miR‐34a is upregulated in AIP‐mutated somatotropinomas and promotes octreotide resistance

Pituitary adenomas (PAs) are intracranial tumors associated with significant morbidity due to hormonal dysregulation, mass effects and have a heavy treatment burden. Growth hormone (GH)‐secreting PAs (somatotropinomas) cause acromegaly‐gigantism. Genetic forms of somatotropinomas due to germline AIP mutations (AIPmut+) have an early onset and are aggressive and resistant to treatment with somatostatin analogs (SSAs), including octreotide. The molecular underpinnings of these clinical features remain unclear. We investigated the role of miRNA dysregulation in AIPmut+ vs AIPmut− PA samples by array analysis. miR‐34a and miR‐145 were highly expressed in AIPmut+ vs AIPmut− somatotropinomas. Ectopic expression of AIPmut (p.R271W) in Aip−/− mouse embryonic fibroblasts (MEFs) upregulated miR‐34a and miR‐145, establishing a causal link between AIPmut and miRNA expression. In PA cells (GH3), miR‐34a overexpression promoted proliferation, clonogenicity, migration and suppressed apoptosis, whereas miR‐145 moderately affected proliferation and apoptosis. Moreover, high miR‐34a expression increased intracellular cAMP, a critical mitogenic factor in PAs. Crucially, high miR‐34a expression significantly blunted octreotide‐mediated GH inhibition and antiproliferative effects. miR‐34a directly targets Gnai2 encoding Gαi2, a G protein subunit inhibiting cAMP production. Accordingly, Gαi2 levels were significantly lower in AIPmut+ vs AIPmut− PA. Taken together, somatotropinomas with AIP mutations overexpress miR‐34a, which in turn downregulates Gαi2 expression, increases cAMP concentration and ultimately promotes cell growth. Upregulation of miR‐34a also impairs the hormonal and antiproliferative response of PA cells to octreotide. Thus, miR‐34a is a novel downstream target of mutant AIP that promotes a cellular phenotype mirroring the aggressive clinical features of AIPmut+ acromegaly.

Fonds d'Investissement Pour la Recherche Scientifique; Helmholtz Alliance; JABBS Foundation together, somatotropinomas with AIP mutations overexpress miR-34a, which in turn downregulates Gαi2 expression, increases cAMP concentration and ultimately promotes cell growth. Upregulation of miR-34a also impairs the hormonal and antiproliferative response of PA cells to octreotide. Thus, miR-34a is a novel downstream target of mutant AIP that promotes a cellular phenotype mirroring the aggressive clinical features of AIPmut+ acromegaly. in the pathogenesis of pediatric-onset PAs. 8,9 Germline AIPmut-associated (AIPmut+) PAs are distinctive, presenting usually as growth hormone (GH)-secreting PAs (somatotropinomas) that are significantly larger, more invasive and occur at a younger age than PAs without AIP mutations (AIPmut-). 10 The firstline treatment for sporadic somatotropinomas is transsphenoidal surgery, which controls hormone hypersecretion in only about 50% of patients. 11 For uncontrolled or recurrent disease, medical therapy with somatostatin analogs (SSA) is used in most cases. 12 As the response to SSA is dependent on several molecular determinants, 10%-30% of patients do not respond to this therapy and can require other therapies such as pegvisomant or radiotherapy; novel therapeutic approaches (eg, metformin and statins) are now being evaluated preclinically and in small patient cohorts (reviewed in Reference 13). AIPmut + somatotropinomas are relatively resistant to first-line medical treatment with SSA. 10 The molecular pathways that lead to this aggressive, treatment-resistant phenotype are of great clinical relevance, but the specific pathway(s) are unclear.
AIPmut+ leads to impaired interaction with phosphodiesterase-4A5 (PDE4A5), 14 dysregulated protein kinase A (PKA) function, 15,16 and disrupted regulation of the inhibitory G protein, Gαi2, which can increase cyclic adenosine monophosphate (cAMP). 17,18 Like many other human tumor types, the development of PAs can be influenced by microRNAs (miRNAs); differential miRNA expression in PAs is related to histological tumor subtypes, clinical behavior and treatment responses. [19][20][21][22][23][24] Interactions between miRNAs and normal, nonmutated AIP have also been noted: miR-107 is overexpressed in somatotropinomas and binds to and represses AIP in GH3 cells in vitro. 25 As the effects of AIPmut per se on miRNA expression in human tumors are unknown, we set out to determine the miRNA signature of AIPmut+ and AIPmut− PAs. Among the miRNAs we identified as being upregulated in AIPmut+ PA tissues was miR-34a. We show that miR-34a has pro-oncogenic functions in PAs, likely mediated by increased cyclic adenosine monophosphate (cAMP) signaling, and that Gαi2 is a direct target of miR-34a and is differentially expressed in human PAs depending on AIPmut status. We report for the first time that miR-34a upregulation leads not only to increased cell proliferation and GH secretion in vitro, but also induces resistance to the antiproliferative and hormonal effects of the first-generation somatostatin analog, octreotide.

| DNA extraction and LOH analysis
To establish germline AIP status, DNA was isolated from peripheral blood leukocytes and AIP was sequenced under conditions described. [8][9][10] Germline DNA was also studied for deletions in AIP using multiplex ligand-dependent probe amplification (MLPA), but no germline deletions were present. DNA was also extracted following microdis- Input was regulated eight miRNAs from somatotropinomas (FC > 1.5x, P < .002; miR-383 was excluded). Heat maps were generated in R.

| Plasmid constructs and antibodies
The p.  strand of the fragment. After oligo annealing, the fragments were cloned in a psiCHECK-2 backbone (#C8021, Promega).
Primers and Oligos for cloning and mutagenesis are listed in Table S2.
Primary antibodies are listed in Table S3 Table S4.  Table S5.
Electroporation was conducted using the pulse code CZ-167. All cell lines were routinely tested for mycoplasma contamination using the PCR Mycoplasma Test Kit (#PK-CA91-1048, PromoKine) and found to be mycoplasma-free.
For therapy response, cells were treated with 10 or 100 nM octreotide (kindly provided by Italfarmaco SpA) in serum-free medium for the indicated times.

| Protein extraction and Western blotting
Cells were collected at the indicated time points and lysed in radio- Clonogenic potential of the cells was monitored by seeding the miRNA overexpressing cells at a low density of 1 × 10 3 cells per well in six-well plates. After 8 days, colonies were stained with 0,3% Crystal Violet in 30% ethanol and those with a diameter >100 μm were counted.
GH3 cells transfected with miRNA mimics were plated in 96-well plate at a density of 1.5 × 10 5 cells per well and six wells per group.
For assessment of apoptosis in the cells, Caspase-Glo 3/7 Assay (#G8090, Promega) was performed according to the manufacturer's instructions.

| Immunoassays
Transfected GH3 cells (7 × 10 5 per well) were plated in six-well plates and 24 hours later, cAMP levels were determined by using an ELISA

| miRNA target prediction
The targets of the differentially expressed miRNAs were predicted by four different prediction tools, TargetScan 7.0, miRANDA-mirSVR, DIANA tools and PITA that take different parameters into consideration (Table S6).

| Reporter gene assays
To conduct reporter gene assays, HEK293 cells were lysed 24 hours after transfection according to the protocol of the Dual-Luciferase Reporter (DLR) Assay Kit (#E1980, Promega). Subsequently the reporter assay was performed according to the manufacturer's instructions.

| Immunohistochemistry and scoring
Immunhistochemical stainings were performed on an automated immunostainer (Ventana Medical Systems) as previously described. 28 Primary antibodies were diluted in Dako REALTM antibody diluent

| Statistical analysis
Results of the cell assays are shown as the mean of values obtained in independent experiments ± SEM. Unpaired two-tailed Student's ttest, Mann-Whitney test, one-way and two-way ANOVA were used to detect significance between two series of data, and P < .05 was considered statistically significant.  Figure 1A). In prolactinomas, three miRNAs were regulated using the same criteria ( Figure S3). One AIPmut+ prolactinoma was excluded due to suboptimal RNA quality, and since only two samples remained, we focused on the somatotropinomas for subsequent analyses. Some of the significant differentially expressed miRNAs have previously been linked to PAs, including miR-26a, downregulated in prolactinomas vs normal pituitary and upregulated in invasive adenomas 19,29 ; miR-143 and miR-145 were downregulated in ACTHsecreting adenomas vs normal pituitary, 19 and miR-383, downregulated in invasive NFPAs. 19 To obtain a functional annotation of the differentially expressed miRNAs, we used Ingenuity Pathway Analysis. The most enriched terms were cancer-related, followed by terms related to cell cycle and cellular movement (Table S7)

| Short-term upregulation of miR-145 and miR-34a promotes oncogenic features in PA cells
We next investigated whether upregulation of miR-34a and/or miR-145 (mimicking AIPmut+ tumors) affects the phenotypic features of PA cells. GH3 cells (derived from a rat GH/prolactin-secreting adenoma) were transfected with specific mimics for mature miR-34a and miR-145, and functional assays were performed assessing proliferation, clonogenic potential, migration and apoptosis (Figure 2A-D).
Short-term overexpression of both miRNA mimics (24 hours) increased GH3 cell proliferation when compared to cells transfected with an unspecific control miRNA mimic (Figure 2A). GH3 cells overexpressing miR-34a migrated significantly more than the control ones, whereas transfection of miR-145 mimic had no effect on cell migration ( Figure 2B). miR-34a and miR-145 overexpression decreased GH3 cells apoptosis, with miR-34a having the stronger effect ( Figure 2D). miR-34a, but not miR-145, upregulation increased by Octreotide suppresses GH and prolactin (PRL) secretion from GH3 cells in vitro. 32 Overexpression of AIP-wt in GH3 cells reduces GH secretion by decreasing cAMP levels. 33 Thus, we measured GH in the supernatant of GH3 cells transfected with the above miRNA 3.7 | Guanine nucleotide-binding protein G(I) subunit alpha-2 (GNAI2) as a novel predicted target gene of miR-34a AIPmut in PAs leads to upregulation of miR-34a, which in turn promotes proliferation and cAMP signaling. To better understand these effects, we identified predicted miR34a target genes using four different prediction tools (Table S6). Following the criteria outlined above (Materials and Methods), we selected several putative target genes (Table S8). As TargetScan considers only seed match and conservation for target prediction (Table S6), we excluded targets predicted only by algorithm alone. We focused on genes involved in cAMP signaling (Table S9), which included several phosphodiesterases (PDE3a, PDE4A, PDE5a, PDE7b) and G-protein alpha subunit family members (GNAO1, GNAI2, GNAI3, GNAQ; Table S9). GNAI2 and GNAI3 are interesting candidates as they encode inhibitory Gαi subunits which lead to decreased cAMP levels. To verify whether these genes were regulated by miR-34a, we transfected GH3 cells with the miR-34a mimic (for overexpression) or a specific anti-miR-34a (for downregulation) and assessed the level of GNAI2 and GNAI3. The modulation of miR-34a levels were confirmed by qRT-PCR at 24 hours and 48 hours posttransfection ( Figure S5). Aip was previously shown to be regulated by miR-34a 37  Then, we tested whether miR-34a directly targets Gnai2 using reporter gene assays. Given that a seed match for miR-34a in the 3'UTR region of rat Gnai2 was predicted, we cloned this sequence into the psiCHECK-2 vector upstream of the firefly luciferase gene ( Figure 3G). This construct was then transfected into HEK293 cells F I G U R E 2 mir-34a promotes tumorigenic behavior in PA cells. A-D, GH3 cells were transfected with an unspecific miRNA mimic (unspecific control) or with specific mimics for mature miR-34a or miR-145. All assays were performed 24 hours after transfection. The values are normalized against the negative control arbitrarily set to 100%. Results are reported as the mean ± SEM. A, Cell viability was assessed by the WST-1 assay. The experiment was independently performed three times each with six technical replicates. B, Migration assays were conducted using the Boyden chamber and the migrated cells were counted. The experiment was independently performed twice each time with three technical replicates. C, Transfected GH3 cells were plated and selected. Seven days later the colonies were fixed, stained and those that reached a diameter >100 μm were counted. The experiment was independently performed three times each with 3 to 4 technical replicates. D, Caspase 3/7 activity was measured to assess apoptosis. The experiment was independently performed three times each with two technical replicates. Results are reported as the mean ± SEM *P < .05; **P < .01; ***P < .001; n.s., not significant (by one-way ANOVA). E-G, effect of miR-34a and miR-145 on cell viability and GH or PRL secretion of GH3 cells following octreotide treatment. GH3 cells were transfected with an unspecific miRNA (unspecific control) or with specific mimics for mature miR-34a or miR-145 and subsequently treated with 10 nM octreotide (Oct) for 72 hours or 100 nM Oct for 48 hours. Cell viability (E) and GH (F) or PRL (G) secretion were then measured using the WST-1 assay (E) or ELISA (F,G), respectively. The experiments were independently performed three times each with six (e) or two (F, G) technical replicates. The values are normalized against the untreated unspecific control arbitrarily set to 100%. Results are reported as the mean ± SEM. *P < .05; **P < .01; ***P < .001; n.s., not significant (by two-way ANOVA). H-L, effect of miR-34a and miR-145 on cAMP levels and ERK1/2 activation. H, GH3 cells were transfected with an unspecific miRNA (unspecific control) or with specific mimics for mature miR-34a or miR-145. Intracellular cAMP levels were measured 24 hours after transfection. The experiment was independently performed five times each with two technical replicates. The values are normalized against the negative control arbitrarily set to 100%. Results are reported as the mean ± SEM. I, In samples parallel to (H), western blot was performed 24 hours after transfection using an anti-p-ERK-1/2 antibody and an anti-ERK-1/2 antibody. The blot shown is representative of 2 independent experiments with similar results. L, Ratio of the band intensities of the blots (n = 2) described in (H). Results are reported as the mean ± SEM. ***P < .001; n.s., not significant (by one-way ANOVA). UT, untransfected; UC, unspecific control together with an unspecific miRNA inhibitor or the anti-miR-34a, and luciferase activity was monitored. A decrease in luciferase activity was detected between cells cotransfected with empty vector and unspecific control miRNA inhibitor vs cells cotransfected with empty vector and anti-miR-34a inhibitor ( Figure 3I). In contrast, anti-miR-34a increased the luciferase activity of the construct containing the Gnai2 F I G U R E 3 Legend on next page. seed match vs unspecific miRNA inhibitor by almost 30% (Figure 3I).
Deletion of the seed region abolished the anti-miR-34a-mediated regulation of luciferase activity, thereby confirming that this miRNA does bind to the seed sequence in Gnai2 ( Figure 3H,I).

| Gαi2 expression is reduced in human AIPmut + PAs
AIPmut+ PAs show an increase in miR-34a expression, which in turn regulates Gnai2 expression. Thus, we next investigated Gαi2 (and AIP) expression by immunohistochemistry (IHC) on AIPmut− and AIPmut+ samples (Table S1). Slides were then scored semiquantitatively for staining intensity ( Figure 4A). In total, 38 human PA tumors were scored for AIP expression and 30 (n = 18 AIPmut−; n = 12 AIPmut+) for Gαi2 expression ( Figure 4B,C). We found that levels of Gαi2 were significantly lower in AIPmut+ vs AIPmut− tumors (P < .05; Figure 4D), whereas there was no statistically significant difference in AIP staining between the two sample groups ( Figure 4E). By only considering the somatotropinomas (n = 13, AIPmut−; n = 9, AIPmut+), decreased Gαi2 staining in AIPmut + tumors was even more pronounced (P < .01; Figure 4F). Again, no significant difference in AIP expression was detected between AIPmut+ and AIPmut− somatotropinomas

| DISCUSSION
Acromegaly is a rare and disfiguring disease that, if inadequately treated, carries significant morbidity and increased mortality. 12 Resistance to medical therapy with SST2-specific SSAs like octreotide or lanreotide is an important impediment to hormonal control in acromegaly. AIPmut in acromegaly leads to such SSA resistance, which has a major clinical impact on patients as AIPmut are also associated with young-onset, large and invasive PAs. 10 The mechanisms to explain this phenotype are therefore of high clinical relevance. We report here for the first time that AIPmut+ somatotropinomas have a distinct miRNA profile of miR-34a upregulation, a well-known miRNA that can function as an oncomiR in multiple cancers. Loss of AIP due to mutation led to increased miR-34a, which was associated with increased cAMP and low Gαi2 expression. Importantly, miR-34a dysregulation also miR-34a, together with miR-34b and miR-34c, are transcriptionally regulated by p53. miR-34a was initially considered a tumor suppressor since it is downregulated in several cancers, and reduces proliferation and induces apoptosis of tumor cells. 40,44,45 However, recent evidence also points to a pro-oncogenic role for miR-34a as it is overexpressed in gastric cancer and brain tumors, among others. 46-48 miR-34a plays a proproliferative and antiapoptotic role in vascular and lymphoid tissues, [49][50][51] and it was also found to induce chemoresistance of osteosarcoma cells, 52 and to promote genomic instability. 53 Therefore, miR-34a overexpression, a feature often acquired during carcinogenesis, can play an oncogenic or tumorsuppressive role in a tissue-and context-specific manner. In the normal pituitary, miR-34a is expressed at a level similar to that seen in AIPmut− somatotropinomas, being overexpressed only in adenomas with an AIP mutation ( Figure S9). Interestingly, miR-34a was previously shown to have increased expression in non-AIPmut somatotropinomas with low AIP protein levels, and to directly target and inhibit AIP in a nonpituitary model (HEK293). 37  In contrast to findings from our group and from Denes et al, 37 it was recently reported that long-term overexpression of miR-34a decreases the proliferation of GH4C1 cells, 54 a clone derived from GH3 cells but that produces negligible amounts of GH. We did not see changes in GH4C1 cell proliferation upon overexpression of miR-34a up to 72 hours posttransfection ( Figure S10) whereas in GH3 cells an increase in viability was already detectable 24 hours posttransfection and lasted for several days (eg, in clonogenic assays). This discrepancy suggests that miR-34a elicits different effects due to underlying secretory and molecular heterogeneity between GH3 and GH4C1 clones. 55,56 Dysregulation of intracellular cAMP levels is a hallmark of functioning endocrine tumors, including somatotropinomas, where cAMP promotes cell division and GH secretion. A link between AIP and cAMP has been established in that overexpression of wild-type AIP in GH3 cells attenuates the forskolin-dependent increase in intracellular cAMP and GH secretion, whereas silencing of endogenous Aip increases cAMP concentration. 33 Mechanistically, AIP interacts with several members of the cAMP signaling cascade, including phosphodiesterases, the enzymes that deactivate cAMP, and G proteins, which can either activate AC and increase cAMP (activating) or have the opposite effects (inhibitory). 16,57 In Aip-deficient MEFs, the inhibitory Gαi2 and Gαi3 proteins do not inhibit cAMP synthesis, which suggests that AIPmut-related pituitary tumorigenesis may occur via Gαi signaling and cAMP. 17 In this context, our finding that in PA cells AIPmut leads to upregulation of miR-34a, which in turn increases intracellular cAMP levels, offers an additional molecular mechanism relating defective AIP to cAMP production ( Figure 5). Among the predicted miR-34a F I G U R E 5 Model of the effect of AIP deficiency on intracellular processes. A, Physiological and B, pathological (AIP mutation or loss) scenario in which AIP deficiency leads to upregulation of miR-34a which in turn decreases Gαi2 levels. The inhibition of AC by Gαi2 is reduces and the AC produces nonphysiological amounts of cAMP, which acts as an mitogenic signal in the cell targets was GNAI2 encoding Gαi2, which was confirmed to be a direct target by reporter gene assays. Accordingly, miR-34a overexpression in PA cells reduces GNAI2 levels, while its downregulation increases them. A reduction intracellular levels of Gαi2 is expected to enhance AC activity and to increase cAMP levels, which we observed.
Although a role has been suggested for Gαi2 in mediating the effects of defective Aip, the mechanism was unknown as loss-of-function mutations in GNAI genes are not seen in PAs. 58 We demonstrate that AIPmut decreases GNAI2 expression via miR-34a upregulation, and this ultimately leads to increased AC activity as well as cAMP levels ( Figure 5).
In line with previous reports, we observed a trend towards a negative correlation between AIP levels and tumor size. 59,60 Also, we confirmed that low AIP staining is associated with higher likelihood of tumor invasion and AIPmut+ status did not predict AIP staining intensity. 61 We also found that AIPmut are associated with a reduction in Gαi2 levels, thereby extending the findings of Tuominen et al. 17 Similar to AIP, the levels of Gαi2 are reduced in invasive vs noninvasive PAs only in the AIPmut+ group.
In summary, we show that AIPmut in GH-secreting PAs leads to dysregulation of a specific subset of miRNAs, including miR-34a, which is induced by AIPmut and has pro-oncogenic functions in somatotropinomas through regulation of Gαi2 expression and increased cAMP concentrations. As increased levels of miR-34a impair the response of PA cells to octreotide, the lack of response of AIPmut+ patients to first-generation SSAs might be mediated by induction of miR-34a expression. These findings further support the hypothesis that failure to inhibit cAMP synthesis due to downregulation of Gαi2 is key event in AIP-mediated pituitary tumorigenesis. Furthermore, this adds another level of complication to the multifaceted role of AIP in PAs given that the increase of miR-34a may lead to the regulation of other genes in addition to Gnai2. As both high miR-34a and low Gαi2 levels are associated with resistance to SSA therapy, they represent potential biomarkers that could be used as evidence to personalize treatment choices and improve outcomes in AIPmut patients.