Biological Behaviour of Craniopharyngiomas

Craniopharyngiomas (CPs) are benign tumours (WHO grade 1) that develop in the sellar region, which is an anatomical structure limited ventrally by the cranial base, dorsally by the dorsum sellae (with the suprasellar cistern and optic chiasm immediately superior to this), laterally by the cavernous sinuses and carotid arteries, and caudally by the brain stem. CPs were first described by the Viennese pathologist Jakob Erdheim in a 200-page-long report published in 1904 [1-3]. Despite their benign histological nature, CPs can be clinically challenging due to their location and their tendency to invade surrounding structures such as the pituitary gland, hypothalamus, and visual pathways. Current treatments are surgery followed by radiotherapy, but these modalities are not always curative and can often contribute to further damage. Overall, CPs are associated with a high degree of morbidity, leading to poor quality of life and increased mortality rate on long-term follow-up [4].

There are two types of CPs: (i) The adamantinomatous form of CP (ACP) is the most frequent pituitary tumour in children and shows a bimodal peak of distribution (5–15 years in the childhood-onset ACP and 45–60 years in the adult-onset ACP); (ii) The papillary form (PCP) is mostly an adult tumour (peak at 40–45 years). Research from the last 10 years has demonstrated that these two tumour types represent distinct identities each with specific genetic, epigenetic, and pathological features. In this minireview, we will discuss the main features that differentiate ACP and PCP, and elaborate how the biological differences have helped identify novel targeted treatments. Further readings are recommended to cover more detailed pathological and clinical descriptions [4-11].

ACPs are tumours that usually contain solid as well as cystic components. The solid part of the tumour comprises epithelial tumour cells, which are highly heterogeneous and include the palisading epithelium, stellate reticulum, and groups of cells forming whorl-like structures [12] (Fig. 1). The palisading epithelium and stellate reticulum form finger-like protrusions near the invasive front, which usually contain a string of cell whorls inside [13]. Surrounding the epithelial tumour, ACPs often contain glial reactive tissue, mostly comprising astrocytes and immune cells. The proportions of tumour epithelium and glial reactive tissue can vary considerably between ACP samples: for instance, some tumour samples may contain mostly tumour epithelium with little or no reactive glial tissue, whilst others may be comprised mostly of glial tissue with little epithelial component [14]. Other histopathological features include calcification, which can be observed by computerised tomography scans, and the presence of nodules of wet keratin (containing cells without visible nuclei). Both of these features help establish a diagnosis of ACP. ACP tumours can hold one or several cysts filled with a dark fluid commonly referred to as machine oil, which is rich in lipids and inflammatory mediators.

PCPs are solid epithelial tumours, characterised by the presence of a well-differentiated non-keratinising squamous epithelium supported by fibrovascular cores (Fig. 1). Fibrovascular cores are tubular structures that contain stroma and blood vessels, lined by a well-defined pseudostratified epithelium (Fig. 1). PCPs are rarely cystic and do not show calcification.

Mutations in CTNNB1 were first reported in ACPs by Sekine et al. [15] in 2002. This finding has been subsequently recapitulated in many independent studies, and CTNNB1 mutations have been identified in 16–100% of the tumours analysed [16]. These mutations, which affect mostly the amino acids encoded by exon 3 of CTNNB1, are predicted to result in the expression of a degradation-resistant form of the protein leading to the activation of the WNT/β-catenin pathway [17]. Failure to identify the mutation in all ACP samples has led to the speculation that other genetic mutations may underlie ACP tumourigenesis. Indeed, coexisting mutations in BRAF(V600E) and CTNNB1 (T41I) have been identified in 2 ACP tumours [18]. Sanger sequencing of specific cell populations has furthered controversy on whether the mutations are clonal or present only in some but not all the epithelial tumour cells [15, 19, 20]. Recently, laser capture microdissection was combined with tagged-amplicon deep sequencing, an ultrasensitive approach that detects very low mutant allelic frequencies, to screen 22 ACP tumour samples. CTNNB1 mutations were identified in all samples including those with very low mutant allelic frequencies [21]. These data suggest that failure to identify CTNNB1 mutations in a low proportion of ACP tumours may be due to the lower sensitivity of the sequencing technology used in previous studies (e.g., Sanger sequencing, single-strand conformation polymorphism analysis, exome sequencing, and targeted next-generation sequencing). Therefore, if there are other oncogenic mutations in human ACPs, these are likely to be rare compared with CTNNB1 mutations.

Murine studies have confirmed that CTNNB1 mutations are oncogenic drivers, i.e., capable of initiating and sustaining tumourigenesis. The expression of a functionally equivalent form of stabilised β-catenin in either pituitary embryonic precursors or SOX2+ adult stem cells results in the formation of ACP-like tumours in mice [22, 23]. These tumours resemble some of the histological and radiological features of human ACP [24], but do not calcify or show wet keratin. A common characteristic in mouse and human ACP is that nucleocytoplasmic accumulation of β-catenin occurs only in sporadic cells, frequently forming cell clusters that overlap with the epithelial whorls previously described or dispersed throughout the tumour as single cells (Fig. 1) [25]. The reason why protein accumulation occurs only in a small cell fraction, despite the presence of the CTNNB1 mutation throughout the tumour, remains unknown. These cell clusters, showing nucleocytoplasmic accumulation of β-catenin, are not present in PCP or any other pituitary tumour [26]. As well as histologically, gene expression profiling has demonstrated that mouse and human clusters are equivalent molecular structures [14]. Moreover, the pattern of gene expression in the clusters resembles the enamel knot, a critical signalling centre that controls epithelial and mesenchymal interactions during tooth development. These similar molecular signatures have provided a molecular paradigm that explains the histological similarities between ACP and tooth development and tumours of the teeth, which have been reported for over a century [2, 27].

ACPs and PCPs have a low mutation rate (non-synonymous mutation rate of 0.9 per Mb), which is expected in benign grade I tumours [28]. They have stable genomes and gains or losses of large chromosomal regions are rare. In one study, more focal losses and gains of unknown function were identified [29]. The methylomes are different between ACPs and PCPs, a feature that facilitates molecular diagnosis [30, 31], but the functional significance of distinct epigenetic landscapes remains unknown.

PCPs are likely to be driven by mutations in BRAF, specifically p.BRAF-V600E. This mutation has been identified in the vast majority of PCP tumours analysed and the expression of the mutant protein confirmed by immunohistochemistry using an anti-BRAF-V600E antibody (Fig. 1) [28, 32]. Although this mutation is predicted to result in the activation of the MAPK/ERK pathway in all tumour cells, immunohistochemistry against phospho-ERK1/2 (pERK1/2), a read-out of the active MAPK/ERK pathway, has revealed that only a small proportion of epithelial cells lining the fibrovascular cores activate this pathway, despite the expression of BRAF-V600E throughout the tumour [32]. In this study, these pERK1/2+ cells were shown to express the pituitary stem/progenitor markers SOX2 and SOX9, suggesting that these lining cells may represent tumour stem cells. Moreover, the vast majority of the Ki67+ proliferative cells are contained within the SOX2/SOX9+ compartment around the fibrovascular cores. Mouse models expressing the p.BRAF-V600E mutation have been generated, but perinatal lethality has prevented assessment of the potential tumourigenic effect [32]. Nonetheless, close examination of these murine models has revealed that the expression of this oncogenic driver in early pituitary precursors leads to the expansion of SOX2/SOX9+ stem cells, which are highly proliferative and show impaired differentiation. Together, studies in the mouse and humans suggest a likely tumourigenic mechanism, by which the activation of the MAPK/ERK pathway within SOX2/SOX9 stem cells may lead to tumour formation.

Molecular profiling and immunohistochemistry analyses have revealed that the cluster cells in both mouse and human ACPs contain senescent cells. Senescence is defined as a cellular state that is characterised by a permanent cell cycle arrest due to the expression of cell cycle inhibitors (e.g., p16 and p21) [33, 34]. Senescence is induced by several stressors that cause DNA damage, among them radiotherapy, chemotherapy, and oncogenic signalling. Despite the fact that senescent cells are unable to re-enter the cell cycle (except if cell cycle arrest pathways are inactivated by genetic or epigenetic mechanisms), these cells are metabolically very active and secrete a plethora of growth factors and inflammatory mediators referred to as the senescence-associated secretory phenotype (SASP) [35]. A bulk of research has shown that senescent cells underlie several ageing-related diseases or even contribute to organismal ageing through SASP activities [36]. In cancer, senescent cells are a double-edged sword that can prevent expansion of cells harbouring DNA damage autonomously but also promote tumour expansion and progression to malignancy in a cell non-autonomous manner [35, 37].

Studies in ACP mouse models have provided insights into the role of senescent cluster cells in initiating tumour formation. Initial experiments, in which SOX2+ pituitary stem cells were targeted to express oncogenic β-catenin and simultaneously a fluorescent reporter (e.g., yellow fluorescent protein) demonstrated that these stem cells are the cell of origin of the β-catenin-accumulating cell clusters, but not of the tumours, which are derived from a different cell lineage [22]. Based on these results, a model of paracrine tumourigenesis was proposed, in which the cluster cells may be able to induce tumour formation in a paracrine manner, but the underpinning mechanisms were not understood (Fig. 2). More recently, it has been shown that mouse and human clusters contain senescent cells with an activated SASP, and that the attenuation of the senescent/SASP response in murine cluster cells, either genetically or in aged mice, result in a significant reduction in tumour-inducing potential [38, 39].

The significant increase in knowledge of tumour biology that has accumulated over the last few years has led to the identification of novel targetable pathways in both PCP and ACP. The presence of p.BRAF-V600E mutations in PCP patients has provided a molecular rationale for the use of MAPK/ERK pathway inhibitors in these patients. Although the pERK1/2+ cells are just a minority of the tumour cells in PCP tumours [32], the inhibition of the MAPK/ERK pathway using BRAF-V600E or MEK inhibitors, alone or in combination, has given excellent results in patients [40, 41]. The success in these small studies has led to a clinical trial in BRAF-V600E-positive PCP patients using a combination of vemurafenib (BRAF-V60E inhibitor) and cobimetinib (a MEK inhibitor) (ClinicalTrials.gov Identifier: NCT03224767).

In ACP tumours, however, the identification of CTNNB1 mutations leading to the activation of the WNT/β-catenin pathway has not been translated into novel targeted treatments due to the difficulty in targeting this pathway without causing unacceptable toxicity. However, gene profiling has revealed other potential targetable pathways downstream of the WNT/β-catenin pathway. Inflammatory mediators (e.g., IL-6 and IL-1) have been identified both in the solid and cystic tumour compartments, suggesting a critical role of these factors in ACP pathogenesis [14, 42, 43]. Supporting this hypothesis, 2 patients have been treated with tocilizumab, an IL-6 inhibitor, leading to a discreet improvement in disease management [44]. Sonic hedgehog, a signalling factor with critical roles during development, was found to be upregulated in mouse and human ACP [45] and further confirmed in other studies [14, 31, 43, 46]. The activation of the SHH pathway can be targeted with several inhibitors, including vismodegib, a clinically approved drug that is used against other human cancers (e.g., medulloblastoma). Unfortunately, pre-clinical data in vitro and in the ACP mouse model as well as patient-derived xenograft mice have shown that vismodegib treatment leads to increased tumour cell proliferation, premature tumourigenesis, and reduced mouse survival [47].

A recent study has revealed that the MAPK/ERK pathway is activated in human and mouse ACP tumours, as evidenced by the expression of p-ERK1/2 [14]. Since ACPs do not carry mutations in MAPK pathway components, these data suggest that the pathway is activated in a paracrine manner. Indeed, cluster cells express many ligands known to signal through this pathway, such as fibroblast growth factors, epithelial growth factors, and platelet-derived growth factors [14, 45]. Interestingly, the inhibition of the MAPK/ERK pathway using the MEK inhibitor trametinib has been shown to result in reduced proliferation and increased apoptosis in both mouse and human ACP tumours in vitro [14]. There is currently an open clinical trial of single-agent tocilizumab (IL-6R inhibitor; ClinicalTrials.gov # NCT03970226), and other multicentre trials are in development.

ACPs and PCPs are relatively simple tumours carrying mutations in either CTNNB1 or BRAF (p.BRAF-V600E), respectively. At the cellular level, senescence has been identified as a potentially pro-tumourigenic mechanism that may initiate ACP tumourigenesis in mice and promote growth and invasion in human ACP. The accumulated knowledge on the biology of these tumours is being translated into clinical and pre-clinical trials testing novel targeted therapies. It is likely these studies will provide efficacious medical treatments against these aggressive tumours.

The authors wish to thank Drs Scott Haston, Romain Guiho, and Gabriela Carreno for their help with the Figures.

Ethical approval was not required since this paper does not concern animal experimentation or the use of human volunteers.

The authors have no conflicts of interest to declare.

Funding for this research was provided by Cancer Research UK, Children’s Cancer and Leukaemia Group, Children with Cancer UK (15/190), MRC (MR/M125/1 and MR/L016729/1), Brain Tumour Charity (SIGNAL and EVEREST), Great Ormond Street Hospital Children’s Charity, the Lister Institute of Preventive Medicine, the Morgan Adams Foundation and National Institute of Health Research Biomedical Research Centre at the Great Ormond Street Hospital for Children NHS Foundation Trust, and the University College London. J.P.M.-B. is a Great Ormond Street Hospital for Children’s Charity Principal Investigator.

This paper was written by both authors.