Cigarette smoke is the major cause of lung cancer, the leading cause of cancer death, and of chronic obstructive pulmonary disease, the fourth leading cause of death in the United States. Using high-density gene expression arrays, we describe genes that are normally expressed in a subset of human airway epithelial cells obtained at bronchoscopy (the airway transcriptome), define how cigarette smoking alters the transcriptome, and detail the effects of variables, such as cumulative exposure, age, sex, and race, on cigarette smoke-induced changes in gene expression. We also determine which changes in gene expression are and are not reversible when smoking is discontinued. The persistent altered expression of a subset of genes in former smokers may explain the risk these individuals have for developing lung cancer long after they have discontinued smoking. The use of gene expression profiling to explore the normal biology of a specific subset of cells within a complex organ across a broad spectrum of healthy individuals and to define the reversible and irreversible genetic effects of cigarette smoke on human airway epithelial cells has not been previously reported.
Deep learning is a powerful tool for whole slide image (WSI) analysis. Typically, when performing supervised deep learning, a WSI is divided into small patches, trained and the outcomes are aggregated to estimate disease grade. However, patch-based methods introduce label noise during training by assuming that each patch is independent with the same label as the WSI and neglect overall WSI-level information that is significant in disease grading. Here we present a Graph-Transformer (GT) that fuses a graph-based representation of an WSI and a vision transformer for processing pathology images, called GTP, to predict disease grade. We selected 4,818 WSIs from the Clinical Proteomic Tumor Analysis Consortium (CPTAC), the National Lung Screening Trial (NLST), and The Cancer Genome Atlas (TCGA), and used GTP to distinguish adenocarcinoma (LUAD) and squamous cell carcinoma (LSCC) from adjacent non-cancerous tissue (normal). First, using NLST data, we developed a contrastive learning framework to generate a feature extractor. This allowed us to compute feature vectors of individual WSI patches, which were used to represent the nodes of the graph followed by construction of the GTP framework. Our model trained on the CPTAC data achieved consistently high performance on three-label classification (normal versus LUAD versus LSCC: mean accuracy = 91.2 ± 2.5%) based on five-fold cross-validation, and mean accuracy = 82.3 ± 1.0% on external test data (TCGA). We also introduced a graph-based saliency mapping technique, called GraphCAM, that can identify regions that are highly associated with the class label. Our findings demonstrate GTP as an interpretable and effective deep learning framework for WSI-level classification.
Background: Tobacco use remains the leading preventable cause of death in the US. The risk of dying from smoking-related diseases remains elevated for former smokers years after quitting. The identification of irreversible effects of tobacco smoke on airway gene expression may provide insights into the causes of this elevated risk. Results: Using oligonucleotide microarrays, we measured gene expression in large airway epithelial cells obtained via bronchoscopy from never, current, and former smokers (n = 104). Linear models identified 175 genes differentially expressed between current and never smokers, and classified these as irreversible (n = 28), slowly reversible (n = 6), or rapidly reversible (n = 139) based on their expression in former smokers. A greater percentage of irreversible and slowly reversible genes were down-regulated by smoking, suggesting possible mechanisms for persistent changes, such as allelic loss at 16q13. Similarities with airway epithelium gene expression changes caused by other environmental exposures suggest that common mechanisms are involved in the response to tobacco smoke. Finally, using irreversible genes, we built a biomarker of ever exposure to tobacco smoke capable of classifying an independent set of former and current smokers with 81% and 100% accuracy, respectively. Conclusion: We have categorized smoking-related changes in airway gene expression by their degree of reversibility upon smoking cessation. Our findings provide insights into the mechanisms leading to reversible and persistent effects of tobacco smoke that may explain former smokers increased risk for developing tobacco-induced lung disease and provide novel targets for chemoprophylaxis. Airway gene expression may also serve as a sensitive biomarker to identify individuals with past exposure to tobacco smoke.
West Nile virus (WNV) exists in nature as a genetically diverse population of competing genomes. This high genetic diversity and concomitant adaptive plasticity has facilitated the rapid adaptation of WNV to North American transmission cycles and contributed to its explosive spread throughout the New World. WNV is maintained in nature in a transmission cycle between mosquitoes and birds, with intrahost genetic diversity highest in mosquitoes. The mechanistic basis for this increase in genetic diversity in mosquitoes is poorly understood. To determine whether the high mutational diversity of WNV in mosquitoes is driven by RNA interference (RNAi), we characterized the RNAi response to WNV in the midguts of orally exposed Culex pipiens quinquefasciatus using high-throughput, massively parallel sequencing and estimated viral genetic diversity. Our data demonstrate that WNV infection in orally exposed vector mosquitoes induces the RNAi pathway and that regions of the WNV genome that are more intensely targeted by RNAi are more likely to contain point mutations compared to weakly targeted regions. These results suggest that, under natural conditions, positive selection of WNV within mosquitoes is stronger in regions highly targeted by the host RNAi response. Further, they provide a mechanistic basis for the relative importance of mosquitoes in driving WNV diversification. West Nile virus (WNV) was introduced into New York state in 1999 and has since spread across the Americas. It is transmitted in nature between adult female mosquitoes and birds and occasionally infects humans and horses. Within the host, WNV exists as a diverse assortment of closely related mutants. WNV populations within mosquitoes are more complex genetically than are those within birds. The reasons for this discrepancy are unknown, but may be related to the host's innate antivirus response. We demonstrate that WNV is targeted by RNA interference, a highly sequence-specific pathway in the mosquito. Further, we present data that correlates the intensity of this targeting with virus mutation under natural conditions. These results provide a mechanistic explanation for the increasead complexity of WNV populations in mosquitoes: the RNAi response creates an intracellular environment where rare genotypes are favored. In addition, our results suggest that genetically diverse WNV populations may have an advantage over less diverse populations because they present a more complex target for the RNAi response. Finally, these data suggest that WNV, and possibly other viruses with high mutation rates, may escape an engineered antivirus intervention that is highly sequence-specific.