Maria Teresa Landi's research while affiliated with National Institutes of Health and other places

Lung cancer in never smokers (LCINS) accounts for up to 25% of all lung cancers and has been associated with exposure to secondhand tobacco smoke and air pollution in observational studies. Here, we evaluate the mutagenic exposures in LCINS by examining deep whole-genome sequencing data from a large international cohort of 871 treatment-naïve LCINS recruited from 28 geographical locations within the Sherlock- Lung study. KRAS mutations were 3.8-fold more common in adenocarcinomas of never smokers from North America and Europe, while a 1.6-fold higher prevalence of EGFR and TP53 mutations was observed in adenocarcinomas from East Asia. Signature SBS40a, with unknown cause, was found in most samples and accounted for the largest proportion of single base substitutions in adenocarcinomas, being enriched in EGFR -mutated cases. Conversely, the aristolochic acid signature SBS22a was almost exclusively observed in patients from Taipei. Even though LCINS exposed to secondhand smoke had an 8.3% higher mutational burden and 5.4% shorter telomeres, passive smoking was not associated with driver mutations in cancer driver genes or the activities of individual mutational signatures. In contrast, patients from regions with high levels of air pollution were more likely to have TP53 mutations while exhibiting shorter telomeres and an increase in most types of somatic mutations, including a 3.9-fold elevation of signature SBS4 (q-value=3.1 × 10 ⁻⁵ ), previously linked mainly to tobacco smoking, and a 76% increase of clock-like signature SBS5 (q-value=5.0 × 10 ⁻⁵ ). A positive dose-response effect was observed with air pollution levels, which correlated with both a decrease in telomere length and an elevation in somatic mutations, notably attributed to signatures SBS4 and SBS5. Our results elucidate the diversity of mutational processes shaping the genomic landscape of lung cancer in never smokers.

Purpose: This study employs cutting-edge deep learning techniques to comprehensively analyze hematoxylin and eosin-stained (H&E) whole slide images (WSIs) to predict which tumors carry driver gene alterations, mutational signatures, and additional molecular features among lung cancers from never-smokers (LCINS)— the seventh leading cause of cancer death worldwide. Method: A total of 464 H&E-stained WSIs of lung adenocarcinomas from the Sherlock-Lung study were utilized, with 325 WSIs for model training and 139 for testing. Alongside the WSIs, genomic data, including mutations (overall nonsynonymous mutations and driver mutations only), fusions, or copy numbers in driver genes (TP53, KRAS, EGFR, CDKN2A, MDM2, ALK, RBM10), mutational signatures (APOBEC), molecular features (Kataegies, Whole Genome Doubling (WGD) status, Tumor Mutational Burden (TMB)), and specific hotspot driver mutation in EGFR (p.L858R and p.E746_A750del) and KRAS (p.G12C, p.G12V, and p.G12D), were included for all tumors. Each WSI was assigned different labels based on the presence or absence of these genomic alterations or features. We used a customized multilabel binary deep learning model based on ResNet50, a residual convolutional neural network, for analyzing the data. The network was trained from scratch, with an initial epoch set to 100. Results: Our methodology demonstrated high predictive performance, measured by the area under the receiver operating characteristic (AUROC). In the presence or absence category, we achieved high AUROC for detecting any alterations in these driver genes: TP53 (0.81), KRAS (0.92), EGFR (0.93), CDKN2A deletion (0.88), MDM2 amplification (0.94), and ALK fusion (0.86). We also evaluated molecular features, obtaining AUROC scores of 0.79 for WGD status and 0.86 for Kataegies. Moderate AUROC scores were observed for tumors with RBM10 mutations (0.67), high TMB (0.64), and APOBEC signatures (0.56). When focusing on driver mutations only, we achieved high AUROC scores for EGFR (0.88), KRAS (0.84), TP53 (0.93), and RBM10 (0.78). In addition, our model successfully predicted tumors with specific mutations in EGFR (p.L858R = 0.79; p.E746_A750del = 0.77) and KRAS (p.G12C = 0.77). Performance was suboptimal for KRAS p.G12V (0.41) and KRAS p.G12D (0.47). Conclusions: Our deep learning network achieves high prediction scores in identifying tumors with critical driver gene alterations and actionable mutations, holding promise for potential clinical use. In the future, the model could be optimized as a screening assay to guide molecular testing and therapeutic management of patients with LCINS. Citation Format: Monjoy Saha, Tongwu Zhang, Praphulla Bhawsar, Wei Zhao, Jianxin Shi, Soo Ryum Yang, Jonas Almeida, Maria Teresa Landi. Deep learning-based molecular characterization of lung cancers from never smokers using hematoxylin and eosin-stained whole slide images [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB243.

Lung cancer is the leading cause of cancer-related mortality worldwide. Understanding lung cancer evolutionary dynamics can help identify tools to intercept its growth and suggest strategies for treatment. Multiple factors can impact the tumors’ natural history and distinctly affect growth rate. However, research on the evolutionary trajectories of lung cancer across demographic or exposure scenarios remains inadequate. Additionally, the roles of mutational processes and complex genomic alterations on the evolution of lung cancer are still largely unexplored. In the largest genomic study of lung cancer to date, we analyzed deep whole-genome sequencing (~ 81x) and other omics data of 1217 lung cancers from the Sherlock-Lung study. To ensure adequate statistical power for identifying subclone architectures and constructing lung cancer evolutionary histories, we utilized a metric known as NRPCC (number of reads per tumor chromosomal copy) to select 542 lung adenocarcinoma (LUAD) samples for clonal evolution analyses, including 186 and 181 samples from never-smoker subjects of European and Asian ancestry, respectively, and 121 samples from smokers of European ancestry. We found that major driver genes and exogenous mutations contribute to tumor initiation, while copy number gains and endogenous processes appear later in tumor evolution. Tumors harboring EGFR mutations in never-smoker females of European descent show long latency, while tumors with KRAS mutations have shorter latency regardless of ancestry and sex. Notably, tumors harboring the mutational signature ID2 have short latency and aggressive phenotype, accompanied by increased genomic instability, elevated hypoxia scores, high CpG methylator phenotype, low neoantigen burden, and propensity to develop metastasis. We show that LINE-1 retrotransposition-induced mutagenesis contributes to the origin of ID2 mutations. The transcriptional factor ZNF695, a member of the KZFP family, up-regulated in LUAD, appears to contribute to LINE-1 retrotransposition through a dominant-negative effect and LINE-1 promoter demethylation. In a multivariate analysis of genomics, exposures and demographic factors, LUAD latency was most significantly associated with ID2, followed by EGFR mutations, KRAS mutations, and sex, suggesting an independent impact of these factors on LUAD evolution. Our findings underscore the complex interplay of ancestry, sex, exogenous mutagenesis, epigenetic regulation, and LINE-1 retrotransposition in shaping LUAD evolutionary trajectories, paving the way for potential targeted therapeutic interventions. Citation Format: Tongwu Zhang, Wei Zhao, Christopher Wirth, Marcos Díaz-Gay, Jinhu Yin, Phuc H. Hoang, Jian Sang, John McElderry, Alyssa Klein, Azhar Khandekar, Caleb Hartman, Jennifer Rosenbaum, Frank Colon-Matos, Kristine M. Jones, Neil E. Caporaso, Robert Homer, Angela C. Pesatori, Dario Consonni, Lixing Yang, Bin Zhu, Jianxin Shi, Kevin Brown, Nathaniel Rothman, Stephen J. Chanock, Ludmil B. Alexandrov, Jiyeon Choi, Maurizio Cardelli, Qing Lan, Martin A. Nowak, David C. Wedge, Maria Teresa Landi. Deciphering lung adenocarcinoma evolution and the role of LINE-1 retrotransposition [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB226.

While most lung cancers are associated with tobacco smoking, prior studies have indicated that 15-20% of cases are found in never smokers. Lung cancer in never smokers (LCINS) shows demographic variations, being more common in females and Asian populations. Epidemiological studies have also identified environmental risk factors for LCINS, including exposure to secondhand tobacco smoke and air pollution. As part of the Sherlock-Lung project (current data freeze of 1217 samples), we generated deep WGS data from a large international cohort of 871 treatment-naïve LCINS recruited from 28 locations across four continents. Our cohort allowed mapping the genomic landscape operative in lung cancer histologies commonly found in never smokers, including adenocarcinomas (n=737) and carcinoids (n=61), as well as other histologies rarely attributed to never smokers (n=73). Information on passive smoking was collected for 458 patients, with 250 being exposed to secondhand smoke. We identified driver alterations and performed mutational signature analyses for single base substitutions (SBS), indels, doublet base substitutions (DBS), copy number alterations, and structural variants. We detected eighteen SBS signatures, with eight not previously linked to LCINS, including SBS4, associated with smoking but also observed in lifelong non-smokers exposed to different environmental mutagens, such as indoor pollution. Several additional signatures were detected in other variant types, including a novel indel signature. Notably, secondhand smoke exposure corresponded to increased SBS burden, although it was not associated with specific signatures. In contrast, we observed a signature-specific association of air pollution (measured as PM2.5 exposure) with SBS, indel, and DBS burden. SBS4 and correlated signatures ID3 and DBS2 were positively associated with PM2.5 exposure. Similarly, clock-like signature SBS5 was highly correlated with PM2.5, potentially acting as a readout of a promotion mechanism where lung cells undergo more cell divisions in individuals living in highly polluted areas. We also observed variability across histologies and ancestries for signatures and driver mutations, including enrichment of signatures SBS17b and SBS40a as well as KRAS mutations in adenocarcinomas from Europeans; aristolochic-acid associated signature SBS22a, and EGFR and TP53 mutations in adenocarcinomas from East Asians; and signature SBS8 and ARID1A mutations in carcinoids from Europeans. The unprecedented size of our cohort allowed us to refine the prevalence and intensity of the driver mutations and mutational signatures involved in LCINS, as well as their variability across ancestries, geographical locations, and histologies. Our results indicate that air pollution may act as both initiator and promoter of neoplastic expansion of LCINS. Citation Format: Marcos Diaz-Gay, Tongwu Zhang, Azhar Khandekar, Burçak Otlu, Shuvro Nandi, Christopher D. Steele, Nathaniel Rothman, Stephen J. Chanock, Qing Lan, Ludmil B. Alexandrov, Maria Teresa Landi. The mutagenic forces shaping the genomic landscape of lung cancer in never smokers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB231.

Purpose of the study Elucidating evolutionary trajectories of cancers allows us to understand the key events, and the order in which they occur, throughout their development. This can help us to find important associations with tumor progression and prognosis. Our aim was to perform de novo identification of the evolutionary trajectories within Sherlock-lung, with a dataset containing the largest collection of lung cancer in never smokers (LCINS) samples ever analyzed. Experimental procedures Our Plackett-Luce ordering model utilized copy number data from Battenberg and mutation cancer cell fraction (CCF) data from DPClust. Frequently-occurring copy number events and driver mutations are ordered within each sample using their copy number states and CCFs. An aggregate ordering is then calculated for a sample set. Mixture model analysis identifies subsets of samples displaying distinct orders of events, uncovering diverse evolutionary trajectories within a tumor set. Dataset The Sherlock-lung whole genome sequencing dataset (n=1217) was filtered to the samples that allowed us to identify subclonal expansions. Samples required at least 10 reads per chromosome copy and a minimum cellularity of 30%. This provided 458 LCINS samples of various histologies. 155 smoker samples were also analyzed for comparison. Results We identified two subsets of LCINS tumors following distinct evolutionary trajectories. The “loss-based” subset commonly saw whole genome duplication (WGD) combined with copy number losses occurring earlier, and at higher prevalence, than gains. Contrastingly, in the “gain-based” subset, WGD was relatively rare but ploidy increased via copy number gains, which were more prevalent than losses. Interestingly, these different trajectories converged on similar overall copy number states. The loss-based subset had a higher mutational burden and a higher proportion of the genome altered, and followed a more smoker-like trajectory than the gain-based subset. Considering these differences alongside the convergence in copy number states, it is intriguing that survival times were similar between the two subsets. Copy number events defined the difference between the two trajectories. However, driver mutations also played important roles in tumor evolution in LCINS. TP53 and EGFR mutations were associated with greater genomic instability. Conversely, KRAS mutations were associated with more stable genomes. Samples with early clonal mutations in TP53, ERBB2, and PIK3CA, as well as those with a copy number gain of ERBB2, exhibited shorter survival times. Conclusions Two distinct evolutionary trajectories of LCINS were identified by de novo Plackett-Luce event ordering analysis. The contrast between the subgroups was defined by different paths of copy number activity, but they ultimately converged on similar overall copy number states and outcomes. Key early driver mutations influenced genomic instability and survival times. Citation Format: Christopher Wirth, Tongwu Zhang, Wei Zhao, Phuc Hoang, Jian Sang, Nathaniel Rothman, Marcos Díaz-Gay, Ruxandra Teslo, Naser Ansari-Pour, Máire Ní Leathlobhair, Iliana Peneva, William Eagles, Lixing Yang, Ludmil Alexandrov, David C. Wedge, Maria Teresa Landi. Evolutionary trajectories of lung cancer in never smokers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB228.

APOBEC enzymes are part of the innate immunity and are responsible for restricting viruses and retroelements by deaminating cytosine residues. Most solid tumors harbor different levels of somatic mutations attributed to the off-target activities of APOBEC3A (A3A) and/or APOBEC3B (A3B). However, how APOBEC3A/B enzymes shape the tumor evolution in the presence of exogenous mutagenic processes is largely unknown. Here, by combining deep whole-genome sequencing with multi-omics profiling of 309 lung cancers from smokers with detailed tobacco smoking information, we identify two subtypes defined by low (LAS) and high (HAS) APOBEC mutagenesis. LAS are enriched for A3B-like mutagenesis and KRAS mutations, whereas HAS for A3A-like mutagenesis and TP53 mutations. Unlike APOBEC3A, APOBEC3B expression is strongly associated with an upregulation of the base excision repair pathway. Hypermutation by unrepaired A3A and tobacco smoking mutagenesis combined with TP53-induced genomic instability can trigger senescence, apoptosis, and cell regeneration, as indicated by high expression of pulmonary healing signaling pathway, stemness markers and distal cell-of-origin in HAS. The expected association of tobacco smoking variables (e.g., time to first cigarette) with genomic/epigenomic changes are not observed in HAS, a plausible consequence of frequent cell senescence or apoptosis. HAS have more neoantigens, slower clonal expansion, and older age at onset compared to LAS, particularly in heavy smokers, consistent with high proportions of newly generated, unmutated cells and frequent immuno-editing. These findings show how heterogeneity in mutational burden across co-occurring mutational processes and cell types contributes to tumor development, with important clinical implications.

Adenosine-to-Inosine (A-to-I) editing represents a crucial RNA modification mechanism, holding the potential to drive carcinogenesis. This sophisticated post-transcriptional process is catalyzed by a series of ADAR (adenosine deaminase acting on RNA) family members, which can profoundly reshape the cellular epitranscriptomic landscape. While previous explorations have illuminated the important role of A-to-I editing in lung cancer tumorigenesis and prognosis, most studies have focused on lung cancer from smokers. To investigate the contribution on RNA-editing in lung cancers in never smokers (LCINS) , we embarked on a rigorous dissection of A-to-I RNA editing events in 603 LCINS. Harnessing transcriptome and whole-genome sequencing technologies, we identified 15,880 A-to-I sites, of which 75% were positively linked to the enzymatic activity of ADAR1. 92.80% of detected editing sites were located at Alu repetitive regions, with the remainder in non_Alu repeat (4.1%) and non-repeat regions (3.1%). The global RNA editing levels, measured as Alu Editing Index (AEI), were found to be significantly associated with ADAR1 expression, Tumor Mutational Burden, and Neoantigens Burden. A subtle correlation between global RNA editing levels and tumor stage was also noted. Delving into cancer-associated editing events, we identified 5,539 differentially edited sites in tumors compared with normal tissues, which were located at 823 genes, including several key tumor suppressors and oncogenes, like MDM2, BRAF, TP53, BRCA2, ATM, and VHL. Subsequent pathway analysis identified TGF-beta, Notch Signaling, and DNA Damage Response as the key pathways for cancer progression. Using the characterized A-to-I editing profiles, we built a prognostic risk stratification model, with high risk scores significantly linked to poor survival outcomes, even after adjusting for other clinical factors. Our study provides a comprehensive A-to-I editing landscape of lung cancer in never smokers, highlighting the potential roles of RNA modifications in the pathogenesis of this lethal disease. Citation Format: Jian Sang, Wei Zhao, Tongwu Zhang, Maria Teresa Landi. The RNA editing landscape of lung cancer in never smokers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6250.

Lung cancer is the leading cause of cancer mortality worldwide and about 10-25% of lung cancers are from never-smokers (LCINS). Previous genomic studies of LCINS identified multiple genomic subtypes with different genetic drivers and evolutionary processes. To dissect the profile of LCINS cell states and tumor microenvironment, and their relationship with genomic lesions, we sequenced and assembled a large RNA-seq data set of 685 samples of lung adenocarcinoma from never-smokers and integrated them with whole genome sequencing data, pathological features and clinical outcomes from the same subjects. We identified three transcriptomic subtypes defined by distinct gene expression patterns. The three subtypes were associated with different pathway activities and showed remarkable heterogeneity in cell composition, lineage fidelity and pathological features. The genomic driver events and mutational signatures were significantly enriched in specific subtypes. Clinical outcomes also differed across the subtypes. For example, one subtype had prolonged overall survival and was associated with predicted response to immune checkpoint blockade. This study emphasizes the importance of transcriptome-based classification of LCINS, which has profound clinical implications beyond those provided by genomic and pathological assessment. Citation Format: Wei Zhao, Tongwu Zhang, Phuc H. Hoang, Jian Sang, Maria Teresa Landi. Transcriptomic profiling of lung adenocarcinoma from never-smokers reveals molecular subtypes with clinical implications [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 5672.

Background Although polygenic risk score (PRS) has emerged as a promising tool for predicting cancer risk from genome-wide association studies (GWAS), the individual-level accuracy of lung cancer PRS and the extent to which its impact on subsequent clinical applications remains largely unexplored. Methods Lung cancer PRSs and confidence/credible interval (CI) were constructed using two statistical approaches for each individual: (1) the weighted sum of 16 GWAS-derived significant SNP loci and the CI through the bootstrapping method (PRS-16-CV) and (2) LDpred2 and the CI through posteriors sampling (PRS-Bayes), among 17,166 lung cancer cases and 12,894 controls with European ancestry from the International Lung Cancer Consortium. Individuals were classified into different genetic risk subgroups based on the relationship between their own PRS mean/PRS CI and the population level threshold. Results Considerable variances in PRS point estimates at the individual level were observed for both methods, with an average standard deviation (s.d.) of 0.12 for PRS-16-CV and a much larger s.d. of 0.88 for PRS-Bayes. Using PRS-16-CV, only 25.0% of individuals with PRS point estimates in the lowest decile of PRS and 16.8% in the highest decile have their entire 95% CI fully contained in the lowest and highest decile, respectively, while PRS-Bayes was unable to find any eligible individuals. Only 19% of the individuals were concordantly identified as having high genetic risk (> 90th percentile) using the two PRS estimators. An increased relative risk of lung cancer comparing the highest PRS percentile to the lowest was observed when taking the CI into account (OR = 2.73, 95% CI: 2.12–3.50, P-value = 4.13 × 10⁻¹⁵) compared to using PRS-16-CV mean (OR = 2.23, 95% CI: 1.99–2.49, P-value = 5.70 × 10⁻⁴⁶). Improved risk prediction performance with higher AUC was consistently observed in individuals identified by PRS-16-CV CI, and the best performance was achieved by incorporating age, gender, and detailed smoking pack-years (AUC: 0.73, 95% CI = 0.72–0.74). Conclusions Lung cancer PRS estimates using different methods have modest correlations at the individual level, highlighting the importance of considering individual-level uncertainty when evaluating the practical utility of PRS.

Objectives: The quantitative job-exposure matrix SYN-JEM consists of various dimensions: job-specific estimates, region-specific estimates, and prior expert ratings of jobs by the semi-quantitative DOM-JEM. We analyzed the effect of different JEM dimensions on the exposure-response relationships between occupational silica exposure and lung cancer risk to investigate how these variations influence estimates of exposure by a quantitative JEM and associated health endpoints. Methods: Using SYN-JEM, and alternative SYN-JEM specifications with varying dimensions included, cumulative silica exposure estimates were assigned to 16 901 lung cancer cases and 20 965 controls pooled from 14 international community-based case-control studies. Exposure-response relationships based on SYN-JEM and alternative SYN-JEM specifications were analyzed using regression analyses (by quartiles and log-transformed continuous silica exposure) and generalized additive models (GAM), adjusted for age, sex, study, cigarette pack-years, time since quitting smoking, and ever employment in occupations with established lung cancer risk. Results: SYN-JEM and alternative specifications generated overall elevated and similar lung cancer odds ratios ranging from 1.13 (1st quartile) to 1.50 (4th quartile). In the categorical and log-linear analyses SYN-JEM with all dimensions included yielded the best model fit, and exclusion of job-specific estimates from SYN-JEM yielded the poorest model fit. Additionally, GAM showed the poorest model fit when excluding job-specific estimates. Conclusion: The established exposure-response relationship between occupational silica exposure and lung cancer was marginally influenced by varying the dimensions of SYN-JEM. Optimized modelling of exposure-response relationships will be obtained when incorporating all relevant dimensions, namely prior rating, job, time, and region. Quantitative job-specific estimates appeared to be the most prominent dimension for this general population JEM.