Novel somatic variants involved in biochemical activity of pure growth hormone-secreting pituitary adenoma without GNAS variant

We aimed to identify somatic genetic alterations in pure growth hormone (GH)-secreting pituitary adenomas without GNAS variants. Patients with GH-secreting pituitary adenoma who underwent transsphenoidal adenomectomy at Severance Hospital, Yonsei University College of Medicine were recruited. Somatic genetic alterations were profiled by whole-exome sequencing (WES) and targeted resequencing. WES was performed using DNA from nine GH-secreting pituitary tumors and corresponding blood samples. Absence of GNAS variant was confirmed by Sanger sequencing. For targeted resequencing of 140 fixed tissues, 48 WES-derived candidate genes and 7 GH-secreting pituitary adenoma-associated genes were included. Forty-eight genes with 59 somatic variants were identified by WES. In targeted resequencing, variants in 26 recurrent genes, including MAST4, PRIM2, TNN, STARD9, DNAH11, DOCK4, GPR98, BCHE, DARS, CUBN, NGDN, PLXND1, UNC5B, and COL22A1, were identified, but variants in previously reported genes were not detected. BCHE, DARS, NGDN, and UNC5B variants were associated with increased GH-secreting pituitary tumor biochemical activity, which was confirmed in vitro. Although recurrent point variants were rare, several somatic variants were identified in sporadic pure GH-secreting pituitary adenomas. Several somatic variants may affect pathways involved in the tumorigenesis and biochemical activities of GH-secreting pituitary adenomas.


WES.
WES was performed to identify novel genes related to pure GH-secreting pituitary adenoma. Based on our aims, nine patients not carrying a variant at codon 201 or codon 227 of GNAS were selected for the cohort (Supplementary Table S1). Tumor tissue and blood samples were sequenced, and the data were analyzed with bioinformatics software (Fig. 2a).
Although the Ti/Tv ratio observed (47/53 = 0.89; Fig. 2b) was lower than the commonly reported ratio of ~ 2, Sanger sequencing of several mutant loci agreed with the data (Supplementary Fig. S1).
Targeted resequencing. We performed targeted sequencing to validate variants in the targeted genes ( Fig. 3a). Among the 140 genomic DNA samples prepared from formalin-fixed, paraffin-embedded (FFPE) tissues, those carrying GNAS variant were confirmed by Sanger sequencing and excluded; 50 samples remained in  In the first stage, "Target Discovery Stage, " blood and tissue samples of 30 patients were analyzed by Sanger sequencing, whole-exome sequencing (WES), and single nucleotide variant (SNV) validation resequencing, and genes with somatic variants were identified as target genes for profiling. In the second stage, "Target Profiling Stage, " target genes from 140 FFPE samples were analyzed by targeted resequencing, and relevant variants were selected for further validation. In the third stage, "In vitro Validation Stage, " selected variants were functionally examined by in vitro validation. Expression vectors containing variants were constructed and transfected into GH3 cell lines (drawn in Microsoft 365 Powerpoint; https:// www. micro soft. com/ ko-kr/ micro soft-365). www.nature.com/scientificreports/ the validation cohort. The samples were captured and sequenced to obtain high-depth sequencing data, which were analyzed to profile variants. Twenty-six genes carrying variants were identified (Fig. 3b, Supplementary Fig. S2). MAST4 was the most frequently mutated gene, found in 8% (4/50) of GNAS-negative patients. PRIM2, TNN, STARD9, DNAH11, DOCK4, and GPR98 were found in 6% (3/50) of the cohort, and BCHE, DARS, CUBN, NGDN, PLXND1, UNC5B, and COL22A1 were found in 4% (2/50) of the cohort. However, no variant was shared between any two samples, although frequently mutated genes were found.   www.nature.com/scientificreports/ quencing were queried, and five high-ranked terms were compared (Supplementary Tables S3 and S4). Amyloidbeta binding (GO:0,001,540) terms in the molecular function category that are associated with GH-secreting pituitary adenomas 14 were commonly found in WES and targeted resequencing, and BCHE and CLSTN1 were related to the terms. Additionally, G-protein coupled receptor-related terms were observed in the WES results. G-protein coupled amine receptor activity (GO:0008227), G-protein coupled neurotransmitter receptor activity (GO:0099528), and G-protein coupled acetylcholine receptor activity (GO:0016907) terms were detected, with CHRM2 and OR6T1 related to the terms. Considering that CHRM3 is a paralog of CHRM2 15 , we predicted that CHRM2 may be important. However, only a small number of genes was related to the GO terms, whereas enriched GO terms analyzed from previously reported genes (Supplementary Table S5) resulted from a much larger number of related genes (Supplementary Table S6). Therefore, we analyzed the relationship between the variants and clinical manifestations to investigate whether the variants were associated with a phenotype, such as GH secretion. The clinical manifestations of 50 patients selected for targeted resequencing are presented in Table 1. There was no significant difference in tumor size, age, sex, or Ki67 index according to the genetic variants. However, some variants were associated with the biochemical activities of GH-secreting pituitary adenoma, such as BCHE, DARS, NGDN, and UNC5B variants (Table 1). Although significance could not be calculated because of the limited number of subjects, the patients presented exceptionally increased GH and IGF-1 levels before transsphenoidal adenomectomy. Variants were validated by Sanger sequencing (Fig. 3c, Supplementary Fig. S3). The patients had no family history of pituitary adenomas or other neuroendocrine tumors.
As transfection of mutants in GH3 cells significantly affected the signaling pathway of GH synthesis and secretion, we analyzed the concentration of GH secreted into the cell-culture media by enzyme-linked immunosorbent assay (ELISA). Among the several mutants, transfection of BCHE p.P320L (16%, p-value = 0.018), NGDN p.S79C (55%, p = 0.027), and NGDN p.R100H (50%, p = 0.047) mutants significantly increased GH secretion in  www.nature.com/scientificreports/ cell-culture media compared with the vector used as a negative control (Fig. 4c). The amount of GH secreted in the media was normalized to the optical density values obtained for the same samples from the MTS assay (Fig. 4c). Furthermore, the effects of each mutant on cell proliferation were analyzed by MTS assay under the same experimental conditions. Compared with negative control-transfected cells, transfection of BCHE p.P320L (6%, p = 0.001), BCHE p.W571C (10%, p = 0.001), NGDN p.S79C (10%, p = 0.012), and NGDN p.R100H (11%, p = 0.030) mutants slightly increased cell proliferation (Fig. 4b). Based on the ELISA and MTS assays, the variants in NGDN and BCHE may affect GH secretion and tumor growth in GH-secreting pituitary adenomas. Finally, these variants were confirmed by Sanger sequencing with human GH-secreting pituitary adenoma (Fig. 3c,  Supplementary Fig. S3).  www.nature.com/scientificreports/

Discussion
In this study, we identified novel somatic variants associated with pure GH-secreting pituitary adenoma from patients with neither GNAS variants nor a family history. The following are the major findings of our study: (a) 48 genes with 59 somatic variants were identified among nine patients using WES; (b) targeted resequencing revealed variants in 26  In GO analysis, only a small number of genes was related to GO terms, whereas enriched GO terms for previously reported genes showed significant overlaps with pituitary adenomas and acromegaly-related terms (p = 1.3 × 10 -60 and 1.6 × 10 -28 , respectively). This indicates that previously reported genes are rarely observed in other disease categories, and that current variants found in this study do not overlap with those found previously. Although previous databases included mostly western populations, only East Asians were analyzed in this study 11,15,16 . In several studies, ethnic differences were significantly associated with clinical manifestations and genetics of GH-secreting pituitary adenoma [17][18][19] . Furthermore, we analyzed pure GH-secreting pituitary adenomas without GNAS variant rather than general sporadic cases. The difference in a subject's ethnicity and inclusion criteria may be responsible for the differences in the GO results. Also, while this manuscript was being reviewed, the variants were searched again in the recently updated COSMIC 20 and ClinVar 21 databases and most of the variants have not been reported or rarely observed and not reported as significantly pathogenic (Supplementary Table S7).
The underlying pathophysiology of pituitary tumorigenesis is characterized by a combination of disordered cell proliferation concomitant with dysregulated hormone hypersecretion. GH-secreting pituitary adenoma formation results in unrestrained somatotroph proliferation associated with intrinsic cell cycle dysfunction as well as activation of cAMP signaling 1 . Most tumors are caused by heterogeneous variants, some of which are involved in the same molecular pathways and can promote tumor development 22 . Although variants with high frequency or hotspots, such as GNAS, were not identified, variants in recurrent genes among 50 GH-secreting pituitary adenomas suggested that these variants are involved in GH production from somatotroph cells. Among the gene lists of targeted resequencing, variants in 14 genes were observed in at least two or more patients. Regarding kinase activity, MAST4, TNN, PLXND1, and COL22A1 are directly involved in the serine/threonine kinase family and PI3K-Akt and Akt signaling pathways, which can induce tumor formation in GH-secreting pituitary adenoma (www. genec ards. org) [23][24][25] . Furthermore, several mutated genes can be modulated by transcription factors associated with GH-secreting pituitary adenomas. For example, PRIM2, DOCK4, and DARS can be modulated by STAT family proteins; COL22A1, NGDN, and UNC5b by GATA family proteins; and DNAH11 by aryl hydrocarbon receptor and translocator, which induce somatotroph tumor formation (www. genec ards. org).
GH secretion is controlled by Ca 2+ and adenosine ATP signaling, and both signals appear to be frequently activated in GH-secreting pituitary adenomas. Activated cAMP and Ca 2+ signaling lead to an increase in the cytosolic free calcium concentration in pituitary tumor cells 26,27 . Moreover, ATP, which is simultaneously released with pituitary hormones, induces an increase in the intracellular free calcium concentration in pituitary cells 28 . For example, Gsp variants induce constitutive activation of adenylate cyclase to convert ATP to cAMP. An increase in the intracellular cAMP concentration, in turn, leads to activation of protein kinase A and further mitogenic signaling in GH-secreting cells 8 . From a clinical perspective, GNAS-mutated GH-secreting pituitary adenomas present significantly increased biochemical activity among patients with acromegaly. Until now, genetic alterations associated with biochemical activity other than GNAS variants have not been reported. In this study, novel gene variants were identified in GH-secreting pituitary tumors and associated with biochemical activities. BCHE encodes butyryl cholinesterase and deacylate ghrelin, which stimulate GH secretion from somatotrophs 29 , and DARS encodes aspartyl-tRNA synthetase and is modulated by CXC chemokine receptor (CXCR) 7 in somatotrophs. Several studies reported that the expression of CXCR7 is increased in prolactinoma, and that GH-secreting pituitary adenomas and CXCR7 reduce the activity of DARS 30 . NGDN encodes the E1F4E binding protein and, in leukemia cell lines, NGDN knockdown activates the mTOR pathway 31 . UNC5B encodes a member of the netrin receptor family and is involved in angiogenesis. Furthermore, UNC5B was recently associated with aggressive behavior in null pituitary adenoma and silent subtype III pituitary corticotroph adenomas 32 . However, there have been no reports on these genetic variants in GH-secreting pituitary adenoma.
This study had several limitations. Although variants in recurrent genes were identified, functional studies of all variants were not performed. Nevertheless, several variants in four genes influencing the biochemical activities of GH-secreting pituitary adenoma were introduced in GH3 cells, and functional studies were performed. Second, the frequency of variants was relatively lower than that of previously reported somatic variants, such as GNAS variants in GH-secreting pituitary adenomas and USP8 variants in Cushing's disease. However, apart from GNAS, acromegaly-associated somatic variants are not well-established. Several studies have failed to demonstrate the clinical significance or recurrent variants of the reported somatic variants in sporadic GHsecreting pituitary tumors 11,15,16 . Furthermore, we found that variants actually promoted GH secretion in vitro in rat GH-secreting pituitary adenoma cells and human mutant vectors, showing high similarity between both human and rat sequences. Third, we excluded patients with GNAS variants and evaluated pure GH-expressing pituitary tumors in patients with acromegaly. In some conditions, GH-secreting pituitary adenomas are associated with somatic variants that combine with GNAS, except for hotspot variants; however, we sought to discover novel variants that function independently. Follow-up studies are needed to investigate the somatic variants found in this research. Finally, we did not evaluate granulation pattern of GH secreting pituitary tumors. Several studies reported that sparsely granulated type is associated with mutations in the extracellular domain of www.nature.com/scientificreports/ the growth hormone receptor and densely granulated type present more frequent GNAS mutation 33,34 . Further studies evaluating genetic alterations according to granulation type would be necessary.
In conclusion, several somatic variants were identified in sporadic pure GH-secreting pituitary adenomas from patients who had neither GNAS variants nor a family history. Furthermore, GH-secreting pituitary tumors with variants in BCHE, DARS, NGDN, and UNC5B presented high biochemical activities including GH and IGF-1. Although no novel recurrent point variants were identified, the observed somatic variants may affect pathways involved in tumorigenesis and biochemical activities of GH-secreting pituitary adenomas. Further studies are needed to determine the functional mechanism of variants in selected genes that affect GH secretion and pituitary adenoma formation.

Methods
Patients. Among patients in the Severance Hospital Pituitary Tumor Cohort, 232 subjects received transsphenoidal adenomectomy following a diagnosis of GH-secreting pituitary adenoma from January 2012 to January 2017 at Severance Hospital, Yonsei University College of Medicine, Seoul, Korea. All the enrolled patients were single ethnicity, Korean. GH-secreting pituitary tumors were confirmed by measuring serum insulin-like growth factor-1 (IGF-1) and nadir GH levels during the 75-g oral glucose tolerance test. Immunohistochemical staining of the operated pituitary tumors was performed for GH, prolactin, luteinizing hormone, follicle-stimulating hormone, thyroid stimulating hormone, and adrenocorticotropic hormone. Pituitary tumors expressing hormones other than GH were excluded from this study. Finally, the tumor tissues of 140 patients with acromegaly expressing only GH were evaluated. No patients had a family history of pituitary adenoma. Pituitary tumors were classified based on dynamic magnetic resonance imaging of the sella turcica and parasellar region according to the modified Hardy radiological classifications, as described previously 35 .
This study was conducted in accordance with the Declaration of Helsinki and approved by our institutional review board and Ethics Committee (Yonsei University Institutional Review Board and Ethics Committee: approval # 4-2011-0740). Written informed consent was required and obtained from all patients.

WES.
For WES, 29 patients agreed to tissue collection during transsphenoidal adenomectomy. Genomic DNA samples from fresh frozen tissue were extracted using a DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany). Variants at codons 201 and 227 in GNAS were tested by PCR followed by Sanger sequencing (Macrogen, Seoul, Korea). PCR and Sanger sequencing primers were designed to cover both hotspot variants of GNAS (Supplementary Table S8). Eight GNAS variant-negative pituitary tumor samples and one additional sample prepared from FFPE tissue were subjected to WES, and each normal genomic DNA sample from blood was extracted using a DNeasy Blood & Tissue Kit (Qiagen). Shotgun DNA libraries for sequencing were constructed, and WES was performed (Celemics, Seoul, Korea) using the SureSelect Human All Exon V4 + UTRs kit (Agilent Technologies, Santa Clara, CA, USA). Each library was pooled and sequenced with a Hiseq2500 (Illumina, San Diego, CA, USA).
Targeted resequencing and data analysis. For single nucleotide variant (SNV) validation, probes were designed to cover 60 bp on either side of the targeted SNV, synthesized from microarray, and in vitro-transcribed to contain biotinylated UTP using a MAXIscript™ T7 Transcription Kit (Thermo Fisher Scientific, Waltham, MA, USA). Target regions were captured from tumor tissue samples used in exome sequencing, which was performed at Celemics, Inc.
Raw sequencing data (FASTQ) were aligned using Novoalign (v2.07.18). SAM files were converted to BAM format using SAMtools (v. 0.1.18). Each genotype of target loci was piled-up using the mpileup command in SAMtools and compared to exome sequencing data. Variants with a variant allele frequency > 0.1, total depth > 15, and variant depth > 2 were selected.
Statistical analysis. Continuous and categorical variables are expressed as the medians (range). The data were analyzed using the SPSS software package for Windows (v. 25.0, IBM; https:// www. ibm. com/ kr-ko/ analy tics/ spss-stati stics-softw are). The Mann-Whitney U test and Kruskal-Wallis test were used for continuous variables, and linear-by-linear association for categorical variables was performed as appropriate. All statistical tests were two-tailed, and a p-value < 0.05 was considered significant 36 .

Data availability
All data generated or analyzed during this study are included in this published article or in the data repositories listed in References. www.nature.com/scientificreports/