Introduction

Pituitary neuroendocrine tumor (PitNET) or previously often described as clinically significant pituitary adenoma is a rare non-metastasizing endocrine neoplasm affecting approximately 1 person out of 1000 in the general population [1]. Hormone secreting PitNET can cause overproduction of growth hormone (GH), adrenocorticotropic hormone (ACTH), prolactin (PRL) or rarely other hormones leading to various systemic endocrine disorders (acromegaly, Cushing’s disease, and others). PitNETs without disruption of patient’s hormone profile (previously called non-functioning pituitary adenoma) is a subgroup of PitNETs which does not cause disbalance of pituitary hormones. They are usually diagnosed later in life than hormones producing PitNETs and usually due to symptoms related to the effects of the proliferation of PitNET cells on surrounding tissue [2, 3].

High genetic heterogeneity has been observed in pituitary PitNET tissue. Newey et al. have sequenced exomes of somatic tissue of seven PitNETs which do not secrete hormones tissues and found 24 potential somatic candidate variants with an average of 3.5 variants per tumor (range from 1 to 7 per neoplasm). They also performed targeted sequencing of potential drivers in PDGFD, NDRG4, ZAK in an independent sample group of 24 tumors, but found no somatic variants in the identified candidates [4]. A study that conducted exome sequencing of 36 somatotroph PitNET tissues discovered 132 somatic variants with average 3 variants per sample ranging from 0 to 13 in the tumor. The GNAS mutations were found in 11 tumors (31.4% from 36) and several somatic variants in genes involved in cAMP and calcium signaling were identified [5]. Genome sequencing of 12 somatotroph PitNET have also found mutations in GNAS, but no other recurrent somatic variants, and in this report authors demonstrated average 129 somatic variants per tumor genome (range 11–273) and 2.3 variants per exome [6].

Song and colleagues have performed one of the largest studies to date analyzing the genetic landscape of 125 PitNETs [7] identifying GNAS and USP8 mutations repeatedly confirming the role of these genes in PitNET development, as reported by other studies [5, 8–10]. This study also indicated variants in genes NR3C1, KIF5A, GRB10 as potential recurrent candidates, and identified MEN1 somatic variant in two patients without familial MEN syndrome. The authors also found enrichment in somatic variants in Raf/MEK/ERK, PI3K/AKT/mTOR, cAMP, and other signaling pathways [7].

Exome sequencing of eight thyropinomas have not revealed convincing recurrent somatic variants [11] in this study group, but exome analysis of larger group of 42 PitNETs indicated several mutations and insertions/ deletions in genes ATAD3B, BHLHE22, KDM2B, OR5M1, TTN, VPS13B that were encounter in more that one PitNET [12]. No recurrent somatic variants were found in the whole-exome sequencing of 12 prolactinomas [13]. Nemeth et al. have studied mitochondrial variants in 44 PitNET cases and found no significant association with clinical features of the tumor, however, they demonstrated that samples with the highest homoplasmic variant rate had the highest Ki-67 index [14].

Recent study of 134 patients demonstrated novel molecular PitNET classification with deviations from cell lineage guidelines of 2017. High amount of somatic variants per exome (range: 14–247) was reported and recurrence of USP8 and GNAS variants confirmed, as well as presence of chromosomal alterations [15].

Available literature data shows that the number of overlapping candidate genes harboring somatic variants in PitNET like GNAS and USP8 have been confirmed in several independent studies [5, 7, 9, 10], but most of the discovered somatic variants have been detected in the single tumor and have not become novel targets for deeper investigation of PitNET development. Genes that are involved in the development of inherited pituitary PitNET (AIP, MEN1, CDKN1B, PRKAR1A) have mostly shown not to contain somatic variants in sporadic cases, with rare exceptions [7, 12]. Additional studies using independent sample groups and larger sample sizes are required to identify overlapping candidate gene to be further investigated for their role in PitNET development mechanisms and as potential targets for improvement of therapy and patient’s health care.

In this study, we performed exome sequencing of fifteen PitNETs and used two somatic variant detection algorithms for improved yield and reduced impact of algorithm specific biases to discover somatic variants. We discovered two recurrent somatic variants and two recurrent genes within our sample as well as 13 genes containing somatic variants which are overlapping with other independent studies. We believe that the sequencing data analysis of PitNET benefits from using multiple variant calling approaches by providing higher confidence in variants that were detected with different methods.

Materials and methods

Library preparation and exome sequencing

Exome sequencing was carried out as an outsourcing service to commercial sequencing provider Genera Ltd. (Latvia). DNA libraries were constructed using Illumina Nextera TruSeq Exome kit (Illumina, USA) and sequenced with Illumina NextSeq 500/550 High Output v2 kit (150 cycles) (Illumina, USA) obtaining 75bp paired-end reads on Illumina NexSeq 500 sequencer (Illumina, USA). Exome sequencing was performed in five batches (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Batch information of PitNET exome sequencing.

https://doi.org/10.1371/journal.pone.0265306.t001

Results

Characteristics of PitNET samples

We obtained whole-exome sequencing data of paired white blood cells and PitNET tissue from 15 patients. Mean depth of coverage was 50.3X (range 15.9–82.2X, SD = 17.2). Individual PitNET patient information is shown in Fig 1. Clinically nine PitNETs were defined as NFPA, three were GH secreting and two PRL secreting tumors. Immunohistochemistry analyses identified that PIT1 cell lineage marker dominates in four PitNETs, gonadotroph PitNET marker NR5A1 (SF1) is the most abundant in five tumours and corticotroph cell lineage marker TBX19 (TPIT) has highest expression in one PitNET (Fig 1 and S1 File). All PitNETs were larger than 10 mm in at least one dimension in MRI data measurement and 60% of patients were men.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Amount of somatic variants, clinical information of PitNET patients (PitNET clinical type, age at diagnosis, sex), % of cells expressing lineage markers and genes with recurrent somatic variants are shown for all 15 PitNET patients.

PitNETs are listed in ascending order from left to right.

https://doi.org/10.1371/journal.pone.0265306.g001

Identification of somatic exome variants

Results of exome sequencing of 15 PitNETs show a total of 144 PitNET specific somatic variants (mean = 9.6, range 0–31 per PitNET) of which all were SNVs. We employed two somatic variant search algorithms (GATK and Strelka2) and reported only somatic variants where both of the employed algorithms agreed.

Most of the discovered variants (N = 91) are located within the exons of genes and 58 of those are changing the amino acid sequence of the protein (range per tumor 0–11, mean 3.9) while 22 are synonymous SNVs, rest being 3 prime UTR variants (N = 4), 5 prime UTR variants (N = 3) and non coding transcript exon variant (N = 4).The most common location of variants outside the coding sequence were introns (N = 44). The rest of the variants were located in untranslated gene regions, non-coding transcripts, predicted splice sites and outside genes altogether (S1 Table).

Assessment of possible functional consequences of SNVs identified eight SNVs from seven PitNETs which were rated within the top 0.1% deleterious SNVs in the human genome according to Combined Annotation Dependent Depletion (CADD) matric (CADD Phred score >30) and further 34 SNVs within the top 1% deleterious SNVs in the human genome (Phred score >20). All but two of the analyzed PitNET samples contain at least one highly deleterious SNV (Phred score >20) (S1 Table). Three PitNETs have a somatic previously known PitNET associated GNAS mutation (position R844C in transcript ENST00000371100.4). Three genes (GNAS, AHNAK and PAML2/AKAP2) have a mutation in at least two PitNETs but in different positions. While transcribed RNA pseudogene AC002519.6 contains the same genetic variant in the same position in two PitNETs (Tables 2 and 3). Additionally, we identified 13 overlapping genes with other PitNET and neoplasm genetic studies (GNAS [8], CRTC3, SMARCA4 [24], KIT [25], PLA2G6, TESK1 [5], TSC2 [4], KLHL4, RYR1 [6], POLR3B, KHDRBS2, LAMA1, TUBGCP6 [7] and AHNAK [15]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Somatic variants detected in more than one PitNET sample.

https://doi.org/10.1371/journal.pone.0265306.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Somatic variants found in the same gene and in more than one PitNET sample.

https://doi.org/10.1371/journal.pone.0265306.t003

Copy number analysis was performed using CoNIFER [21] and XHMM [22] algorithms did not reveal presence of copy number alterations in analysed samples.

Discussion

This is the first study demonstrating the advantages of exome sequencing analysis using multiple data analysis workflows in the research of PitNET genetic composition. Despite the existing knowledge that in the exome and genome sequencing variant calling results from the single pipeline are incomplete [19, 26], no current literature data is examining the consequences of this for the discovery of PitNET somatic variants. It has been demonstrated that non-overlapping variants from different pipelines mostly are true positive nucleotide changes [19] yet most raw non-overlapping calls of both GATK and Strelka2 in our data showed low alternate allele frequency and other variant calling issues. Therefore, our results provide a more comprehensive approach to identify the genetic composition of PitNET. Exome analysis of blood–tumor pairs from 15 PitNET patients using two distinct data analysis workflow showed increased mean number of somatic variants per PitNET (9.6) compared to previous publications (range 2.3–3.5) [4, 6, 7] and is supported by two of the latest reports [12, 15] showing number of somatic variants often above 10 per tumor. Such an improved approach can lead to the identification of novel mechanisms of PitNET development. Although studies published in literature already encompass several hundred exomes and genomes, recurrent somatic variants beyond established risk factors (GNAS, USP8) are virtually non-existent in sporadic PitNET cases. This could be explained that PitNETs form according to Knudson’s two hit hypothesis [27] and large amounts of nucleotide change combinations can result in a PitNET. Also it is possible that constraints of current exome and genome sequencing technology and analysis algorithms are unable to find certain amounts of somatic variants, therefore important findings are missing from literature. The third reason for low overlap in results of the published PitNET genomic studies could be bias introduced by sampling as surgery materials are obtained mostly from patients with PitNETs which are unresponsive to drug treatment.

Regarding IHC results confirmed that most of the PitNET presented clinical phenotype according to their expressed lineage markers and hormones. Only one of the IHC tested clinically non-functional PitNET (PN02) differed from the rest by having TPIT cell lineage marker ekspressed in 30%– 70% of observed cells rather than NR5A1. On the other hand, PN10 was presenting hyperprolactinemia phenotype, but the highest proportion of cells (>70%) was expressing CGA and NR5A1 indicating gonadotroph PitNET [28], nevertheless, PN10 was also expressing POU1F1 cell lineage marker and PRL in up to 30% of cells, probably causing more noticeable hyperprolactinemia induced clinical phenotype. Also we observed that four out of five PRL or GH secreting PitNETs were expressing more than one cell lineage marker while the same was true in two out of six PitNETs which were designated as non-functioning PitNET. Similar differences between secretion subtype and cell lineage markers has been observed in other studies [15] and they are of low concern regarding detection of somatic tumor variants.

We have, for the first time, detected recurrent somatic variant located in exactly the same position in an AC002519.6 of the two PitNETs (PN01 and PN07) (Table 2). The variant is listed in dbSNP with identifier rs111534922, but has 0 MAF across all populations. As a transcribed RNA pseudogene, there is little information about it in literature [29].

Total transition to transversion ratio was 1.3. This is in line with literature data where variability of transition/transversion ratio across PitNET is high, although the majority have it between 1.2 and 3 [6].

It is fairly common to observe somatic mutations in GNAS, USP8 and MEN1 when performing WES or WGS in PitNET patients [6, 7, 10]. We also detected common GNAS somatic mutation at chr20:57484420, rs11554273 (also known as GNAS 201 R>C) in three PitNET patients (all expressing PIT) (Table 3). GNAS rs11554273 has virtually zero frequency of a minor allele in the general population sample, but is repeatedly found in PitNET tissues [6, 7] and it is heterogeneously impacting clinical properties of somatotroph PitNET [30].

The only other recurrent gene with somatic variants in more than one tumor was AHNAK (PN01 and PN02), but impact on gene function for both variants is hard to pinpoint due their supposedly benign consequence (synonymous variant and 3’ untranslated region variant) and low predicted consequence phred score by CADD (0.009 and 0.59 respectively). Interesting that somatic variants in AHNAK have been reported before in PitNET [15]. AHNAK previously has been implicated in both pro-tumorigenic [31–33] and anti-tumor [34] roles of which the most interesting in PitNETs would be its involvement in epithelial-mesenchymal transition [35]. Nevertheless, its role in PitNETs should be defined in additional, more focused studies. Another way to perform the research of somatic variants of PitNET is to evaluate their presence in the analyzed sample as an allelic fraction. Although limitations of this approach are clear (random distribution of allelic fraction, higher false discovery rate, allele-specific library amplification and PitNET tissue sample pollution with non-tumor cells), we speculate that at least part of somatic variants with allelic fraction below 20% (designated as “subclonal”) arose as a result of increased rate of cell division and diminished capability to repair DNA replication errors in an already established PitNET. This may indicate that the majority of true subclonal variants are present only in part of PitNET cells and promotes a question if these genetically heterogeneous cell groups have an impact on overall PitNET properties and clinical characteristics. Fifty-eight (40.35%) of detected somatic variants in this study were subclonal, showing that multiple variation detection algorithms identify such variants and that they are common in PitNETs.

Several genes that have been reported with somatic variants (without dbSNP identifier and gnomAD frequency) in previous PitNET exome and genome sequencing studies, also harbor somatic variants in tumors of our PitNET patients. These genes are (GNAS [8], TSC2 [4], KLHL4 [6], POLR3B, KHDRBS2, LAMA1, TUBGCP6 [7], RYR1 [7, 6], PLA2G6, TESK1 [5], AHNAK [15] but their somatic variants are all novel and in all cases distant enough to exclude variant hotspots or disruption of certain part of the protein. These genes could be potential candidates for further studies as functional pathways impacted by these genes lead to the formation of PitNETs. RYR1 has been reported in previous PitNET studies at least twice [6, 7] and had not been highlighted as PitNET associated. RYR1 is not expressed in rat pituitary [36], no data about this gene expression in human pituitary have been reported, if indeed RYR1 is related to PitNET development this could also be due to an indirect effect through gene-gene interaction or linkage disequilibrium with another causative factor. RYR1 functions as a regulator of inner cell Ca²⁺ flow, the dysfunctions of this factor have been implicated in heart conditions, myopathies and neurodegenerative disease [37]. It has been shown that somatic variation in calcium regulation related genes are common in PitNETs [5], therefore this is an interesting target for potential functional studies in PitNET development.

LAMA1 [38, 39] and AHNAK [35] has been implicated in tumor metastasis while TSC2 is considered tumor suppressor (reviewed in [40]). Relation of other overlapping genes with PitNET is more obscure.

Tumor-related gene KIT [25], was also detected to have a somatic variant. (synonymous and splice site variant) was found in PN02. This somatic variant was estimated to be relatively benign.

Several limitations can be identified for our study. The deviation of PitNET type distribution in our sample from general distribution of clinical PitNET patient type in our population and literature [41] could be explained with relatively small sample size and/ or overrepresentation of large PitNET (over 10 mm in largest dimension)which are more likely to be treated with surgery and therefore providing tumor cells for genetic research. The unusual gender disbalance also is most likely attributed to chance as the material from surgeries from March 2016 to February 2018 were collected. And finally, chromosomal aberrations in PitNET are often investigated by using array genotyping which could add information on copy number variations across PitNET genomes. We did not detect high confidence copy number variations in our samples confirming problems in such analysis from exome data mentioned in M. Fromer and S. M. Purcell, 2015 [22].

In conclusion, besides GNAS and AHNAK, we have identified PitNET 11 more genes overlapping from previous PitNET studies which contain somatic variants of which several (RYR1, LAMA1, AHNAK, TSC2) could be highlighted as potential candidates for further PitNET research. The role of these genes and somatic variants in PitNET development should be investigated in further studies. We propose the use of multiple sequencing data analysis pipelines as they can provide more accurate identification of somatic variants in PitNETs.

Supporting information

Acknowledgments

The authors acknowledge the Latvian Biomedical Research and Study Centre and the Genome Database of the Latvian Population for providing infrastructure, biological material and data.