2026 Proffered Presentations
S039: LINKING 7T MRI TO CORTICAL GENE EXPRESSION: A NOVEL PATH TOWARD IDENTIFYING MOLECULAR MECHANISMS UNDERLYING PITUITARY ADENOMA
Matt Jogodnik, BS; Alex Devarajan, BS; Mackenzie T Herb, PhD; Alan C Seifert, PhD; Bradley N Delman, MD, MS; Priti Balchandani, PhD; Joshua Bederson, MD; Raj Shrivastava, MD; Icahn School of Medicine at Mount Sinai
Introduction: While transcriptomics have proven indispensable in identifying tumor-intrinsic features of pituitary adenomas, much less is known about the mechanisms underlying their effects on whole-brain structure and function. Patients experience neurological disturbances that suggest involvement beyond the sella, yet the molecular mechanisms driving these effects remain poorly characterized due to the inability to sample healthy tissue. By integrating high field, 7 tesla, magnetic resonance imaging (7T MRI) with spatially resolved gene expression data, imaging transcriptomics enables the linking of observed structural changes to underlying molecular pathways. Imaging transcriptomics has not yet been applied to skull base tumors, and, to our knowledge, no prior study has leveraged 7T MRI within this framework. Herein, we sought to identify molecular signatures characterizing cortical regions disproportionately affected in patients with pituitary adenoma.
Methods: 18 patients with pituitary adenomas and 27 controls without prior neurological or neuropsychiatric history underwent brain MRI including T1-weighted MP2RAGE imaging. T1 image data were subsequently processed with FreeSurfer 7 which included skull stripping, motion correction, and normalization. Cortical thickness was extracted bilaterally across Desikan-Killiany parcels. Regional microarray expression data from 6 postmortem brains in the Allen Human Brain Atlas were normalized with abagen and mapped to the same parcellation. Cohen’s d effect sizes were calculated to generate a cortical thickness difference map between patients and controls after regressing out the effects of age, sex, and total intracranial volume. Partial least squares (PLS) regression was then used to identify principal components linking transcriptomic variation to case-control differences. Spatial autocorrelation was addressed using spin-test null models (10,000 rotations), with multiple comparisons corrected by false discovery rate (FDR). Gene set enrichment analysis was performed with the Hallmark database. Cortical surface maps were visualized in FreeSurfer 7, and statistical plots were generated in R and Python.
Results: Cortical thickness differences between patients and controls were highest in the right superior temporal cortex (Cohen’s d = 2.46) and right caudal anterior cingulate (d = 1.81) (Figure 1). The first principal component of PLS regression (PLS1) significantly associated spatial patterns of transcriptome variation with these case-control differences (ρ = -0.56, spin p < 0.002) (Figures 2 & 3). Regions with high alignment to this axis were enriched for genes belonging to immune and metabolic pathways (Figure 4).
Conclusion: Pituitary adenomas were associated with distributed differences in cortical thickness, particularly within temporal and cingulate regions, that aligned with specific transcriptomic signatures. These findings suggest that structural alterations in patients with pituitary adenoma preferentially occur in cortical regions characterized by distinct molecular patterns. This work represents the first application of imaging transcriptomics to skull base tumors and the first integration of 7T MRI within this framework, establishing a novel paradigm with potential to identify molecular mechanisms that underlie neurological disease.
