Identifying the impact of autocrine LIF signaling on myeloid cells.

Autocrine LIF signaling enhances tumorigenesis and inflammation.

Depletion of interstitial macrophages in Lkb1 mutant lung tumors after clodronate administration increases effector T cell function.

Autocrine LIF signaling induces Lkb1 mutant tumor plasticity.

Abstract Immune checkpoint inhibitors (ICI) improved outcomes in metastatic urothelial carcinoma (mUC), but primary and acquired resistance remain poorly understood. We performed single-nuclei RNA sequencing on sequential metastatic biopsies from ICI-treated mUC patients. Tumor cells showed transcriptomic heterogeneity within individual lesions, basal cells being associated with increased immune infiltration and response. Myeloid and lymphoid compartments exhibited features of immune dysfunction in non-responders. Longitudinal analyses revealed convergent adaptive resistance mechanisms, dominated by polarization toward pro-tumoral macrophage states, but also including downregulation of the antigen presentation machinery in tumor cells, increased checkpoint expression with loss of cytotoxicity in T cells. Individual trajectories point to distinct evolutionary routes under ICI pressure. Across pivotal ICI trials, bulk expression of the M2-like macrophage marker HES1 predicted ICI resistance. Our study provides the first single-cell longitudinal atlas of ICI-treated mUC, revealing macrophage reprogramming as a dominant driver of resistance, establishing a framework for individualized immunotherapy strategies.

Inhibition of autocrine LIF signaling enhances T cell effector function.

LIF neutralization eliminates a dedifferentiated stem-like tumor cell state.

Lkb1 mutant tumors transcriptionally upregulate the cytokine LIF.

Immune Analysis of Lkb1 mutant lung tumors by scRNA-seq.

Immune analysis of murine Lkb1 mutant lung tumors by flow cytometry.

LKB1 mutations in lung cancer promote an immunosuppressive tumor microenvironment, but the underlying mechanisms remain unknown. Using genetically engineered mouse models and human tumor samples, we demonstrate that LKB1 loss leads to high expression of the cytokine leukemia-inhibitory factor (LIF), which through a cancer cell-autonomous autocrine loop, orchestrates the infiltration of immunosuppressive SiglecFHi neutrophils and Arg1+ interstitial macrophages. Genetic deletion of Lifr, the receptor for LIF, on Lkb1-mutant lung tumors revealed that autocrine LIF signaling induces tumor plasticity and the emergence of a Sox17+ dedifferentiated inflammatory cell state. Antibody-mediated LIF neutralization selectively eliminates the Sox17+ tumor cell state, reduces immunosuppressive myeloid cells, and enhances antitumor T-cell responses. Our study uncovers a novel LKB1-LIF axis driving immune evasion and identifies LIF as a potential therapeutic target in LKB1-mutant lung cancer. This work highlights the interplay between tumor genetics, cellular plasticity, and immune regulation in lung cancer progression. SIGNIFICANCE:LKB1-mutant lung cancers express LIF, which induces an immunosuppressive Sox17+ tumor state. Anti-LIF therapy eliminates this state and restores antitumor immunity, revealing a novel vulnerability in this aggressive cancer subtype lacking effective targeted therapies.

Loss of function mutations in Lkb1 enhance tumorigenesis.

Loss of autocrine LIF signaling alters the immune infiltration of Lkb1 mutant tumors.

Immune analysis of human LKB1 mutant lung tumors.

Abstract Renal cell carcinoma (RCC) is amongst the most immune-infiltrated solid tumours, but only a small subset of patients achieves durable response to immune checkpoint blockade therapy. Efforts to characterize the immune microenvironment and molecular regulators responsible for treatment responses have explored numerous facets of disease biology using compartmentalized genomics, transcriptomics, and proteomics datasets, yielding many important yet context and data specific insights. Therefore, to provide a more integrated approach to informing future precision medicine strategies, we combined the complementary strengths of multiple technological platforms to profile multi-regional, spatially annotated surgical biospecimens from 65 RCC patients by single-cell RNA sequencing with paired TCR and BCR repertoire analysis, imaging mass cytometry, suspension mass cytometry, spatial transcriptomics and deconvolved bulk RNA sequencing. With this resource dataset, we explored patient subgroups and precision immunotherapy strategies using an integrated analysis of transcripts and proteins across single cell and spatial modalities. Proximal cell interactions and distinct receptor-ligand pairings identified 7 recurrent cellular communication networks. Robustly mapping reproducible gene signatures across technologies and to a variety of publicly available datasets, we show these highly refined immune subgroups stratify patients with tumour microenvironments associated with prognosis and immunotherapy response. Notably, this reveals that highly infiltrated environments with the potential for immunotherapy response may in fact comprise two distinct communication networks, with differing modes of T cell clonal expansion and immune evasion axes associated with T cell exhaustion or myeloid and NK reprogramming, which could inform targeted combination therapeutic strategies to improve outcomes. Overall, we provide a high-dimensional multi-modal resource dataset that enables cross-platform integration, links stages of T cell clonal expansion with enabling or suppressive RCC immune cell communication networks and nominates rational strategies for combinatorial precision immunotherapy. (Funded by University Health Network, Toronto; REMEDY ClinicalTrials.gov number, NCT04005183 .)

While three major genetic alteration subsets, characterized by mutations in STK11, KRAS , and EGFR , are seminal in driving tumorigenesis in LUAD, their distinct effects on tumor cells and the tumor microenvironment are not fully understood. Here, we map critical oncogenic subset-specific vulnerabilities by identifying conserved cell-type-specific reprogrammings between human and mouse LUAD. Through harmonized scRNA-seq analysis of 57 human and 18 mouse specimens, we unveil that genetic alterations impose genotype-specific immune imprints on the tumor microenvironment: KRAS is associated with a transitional immune state, whereas STK11 and EGFR mutations define discrete and contrasting immune phenotypes. We find that STK11-mutant tumors exhibit complement and interferon-rich immune microenvironments while EGFR-mutant tumors harbor a naive T cell-rich phenotype accompanied by a global HLA downregulation, stress-responsive alveolar macrophages marked by MARCO. In the epithelial compartments, cross-species analysis reveals metabolic dependencies, including OGDH in EGFR- and PDE4D in STK11-mutant cells. Statement of Significance:This study delineates oncogenotype-specific molecular and immune imprints in lung adenocarcinoma, revealing STK11-, KRAS-, and EGFR-mutated tumors as transcriptionally and immunologically discrete ecosystems. By mapping epithelial vulnerabilities alongside immune imprints, we unveil genotype-imposed dependencies that inform the rational development of precision therapies with direct translational relevance.

Renal Cell Carcinoma (RCC) has been characterized as being amongst the most immune infiltrated solid tumors with a highly heterogenous immune landscape. Within spatially organized cellular networks (CNs) of the tumor immune microenvironment (TIME), key immune cell-cell interactions (CCIs) impact immune cell function and organization ultimately impacting the patient’s overall response. Multiple studies have observed an association of tertiary lymphoid structures, a commonly observed spatial CN, with positive clinical outcomes in RCC, however additional CNs and the CCIs that control these networks need to be identified to better harness and potentially reprogram immune responses to improve patient outcomes. The recent explosion of interest in the heterogeneity of the immune landscape in RCC has led to numerous publications using the latest technologies in spatial transcriptomics, proteomics, and metabolomics. However, many of these studies use this data in isolation and therefore, may be hindered by the technological biases inherent in each method. Here, we have developed a novel approach to integrate spatial and single cell multi-omic data harnessing the strengths of each technology to better interrogate CNs that exist in the RCC TIME. Fresh surgical samples were procured at the University Health Network (Toronto, Canada) through the REnal cancer MicroEnvironment DiscoverY (REMEDY) study. Bulk RNA sequencing (RNA-seq), single cell RNA sequencing (scRNA-seq), single cell suspension mass cytometry (SMC), whole transcriptome digital spatial profiling (DSP), and imaging mass cytometry (IMC) was performed on spatially concordant tumor regions across 54 patients. scRNA-seq enabled the identification of immune, stromal, and malignant high-resolution cell states, which informed a tailored antibody panel design for SMC and IMC and served as a reference framework for harmonized cell-type annotation across modalities. This enabled the integration of our transcriptomic and proteomic data to delineate RCC-specific CNs enriched for defined CCIs across unique biological pathways. Using this integrative approach, we identified seven high-resolution patient immunophenotypes. To evaluate their prognostic and predictive relevance, we derived representative gene signatures using a linear mixed model (Flash-MM) to interrogate publicly available bulk RNA-seq datasets, including TCGA, JAVELIN, and IMmotion151. This analysis revealed immunophenotype-specific associations with survival following surgery or systemic therapy in univariate models. To explore potential biological mechanisms associated with these divergent clinical outcomes, we incorporated spatial information into traditional CCI analyses and performed pathway analyses to define functional relationships. This revealed that patient subtypes with high lymphoid infiltration exhibit greater spatial heterogeneity, potentially reflecting the coexistence of multiple activated immune pathways. In contrast, patient subtypes with low immune infiltration were enriched in more developmental signaling pathways. In addition to our biological observations, we were also able to assess the technological differences between patient matched samples and compare the ability of each technology to capture inter-patient and intra-patient heterogeneity. Collectively, this work identifies clinically distinct subgroups defined by CNs and details the interpatient cellular heterogeneity that exists in RCC, providing the foundation for future personalized therapeutic interventions against this disease.

Background Bulk RNA sequencing (RNA-seq) is an essential research and clinical diagnostics tool capable of uncovering biological insights across experimental conditions, sample types and diseases at scale. However, RNA-seq data is sensitive to batch effects introduced by technical variation. Harmonizing expression (or batch-correcting) is critical when analyzing measurements from different RNA-seq platforms. In the context of oncology, efficiently and accurately harmonizing expression at scale is important for harnessing massively large datasets (hundreds of thousands of samples) of tumor molecular profiles from different assay platforms. We aimed to develop a method for harmonizing expression in this challenging context. Results Here, we extend the widely used ComBat-seq method, as implemented in the pyComBat tool, to enable three key advances: (i) directionally adjusting counts from one batch towards a reference batch, rather than an average expression profile, (ii) separating the training and adjusting steps so that newly profiled samples not available at the time of initial model training can be harmonized, and (iii) flexibly handling outliers to improve the quality of harmonized counts. The resulting model correctly learns gene-specific differences between assay platforms and can near-instantaneously harmonize individual samples. We validated the use of the Caris-ComBat-seq tool to harmonize RNA-seq measurements on a benchmarking dataset of ∼10,000 TCGA tumor samples. Finally, we demonstrated its strength for very large datasets, by harmonizing RNA-seq data from nearly half a million tumor samples profiled by two different next-generation sequencing assays in Caris Life Science’s clinical laboratory. Source code, tutorials and manuscript data are available at: and . Conclusions We present Caris-ComBat-seq, a new variant of the ComBat-seq algorithm designed to harmonize count-based expression data in the context of high-throughput profiling laboratories. It offers the same ability to ameliorate batch effects and retain biological signal as the original ComBat-seq, with additional operational flexibility and scaling that benefits its high-throughput application. ### Competing Interest Statement All authors are current or former employees of Caris Life Sciences.

Non-small cell lung cancer (NSCLC) is one of the deadliest and most prevalent cancers worldwide, with 5-year survival rates of ~28%. The molecular heterogeneity within NSCLC encompasses several types of genetic alterations, such as mutations, amplifications, and rearrangements, and can drive aggressive tumor behavior and poor response to therapy. Among these genetic alterations are ALK and ROS1 fusions. Though these fusion events are relatively rare, their identification is crucial for selecting effective targeted treatments and avoiding therapies with significant side-effects. Fluorescent in situ hybridization (FISH), immunohistochemistry (IHC), and sequencing of DNA and RNA are standard methods to detect ALK and ROS1 fusions, but they are costly, time-consuming, and require adequate tumor tissue. Here we employ deep learning models using whole slide images (WSIs) of hematoxylin and eosin (H&E)-stained formalin-fixed paraffin embedded (FFPE) NSCLC tumor specimens to identify tumors most likely to harbor ALK and ROS1 fusions in a cohort of 33,014 patients, out of which 306 and 697 patients are positive for ROS1 or ALK fusions, respectively. A vision transformer model (MoCo-V3) was trained as a feature extractor, followed by training transformer-based models to predict the presence of ROS1 and ALK fusions. Due to the limited positive sample size for ROS1, a two-step specialized training procedure was implemented to enhance prediction performance during cross-validation. Our approach achieved receiver-operating characteristic areas under the curves (ROC AUCs) of 0.85 for ROS1 and 0.84 for ALK on a holdout dataset, demonstrating the effectiveness of this method. This framework holds significant potential for clinical application by offering a scalable, accurate, and cost-efficient method for detecting ALK and ROS1 fusions. Furthermore, it may serve as a pre-screening tool to identify candidates for confirmatory diagnostic testing and clinical trials, ultimately improving the efficiency of selecting appropriately targeted therapies for NSCLC patients.