AIP gene germline variants in adult Polish patients with apparently sporadic pituitary macroadenomas

Introduction Up to 5% of all pituitary tumors are hereditary e.g. due to MEN1 or aryl hydrocarbon receptor-interacting protein (AIP) genes mutations. Objectives The study was aimed at the assessment of the frequency and characteristics of AIP-mutation related tumors in patients with apparently sporadic pituitary macroadenomas in the Polish population. Materials and methods The study included 131 patients (57 males, 74 females; median age 42 years) diagnosed with pituitary macroadenomas, and with a negative family history of familial isolated pituitary adenoma (FIPA) or multiple endocrine neoplasia type 1 (MEN1) syndromes. Sanger sequencing was used for the assessment of AIP gene variants. The study was approved by the Ethics Board of JUMC. Results AIP variants were identified in five of the 131 included subjects (3.8%): one diagnosed with Cushing’s disease, two with acromegaly, and two with non-secreting adenomas. Patients harboring hereditary AIP gene alterations did not differ from the rest of the study group in median age at diagnosis (41.0 vs. 42.5 years, P=0.8), median largest tumor diameter (25 vs. 24 mm, P=0.6), gender distribution (60.0% vs. 56.3% females, P=0.8), secreting tumor frequency (60.0% vs. 67.5%, P=0.7), or acromegaly diagnosis frequency (40.0% vs.37.3%, P=0.9). Conclusions In our series of apparently sporadic pituitary macroadenomas, AIP gene variant carriers did not differ substantially from patients with negative genetic testing. A risk factor-centred approach to AIP genetic screening may result in missing germline variants. Considering the clinical impact of such genetic variants and their relatively low penetrance, it is, however, doubtful if general genetic screening benefits the whole cohort of pituitary macroadenoma patients and their families.


Introduction
Pituitary tumors, if autopsy and radiological imaging are taken into consideration, occur in about 16.7% of the general population (1). The prevalence of clinically significant lesions is estimated at one case per approximately 1000 individuals (2,3). Most of pituitary adenomas are sporadic (4). Only up to 5% of them are hereditary and can be caused, e.g. by MEN1 or aryl hydrocarbon receptor-interacting protein (AIP) gene mutations, and present as a part of multiple endocrine neoplasia type 1 (MEN1) or familial isolated pituitary adenoma (FIPA) syndromes (4,5).
The AIP gene is a suppressor gene encoding a 330 amino acid protein involved in the cAMP-phosphodiesterases pathway (6)(7)(8)(9)(10)(11). The most common AIP variants (AIPvar) are nonsense and missense mutations, deletions, insertions, splice-site and promoter mutations, and large deletions (6,7). Most of them may result in a truncated protein or, less frequently, affect the tetratricopeptide repeat (TPR) domains or the C-terminal a-helix (6-8, 12, 13). Furthermore, in patients with germline AIPvar, loss of heterozygosity (LOH) has been found in the tumor tissue at the site of the AIP gene in the 11q13 region (6,8). Some of the AIPvar are rare alterations without pathogenic effects, and no impact on protein function. Differentiating between these issues is important because rare genetic changes are also found in healthy controls (7). Whenever AIPvar is used in this manuscript, it refers to pathogenic or possibly pathogenic variants of AIP.
Our investigation was aimed at the assessment of the frequency and characteristics of pathogenic AIPvar-related tumors in a population of Polish patients with apparently sporadic pituitary macroadenomas.

Materials and methods
This was a single-center study. Inclusion criteria were: (a) diagnosis of pituitary macroadenoma, (b) age >18 years, (c) negative family history of MEN1 syndrome, familial isolated pituitary adenomas (FIPA), and other hereditary syndromes with pituitary involvement, (d) informed consent to genetic testing. Between 2013 and 2019, 1134 patients with pituitary adenomas were hospitalized at the Department of Endocrinology specializing in adult care. Finally, after assessment for eligibility, 131 patients were included ( Figure 1). Data on patients' sex, age at diagnosis, tumor size, tumor type (clinical manifestation), pituitary surgery, and other treatments were recorded.
Written informed consent was obtained from all participants. The study design was approved by the Ethics Board of the Jagiellonian University (KBET/119/B/2013).
Genetic testing was performed between 2013 and 2019. DNA was isolated from peripheral blood samples. DNA sequencing was done using the Sanger method on an ABI3500 Genetic Analyzer. Analysis of the results was performed in SeqScape v2.7. All AIP coding exons and adjacent splice sites were analyzed. For variants, the clinical significance and prevalence in the population were determined based on publicly accessible databases (NCBI ClinVar, HGMD, VarSome), with NM_003977 used as a reference sequence (19)(20)(21).
Statistical analysis was performed with IBM SPSS Statistics 28.0. The Mann-Whitney U test was used for comparing the groups due to non-Gaussian distribution of data. P values <0.05 were considered statistically significant.

Patients with AIP gene germline variants
A germline AIPvar was identified in five patients (3.8%): two with c.47G>A (p.Arg16His), two with c.911G>A (p.Arg304Gln) and one with c.684G>A (p.Gln228=). One patient was diagnosed with Cushing's disease (9.1% of 11 ACTH-producing adenomas), two with acromegaly (4.8% of 42 GH-producing adenomas), and two with non-secreting adenomas (4.7% of 43 non-secreting tumors). There was no difference in secreting tumor (P=0.7), or acromegaly frequency (P=0.9) as compared to the group without pathogenic AIPvar. This group consisted of two males and three females (no difference in gender distribution from the AIPvar(-) group, P=0.8). The patients' median age at diagnosis was 41 years (range 23-74 years) and did not differ from the AIPvar(-) group (p=0.8). The median largest tumor diameter (25 mm) also did not differ significantly (P=0.6). Four out of 5 patients were treated surgically (80%), and the surgery was curative in one patient. One out of two patients diagnosed with acromegaly was treated with somatostatin analogues.
A brief description of the AIPvar positive (AIPvar(+)) patients and the predicted impact of the detected AIPvar are presented in Table 1 and Table 2, respectively.

Family screening
Two of the five AIPvar(+) patients agreed for their families to be offered AIP genetic testing. The AIP alteration c.911G>A was found in the asymptomatic mother of patient III, who proved negative on hormonal check-ups and imaging. The second tested person was the underaged son of patient II. He carried an AIPvar of unknown clinical significance, c.47G>A, and has not been clinically screened yet.

Discussion
In this study, we wanted to investigate the frequency and characteristics of germline AIPvar in Polish patients with apparently sporadic pituitary macroadenomas, followed up in the tertiary academic clinical center. We have found AIP variants that are very rare and/or have been described in the databases NCBI ClinVar or HGMD as pathogenic or of unknown significance in 3.8% of the studied group. Similar frequencies of germline AIPvar were found in a study by Cazabat et al. They have found germline AIPvar in 16 (3.6%) out of 443 (aged 4-87 years, both with micro-and macroadenomas) (22). The prevalence of AIPvar(+) patients might be higher in younger (≤30 years) cohorts. In a study by Hernandez et al., 8.4% of young-onset sporadic pituitary adenoma cases harbored pathogenic or likely pathogenic AIPvar (15). In a sporadic cohort of patients diagnosed with macroadenomas ≤30 years or with pituitary adenomas ≤18 years of age, Marques et al. found AIP alterations in 6.8% of cases (23). In contrast, in a Spanish cohort of 235 apparently sporadic adenomas in patients ≤30 years of age, pathogenic AIPvar were detected in 3.8 of cases (24). The higher incidence of pathogenic mutations in some of the above-mentioned studies is mainly because the pediatric population was not included in our analysis. It should be noted that pathogenicity assessment may be based on different criteria, parameters, and tools. The missense variants identified in our study, designated as pathogenic or of unknown significance in the Flow chart of the patients' inclusion. Based on the collected data, we have found that our patients with AIPvar did not differ substantially from the rest of the study group. In our study, the median age at diagnosis in AIPvar(+) patients was 41.0 vs. 42.5 years in the rest of the group. This is inconsistent with most of the available data. In the series by Marques et al., patients with pathogenic AIPvar were 8 years younger at the onset of symptoms and 6 years younger at diagnosis (23). 65% of them were younger than 19 years, and 87% were younger than 30 years. Daly et al. compared patients with or without AIPvar and recorded a larger difference in age at diagnosis: 25.7 vs. 38.8 years, respectively (13). A similar age difference was found by Cazabat et al. (22): 23.5 vs. 40.9 years, respectively; none of the AIPvar carriers was older than 40 years. This contrasts with our group, in which 3 out of 5 patients were older than 40 years. It may be argued that patients were younger at the onset of the disease, nevertheless, the oldest was over 70 years old when diagnosed with Cushing's disease. Data may also be distorted by a large number of patients excluded due to lack of consent to participate in the study, exclusion of children and adolescents, and a delay in referring patients to the Endocrinology Department by other medical specialists due to the lack of specific symptoms.
We noted no difference in median tumor diameter, whereas, in published studies, the size difference is significant in favor of AIPvarrelated tumors, 24.6 ± 10.7 mm vs. 14.5 ± 10.1 mm in genetically unaltered cases (13). This may be because we only enrolled macroadenoma patients, and large tumors are predictive of germline AIPvar (26). In the sporadic pituitary tumors cohort reported by Hernandez-Ramirez et al., there was no difference in the proportion of giant adenomas between AIPvar positive and negative patients (15). All AIPvar(+) sporadic cases had macroadenomas (in contrast to 86.3% in AIPvar(-) group), and presented more frequently extrasellar extension (95% vs. 58.9%, respectively). The available data indicates a small predominance of males (12,16,26) or an equal number of males and females (15, 27) among AIPvar carries. In a large international collaborative study, AIPvar carriers were predominantly males (63.5%) (16). Similar results were obtained in a smaller study of sporadic pituitary tumors (around 61% AIPvar carriers being males) (12). Interestingly, male and female patients showed no phenotypic differences (12,16). The gender proportions in our study were inverted: 3 AIPvar(+) females to 2 males (similarly as in the whole screened group).
We have also noticed a deviation from other cohorts in the tumor types distribution. In our study, GH-secreting tumors were the most common (40% of AIPvar(+) patients and 37% of the rest of the group). The data indicate that GH-secreting tumors appear more often in AIPvar(+) cases, up to 78.1% of some studied cohorts (14,16,26). The proportion of AIPvar(+) acromegalic patients was similar (4.8% vs. 4.1%) as reported by Cazabat et al., and higher in the case of ACTH-producing (9.1% vs. 6.8% in Cazabat's study) or nonsecreting/gonadotropin-secreting adenomas (4.7% vs. 0.9%, respectively) (22). We have not found any AIPvar(+) patient with prolactinoma (in contrast to 4.6% of all prolactin secreting tumors in Cazabat study (22)). The group of Hernandez-Ramirez did not find any AIPvar(+) patients with Cushing's disease, functioning gonadotropinomas, or TSH-omas (15). In their cohort of sporadic AIPvar(+) cases, all patients were diagnosed with acromegaly. In a sporadic cohort of young patients, AIPvar-related tumors accounted for 10.5% of somatotropinomas, 1.3% of prolactinomas, and none of the non-functioning adenomas (23).
Our group of AIPvar(+) patients differs from those reported in the literature in the clinical presentation ( Table 1) and age at diagnosis. This may be explained by the small group size and selection bias, particularly screening-out patients diagnosed during childhood, who comprise a large proportion of the other cohorts (22). Most of the patients who agreed to participate were referred for hospital workup, which may explain the underrepresentation of prolactinomas.
Surgery and SSA treatment are less effective in AIPvar(+) cases (5,16). Leontiou et al. evaluated the response to SSA in AIPvar(+) patients and found that 53% had a poor response to therapy (27). In contrast, resistance to SSA therapy is noted in about 25% of AIPvar (-) patients (27)(28)(29). In another study, more than one-third of patients with AIPvar-related somatotropinomas underwent two or more surgical interventions. Furthermore, only 11% achieved disease control during post-operative SSA treatment; tumor shrinkage was observed in 16% of cases (5). In Marqes et al. study AIPvar-related adenomas more frequently required multimodal and multiple treatments (23). Similar tumor behavior was noticed in our group. Only one patient was followed without any interventions so far. Most of the presented cases, like in other reports (4,5,14,16), required a multimodal therapeutic approach, which frequently did not result in proper disease control. The considered reasons for treatment resistance in AIPvar(+) tumors are defective Gai2 or ZAC1 pathways, mediating SST2 receptors function (18,30). It is likely that decreased AIP expression is the cause of the poor response of GHreleasing tumors to SSA treatment (18,30,31). In our study, we also wanted to assess the impact of the three AIPvar types detected in our group, on the course of the disease and compare the results with the available literature.
Patients I and II harbored the c.47G>A (p.Arg16His) missense AIPvar (rs145047094). In the non-Finnish European population, the variant occurs with a frequency of 0.34%, according to the gnomAD v2.1.1 database (32). The pathogenicity interpretation according to ACMG guidelines (25) (assessed by VarSome (21)), ClinVar and HGMD databases, as well as in silico classifications of the variant are summarized in Table 2.
The summary of published data on the clinical significance of the c.47G>A variant is presented in Table 3. Most authors have interpreted this alteration as a very rare variant without a clear causative effect (Table 3). Interestingly, this variant was noted in patients with Cushing's disease from Poland, as well as in a young  (22,33). Similarly, patient I in our group was diagnosed with Cushing's disease due to an aggressive type of adenoma (Crooke cells in IHC staining: ACTH+, Ki67 about 3%), refractory to treatment. On the other hand, in patient II with acromegaly, the adenoma showed no signs of increased aggressiveness on histopathological examination; however, he did not fulfil all acromegaly remission criteria. This variant was also found in the patient's son, but he is yet to be evaluated clinically. Regarding offspring age, only longitudinal observation may prove if the variant will be pathogenic. Therefore, the impact of the c.47G>A variant on pituitary tumorigenesis still needs to be elucidated. While examining the mechanisms of the pathogenicity of the rs145047094 variant, Baciu et al. concluded that pathogenic variants which lead to premature stop codons, cause truncating AIP proteins and affect important functional domains; however, missense changes can have both pathological and mild effects (37). Pituitary tumorigenesis may require other pathology, for example, the LOH in adenoma cells, which was not observed in the case of the abovementioned variant (13, [33][34][35]. It has also been suggested that this variant lacks functional impact due to the weak binding effect of PDE4A5 (6). In a latest work of Garcia-Rendueles et al., who performed a series of functional in vitro analyses, N-term and C-term AIP point variants were proven to impact the molecular interactions of AIP and block the RET-apoptotic pathway. Based on those criteria, the variant p.Arg16His was classified as pathogenic (42).
Another missense variant, c.911G>A (p.Arg304Gln), was detected in our study in a 26-year-old acromegalic woman and a 56-year-old man with non-secreting adenoma (patients III and IV). It is registered in NCBI dbSNP under the accession number rs104894190. For pathogenicity classification and in silico analyses, see Table 2. This variant occurs at a CpG island hotspot (5-7). The clinical significance of the c.911G>A variant is equivocal (Table 4).
Dal et al. have postulated the role of other genes in the development of pituitary tumors in AIPvar carriers, namely PDE11A (associated with adrenal tumorigenesis) and ALG (coding a protein essential for glycoprotein folding and stability) (48). Patient III from our study had a treatment-resistant tumor and required two surgeries, radiotherapy, and SSA treatment. The mother of the patient had the c.911G>A (p.Arg304Gln) variant but she was negative on biochemical check-up and imaging, which may question the variant's clinical pathogenicity. Patient IV required double surgery and stereotactic radiotherapy, which may suggest that the variant impacts the clinical course of pituitary adenomas. While examining the mechanism of the mutation some authors found that this variant did not significantly reduce PDE4A5 binding (6), other concluded that it may impact AIP and aryl hydrocarbon receptor (AHR) interactions (33) or may interfere with protein stability or folding (49) without directly affecting the protein-protein interaction (6). Other authors observed that this AIP alteration did not disrupt chaperone binding and did not show a significant reduction in bgalactosidase activity, which may reduce its pathogenic effect (6,50).  Proband's older sister and 14 years old daughter (asymptomatic at testing).
(Continued)  (48). In the study by Garcia-Rendueles et al., the functional outcome in transfected cells was similar as in the case of the variant p.Arg16His, therefore, the disease-causing potential of the variant p.Arg304Gln was also proven positive in this study (42). The last variant was c.684G>A (rs1365555914), which was found in patient V. The pituitary tumor in this case was smaller than in the previous patients (I-IV), and subsequent pituitary MRIs did not reveal its progression over the years. The c.684G>A variant itself does not introduce an amino acid change (CAG>CAA, Lys>Lys), however, due to the additional common variant rs641081 (c.682C>A) in homozygous state in the patient, the final codon at position 228 in the protein in this patient changed from CAG to AAA, leading to the amino acid alteration p.Gln228Lys. The p.Gln228Lys alteration caused by the common variant rs641081 alone has been classified as benign, and the more prevalent allele A (observed in a homozygous state in the patient) occurs in the non-Finnish European population with a frequency of 99.8%. The variant rs1365555914 detected in the patient is extremely rare, with no information on its frequency available and zero allele counts reported in the gnomAD database for any population. This rare variant does not change the coded amino acid with any of the alleles of the rs641081 variant and, although there is a noticeable difference in the codon usage frequency for the reference glutamine (35.5 vs 14.1 per thousand for CAG and CAA, respectively), the usage frequency for lysine codons with the rs1365555914 variant, as observed in the patient, are very similar: 31.8 vs 27.5 per thousand for AAG and AAA, respectively (53). It is, therefore, not clear whether or not this variant is of clinical significance for the patient. However, we report it due to its rarity and because it was the only suspected variant identified in the genetic screening of the patient diagnosed at a young age.
The above-discussed AIP alterations illustrate the difficulties in genetic testing results interpretation. The inconsistency in assessing the pathogenic impact of AIP variants is particularly seen in simplex cases (54). Negative family history may be caused by reduced penetration of the variant or lack of information about the family (55). The penetration of pituitary adenoma in AIPvar carriers is described in the range 12-30% (14,55). Some studies even indicate that the pathogenic variant type has little effect on penetration (15,55).
The question remains, what significance for clinical management each type of AIPvar has. Daly et al. asked whether dividing AIP variants into non-pathogenic or pathogenic is useful (56). It is unclear whether AIP alterations are always the main factor responsible for tumor development. It may be that AIP variants only facilitate tumor formation. Variants considered clinically pathogenic may be reclassified as innocent on further analysis (56). Therefore, rare alterations deserve special attention. In our study, regardless of the specific type of the variant and its pathogenicity described in literature, most of the AIPvar-related tumors were aggressive and usually resistant to standard treatment. Other data suggest that the type of changes in the AIP-encoded protein caused by mutation may be relevant for disease course and the decision on family member screening. According to Hernandez-Ramirez et al., truncating mutations are related to a younger age at diagnosis and the onset of symptoms, and a more common occurrence of pediatric cases (15). In this study, there was no difference between truncating and nontruncating mutations in the proportion of acromegaly, the number of patients per family, maximum tumor diameter, or extrasellar expansion. During the genetic screening, new variants/mutations of the AIP gene may be found, and their pathogenicity usually needs to be proven. For example, in the study by Cazabat et al. (22), previously not described variants were found in half of sporadic AIPvar(+) patients. The authors performed parental screening in 7 out of 16 AIPvar(+) patients. In all of them, one of the parents was an asymptomatic carrier. Of note, no pituitary adenoma was found in parents-carriers, who agreed to clinical evaluation (22). Ten new, likely pathogenic mutations were also reported by Hernandez-Ramirez' group (15).
The important question is the probability of pituitary adenoma development in asymptomatic AIPvar carriers. In a group of 160 apparently unaffected AIPvar carriers in the Hernandez-Ramirez study, pituitary adenoma was established in 11.3% of patients (15). Half of them exhibited GH oversecretion, the rest were diagnosed with non-functioning adenomas. Only five out of 18 prospective cases were diagnosed with macroadenomas. The authors concluded that depending on the applied clinical screening, up to 25% of apparently unaffected carriers may develop pituitary adenoma. Marques et al. prospectively followed 187 apparently unaffected AIPvar carriers identified by testing of the first-degree relatives of AIPvar(+) FIPA and sporadic adenoma patients (23). 88.2% of them did not develop a pituitary adenoma during the mean 5.9 ± 3.3 years follow-up. 19 of 22 adenomas were diagnosed at first screening (in 8 cases, retrospective signs which may be attributed to the pituitary tumor were noted), and 3 cases were recognized during subsequent follow-up. The prospectively recognized pituitary adenomas were smaller, 68% being microadenomas, and were associated with lower rates of hypopituitarism at diagnosis, extrasellar extension, or cavernous sinus invasion. Such patients were less frequently operated and none of them required radiotherapy.
Finally, it is worth asking which patients should be screened, considering the varying course of the disease and low disease penetration in patients with AIPvar. Genetic screening was most commonly indicated in the case of (12,14,26): meeting the criteria of FIPA, or pituitary adenoma diagnosed in <18 year-olds, or pituitary macroadenoma diagnosed in <30 year-olds. The probability of detecting new AIP alterations in the fifth decade of life is low (2,12,15). Published data suggest that only 13.2% of AIPvar(+) patients had an onset of illness after 30 years of age (15). Therefore, unselected screening is probably not a cost-effective method (22,27,(57)(58)(59). In large unselected cohorts, AIP pathogenic or likely-pathogenic variants occurred in 3.6%-8.3% of included patients (22,27,57,59,60). If young populations were considered, the incidence increased from 11.7% in patients <30 years to 20.5% in pediatrics (5). Although in our group, AIP variants were detected in older patients with an aggressive course of the disease, the benefit for the individual patient over 40-50 years of age from genetic screening is negligible if the family history is negative. It seems that the detection of an AIPvar in this setting does not impact the patients' management, as the AIP-encoded protein is not a therapeutic target currently. Even if a more aggressive course of adenomas in AIP germline variant-carriers may be predicted, the treatment modalities did not differ from those applied in large to giant, invading or drug-resistant pituitary tumors unrelated to AIP alterations.
In a systematic review by van den Broek et al. (26), the following recommendations were proposed based on available studies. The authors strongly recommend against routine genetic testing in sporadic pituitary adenoma. A weak recommendation for AIP mutation analysis in patients with sporadic pituitary adenomas 30 years old or younger, especially those diagnosed with acromegaly and gigantism, was made based on low-quality evidence.

Conclusion
In conclusion, in our series of apparently sporadic pituitary macroadenomas, AIPvar carriers were identified in 3.8% of the study group and did not differ substantially from patients with negative genetic testing. Therefore, routine genetic screening for AIP variants in non-selected adult pituitary adenoma patients seems currently ineffective. It seems that in clinical practice, a targeted screening approach limited to patients at risk of AIPrelated pituitary adenomas should be applied. If an AIP variant is detected, genetic testing should be discussed with the proband's family members (particularly symptomatic and younger ones), as additional data may improve our understanding of the clinical significance of detected genetic alteration.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Funding
The study was funded by the Jagiellonian University Medical College grant No K/ZDS/003795.