Jennifer E. Beane-Ebel PhD

Biography

1. Smoking Effects on Airway Epithelial Gene Expression: Given the substantial lung disease burden caused by cigarette smoking, surprisingly few human studies had examined how smoking alters the pulmonary airway epithelium, which is exposed to the highest concentrations of cigarette smoke. My research characterized smoking- and cessation-related changes in the airway transcriptome, providing a global molecular view of the airway response to tobacco exposure. By analyzing the behavior of these changes among former smokers, we classified gene expression changes as irreversible, slowly reversible, or rapidly reversible after cessation, yielding insights into the mechanisms underlying both reversible and persistent effects of tobacco smoke. More recent single cell RNA-seq of airway epithelium from current and never smokers have attributed many of these earlier bulk transcriptomic changes to smoking-associated shifts in epithelial cell composition and state. For example, compared with never smokers, current smokers showed expression of toxin metabolism genes within ciliated cells, evidence of tissue remodeling with loss of club cells and goblet cell hyperplasia, and emergence of a previously unrecognized epithelial cell population expressing genes that remain persistently altered by smoking (Duclos 2019). Collectively, this work demonstrates that smoke exposure drives a complex landscape of molecular and cellular alterations that may prime the human bronchial epithelium for disease.

2. Biomarker Development for the Early Detection of Lung Cancer: Lung cancer is the leading cause of cancer-related death worldwide, and mortality remains high in part because effective tools for early-stage detection are limited. In smokers undergoing bronchoscopy for suspected lung cancer, we profiled histologically normal airway cells using microarrays to develop a gene expression-based diagnostic biomarker for lung cancer. I subsequently demonstrated that this biomarker adds predictive value beyond standard clinical risk factors and maintains strong performance in individuals at intermediate risk. The biomarker was patented and commercialized, and it has been used to help rule out lung cancer in high-risk patients undergoing bronchoscopy when the procedure was non-diagnostic. Building on these studies, our group and industry partners developed a less invasive approach using nasal brushings. As part of the NIH Early Detection Research Network, we are now developing a multimodal biomarker to discriminate benign from malignant indeterminate pulmonary nodules detected by CT. The biomarker will integrate imaging features from CT, gene expression from nasal brushings, and blood-based cell free DNA methylation to improve risk stratification and assign patients with intermediate-risk nodules to clinically actionable groups.

3. Determinants of Lung Premalignant Lesion Development and Progression: Lung squamous cell carcinoma (LUSC) premalignant lesions (PMLs) arise within the epithelial layer of the bronchial airways, whereas lung adenocarcinoma (LUAD) PMLs are localized proliferative lesions of atypical pneumocytes that expand along alveolar walls. We currently lack robust tools to identify—and intervene on—lung PMLs at highest risk of progression to invasive cancer. To address this gap, we are building a high-resolution, multidimensional genetic, molecular, and cellular atlas of precancerous lung lesions and their surrounding microenvironment. Using this atlas in LUSC, we identified four molecular subtypes of PMLs with distinct epithelial and immune programs. One subtype (the Proliferative subtype) was enriched for bronchial dysplasia and showed upregulation of metabolic and cell-cycle pathways. Importantly, progressive/persistent Proliferative PMLs exhibited reduced interferon signaling and antigen processing/presentation, accompanied by depletion of both innate and adaptive immune cells compared with regressive lesions. We further showed that decreased antigen processing and presentation in Proliferative lesions may be regulated by an epithelial miRNA and Hippo signaling, highlighting potential interception targets (Beane 2019, Moaz 2021, Ning 2023, Ning 2025). In LUAD, we identified PML “archetypes,” including one marked by higher proliferation, a pro-tumor immune milieu, and enrichment for EGFR driver mutations. LUAD tumors nearest this archetype showed aggressive features (high grade, lymph node invasion) and worse disease-free survival, suggesting a high-risk PML subset that could improve stratification and reveal targets for interception (Anderson 2025). Overall, molecular profiling of lesions and adjacent non-lesion airway regions is revealing early, testable events in lung carcinogenesis for interception trials.

4. Lung Cancer Chemoprevention: Lung cancer chemoprevention uses dietary, pharmacologic, and immunologic interventions to slow or reverse the progression of lung premalignant lesions (PMLs) to invasive cancer. Because cancer-incidence endpoints require very large trials, our work has emphasized intermediate endpoints—including bronchial dysplasia histopathology, proliferation biomarkers (e.g., Ki-67), eicosanoid metabolites, and airway “field-of-injury” transcriptomic signatures measured in paired bronchial and nasal samples. Across multiple clinical studies, we characterized phenotypic and molecular responses to candidate agents (e.g., aspirin, zileuton, myo-inositol, sulforaphane, and CIMAvax-EGF) and evaluated their ability to favorably modulate previously derived gene-expression signatures of smoking, lung cancer, COPD, and squamous dysplasia. Low-dose aspirin produced minimal changes in predefined carcinogenesis signatures in nasal epithelium, while inducing broad pathway-level transcriptional changes—supporting nasal brushings as a scalable surrogate tissue for monitoring chemopreventive response. Aspirin plus the 5-LOX inhibitor zileuton similarly had minimal effects on smoking-, lung cancer–, and COPD-related signatures, but favorably modulated a bronchial squamous dysplasia signature and suppressed urinary leukotrienes. In smokers with bronchial dysplasia, myo-inositol was safe but did not significantly improve dysplasia overall, while molecular profiling suggested pathway modulation and potential benefit in subsets. Complementing these efforts, a randomized phase II trial in former smokers showed that 12 months of oral sulforaphane significantly reduced bronchial Ki-67, and we are evaluating associated changes in the airway transcriptome. Finally, in an ongoing lung cancer prevention study of CIMAvax-EGF, we are assessing relationships between serum anti-EGF antibody responses and airway gene expression to define biomarkers of efficacy.

5. Integration of multimodal data for lung premalignant lesion and tumor stratification: Many lung studies still depend on pathologist review of resection specimens or small biopsies to stage disease, but the quantitative and spatial information embedded in H&E slides is rarely incorporated into predictive models. To address this gap, we are developing computational pathology approaches that extract reproducible, slide-level representations from digitized whole slide images (WSIs) and use them to stratify lung tumors and bronchial premalignant lesions (PMLs) across the continuum from normal epithelium to carcinoma in situ and invasive cancer. Using graph-based deep learning, we can model spatial relationships among tissue regions and generate interpretable, class-specific heatmaps that align with pathologist-annotated areas (Zheng 2022); notably, WSI-derived features can differentiate carcinoma in situ lesions by subsequent progression status, supporting potential use in risk stratification for surveillance and interception trials (Gindra 2024). Critically, we are extending these models beyond “image-only” prediction by integrating molecular measurements to better capture biology that is not fully apparent from morphology alone. An attention-based fusion framework that combines graph representations of pathology images with gene expression signatures improves outcome modeling in non–small cell lung cancer. In the premalignancy setting, a transformer based multimodal framework that jointly leverages WSIs and bulk gene expression improves classification of dysplasia (or worse) versus non-dysplasia compared with either modality alone, can be trained across multiple cohorts, and supports inference when only one modality is available (Xu 2026) —enabling mapping of PMLs onto a continuous spectrum of disease that may ultimately inform progression prediction and chemoprevention response.