Molecular-Genetic and In Silico 3D Characterization of an Adamantinomatous Craniopharyngioma, the ‘Chameleon’ Among Pediatric Brain Tumors

1. Introduction

Craniopharyngiomas can be divided in adamantinomatous craniopharyngiomas (90%), affecting children between 5 and 15 years and adults between 45 and 60 years; and in papillary craniopharyngiomas (PCPs, 10%), affecting mostly adults between 40 and 55 years (Otte and Müller, 2021; Pfister et al, 2022). ACPs are slow-growing, non-cancerous epithelial tumours (WHO Grade 1) originating from cell remnants of the embryonic craniopharyngeal duct and Rathke’s pouch. They have an incidence of 0.5-2.5 per million population and account for less than 5% of all brain tumors (Karsonovich et al, 2026). Due to their initially non-specific clinical presentation, it can take several years before the diagnosis is made. Their proximity to the pituitary gland and hypothalamus leads to multiple hormone deficiencies which often worsen after treatment with surgery and/or radiotherapy (De Alcubierre et al, 2025; Zhou et al, 2021). In histological sections, tumor tissue typically stains positive for β-catenin, whose overall structure is represented in Figure 1, and negative for BRAF(V600E). By contrast, BRAF(V600E) is positive in PCP. ACP is caused by somatic driver mutations in the CTNNB1 gene on chromosome 3p22.1 (Sekine et al, 2002, Apps et al, 2020). Clinical management requires a multidisciplinary team including general paediatrician, endocrinologist, dietitian, ophthalmologist, neurooncologist, neuroradiologist and clinical psychologist. Patients require long-term follow up and close monitoring for relapse which occurs in 20 – 40% of cases. Overall, prognosis is good with a 30-year survival rate of up to 80% (Gan et al, 2023). Here, we present a young girl with ACP and panhypopituitarism who was investigated with biochemical (data not shown), radiological, histopathological, molecular-genetic and bioinformatics techniques.

2. Methods

1.1. Magnetic Resonance Imaging

Images were obtained on a GE HealthCare SIGNA Voyager 1.5T 70cm wide-bore MRI scanner using AIR Recon DL (AI-based reconstruction) and AIR coils. The following sequences were applied: T1W in coronal plane, T2W PROPPELLER in axial and sagittal planes, FLAIR PROPELLER in axial plane, DWI-ADC and SWAN. Slice thickness was 5 mm for all sequences except for SWAN which was 2.9 mm thick.

1.2. Histopathology and Immunohistochemistry

Standard tissue preparation and staining techniques were used for H&E, β-catenin, BRAF(V600E), CKC (AE1/AE3) and MGMT stains (Kazmi and Schuerch, 2022). Immunohistochemistry was carried out on an automated Roche BenchMark Ultra or Ultra Plus platform.

1.3. Single Nucleotide Polymorphism Array

DNA was extracted from patient blood leukocytes according to standard techniques and processed with Infinium CytoSNP-850K v1.4 Illumina BeadChip array (Illumina, San Diego, CA). Data were analyzed using the dedicated cytogenetics software BlueFuse Multi v4.5 (GRCh 38).

1.4. DNA Methylation Array

DNA was obtained from formalin-fixed, paraffin-embedded tumor curls (FFPE, tumor content 40-50%) and treated with sodium bisulphite (EZ DNA Methylation Kit, Zymo Research). Somatic DNA was further processed with the Infinium MethylationEPIC v2.0 Kit (Illumia, San Diego, CA). Data were sent for methylation profiling to Heidelberg Epignostix GmbH and analyzed with the CNS Tumor Methylation Classifier v12.8 (Capper et al, 2018).

1.5. Next Generation Sequencing

DNA was extracted from FFPE curls. Roche KAPA HyperCap Custom Probes (IDH1, IDH2, APC, CTNNB1, GNAS, H3-3A, H3-3B, BCOR, SMARCB1) were used for library preparation and Illumina NovaSeq for massive parallel sequencing by synthesis (SBS). DNA reads were interpreted with Mutect2-GATK (Broad Institute, Cambridge, MA) and Pindel (Wellcome Sanger Institute, Hinxton, UK).

1.6. In Silico Bioinformatic Analyses

The Ser33 to cysteine substitution was modeled in silico using both the structure of the Leu31–Thr40 β-catenin peptide fragment within the β-TrCP/Skp1/NRX-2776 ternary complex (PDB entry: 6M90) (Simonetta et al, 2019) and the AlphaFold-predicted model of full-length β-catenin (AF-P35222-F1-v6) (Jumper et al, 2021, Bertoni et al, 2025). The latter model was employed to validate the molecular dynamics results obtained for the short β-catenin peptide based on previously reported crystallographic data (Simonetta et al, 2019) and to avoid overfitting caused by the reduced length of the short peptide present in the X-ray structure. All modeling procedures were carried out in Coot 0.9.8.96 (Emsley et al, 2010). The resulting models were refined by simple molecular dynamics simulations using phenix.dynamics (Burnley et al, 2012) within the Phenix 2.0.5936 suite (Liebschner et al, 2019). Subsequent analyses, including bond distance measurements from the simulation models, rotamer evaluation according to minimal energy criteria, and figure preparation, were performed by PyMOL 2.5 (The PyMOL Molecular Graphics System, 2021).

3. Results

3.1. Case

An 8-year-old girl was referred to paediatric outpatients with a 6-month history of tiredness, nausea, abdominal pain and slow bowel movements. She took iron supplements for mild iron deficiency anaemia. On examination she looked extremely pale and had very fair hair, hypertelorism and diastemas. Weight and height were on the 9th centiles. She was referred to tertiary pediatric gastroenterology where she was extensively investigated with an abdominal MRI scan and endoscopies. Diagnoses of non-specific colitis and small intestinal bacterial overgrowth (SIBO) were made. The girl was referred to the regional genetics service who took detailed photographs and reviewed early childhood images. She was noted to have difficulties with sensory processing prompting involvement of primary mental health services (PMHS). Her school attendance dropped to 50%. One year later, she complained of intermittent headaches and sensitive scalp skin. Magnetic resonance imaging (Figure 2) and computed tomography (CT) of the brain revealed a large mass originating in the sella turcica. She was referred to the regional children’s hospital for further management. Preoperatively she was found to have secondary hypothyroidism and adrenal insufficiency which were treated with Levothyroxine and Hydrocortisone. Ophthalmology review revealed mild optic disc atrophy and restricted peripheral vision bilaterally. At the age of 10 years, she underwent a right pterional craniotomy and debulking of the tumor. Immunohistopathology was consistent with an ACP (Figure 3A-D). Postoperatively, she developed central diabetes insipidus with polyuria and polydipsia which was treated with Desmopressin. She gained weight rapidly (50th centile) caused by her increased appetite (hypothalamic hyperphagia) and steroid replacement therapy. Her post-surgery MRI brain scan demonstrated a high T1-weighted signal in the inferior sella turcica, likely to represent residual tumour tissue. However, a repeat pituitary MRI scan with contrast performed 4 months after surgery showed no significant residual disease. At 11 years of age, her weight gain slowed down, and she was commenced on daily subcutaneous growth hormone injections for growth hormone deficiency. The girl was referred to a dietitian and a clinical psychologist for advice and support.

3.2. SNP Array

SNP analysis detected a normal female profile [arr (X,1-22)] without clinically significant imbalances.

3.3. DNA Methylation Profiling

The copy number variation profile of the patient covering chromosomes 1 - 22, X, showed no evidence of copy number changes. No match was found among 184 tumor classes. The status of the MGMT promotor was unmethylated (Figure 4).

3.4. CTNNB1 Gene Sequencing

Sequencing of the CTNNB1 gene (NM_001904.4) identified an oncogenic variant, c.98C>G p.(Ser33Cys), with a variant allele frequency (VAF) of 6%. This variant is the cause of the excessive cell growth in this patient (ACP, ICD-O 9351/1).

3.5. In Silico Bioinformatic Analyses

Ser33 is located within the N-terminal domain of β-catenin (Figure 1), a highly flexible region that has not been fully structurally resolved in previous studies, including those of the full-length protein. Nevertheless, the serine residue at this position appears to play a key role in targeting the protein for proteasomal degradation, as Ser33, together with Ser37 and Thr41, is a phosphorylation site required for recognition by the SCF (Skp1–Cullin1–F-box) complex (Wu et al., 2003). The SCF complex is a multi-protein E3 ubiquitin ligase that mediates the ubiquitination of phosphorylated substrates, marking them for degradation by the 26S proteasome. In this complex, the F-box protein β-TrCP1 acts as the substrate recognition subunit, specifically binding phosphorylated degron motifs present in target proteins such as β-catenin (Bi et al., 2020; Baek et al., 2023; Colding-Christensen et al., 2023).

Figure 5A displays the complex structure of the solved Leu31–Thr40 β-catenin peptide fragment with the β-TrCP/Skp1/NRX-2776 ternary complex (PDB entry: 6M90) (Simonetta et al., 2019), where the location of Ser33 could be appreciated. Notably, NRX-2776 is a synthetic stabilizing ligand enhancing complex formation and facilitating structural characterization.

A focus on the interaction between the Leu31–Thr40 β-catenin peptide with β-TrCP (Figure 5B) shows that Ser33 is phosphorylated (pSer33) and, through this post-translational modification, it establishes a well-defined interaction network within the β-TrCP binding pocket, characterized by multiple hydrogen bonds with key residues such as Arg285, Tyr271, and Ser309. These residues directly participate in substrate binding and are conserved within the β-TrCP substrate recognition interface (Wu et al., 2003). In addition, they contribute to phospho-degron recognition by forming specific contacts with the negatively charged phosphate group, as the adjacent Asp 32. Thus, the phosphorylated serine 33 plays a central role in stabilizing the complex by promoting electrostatic and hydrogen-bonding interactions that ensure high-affinity and specific recognition, as already mentioned in literature (Wu et al., 2003).

The p.Ser33Cys variant was modeled in silico, as reported in the Methods section, based on the pSer33 b-catenin peptide complexed with β-TrCP (Figure 5C) and shows an altered binding mode to β-TrCP1 (Figure 5D). In detail, the substitution of Ser33 with cysteine disrupts the canonical interaction pattern, leading to increased distances between the peptide and the critical binding residues of β-TrCP1 (approximately 4.0 Å) and a significant reduction in stabilizing contacts as well as to a more relaxed relationship with the ubiquitination complex if compared to the phospho-Ser33 containing peptide. The Ser33Cys variant peptide behaviour was confirmed also for the structure modeled starting from the AlphaFold prediction of the full-length protein. The structural rearrangement highlights how the absence of phosphorylation at position 33 could compromise the molecular recognition by β-TrCP, ultimately affecting the stability and efficiency of the interaction. Namely, the destruction motif (also known as the DSGXXS phospho-degron sequence) of β-catenin is no longer able to be recognized and be bound by the β propeller of β-TrCP1 (Winston al., 1999), possibly leading to the accumulation of β-catenin and the failure of its ubiquitination.

4. Discussion

Headaches and vomiting caused by raised intracranial pressure, and visual impairment are the commonest symptoms of children with ACP. Our patient was atypical as she complained of abdominal pain, nausea and delayed defecation which triggered the referral to a paediatric gastroenterologist. Her tiredness was attributed to mild iron deficiency anaemia. Other signs and symptoms observed in this group of patients include unexplained variations in weight, height or appetite, early or delayed onset of puberty, polydipsia, polyuria, sleep disturbance, intolerance to low temperatures, behaviour difficulties, hypotension, muscle weakness and seizures (Child Growth Foundation, 2000). The time from onset of symptoms to diagnosis of ACP can vary from 8 months to 8 years (Children’s Cancer and Leukaemia Group, 2021). Gross total resection (GTR) of the tumor is the preferred therapeutic option. However, this goal is frequently unachievable due to its sticky, oily nature and proximity to vital neurovascular structures. Instead, subtotal resection (STR), which was performed in this patient, can be combined with radiation therapy if deemed necessary. In children proton beam therapy is superior to radiotherapy as the brain tumor can be targeted more precisely, and there is no associated exit dose that damages healthy tissue (Bidur and Prasad, 2015; Graffeo et al, 2018). During a psychological assessment three months post-surgery our patient described her quality of life (QoL) as improved compared to pre-surgery. Steinbok (2015) reports long-term QoL based on questionnaire survey results as good in 50% of patients. ACP is considered a chronic disease, and survivors require annual MRI brains scans for at least 10 years (European Society for Paediatric Oncology, 2021).

Different state-of-the-art genetics and bioinformatics methods were applied in this case. DNA methylation is an epigenetic process that impacts on gene expression. Through addition of a methyl group to the cytosine ring at a cytosine-phosphate-guanine (CpG) island, transcription factors (TFs) can no longer bind to DNA, and the respective gene is inactivated (‘silenced’). Genome-wide methylation profiles, specifically hypomethylation, and the methylation status of tumor-specific genes are important for cancer diagnostics (Ibrahim et al, 2022; Marrero-Gutiérrez et al, 2025). Methylated MGMT and/or low expression of MGMT (O-6-methylguanine-methyltransferase) make ACPs more sensitive to treatment with the alkylating prodrug Temozolomide (Zuhur et al, 2011). Cavalheiro and co-workers (2026) reported a teenage boy with recurrent ACP who underwent treatment with monthly 7-day courses of oral Temozolomide for 1 year during which his large tumor shrank by 50%. It is noteworthy that Temozolomide has significant side effects. Therefore, it cannot be considered as first-line treatment for patients with ACP. Our patient’s MGMT promotor was unmethylated. Using FFPE curls instead of flash-frozen tumor tissue can lead to a falsely negative genome wide methylation profiling result.

The human CTNNB1 gene has 16 exons and spans ca. 65kb (3:41194741-41260096, GRCh38). Exons 1 and 16 contain untranslated regions (UTRs). Germline loss of function mutations of this gene cause autosomal dominant exudative vitreoretinopathy type 7 (OMIM 617572) and neurodevelopmental disorder with spastic diplegia and visual defects (OMIM 615075). Somatic, de novo missense mutations of CTNNB1 cause ACP, colorectal cancer, hepatocellular carcinoma, medulloblastoma, ovarian cancer and pilomatricoma (Foreman et al, 2023; Gao et al, 2017). The pathogenic single nucleotide variant identified in this patient (Chr3: 41224610, GRCh38) has been described before in different cancerous and non-cancerous tumors (Koch et al, 2001; Campanini et al, 2010; Cani et al, 2011; Ohata et al, 2018; Seki-Soda et al, 2022; Fonseca YG et al, 2026; Rheki et al, 2026). It resides in exon 3 of CTNNB1 which is a hotspot for tumor-inducing variants (Landrum et al, 2016; Kim and Jeong, 2019).The CTNNB1 protein (catenin beta 1, beta-catenin, β-catenin) belongs to the armadillo family of proteins which have important structural as well as functional roles. The name stems from the armadillo (ARM) repeat domains which consist of 40-45 amino acids that form 3 alpha helices arranged in a hairpin structure. They have been likened to the flexible armor of the placental mammal armadillo and facilitate intercellular adhesion (Hatzfeld, 1999). Catenin beta 1 is 781 amino acids (aa) long and has a molecular mass of 85.5 kDa. It can be divided in 3 segments (Figure 1): an N-terminal domain of ca. 150 aa which contains phosphorylation sites, a core domain with 12 ARM repeats (524 aa) and a C-terminal domain of ca. 100 aa which acts as a transcription transactivation domain (TAD). The serin to cysteine conversion observed in our patient is located within the N-terminal domain of β-catenin (Seal et al, 2026; Stelzer et al, 2016).

Beta-catenin is a subunit of the cadherin protein complex and plays an important role in the Wnt (wingless/int-1) signalling pathway which regulates embryonic development, cell determination and cell migration. In the absence of a Wnt ligand, β-catenin binds to the destruction complex and is degraded. The degradation can be triggered by β-TrCP binding to a phosphorylated form of b-catenin, ubiquitination and proteosome targeting. In silico modeling and molecular dynamics simulations highlight that the Ser33-to-Cys substitution markedly disrupts the canonical binding geometry within the β-TrCP pocket, resulting in suboptimal positioning of the b-catenin moiety, increased intermolecular distances (~4.0 Å), and a substantial loss of stabilizing interactions, thereby emphasizing how the absence of pSer33 severely compromises β-TrCP recognition and binding efficiency, thereby its degradation.

In the presence of a Wnt ligand or in the absence of a phosphorylation reaction, β-catenin accumulates in the cytosol and translocates to the nucleus where it activates the transcription of various genes. If β-catenin is defective, it becomes resistant to degradation and switches on nuclear genes in an unregulated, tumorigenic fashion (Campanini et al, 2023; Lainšček et al, 2025). Based on a multi-omics analysis Wang et al. (2023) distinguish 3 molecular ACP subgroups (Wnt, ImA and ImB) where the Wnt group has longer event-free survival and where the ImA/ImB groups show a better response to immune checkpoint blockade (Gonzalez-Meljem et al, 2025; Cacciotti et al, 2020).

To our knowledge, this is the first detailed case report of a child with ACP and c.98C>G p.(Ser33Cys). The patient presented with atypical, chronic symptoms providing further evidence for the ‘chameleon-like’ nature of this rare brain tumor. Every ACP should be genetically confirmed by NGS, in addition to immunohistochemical staining. 3D modeling suggests that the variant determines a failure of b-catenin degradation with the consequence of its possible accumulation and translocation to the nucleus for activation of Wnt genes. Novel treatment with immune checkpoint inhibitors (ICI) may become available in future for patients with ACP.