Introduction
Adamantinomatous craniopharyngioma (ACP) is the most common non-neuro-epithelial brain tumour in children [
31,
43,
47]. Although not metastatic and histologically benign, ACP is invasive and prone to recurrence after surgery, the conventional mode of treatment. Adamantinomatous craniopharyngioma often behaves aggressively with invasion of the hypothalamus and visual pathways. Therefore, total resection of the tumour without damage to vital surrounding structures such as the hypothalamus and optic chiasm is not always possible. In these children, radical excision is associated with unacceptable morbidity and mortality whilst subtotal resection without adjuvant radiotherapy predisposes to a high (>60%) 3-year recurrence risk and further hypothalamic and visual compromise [
34,
38,
44]. Although a conservative surgical approach with adjuvant radiotherapy for residual tumour has been recently adopted, both tumour recurrence and treatment-associated morbidity are still high. Consequences of the tumour and its treatment include obesity with associated Type 2 diabetes mellitus, learning difficulties, visual impairment and panhypopituitarism, which can be life-threatening. This poses a heavy burden to parents and carers as well as a heavy cost for health services.
A crucial role for Wnt/β-catenin signalling in the aetiology of ACP has been firmly established. Activating mutations in the gene encoding β-catenin (
CTNNB1) have been identified in the majority of samples of human ACP [
9,
51]. Recently, phenotypic analysis of a mouse model (
Hesx1
Cre/+;
Ctnnb1
lox(ex3)/+) expressing a mutant form of β-catenin that cannot be degraded, leading to over-activation of the pathway, has confirmed that these mutations, rather than a second hit, are causative of the tumours [
18]. A characteristic histological finding in both human and mouse ACP is the restricted nucleocytoplasmic accumulation of β-catenin and over-activation of the Wnt/β-catenin pathway in very few cells that form clusters (β-cat
nc clusters). Despite harbouring the tumorigenic mutation in the β-catenin gene, all other cells only show the normal β-catenin staining in the cytoplasmic membrane without any nucleocytoplasmic accumulation (non-cluster β-cat
m cellular component of ACP) [
18,
25,
26,
32]. Beyond the relevant diagnostic value of this unique histological feature to distinguish ACP from other pituitary tumours, little is known about the reason for this specificity or the relevance of the cluster cells in the disease. Recent evidence points towards a role in tumour progression and invasiveness into the brain [
8,
27‐
29]. A deeper analysis of these cellular structures may provide novel insights into the pathogenesis of ACP resulting in the identification of new disease biomarkers and pharmacological targets.
An open question is the contribution of pituitary progenitors/stem cells (PSCs) in the aetiology and pathogenesis of ACP. Previously, we have demonstrated that one of the initial effects of mutated β-catenin is the increase of PSCs in the ACP murine model compared with control pituitaries. At late gestation and early postnatal stages, a proportion of β-catenin-accumulating cells within the clusters express the stemness marker SOX2 in the murine pre-tumoral pituitary [
14,
18]. However, SOX2 is not expressed in the β-catenin accumulating clusters in human ACP, but rather in sporadic cells within the tumour [
17]. This raises the question of what the connection is between the human and mouse β-catenin-accumulating clusters.
In this study, we have used the ACP mouse model to investigate the pathogenesis and the possible involvement of PSCs in the aetiology of human ACP. We demonstrate that β-catenin-accumulating cluster cells have functional and molecular characteristics of pituitary progenitors/stem cells. We present their global gene expression profile and reveal novel genes and signalling pathways expressed in both mouse and human ACP. This molecular analysis highlights interplay between clusters and surrounding cells through the secretion of signals involved in proliferation, survival, stem cell maintenance, cell migration and tumorigenesis.
Discussion
In this paper, we have utilised a recently generated mouse model for ACP and carried out an unbiased molecular screen demonstrating the de-regulation of numerous genes and signalling pathways with tumorigenic potential in both mouse and human ACP. We provide molecular and expression data indicating that β-cat
nc cluster cells act as a source of mitogenic and pro-survival signals for themselves and surrounding β-cat
m tumour cells. Our global gene profiling analysis has revealed that members of the SHH, FGF and BMP family of morphogens, which are critical during normal pituitary development [
23,
30,
39], show higher expression levels in the β-cat
nc cluster cells in both mouse and human ACP. As well as providing novel insights into the pathogenesis of human ACP, this research has identified potential therapeutic targets for these tumours.
SHH is required during the normal development of several organs and in adulthood this pathway plays an important role in the maintenance of stem-cell niches [
1,
30,
53]. Over-active HH signalling occurs in numerous human cancers and can be caused by either a mutation-driven (ligand-independent) mechanism or ligand-dependent signalling (i.e. autocrine/paracrine signalling). Inactivating mutations in PTCH1 or more rarely, activating mutations in SMO have been identified in most sporadic medulloblastomas [
19,
48,
64]. Loss-of-function mutations in PTCH1 underlie the molecular cause of Gorlin syndrome (also known as nevoid basal cell carcinoma syndrome), a rare condition characterised by an increased risk of developing various tumours, commonly medulloblastoma, rhabdomyosarcoma and basal cell carcinoma [
58]. In contrast, many epithelial cancers, including small-cell lung cancer, pancreatic, prostate and gastrointestinal malignancies, also exhibit over-active HH pathway, but this is caused by increased expression of SHH ligand without known mutations in pathway components. In these tumours, SHH is released to the stroma (paracrine action) where it promotes tumour growth, infiltration and angiogenesis. In addition, activation of the SHH pathway in stromal cells is proposed to feedback signals to induce a more suitable environment for the SHH-expressing cells [
58]. SHH autocrine signalling within the epithelial cells also appears to be important for cell renewal of cancer stem cells in breast cancer, multiple myeloma and chronic myelogenous leukaemia stem cells [
64]. Our expression data strongly support a key role for HH signalling in the pathogenesis of mouse and human ACP, whereby SHH expression in the β-cat
nc cell clusters activates the pathway in both β-cat
nc (autocrine action) and β-cat
m cells (paracrine signalling). Whether over-activating mutations in the HH pathway components may underlie ACP tumorigenesis requires further research.
Over-activating mutations in FGF receptors have been identified in several human cancers, including breast, bladder, prostate, endometrial and lung cancers as well as haematological malignancies [
62]. Expression levels of the FGF receptors 1–4 remained essentially unchanged between β-cat
nc clusters and surrounding β-cat
m cells. We noticed, however, that expression of
Fgfrl1 (FGF receptor like-1), a recently identified receptor lacking the critical intracellular domain responsible for signal transduction and thought to act as a decoy receptor able to inhibit FGF signalling in the expressing cells [
55], was expressed 8.99-fold higher in the cluster cells. Although bestowed with several functional activities, FGFs are potent mitogenic signals in a variety of cell contexts, including the pituitary gland. It is tempting to speculate that β-cat
nc cell clusters may act as a source of FGFs inducing surrounding β-cat
m cells to actively divide, while protecting themselves by expressing higher levels of
Fgfrl1. This could explain the paradoxical observation that cluster cells in both mouse and human ACP remain quiescent (e.g. Ki67 negative), but cells in the immediate vicinity are mitotically active [
18]. Related to this notion, β-cat
nc cells express high levels of anti-apoptotic proteins of the Bcl2/Bcl-xL family.
An interesting concept that can be inferred from our data relates to the origin of the β-catenin accumulating (β-cat
nc) cluster cells. In mouse, these cells have an embryonic origin in Rathke’s pouch undifferentiated precursors (i.e. HESX1 and SOX2 expressing cells) [
18]. However, SOX2 is not uniformly expressed in all clusters or every cell within the clusters in the mouse pre-tumoral pituitary at late gestation or early postnatal life (Fig.
1b) [
18]. Subsequently, in advanced mouse tumours, SOX2 expression is rarely observed, although at this stage, β-cat
nc clusters are not identifiable and most of the tumour cells exhibit accumulation of β-catenin in the nucleus, cytoplasm or both [
18]. Similarly, although SOX2 positive cells have been identified in human ACP [
17] (our unpublished observations), they are not present in the β-cat
nc cluster cells. This suggests that β-cat
nc cluster cells in human ACP and β-catenin-accumulating cells in advanced mouse ACP may derive from undifferentiated precursors/stem cells present in the embryonic or postnatal pituitary, but have lost cell stemness and down-regulated SOX2 expression.
Our data on telomere length on mouse and human β-catnc cluster cells are compatible with the idea that this similar cellular component of the tumours may correspond to different temporal stages of ACP development. Mouse β-catnc cluster cells in the pre-tumoral pituitaries exhibit longer telomeres than surrounding β-catm cells, suggesting the presence of stem cells in these structures. In contrast, β-catnc cluster cells in fully established human ACP have shorter telomeres than the rest of the non-cluster β-catm cells, suggesting that they do not contain stem cells. This could indicate the cellular ontogenesis of the β-catnc clusters in ACP, whereby SOX2+ve stem cells with long telomeres are initially present but differentiate at later stages of tumour development losing SOX2 expression and shortening their telomeres. Unfortunately, species differences in ACP prevents us from testing this idea experimentally as fully established tumours in our mouse model do not contain β-catnc clusters and early stages of human ACP are not available.
Regardless, it is tempting to speculate that human ACP may be a tumour of stem cell origin, in which pituitary progenitors/stem cells played a role solely at an early stage of tumorigenesis. Advanced and clinically relevant human ACP would be devoid of such cells. This is different from the general dogma for cancer stem cells, where they self-renew and give rise to progeny that populate the tumour bulk. Instead, the contribution of the stem cell could be to initiate a cascade of signalling events leading to the perpetuation of a pathogenic unit (β-cat
nc cluster cells and microenvironment), without further need for the original stem cells. Therefore, the tumour-initiating mutation may occur in a progenitor/stem cell but the propagation of the tumour may require a different cell type. This is in agreement with findings in other tumours such as glioma and medulloblastoma [
37,
50]. The data presented here strongly suggest an important autocrine/paracrine function of the cluster cells as signalling centres within the tumour and the interplay with the stromal cells, which is maintained after loss of SOX2 expression.
Current treatments for human ACP are far from ideal and associated with high morbidity and significant mortality [
31,
43]. Our data highlight several genes and pathways likely to play essential roles in the pathogenesis of human ACP, as they do in several other human cancers. For some of these pathways, specific small-molecule inhibitors have been designed and their efficacy is currently being tested in a variety of clinical trials [
35,
57]. The research presented here is expected to promote the development of chemical-based therapies leading to more efficient and safer treatments for these childhood tumours.