Antibody-drug conjugates (ADC) as a new treatment modality have enabled novel, promising treatment options in lung cancer. However, biomarkers for the selection of appropriate treatments are still under development, posing novel challenges for tissue selection and development of companion diagnostics (CDx). In this review, we address the challenges and discuss best practice for fast, reliable and robust implementation of novel biomarkers in treatment selection of ADCs in lung cancer.

Circulating tumor DNA (ctDNA) has emerged as a transformative tool in cancer diagnostics, enabling the non-invasive detection of tumor-derived DNA fragments released into the bloodstream through cellular lysis or active secretion. ctDNA measurement has demonstrated its clinical usefulness, including early cancer detection, identification of resistance mechanisms, and screening of asymptomatic individuals. In addition to prognosis, ctDNA analysis is increasingly used to guide adaptive treatment strategies by detecting minimal residual disease and tracking tumor evolution in real time. Recent advances in artificial intelligence are poised to further enhance the clinical impact of ctDNA, transforming it from a passive monitoring biomarker into a dynamic molecular sensor integrated into predictive clinical decision models. However, broad implementation of ctDNA-based assays in routine practice requires rigorous prospective validation, cross-platform standardization, and regulatory approval to unlock its full potential in precision oncology.

Exposure of cells or tissues to chemical compounds can be analyzed through transcriptomic signatures, which can be used to classify chemical agents. This information can also enrich Adverse Outcome Pathways (AOP). Transcriptional signatures have generally been obtained using “bulk” analysis, by which the global gene expression pattern of an entire tissue is determined. Although this approach has been useful in toxicology, some information is lost, especially when tissues containing multiple cell types are considered. With the advent of single-cell transcriptomics (scRNA-seq), it is now possible to obtain higher resolution, cell type–specific responses in complex tissues. The aim of the present study was to evaluate the added value of scRNA-seq in analysis of the acute response of human bronchial epithelial cells grown at the air/liquid interface (ALI) to a known toxic compound, CdCl2, with well described transcriptional signatures of exposure. Fully differentiated mucocilliary epithelia obtained from three independent donors were exposed to 10 µM CdCl2 and scRNA-seq analysis was performed on a total of 18,255 cells to obtain cell type–specific signatures. Our results show that the contribution of each cell type to the overall transcriptomic bulk response varies. For example, the classical heavy metal detoxification response was only detected in multiciliated and secreting cells, while absent in basal cells. The data demonstrate that scRNA-seq provides high-resolution transcriptional signatures with unexpected features. This added information is likely to have implications for the refinement of AOPs and could serve as a basis for a new generation of tests in predictive toxicology.

Interstitial lung diseases, particularly idiopathic pulmonary fibrosis (IPF), have dismal prognoses, with a median survival of 3-5 years, owing to a lack of early biomarkers or effective treatments. This review highlights the lung microbiome as a key biological factor in IPF pathogenesis and a promising therapeutic target. Elevated burdens of pathogenic bacteria, including Streptococcus and Staphylococcus, in bronchoalveolar lavage fluid correlate with accelerated progression and higher mortality. These bacteria release toxins and activate Th17-driven inflammation, providing mechanistic links to alveolar injury and fibrosis. Host genetics and systemic factors, including oral-gut-lung interactions, further shape disease progression. Although antibiotic trials have been unsuccessful, embracing the microbiome as an active participant in IPF may open unprecedented opportunities for personalized interventions.

Shared epitopes pose safety and efficacy issues for T-cell immunotherapy. To characterize the extent of this problem, we performed a computational analysis establishing a complete atlas of shared identical sequences across the human and murine proteomes. Unlike bacterial or viral antigens, self-antigens, including tumor-associated antigens (TAAs), frequently contain sequences of sufficient length to generate identical epitopes in other self-proteins. Epitopes from these shared sequences can theoretically reduce target specificity, confound immunomonitoring studies, and contribute to pre-existing immune tolerance toward TAAs. Notably, a subset of TAAs identified in this atlas is free of this drawback, providing a new criterion for antigen prioritization in cancer immunotherapy. To facilitate the detection of shared sequences, a web server has been made available at https://epitopemapper.ircan.org/ and the open-source code at https://github.com/IRCAN/EpitopeMapper.

Processing bodies (P-bodies) are cytoplasmic, membraneless organelles that play a key role in regulating RNA translation. To identify new pathways controlling their formation, we conducted a Food and Drug Administration (FDA)-approved drug screen. We found that glucocorticoids, among the most prescribed medicines, significantly increase P-body numbers across diverse epithelial cell types. This effect was fully reversible after glucocorticoid withdrawal, illustrating the adaptive dynamics of P-bodies. Using genetic invalidation and rescue approaches, we demonstrated that this accumulation requires the Glucocorticoid Receptor alpha isoform. P-body accumulation was associated with the sequestration of P-body-specific-targeted mRNAs, altering their translation yield. Notably, this translational regulation depends on transcript sequence features rather than abundance, with AU-rich mRNA transcripts being sequestered and GC-rich mRNAs preferentially translated under glucocorticoid treatment. Furthermore, we linked the decrease of LSM14B, a negative regulator of P-bodies, under glucocorticoid treatment to P-body reshaping. Our results reveal that, beyond their known transcriptional activity, prolonged exposure to glucocorticoids influences mRNA post-transcription and translation through a nucleotide composition-based mechanism. ### Competing Interest Statement The authors have declared no competing interest. Agence Nationale de la Recherche, ANR-11-LABX-0028-01, ANR-15-IDEX-01, ANR-23-IAHU-0007, ANR-10-IDEX-0002, ANR-20-SFRI-0012 Fondation ARC pour la Recherche sur le Cancer, https://ror.org/0489qz649, Canc’air GENExposomics, PJA-20191209562 Cancéropôle PACA, https://ror.org/01mwvah42 Institut National du Cancer, INCa_18414 Institut Thématique Cancer, 18CN045 Ministère de l’Enseignement Supérieur et de la Recherche, https://ror.org/03sjk9a61

Erythropoietin (EPO) is a hormone mainly produced by the kidney. EPO stimulates red blood cell production in response to hypoxia. EPO represents one of the most compelling paradoxes in medical oncology. For decades, recombinant EPO has been prescribed as an effective therapy for cancer-related anemia and fatigue, significantly improving patients' quality of life while reducing their dependence on blood transfusions. Yet, accumulating scientific evidence knowledge has highlighted the complex interactions between EPO and tumor biology that extend well beyond its hematopoietic functions. The putative expression of EPO receptors on tumor cells has raised concerns about a potential role of EPO in promoting tumor progression, although conclusive clinical evidence remains lacking. Recent studies have revealed that EPO may contribute to immunosuppressive mechanisms within the tumor microenvironment, thereby reshaping our understanding of its risk/benefit profile in oncology. This evolving body of evidence calls for an objective reassessment of EPO's dual nature as both a therapeutic ally and a potential oncological threat, in order to bring more caution around decision-making in the current era of strongly evolving cancer care.

Tumor-associated antigens (TAAs) are the targets of several therapeutic cancer vaccines. However, many TAAs contain epitopes identical to unintended targets, creating shared epitopes with other human proteins in normal tissues. Moreover, for some TAAs like ASCL2, KLK2, TPTE, CLDN6, and PSMA, the off-targeted proteins are often expressed at a higher level in healthy tissues than the target in cancer, potentially impacting both the safety and the efficacy of T cell immunity. Altogether, our analysis indicates a suboptimal design of several cancer vaccines currently in clinical development: ATP128, BNT111, BNT112, BNT116, INO-5401. We recommend that next-generation cancer vaccines should integrate rigorous epitope filtering strategies to eliminate shared sequences in TAAs.

Programmed Death-Ligand 1 (PD-L1) is a major target for immunotherapy using checkpoint inhibitors (CPIs), particularly in lung cancer treatment. Tumoral PD-L1 expression has been recognized as a natural predictor of CPI response. This predictive relationship is primarily due to its upregulation by interferon-gamma, which is released by immune cells (mainly T lymphocytes and natural killer cells) in proximity to tumor cells, driving an immune resistance mechanism. However, PD-L1 expression is modulated at multiple levels, including oncogenic signaling pathways, and transcriptional and post-transcriptional regulations, potentially leading to false positive predictions. Conversely, variable glycosylation of PD-L1 may compromise the accuracy of immunohistochemical measurements, resulting in false negative predictive data. In addition, PD-L1 expression demonstrates relative instability throughout treatment courses (e.g., chemotherapy and tyrosine kinase inhibitors), further limiting its clinical utility. In this review, we focused on the molecular mechanisms governing PD-L1 expression with a special emphasis on lung cancer. We also discussed biomarker strategies for optimizing patient selection for checkpoint inhibitor therapy where multimodal/multi-omics meta-biomarker approaches are emerging. Such comprehensive PD-L1-enriched biomarker strategies require evaluation through large-scale prospective studies, particularly in lung cancer, where numerous competing predictive candidates exist for CPI response.

The emergence of digital twins (DTs) in the arena of anticancer treatment echoes the transformative impact of artificial intelligence in drug development. DTs provide dynamic, accessible platforms that may accurately replicate patient and tumor characteristics. The potential of DTs in clinical investigation is particularly compelling. By comparing data from virtual trials with conventional trial results, medical teams can significantly enhance the reliability of their studies. Moreover, a significant breakthrough in clinical research is the ability of DT to augment patient data during ongoing trials, enabling adaptive trial designs and more robust statistical analyses to be performed even with limited patient populations. The development of DTs faces however several technical and methodological challenges. These include their tendency to produce unreliable predictions, non-factual information, reasoning errors, systematic biases, and a lack of interpretability. Future research in this field should focus on an interdisciplinary approach that brings together experts from diverse fields, including mathematicians, biologists, and physicians. This collaborative strategy promises to unlock new frontiers in personalized cancer treatment and medical methodologies.

PROTACs are bifunctional small molecules that simultaneously bind a target protein and a component of the ubiquitin–proteasome system, thereby inducing selective degradation of the target. They represent a class of compounds capable of achieving the complete elimination of disease-relevant proteins. Molecular glues, by contrast, enhance existing surface complementarity between an E3 ligase and a target protein, promoting its ubiquitination and subsequent degradation. Both approaches are at the forefront of current efforts to overcome the long-standing challenge of undruggable tumor targets. In this context, AI-based strategies offer a powerful means to accelerate the discovery, optimization, and production of highly selective protein binders, streamlining access to potent degraders and maximizing therapeutic potential. These capabilities open new horizons for targeting a wide spectrum of previously inaccessible molecular pathways involved in cancer progression. Altogether, these advances position PROTACs and molecular glues as transformative agents for personalized oncology, particularly within the emerging paradigm of molecular tumor boards, where tailored therapeutic decisions and tumor-adapted drugs could be made rapidly accessible for a given patient.

Shared epitopes create safety and efficacy issues for T-cell immunotherapy. In order to facilitate the monitoring of immune responses and the engineering required to solve this problem, we performed a computational proteome-wide epitope screening to establish the complete atlas of shared epitopes in the human and murine proteomes. Unlike bacterial or viral antigens, self-antigens like tumor-associated antigens (TAAs) frequently contained a high level of shared MHC-II epitopes identical to unintended other self-proteins. Therefore, shared epitopes should be a mandatory and systematic concern in studies using TAA. Noticeably, a subset of TAAs identified in this atlas is free of this drawback. Therefore, this dataset will be essential for immunologists designing cancer vaccines, but also to interpret immunomonitoring studies against self-antigens in oncology and autoimmunity. To facilitate the detection of common epitopes, a web server has been made available at . ### Competing Interest Statement GK is shareholder of Telomium, a company developing cancer vaccines.

The conventional rules for anti-cancer drug development are no longer sufficient given the relatively limited number of patients available for therapeutic trials. It is thus a real challenge to better design trials in the context of new drug approval for anti-cancer treatment. Artificial intelligence (AI)-based in silico trials can incorporate far fewer but more informative patients and could be conducted faster and at a lower cost. AI can be integrated into in silico clinical trials to improve data analysis, modeling and simulation, personalized medicine approaches, trial design optimization, and virtual patient generation. Health authorities are encouraged to thoroughly review the rules for setting up clinical trials, incorporating AI and in silico methodology once they have been appropriately validated. This article also aims to highlight the limits and challenges related to AI and machine learning.

Artificial intelligence (AI) is progressively spreading through the world of health, particularly in the field of oncology. AI offers new, exciting perspectives in drug development as toxicity and efficacy can be predicted from computer-designed active molecular structures. AI-based in silico clinical trials are still at their inception in oncology but their wider use is eagerly awaited as they should markedly reduce durations and costs. Health authorities cannot neglect this new paradigm in drug development and should take the requisite measures to include AI as a new pillar in conducting clinical research in oncology.

The use of companion diagnostics has become a standard in precision oncology in the context of ongoing therapeutic innovation. However, certain limitations make their application imperfect in current practice. This position paper underscores the need to broaden the notion of companion testing, considering the potential of emerging technologies, including computational biology, to overcome these limitations. This wave of progress should impact not only our representation of the analytical tool itself but also the nature of the tumoral sample under analysis (liquid biopsies). The complex inter-relationship between companion test guided-personalized therapy, and health agency policies for new drug agreements will also be discussed.

The past decade has witnessed a revolution in cancer treatment, shifting from conventional drugs (chemotherapies) towards targeted molecular therapies and immune-based therapies, in particular immune-checkpoint inhibitors (ICIs). These immunotherapies release the host’s immune system against the tumor and have shown unprecedented durable remission for patients with cancers that were thought incurable, such as metastatic melanoma, metastatic renal cell carcinoma (RCC), microsatellite instability (MSI) high colorectal cancer and late stages of non-small cell lung cancer (NSCLC). However, about 80% of the patients fail to respond to these immunotherapies and are therefore left with other less effective and potentially toxic treatments. Identifying and understanding the mechanisms that enable cancerous cells to adapt to and eventually overcome therapy can help circumvent resistance and improve treatment. In this review, we describe the recent discoveries on the onco-immunological processes which govern the tumor microenvironment and their impact on the resistance to PD-1/PD-L1 checkpoint blockade.

Supplementary Figure S3 from Disruption of Autophagy at the Maturation Step by the Carcinogen Lindane Is Associated with the Sustained Mitogen-Activated Protein Kinase/Extracellular Signal–Regulated Kinase Activity

AVI file - 5901K, Video 5 (related to Fig. 2G) is a time-lapse recording of a mononucleate a3-/- cell that formed a tetranucleate cell after two rounds of abortive mitosis.

<p>AVI file - 875K, Video 2 (related to Fig. 2D) shows the delayed abscission of an a3-/- cell that remained connected by an intracellular bridge for up to 8 h 45 min before separating.</p>

Nearly fifty million older people suffer from neurodegenerative diseases, including Alzheimer (AD) and Parkinson (PD) disease, a global burden expected to triple by 2050. Such an imminent "neurological pandemic" urges the identification of environmental risk factors that are hopefully avoided to fight the disease. In 2022, strong evidence in mouse models incriminated defective lysosomal acidification and impairment of the autophagy pathway as modifiable risk factors for dementia. To date, the most prescribed lysosomotropic drugs are proton pump inhibitors (PPIs), chloroquine (CQ), and the related hydroxychloroquine (HCQ), which belong to the group of disease-modifying antirheumatic drugs (DMARDs). This commentary aims to open the discussion on the possible mechanisms connecting the long-term prescribing of these drugs to the elderly and the incidence of neurodegenerative diseases.Abbreviations: AD: Alzheimer disease; APP-βCTF: amyloid beta precursor protein-C-terminal fragment; BACE1: beta-secretase 1; BBB: brain blood barrier; CHX: Ca2+/H+ exchanger; CMI: cognitive mild impairment; CQ: chloroquine; DMARD: disease-modifying antirheumatic drugs; GBA1: glucosylceramidase beta 1; HCQ: hydroxychloroquine; HPLC: high-performance liquid chromatography; LAMP: lysosomal associated membrane protein; MAPK/JNK: mitogen-activated protein kinase; MAPT: microtubule associated protein tau; MCOLN1/TRPML1: mucolipin TRP cation channel 1; NFE2L2/NRF2: NFE2 like bZIP transcription factor 2; NRBF2: nuclear receptor binding factor 2; PANTHOS: poisonous flower; PD: Parkinson disease; PIK3C3: phosphatIdylinositol 3-kinase catalytic subunit type 3; PPI: proton pump inhibitor; PSEN1: presenilin 1, RUBCN: rubicon autophagy regulator; RUBCNL: rubicon like autophagy enhancer; SQSTM1: sequestosome 1; TMEM175: transmembrane protein 175; TPCN2: two pore segment channel 2; VATPase: vacuolar-type H+-translocating ATPase; VPS13C: vacuolar protein sorting ortholog 13 homolog C; VPS35: VPS35 retromer complex component; WDFY3: WD repeat and FYVE domain containing 3; ZFYVE1: zinc finger FYVE-type containing 1.