Supplementary Figure 3 show schematic of diversification of the Lenti-sgTSG19-sgV1/Cre and Lenti-sgTSG19-sgV3/Cre pools and tumor initiation in Cas9-negative mice, additional explanation of adaptive sampling method used in assessing the impact of Lenti-sgRNA/Cre vectors in V1- and V3-driven models and additional data on tumor suppressor function in V1- and V3-driven lung cancer including tumor burden, tumor number and effect.

Abstract Tumor mutational burden (TMB) shapes tumor transcriptional state, but studies typically describe this response as an average effect pooled across cancer types. Whether that average reflects a consistent response present within individual cancer types, or is an artifact of merging heterogeneous, tissue-specific responses, remains unresolved. Here we analyze ∼9,100 tumors across 32 TCGA cancer types to test whether the transcriptional response to TMB is genuinely consistent across tissues. We construct a TMB axis score from TMB-associated genes upregulated with increasing TMB, yielding a sample-level measure of response strength, and subsequently decompose it at the component and pathway/complex levels. The pooled transcriptional response to TMB stays largely consistent within each cancer type, and no single cancer is driving the pooled signal. This consistency was also observed at the component and pathway/complex levels. These findings support TMB as a promising tissue-agnostic signature, with implications for tissue-agnostic therapeutic targeting. Summary Increasing tumor mutational burden (TMB) reflects the accumulation of mutations that impose a broad physiological burden on cancer cells, including increased DNA damage responses, protein misfolding, metabolic stress, and immune signaling. Here, we quantify TMB-associated transcriptional responses across and within cancer types to test whether this response is consistent. We construct a TMB axis score capturing genes upregulated with increasing TMB, generating a sample-level measure of response strength. Most cancer types exhibit a consistent response magnitude as TMB increases. TMB variance, more than sample size, constrains the detectability of this response within individual cancer types. Because baseline expression of TMB-associated genes differs across cancer types, pan-cancer comparisons can mask this signal. In contrast, within-cancer-type normalization reveals a transcriptional response spanning multiple cellular pathways that is consistent across cancer types despite differences in baseline gene expression.

Supplementary Table 9 shows that number of sgNT1 tumors sampled in Cas9-EGFP V1 cohort for the Lenti-sgTSG19 and Lenti-sgTSG75 pools, Kras control V1 cohorts, Cas9-EGFP V1 cohort in the drug experiment and Kras control V1 cohort in the drug experiment.

Supplementary Figure 8 shows histological analysis and quantification of cancer type marker gene expression and signaling protein phosphorylation in EML4-ALK V1 and V3 tumors of different tumor suppressor genotypes

The increasing accessibility of long-read sequencing and the rapid development of automated variant callers are promoting the generation of population-level structural variation data. However, the effect of the length of long-reads on automated variant callers is not well understood, especially for non-human species. Here we show that only ultra-long long-reads, with read N50s greater than 50 kb, are capable of accurately calling structural variants of any size in Drosophila melanogaster euchromatin. We used Oxford Nanopore Technologies to long-read sequence eight, inbred D. melanogaster strains to extremely high coverage (mean 238×), and we then downsampled the reads to create read pools of different length distributions. We assembled genomes from these different read-length pools and used both read-based and assembly-based structural variant callers to call variants in each strain before merging the calls into population-level datasets. We manually validated over 2,300 putative structural variants to assess the precision of the variant calls across the different read-length distributions and to determine the cause and rates of false positive errors. We found that more than half of all structural-variant-calling errors stem from misaligned reads that contain mobile elements or are located in repetitive and complex regions. Overall, our results show that long reads should be at least three times longer than the largest transposable elements found in the genome in order to accurately call structural variants at the population level.

Supplementary Table 8 shows the Odds ratios (OR) and P-values from two-sided Fisher's exact tests comparing the occurrence of alterations in samples with EML4-ALK V1 relative to samples with EML4-ALK V3 for genes with at least 10 alterations across both cohorts

Diverse fusions of echinoderm microtubule-associated protein-like 4 (EML4) and anaplastic lymphoma kinase (ALK) are oncogenic drivers in lung adenocarcinoma. EML4-ALK variants have distinct breakpoints within EML4, but their functional differences remain poorly understood. In this study, we use somatic genome editing to generate autochthonous mouse models of EML4-ALK-driven lung tumors and show that variant 3 (V3) is more oncogenic than variant 1 (V1). By using multiplexed genome editing and quantifying the effects of 29 putative tumor-suppressor genes on V1- and V3-driven lung cancer growth, we show that many tumor-suppressor genes have variant-specific effects on tumorigenesis. Pharmacogenomic analyses further suggest that tumor genotype can influence therapeutic responses. Analysis of human EML4-ALK-positive lung cancers also identified variant-specific differences in their genomic landscapes. These findings suggest that EML4-ALK variants behave more like distinct oncogenes than a uniform entity and highlight the dramatic impact of oncogenic fusion partner proteins and coincident tumor-suppressor gene alterations on the biology of oncogenic fusion-driven cancers. SIGNIFICANCE:EML4-ALK-driven lung cancer is treated as a uniform disease despite the presence of distinct fusion variants in patients. Our findings show that EML4-ALK variants are functionally distinct, which may have implications for the treatment of this cancer type and highlights the need to consider differences among variants of other oncogenic fusions.

Expression of Ink4a and a p53 tumor suppression-associated gene (TSAG) signature

Trade-offs are an inherent feature of organismal biology and fundamental to the evolution of natural populations. Here, we use experimental evolution in large, genetically diverse populations of Drosophila melanogaster to directly measure the manifestation of trade-offs in response to fluctuating selection on ecological timescales. We first conducted a lab-based selection experiment to quantify a genome-wide signal of fluctuating selection elicited in response to shifting population densities and in the absence of fluctuating abiotic conditions. We then conducted an independent experiment to show that lab-based manipulations of population density can identify loci relevant to selection during population expansion and collapse in an outdoor setting, where multiple biotic and abiotic conditions fluctuate simultaneously. In concert, our data indicate a role of eco-evolutionary feedbacks and generic fitness trade-offs in the maintenance of variation in natural populations and show how a coarse-grained genetic architecture of adaptation can lead to predictable evolutionary change across settings.

The CDKN2A locus, which is frequently deleted in pancreatic ductal adenocarcinoma (PDAC), encodes two tumor suppressors, ARF and INK4A, that may influence tumorigenesis through distinct mechanisms. Distinguishing their individual contributions to cancer could help improve the understanding of PDAC pathogenesis and potentially uncover targetable vulnerabilities. Moreover, whereas ARF is known to enhance p53 function, defining its p53-independent activities could elucidate new processes that drive PDAC development. In this study, we sought to understand ARF function in PDAC suppression. Analysis of gene expression and mutational patterns in human PDAC TCGA data indicated that CDKN2AARF and CDKN2AINK4A are commonly both affected by point mutations and/or deletions, suggesting that their combined inactivation contributes to PDAC development. In genetically engineered mouse models, Arf inactivation accelerated KRASG12D-driven PDAC development, both in the presence and absence of Trp53, demonstrating that ARF is a PDAC-suppressor and can act in a p53-independent manner. Transcriptomic analyses of PDACs supported a p53-independent role for ARF, with ARF deficiency promoting extracellular matrix, collagen synthesis/assembly, and epithelial-mesenchymal transition gene expression programs. Accordingly, ARF-deficient PDACs displayed extensive remodeling of the tumor microenvironment (TME), associated with collagen deposition, increased tissue stiffness, and higher fibroblast content-hallmarks of aggressive and treatment-resistant PDAC stroma. Together, this study shows how ARF deficiency associated with CDKN2A inactivation sculpts the PDAC TME in a p53-independent fashion. Given the central role of the TME in PDAC progression and therapeutic resistance, these findings may provide insight critical for improving therapeutic interventions for PDAC. SIGNIFICANCE:ARF deficiency induced by CDKN2AARF alterations promotes remodeling of the pancreatic cancer microenvironment, which could provide a genotype-specific therapeutic vulnerability to improve outcomes of pancreatic cancer patients. See related commentary by Destefanis and Mulvaney, p. 3101.

Abstract Oncogenic alterations in exons encoding the kinase domain of the Epidermal Growth Factor Receptor (e.g. EGFR L858R mutations) occur frequently in lung adenocarcinomas (LUADs) and promote tumor growth. EGFR tyrosine kinase inhibitors (TKIs), like osimertinib, have greatly improved lung cancer outcomes, yet EGFR TKI resistance remains inevitable. In addition to mutations in oncogenes, co-occurring genomic alterations in tumor suppressor genes (TSGs) have emerged as core determinants of LUAD tumor fitness and therapeutic response. Moreover, recent work suggests that the oncogenic driver dictates the effect of putative TSG inactivation on the fitness landscape of tumorigenesis. To study the effects of co-occurring TSG mutations in vivo, we leveraged autochthonous, immunocompetent genetically engineered mouse models (GEMMs) of EGFR L858R-driven LUAD, in Trp53 proficient and deficient settings, carrying a conditional Cas9 allele for CRISPR-Cas9 genome editing. We investigated the effect of inactivation of 58 putative TSGs on EGFR-driven LUAD tumor growth, tumor initiation, and osimertinib sensitivity. In parallel, we also induced tumors using the same lentiviral pool in models of Kras G12D, Kras G12D;p53-deficient, and Kras G12C-driven LUADs. In mutant EGFR-driven tumors, we identified genes that when inactivated: (i) promote tumor growth, (ii) suppress tumor growth, and (iii) reduce sensitivity to osimertinib. Inactivation of Tsc1 or Tsc2, negative regulators of mTOR-complex signaling, and the ubiquitin ligase associated genes Cul3 and Rnf43 significantly promoted tumor growth in addition to Apc, Rbm10, Rb1 described in a prior screen. Surprisingly, we also identified a set of genes, enriched in chromatin modifiers, that decreased tumor fitness in mutant EGFR-driven LUADs, including Crebbp and Smarca4. Conversely, loss of these same genes did not affect the growth of Kras G12C and G12D-driven tumors, suggesting that fitness effects of gene inactivation can vary across oncogenic contexts, even within what is conventionally considered a linear signaling axis. Indeed, loss of Setd2, Kmt2d, Ep300, and Stk11 all had significant detrimental effects on tumor growth in an EGFR context but had significant effects promoting tumor growth in a Kras G12C context. Through this screen we also identified genes that when inactivated contribute to reduced osimertinib sensitivity in mutant EGFR-driven tumors including Nf1, Kmt2d, and Pten in Trp53 proficient and deficient settings, whereas loss of Nf2 and Kdm6a only reduced sensitivity in a Trp53 deficient setting. These results inform the biology of tumor growth and reveal new genetic interactions in EGFR-driven LUADs with therapeutic implications. Citation Format: Mariana Do Carmo, Matthew Martin, Michael Rosen, Lily Blair, Anna Tribe, Keita Maemura, Giorgia Foggetti, Francisco Exposito, Zeynep Ugur, Lafia Sebastian, Vy Tran, Ian Lai, Alyna Katti, Ian Winters, Dmitri A. Petrov, Nicolas Floc'h, Monte M. Winslow, Katerina A. Politi. Tumor suppressor gene inactivation shapes the landscape of EGFR-mutant lung adenocarcinoma progression with therapeutic implications [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6058.

Supplementary Figure 10 shows Gene Ontology (GO) and GSEA analysis for molecular functions enriched in genes that are differentially expression between V1 and V3 tumors that are Setd2-proficient and Setd2-deficient.

Canine distemper-a measles-like disease with high mortality and presently without a cure-poses a major threat to wild and domestic carnivores globally. Domestic dogs are generally considered to be the main reservoir and long-range transmitter of the disease, but the role of wildlife is likely underestimated and the long-term persistence of CDV in wildlife has never been assessed. We sequenced canine distemper virus (CDV) full and partial genomes that were sampled over a decade (2012-2021) from Arctic foxes and other canids in Alaska and Yellowstone, and compiled a dataset of all published CDV genomes sampled across 32 species globally. We show the first ever evidence of persistence of CDV in wildlife (for almost a decade) with explosive transmission dynamics crossing host species barriers. Strains sampled from the Arctic for the first time connect North America to Eurasia, and are distinct from the Yellowstone strain and other known North American lineages. This suggests that separate wildlife outbreaks occur concurrently in North America, with possible introductions from Eurasia. The long-term persistence, long-range movement and explosive spread of this devastating panzootic virus that we document within wildlife is alarming and highlights the need for increased monitoring efforts to better protect wildlife populations globally.

High-quality genome annotations are essential if we are to address central questions in comparative genomics, such as the origin of new genes, the drivers of genome size variation, and the evolutionary forces shaping gene content and structure. Here, we present protein-coding gene annotations for 301 species of the family Drosophilidae, generated using the Comparative Annotation Toolkit (CAT) and BRAKER3, and incorporating available RNA-seq and protein evidence. We take a comparative phylogenetic approach to annotation, with the aim of improving consistency and accuracy, and to generate a robust set of gene annotations and orthology assignments. We analyze our annotations using a phylogenetic mixed-model approach and find that gene number and CDS length exhibit moderate phylogenetic heritability (40% and 9.7%, respectively). For comparison, we also present analyses using a subset of the 215 highest quality genomes, although the findings were not markedly different. Our work suggests that while evolutionary history contributes to variation in these traits, species-specific factors-including assembly error-play a substantial role in shaping observed differences. To illustrate the utility of our annotations for comparative analyses, we investigate codon usage bias and amino acid composition across Drosophilidae. We find that codon usage is correlated with overall GC content and evolves slowly, but that it is also strongly shaped by selection-such that, in general, species with the strongest selection on synonymous codon usage show the lowest GC bias in third codon positions. This comparative annotation dataset forms part of an on-going collaborative project to sequence and annotate all species of Drosophilidae, with data and annotations being made rapidly and freely available on an on-going basis. We hope that this effort will serve as a foundation for studies in evolutionary and functional genomics and comparative biology across Drosophilidae.

Supplementary Figure 13 shows additional data on the differences in tumor suppressor effects across different oncogenic contexts including an example plot of correlation analysis with an explanation of comparisons, correlation in the rank order of tumor suppressive effects across different oncogenic contexts calculated across various values of Ni=basal,j=basal, and correlation in the rank order of tumor suppressive effects across different oncogenic contexts across various values of Ni=basal,j=basal with genes ranked by fold change in adaptively sampled 95th percentile tumor size.

Supplementary Table 12 shows differential gene expression (log2 fold change and significance) between V1 and V3 tumors as well as between Setd2-deficient and -proficient tumors.

Supplementary Table 1 shows the frequency of mutations in the selected genes in EML4-ALK patients (from AACR Project GENIE) as well as inclusion criteria and major pathway or function

Abstract The impact of cancer driving mutations on immunosurveillance throughout tumor development remains poorly understood. To better understand the contribution of tumor genotype to immunosurveillance, we generated and validated lentiviral-based vectors that create increasingly immunogenic neoantigens. This vector system is compatible with autochthonous Cre-regulated cancer models, CRISPR/Cas9-mediated somatic genome editing, and tumor barcoding. Here, we show that in the context of oncogenic KRAS-driven lung cancer and strong neoantigen expression, tumor suppressor genotype dictates the degree of immune cell recruitment, positive selection of tumors with neoantigen silencing, and tumor outgrowth. By quantifying the impact of 11 commonly inactivated tumor suppressor genes on tumor growth across neoantigenic contexts, we show that the growth-promoting effects of tumor suppressor gene inactivation correlate with increasing sensitivity to immunosurveillance. Importantly, some genotypes also dramatically changed sensitivity to immunosurveillance independently of their growth-promoting effects. We propose a model of immunoediting in which tumor suppressor gene inactivation works in tandem with neoantigen expression to shape tumor immunosurveillance and immunoediting such that the same neoantigens uniquely modulate tumor immunoediting depending on the genetic context.

Supplementary Figure 14 shows the the response of EML4-ALK-driven lung tumors to lorlatinib in vivo, including tumor volume measured by µCT and histology, adaptively sampled mean tumor size of tumor suppressor knockouts, Kolmogorov–Smirnov (KS) distance between the cumulative density plots of the lorlatinib-treated and the “shrunk” vehicle-treated sgInert tumor sizes, estimated optimal shrinkage values of sgInert tumors, and the impact of each tumor suppressor targeting vector on lorlatinib response.

Supplementary Figure 9 shows a schematic of the experimental workflow to isolate Setd2-proficient and Setd2-deficient V1- and V3-driven cancer cells, quality control metrics of the sorted cells, principal components analyses, and a heatmaps of differentially expressed genes.