Abstract
J. Chakrabarti: None. N.E. Garrett: None. K. Hurth: None. J.D. Carmichael: None. M. Shiroishi: None. D. Weisenberger: None. E.R. Laws: None. C. Horbinski: None. A.B. Castro: None. J. Eschbacher: None. A.S. Little: None. Y. Zavros: None. G. Zada: None. Background: Cushing's disease (CD), a serious complex disease characterized by excessive cortisol production caused by an adrenocorticotropic hormone (ACTH)-secreting Pituitary Neuroendocrine Tumor (PitNET). CD encompasses subtypes, exhibiting unpredictable invasiveness and recurrence. The molecular complexity of PitNETs requires in-depth multi-omic approaches yet to be clearly implemented. High-throughput technologies such as methylation profiling, RNA sequencing, and spatial transcriptomics delineate a unique opportunity to dissect the complex molecular architecture of these tumors. Objective: To leverage spatial multi-omic approaches to construct a prediction model related to the invasiveness and recurrence of CD in patients. Methods: Tumors were collected during transsphenoidal endoscopic surgery from CD patients. A genome-wide methylation analysis was performed using bisulfite sequencing. Simultaneously, RNA sequencing was conducted to profile differential gene expression. An integrated analysis pipeline, ChAMP was used for analyzing the methylation data. Unbiased hierarchical clustering was performed with the pheatmap R package using Pearson's correlation to show gene sets with analogous expression patterns and samples with similar gene expression profiles. NanoString CosMx™ Spatial Molecular Imaging (SMI) was performed using formalin-fixed paraffin-embedded (FFPE) PitNET tissues collected from both tumor and dural region. Results: Clustering of differential methylation patterns revealed notable hypermethylated CpGs in silent CDs, including FAM172A, LINC00261, and GOT1L1. Notably, CD patients reported to have invasive PitNETs showed higher hypermethylation in tumor suppressor gene MBP when compared to noninvasive tumors. Integrating differential methylation patterns with RNA sequencing data indicated enrichment in the PI3K-AKT and TGF-β signaling pathways. There was an overlapping subset (ANXA1, COL6A2, EFNA5, FGFR1, and SMAD3) of genes displaying aberrant methylation patterns that coincided with genes correlating with those identified by SMI. COL6A2, SMAD3, and EFNA5 displayed hypomethylation in the invasive and recurrence corticotroph PitNET subtypes, corresponding to upregulation in the invasive and silent PitNET genes spatially expressed in the SMI data. We found that COL6A2, SMAD3, and EFNA5 genes are of significance given their associated role as regulators of the epithelial-to-mesenchymal transition pathway during tumor progression. Conclusion: Integrating spatial multi-omics-based signatures with traditional histopathological evaluation represents a promising avenue for refining the classification of pituitary tumors and holds the potential to predict patient outcomes that will ultimately guide targeted treatments in the management of PitNETs associated with CD. Saturday, June 1, 2024