Introduction

Pituitary tumors include neoplasms that arise in and around the sella turcica [1, 2]. The majority arise from the hormone-secreting adenohypophysial cells that are members of the family of neuroendocrine epithelial cells. Other tumors include those arising from pituicytes (the modified glia of the posterior lobe), tumors that arise from the meninges, stromal elements, and adjacent hypothalamus, as well as craniopharyngiomas and many others [3].

Tumors of adenohypophysial cells exhibit a wide spectrum of hormonal and proliferative characteristics; small tumors may be incidental findings [4] or may be the cause of significant morbidity and even mortality, as in the case of untreated Cushing disease [5]. A large number of tumors are prolactinomas (functional lactotroph tumors) that respond to medical therapy with normalization of hormone levels and tumor shrinkage in the majority of cases. However, a significant proportion, estimated to be more than half of all patients with clinically diagnosed anterior pituitary tumors [6], require surgical intervention, some with curative intent, others for debulking and avoidance of compressive effects. Up to one quarter of patients who undergo surgery have residual and persistent disease that may entail complex management for hormone regulation as well as additional surgical intervention, radiotherapy, and chemotherapeutic attempts to restrain tumor growth.

In this review, we summarize the features of adenohypophysial hormone-producing cells that represent neuroendocrine cells, we discuss the terminologies that are relevant for the classification of the neoplasms that arise from these cells, and we document the epidemiology of these lesions. The large body of work reviewing prognostic features of the various tumor types is summarized and the importance of tumor subtyping is discussed as a model for other neuroendocrine tumors (NETs).

Hormone-secreting pituitary adenohypophysial cells are neuroendocrine cells

Endocrine cells that produce hormones are found throughout the body, both in endocrine glands and dispersed among other cells of non-endocrine organs. They fall into three categories based on cell differentiation and the biochemical features of the hormone(s) they produce [7]:

  1. (1)

    Steroid hormone-producing cells are found mainly in the adrenal cortex and gonads. They are thought to be of mesodermal embryonic origin and their biochemistry involves the uptake of cholesterol for conversion by complex enzymatic processes into glucocorticoid, mineralocorticoid, and sex steroid hormones. They are characterized structurally by an abundance of smooth endoplasmic reticulum and fat globules.

  2. (2)

    Thyroid follicular cells, of endodermal origin, are epithelial cells that synthesize thyroglobulin, store it in sequestered follicles, and resorb it for iodination and cleavage into thyroid hormones. They have extremely well-developed intercellular junctions, rough endoplasmic reticulum, and complex microvilli at the apical lumen that are critical for resorption of thyroglobulin.

  3. (3)

    The largest family of endocrine cells have neuroendocrine differentiation and are known as neuroendocrine cells (Fig. 1). These cells all are involved in the production of peptide hormones that are produced in the rough endoplasmic reticulum, processed in the Golgi complex and packaged into membrane-bound neurosecretory granules where they are stored until hormone regulatory signals cause them to move to the cell membrane for exocytosis-mediated release into the bloodstream (Fig. 2A). Their common structural and functional features gave rise to the APUD (amine precursor uptake and decarboxylation) theory by Pearse in 1974 [8]. There are two subfamilies of neuroendocrine cells: epithelial neuroendocrine cells, derived mainly from the embryonic endoderm, and non-epithelial neuroendocrine cells (paraganglia, also called “paraneurons”), derived from the embryonic neuroectoderm. Epithelial neuroendocrine cells include pituitary adenohypophysial cells, parathyroid cells, parafollicular C-cells, pancreatic islet cells, and the cells of the so-called “dispersed neuroendocrine system” that populate the mucosa of the respiratory and gastrointestinal tracts from the nose and mouth to the rectum, as well as the genitourinary system [7]; these cells are also present in the skin where they are known as Merkel cells. Paraganglia are present in soft tissue from the base of the skull to the bottom of the vertebral column; the largest form the adrenal medulla bilaterally, and they include the para-aortic paraganglia, the carotid body, the jugulotympanic paraganglion (also called “glomus jugulare”), the organ of Zuckerkandl, and paraganglia of the lungs and porta hepatis [9]. All neuroendocrine cells express a number of proteins that are characteristic of their differentiation, including neural cell adhesion molecule that is recognized by the antibody CD56, neuron-specific enolase, synaptophysin, PGP 9.5, chromogranins, and the transcription factor insulinoma-associated protein-1 (INSM1); several of these are not specific for neuroendocrine cells and only chromogranins and INSM1 represent specific biomarkers of neuroendocrine differentiation (Fig. 2B, C) [10] as they are only also expressed by neural tumors [11]. In addition to these common biomarkers, most epithelial neuroendocrine cells express keratins (that are not present in paraganglia) and specific transcription factors and hormones that allow cell type characterization, whereas paraganglia express the transcription factor GATA3 and tyrosine hydroxylase, a critical enzyme early in the biosynthetic pathway of catecholamines.

    Fig. 1: The Families of Neuroendocrine Cells.
    figure 1

    These highly specialized hormone-secreting cells are divided into two families, the epithelial group that usually express keratins and the paraganglia that derive from the sympathetic and parasympathetic autonomic nervous system and usually do not express keratins. Epithelial neuroendocrine cells can comprise a neuroendocrine organ, such as the pituitary or parathyroid, or they can be dispersed throughout non-endocrine or non-neuroendocrine tissues. Each group of epithelial neuroendocrine cells is composed of specific cell types that have distinct structure and function.

    Full size image
    Fig. 2: Neuroendocrine features of pituitary neuroendocrine tumors.
    figure 2

    A The ultrastructure of this PitNET from a patient with acromegaly illustrates the neuroendocrine features of adenohypophysial cells. The tumor cell cytoplasm contains rough endoplasmic reticulum (*) for hormone synthesis, a prominent Golgi complex (G) where hormone is packaged into forming secretory granules, and numerous membrane-bound secretory granules that store hormone for release into the extracellular space. B The tumor cells show cytoplasmic chromogranin reactivity. C Nuclear INSM1 positivity is a feature of pituitary neuroendocrine tumors.

    Full size image

Pituitary adenohypophysial cells are a complex family of neuroendocrine cells that, like other related cells, express transcription factors that determine their ability to synthesize specific hormones [12]. Studies of cytodifferentiation have demonstrated three major families of pituitary adenohypophysial cells, with six terminally differentiated cell types (Fig. 3). In the mature gland, there is evidence to suggest that transdifferentiation occurs physiologically between the four cell types of the PIT1 family; it is unclear if transdifferentiation occurs between corticotrophs, gonadotrophs and the PIT1-lineage somatotrophs, lactotrophs, mammosomatotrophs, and thyrotrophs. In general, PitNETs are classified as originating from these various cell types, but in addition, morphologic variants of these mature cells have different patterns of granulation, and some tumors show differentiation along a lineage without the features of terminally differentiated mature cell phenotypes (Fig. 3). Null cell tumors do not show evidence of adenohypophysial lineage differentiation; with improved techniques they represent as low as 0.6% of all adenohypophysial tumors [13], consistent with the initial proposal that with advances in knowledge, they would all but disappear [14] and now their continued existence has been questioned [15]. Mixed tumors also occur; they may be of a single lineage, such as the mixed tumors of somatotrophs and lactotrophs that may represent divergent differentiation of a single cell type, or they may be synchronous distinct tumors of different lineages [16]. There are also unusual plurihormonal tumors that show mixed lineage [1, 17]. Very rare and unusual pituitary tumors of primitive adenohypophysial cells (composed of a mixture of adenohypophysial neuroendocrine cells, primitive Rathke’s cleft cells, and primitive folliculostellate cells) are known as pituitary blastomas [18].

Fig. 3: Pituitary neuroendocrine cells and their tumors.
figure 3

Adenohypophysial hormone-secreting cells differentiate along three lineages to form multiple mature cell types. Each mature cell type is associated with at least one tumor type, but several cell types give rise to multiple tumor variants. In addition, there are tumors composed of cells that lack terminal differentiation; some of these show lineage differentiation but others, most commonly those classified as null cell tumors, lack evidence of any lineage specificity.

Full size image

Terminologies in classification of pituitary NETs

Tumors composed of adenohypophysial hormone-secreting cells have been classified as “adenomas” since their early descriptions in the late 19th century and early 20th century. This terminology has been applied even to tumors that extend upwards to compress the optic chiasm and brain, and to those that infiltrate dura and invade laterally to involve the cavernous sinuses, as well as to tumors that penetrate the sphenoid bone, clivus, and/or paranasal sinuses. The term “carcinoma” has been restricted to PitNETs that show evidence of distant metastasis to cervical lymph nodes, lung, liver and bone, and/or discontinuous cerebrospinal fluid dissemination within the cranium or spinal canal [1, 2]. This represents a divergence from some other tumor classification schemes in which invasion represents the defining characteristic of a malignancy, for example, epithelial neoplasms including squamous and basal cell carcinomas, breast carcinomas, and intestinal malignancies.

By definition, an “adenoma” is a “benign tumor of glandular epithelial cells.” The definition of “benign” is “a disease that is not harmful in effect.” The Merriam-Webster dictionary defines benign as “of a mild type or character that does not threaten health or life,” as “having no significant effect.”

To address the discrepancy between benign behavior and the morbidity of patients with invasive tumors that cannot be resected surgically, efforts were made to classify some tumors as “atypical adenomas” [19]. However, the criteria for this diagnosis suffered from a lack of reproducibility, and this terminology has since been abandoned [2].

In other sites, tumors of neuroendocrine cells have had varying terminologies applied and they have undergone changes over time. In the pancreas, “islet cell adenoma” was replaced by “islet cell tumor” when it became clear that they can behave aggressively; subsequent studies that showed origin of some of these tumors from neuroendocrine cell precursors in pancreatic ducts rather than in islets resulted in a further change to “pancreatic neuroendocrine tumors” (pNETs or PanNETs) [19, 20]. Oberndorfer initially classified small bowel NETs as “carcinoid” tumors because they looked “carcinoma-like” [21]; once their behavior was clarified as potentially malignant, the term “carcinoid” was mostly restricted to the clinical syndrome of serotonin excess that is a feature of the enterochromaffin (EC) cell neoplasms, and these tumors assumed the terminology of intestinal NETs along with their counterparts, gastric, and rectal NETs. In the lung, the terminology of “typical and atypical carcinoid tumors” remains embedded in the literature [22], but is likely to change over time [23].

Molecular analysis combined with careful clinical studies have shown that well-differentiated NETs have unique genetic and epigenetic alterations that are not infrequently germline, giving rise to familial syndromes such as multiple endocrine neoplasia types 1, 2, 4, and 5, von Hippel Lindau syndrome, tuberous sclerosis, neurofibromatosis syndromes, and the succinate dehydrogenase (SDH)-related [24,25,26,27]. In contrast, poorly differentiated neuroendocrine neoplasms (NENs), including high-grade malignancies in the lung and upper respiratory tract, pancreas and gut, breast, and genitourinary system, are more closely related to non-endocrine carcinomas, both genetically and clinically [28, 29] with some exceptions in the lung [30].

To address the issues of terminology, the World Health Organization (WHO) proposed the implementation of a unified approach to the classification of neuroendocrine neoplasia at all body sites. The proposal includes the use of the term “NEN” to describe all members of this large family of proliferative lesions of neuroendocrine cells [23]. The term “neuroendocrine carcinoma” (NEC) was proposed for the aggressive poorly differentiated high-grade malignancies; in contrast, the well-differentiated and generally low-grade neoplasms were classified as “NETs” in keeping with the terms in place for gastro-entero-pancreatic lesions.

The term “tumor” is indeed problematic. Etymologically derived from the Latin words “tumere,” meaning “to swell” and/or “tumor,” meaning “swelling,” this word has been used since the 16th century to mean a mass that can be inflammatory, neoplastic, or vascular. However, in modern medical vernacular, it has taken on the context of neoplasia, and in the case of “NET,” it has a specific connotation, implying a neoplasm composed of well-differentiated neuroendocrine cells, usually with low proliferative activity [23]. There is no specific indication of benign or malignant classification, as it is recognized that these lesions may show a wide spectrum of behaviors, including a potential for local invasion and metastasis.

In 2016, to address the clinical need for recognition of the morbidity caused by invasive and aggressive pituitary “adenomas” and their metabolic disturbances, and anticipating a convergence of classifications of NETs at various body sites, the International Pituitary Pathology Club proposed a change in nomenclature for tumors of hormone-secreting adenohypophysial cells from “adenoma” to “NET” [31]; in parallel with the pancreatic abbreviation “PanNETs,” the term was abbreviated to “PitNETs.” This terminology is consistent with the WHO proposal. It recognizes these pituitary neoplasms as members of the neuroendocrine family. The term “tumor” replacing adenoma provides for the spectrum of behavior from indolent lesions to those with significant hormonal impact and/or invasive growth, thus recognizing the morbidity of aggressive tumors without implying overt malignancy. It also avoids the contradiction when a pituitary tumor develops a metastasis, which, although rare, usually follows a prior diagnosis of “adenoma.”

Another important aspect related to terminology and tumor classification is the categorization of histological tumor subtypes based on clinical and biochemical features. The term “pituitary NET” is a classification for the family of tumors, and each tumor subtype should be classified based on cell type, e.g., corticotroph, lactotroph, somatotroph, mammosomatotroph, thyrotroph, or gonadotroph tumor (Fig. 3). The clinical presentation is important. Most somatotroph and lactotroph tumors are clinically functioning and the classical presentation of a corticotroph tumor is Cushing disease; however, some of these differentiated hormone-producing tumors are clinically “silent.” In contrast, the vast majority of gonadotroph tumors are clinically silent. Clinicians should be aware that among tumors that are clinically classified as nonfunctioning, the majority are gonadotroph tumors that tend to be indolent, but there are also tumors that are morphologically classified as silent corticotroph, silent somatotroph, and other silent tumors of PIT1 lineage, and very rarely they can be classified as true null cell tumors with no evidence of pituitary lineage differentiation.

Unlike in other sites, detailed structure–function correlations in pituitary tumors have shown that there are structural variants of hormone-producing cell types that are clinically relevant. For instance, the clinical entity known as acromegaly is not a single disease [32]; densely and sparsely granulated variants of somatotroph tumors have distinct clinical and radiologic features and different responses to therapy, and several other tumors of PIT1 lineage can also manifest with acromegaly. Similarly, a TSH-producing PitNET is not always a thyrotroph tumor, since both mature and poorly differentiated plurihormonal PIT1-lineage tumors can also result in central hyperthyroidism [33, 34]. A substantial advance in knowledge has been achieved in PitNETs by clinicopathologic and genomic studies that have applied rigid criteria to distinguish histological tumor subtypes.

Epidemiology

Like other NETs that have been increasing in incidence [35], PitNETs that were once considered to be rare are increasingly diagnosed. They are frequently identified incidentally at autopsy or on routine radiologic examination for other reasons. Autopsy studies have reported a prevalence as high as 22.5 [36] and 27% [37] of cases. Approximately 20% of “normal” pituitary glands harbor an incidental radiologically detectable lesion measuring 3 mm or more in diameter [4, 38]. The majority of these incidental findings are clinically nonfunctioning pituitary tumors, but even these “asymptomatic” tumors can cause hypopituitarism [39] while others are prolactinomas that cause undiagnosed clinical symptomatology [40, 41]; occasionally they are somatotroph or corticotroph tumors causing clinically undiagnosed acromegaly and Cushing disease, respectively.

In several population studies from the last 15 years, the prevalence of clinically diagnosed pituitary tumors ranges from ~78 to 116 cases per 100,000 people [6, 42,43,44]; one study from a Swedish group reported an incidence of 3.9/100, 000 people [45]. Data from the Central Brain Tumor Registry of the United States from 2011 to 2015 show pituitary tumors as 17.5% of reported brain tumors; they estimate an incidence rate of 4.12 sellar tumors/100,000 population per year, but this figure is clearly low, since these statistics are based only on surgically resected tumors [46]. In a recent Korean nationwide population-based study, the annual incidences per 100,000 population were estimated to be 3.5 for nonfunctioning pituitary tumors, 1.6 for lactotroph tumors, 0.5 for somatotroph tumors, and 0.2 for corticotroph or thyrotroph tumors [47]. Independent of country-specific differences, epidemiologic studies considered together show that clinically apparent PitNETs are increasing in both annual incidence (between 3.9 and 7.4 cases per 100,000) and prevalence (76–116 cases per 100,000 population), the latter corresponding roughly to 1 case per 1000 people, with nonfunctioning tumors and prolactinomas as the most frequent subtypes [48].

Prolactinomas are consistently the most common type of PitNET [4, 6, 42,43,44,45, 48]; they are usually treated medically [49,50,51,52] and most are therefore not captured in cancer or surgical statistics, so their exact incidence is unknown. In one prevalence study, surgical resection was performed for 56% of patients with clinically diagnosed PitNETs [6]. The relative frequency of the various tumor types resected surgically shows that more than a third are hormonally inactive mass lesions of SF1 lineage [53, 54]. About 30% are of PIT1 lineage and more than half of those give rise to growth hormone excess, while ~15% are TPIT lineage tumors [6, 42, 49, 52, 54, 55].

Prognostic biomarkers in pituitary NETs

The prognosis of a patient with a PitNET depends on multiple factors, which are collectively described as “aggressiveness.” The relevant parameters that define aggressiveness include hormone secretory activity, tumor size, site, extent of invasiveness, and rate of growth into surrounding structures as well as refractoriness to therapy. Together, these parameters influence the success of surgical resection, metabolic disorders caused by the tumor, along with other intrinsic tumor characteristics that collectively impact disease-specific morbidity.

While the vast majority of prolactinomas are clearly not aggressive, some are resistant to dopamine agonist therapy and the rare very large tumors in men, can be more aggressive, not only because they are larger at presentation but possibly because of altered expression of genes that are mediated by estrogen receptor signaling [56]. In the following paragraphs, we review the evidence that is available to help define aggressiveness based on the extent of disease and tumor type.

Of patients who undergo surgery, the percent that have persistent disease varies depending on the size, invasiveness of the tumor, and extent of resection [57]. Recurrence after surgery has been reported to vary from 0 to 40% in those with 5 or more years of follow-up; the variability is often due to the duration of follow-up. In one series, progression occurred in 27.8% of patients and the mean time to progression was 23.7 months (range, 3.7–52.4 months), both associated with tumor size and extent of resection [58]. Neurosurgeons have defined the extension of tumors into adjacent neuroanatomical regions (Fig. 4) such as the degree of lateral extension into the cavernous sinus (Knosp classification) [59] and also the degree of extra-sellar and vertical extension into suprasellar regions (Hardy classification) [60]. Many patients with Knosp grade 3 or higher cavernous sinus invasion eventually require multimodal therapies including repeat operations and/or radiotherapy [61]. Despite efforts to predict cavernous sinus invasion on magnetic resonance imaging (MRI), the precise confirmation of invasion frequently requires surgical and/or histological examination. Clearly there is a need for better biomarkers to help predict invasiveness [62].

Fig. 4: Magnetic resonance imaging of an aggressive pituitary neuroendocrine tumor.
figure 4

Contrast enhanced T1-weighted MRI of sella of a locally invasive macrotumor that was associated with visual loss and weight gain. On investigation, the ACTH level was 237 pg/mL (normal, <50) and morning cortisol 28 μg/dL (normal, 4–22), providing a biochemical diagnosis of Cushing disease. The tumor surrounds the pituitary gland, which is seen as an ill-defined focus of enhancement in the middle of the sella, and fills both cavernous sinuses (both Knosp grade 4), with suprasellar extension through the oculomotor triangle into the basal cistern just above the right cavernous sinus, involvement of the clivus, and invasion into the sphenoid sinus. A diagnosis of pituitary carcinoma cannot be assigned because of the absence of any noncontiguous metastatic tumor. This highly invasive tumor shows the incoherence of diagnosis as a grade I “adenoma” in the face of such aggressive clinical behavior.

Full size image

Surgical remission varies by tumor type [1, 63]. In a surgical series that included multiple hormonally active tumor types, at a median follow-up of 56 months, tumor recurrence was 0, 11, and 14% for GH-, ACTH-, and PRL-secreting tumors [64]. In another study, the recurrence-free survival at 10 years was 78.2% in acromegalic patients, 68.1% in prolactinomas, 74.3% in Cushing disease, 70.3% in TSH-secreting tumors and 75.3% in clinically hormonally inactive tumors [65]. In yet another series, corticotroph tumors had a higher risk of recurrence than other tumor types [66].

The Knosp classification has been reported to correlate with surgical outcomes in GH tumors [67]. When surgery is performed by an experienced surgeon, remission is >85% for microtumors <1 cm [68,69,70] but only 40–66% for macrotumors [69,70,71,72]. Because GH-secreting tumors can invade into sellar dura and/or the cavernous sinus at the time of diagnosis, the overall control rate with strict hormonal criteria is 50–80% [67, 73,74,75]. Occult dural foci of tumor may provide the basis for persistent acromegaly after gross total resection and a clear postoperative MRI. Many studies report the outcome of transsphenoidal surgery as the first-line treatment; the impact of preoperative medical therapy for GH-secreting tumors to address comorbidities and reduce soft tissue swelling to facilitate the operative approach remains the subject of debate.

For ACTH-secreting tumors, reported remission rates vary from 60 to 90%, but up to 15–20% eventually recur [76,77,78,79]. Pivonello et al. [80] reviewed 74 studies published between 1976 and 2014 involving 6134 patients with Cushing disease and a mean follow-up duration of 64.3 months. In their review, they found that the overall initial remission rate ranged from 25 to 100%, with a mean remission rate of 77.8% (median, 78.7%); the recurrence rate ranged from 0 to 65.6%, with a mean recurrence rate of 13.2% (median, 10.6%). The remission rates in microtumors ranged from 48.7 to 100% (mean, 82.1%; median, 85.7%), whereas in macrotumors, they ranged from 30.8 to 100% (mean, 62.3%; median, 64.1%). The recurrence rates in patients with microtumors and macrotumors were 0–36.4% (mean, 11.7%; median, 10.9%) and 0–59% (mean, 18.8%; median, 13.9%), respectively. Although en bloc resection achieves the best rates of remission (98.5%), it applies only to smaller tumors not invading dura or other parasellar structures [81]. Dickerman and Oldfield noted the clinical importance of occult dural invasion, i.e., invasion of the dura or cavernous sinus that is not evident on imaging studies and is not obvious to the surgeon, as the basis of recurrence or persistent tumor and endocrinopathy in their series of Cushing disease [82].

After surgery, the probability of recurrence of apparently completely resected tumors and the time to clinically relevant regrowth of known residual tumor correlate with radiologic tumor doubling-time [83]. However, there is controversy about the role of proliferation markers including mitoses (conventional or phosphohistone-H3-assisted count), Ki-67 labeling index, and p53 staining [84, 85].

While transsphenoidal surgery is the initial treatment of choice for the great majority of large pituitary tumors, many cases of functional tumors cannot be hormonally controlled, and the recurrence rate of nonfunctioning tumors is high. In tumors that cannot be fully resected, the prognosis depends on multiple factors that mediate tumor responsiveness to other therapeutic modalities. These include therapies used to control hormone hypersecretion as well as those intended to target structural tumor growth. Patients with acromegaly and residual tumor postoperatively are more likely to respond to first-generation somatostatin analogues if their disease is due to a densely granulated somatotroph or mammosomatotroph tumor than if they have a sparsely granulated somatotroph tumor [86,87,88,89,90]. It has been reported that patients with persistent or recurrent Cushing disease are more likely to respond to pasireotide if they have a tumor with a USP8 mutation [91], a molecular alteration that is more frequent in densely granulated microtumors [91, 92], again emphasizing the importance of tumor histotype in determining therapeutic responsiveness.

Among patients with clinically nonfunctioning tumors, classification by cell type is equally important [13]. Among clinically nonfunctioning tumors, gonadotroph tumors are more indolent whereas true null cell tumors and silent corticotroph tumors are more likely to behave aggressively [13, 93,94,95]. For example, in one report of silent corticotroph tumors with a follow-up longer than 5 years, 31% recur (CI 13–38%) [96]. Approximately 50% of patients with detectable postoperative residual tumor develop progression within a 5-year period [97,98,99]; however, the most common gonadotroph tumors tend to grow slowly, whereas silent corticotroph tumors, while not always reported to have a higher recurrence rate [96], have been shown in most studies to be more proliferative and invasive, resulting in relatively more frequent compressive features with visual field loss, hypopituitarism, and other symptoms and signs of tumor progression [100,101,102,103,104].

Certain tumor types are considered to be inherently aggressive. The so-called “poorly differentiated” PIT1-lineage tumors (formerly known as “silent subtype 3 pituitary adenomas”), which are thought to lack terminal PIT1-directed cell maturation, have been associated with frequent disease progression in three clinical series [33, 105, 106]. Crooke cell tumors have been well-recognized for their adverse biology [107,108,109]. In the group of somatotroph tumors, sparsely granulated tumors are more aggressive than densely granulated tumors; they present with larger and more invasive lesions [87, 89, 110,111,112]. Among prolactin-producing tumors, acidophil stem cell tumors have a higher risk of aggressive disease [113, 114] and are often resistant to medical therapy [115]. TSH-secreting tumors historically showed more invasive behavior and were more difficult to fully resect [116]; however, this may relate to delays in diagnosis, and inappropriate treatment with thyroid ablation that resulted from misdiagnosis as primary hyperthyroidism.

It has been estimated that of pituitary tumors requiring surgery, some 15% can be considered as highly aggressive based on their surgical, radiological, and pathological characteristics [117, 118]. Such patients may be resistant to both conventional radiotherapy and radiosurgery. Newer medical agents adopted in the broader field of NETs are beginning to make their way into the therapeutic armamentarium for a subset of unresectable or especially aggressive PitNETs. These include the mTOR inhibitor everolimus [119, 120] as well as the DNA methylation inhibitor temozolomide [121,122,123,124]. In other tissues, responsiveness to temozolomide has been shown to be predicted by the level of expression of O6-methylguanine-DNA methyltransferase but it is unclear if this is useful in the case of PitNETs [122, 123, 125, 126]. Equally as important, the optimization of therapeutic responsiveness of temozolomide-based therapy through the combined use with capecitabine has resulted in the now widely adopted CAP/TEM regimen in NET management [127, 128]; it is expected that CAP/TEM will similarly be used for temozolomide monotherapy escape or resistant PitNETs [128,129,130]. Peptide radio-receptor therapy with 177lutetium (177Lu)-based protocols represents another major advance in NET management that will likely extend to use in PitNETs [129, 131]. Such NET management strategies, alone or in sequence, will prove to be critically important in directing their extended use for aggressive PitNETs.

These data show the importance of determining a dynamic multifactorial clinicopathologic risk stratification. This has been attempted [57]; however, it is important to distinguish tumor grade from risk score and tumor stage. It is important to apply grading based not only on proliferation but also on differentiation, and there is clearly a need for appropriate staging of PitNETs to allow the determination of prognosis based on surgical resectability. The prognosis for patients with inoperable disease is dependent on response to medical therapy that is mainly a reflection of cell type and subtype.

Genetic predisposition and pathogenesis-related biomarkers

Like other NETs, PitNETs are members of several families of tumors associated with a genetic predisposition. They form one of the three “P”s of multiple endocrine neoplasia type 1 (MEN1), originally described by Wermer [132]. This autosomal dominant disorder with incomplete penetrance is due to inheritance of a mutant copy of the MEN1 gene that encodes menin. Approximately half of the affected patients develop a PitNET; the most common types produce prolactin and/or GH [4, 133,134,135] and poorly differentiated tumors of PIT1 lineage have been reported in patients with MEN1 [33]. The tumors may be multiple and multicentric [134, 136], but this can be hard to prove. Patients with MEN1 generally present at a younger age than those with sporadic tumors [135]. Other manifestations of MEN1 include parathyroid and pancreatic neoplasms as well as frequent adrenal cortical tumors [137] and, more recently, breast cancer has been shown to be more frequent in patients with MEN1 [138].

Defects in CDKI genes are responsible for an MEN1-like syndrome known as MEN4. Patients have mutations in the CDKN1B/p27Kip1 gene [139, 140] and the CDKN2C/p18INK4c gene [141, 142]. MEN4 is characterized by parathyroid proliferations, PitNETs, and pNETs as well as small intestinal NETs, lymphoma, and breast cancer [143].

Cases of MEN1-like acromegaly associated with hyperparathyroidism and pNETs have been described in patients with germline mutation of CDC73 [144, 145]. Germline pathogenic MAX variants have been associated with PIT1-lineage tumors along with other endocrine and non-endocrine tumors in a syndrome proposed as MEN5 [27].

A small number of patients with pituitary GH-secreting tumors, most often those presenting at a young age and associated with gigantism, have Carney’s complex [146], an autosomal dominant disorder associated in around 50% of cases with germline mutations in the PRKAR1Aα gene that encodes the PKA regulatory subunit 1Aα, causing unrestrained cAMP signaling [147, 148]. The syndrome is characterized by spotty pigmentation due to nevi affecting mucosal surfaces and the lips, malignant melanotic nerve sheath tumors, cardiac and other myxomas, and endocrine tumors including not only pituitary tumors but also bilateral pigmented nodular adrenocortical disease and testicular tumors [146].

Mutations in genes encoding the various components of the SDH complex are implicated in several types of familial neoplasia [149] including pheochromocytomas and paragangliomas, gastrointestinal stromal tumors, and renal carcinomas as well as rare pituitary tumors [150].

Lynch syndrome is a well-recognized autosomal dominant familial cancer predisposition syndrome that increases the risk of many types of cancer, mainly colorectal carcinomas but also tumors of the stomach, small intestine, liver, gallbladder, upper urinary tract, brain, skin, and adrenal cortex; women with this disorder have a high risk of ovarian and endometrial carcinomas. It is due to mutation in one of the genes encoding mismatch repair proteins (MLH1, MSH2, MSH6, or PMS2). However, recently NETs, including pancreatic and pituitary tumors, have also been reported in patients with this disorder [151, 152].

In some cases, familial transmission is unique to pituitary tumors without other associated NETs. Predisposition to develop pituitary GH-producing tumors without other endocrine tumors was originally described as the isolated familial somatotropinoma syndrome [153]. Subsequent studies identified predisposition for other types of pituitary tumors as well, and the nomenclature was altered to the familial isolated pituitary adenoma syndrome [154]. The pituitary tumors in these families arise at a younger age and are larger than tumors in matched sporadic pituitary tumor cohorts. This syndrome is associated in around 20% of cases with germline mutations in the AIP gene that encodes aryl hydrocarbon-interacting protein (AIP), which functions as a tumor suppressor [154,155,156]. AIP is a chaperone protein that modulates transcription of the aryl hydrocarbon receptor, which was originally described to mediate toxicological and carcinogenic dioxin effects [157, 158].

A rare form of early childhood X-linked acrogigantism attributed to Xq26 microduplications and GPR101 mutation [159] may primarily predispose to hyperplasia but somatotroph and/or mammosomatotroph PitNETs have also been described [160].

Sporadic tumors rarely have mutations in the genes that are implicated in these familial syndromes [1, 161]. Instead, mutations have only been identified only in small subsets of tumors. For example, a subset of somatotroph tumors have activating mutations of GNAS resulting in tumors that have high cyclic AMP levels and responsiveness to somatostatin analogues; several studies have shown that this mutation is most often seen in tumors with a densely granulated somatotroph phenotype [32, 162] but not all densely granulated tumors harbor GNAS mutations and the correlation is not perfect [110, 111]; histopathology remains the best predictor of therapeutic responsiveness [90]. Some corticotroph tumors have mutations of USP8, a gene encoding a member of the ubiquitin protease family and is thought to alter EGFR signaling; this feature is predictive of pasireotide response [91] and tends to be identified in small densely granulated corticotroph tumors [91, 92, 163, 164]. Interestingly, a recent study has identified ATRX mutations in aggressive pituitary tumors, mostly of corticotroph type, including those with metastases [165], similar to pNETs [166] and paragangliomas [167]. Nevertheless, the majority of sporadic PitNETs lack recurrent mutations, showing only variable genomic instability with copy-number variations [168,169,170,171,172], and like small bowel NETs [173], epigenetic dysregulation has been implicated in their pathogenesis [174].

Clinical management: the need for centers with sustained experience

Endocrine tumors are complex and require broad multidisciplinary expertise. Unlike other solid tumors, they produce a number of hormonal disturbances that have significant importance to the patient and that often are as important or perhaps more so than tumor growth. In the case of PitNETs, the clinical team includes endocrinologists, neuroradiologists, neurosurgeons, pathologists, medical and radiation oncologists, specialized nurses, and psychosocial experts who are knowledgeable about the emotional disorders that are common in patients with these tumors [175,176,177]. The need for psychosocial care in oncology is now becoming widely recognised, but the recognition of the hormonal aspects of endocrine neoplasia adds another dimension to this problem. A strong and integrated multidisciplinary team is required to allow for personalized approaches to care and ongoing support.

Throughout the various areas of medicine, subspecialisation has been proven to result in significant quality improvement, error reduction, and cost avoidance [178,179,180,181,182,183,184,185,186,187,188,189,190]. Data show consistently that all members of the care team have better outcomes with increasing numbers of patients. Experts are able to create standards based on evidence from their practices. For example, pathologists who see many cases of pituitary tumors each year have formulated a synoptic report that provides a standard approach to morphologic diagnosis and structure–function correlations [191]. Although the designation of high-volume centers as “Centers of Excellence” has been proposed [192], as yet no criteria have been created for such a designation, nor any external assessment protocols that provide objective measures of “excellence.”

PitNETs as a model for other NETs

As we have shown in the previous paragraphs, the careful categorization of PitNETs into different morpho-functional subtypes according to cell lineage and specific structural properties of neoplastic cells leads to a classification with important predictive and prognostic implications. In fact, such a degree of insight into structure–function–behavior relationships has not been reached for NETs arising in other organs. Nevertheless, it should be recognized that, in almost every anatomical site in which neuroendocrine cells occur, except for thyroid and parathyroid where only a single neuroendocrine cell type is recognized, NENs exhibit a wide morpho-functional spectrum, reflecting and sometimes amplifying the diversity of normal neuroendocrine cells to which they are related.

In particular, it is known that gastrointestinal and pNETs include many different subtypes, defined according to cell morphology and hormone production [193]. However, in the current WHO classifications, these features are only taken into account for functioning PanNETs [2, 194], mainly in relation to the clinical syndromes associated with these neoplasms, whereas the core of the classification is based on the degree of histological differentiation and the proliferative grade. In the small bowel, the pioneering work of Swedish investigators identified more than 20 different cell types with distinct hormone products and ultrastructure in the late 1970s and early 1980s [195]; this work has been largely ignored, likely because the vast majority of tumors are EC cell serotonin-producing lesions that can potentially cause the carcinoid syndrome and its sequelae. It is unclear why EC cells alone, from among the many others scattered throughout the ileum, give rise to tumors. The lungs have at least three neuroendocrine cell types known to produce several peptides, including bombesin, calcitonin, and calcitonin-gene-related peptide, as well as serotonin, and although the hormonal profiles of pulmonary NETs have been investigated [196], this subject has been disregarded, possibly because of a suggested lack of clinical impact. The same is true also for other rare anatomical sites including the upper respiratory tract, genitourinary tract, and others. In general, the field of NETs seems to suffer from an under-appreciation of the clinical significance of the cells of origin and their specific alterations in tumors. The very few exceptions are represented by the work identifying the four cell types in duodenal NETs [197], publications showing that L cell rectal NETs have a distinct morphology and more indolent behavior than non-L cell rectal NETs [198,199,200], and data reporting ectopic hormone secretion in lung [201, 202] and pNETs [201, 203,204,205]. Lessons learned from PitNETs should prompt a re-evaluation of the significance of hormone profiles and careful morphological examination of specific cell types in NETs of other body sites. Documentation of tissue-specific transcription factors and hormone profiles can advance the diagnostic management of NETs, improve the diagnosis of metastatic disease from an unknown primary site [206] and provide biomarkers for clinical surveillance. Moreover, this information can also provide useful tools for the therapeutic choices and prognostic evaluation of patients with these tumors [207, 208].

If the pituitary is to be used as the paradigm of neuroendocrine neoplasia, the field must also align with other NETs in more than just nomenclature. Indeed, the concept of pituitary carcinoma, traditionally applied to metastatic lesions, should be relinquished in light of two important considerations. First, the term NEC is a highly specific term, which is applied only to high-grade neoplasms with distinct genetic alterations more akin to those of other epithelial malignancies [23]. Curiously, there are almost no examples of such lesions in the pituitary with the possible exception of rare TP53-mutant carcinomas. Even the so-called “poorly differentiated” tumors are only considered to be poorly differentiated in the context of the complex and elegant system of cytodifferentiation of adenohypophysial cells, but not in the context of high-grade histopathology; this terminology should be reconsidered in light of the WHO classification and perhaps warrants alternate terminology, such as “immature” rather than “poorly differentiated.” The second consideration is that the term PitNET, in analogy with NETs of other anatomical sites, implies a spectrum of behaviors including potential metastatic spread and, as such, does not require a specific term to distinguish locally aggressive and/or metastatic lesions.

Building on the WHO proposal for a common classification for NENs, we propose a common approach to the investigation of structure–function correlations in the endocrinology and pathology of NETs at all body sites as has been done in the pituitary gland.

Conclusions

  • Pituitary adenohypophysial neoplasms are epithelial and neuroendocrine in nature, thus are fully embedded in the concept of neuroendocrine neoplasm.

  • The vast majority are well-differentiated and fall into the category of NETs, which show a spectrum of behavior from indolent to locally and hormonally aggressive, and occasionally metastatic.

  • Recognition of adenohypophysial cell differentiation provides meaningful biological, clinical, prognostic, and predictive information that informs appropriate management.

  • Dynamic risk stratification for patients with PitNETs currently relies on the tumor cell type and subtype, proliferative activity and invasiveness; there is a need to implement grading and staging for these neoplasms to provide a predictive model of tumor behavior.

  • Standardized data collection and synoptic reporting are strongly recommended.

  • A multidisciplinary approach is essential for optimal outcomes.

  • PitNETs represent an advanced model of structure–function correlations that can be used for research in other NETs.