miR or scrambledanti-miR (AM17010, Life Technologies) at a final concentration of 50 nM. Twenty-four and forty-eight hours later cells were harvested and proteins and RNA extracted. RT-qPCR was performed with the TaqMan system using ready made AIP-probe primer kits (Hs_00610222_m1, Rn_00597273-m1, Life Technologies). Reactions were performed in triplicate using glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as endogenous control. Data were analyzed as previously described [27]. Protein extraction and quantification were performed as described previously [29]. Twenty to 80 μg of cell culture lysates were separated by electrophoresis and transferred onto nitrocellulose membranes. Membranes were incubated with mouse monoclonal AIP antibody (Novus NB100-127) at 1:1000 dilution and GAPDH rabbit antibody (sc-25778 Santa Cruz Biotechnology, Dallas, USA) at 1:1000 dilution was used as loading control. Infrared fluorescent-labeled anti-rabbit or anti-mouse secondary antibodies (IRDye 680 and 800, Li-Cor Biosciences, Cambridge, UK) were used at a 1:8000 dilution. Immunoblot detection and density measurements were performed on the Odyssey infrared-imaging system (Li-Cor).

Statistical analysis
The statistical analysis was performed using SPSS version 16.0 for Windows (SPSS, Inc., Chicago, IL, USA) or StatsDirect software (Addison-Wesley-Longman, Cambridge, UK). The results were reported as median values (minimum-maximum) or mean ± SEM of two to ten independent experiments, each performed in triplicate. The Student t-test or the Mann-Whitney test was used as appropriate to compare numerical variables. The chi-squared test was used to compare categorical variables. P values < 0.05 were considered statistically significant.

Analysis of AIP mutations
A total of 34 tumors from acromegalic patients were selected for the study. Tumor genomic DNA (gDNA) from 28 patients and tumor cDNA from six patients were sequenced for AIP mutations. Two patients were identified with truncating AIP mutations (p.Y268Ã, and p. R304Ã) and one patient with a variant with controversial significance (p.R16H) [30][31][32]. Leukocyte-derived DNA from these patients confirmed heterozygous germline mutations and these three patients were excluded from the study; therefore only data from 31 adenomas were included in following experiments.
Demographic, radiological, biochemical and pathological data of the patients with acromegaly The demographic, biochemical and pathological data of the 31 patients included in the study are summarized in Table 2. The median age at diagnosis was 43 years (range 23-63), 15 patients (48%) were male. The median GH at diagnosis was 23.0 ng/mL (range 1.6-392.5) and the median IGF-I was 408% of the upper limit of normal range (range 165-1139). Twentyseven tumors (87%) were macroadenomas and 15 (48%) co-expressed GH and prolactin upon immunostaining.
AIP protein levels and correlation with AIP mRNA levels All tumors expressed AIP, with low expression levels (score 1-2) observed in 13 cases (42%) (Fig. 1). Interestingly, there was no difference in the AIP mRNA expression between tumors with low or high AIP protein levels: in the low AIP protein group the median AIP mRNA expression was 0.91 (range 0.48-1.95) and in the high AIP protein group was 1.14 (0.45-2.34, low vs. high protein group P = 0.123). These data lead us to the hypothesis that post-transcriptional regulation, such as that exerted by miRNAs, may be the cause of the low AIP protein expression. Based on in silico predictions, we selected 11 miRNAs for analysis by real-time qPCR: let-7a, let-7b, miR-202, miR-22, miR-34a, miR-34c, miR-449b, miR-510, miR-612, miR-639 and miR-671 (Table 3). Two miRNAs showed higher expression in tumors with low AIP protein levels compared to tumors with high AIP protein levels ( Fig. 2 . There was no difference in the expression of the other miRNAs between tumors with low or high AIP protein levels (Fig. 2).
A total of 26 patients were initiated on OCT-LAR after surgery. In 10 patients (39%) acromegaly was considered controlled after OCT-LAR therapy. Only one out of nine patients (11%) whose tumors presented low AIP expression achieved disease control with medical treatment, while nine out of 17 patients (53%) harboring tumors with high AIP expression achieved disease control with OCT-LAR therapy (P = 0.045). The miR-34a levels were lower in those patients controlled with OCT-LAR therapy than in the uncontrolled patients [0.14 (range 0.08-4.96) and 1.12 (range 0.72-2.19), respectively, P = 0.003].
There was no correlation of miR-22 levels with tumor invasiveness, granulation pattern or response to OCT-LAR therapy.

miR-34a effect on regulation of AIP expression in vitro
To verify the in silico predicted interaction between miR-34a and miR-22 and AIP, we used a pGL3 vector containing the human wild type (WT) AIP-3'UTR downstream of the coding sequence of Firefly luciferase. Transfection of pre-miR-34a precursor and WT-AIP-3'UTR into GH3 cells resulted in a 31±4% reduction of luciferase activity compared with the control scrambled miR (P<0.0001) (Fig. 4). To confirm that this effect was caused by miR-34a interaction with the cloned fragment and not by nonspecific binding, we compared the effect exerted by the pre-miR-34a precursor and the scrambled miR on the empty pGL3 vector. miR-34a did not change the luciferase activity of the empty vector compared with the control scrambled miR (Fig. 4). As the endogenous level of miR-34a in GH3 cells was low (Fig. 5), we predict that the endogenous miR-34a did not interfere significantly in our experimental setting. Transfection of pre-miR-22 precursor and WT-AIP-3'UTR into GH3 cells resulted in no reduction of the luciferase activity (data not shown).  MiR-34a in Somatotropinomas

Confirmation of predicted miR-34a binding sites
To confirm the importance of miRNA binding and to investigate which predicted binding site of miR-34a is involved in the miR-34a effect we used deletion mutants targeting the three different bindig sites: MUT_A for the mutated binding site A, MUT_B for site B and MUT_C for site C. MUT_A leads to a complete loss of miR-34a effect on luciferase activity (Fig. 4), while MUT_B did not change the inhibitory effect of miR-34a on the luciferase assay (Fig. 4). Although MUT_C overall did not change significantly the inhibitory effect of miR-34a, in some of the experiments a small effect was observed. Therefore we created a combined mutant of site A and site C: MUT_A+C to see if an additional effect of site A and C could be observed. The data with the combined mutant was similar to the one with MUT_A only (Fig. 4).

Regulation of endogenous AIP expression by miR-34a in vitro
To further characterize the interaction of miR-34a and AIP in vitro we measured mRNA and protein levels of endogenous AIP after miR-34a overexpression and inhibition in HEK293 cells. Significant decrease in AIP protein level was observed 48h post-transfection with miR-34a compared to scrambled miR control (n = 7, 17±3%, P = 0.001, Fig. 6A), suggesting that high levels of miR-34a can suppress endogenous AIP protein expression in vitro. Transfection with anti-miR-34a did not change AIP protein levels (n = 3, 0±7%, P = 0.998, Fig. 6B). We have also observed a significant decrease in endogenous AIP protein levels in GH3 cells after miR-34a overexpression (n = 4, 25±1%, P = 0.0005, Fig. 6C).
Although miR-34a overexpression induced a significant decrease in endogenous AIP protein levels, no significant change was seen at the mRNA level in HEK293 and GH3 cells, matching observations in our human adenomas. After miR-34a overexpression in HEK293 and GH3 cells we have observed no significant change in AIP mRNA levels compared to scrambled miR control (Fig. 6D, 6E).

Discussion
In this study we showed that low AIP protein levels in human sporadic somatotropinomas is associated with high miR-34a expression and that miR-34a can down-regulate AIP protein levels in in vitro experiments. In addition, we showed that high miR-34a levels are associated with a lower chance of acromegaly control with SSA therapy and we confirmed our previous MiR-34a in Somatotropinomas findings [7] that low AIP protein expression is associated with a poor response to SSA. Our data demonstrates that the inhibition involves AIP translation repression without reduction in AIP mRNA, as there was no difference in the AIP mRNA levels between tumors with low or high AIP protein levels.
Thirteen out of 31 (42%) tumor samples showed low AIP protein levels, a percentage similar to the data previously published by Jaffrain-Rea et al. (48%) [9] and also by our group in two different sets of samples (55% and 51%) [7,8]. Most of the tumors exhibiting low AIP expression (with or without germline AIP mutations) were invasive [8,9], and patients harboring those tumors have a poor response to the medical treatment with SSAs, the mainstay of the medical treatment of acromegaly [7]. Recent data suggest that the expression of AIP is important in the mechanism of action of this class of drugs [3,17]. In addition, we showed that the majority of the tumors showing low AIP expression are sparsely granulated adenomas. Tumors with this cytokeratin pattern are known to be more invasive and associated with a poor response to SSA therapy [33]. Therefore, one of the possible explanations for this adenoma phenotype is the low AIP expression in these tumors.
As our data showed that there is no correlation between AIP mRNA expression and protein levels, we hypothesized that the low AIP protein levels could be explained by miRNA regulation. Therefore, we studied the miRNAs predicted to regulate the AIP gene. After a careful selection using in silico prediction, we studied 11 miRNAs. Although miR-107 has been previously shown to inhibit AIP in vitro [27], this miRNA did not fulfill the strict selection criteria in this study and was not included in the analysis. Out of our selected 11 miRNAs, two (miR-22 and miR-34a) were significantly overexpressed in the tumors with low AIP protein levels. We note that our sample size is the largest to date of the studies concentrating on miRNA expression in somatotropinomas. According to our results, we hypothesized that these miRNAs could be responsible for the reduced AIP levels. Our findings showed that miR-34a down-regulates AIP protein levels, while miR-22 had no inhibitory effect. We also showed that higher miR-34a levels are associated with a lower chance of disease control with SSA therapy. Therefore, elucidation of the mechanism involved in reducing AIP protein levels in sporadic somatotropinomas may help to predict response to SSAs and help with the development of novel therapeutic options for the treatment of this subgroup of invasive tumors.
To study the effect of miR-34a on AIP we have used a reporter vector in which the entire human AIP-3'-UTR has been fused to a luciferase reporter plasmid. We observed that high MiR-34a in Somatotropinomas levels of miR-34a reduce the luciferase activity using the vector containing the WT AIP-3'UTR suggesting that miR-34a can bind to the AIP-3'UTR. To validate the three predicted target sites for miR-34a within the 3'-UTR of human AIP we created mutants for each predicted miR-34a binding sites. MUT_B and MUT_C plasmids exerted a significant inhibitory effect on the luciferase activity, similar to the wild-type plasmid, suggesting that these sites are not important for miR-34a binding. SITE_A plasmid, however, failed to induce inhibition of luciferase activity, suggesting that this binding site is responsible for miR-34a-mediated AIP repression. To study the regulation of endogenous AIP expression by miR-34a in vitro we have used rat GH3 cells and human HEK293 cells. We found that overexpression of miR-34a significantly decreases AIP protein levels measured 48h after transfection, while there was no significant change in AIP mRNA levels. When we transfected the cells with anti-miR-34a we could not see any change in AIP protein levels, which might be due to the fact that these cell lines have very low endogenous miR-34a levels therefore antagonism with anti-mir-34a may not produce detectable effects (Fig. 5). We have attempted to see if cells would proliferate more when transfected with miR-34a, and although a trend was observed, this has not reached significance (data not shown).
The miR-34 family consists of three miRNAs: miR-34a, miR34b and miR-34c. miR-34b and -34c share a common primary transcript (located on chromosome 11q23), while miR-34a is encoded by its own transcript on chromosome 1p26 [34]. miR-34a is an intergenic miRNA located between the genes coding for the G-protein coupled receptor 157 and hexose-6-phosphatase dehydrogenase [34]. Recent miRNA profiling analysis did not find differences in overall miR-34a expression in somatotropinomas compared to normal pituitary [11,35]. This is in line with our findings that showed that miR-34a was overexpressed in tumors with low AIP expression, but not in all somatotropinomas. There were a few cases with high miR-34a level and high AIP expression. We have not identified sequence variants in the 3'UTR of the AIP gene in these samples. Therefore as high miR-34 expression does not always correlate with a low AIP protein level, other factors could also influence AIP expression. To elucidate the complex regulation of AIP expression will need future studies.
miRNAs have unique tissue-specific patterns and the same miRNA can behave as an oncogene or tumor suppressor gene depending on their target mRNAs in that particular organ [13,36]. miR-34a has been shown to behave as a tumor suppressor miRNA by reducing proliferation and enhancing apoptosis in many human neoplasias, including osteosarcoma, colorectal, pancreatic and ovarian cancer [37,38]. In contrast, miR-34a has also been shown to have oncomiR properties in other studies, including a pro-proliferative role in follicular lymphoma cell lines and an anti-apoptotic effect in B-lymphoid cells [39,40]. It has also been shown to antagonize the anti-tumoral effects of docetaxel in human breast cancer cells [41]. Our data show higher miR-34a levels in human adenomas with low AIP protein levels. In addition, we demonstrated that miR-34a can bind to AIP and decrease its expression in vitro. Interestingly, miR-34a overexpression in HCT116 human colon carcinoma cells also showed down-regulation of AIP expression [42] (http://www.ncbi.nlm.nih.gov/geoprofiles/39833130), suggesting that the putative role of miR-34a in regulating AIP expression may also be present in other tumors. Patients with loss-of-function AIP mutations usually harbor large and invasive somatotropinomas [4,5]. In the absence of mutations, low AIP protein level is associated with a similar phenotype [8,9]. In this study we postulated that miR-34a acts as an oncomiR in somatotropinomas, being an inhibitor of AIP, a tumor suppressor gene, and therefore miR-34a might be implicated in the pathogenesis of these tumors. On the other hand, as miR-34 is involved in a wide range of tumorigenesis, the role of miR-34 on somatotropinomas may not solely rely on AIP.
The involvement of miR-34a in the pathogenesis of sporadic somatotropinomas may allow the development of new therapeutic strategies for the treatment of these tumors. The therapeutic inhibition of miR-34 has previously been attempted in the context of heart disease. This resulted in the attenuation of pathological cardiac remodeling and improvement in heart function in a mouse myocardial infarct model [43]. In the same study, the authors used a 15mer locked nucleic acid (LNA) anti-miR-34a and observed that a single dose in three consecutive days inhibited the miR-34a as early as day one and that the inhibition persisted for two months after the last dose [43]. Therefore, future studies addressing the use of LNA anti-miR-34a in the setting of invasive somatotropinomas with low AIP protein levels may provide a new approach for the treatment of these tumors. The use of anti-miRNAs has been previously described in other tumor cell lines, for example, in an orthotopic xenograft breast cancer model with systemically injected liposomes that delivered 2'-O-Me anti-miRNAs against miR-132, resulting in delayed tumor growth and suppressed angiogenesis [44].
In conclusion, we have demonstrated that miR-34a is overexpressed in sporadic somatotropinomas with low AIP protein levels in the absence of mutations in this gene and that this overexpression is inversely correlated to the response to SSA. Functional studies confirmed that miR-34a down-regulates AIP expression, suggesting the possible involvement of miR-34a in the pathogenesis of sporadic somatotropinomas.