Formalin fixation and paraffin embedding (FFPE) have a negative effect on nucleic acids, leading to RNA fragmentation, cross-linking and reduced RNA yield and integrity. These factors make RNA extraction and downstream applications challenging [8, 30, 37, 44]. Despite these challenges, FFPE samples are a valuable resource for retrospective studies, especially for rare diseases, long-term outcomes or under-researched patient cohorts [8, 30, 37, 44]. In addition, FFPE samples can be stored long-term at low cost, making them ideal for linking RNA expression data to clinical outcomes. While issues such as RNA degradation, fragmentation and low yields can affect the accuracy of transcriptome sequencing, advances in extraction techniques, sequencing technologies and bioinformatics are helping to overcome these limitations [37, 38]. Previous studies [8, 30, 37, 44] have shown that FFPE samples are suitable for RNA isolation and whole-transcriptome sequencing, and our results support these conclusions.

We observed the overexpression of several keratin genes (KRT5, KRT7, KRT8, KRT10, KRT14, KRT17, KRT19, KRT20 and KRT75) in CPs compared to normal pituitary gland. The keratins overexpressed in CPs are primarily soft epithelial keratins, reflecting the tumor’s origin from the epithelial remnants of Rathke’s pouch. These tumors often include cysts lined by squamous epithelium, which explains the increased keratin expression and supports previous findings [5, 25, 35]. In addition, the adamantinomatous subtype (ACP) commonly exhibits a histopathological feature known as “wet keratin”, which is associated with a high accumulation of keratin proteins. KRT75, a hair follicle-specific keratin, is not thought to contribute to the pathology of CPs. However, a previous study [35] demonstrated IHC positivity for human hair keratin in ACP, suggesting possible follicular differentiation. Generally, the expression of keratin genes is low in normal pituitary gland, except KRT8, KRT10 and KRT18, according to a study researching tissue-specific expression patterns of keratin genes [18]. Understanding keratin expression profiles may enhance the diagnosis of CPs, especially when limited tissue samples are available. Recent single-cell and spatial studies confirm that ACPs exhibit cellular diversity, encompassing classic epithelial and whorl-like cells as well as senescent and stem-like tumor cell populations. Our findings of distinct keratin gene expression and Wnt pathway activity are supported in a recent study [39], which highlighted the spatial arrangement and potential functional roles of these diverse cell types.

In addition to the keratin overexpression, we observed elevated expression of genes encoding several important signaling factors, including sonic hedgehog (SHH), fibroblast growth factors (FGF), transforming growth factor-beta (TGF-β) and bone morphogenetic proteins (BMPs). These factors are secreted by epithelial whorls, which are a hallmark of ACP [2, 3, 9, 43]. Notably, ACPs exhibited evidence of active hedgehog signaling, as shown by the significant overexpression of SHH compared to normal pituitary gland and PCP. The overactivation of SHH signaling may be linked to cross-regulatory interactions between the hedgehog and Wnt/β-catenin pathways, which are known to influence each other. Moreover, our results show upregulation of several targets of the Wnt signaling pathway, including LEF1, WNT3A, WNT7A, AXIN1, AXIN2, APC, and NOTUM, further supporting the involvement of Wnt/β-catenin signaling in ACPs. The upregulation of LEF1, a transcription factor activated by β-catenin, highlights the pathway’s contribution to the transcriptional changes seen in ACPs. Similarly, overexpression of WNT3A and WNT7A, key ligands in the Wnt signaling pathway, likely enhances the activation of downstream Wnt signaling cascades. ACP, an important tumor suppressor gene, and NOTUM, a Wnt inhibitory factor, were similarly upregulated, further reflecting dysregulation of Wnt signaling in ACPs.

We also observed downregulation of genes related to pituitary hormone production, including LHX3, FSHB, GH1, and TSHB, all of which play an important role in regulating pituitary function. LHX3 is a critical transcription factor required for pituitary development and the expression of various pituitary hormones. FSHB encodes for the beta subunit of FSH, which plays a vital role in reproduction. GH1 is a key gene that encodes the growth hormone, which is crucial for growth and metabolism, while TSHB encodes the beta subunit of TSH, which is essential for the regulation of thyroid function.

The WHO CNS Tumor Classification also specifies the expression of p63 in epithelial layers when determining the immunophenotype. Our DGE supports this, since the TP63 gene was upregulated in ACP and PCP compared to normal pituitary gland. The WHO CNS Classification also states variable expression of SOX2 and SOX9. In our study, we observed only a weak statistical difference in expression of SOX2 and SOX9 between normal pituitary tissue and ACP. However, SOX9 expression varied significantly between ACP and PCP and between PCP and normal pituitary gland. These results are consistent with the complex immunophenotypic profile of CPs and further support the diagnostic utility of these markers.

The detection of BRAF p.V600E mutations in all PCP samples is consistent with previous studies and underlines the role of the MAPK signaling pathway in the pathogenesis of PCP. This mutation is a potential target for molecular therapies with BRAF inhibitors, which are already the standard of care in other cancers. All ACP samples had mutations in exon 3 of the CTNNB1 gene, which is consistent with activation of the Wnt signaling pathway. The presence of these mutations in all ACP samples confirms their important role in tumor development and progression. Interestingly, our study found that some mutations were only detectable by specific sequencing technologies, such as RNA sequencing and targeted DNA sequencing, emphasizing the importance of using comprehensive molecular techniques for accurate mutation detection, and critical evaluation of methods used for mutation detection. One of the most notable observations was the occurrence of specific CTNNB1 mutations in recurrent ACP samples, suggesting that these mutations could be a prognostic factor in defining tumor recurrence, as suggested in previous studies [13, 15]. This could lead to improved monitoring and management of patients harboring these mutations, and potentially assist in the early detection of ACP recurrence. However, further investigations in a larger patient cohort are needed to confirm the prognostic significance of these specific mutations. We also recommend conducting longitudinal studies to examine how such mutations evolve over time and contribute to tumor progression and recurrence.

Interestingly, among the six ACP samples that either presented as recurrent tumors or where the patients developed ACP again later in life, five exhibited two distinct mutations in the CTNNB1 gene: T41I and S37Y. Notably, mutations affecting T41 have previously been linked to a poorer prognosis [13]. This association suggests that specific alterations in the CTNNB1 gene, particularly those involving T41, may contribute not only to the recurrence of ACP but also to the overall severity and progression of the disease.

To our knowledge, this is the first study to utilize spatial transcriptomic based analysis for CPs. Spatial transcriptomics was chosen to capture the spatial heterogeneity of gene expression within CPs, providing insights that bulk transcriptomics could not reveal, and enhancing our understanding of tumor biology, pathology and therapeutic potential [4, 23, 26]. Our Xenium-based spatial transcriptomics analysis was limited to 377 genes included in the Human Multi-tissue and Cancer Panel. Even with this limitation, we were able to show different expression profiles between ACP and PCP. Clustering based on gene expression levels revealed 13 distinct cell groups, some of which we were able successfully to annotate using a publicly available reference.

Xenium analysis revealed 41 differentially expressed genes between the ACP and PCP groups. Notably, genes associated with tumor growth, such as PPP1R1B and PROX1, were upregulated in the ACP group, indicating a possible enhancement of tumor proliferation and survival mechanisms. Conversely, downregulation of MS4A1 and EHF may suggest an immune escape strategy employed by ACP tumors. Furthermore, the immune evasion observed in ACP, marked by downregulated genes such as CD274, aligns with recent findings in a study [45], which reported fewer T cells and an increased infiltration of immune-suppressing macrophages in ACP. This immune landscape further supports the potential for targeted immunotherapies in ACP. In addition, another study [7] demonstrated that AXL signaling contributes to immune escape. AXL is a receptor tyrosine kinase that promotes immune suppression, cell survival, and tumor progression. Inhibition of AXL was shown to enhance immune responsiveness, highlighting its potential as a therapeutic target. These findings highlight how both immune cells and tumor signals help ACP evade the immune system, pointing to new treatment possibilities. Furthermore, the metabolic alterations observed in ACP, highlighted by the upregulation of HPX and CYP2B6, point to the adoption of unique metabolic pathways by the tumor. The small sample size for PCP (n = 4 transcriptomic analysis, n = 1 spatial transcriptomics) reduces statistical power, especially for immune and microenvironment analyses. As shown in studies [4, 23], robust characterization of immune cell diversity and spatial architecture in rare tumors requires larger cohorts for reliable conclusions. Therefore, our conclusions regarding immune signatures, tumor biologic behavior, and microenvironment features in ACP and PCP should be viewed as preliminary and hypothesis-generating, pending further validation in larger, more comprehensive studies.

Interestingly, APCDD1 was upregulated in ACP when compared to PCP. APCDD1 is a membrane protein that acts as a negative regulator of Wnt signaling by lowering the concentration of available Wnt ligands [20]. APCDD1 is known to be abnormally expressed in colon cancer and Wilms’ tumors [46]. However, the role of APCDD1 in the biology of CPs has not been yet described. Despite its inhibitory function, APCDD1 is a target of Wnt signaling and thus gets upregulated when Wnt is hyperactive. The upregulation represents an attempt to balance the excessive Wnt activity caused by CTNNB1 mutations. Recent study [20] shows that APCDD1 limits Wnt signaling by directly binding and neutralizing lipid-modified Wnt ligands at the cell surface. Its upregulation in ACP likely reflects a negative feedback response to persistent Wnt pathway activation, but this mechanism may be insufficient to fully counteract oncogenic signaling. This complex interplay highlights the regulatory challenges in Wnt-driven tumors such as ACP, although it is also possible that these differences are not only because of the Wnt/β-catenin pathway, but are a consequence of changed regulation of different branches of Wnt signaling.

Our application of spatial transcriptomics to CPs provides a valuable layer of insight that complements high-resolution single-cell data and functional studies like [7, 39, 45]. Integrating spatial and molecular profiling with clinical outcomes, as presented in these recent works, will be critical for translating biomarker findings into improved diagnostics and personalized therapies, particularly those targeting pathways such as Wnt, AXL, and the tumor immune microenvironment. Future studies should link spatial cellular context and molecular profiles with treatment response and prognosis in larger well-annotated cohorts.

We would also like to mention obstacles we encountered in this study so that other researchers can take them into account when planning their research. FFPE samples are an important, and often the only source of RNA, especially in rare tumors, and must be considered for transcriptional analysis. The successful extraction and sequencing of RNA from FFPE samples has shown that it is possible to use archival tissue for molecular studies, which is crucial for the investigation of rare diseases with limited sample availability. DNA contamination was detected in six samples, even though DNase treatment was performed during the isolation step. For this reason, we recommend that, if possible, all samples for cDNA library preparation are additionally treated with DNase prior to any other steps.

Further validation by IHC studies and additional functional analyses of the mentioned genes could provide deeper insights into their role in CPs. Exploration of therapeutic interventions targeting the Wnt, SHH or other signaling pathways may offer new treatment options, especially in cases in which surgical resection is challenging. Despite the significant findings, our study is limited by the small sample size, due to the rarity of CPs. Larger studies are needed to validate these results and further explore the molecular mechanisms underlying these tumors.

The consistent presence of BRAF p.V600E mutations in PCP and CTNNB1 exon 3 mutations in ACP emphasizes the unique pathogenesis of each subtype. The identification of specific mutations and distinct gene expression profiles provides a foundation for developing targeted therapies and improving diagnostic accuracy. Despite the challenges of working with FFPE samples and the limited sample size, our research lays the groundwork for future studies to improve diagnosis and treatment strategies for these rare but impactful tumors. In addition, our study is one of the first to look at spatial transcriptomic of these tumors, providing deeper insight into the molecular background. This study underscores the importance of comprehensive molecular profiling in rare tumors and highlights the potential for personalized medicine in the management of CPs.