Mice with Inactivation of Aryl Hydrocarbon Receptor-Interacting Protein (Aip) Display Complete Penetrance of Pituitary Adenomas with Aberrant ARNT Expression

Abstract

Mutations in the aryl hydrocarbon receptor-interacting protein (AIP) gene have been shown to predispose to pituitary adenoma predisposition, a condition characterized by growth hormone (GH)-secreting pituitary tumors. To study AIP-mediated tumorigenesis, we generated an Aip mouse model. Heterozygous mice developed normally but were prone to pituitary adenomas, in particular to those secreting GH. A complete loss of AIP was detected in these lesions, and full penetrance was reached at the age of 15 months. No excess of any other tumor type was found. Ki-67 analysis indicated that Aip-deficient tumors have higher proliferation rates compared with Aip-proficient tumors, suggesting a more aggressive disease. Similar to human AIP-deficient pituitary adenomas, immunohistochemical studies showed that expression of aryl hydrocarbon receptor nuclear translocator 1 or 2 (ARNT or ARNT2) protein was lost in the mouse tumors, suggesting that mechanisms of AIP-related tumorigenesis involve aberrant ARNT function. The Aip^+/− mouse appears to be an excellent model for the respective human disease phenotype. This model constitutes a tool to further study AIP-associated pituitary tumorigenesis and may be potentially valuable in efforts to develop therapeutic strategies to treat pituitary adenomas.

Pituitary adenomas are common, benign, monoclonal neoplasms of the anterior pituitary gland. They account for approximately 15% of intracranial tumors. Approximately two thirds produce pituitary hormones in excess; among these, prolactin (PRL) and growth hormone (GH)-oversecreting adenomas are the most common. The significant morbidity associated with these lesions arises from the adverse effects of the hypersecretion, such as acromegaly or gigantism in the case of GH secreting adenomas, and/or local compressive effects.¹

Mutations in the aryl hydrocarbon receptor interacting protein (AIP) gene have been identified as an underlying cause in human pituitary adenoma predisposition (OMIM 102200), characterized mainly by GH secreting adenomas (somatotropinomas), although susceptibility to PRL (prolactinomas), adrenocorticotropic hormone (ACTH), and nonsecreting adenomas is also part of the disease phenotype.^2–5 So far AIP mutations have not been associated with any other tumor types.^6–8 Typically, pituitary adenoma predisposition patients have a young age at disease onset but do not necessarily display a strong family history of pituitary adenomas. AIP mutation positive tumors seem to be larger and may have a worse response to somatostatin analogs as compared to sporadic tumors.^4,5,9

Inactivating germline mutations, loss of the normal allele in tumors, as well as recent functional evidence implicate the tumor suppressor role of the AIP gene.^2,5,10 Many of the proteins known to interact with AIP can be linked to tumorigenesis. The best known function of AIP is to stabilize aryl hydrocarbon receptor (AHR) (also known as dioxin receptor) in a cytoplasmic chaperone complex together with heat shock protein 90 (HSP90) and p23. The binding of dioxins or dioxin-like chemicals leads to shuttling of AHR to the nucleus, where it forms a complex with the aryl hydrocarbon receptor nuclear translocator 1 [ARNT, also known as hypoxia-inducible factor (HIF)-1β]. The AHR/ARNT heterodimer regulates the expression of several xenobiotic metabolizing enzymes.¹¹ ARNT is also required by HIF-1α in the nucleus. ARNT/HIF-1α heterodimer regulates several genes involved in tumorigenesis under hypoxia, and HIF-1α is present in many tumors, contributing to angiogenesis, proliferation, metastasis, and resistance to radiation therapy.¹² Thus, the ubiquitously expressed ARNT is an essential partner in the physiological response to chemical toxicants and hypoxia. ARNT participates also in the regulation of estrogen receptor signaling.^13,14 Interestingly, it has been shown that ARNT2, an ARNT homolog, can compensate the lack of ARNT and form a functional complex with HIF-1α under hypoxia. However, it seems that ARNT2 is not capable of cooperating with AHR in the activation of xenobiotic responsive element–dependent genes.¹⁵ ARNT2 was initially classified as being expressed in neural tissue and the kidney.¹⁶ While much is known about the function of ARNT, the expression pattern and dimerization partners of ARNT2 are less clear. Recently, we showed that expression of ARNT is significantly reduced in human AIP-deficient pituitary adenomas, suggesting a link between AIP and AHR/ARNT signaling in pituitary tumorigenesis.¹⁰ However, further studies are needed to unravel the mechanism by which AIP exerts its tumor suppressive action in the pituitary.

Mouse models have been used to study various aspects of pituitary development, function, and disease. A recently published Aip (Ara9) mouse model revealed that homozygous germline Aip mutations are embryonic lethal and homozygous mutant embryos die due to various cardiovascular malformations. In addition, most mice with reduced Aip expression showed failure of liver vein inclusion resulting in reduced liver size.^17,18 However, possible tumor predisposition was not approached in either study.

Here we report a novel conventional Aip knockout mouse, which models the tumor susceptibility caused by human AIP germline mutations. We show that Aip^+/− mice are extremely prone to pituitary tumors, in particular GH secreting adenomas. In addition, we provide evidence that Aip-associated tumors present a more aggressive disease profile and that Aip deficiency in pituitary tumors has a striking effect on the presence of ARNT and ARNT2 proteins, the heterodimerization partners of HIF-1α. Overall, these results indicate that the Aip mouse model provides an excellent model for the human phenotype and suggest that mechanisms of AIP related tumorigenesis involve aberrant ARNT function.

Materials and Methods

Ethics Statement

All of the animal experiments were authorized by the appropriate review committee, and regulations concerning the use of animals in research were adhered to.

Generation of Aip^+/− Mice

To generate the Aip mouse model, embryonic stem (ES) cells containing the gene trap vector construct in an intronic region of genomic DNA between Aip exons 2 and 3 (ENSMUST00000117831) were used (BayGenomics, University of California, Davis, CA).¹⁹ Embryonic Stem (ES) cells were injected into blastocysts, and chimeras were identified. For the generation of congenic mice, inbred mouse strain C57BL/6Rcc was used. Mice were genotyped by multiplex PCR from cDNA and genomic DNA. Total RNA or DNA was extracted from an ear piece using RNeasy mini kit (Qiagen, Hilden, Germany) or DNeasy Blood & tissue kit (Qiagen, Hilden, Germany). cDNA was produced according to standard protocols. Primers for cDNA genotyping were as follows: forward primer cF (5′-GAGAGTTGCCGGACTTTCAG-3′), and two reverse primers; cRa (5′-TCACAGAGGAACTGGGCAAT-3′) and cRb (5′-ATTCAGGCTGCGCAACTGTTGGG-3′) in GeneTrap vector (Figure 1). Length of the amplified wild-type (WT) allele was 205 bp, and mutant allele 278 bp. Primers for genomic genotyping were; one forward primer gF (5′-TGTCTGACTCACTCTTTTCTTTATTCC-3′), and two reverse primers; gRa (5′-AGGAGAGCCAGACAAAACCA-3′), and gRb (5′-CTGGTGAGGCCAAGTTTGTT-3′). WT allele was 177 bp and mutant allele 193 bp. PCR conditions are available on request.

Figure 1.

Construct and validation of Aip-deficient mice. A: Genomic structure of Aip (transcript ENSMUST00000117831) and the gene trap vector construct inserted in the intron between exons 2 and 3. The gray triangle represents the splice acceptor site after mouse En2 intron 1 sequence. B-geo is a fusion gene of β-galactosidase and neomycin, and pA is a polyadenylation signal. Black arrowheads designate genomic PCR primers, and gray arrowheads designate cDNA PCR primers used in genotyping. B: Results of genomic and cDNA PCR genotyping with wt (wild-type) and mut (mutant) bands.

Tissue Preparations

Heterozygous and WT mice were followed up in a cohort study, in which necropsies were performed at 3-month intervals, beginning at 3 months up to 21 months. Mice used in the study had 89% to 100% C57BL/6Rcc genetic background. Analysis in each age group included heterozygous mice and age-matched WT controls from the same litters (2–3 litters per age group). In each age group and in both genotype classes both genders were represented close to 1:1. After CO₂ anesthesia and neck dislocation, tissues were collected. Pituitary gland, thyroid/parathyroid, adrenal glands, pancreas, brain, kidneys, and liver were collected; other tissues were also collected if macroscopic abnormalities were detected. Total weights of the mice and the relative weights of their liver, spleen, and kidneys (organ weight/total weight of mouse × 100) were measured. Small tissues were fixed up to 2 hours and larger ones overnight in cold 4% paraformaldehyde. The pituitary gland was retained on top of the skull and fixed for 1.5 hours followed by decalcification of 2 hours. Fixed tissues were embedded in paraffin and sectioned at 5 μm. Hematoxylin and eosin (H&E) staining of the pituitary gland was performed in approximately every 50 μm to thoroughly analyze the histopathological features of the organ. From other tissues all macroscopic abnormalities were further examined at microscopic level.

Immunohistochemical Analyses

Hormonal status of the tumors was assessed by GH (1/400 dilution, A0570, DakoCytomation, Glostrup, Denmark), PRL (1/4000, A0569, DakoCytomation, Glostrup, Denmark), and ACTH (1/2000, PA1-36035, AH diagnostics, Århus, Denmark) immunohistochemistry. In addition, expression of AIP (1/100, ab48833, Abcam, Cambridge, UK), ARNT (1/50, ab14829, Abcam, Cambridge, UK), ARNT2 (1/100, sc5581/clone M-165, Santa Cruz, Santa Cruz, CA), HIF-1α (1/200, NB100-479, Novus Biologicals, Littleton, CO), estrogen-receptor α (ERα) (1/100, ab80922, Abcam, Cambridge, UK), and Ki-67 proliferation marker (1/250, ab15580, Abcam, Cambridge, UK) was investigated. Heat-induced antigen retrieval was performed in a microwave oven (20–25 minutes) primarily in (pH 6) citrate or alternatively in (pH 9) TE buffer (AIP and ARNT2). Antibodies in the above mentioned dilutions were incubated on slides for 1 hour in room temperature or alternatively overnight in +4°C (GH, PRL, and ACTH). Power Vision rabbit or rabbit/mouse Poly-HRP immunohistochemistry (IHC) Kit with DAB as a chromogen (ImmunoVision Technologies, Norwell, MA) was used for antibody detection. Finally, sections were counterstained with hematoxylin.

The AIP protein and the hormone IHCs were scored either negative or positive. The staining intensity of ARNT, ARNT2, and HIF-1α was scaled as negative (0), weak (1), intermediate (2), or high (3). In the case of macroadenomas the Ki-67 proliferation index (PI = the number of Ki-67–positive cells among the total number of resting cells) was evaluated from 100 to 500 tumor cells in the area of strongest expression. If the pituitary tumor contained less than 100 cells, all of the cells were counted. Only distinctly stained nuclei were considered as immunopositive. ERα intensity was scaled as negative (0), weak (1), intermediate (2), or high (3). The percentage of ERα-positive cells was evaluated on a scale of 0 to 4 (0% = 0, 1 to 25% = 1, 26–50% = 2, 51–75% = 3, and >75% = 4). Finally, the Q-score method (intensity score + % cells stained, range 0–7) was used to quantify the ERα expression.²⁰

LOH Analysis

To assess loss of heterozygosity (LOH), fresh pituitary tumor DNA from two Aip^+/− mice was sequenced. Multiplex PCR was performed with a forward primer targeted in the wild-type sequence (TGTGTGCTTTTGTACCTGTTGT), a forward primer in the insertion (ATGGCAGCACTGCATAATTC), and a reverse primer in the wild-type sequence (AGCATTTTGAGAAAAGAAAAATTAACA). The amplified WT allele was 166 bp and the mutant was 162 bp. Allelic imbalance was scored by comparing the ratios of the allele peak heights in sequencing graphs of 30-bp-long sequence stretches between normal and tumor samples as described previously.^21,22 The cutoffs for LOH were <0.60 and >1.67.

Q-PCR

RNA was extracted from livers of 18-month-old Aip^+/− and Aip^+/+ mice with RNeasy Mini Kit (Qiagen, Hilden), and cDNA was produced by standard methods. The relative expression of Igf-1 (insulin-like growth factor 1) was determined using TaqMan chemistry and 7500 Fast Real-Time PCR system (Applied Biosystems, Foster City, CA). The Igf-1 probe was Mm00439560 (Applied Biosystems) and the relative mRNA copy numbers were normalized against β-actin housekeeping gene (4352341E; Applied Biosystems).

Statistical Analyses

The χ²⁻statistic was used to investigate the deviation of live Aip embryos from the expected Mendelian 1:2:1 ratio. Comparisons between the groups were drawn through the Student's t-test, and in cases of asymmetric variables the Wilcoxon-Mann-Whitney U-test. The Fisher's exact test or Fisher Freeman-Halton extension test was used to test statistical significances. Correlations were assessed by Spearman rank correlation.

Results

Generation of Aip^+/− Mice and Genotyping

The Aip mutation was generated by inserting a gene trap vector construct into an intronic region of genomic DNA between Aip exons 2 and 3 (ENSMUST00000117831) (BayGenomics, University of California, Davis, CA).¹⁹ The inserted vector construct creates an artificial splicing site after 34 codons (34/331) (Figure 1A). No leakage of the construct was observed when tested by Q-PCR and Western blotting.¹⁰ Mice were genotyped by multiplex PCR either from cDNA or genomic DNA (Figure 1B).

The crossings of heterozygous mice yielded one live Aip^−/−, 37 Aip^+/−, and 22 Aip^+/+ embryos when analyzed at the E12.5 stage. This deviates significantly from the expected Mendelian 1:2:1 ratio (χ² = 17.97, P < 0.001). No living knock-out (Aip^−/−) pups were born. The result is consistent with the study of Lin et al where half of the Aip knockout embryos died before E10.5 and the remaining knockout fetuses survived no later than E14.5.¹⁷

Aip^+/− Promotes GH Secreting Pituitary Adenoma Development in Vivo

Altogether 88 heterozygous Aip^+/− and 58 WT Aip^+/+ mice were examined. Among the 88 Aip^+/− mice, 69 mice (78.4%) developed one or more pituitary tumors, whereas only 12 (20.7%) of the 58 WT littermates displayed this tumor type (Figure 2A). Thus, pituitary tumors were significantly more frequent among Aip^+/− mice (P < 10⁻⁶). Aip^+/− mice showed first pituitary lesions, localized in the pars distalis corresponding to the human anterior pituitary, at six months of age. Several macroscopically visible macroadenomas were detected among heterozygous mice at the older age groups (Figure 3, A and B). The pituitary tumor phenotype reached full penetrance at the age of 15 months (Figure 2A). The purity of genetic background did not have an effect on the incidence of pituitary adenomas, as Aip^+/− mice with variable C57BL/6Rcc genetic background (89% to 100%) all developed pituitary adenomas by 15 months of age (data not shown). No differences in pituitary adenoma formation between sexes were detected (P = 0.21, Fisher's exact test, see Supplemental Table S1 at http://ajp.amjpathol.org).

Figure 2.

Incidence of pituitary tumors in Aip^+/− and Aip^+/+ mice. A: Incidence curves of pituitary adenomas detected in 3-month interval groups of Aip^+/− and Aip^+/+ mice. B: Proportions of GH-secreting adenomas in the analyzed groups.

Figure 3.

Normal pituitary gland and pituitary with macroadenoma. A: The normal pituitary gland of a wild-type mouse. B: Macroadenoma of a heterozygous Aip^+/− mouse. The pituitary gland is depicted by a white arrow.

Aip^+/− mice were prone to develop multiple primary pituitary tumors already at the age of six months, whereas multifocal pituitary tumors were detected among the WT mice at much older age groups (15–21 months) (see Supplemental Table S1 at http://ajp.amjpathol.org). The multifocal origin of adenomas was confirmed by sectioning through pituitary glands.

The hormonal status of the tumors was defined by IHC analyses. The majority of the Aip^+/− mice developed GH secreting adenomas (61/69, 88%) (Figure 2B). Also, prolactinomas were relatively common among heterozygous mice (Figure 4, A–C). In addition, two mixed GH/PRL and one ACTH secreting tumor were seen (see Supplemental Table S1 at http://ajp.amjpathol.org). The predominance of GH-secreting adenomas is consistent with the finding that AIP mutations predispose primarily to somatotropinomas in humans. The majority of the adenomas in Aip^+/+ mice were prolactinomas (25/27, 92.6%), but also two GH/PRL tumors were detected (7.4%). WT mice did not develop purely GH secreting adenomas.

Figure 4.

Immunostaining of GH, PRL, and AIP. A: H&E staining of a pituitary gland with two pituitary adenomas. B: GH immunostaining; one GH-positive and one GH-negative pituitary tumor observed in the pars distalis of a heterozygous Aip^+/− mouse. C: PRL immunohistochemistry of the corresponding lesions. Scale bars = 200 μm. Negative AIP staining of GH-positive (D) and PRL-positive (E) adenomas from an Aip^+/− mouse. Scale bars = 100 μm. Tumors are depicted by white arrows. HE, GH, and PRL staining was performed using serial sections from the same Aip^+/− mouse.

Loss of Aip in Pituitary Tumors

Fresh tumor tissue from two Aip^+/− mice was available for LOH studies. The cutoffs for LOH were <0.60 and >1.67 for mutant and WT allele, respectively.²¹ The allele peak ratios in the normal/tumor pairs were 2.19 and 2.08, indicating reduction of the WT allele in these two tumor samples. Complete loss of the WT allele was never seen due to normal tissue contamination.

The expression of the AIP protein in pituitary tumors was studied with IHC. All of the 38 stained GH-positive adenomas from Aip^+/− mice revealed negative AIP immunostaining (Figure 4D). Also the majority of the PRL-positive adenomas (22/26, 84.6%) (Figure 4E) as well as one ACTH immunopositive tumor showed lack of AIP protein. The finding of four prolactinomas with positive AIP staining suggests that some or all of these lesions may have arisen incidentally and were not related to the germline defect. All of the pituitary adenomas in the WT mice showed positive AIP protein expression.

Other Phenotypic Effects of Aip^+/−

Viability or total weight did not differ between WT and heterozygous mice (data not shown). The relative weights (organ weight/total weight × 100) of liver, kidney, and spleen were the same between Aip^+/− mice and their Aip^+/+ littermates up to 12 months. Aip^+/− mice ≥15 months showed a trend toward increased relative organ weights compared to the WT mice, but these differences were not significant (Figure 5, A and B). Twenty-one months observation did not reveal excess of any other tumor type in Aip^+/− mice. Aip IHC for one lung, five liver, and two kidney tumors from Aip^+/− mice was performed. All tumors showed AIP protein expression, suggesting an incidental association. A slight excess of macroscopically visible hyperplasia of adrenal glands was detected in Aip^+/− mice (Aip^+/−; 7/87 versus Aip^+/+; 2/58, P = 0.16). Histopathological examination, however, did not reveal any neoplastic growth.

Figure 5.

Relative organ weights and Igf-1 expression levels of Aip^+/− and Aip^+/+ mice. Weights of kidney, liver, and spleen in (A) male (M) and (B) female (F) mice. Calculations were made by dividing organ weight by total weight and multiplying by 100. Error bars indicate standard deviations. C: Relative Igf-1 expression levels from livers of 18-month-old mice. Aip^+/−, heterozygous mice with GH-secreting adenomas; Aip^+/+GH+, WT mice with GH-secreting adenomas; Aip^+/+ GH-, WT mice without GH-secreting adenomas.

Mice with Aip-Deficient Somatotropinomas Have Elevated Igf-1 Levels

To assess the functionality of GH secreted by Aip-deficient somatotropinomas, expression of Igf-1 in liver was measured by quantitative PCR. Seven 18-month-old Aip^+/− mice and 11 WT mice were studied. Among the WT mice there were three animals having GH/PRL secreting adenoma. The mean Igf-1 expression for seven Aip^+/− mice with GH adenomas was 1.9 ± 0.26 (SD), and for the eight WT mice not having GH secreting adenomas 1.4 ± 0.22 (SD). The relative Igf-1 expression value for the WT mice with GH secreting adenoma was 1.8 ± 0.11 (SD), thus being in line with the expressions measured from the heterozygous mice (Figure 5C). Altogether the Aip^+/− mice with somatotropinomas had significantly elevated Igf-1 expression levels compared to the WT mice not having GH secreting adenomas (P = 0.002, Student's t-test).

Aip-Deficient Pituitary Tumors Show High Proliferation Rate

To evaluate the proliferation rate in Aip mutation positive and negative tumors, Ki-67 IHC analysis was performed. The Ki-67 protein is expressed in all phases of the active cell cycle (G1, S, and M phase) but is absent in resting (G0) cells. The Aip mutation positive tumors had a significantly higher proliferation rate compared to wild-type adenomas (P = 0.014) (Table 1; also see Supplemental Table S2 at http://ajp.amjpathol.org). No correlation between age and proliferation rate was detected (Spearman rank correlation; rho = −0.12, P = 0.59). The hormonal status of Aip-deficient tumors did not unambiguously correlate with the PI values (P = 0.05, Student's t-test), although prolactinomas showed higher average PI when compared to GH tumors, 10.1 ± 3.6 (SD) versus 6.1 ± 4.7 (SD), respectively.

Table 1.

Ki-67 Proliferation Indices (PI) Observed in Aip-Deficient and -Proficient Pituitary Tumors

Genotype	Tumors (n)	Average PI (SD)
Aip+/−			P = 0.01*
GH	16	6.1 (±4.7)
PRL	7	10.1 (±3.6)
Aip+/+
PRL	13	3.6% (±3.1)

^*Two-sided P value with Student's t-test; difference between mutation-positive and mutation-negative tumors.

Estrogen-Receptor α Expressions in Aip-Deficient and -Proficient Tumors

IHC was used to examine the expression of ERα in Aip-related tumorigenesis. The study comprised 30 Aip-deficient adenomas (14 PRL and 16 GH tumors) and eight Aip-proficient prolactinomas. All Aip-proficient tumors and 29/30 of Aip-deficient tumors were positive for ERα. ERα showed a distinct nuclear expression (see Supplemental Figure S1 at http://ajp.amjpathol.org). The Aip-proficient tumors showed significantly higher ERα expressions compared to the Aip-deficient adenomas (GH and PRL), mean 4.6 ± 0.70 (SD) versus 3.8 ± 1.37 (SD), respectively (P = 0.02, Student's t-test). However, no statistically significant difference between Aip-deficient and -proficient tumors was found when only prolactinomas were compared (P = 0.11). ERα expressions did not correlate with proliferation rates or gender (rho = −0.31, P = 0.65, Spearman rank correlation; P = 0.34, Student's t-test, respectively). No statistical significance between ARNT or ARNT2 deficiency and ERα expression was detected (P = 0.37, Student's t-test). ERα was present in all normal pituitaries, and the expressions were similar between Aip^+/− and Aip^+/+ mice, 5.7 ± 0.8 (SD) and 5.6 ± 0.7 (SD).

ARNT Protein Imbalance in Aip-Deficient Pituitary Tumors

ARNT and ARNT2 IHCs showed a total lack of ARNT in 14 and the total lack of ARNT2 protein in 40 Aip-deficient tumors (Figure 6, A–D, Table 2). GH-secreting Aip-deficient tumors showed more often lack of ARNT2 than ARNT (χ² = 7.28, P < 0.007). Remarkably, almost all Aip-deficient tumors expressed only either ARNT or ARNT2 (49/53, 92.5%, P < 10⁻⁵, Fischer's exact test). In contrast, both proteins were present in all Aip-proficient tumors (10/10 in WT animals and 4/4 in heterozygous animals, 100%). No differences in proliferation rates between ARNT (n = 4) and ARNT2-deficient (n = 9) tumors were detected; median 4% (range, 2 to 8%) versus 4% (2 to 10%), respectively (P = 0.73, Mann–Whitney U-test).

Figure 6.

ARNT and ARNT2 immunohistochemical staining in Aip-deficient tumors. A and B: A GH-positive adenoma showing negative ARNT (A) and positive ARNT2 (B) staining. C and D: A PRL-secreting pituitary tumor showing positive ARNT (C) and negative ARNT2 (D) immunoreaction. Scale bars = 50 μm. Tumors are depicted by white arrows.

Table 2.

ARNT and ARNT2 Protein Imbalance in Aip-Deficient and -Proficient Pituitary Tumors

Genotype	Tumors Lacking ARNT, n (%)	Tumors Lacking ARNT2, n (%)
Aip+/−
GH	11/47 (23.4%)	36/47 (76.6%)	P < 0.007*
PRL	3/7 (43.9%)	4/7 (57.1%)
Aip+/+
PRL	0/14	0/14

^*The χ² statistic was used to investigate the loss of either ARNT or ARNT2 in Aip-deficient GH tumors.

To examine the hypoxia response in Aip-deficient and -proficient tumors, HIF-1α IHC was performed. A total of 50 Aip-deficient and 13 Aip-proficient tumors were studied. HIF-1α was present in 46/50 of Aip-deficient tumors and in all proficient samples, suggesting the activation of this oncogenic pathway (see Supplemental Figure S1 at http://ajp.amjpathol.org). Staining intensity averages were 1.8 ± 0.9 (SD) and 2.0 ± 0.4 (SD) in Aip-deficient and -proficient tumors, respectively (P = 0.26, Student's t-test). PRL secreting adenomas showed significantly higher HIF-1α expression when compared to GH-positive adenomas (P = 0.002). Expression of HIF-1α was found to be even between ARNT and ARNT2 negative tumors. (P = 0.92, Fischer's exact Probability with Freeman–Halton extension). Similarly, there was no correlation between HIF-1α intensities and proliferation rates (P = 0.51, Mann–Whitney U-test). For reference, adjacent normal pituitary tissue had a weak cytoplasmic expression or did not display HIF-1α staining.

Discussion

In the present study, an Aip^+/− mouse model was generated to depict pituitary tumorigenesis caused by human germline AIP mutations. Heterozygous Aip mutation increased dramatically the incidence of pituitary adenomas in C57BL/6Rcc mice. First tumors were detected at six months of age. No tumors were detected at the age of three months. It is, however, possible that pituitary adenomas in this age group are relatively rare and the lesions might be too small to detect with routine HE-staining. No excess of any other tumor type was detected. GH-secreting adenomas dominate in Aip^+/− mice even though prolactinomas, two mixed GH/PRL, and an ACTH secreting adenoma were also detected. As compared with human AIP mutation carriers, mixed GH/PRL adenomas were proportionally less frequent in our mouse model. It has been estimated that mammosomatotrophs (cells releasing both PRL and GH) are relatively common in human pituitary (25 to 50%). In mice the percentage of these cells has been reported to be considerably smaller (0.6% to less than 20%), which may explain the relatively low frequency of mixed adenomas in Aip^+/− mice.²³ Overall, the Aip^+/− mouse model greatly resembles the human phenotype reproducing a close to identical tumor phenotype, suggesting that the factors underlying Aip-deficient tumorigenesis are similar in mice and humans.^2,3,5,9 The only major difference observed between the human and mouse phenotype was the complete penetrance of pituitary tumors in mice; all heterozygous mice developed pituitary tumor(s) when aged up till 15 months (Figure 2). To our knowledge, Aip^+/− mouse is by far the most pituitary adenoma prone model currently available, emphasizing the fundamental importance of AIP for tumorigenesis in this organ.^24–26

Acromegaly, resulting from GH secretion, is also known to cause an overgrowth of all organ systems, bones, joints, and soft tissues.¹ GH is known to regulate the abundance of Igf-1 expression in the liver.²⁷ The present study showed that liver Igf-1 expression of somatotropinoma bearing Aip^+/− mice was increased as compared with control animals, supporting the view that the GH secreted by Aip-deficient somatotropinomas is functional (Figure 5C). Along the same lines, we detected signs of increased relative organ weights in Aip^+/− mice ≥15 months (Figure 5, A and B). However, the weight differences were not statistically significant, perhaps due to the relatively small number of mice in each group.

Lin and coworkers reported that 56% of heterozygous Aip mice displayed a reduced liver size and weight due to a patent ductus venosus.¹⁸ In our study, Aip^+/− mice had the same or even slightly increased relative liver weight compared with WT littermates (Figure 5). The reason for the discrepancy between these studies can be differences in the placement of the germ line mutation to produce the Aip inactivation or possibly the different C57BL substrains used for inbreeding.²⁸

Ki-67 analysis showed that Aip-associated tumors had higher proliferation rates as compared with the WT pituitary adenomas (P = 0.014, Table 1, see Supplemental Table S2 at http://ajp.amjpathol.org). Because of lack of tumor material we were not able to perform Ki-67 IHC in human AIP mutation positive tumors. However, the proliferation average detected in tumors of WT mice (3.6%) is well comparable to values detected in sporadic human pituitary tumors (1 to 4%).^29–31 In humans the AIP mutation positive tumors have been reported to be larger and to have a poorer response to somatostatin analogues.^5,9 Our result supports the view that AIP mutation positive adenomas have a more aggressive disease profile. AIP-associated tumors may require more frequent follow-up and development of tailored therapeutic strategies. The estrogen receptor signaling pathway is known to act in biosynthesis and secretion of hormones of the anterior pituitary and to stimulate the proliferation of lactotropes and gonadotropes.³² Protein expression of ERα was uniform between Aip-deficient and -proficient prolactinomas. In contrast, the Aip-deficient GH-positive tumors had lower ERα expression when compared to prolactinomas. This is in accordance with earlier studies reporting that prolactinomas have a tendency to show higher ERα levels as compared with GH-secreting adenomas.^33,34 Although we were not able to compare ERα expression between Aip-deficient and -proficient GH-secreting tumors due to the rarity of WT GH-secreting tumors, these results suggest that ERα may not be a key factor in Aip-associated pituitary tumorigenesis.

Presence of HIF-1α indicated the activation of the hypoxia response both in Aip-deficient and -proficient pituitary adenomas. The level of HIF-1α was significantly higher in Aip-proficient prolactinomas (P = 0.01). It has been, accordingly, reported in humans that prolactinomas have a tendency to show higher HIF-1α protein expression levels as compared with GH secreting adenomas.³⁵

ARNT is a known heterodimerization partner of HIF-1α to form an active HIF-1 complex, but ARNT2 has been shown to be able to compensate the lack of ARNT through binding with HIF-1α.^36–38 Our ARNT and ARNT2 IHCs revealed the total lack of either ARNT or ARNT2 in Aip-deficient mouse pituitary tumors (Table 2). This finding is in agreement with our earlier work where ARNT protein was significantly reduced in human AIP mutation positive pituitary tumors (ARNT2 was not studied in that work).¹⁰ Strikingly, we found that almost always there was loss of either ARNT or ARNT2, but not both, in the Aip-deficient mouse lesions (P < 10⁻⁵). While the mechanisms behind this form of haploinsufficiency remain to be elucidated, this result suggests that the signaling through ARNT and ARNT2 is a key factor in the genesis of pituitary tumors after AIP function is lost.

In conclusion, Aip^+/− mice display a disease phenotype which is strikingly similar to that reported in humans. AIP/Aip germline mutations appear to associate only with pituitary adenomas. In both human and mouse, GH-secreting adenoma is the most common tumor type accounting for 80% and 88% of AIP/Aip associated pituitary tumors, respectively. The dramatically increased somatotropinoma risk associated with heterozygous AIP germline mutations in mice and humans makes AIP an attractive candidate gate keeper gene in somatotrophs. In addition, our data support the previously presented notion of a more aggressive disease profile in AIP-deficient pituitary tumors.

The generation of this disease model provides an important tool to further dissect and elucidate the molecular basis of pituitary tumorigenesis. It is also potentially valuable in efforts to develop therapeutic strategies for management of patients with treatment resistant pituitary adenomas.