Molecular Markers May Predict Pituitary Tumor Invasion

Pituitary neuroendocrine tumors, known as PitNETs, represent the third most common primary central nervous system tumor in adults, accounting for approximately 10% to 15% of all intracranial tumors. While the majority of these growths follow an indolent clinical course, a substantial subset exhibits local invasion into surrounding structures. This anatomical invasiveness can limit the ability of surgeons to achieve gross total resection, thereby increasing the risk of residual disease and recurrence. A critical distinction in current clinical practice is that anatomical invasiveness should not be equated with clinical aggressiveness. Aggressive PitNETs are defined by a longitudinal pattern of repeated or multiple local recurrences and frequently require multiple therapeutic interventions. The 2022 Fifth Edition of the World Health Organization Classification formally renamed pituitary adenomas to PitNETs, establishing a lineage classification framework centered on core transcription factors, including PIT1, TPIT, and SF1. This shift drives the transition from traditional pathological diagnosis to molecular subtyping, aiming to identify invasive subpopulations early in the pre- and postoperative period.

Recent advances in high-throughput sequencing, liquid biopsy, and single-cell omics have enabled the emergence of numerous novel molecular biomarkers. A comprehensive review published in Frontiers in Endocrinology systematically outlines recent progress in multidimensional biomarkers, including DNA methylation, protein markers, non-coding RNAs, liquid biopsy indicators, and tumor microenvironmental features. The study emphasizes that molecular taxonomy, which classifies tumors according to lineage and cellular identity, is distinct from molecular prognostication, which predicts clinically relevant outcomes such as local invasion or treatment resistance. Although these dimensions may overlap, they are not interchangeable. The review argues that integrating heterogeneous molecular findings within a framework that considers tumor lineage, biological phenotype, and clinical endpoint is essential for accurate risk stratification. This approach minimizes the conflation of molecular classification with mechanistic drivers and clinically oriented prognostic biomarkers.

Specific somatic alterations have emerged as candidate tissue markers that may contribute to risk stratification. TP53 mutations are frequently detected in invasive subpopulations, with mutation rates reaching up to 24% in invasive corticotroph tumors. These mutations typically herald larger tumor volume, enhanced invasiveness, and significantly shortened overall survival. Functional inactivation of the ATRX gene, a key regulator of chromatin remodeling, triggers the alternative lengthening of telomeres mechanism, granting tumor cells unlimited replicative potential. This phenomenon is predominantly observed in functional corticotroph tumors and frequently co-occurs with TP53 mutations. Pathogenic ATRX alterations may result in the loss of nuclear ATRX protein expression, which can be detected by immunohistochemistry. However, loss of ATRX immunoreactivity represents a protein-level surrogate of underlying dysfunction and should not be considered synonymous with an ATRX mutation. Consequently, routine ATRX immunohistochemical screening for PitNETs with invasive features is recommended to identify high-risk patients requiring intensive follow-up.

In lactotroph PitNETs, SF3B1 mutations have been identified as a genomic alteration linked to aberrant RNA splicing. The somatic hotspot mutation R625H occurs in approximately 3% of standard lactotroph PitNETs but rises to 42% in metastatic cases. SF3B1-mutated tumors display distinct aberrant splicing profiles, which correlate with multiple therapeutic interventions, heightened invasiveness, and shorter disease-free and overall survival. Together, TP53, ATRX, and SF3B1 alterations can be assessed in surgical tissue and may provide lineage-specific risk-enrichment information. However, their independent predictive performance for anatomical invasiveness or clinically defined aggressiveness remains unestablished. Among these alterations, TP53 dysfunction and ATRX loss provide a biologically coherent paired model for clonal evolution in corticotroph PitNETs. Function-altering TP53 variants may attenuate P53 transcriptional output, weakening checkpoint control and allowing DNA-damaged cells to survive. In parallel, ATRX deficiency may impair telomere maintenance, favoring the persistence of replication-fork stalling and telomeric DNA breaks in an ALT-permissive cellular context.

Epigenetic modifications also play a significant role in PitNET biology. Genome-wide DNA methylation profiling has demonstrated distinct clustering patterns among benign, aggressive, and metastatic PitNETs. Aggressive and metastatic tumors exhibit more extensive chromosomal copy-number alterations than benign tumors. These findings suggest that DNA methylation profiling may provide candidate molecular information for risk stratification. However, its ability to prospectively predict aggressive clinical behavior or guide individualized adjuvant treatment remains to be established. CHST7 promoter methylation represents a lineage-associated epigenetic alteration, with SF1-lineage tumors exhibiting a higher frequency of methylation than PIT1- and TPIT-lineage tumors. Lower CHST7 expression is associated with a higher proliferative index, suggesting a potential relationship between CHST7 dysregulation and tumor proliferative biology. Nevertheless, the available evidence does not establish CHST7 as a circulating biomarker or as a specific predictor of invasive behavior in non-functioning PitNETs.

Transcriptomic analyses have identified GALR1 and PENK as candidate tissue biomarkers associated with invasive behavior in non-functioning PitNETs. Both genes were significantly downregulated in invasive tumors, while functional experiments demonstrated that their reduced expression promoted pituitary tumor cell proliferation and invasion. These findings suggest that decreased GALR1 and PENK expression may characterize an invasive phenotype, although further validation in independent prospective cohorts is required. Recent evidence also indicates lineage-dependent transcriptional dysregulation of the mismatch repair pathway in invasive PitNETs. Significantly lower transcript levels of MSH2, MSH6, MLH1, and PMS2 were reported in invasive PIT1-lineage tumors, whereas PMS2 expression was significantly higher in invasive SF1-lineage tumors. These contrasting patterns suggest that the association between mismatch repair gene expression and invasiveness is lineage-dependent rather than uniform across PitNETs. Given the rarity of overt mismatch repair deficiency, these transcriptional changes should currently be regarded as exploratory molecular correlates of invasive behavior rather than validated prognostic biomarkers.

Post-transcriptional modifications, particularly alternative splicing, represent another layer of regulation associated with PitNET invasiveness. The recurrent SF3B1 R625H mutation can induce aberrant RNA splicing, promoting the downregulation of DLG1, which is accompanied by enhanced cell migration and invasion. Alternative splicing of NTRK1/TrkA has also been identified as a potentially relevant molecular mechanism. TrkAIII, an alternative TrkA receptor variant capable of ligand-independent intracellular activation, was detected in all invasive PitNETs in a study of 53 tumors. The relative TrkAIII to full-length TrkA ratio was significantly higher in invasive than in non-invasive PitNETs, a difference particularly evident within the PIT1 lineage. However, because TrkAIII mRNA was also detected in most non-invasive tumors, its widespread expression precludes its interpretation as a specific biomarker of clinical aggressiveness. It may represent a potentially targetable molecular alteration, although its diagnostic and therapeutic utility requires validation in larger independent cohorts.

Liquid biopsy provides a minimally invasive and potentially repeatable approach for characterizing tumor-derived molecular alterations. Plasma cell-free DNA methylation profiling has shown promise as a preliminary approach. A panel of three methylation markers, MIR4535, SLC8A1-AS1, and TTC34, collectively discriminated invasive from non-invasive PitNETs with an area under the receiver operating characteristic curve of 0.980. These findings provide proof-of-concept that tumor-associated epigenetic alterations can be detected in peripheral blood. However, the clinical utility and prognostic performance of this panel require confirmation in independent prospective cohorts. Extracellular vesicles, particularly exosomes, represent another potential source of circulating tumor-derived molecular information. Exosomal cargo includes non-coding RNAs and proteins that may reflect tumor phenotype. For example, exosomal circMFN2 has been reported to promote pituitary tumor cell proliferation and migration through regulation of the miR-146a-3p/TRAF6/NF-κB signaling axis. Nevertheless, circulating exosomal molecules remain exploratory biomarkers rather than established predictors of invasive PitNET behavior.

Radiomics has emerged as a complementary approach for extracting high-dimensional quantitative information from routinely acquired magnetic resonance imaging. In contrast to conventional radiological evaluation, radiomic analysis can quantify tumor shape, signal intensity, texture, and spatial heterogeneity. A retrospective study of 200 patients with macro-PitNETs demonstrated that an XGBoost model combining five radiomic features achieved an area under the receiver operating characteristic curve of 0.85 for distinguishing invasive from non-invasive tumors. More recently, a deep learning radiomics framework developed using data from 1,089 patients across three medical centers generated a radiomics score strongly associated with conventional invasion classifications. This score also outperformed conventional anatomical grading systems in predicting tumor recurrence and showed associations with molecular pathways including MAPK and TGF-β signaling. These findings suggest that radiomics may extend beyond anatomical invasion assessment and potentially capture imaging phenotypes associated with adverse biological behavior.

Current evidence indicates that molecular prognostication in PitNETs is moving beyond conventional histopathological assessment toward biologically informed risk stratification. No single molecular biomarker currently has sufficient validation to independently predict invasive PitNET behavior. The most promising near-term strategy is to integrate candidate markers with established clinical, radiological, and pathological variables, including longitudinal tumor growth, treatment response, and proliferative indices. Future studies should prioritize prospective multicenter validation, standardized analytical workflows, and clinically defined endpoints that distinguish invasion, recurrence, treatment resistance, and true aggressiveness. Such prospectively validated clinicoradiological, pathological, and molecular models may ultimately improve individualized risk stratification and help inform surveillance and management strategies for patients at risk of invasive PitNET progression.

Subject of Research: Molecular biomarkers for predicting invasiveness in pituitary neuroendocrine tumors

Article Title: From histopathology to molecular taxonomy: predicting pituitary neuroendocrine tumors invasiveness

Article References: Zhang, B., Sun, Y., Guo, H., Pei, X., Wu, C., Yang, Y., Yang, Y., & Wan, D. (2026). From histopathology to molecular taxonomy: predicting pituitary neuroendocrine tumors invasiveness. Frontiers in Endocrinology, 17, Article 1907825. https://doi.org/10.3389/fendo.2026.1907825

Image Credits: AI Generated

DOI: 10.3389/fendo.2026.1907825

Keywords: PitNETs, Molecular Biomarkers, DNA Methylation, Liquid Biopsy, Pituitary Tumors, Transcriptomics, Radiomics, WHO Classification, Tumor Invasion, Epigenetics, histopathology, molecular