Lung cancer predominantly affects older individuals, yet how physiological ageing influences tumour evolution remains poorly understood1. Here we show that ageing reprograms the evolutionary trajectory of KRAS-driven lung adenocarcinoma, limiting primary tumour growth while promoting metastatic dissemination through epigenetic activation of the integrated stress response (ISR). The ISR effector ATF4 drives epithelial and metabolic plasticity, conferring metastatic competence. Mechanistically, aged tumour cells show increased sensitivity to the PERK-eIF2α arm of the unfolded protein response, sustaining persistent ATF4 signalling. Targeting ISR-ATF4 genetically or pharmacologically abolishes these adaptations and limits dissemination, whereas ATF4 overexpression alone is sufficient to induce metastasis. The ageing-ATF4 axis imposes a dependency on glutamine metabolism, revealing a therapeutically actionable vulnerability. Clinical analyses confirm that ATF4 is enriched in aged tumours and correlates with poor survival and advanced-stage disease. Collectively, these results define epigenetic ISR-ATF4 activation as a causal driver of lineage plasticity and metastasis in aged tumours, revealing a therapeutic opportunity in older patients with lung adenocarcinoma, the most common yet understudied subset of lung cancer.

Background T cell receptor (TCR) binding properties have been related to a wide range of pathological conditions, including infections, autoimmunity and cancer. Characterising the TCR repertoire is of great biomedical interest but it has been challenging due to its high structural diversity. Methods In situ sequencing (ISS) is a suitable technique for spatial cell typing and linking gene patterns directly to specific histopathological features of large biopsy areas. We applied ISS through the commercial Xenium platform, with the addition of a custom panel specifically designed for TCR gene detection. Based on the IMGT database, we selected unique target sequences for TCR genes encoding the constant, variable and joining TCR chains. Additionally, we developed an analysis pipeline for the assignment of putative clonotypes based on simultaneous expression of alpha and beta variable TCR chains (TCRVβ/Vα pairs) at the single-cell level. Findings Our approach captured specific immune cell distributions in relation to the individual sample clonality, as well as regional dominance of certain TCRVβ/Vα pairs in surgical non-small cell lung cancer (NSCLC) specimens and matching lymph node samples. Furthermore, we were able to study the spatiotemporal evolution of TCR repertoire on longitudinal FFPE biopsies from patients with breast cancer, during neoadjuvant treatment. Interpretation This study highlights the implementation of target-based spatially resolved transcriptomics for the spatial characterisation of TCRVβ/Vα pairs at the single-cell level, without the need for prior sequencing. Our approach allows for spatial immune characterisation of diagnostic tissue samples with emphasis on T cell biology and accompanying T cell diversity. Funding This study was supported from Cancerfonden (CAN 2021/1726), Swedish Research Council (Dnr: 2019-01238), U-CAN and the Trond Mohn Foundation.

Purpose: The prognostic role of tumor-infiltrating lymphocytes in luminal breast cancer remains uncertain, partly because density-based metrics do not capture spatial interactions between immune cell subsets. We developed a density-independent spatial metric quantifying macrophage-T cell proximity and assessed its prognostic value. Experimental Design: Using multiplex immunohistochemistry across three breast cancer cohorts (exploratory, n = 17; discovery, n = 687; validation, n = 305), we measured nearest-neighbor distances from T cells to M1-like and M2-like macrophages, benchmarked against a randomly subsampled total macrophage pool. We defined the Macrophage Spatial Polarity Index (MSPI) as the difference between M2-to-T cell and M1-to-T cell affinity scores, where higher values reflect an M2-dominated spatial phenotype. Cox regression was used to assess associations with distant disease-free survival (discovery) and overall survival (validation). Results: M2-like macrophages preferentially localized near T cells, independent of cell density. Higher MSPI was associated with shorter survival in luminal cancers (discovery: HR = 1.45, p < 0.001), with the strongest effect in young women with early-stage disease (HR = 2.16, p < 0.0001). MSPI remained independently prognostic after adjustment for stage, systemic treatment, and diagnosis period (HR = 2.31, 95% CI 1.73-3.09, p < 0.0001) and was non-significant in HER2-positive and triple-negative subtypes. Validation in an independent ER-positive cohort confirmed the finding (HR = 1.30, p = 0.004). Pooled analysis yielded HR = 2.13 (95% CI 1.68-2.70, p = 3.45 x 10-10). Conclusions: MSPI is a robust prognostic biomarker in luminal breast cancer, particularly in young women with early-stage disease, warranting further validation for risk stratification and therapeutic guidance. ### Competing Interest Statement Ioannis Zerdes has received institutional research grants from Gilead Sciences, honoraria paid to his institution from Novartis, personal honoraria from BioMed Central, and part of Springer Nature Group (editorial tasks), all outside of the submitted work. The authors have no conflicts of interest to declare. ### Clinical Trial NCT04168528 ### Funding Statement AM was supported by a grant from the Spanish Carlos III Health Institute (Contratos Miguel Servet 2023: CP23/00133), a postdoctoral grant from the Swedish Cancer Society (CAN 2017/1066), and a postdoctoral grant from the Public Agency of the Government of Catalonia AGAUR (Beatriu de Pinos, 2021). IF was funded through the Erik och Majje Nasstrom donation to Karolinska Institutet. This work was also supported by the regional agreement on medical training and clinical research (ALF) between the Stockholm County Council and Karolinska Institute (HF, IF), the Swedish Breast Cancer Association (BRO/Brostcancerforbundet) (HF, IF), the BRECT Theme Network (HF, IF), the Swedish Cancer Society (AM, HF, IF, FP, PM, NS, JL), the Knut and Alice Wallenberg Foundation (FP), and the Swedish Research Council (NS). The funding sources were not involved in the study design, collection, analyses, or interpretation of data, writing of the report, or decision to submit the article for publication. The Phase IIa segment of the 68Ga-NOTA-Anti-MMR-VHH2 clinical trial ([NCT04168528][1]) was funded by Kom op Tegen Kanker. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: All patient-derived tissue samples and clinical data used in this study were fully de-identified prior to analysis. Each cohort was handled according to local data-protection requirements and the corresponding ethical approvals: - Exploratory cohort: tissue specimens and clinical records were obtained through the Phase IIa segment of the [68Ga]Ga-NOTA-Anti-MMR-VHH2 clinical trial ([NCT04168528][1]). Samples were coded with a study-specific identifier before being transferred to the analysing centre; the investigators performing the spatial and statistical analyses had no access to any direct identifiers (name, personal identification number, address, or date of admission). In the supplementary patient table, exact ages have been replaced with 5-year non-overlapping age bands to further reduce the risk of indirect identification. - Discovery cohort: a Swedish population-based registry cohort of women diagnosed 1992-2005. Clinical and follow-up data were retrieved from the regional cancer registry and medical records under approvals 2009/1174-31/1 and 2010/586-32 (Research Ethics Committee, Karolinska Institute, Stockholm). All variables were pseudonymised by the data custodians before release. Only aggregated demographic, treatment, and outcome variables are reported. - Validation cohort: a population-based breast cancer tissue microarray collection from Uppsala University Hospital and Vasteras Central Hospital (women diagnosed 1985-2004). Samples and clinical data were de-identified by the source biobanks and released under Regional Ethics Committee of Uppsala approvals 99 422, 2005:118, and 2005:118/2. Investigators received only coded data without direct identifiers. No individual-level identifiers (names, personal identification numbers, dates of birth, dates of admission, precise geographical location, or family relationships) were available to the research team or appear in the manuscript or supplementary materials. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes All data produced in the present study are available upon reasonable request to the authors [1]: /lookup/external-ref?link_type=CLINTRIALGOV&access_num=NCT04168528&atom=%2Fmedrxiv%2Fearly%2F2026%2F05%2F24%2F2026.05.17.26352909.atom

Abstract [Background] Integrins are transmembrane receptors that mediate cell adhesion to adjacent cells and the extracellular matrix, thereby regulating fundamental cellular processes. Abnormal integrin expression has been linked to tumor progression and metastasis across multiple cancer types, primarily through their interactions with components of the tumor microenvironment. Integrin alpha-5 (ITGA5) typically forms a heterodimer with the commonly shared beta-1 subunit and functions as a receptor for fibronectin and fibrinogen. This study aimed to investigate the clinical and biological significance of ITGA5 in non-small cell lung cancer (NSCLC). [Methods] Survival analysis was based on RNA-seq data sets from NSCLC patient cohorts of the University of Tokyo Hospital (n = 100) and The Cancer Genome Atlas (TCGA; n = 986). Protein expression was analyzed by immunohistochemistry (IHC) in diagnostic tissue samples from NSCLC patients from The University of Tokyo Hospital (n = 20) and from the Uppsala University Hospital (Sweden; n = 312), which also included extensive mutation data. The biological relevance of ITGA5 was experimentally evaluated in a xenograft model with Calu-1 cells using an ITGA5 inhibitor (GLPG0187), and the effect of ITGA5 knockdown derived from the same cell line was analyzed by bulk RNA-seq analysis. [Results] Both RNA and protein-level survival analyses consistently revealed that high ITGA5 expression was associated with shorter survival across multiple cohorts (TCGA RNA: p=0.011; University of Tokyo Hospital RNA: p=0.025; University of Tokyo Hospital protein: p=0.038; Uppsala University Hospital protein: p=0.013). Notably, protein IHC analysis indicated a greater impact of ITGA5 in tumors than in stromal cells. ITGA5 inhibition suppressed tumor growth in vivo compared to controls. Bulk-RNA seq data showed that ITGA5 knockdown altered the expression of genes involved in integrin-mediated signaling. Genes such as FERMT2, SEMA7A, and CCN1/CCN2 showed decreased expression following knockdown, consistent with their roles in integrin activation, ECM remodeling, and pro-invasive signaling. In addition, this analysis showed downregulation of pathways related to inflammation and tumor immunity (e.g., CSF3, CXCL8, IL6ST), EMT (e.g., TGFBR1, CDH2, PXN), and mTOR signaling (e.g., PIK3R2, STAT3, HRAS). [Conclusion] Our study provides a comprehensive analysis of ITGA5 expression on RNA and protein levels in the clinical context. The results indicate an essential role of ITGA5 in promoting tumor progression and poor prognosis in lung cancer, providing the rationale for further studying ITGA5 as a target with biomarker and therapeutic potential for NSCLC patients with elevated ITGA5 expression. Citation Format: Mirei Ka, Takahiro Ando, Munetoshi Hinata, Kousuke Watanabe, Akiko Kunita, Masanori Kawakami, Masaaki Sato, Hiroyuki Okada, Hironori Hojo, Tetsuo Ushiku, Cecilia C. Krona, Patrick Micke, Katsutoshi Oda, Hidenori Kage. In situ and functional analysis of Integrin-alpha 5 reveals its role in tumor progression in non-small cell lung cancer [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 2984.

Abstract Immune infiltration shapes anti-tumor responses and is associated with patient outcomes. Yet, comprehensive studies comparing infiltration patterns and their clinical impact across cancer types are scarce.We profiled the tumor immune microenvironment across 16 solid tumor types from more than 2,700 patients using multiplex immunofluorescence with pathologist-curated image analysis to quantify major lymphoid and myeloid subsets and their spatial organization in situ.Across all cancer types, lung, endometrial, and high-grade serous ovarian cancers were highly infiltrated, whereas prostate and ER-positive breast cancers were comparatively “immune cold”. Cancer-agnostic consensus clustering resolved four reproducible immune archetypes with distinct outcomes, including an immune-hot group enriched for CD4, CD8, and B cells that associated with the best survival (p<0.001). Notably, the tumor type accounted for only a part of the immune contexture, as most cancers spanned multiple immune signature groups. Nevertheless, a machine-learning classifier trained on immune compositions distinguished tumor types with high accuracy (AUC = 0.96), confirming that each cancer maintains a recognizable immune imprint even among pan-cancer archetypes.Analysis of survival associations of individual cell classes revealed a rare CD8+FOXP3+ T-cell phenotype that emerged as a strongest positive prognostic factor across cancers (HR=0.79, 95%CI:[0.70-0.89], p=0.002). Single-cell RNA sequencing data indicated that CD8+FoxP3+ cells exhibit two distinct gene programs, with regulatory and cytotoxic capacities. Spatial mapping suggested context-dependency in the function of CD8+FoxP3+ cells, as their proximity to CD8 T cells was linked to better survival (enrichment at 10um: HR=0.76, 95%CI:[0.67-0.90], p=0.0006), whereas proximity to tumor cells was associated with worse prognosis (enrichment at 10um: HR=1.29, 95%CI:[1.12-1.49], p=0.0005).Together, our findings define pan-cancer immune archetypes that are both shared and disease-specific, explaining why “immune hot” and “cold” states are not synonymous with pure infiltration quantities. We identified CD8+FoxP3+ immune cells as a cell type with strong biomarker potential independent of cancer type. Citation Format: Artur Mezheyeuski, Emma Sandberg, Max Backman, Ali Teymur Kahraman, Amanda Lindberg, Carina Strell, Hans Brunnström, Jutta Huvila, Malin Sund, Fredrik Wärnberg, Bengt Glimelius, Ina Hrynchyk, Siarhei Mauchanski, Salome Khelashvili, Klara Hammarström, Margret Agnarsdottir, Gemma Garcia-Vicién, DAVID G. MOLLEVI, Aine O´Reilly, Sara Corvigno, Hanna Dahlstrand, Johan Botling, Ulrika Segersten, Agnieszka Krzyzanowska, Anders Bjartell, Jacob Elebro, Margareta Heby, Sebastian Lundgren, Charlotta Hedner, David Borg, Jenny Brändstedt, Hanna Sartor, Per-Uno Malmström, Martin Johansson, Anna Portyanko, Björn Nodin, Cecilia Lindskog, Karin Leandersson, Karin Jirström, Tobias Sjöblom, Patrick Micke. Tumor-agnostic analysis identified pan-cancer immune archetypes and CD8+FoxP3+ cells as novel in situ predictors of survival [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 193.

This article presents an intelligent differential capacitive bioelectronic sensing system that provides an experimental foundation for future AI-assisted reliable microfluidic reagent delivery in automated pathology. The proposed platform integrates a slot-type microfluidic chamber, a differential slot-line capacitive sensor, embedded readout and signal-conditioning electronics, and a supervisory state assessment concept within a unified architecture. Its purpose is to support stable microliter-scale reagent exchange together with non-invasive process observability in automated staining workflows. The experimental study included flow calibration, analysis of feed direction and chamber tilt angle, preliminary vibration-assisted bubble mobilization, and evaluation of the sensing subsystem. The results showed that reliable operation is achieved only within a practically admissible regime in which fluidic stability and sensing informativeness overlap. In the investigated setup, upper-feed delivery and low chamber tilt angles provided the most favorable filling conditions, while the differential capacitive subsystem enabled stable detection of liquid-state changes in narrow microtubes. The reported results establish a foundation for future AI-assisted transport-state recognition and adaptive monitoring in automated pathology platforms.

Abstract Introduction: Lung cancer symptoms typically do not appear until advanced stages, leading to late diagnoses. This delay is a major contributor to its poor prognosis, resulting in lung cancer being the leading cause of all cancer-related deaths worldwide. Essential biological understanding of how lung cancers arise and progress is still lacking. Studying premalignant lesions that eventually develop into invasive carcinomas helps to bridge this knowledge gap. We are focusing on adenocarcinomas (ACs), the most common subtype of lung cancer, and especially, activating mutations in the Kirsten rat sarcoma virus oncogene (KRAS), which is among the most common oncogenic mutations and has previously been reported in premalignant lung lesions. Given that KRAS signaling is growth-promoting, it is reasonable to assume that KRASmut promotes tumor formation and progression. However, the extent of its influence, the timing of when it arises, and particularly its connection to histologic and phenotypic changes driving tumor evolution, remain unclear. Methods: We have collected a cohort from biopsy material, initially collected for diagnostic purposes from AC patients at varying stages of tumor evolution, with some matched ‘normal-histology’ samples when available. In total, tissue from 38 patients with AC premalignancies is available for investigation. To map KRAS mutations in situ, we employed a BaseScope assay with commercial detection probes against the different KRASmut transcripts. To map KRAS statuses to cellular phenotypes in the microenvironment, we are running a 6K panel on the CosMx platform (Bruker) on a subset of patients, which is a highly sensitive spatial transcriptomics technique at single-cell resolution. Spatial Mass Spectrometry (MS) tracking metabolic features on the timsTOF flex MALDI-2 (Bruker) is also being performed on a subset of patients. Results: Initial tests of the BaseScope assay have revealed specific, yet not sensitive, signals of KRASmut transcripts in premalignant lesions. While advanced cancers display distinct signals, premalignant lesions have fewer, if any, signals. This might be due to the lower expression levels of KRAS at the earlier stages. To address this, we are now applying a BaseScope assay with signal amplification to be able to track KRAS expressed at lower levels. Preliminary analyses of the Spatial MS matrix have revealed highly individual metabolomics features, with great inter-patient heterogeneity. This pattern is also reflected in bulk RNA-seq data, revealing expression profiles more similar across, e.g., Adenocarcinoma in Situ (AiS) and invasive AC from the same patient, than between the AiSs from different patients. Conclusions: By applying this state-of-the-art multi-omics approach to premalignant adenocarcinomas, we hope to identify key regulators of early lung cancer progression and to translate these findings to clinically relevant basic molecular diagnostics. Citation Format: Amanda Lindberg, Bartosz Sobocki, Patrick Micke, Carina Strell. In situ KRAS-mapping and spatial -omics characterization in lung premalignancy [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 687.

Abstract Background: Immune checkpoint inhibitors have transformed non-small cell lung cancer (NSCLC) treatment, yet durable benefit remains limited. Current biomarkers, such as PD-L1 immunohistochemistry (IHC), measure protein abundance rather than functional PD1-ligand engagement and therefore provide limited predictive value. The second PD1 ligand, PD-L2, has often been ignored, but may influence immune regulation in the cellular context. Therefore, the activation of the PD1-PD-L1 axis is not sufficient to explain clinically relevant immune mechanisms. Methods: We analyzed a tissue microarray of early-stage, surgically resected NSCLC (N=345), a cohort previously extensively immunoprofiled, including PD-L1 IHC and PD1-PD-L1 proximity ligation assay (PLA) [1]. We extended this work using fluorescent PLA (co-stained for CD3 and pan-cytokeratin) to visualize PD1-PD-L2 interactions across tumor, stromal, and T-cell compartments, alongside PD-L2 protein expression by IHC. Image analysis was performed in QuPath. Results: In normal lung, PD-L2 was primarily expressed on macrophages and occasional granulocytes. Among tumors, squamous cell carcinomas (SqCC) showed higher PD-L2 expression than adenocarcinomas (AD) (p=0.003) and increased PD1-PD-L2 interactions (p=0.047), consistent with greater immune infiltration. Notably, amplification of 9p24.1, including the PD-L1 and PD-L2 locus, was also observed more frequent in SqCC. PD1-PD-L2 interactions displayed a distinct immune contexture from PD1-PD-L1. Within tumor cell compartment, PD1-PD-L2 correlated only weakly with CD4+, CD8+, CD163+, FoxP3+, PD1, and PD-L1, while PD1-PD-L1 showed strong associations with these immune populations, indicating partly distinct mechanisms of immune regulation. In the stroma, both ligands showed more similar immune-association patterns. Clinically, high PD1-PD-L1 interaction was generally associated with better prognosis. In contrast, PD1-PD-L2 interaction in SqCC showed only a trend toward better survival (p=0.061), whereas in AD it demonstrated a trend toward shorter survival (p=0.092). Also, compartment-specific PD1-PD-L2 interactions in the tumor stroma were associated with poor prognosis, particularly in SqCC with T cell-rich stroma (p=0.033). Conclusion: Our findings suggest that PD1-PD-L2 represents a spatially and biologically distinct checkpoint axis in NSCLC, differing from PD1-PD-L1 in immune phenotypes and clinical associations. Including PD-L2-specific interaction analyses may enhance understanding of checkpoint biology and support refined patient stratification, particularly in SqCC, where PD-L2 co-amplification and upregulation appear to be more common. References: [1] Lindberg A, Muhl L, Yu H, et al. J Thorac Oncol. 2025;20:625-640. Citation Format: Anna Gorbunova, Amanda Lindberg, Lars Muhl, Ghazal Lessan Toussi, Hui Yu, Patrick Micke, Carina Strell. Spatial mapping of PD1-PD-L2 interactions suggests an additional checkpoint axis in non-small cell lung cancer [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 2820.

BACKGROUND:Non-small cell lung cancer (NSCLC) is a heterogenous disease with challenging prognosis despite clinical improvements in the last years. The tumor-microenvironment (TME) is a major factor with many cell types involved as well as distinct spatial organization. We assessed the global immune landscape and spatial architecture of excluded immune cell types within the stroma of NSCLC tissues. METHODS:Multiplex Immunofluorescence (mIF) staining was used to detect major immune cell lineages in tissues of 674 NSCLC patients. AI-based digital image analyses were conducted to stratify NSCLC tissues into stroma and tumor compartments as well as identify and classify single cells based on mIF marker combinations into major immune cell phenotypes. Quantitative single-cell resolved immune cell phenotypes within the stroma area were used to cluster patients based on their excluded immune signature and spatial architecture and explored for impact on overall-survival. RESULTS:NSCLC stroma harbored distinct immune cell as well as spatial organizational patterns with varying combinations. Among these, a mixed immune landscape together with a stromal organization dominated by long distances to B cells and short distances to Helper T cells was associated with long-term overall-survival. This was successfully reduced to a two-variable signature comprising three different cell types, where short distances of Helper T cells to Regulatory T cells in conjunction with high B cell densities provided superior prognostic value. CONCLUSIONS:The excluded immune cells within the stromal compartment are associated with strong prognostic effects based on their spatial architecture and represent a previously underappreciated cell population.

Abstract Background Sensitive and specific blood biomarkers for early detection of colorectal (CRC), lung (LuCa) and ovarian (OvCa) cancers are highly warranted. The current blood tests often need to be complemented with other clinical methods to achieve adequate diagnostic performance. Recent progress in the molecular profiling of plasma from cancer patients holds potential for improved non-invasive screening. Here, we aimed to identify composite proteomic and metabolomic plasma biomarkers for early cancer detection with performances that exceed those of existing FDA-approved blood and stool-based diagnostic tests for CRC, LuCa and OvCa. Methods In a case-control study using samples from the U-CAN and EpiHealth biobanks, we measured plasma levels of 165 proteins and 244 metabolites in 818 patients with CRC, LuCa and OvCa at diagnosis, 119 patients with non-malignant conditions of the corresponding organs, and 1,129 healthy individuals. We performed an exhaustive search over all cut-off values of the ROC for all combinations of up to 4 proteins and metabolites, implementing measures to minimize the impact of cross-cohort comparisons. Ultimately, we benchmarked candidate biomarkers performance to FDA approved blood tests in clinical use for detection of cancer. External validation was performed using publicly available datasets. Results We found biomarkers composed of 2–4 proteins separating cases of each tumor type from healthy controls with ROC AUC, respectively for CRC: CEACAM5, FLT1, IL19, Ferritin (AUC 0.89), LuCa: FNDC5, MDK, PLAUR, CEACAM5 (AUC 0.91) and OvCa: MUC16/CA125, PLG (AUC 0.97). The diagnostic performance of these biomarkers was comparable to, and in some instances surpassed, the performance of established tests, such as Epi proColon (AUC 0.82) and Cologuard (AUC 0.93). Metabolites were informative for tumor stage discrimination, especially in LuCa and OvCa. External validation in the CancerSeek dataset showed strong agreement in biomarker performance. Conclusions The composite protein biomarkers identified in this study represent a novel opportunity for detection, staging and differential diagnosis of common tumor types. Metabolites are more relevant for tumor staging than early detection. A limited number of well-performing analytes enables cost-effective implementation in clinical diagnostics and merits further evaluation.

Increased stromal Platelet-derived growth factor receptor beta (PDGFRβ) expression is a hallmark of the desmoplastic tissue reaction in cancer and marks subsets of cancer-associated fibroblasts, pericytes, and smooth muscle cells. However, its functional status in situ has been anticipated from static expression measures, which cannot determine whether high receptor abundance reflects active signaling. We established two second-generation proximity ligation assays (PLAs) to quantify PDGFRβ activation in the in situ environment of human lung cancer by detecting either phosphorylated PDGFRβ or its interaction with the adaptor protein Grb2. The immunofluorescence-based assays were applied to tissue-microarrays including diagnostic samples from over 600 non-small cell lung cancer (NSCLC) patients. In lung cancer tissue, activation scores correlated with PDGFRβ expression but revealed a more nuanced receptor status, indicating variable activation despite similar expression levels. Higher PDGFRβ activation was associated with increased recurrence risk exclusively in squamous cell carcinoma, a finding not captured by conventional immunohistochemistry. This activation was accompanied by a specific stromal profile enriched for LRRC15- and FAP-positive cells, a pattern absent in adenocarcinomas. PDGFRβ activation status provides functional information beyond receptor expression, uncovering clinically relevant, otherwise overlooked, stromal phenotypes. The approach illustrates the diagnostic potential of functional protein assays in the era of precision medicine.

Background Multiplex immunofluorescence imaging enables detailed characterization of the tumor immune microenvironment, but whether immune cell densities add prognostic value beyond established clinical factors in non-small cell lung cancer (NSCLC) remains unclear. Methods Tissue samples from an NSCLC cohort (n = 298) were stained with a multiplex immunofluorescence panel targeting immune cell markers (CD4, CD8, FoxP3, CD20), cancer cells (pan-cytokeratin), and cell nuclei (DAPI). We quantified immune cell densities, nuclear pleomorphism features, and clinical variables, and trained four machine learning models (logistic regression, random forest, support vector machine, and k-nearest neighbors) to predict overall survival. Results Clinical parameters consistently demonstrated the strongest performance in predicting long and short-term survival (logistic regression mean accuracy 0.60 ± 0.01, AUC 0.66 ± 0.01). The addition of immune cell densities revealed a small, statistically significant improvement in survival prediction (accuracy 0.62 ± 0.01, p < 0.01, AUC 0.67 ± 0.01, p = 0.04), while nuclear pleomorphism features did not improve prediction. When combined with clinical parameters, immune cell densities also improved survival stratification in Cox regression analyses numerically (HR = 0.51 vs. 0.55 for clinical parameters alone). Model interpretation analyses showed that stage and performance status have the largest effect on model performance. Selected immune cell densities (tumor CD4-helper and stroma B-cells) have a limited but consistent effect. Conclusion Clinical parameters remain the dominant predictors of outcome in NSCLC, with immune cell densities providing only limited prognostic value for clinical stratification. The openly available code and datasets present a unique resource for method development or focused analysis.

Abstract The role of nerve innervation in cancer progression remains highly debated. While various in vitro and in vivo models suggest that tumor cells can actively induce neoneurogenesis, tissue-based evidence remains sparse. In this project, we critically examine the hypothesis of tumor-associated nerve innervation by seeking direct, tissue-based evidence of nerve innervation and sprouting in relation to histopathological tumor features.Our study cohort comprises diagnostic FFPE whole-tissue samples from six solid tumor types: breast cancer (Luminal A/B, HER2+, TNBC), non-small cell lung cancer (adenocarcinoma, squamous cell carcinoma), colorectal cancer, pancreatic cancer (PDAC and periampullary adenocarcinoma), prostate cancer, and urinary bladder cancer.Nerve structures are identified via multiplexed immunofluorescence and multispectral imaging, targeting neurofilament light chain (NFL, 70 kDa) and growth-associated protein 43 (GAP43). To delineate the tumor microenvironment, additional markers include CD34 (perineural sheath), CD31 (endothelial cells), CD3 (T-cells), and pan-cytokeratin (tumor cells). We developed both a deep-learning-based algorithm and a thresholding approach to segment nerve fibers and characterize their spatial organization. A custom Python script quantifies nerve density (nerves/tumor area), nerve size (μm2), and the integrity of the perineural sheath via CD34 staining. In parallel, we are constructing 3D nerve reconstructions by aligning and analyzing 30 consecutive 4 μm sections using tailored Python tools.Perineural invasion, as defined as direct contact between tumor cells and nerves, is most frequently observed in tumors from highly innervated organs such as PDAC and prostate cancer, while appearing only sporadically in the other tumor types.In subsequent analyses, nerve features will be systematically correlated with histopathological tumor characteristics, microenvironmental profiles, and clinical data. Ultimately, this study aims to generate a comprehensive atlas of nerve-tumor interactions in human cancer and to delineate both shared and tumor-type-specific patterns of innervation and perineural invasion. Citation Format: Hui Yu, Aglaia Schiza, Victor Ponten, Viktoria Thurfjell, Amanda Lindberg, Julia Sidenius Johansen, Ulrike Segersten, Anca Dragomir, Bengt Glimelius, Artur Mezheyeuski, Astrid Børretzen, Yun-Fan Sun, Lars A. Akslen, Patrick Micke, Carina Strell. Nerve innervation in solid tumors [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 720.

Abstract Introduction: Immune checkpoint inhibitors (ICIs) have transformed cancer therapy, yet durable clinical benefit remains limited to a subset of patients. Current biomarkers, such as PD-L1 immunohistochemistry (IHC), reflect protein abundance rather than functional receptor-ligand engagement and therefore provide only modest predictive value. To address this, we previously demonstrated that mapping PD1-PD-L1 interactions using a proximity ligation assay (PLA) outperforms PD-L1 IHC in predicting immunotherapy response in non-small cell lung cancer (NSCLC) [1]. Methods: Building on these findings, we developed a triplex PLA detecting PD1-PD-L1, PD1-PD-L2, and CD8-MHC I interactions and integrated it with Imaging Mass Cytometry (IMC). This approach enables high-plex (≈40 markers) spatial quantification of active immune signaling pathways alongside detailed immune phenotyping. The platform was applied to human tonsil tissue as a biological control and to pre-treatment NSCLC and triple-negative breast cancer (TNBC) biopsies to map checkpoint engagement, antigen recognition, and T-cell functional states within the tumor microenvironment. Results: In tonsil, PD1-PD-L1 and PD1-PD-L2 interactions were observed primarily between CD8+ T cells and follicular B cells, while additional PD1-PD-L1 interactions occurred between CD8+ T cells and CD68+ macrophages. These patterns suggest partly distinct cellular contexts for PD-L1 and PD-L2 engagement in lymphoid tissue. Consistent with this, preliminary tumor data indicate that PD1-PD-L2 interactions are relatively rare in TNBC compared with NSCLC, suggesting that PD-L2 involvement varies across tissue types and immune environments. Conclusion: PLA-IMC provides a functional, spatially resolved platform for screening and quantifying active immune checkpoint signaling directly in tissue. This approach holds promise to identify functional biomarkers for refined patient stratification and mapping of context-specific checkpoint activity to guide rational immunotherapy combinations in solid tumors. References: [1] Lindberg A, Muhl L, Yu H, et al. In situ detection of programmed cell death protein 1 and programmed death ligand 1 interactions as a functional predictor for response to immune checkpoint inhibition in NSCLC. J Thorac Oncol. 2025; 20:625-640. Citation Format: Ghazal Lessan Toussi, Lars Muhl, Anna Gorbunova, Austin James Rayford, Amanda Lindberg, Neda Hekmati, Viktoria Thurfjell, Aglaia Schiza, Agata Zieba Wicher, Patrick Micke, Carina Strell. Development of a proximity ligation-imaging mass cytometry platform for spatially resolved immune checkpoint analysis [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 1218.

Abstract Background: Hepatic metastases from colorectal cancer represent one of the leading causes of mortality associated with this disease, and their management remains a clinical challenge that requires increasingly precise diagnostic and therapeutic strategies. We have previously reported colorectal liver metastases (CRCLM) with desmoplastic histological growth pattern (dHGP) to have more cytotoxic immune environment, enriched by CD8+ lymphocytes over tumor nests. FoxP3+ cells followed the CD8+, but CD4+FoxP3+ (Tregs) did not demonstrate difference. Here we focused on rare lymphocyte subset: CD8+FoxP3+ cells. Methods: We applied multiplex IHC panel with immune cell markers for tissue samples from 100 patients with CRCLM, characterized by HGP. Cell classes included Cytokeratin (CK)-positive cells, CD8, CD4, CD4+FoxP3+, CD8+FoxP3+, CD20+ and HSA (Hepatocyte specific antigen). Cell density defined as number of cells per mm2. To resolve spatial patterns, we computed neighborhood enrichment (ENR) and the pair correlation function (PCF). ENR, evaluated distance to k-nearest neighbors (k=5-100), and quantified attraction or avoidance between cell classes while controlling for cell density. PCF, computed over radial distances (10-50 µm). Results: The initial cell density analysis demonstrated that CD8+FoxP3+ cells were significantly more abundant in dHGP (p=0.003), following similar arrangement of larger CD8+ cell set (p=0.007). Spatial analysis, adjusted to cell density, demonstrated more complex patterns. First, CD8+FoxP3+ showed strong spatial attraction to CD8 cells in non-dHGP, while in dHGP, where both cell types were significantly more abundant, they tended to disperse (confirmed by agreement across five spatial metrics, with k=5-50, r=50, p=0.01-0.004). Second, In dHGP, the CD8+FoxP3+ cells had strong tendency to neighbor to CD20+ (B-cells) (confirmed by agreement across five spatial metrics, with k=5-25, p=0.05-0.005). Third, CD8+FoxP3+ cells demonstrated attraction to cancer cells, in dHGP, detected by ENR at k=5-25 (p=0.01), indicating nearest vicinity or direct contact. Conclusions: The neighboring tendencies observed between CD8+FoxP3+ and CD8+ cells in non-dHGP, suggest either a strong biological interaction between these two T-cell subsets or dynamic conversion between states of CD8+ linage. In dHGP, CD8+FoxP3+ cells were instead redirected toward CD20+ B-cell rich areas and showed nearest-neighbor attraction to CK-positive tumor cells, consistent with positioning in or near tertiary lymphoid structure (TLS)-like niches and at the tumor-stroma interface. Altogether, this provides the first spatially resolved description of CD8+FoxP3+ cells in CRCLM and highlights their abundance and neighborhood context as potential contributors to the distinct immune architectures and clinical behavior of dHGP versus non-dHGP liver metastases. Citation Format: Artur Mezheyeuski, Gemma Garcia-Vicién, Núria Ruiz, Jose C. Ruffinelli, Kristel Mils, María Bañuls, Natalia Molina, Miguel A. Pardo, Laura Lladó, Patrick Micke, David G. Mollevi. Histologic growth pattern dictates immune cell spatial topography in liver metastases [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 6120.

Overtreatment of breast ductal carcinoma in situ (DCIS) remains a major clinical problem because clinicopathologic criteria are insufficient to define the invasive potential. We evaluated whether periductal stromal morphology in diagnostic hematoxylin-eosin sections predicts invasive recurrence among 711 DCIS patients from the randomized SweDCIS trial, who underwent breast-conserving surgery with or without radiotherapy. Two raters scored the proportion of DCIS ducts surrounded by myxoid stroma. Eighty patients developed ipsilateral invasive recurrence as first event post surgery. Cause-specific Cox regression revealed that an increase in myxoid stroma was associated with a higher risk of invasive recurrence (adjusted HR per 10% increase = 1.16 [1.04-1.28]; p = 0.005). The 20-year cumulative incidence of invasive recurrence was more than doubled for lesions with at least 10% myxoid stroma compared with those with less than 10% myxoid stroma (15.7% vs 7.1%). Periductal myxoid stroma may be a simple histologic marker for the invasive potential of DCIS.

Automated multiplex immunohistochemistry (IHC) and in situ hybridization (ISH) require staining platforms that combine stable reagent exchange, low-volume operation, process observability, and protocol flexibility. Existing autostainers are often rigid and costly, whereas microfluidic and sensing solutions remain largely component-specific rather than system-oriented. This study proposes and partially validates a layered cyber-physical architecture for multiplex histological staining. The architecture integrates five functional layers—biochemical workflow, fluidic processing, capacitive sensing, protocol-driven control, and software-based process representation—within a unified formal framework and is supported at the subsystem level by experimental characterization of its fluidic and sensing layers. Fluidic experiments on a slot-type microfluidic chamber identified a practical operating window in which upper-feed operation, moderate calibrated flow conditions, and low chamber angles between 10° and 40° provide stable filling and acceptable drainage. The differential slot-line capacitive sensing subsystem detected liquid volumes as low as 0.5 µL, with stable threshold-based interpretation at a practical detection threshold of approximately 5 fF after digital filtering. The control and software layers are specified at the architectural and formal model level; their hardware implementation and closed-loop validation remain subjects of future work. Together, the reported results demonstrate that controlled reagent transport and sensing-based process observability are jointly feasible within the proposed modular framework, establishing a conceptual and experimental foundation for scalable, flexible, and resource-efficient multiplex IHC/ISH systems.

Abstract Background: Accumulating evidence suggests that the spatial organization of tumor-immune ecosystems determines the benefit of checkpoint inhibitors. Therefore, topography-aware metrics that adjust for baseline cell abundance are better suited to quantify spatial immune patterns, such as attraction, repulsion, and clustering, across scales. We hypothesized that such multiscale spatial metrics would outperform immune cell density alone for explaining therapy benefit and survival in advanced non-small cell lung cancer (NSCLC) patients. Methods: We applied a multiplex immunofluorescence pipeline on diagnostic tissue of 54 NSCLC patients treated with checkpoint inhibitors. Annotated cell classes included cancer cells, CD8+, CD8+FoxP3+, M1 and M2 macrophages (-/+PDL1), CD68-CD163+ cells and stromal non-immune cells. To spatially resolve the immune patterns, we computed neighborhood enrichment (ENR) and radial local enrichment (RLE). ENR was computed over k-nearest neighborhoods (k=5-100) and provided information about the attraction tendency between two cell classes, independent from cell density. RLE was evaluated over 10-50 µm radii and reflected the spatial cell clustering tendency. The spatial metrics were associated with therapy response and overall survival. Results: CD8+FoxP3+ cells stratify risk. Responders demonstrate neighboring of CD8+ or CD8+PD1+ cells to CD8+FoxP3+ at different scales (ENR k=5-100, p=0.04-0.007), strongest at shorter distances. In the survival analysis, CD8+FoxP3+ cells were associated with shorter survival when they were closer to PDL1+ cancer cells (ENR: HR 3.7-7.5, p=0.01-0.04). CD8 geometry in stroma tracks response but not survival. Increased clustering of CD8+PD1+ associated with higher response to therapy (ENR k=10-100, p=0.05- 0.002). Notably, CD8+PD1- cells did not show such associations. This positive impact on therapy response did not translate into a longer overall survival in this cohort. Tumor purity and compactness is associated with shorter survival. Local enrichment of cancer cells within 10-50 µm was associated with poor response and shorter OS, (RLE: HR 2.2-2.5, p=0.03-0.007), pointing to denser, homogeneous tumor nests as a risk factor. Conclusion: Topography-aware spatial metrics capture clinically meaningful organization beyond simple density. These findings supported the cellular context-dependent evaluation of biomarkers. Furthermore our findings also provide in situ evidence for a focused evaluation of a rare, hitherto poorly described subtype of CD8+FoxP3+ immune cells. Citation Format: Artur Mezheyeuski, Neda Hekmati, Ina Hrynchyk, Amanda Lindberg, Jakob Friedrich, Johanna Mattsson, Miklos Gulyas, Klas Kärre, Johan Isaksson, Carina Strell, Patrick Micke. Beyond density: Topography-aware spatial metrics predict immunotherapy outcomes [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 6546.

Leaky blood vessels are a hallmark of solid tumors. However, the molecular mechanisms and clinical implications of vascular leakage in human cancer remain unexplored. Here, we identified fibrinopeptide-A (FpA) as a robust in-situ marker of vascular leakage, analyzing diagnostic specimens from two non-small cell lung cancer (NSCLC) cohorts (N = 327 and N = 200). Mechanistically, FpA+ staining localized to discrete stromal niches characterized by increased endothelial VEGF receptor-2 phosphorylation, elevated VEGFA production by tumor cells and loss of the endothelial tyrosine phosphatase PTPRB. Immune profiling revealed reduced density of mature dendritic cells and granzyme B-expressing cytotoxic T cells in tumors with high-leakage. This leakage-associated immunosuppression was linked to the presence of tertiary lymphoid structures (TLS) in lung adenocarcinoma (LUAD). However, compared to low-leakage, TLS in high-leakage tumors were enriched in regulatory T cells and conferred no survival advantage. Patients with high vascular leakage exhibited a reduced overall survival. Moreover, in a separate immunotherapy cohort (N = 64), high-leakage was associated with poor response to anti-PD-1/PD-L1 treatment. This study establishes vascular leakage as an important prognostic factor in NSCLC, and provides mechanistic understanding and a methodological framework to stratify NSCLC patients with regard to responsiveness to immunotherapy.

The tumor microenvironment (TME) has emerged as a promising source of prognostic biomarkers. To fully leverage its potential, analysis methods must capture complex interactions between different cell types. We propose HiGINE – a hierarchical graph-based approach to predict patient survival (short vs. long) from TME characterization in multiplex immunofluorescence (mIF) images and enhance risk stratification in lung cancer. Our model encodes both local and global inter-relations in cell neighborhoods, incorporating information about cell types and morphology. Multimodal fusion, aggregating cancer stage with mIF-derived features, further boosts performance. We validate HiGINE on two public datasets, demonstrating improved risk stratification, robustness, and generalizability.