Pancreatic cancer remains a malignancy characterized by an exceptionally high mortality rate, primarily driven by significant intratumoral heterogeneity and a profound immunosuppressive tumor microenvironment. To date, the limited efficacy of current standard-of-care chemotherapy and single-agent immunotherapy underscores an urgent, unmet clinical need for innovative therapeutic strategies. XON11, a new developed polyclonal antibody, targets simultaneously multiple synergistic antigens expressed by pancreatic cancer cells. The aim of these studies was to assess the anti-tumoral efficacy of XON11 and characterize mechanistic impact on cancer stem cell (CSC)-driven tumor heterogeneity, in non-clinical pancreatic cancer models.

Pancreatic cancer heterogeneity was investigated in a panel of five phenotypically diverse human PDAC cell lines (MiaPaca-2, Aspc-1, Capan-1, Panc-1 and BxPC-3). CSC subpopulations were defined and monitored using established markers, including EpCam, CD24, CD44, CD133. We quantified XON11's ability to induce Complement Dependent Cytotoxicity (CDC) and apoptosis over 24- and 72-hour incubation periods, comparing its activity against gemcitabine. The functional capacity of XON11 to impair tumorsphere formation and growth, a key indicator of stemness, was tested over 10 days. In vivo XON11 tolerance and efficacy were evaluated in AsPC-1 and BxPC-3 xenograft mice models after intraperitoneal dosing at 40 mg/kg twice a week or gemcitabine over an 8- to 9-week period.

Flow cytometric analysis confirmed strong cellular heterogeneity among the PDAC cell lines. Mechanistically, XON11 selectively reduced the proportion of the highly aggressive EpCam+ CD24+ CD44+ subpopulation in the BxPC-3 and Panc-1 lines, an effect that was absent following gemcitabine treatment in all tested lines. XON11 exhibited potent and selective anti-proliferative activity across the panel, notably retaining efficacy in cell lines demonstrating resistance to gemcitabine. Crucially, XON11 significantly decreased tumorsphere viability in Aspc-1, Panc-1, and Mia-Paca-2 cells following treatment, achieving complete inhibition of sphere formation at concentrations as low as 33 µg/ml. Translational findings were confirmed in vivo: XON11 demonstrated significant efficacy against tumor growth in both xenograft models. Furthermore, in the BxPC-3 model, XON11 exhibited superior efficacy and tolerance compared to gemcitabine, achieving a Tumor Growth Inhibition (TGI) of 61.9% ± 1.9% versus 34.2% ± 4.1% for gemcitabine.

Based on its potent anti-proliferative and CSC-targeting mechanisms, coupled with its observed superior in vivo efficacy and favorable tolerance profile, XON11 represents a highly promising, novel multimodal therapeutic strategy for overcoming tumor heterogeneity and chemoresistance in recurrent pancreatic cancer.

The era of single-cell analysis has demonstrated the importance of accurate characterization of all organ-specific cell subtypes. In the context of malignancies, this has greatly improved our understanding of cell interactions and reciprocal influence. However, metastatic tissues are often forgotten, even though they are the real cause of cancer-related mortality, especially for colorectal cancer (CRC), whose prognosis is directly related to the presence of liver metastases (lm). One reason for this is the inherent difficulties in analyzing lm-CRC tissues with single-cell resolution: first, the liver is composed of cells that are extremely fragile and poorly recoverable after enzymatic digestion; second, most surgical specimens are from patients who have received chemotherapy and are therefore highly necrotic, fibrotic, and fragile. Due to these limitations, most studies have focused on the immune system characterization, ignoring the role of the other components of a however very rich and complex ecosystem.

To address this gap, we performed one of the first in-depth and representative characterizations of the ecosystem of lm-CRC using single-nuclei RNA sequencing. The global landscape, subpopulation composition, cell-cell interactions, tumor cell inference were analyzed in the entire population and regarding relapse.

We generated more than 26,000 high-quality cells from 9 patients for a total of 30 clusters, including malignant epithelial cells from CRC, immune cells, fibroblasts, and normal liver cells. We found a subset of cancer-associated fibroblasts (CAF) that was positively correlated with early-relapse risk. These CAFs expressed extracellular matrix remodeling and hypoxia features and interacted strongly with the immune, endothelial and epithelial clusters. These preliminary results lay the foundation for an accurate description of a pro-metastatic, immunosuppressive, and chemotherapy-resistant microenvironment that may be the cause of relapse.

Some of the identified populations could serve as relapse-biomarkers and could lead to a reconsideration of therapeutic strategy.

Breast cancer remains the most common cancer in women and the leading cause of death among women, despite recent therapeutic advances. Radiotherapy is a standard treatment, but recurrences persist after irradiation, revealing the existence of radiation-resistant tumour cells such as cancer stem cells (CSCs). These cells have a high capacity for survival and plasticity and are derived from the dedifferentiation of non-stem cancer cells into CSCs known as therapy-induced cancer stem cells (iCSCs). This radio-induced cellular plasticity is a process in which the Hippo pathway, and in particular the LRP4-YAP axis, seems to play a central role. This work aims to explore the precise role of the YAP protein in the radioresistance of breast cancer cells. Inhibiting YAP could make breast cancer cells more sensitive to radiotherapy, thereby reducing the risk of recurrence.

The experiments were performed on the SUM159 cell line, derived from triple-negative human breast carcinoma. Radiosensitivity was tested by genetic (siRNA) and pharmacological (Verteporfin) inhibition of YAP, followed by colony formation assays after irradiation at different doses. The expression of YAP target genes was quantified by RT-qPCR. The efficacy of the YAP inhibitor on the cell line was determined by an MTS assay to calculate the IC50. The proportion of CSCs after exposure to Verteporfin was assessed by an Aldefluor assay (ALDH activity) and by the ability of cells to form spheres under non-adherent conditions. YAP activation after irradiation was monitored by immunolabelling and image analysis.

Genetic inhibition of YAP leads to a significant decrease in colony formation after irradiation, in a dose dependent manner. Pharmacological inhibition of YAP by Verteporfin at its IC50 of 0.8 µM decreases the expression of YAP target genes, reduces the pool of ALDH+ CSCs and abolishes their ability to form spheres, confirming its effectiveness on this cell pool, by targeting YAP. Combination tests with Verteporfin + Irradiation show a synergistic effect on colony formation, even with the addition of low doses of radiotherapy. Inhibition of YAP, whether genetic or pharmacological, radiosensitises breast cancer cells. YAP activation is an early post-irradiation event, supporting its role in the emergence of radioresistant cells.

These results identify YAP as a regulator of radiation-induced cellular plasticity and a promising target for sensitising breast tumours to radiotherapy. However, in vivo validation is still needed before clinical strategies combining radiotherapy and YAP inhibition can be considered.

YAP appears to play a central role in breast cancer radioresistance.

RNA-based therapeutics hold significant promise for precision cancer therapy thanks to their ability to modulate disease-relevant genes. However, effective RNA delivery remains a major challenge, particularly in the context of tumor heterogeneity and the dynamic evolution of the tumor microenvironment. Conventional delivery systems often fail to achieve deep tumor penetration, stability, and efficient cellular uptake, limiting their clinical translation. We report here that self-assembling dendrimer nanosystems can serve as an adaptable RNA delivery platform by hijacking in situ tumor-secreted extracellular vesicles (EVs), thereby enhancing RNA stability, cellular uptake, and propagation throughout heterogeneous and evolving tumors via intrinsic endogenous EV-mediated inter-cellular delivery for effective treatment.

Extracellular vesicles (EVs) were isolated from cells treated with dendrimer/RNA complexes and characterized using TEM and biomarkers. The presence and delivery of RNA within these EVs were evaluated using fluorescence imaging and flow cytometry (FACS) to confirm successful repackaging. EV-mediated propagation of RNA delivery to neighboring cells was analyzed using fluorescently labeled RNA and imaging assays in co-cultured 2D- and 3D cancer cell models. In vivo tumor-xenograft models were employed to assess biodistribution, EV-mediated RNA dissemination, and therapeutic efficacy of RNA-loaded dendrimer nanomicelles.

Following cellular uptake of dendrimer/RNA complexes, RNA was successfully repackaged into tumor-derived EVs, which were isolated and analyzed for their RNA cargo using fluorescence imaging and FACS, confirming effective loading. These EVs mediated further transfer of RNA to neighboring cells, establishing a self-propagating delivery network. In vivo studies demonstrated deep tumor penetration and widespread dissemination of RNA within tumors, correlating with improved therapeutic outcomes.

By combining dendrimer-nanotechnology based nucleic acid delivery with tumor EV trafficking, we developed a smart, self-amplifying RNA delivery platform. This system overcomes key limitations of conventional RNA therapeutics, including instability, poor and inefficiency penetration in heterogeneous tumors. The successful repackaging of RNA into EVs and their propagation to neighboring cells via intrinsic endogenous delivery mechanism demonstrates a natural amplification mechanism, enabling deep tumor penetration and adaptive delivery.

The dendrimer-EV platform protects and propagates RNA effectively, adapts to tumor heterogeneity and evolution, and improves therapeutic outcomes. This strategy represents a versatile and promising avenue for the development of personalized RNA therapeutics.

Tumor recurrence remains a major challenge in cancer treatment, often associated with acquired resistance to conventional chemotherapy. Patient-derived xenografts (PDX) provide a physiologically relevant in vivo model to investigate tumor evolution and the mechanisms underlying therapy resistance. Understanding these mechanisms is critical to identify potential therapeutic vulnerabilities in resistant tumors.

PDX models were analyzed at three stages: treatment-naïve tumors, early recurrence, and advanced resistant recurrence. Single-cell RNA sequencing combined with CITE-seq was performed to characterize transcriptional states at single-cell resolution. InferCNV was applied to predict major genomic alterations. Additionally, an in vivo CRISPR screen was implemented to identify stage-specific molecular dependencies and essential survival pathways.

Transcriptomic analysis revealed that cells from the advanced resistant recurrence exhibit a distinct transcriptional profile, while cells from treatment-naïve and early recurrence stages remain transcriptionally similar. InferCNV analysis showed that the dominant clone in the resistant recurrence was already prevalent in the naïve tumor, indicating that resistance is not strictly driven by genomic selection but rather by non-genetic adaptive mechanisms. Preliminary results from the scRNAseq analysis suggest that specific survival pathways become essential in the advanced resistant stage

These findings demonstrate that tumor resistance evolves primarily through transcriptional and metabolic adaptation under therapeutic pressure, independent of strict clonal selection. The distinction between transient and stabilized resistance highlights the importance of single-cell resolution analyses. Integration of transcriptomic and functional screening approaches provides a powerful framework to uncover stage-specific vulnerabilities that could be therapeutically exploited.

Our study emphasizes that resistance in recurrent tumors is driven by adaptive non-genetic mechanisms rather than exclusive genomic selection. Single-cell analyses combined with in vivo functional screens in PDX models will reveal potential molecular targets and pathways that could be leveraged to overcome or reverse resistance in advanced recurrent tumors, providing a roadmap for the development of precision therapeutic strategies.

Triple-negative breast cancer (TNBC) remains the most aggressive breast cancer subtype with limited treatment options and high recurrence rates. Drug-tolerant persister (DTP) cells, a rare subpopulation that survives chemotherapy through phenotypic plasticity, are increasingly recognized as the seed of relapse. However, systematic approaches to identify molecular vulnerabilities in DTPs and reverse their tolerant state remain lacking, representing a critical gap in developing effective anti-relapse strategies.

We induced a DTP model in SUM159 cells (doxorubicin: docetaxel = 1:1000 dosing), capturing naive, DTP, and relapse states, and performed single-cell RNA sequencing. We adapted the REVERT (REVerse Transitions) framework to infer Boolean regulatory networks and identify attractor states representing stable cellular phenotypes. This systems biology approach enables computational prediction of perturbation targets that can destabilize DTP states, thereby restoring drug sensitivity.

Boolean network inference identified a 20-gene regulatory network governing naive-to-DTP transitions, encompassing cell cycle regulators (E2F1/2/7, TFDP1, BRCA1, RAD21), chromatin modifiers (EZH2), transcriptional regulators (MYC, NFKB1, NFE2L2), and plasticity factors (ZEB1, JUNB, FOXP4). Attractor analysis revealed 12 stable states with basin distributions ranging from 6-13%, suggesting a complex multistable landscape. Key nodes, including E2F family members, MYC, NFKB1, and ZEB1, exhibited high state-switching variability across attractors, marking them as potential intervention targets.

This study provides a systems-level map of regulatory vulnerabilities underlying DTP formation and plasticity in TNBC. The identified network nodes represent candidate targets for experimental validation to develop combination strategies that collapse persister states and prevent relapse. Our REVERT-based approach offers a generalizable framework for rational therapeutic target discovery in chemoresistant cancers.

The REVERT-based approach offers a generalizable framework for rational therapeutic target discovery in TNBC drug-tolerant persister cells.

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most aggressive malignancies, with a median survival of approximately six months and a five-year survival rate of only 5–7%. Post-translational modifications (PTMs), such as ubiquitination, play key roles in protein function regulation, yet their involvement in PDAC chemoresistance remains poorly understood. Mass spectrometry analysis of 60 PDAC samples identified 37 ubiquitination profiles associated with resistance to gemcitabine, 5-FU, oxaliplatin, and irinotecan. We hypothesized that these ubiquitination profiles could serve as theranostic markers and molecular targets in PDAC and aimed to validate their functional relevance.

Seven candidate proteins were selected based on their biological function and reported association with chemoresistance. The type of ubiquitination (mono-, multi- or polyubiquitination) was characterized using nickel pull-down, immunoprecipitation, and western blot analyses. The correlation between ubiquitination and drug resistance was assessed by PLA on TMAs derived from PDX models sensitive or resistant to chemotherapy. Site-directed mutagenesis (lysine-to-arginine substitution) was used to inactivate specific ubiquitination sites, and functional effects were evaluated by overexpression in primary PDAC cell lines.

PSMD2, RPS20, and SLC3A2 were validated as potential theranostic markers. Their ubiquitination was increased in resistant tumor cells, correlating positively with chemoresistance for PSMD2 and RPS20, and negatively for SLC3A2. Inactivation of two critical ubiquitination sites, including RPS2 K58R and CDC42 K133R, partially restored chemosensitivity, highlighting a functional role of ubiquitination in maintaining the resistant phenotype

The identified ubiquitination events appear to modulate protein stability and signaling pathways essential for maintaining chemoresistance in PDAC. The functional rescue observed after site-specific ubiquitination loss suggests that these PTMs actively regulate cellular stress responses and survival mechanisms, positioning them as key drivers of resistance biology.

These findings reveal novel resistance mechanisms involving ubiquitination and support a theranostic and personalized treatment approach in PDAC through the targeting of specific PTMs

Breast cancer patients can benefit from different treatment options depending on the stage and type of tumor. Despite similar indications, not all tumors respond equally well to a particular treatment, and patients who respond poorly are at higher risk of early recurrence.Therefore, there is an urgent need to develop tools to identify the appropriate “tumor/treatment response” pair early after treatment initiation. The search for easily accessible and sensitive biomarkers of early response to cancer therapies is one of the promises of liquid biopsy.

This is the focus of the Neo-R clinical trial, in which blood samples were collected from 70 patients with early breast cancer before, during, and after neoadjuvant chemotherapy. Assuming that the number of circulating tumor cells (CTCs) remains stable or increases at all time points in non-responsive patients, we isolated these cells using size-based and low-deformability methods. We then counted and characterized them at the cytological, phenotypic, and molecular levels.

Surprisingly, in addition to “classical” CTCs, we also observed isolated subsets of atypical circulating cells (aCCs). In the literature, these are anecdotally referred to as “giant cells,” and two hypotheses have been proposed, but not yet tested in detail, regarding their possible origin: aCCs could be either cancer-associated macrophage-like cells or megakaryocytes. Interestingly, these aCCs occur more frequently in patients who do not respond to neoadjuvant treatment.

Using ex vivo (multiplex immunofluorescence), molecular (single-cell RNA sequencing), and in silico approaches, my work aims to provide new insights into these aCCs with the goal of elucidating their role during the metastatic cascade.

Despite aggressive and multimodal treatments, Glioblastoma (GBM) remains the deadliest primary brain tumor with a median overall survival of approximately 18 months and a 5-year survival rate below 5%. Multiple mechanisms contribute to therapeutic failure, including the great intra-tumoral heterogeneity of GBM. Three major molecular states have been described, namely the mesenchymal-like (MES), the astrocyte-like (AC) and the oligodendrocyte progenitor (OPC) states, each of them displaying either stem-like (GSC) or differentiated-like (DGC) features. Among these subpopulations, MES state and GSC cells are considered the most aggressive and seems associated with GBM recurrence. In this context, we developed an immunotherapy approach based on engineered Vδ2 T lymphocytes expressing a Chimeric Antigen Receptor (CAR) targeting the ganglioside GD2 or its O-acetylated form OGD2, with the aim of targeting all GBM subpopulations.

To address GBM heterogeneity, we used human GBM cell lines cultured under GSC- or DGC-promoting conditions, as well as primary GBM cultures representing distinct molecular states. GD2 and OGD2 expression were assessed by flow cytometry. Vδ2 CAR-T cell activation was evaluated by degranulation assay and tumor cell killing, using flow cytometry and live-cell videomicroscopy, respectively.

We first confirmed that Vδ2 T cells recognized and killed MES GBM cells through innate immunoreactivity involving Vδ2 TCR recognition of butyrophilins. Depending on the cell line, innate recognition was also observed with some GSC and DGC cells. We then demonstrated that all GSC cells, as well as AC/OPC cells, express significant levels of GD2, and to a lesser extent, OGD2, allowing their recognition and killing by Vδ2-CAR T cells targeting either ganglioside. Specific CAR-mediated recognition was supported by the loss of Vδ2-CAR T cell cytotoxicity upon abrogation of GD2 or OGD2 expression by genetic silencing of the respective biosynthetic enzymes. In contrast, DGC cells were not efficiently recognized by either innate nor CAR-mediated targeting.

Our results indicate that innate Vδ2 T cell cytotoxicity combined with CAR-mediated ganglioside targeting overcomes most of GBM heterogeneity. Importantly, aggressive subpopulations as MES cells and GSC cells subpopulation were efficiently eliminated. However, our strategy remains ineffective against some DGC cells, highlighting a potential niche for relapse and the need for complementary approaches.

In conclusion, our results support the use of Vδ2-CAR T cells as a promising approach to improve immunotherapy efficiency in GBM.

Due to their variable expression during pancreatic carcinogenesis, cadherins could serve as markers for pancreatic ductal adenocarcinoma (PDAC). A poorly studied cadherin named CDHX for confidentiality, not detected in most healthy tissues is expressed early in pancreatic ductal cancer. High level expression of this molecule is associated with reduced patient survival. CDHX may be used as a tool for prognostic, diagnostic, and therapeutic of PDAC. CDHX could drive gene dysregulation and disrupt signaling pathways, contributing to tumor invasion and promoting cancer aggressiveness.

CDHX expression was forced in the pancreatic cancer cell lines PANC-1 and MIA PaCa-2. Aggregation assays were performed to assess the adhesive potential of CDHX. Invasion and migration assays were performed in order to determine the impact of CDHX on PDAC aggressiveness. Finally, proteomic analysis was conducted to identify CDHX-interacting proteins, potentially involved in cell adhesion and invasion.

CDHX expression slightly alters the adhesive capabilities of cells. However, it affects their invasive properties by promoting the invasion of isolated cells. Expression of several genes involved in cell invasion is deregulated upon CDHX expression potentially explaining the observed changes in invasion.

Further validation in primary cultures and in vivo models is planned. The study indicates novel stage-specific signaling pathways and biomarkers, which could help define a molecular signature of tumor invasion and guide development of CDHX-targeted therapies.

These findings suggest that CDHX may contribute to PDAC aggressiveness and represent a potential therapeutic target.

The molecular heterogeneity of hepatocellular carcinoma (HCC) and the lack of oncogenic addictions are among the features underlying unsatisfactory treatment options in the clinic.

Transcriptomic data from distinct cancer patient cohorts were bioinformatically analysed to assess gene expression and explore pathway enrichments in different groups. A targeted delivery system was engineered to silence the ADAMTSL5 oncogene and its stability was determined under biological conditions. Its effects on HCC cells, tumoroids, and xenografts were evaluated using a range of approaches, including RT-qPCR, immunostaining, flow cytometry, and viability assays, both alone and in combination with anticancer agents. Tumour microarrays were used to assess ADAMTSL5 and NCL protein levels.

We report that approximately half of HCC patients exhibit simultaneous upregulation of two oncogenes, ADAMTSL5 and Nucleolin (NCL). This molecular signature also characterises large proportions of patients with other cancer types. As increased NCL levels result in its localisation to the membrane of cancer cells, we engineered a targeted delivery system, named aptAdamtsl5, based on the conjugation of a nucleic acid aptamer targeting NCL with a shRNA sequence targeting ADAMTSL5. We show that aptAdamtsl5 selectively targets ADAMTSL5 in cancer cells expressing NCL at the plasma membrane, leading to marked cellular and molecular changes and reduced proliferative capacity. We further demonstrate that aptAdamtsl5 sensitises cancer cells to receptor tyrosine kinase inhibitors, to which the cells are otherwise resistant. Moreover, aptAdamtsl5 confers druggable AXL addiction to liver cancer. The increased stability of aptAdamtsl5 in biological samples enables evaluation of its therapeutic effectiveness on mouse and patient-derived tumoroids as well as in HCC in vivo models. A subgroup of HCC patients potentially responsive to aptAdamtsl5 can be identified based on elevated ADAMTSL5 and NCL levels.

Our results identify a new subgroup of cancer patients with elevated ADAMTSL5 and NCL, who could benefit from the therapeutic potential of aptAdamtsl5 to overcome resistance to current clinical treatments.

Histone deacetylases (HDACs) are epigenetic regulators frequently altered in cancer. The relevance of their targeting is highlighted by the use of HDAC inhibitors (HDACi) for anticancer treatment in preclinical and clinical investigations for solid tumours. Among complex cancer types, hepatocellular carcinoma (HCC), the most common type of liver cancer, is characterised by aggressiveness and resistance to available therapies, and displays several epigenetic alterations whose functional relevance has been reported in recent studies. We hypothesised that epigenetic drugs targeting HDACs may confer vulnerability to HCC, including to clinically relevant agents as receptor tyrosine kinase inhibitors (RTKi).

We bioinformatically evaluated the relevance of class-I HDACs in HCC patients using available datasets. The effects of romidepsin, a class-I HDACi, were assessed in HCC cells, patient-derived tumoroids, and HCC mouse models. Molecular alterations were determined using proteomics, biochemistry, lipidomics, and immunostaining.

We showed that overexpression of HDAC1 and HDAC2 occurs in HCC patients across eight cohorts and is correlated with decreased overall survival. We documented that romidepsin perturbs cell cycle and survival signals in HCC cells. Romidepsin alters the expression of lipid metabolism regulators, reshaping the composition of distinct lipid species. Additionally, romidepsin affects the mitotic spindle machinery, leading to monopolar spindle formation and cell cycle arrest. These alterations render HCC cells vulnerable, conferring dependency on RTK support. Combined treatment of HCC cells with romidepsin and the RTKi cabozantinib (RomiCabo) converts the cytostatic effect of romidepsin, into apoptosis. We reported the therapeutic effectiveness of RomiCabo treatment on HCC-patient derived tumoroids and on the Alb-R26Met mouse model, which recapitulates HCC resistance and heterogeneity. Furthermore, we documented that RomiCabo leads to immune remodelling in the tumour microenvironment, conferring an immune-stimulatory profile.

Our findings highlight the intricate crosstalk between epigenetics, metabolism, and immune response in cancer. The broad action of romidepsin on distinct cellular functions underscores its therapeutic potential for HCC treatment.

Glioblastoma is the most aggressive primary brain tumor in adults. The overexpression of anti-apoptotic mechanisms and the presence of cancer stem cells (CSCs) impair treatments efficiency. Among the family of inhibitor of apoptosis proteins (IAPs), we have shown that ML-IAP is a marker of poor prognosis in glioblastoma patients and is overexpressed in CSCs (Tchoghandjian, 2016). The anti-apoptotic activity of ML-IAP relies on caspase inhibition and the sequestration of endogenous SMAC. However, its precise mechanism of action has not yet been fully characterized. To target IAPs, inhibitors called SMAC mimetics (SMs) have been developed. We have demonstrated that GDC-0152, a SM with high affinity for ML-IAP, enhances apoptosis and improves survival in glioblastoma mouse models. It also induces a remodelling of the tumor microenvironment (Snacel-Fazy,2024). In order to optimize its efficacy for a potential therapeutic application, we aim to determine whether its effects are primarily due to a specific interaction with IAPs or if other protein targets are also involved. The objective of this study is to better characterize the anti-apoptotic mechanism of ML-IAP and the mode of action of its inhibitors.

To explore ML-IAP interactome we transfected a Flag-ML-IAP cDNA sequence into a CSC line and we performed an anti-Flag immunoprecipitation followed by proteomic analysis via mass spectrometry. To identify additional potential protein targets of GDC-0152 and its oral derivative GDC-0917 (GDCs), a Thermal Proteome Profiling (TPP) experiment was conducted. This technique is based on the principle that ligand binding to its target can alter the protein’s denaturation temperature. Cell lysates treated with GDCs or vehicle were subjected to a temperature gradient and then ultracentrifuged. The supernatant, containing folded proteins, was collected and analysed by mass spectrometry to identify GDCs protein targets.

Flag-ML-IAP overexpression increased proliferation and clonogenicity in a CSC line but did not alter the response to GDCs. Proteomic analysis of the immunoprecipitated proteins identified potential ML-IAP partners. Among them, BIRC2, HADHB, HTR2, and PGAM5 are already known as partners of other IAP family members, validating our experimental approach. CPT1A and DCLK2 have been identified as potential new partners. TPP revealed 22 proteins with significant changes, suggesting that they may constitute potential targets of GDCs treatment. Their functional implications are now under investigation.

By identifying both the ML-IAP interactome and the proteins potentially modulated by the GDC compounds, this study provides a foundation to improve the development of therapeutic strategies targeting ML-IAP.

Homeostasis preserves tissue function through the coordinated regulation of cells and their interactions. Although essential for physiological balance, its plasticity can favour deregulated (epi)genetic and signalling programmes that initiate and sustain cancer. These alterations remodel the tissue environment to form “cancer communities” composed of malignant, immune, endothelial, and stromal cells, whose reciprocal crosstalk drives heterogeneity and adaptation. At the centre lies the tumour microenvironment and its extracellular matrix (ECM), which controls tissue composition and mechanics. Cancer and stromal cells reshape the ECM through proteases and matricellular proteins that influence signalling, proliferation, invasion, and immune and vascular responses. Such ECM-centred dynamics define key therapeutic vulnerabilities. Hepatocellular carcinoma (HCC), a leading cause of cancer mortality, exemplifies this complexity. HCC displays marked molecular heterogeneity sustained by diverse (epi)genetic changes. ADAMTSL5, a matricellular protein overexpressed in most HCCs, correlates with poor outcomes. Its inhibition suppresses oncogenic pathways, tumour aggressiveness, and drug resistance, while modulating tumour immune interactions. We hypothesise that the matricellular protein ADAMTSL5 ensures the functionality of cancer communities. We have generated engineered agents, such as nanobodies for ADAMTSL5 detection and targeting. We hypothesise that agents targeting ADAMTSL5 perturb the functionality of cancer communities.

Recombinant ADAMTSL5 was used to generate llama nanobodies. Their ability to bind ADAMTSL5 is assessed by ELISA, western blot, immunostaining, and immunohistochemistry. Their ability to block ADAMTSL5 function is explored by testing their effects on HCC cells using viability assays.

Among several nanobodies against ADAMTSL5 that we have generated, seven targeting the C-terminus and four targeting the N-terminus were selected for further characterisation. All detect ADAMTSL5 on cancer cells. Two exhibit blocking functions, shown by altered cell morphology, exemplified by changes in stress fibre formation. These blocking nanobodies impair cancer cell viability when used alone and induce AXL dependency, as demonstrated by increased sensitivity to the AXL inhibitor bemcentinib.

These ongoing studies underline the diagnostic and therapeutic relevance of these nanobodies against ADAMTSL5. Furthermore, they represent unique tools to uncover how ADAMTSL5 functions to regulate cancer communities.

Hepatocellular carcinoma (HCC) is a leading cause of cancer mortality, characterised by extensive heterogeneity at the (epi)genetic and immune microenvironment levels. This heterogeneity contributes to HCC resistance and the limited benefit of current therapies. Previous work from our laboratory has identified key vulnerabilities in HCC following blockade of signals such as MEK, BCL-XL, HDACs, and receptor tyrosine kinases (RTKs). We designed an experimental setting to assess whether combined inhibition of HDAC1/2 (by romidepsin) and anti‑apoptotic signals (by navitoclax or venetoclax) elicits therapeutic effects on HCC cells while specifically remodelling the tumour immune microenvironment.

Viability of HCC cells and Alb-R26Met tumoroids was assessed following romidepsin plus navitoclax or venetoclax (RomiNavi/Vene) treatments. Mechanistic effects were analysed by Western blot. Drug-induced immune remodelling was investigated by spectral cytometry.

We found that the viability of heterogeneous human HCC lines, primary murine Alb‑R26Met HCC cells and tumoroids is severely affected by RomiNavi and RomiVene at relatively low doses. Preliminary mechanistic studies reveal that romidepsin increases p21 while decreasing MYC, Cyclin‑D1, and CDK1 phosphorylation, consistent with disrupted cell cycle progression. Romidepsin also reduces multiple anti‑apoptotic proteins (including BCL‑XL, Survivin, XIAP), while upregulating MCL1, likely as an attempt to counteract unbalanced anti-apoptotic signals. Notably, DNA damage signalling, monitored by γH2AX, is robustly induced only in RomiNavi and RomiVene conditions, indicating that only these combinations induce strong DNA damage. Further mechanistic aspects are currently being explored through -omics. Spectral cytometry revealed that each regimen induces distinct remodelling of immune cell type populations in control livers. RomiNavi prominently increases monocytes, while all treatments reduce B cells, most markedly under RomiNavi or RomiVene. Furthermore, both combinations expand cDC1s and decrease pDCs, whereas only RomiVene elevates cDC2s. Macrophages are decreased in navitoclax‑ and venetoclax‑treated groups, whereas CD44 and PD1 levels on CD90.2⁺ cells remain stable. The remodelling process triggered by RomiNavi/Vene in the tumour microenvironment is ongoing.

Our findings reveal a dual mechanism elicited by RomiNavi/Vene: direct induction of tumour cell cytotoxicity and reshaping of immune cell types. The latter offers the possibility to pair RomiNavi/Vene with optimal immunotherapies.

RomiNavi/Vene elicit dual antitumor effects: direct cytotoxicity via cell cycle arrest, apoptosis, and DNA damage in heterogeneous HCC models, plus distinct liver immune remodeling (e.g., cDC1 expansion, macrophage reduction). Tumor microenvironment validation will guide immunotherapy pairings.

The Wnt/PCP (Planar Cell Polarity) pathway is highly deregulated in TNBC, thereby promoting tumor proliferation and migration. We have identified the serine-threonine kinase MINK1, as a key regulator of this pathway and shown that MINK1 hyperactivates this pathway by phosphorylating PRICKLE1 and LL5β, two pro-metastatic proteins associated with poor prognosis in TNBC. My project aims to characterize the functions and molecular organization of MINK1 and its pro-metastatic complex, as well as to assess its potential as a therapeutic target in TNBC. To do this, an inhibitor of MINK1’s catalytic activity has been developed. This inhibitor is referred to here as “Compound2”

The efficiency of Compound2 was validated thanks to functional assays, including western blots, immunoprecipitations, and cell migration assays performed on TNBC cell lines. Its effect was also confirmed in 3D cellular models, including spheroids formed from TNBC cell lines and organoids generated from primary breast tumors xenografted into humanized mammary glands (PDX). Moreover, proteomic and phosphoproteomic approaches were also carried out to validate the selectivity of Compound2 for MINK1.

Compound2 belongs to the family of ATP analogs, therefore it binds to the kinase domain of MINK1, specifically within the ATP-binding pocket. The crystal structure of the MINK1 kinase domain bound to Compound2 has been resolved, providing invaluable insights into the mechanism of inhibition. Functional and signaling assays have demonstrated that Compound2 inhibits the phosphorylation of MINK1 substrates (PRICKLE1, LL5β and AKT) and significantly reduces the migration of TNBC cells. Its effectiveness was also confirmed in 3D cellular models. A strong and significant correlation was observed between MINK1 expression levels and the response of TNBC cell lines to Compound2. Moreover, the selectivity of Compound2 for MINK1 was confirmed by proteomic and phosphoproteomic analyses. These approaches have also identified pathways regulated by MINK1, thereby opening the door to new combinatorial therapeutic strategies.

The main objective focuses on the generation of MINK1 mutants able to make MINK1 constitutively active or inactive, to better understand its function and role in the tumorigenic process. Other mutations will be introduced within the MINK1 sequence to further confirm the selectivity of Compound2. Another objective is to investigate combination therapy strategies with Compound2 to enhance its efficiency, given MINK1’s involvement in therapeutic resistance.

This project demonstrates the therapeutic potential of targeting MINK1 in TNBC.

Ovarian cancers (OC) represent the leading cause of death from gynecological cancers worldwide. First-line treatment combines surgery with carboplatin-based chemotherapy, and PARP inhibitors (PARPi) for eligible patients. PARPi efficacy rely on defects in the Homologous Recombination DNA repair pathway (HRD), present in 50% of OC cases. Double stranded DNA breaks can be repaired by either Non-Homologous End Joining (NHEJ) or Homologous Recombination (HR) pathways. While HR is considered faithful, NHEJ is error-prone and can lead to cell death. Thus, HRD increases tumor sensitivity to DNA-damaging agents. Our project aims to induce an HRD-like phenotype by inhibiting UBE2N, a key HR mediator, to sensitize ovarian tumors to carboplatin and PARPi.

We used an ovarian cancer cell line and PDTO (Patient-Derived Tumor Organoids) models generated from tumor of patients grown in 3D in extracellular matrix. UBE2N was inhibited either pharmacologically (NSC697923) or genetically (Cas9 knockout). HR proficiency following UBE2N inhibition (UBE2Ni) was assessed using multiple complementary approaches: comet assay (DNA damage), RECAP test (HR activation), DR-GFP reporter assay (effective HR DNA repair), and micronuclei quantification (genomic instability). In parallel, we investigated a potential correlation between the increase in DNA damage and/or the decrease in DNA repair by HR induced by UBE2N inhibition and its ability to sensitize cells to the action of DNA-damaging therapies (carboplatin, PARPi). Sensitization was assessed by monitoring cell morphology using the IncuCyte S3 and CellDiscoverer 7 real-time imaging systems, as well as through viability assays and colony-forming assays.

UBE2N inhibition, through NSC697923 or knockout, increased micronuclei and DNA damage while decreasing HR DNA repair. In addition, UBE2Ni sensitized an ovarian cancer cell line and multiple HR-proficient PDTO models to PARPi.

These findings support the therapeutic potential of UBE2Ni. Further validation in additional PDTO models is ongoing and the effects on DNA repair will be more finely characterized.

Furthermore, this strategy could be extended to other cancer locations where DNA-damaging therapies are currently used such as pancreatic cancer (oxaliplatin-based chemotherapy and PARPi), breast cancer (carboplatin, PARPi), and head and neck cancers (radiotherapy and cisplatin), broadening its relevance in oncology.

Colorectal cancer (CRC) is among the most common cancers. When it progresses to the metastatic stage (mCRC), especially with liver involvement, it becomes particularly difficult to treat. Perioperative therapies are often nonspecific, highly toxic, and frequently fail to eliminate all cancer cells. However, a new class of drugs, called antibody-drug conjugates, offers hope for more effective treatment of aggressive tumors. Their effectiveness depends on the expression of the ADC target in cancer tissues. These treatments have already improved outcomes for certain solid tumors, but in mCRC, their potential remains largely unexplored.

We analyzed the transcriptomic expression of 56 potential ADC targets, whose corresponding ADCs are used in the clinic or are currently in development. We examined their expression in bulk and at the single-cell level (by single nucleus RNAseq) in precancerous stages (inflammatory bowel diseases, polyps), primary CRC, and mCRC (liver and lung). ADC target expression in each type was compared with their expression in normal colon tissue, as well as in other normal tissues. We then assessed their expression at the single-cell level in our cohort of mCRC (B-Org cohort NCT05384184). To confirm the results at the protein level, we performed immunostaining for the ADC targets of interest on samples from the B-Org cohort.

Although transcriptomic expression profiles varied across the tissues analyzed, some genes, such as CD276, NECTIN4, PTK7, LGR5 and MET, emerged as promising targets. Their marked overexpression in mCRC highlights them as new potential therapeutic opportunities, which have not always been explored in clinical trials. We then refined our analysis using single-“cell” data, revealing interesting expression patterns of ADC targets. Some ADC targets showed clear differential expression compared to normal epithelial colon cells, positioning them as key candidates; others were unexpectedly expressed in the stromal compartment rather than in malignant cells, introducing a new microenvironment-based tumor-targeting strategy. Immunostaining of selected ADC targets on mCRC samples confirmed these expression patterns.

Our assessment of ADC targets landscape based on in-silico analysis allows us to identify ADC targets of potential interest for treating mCRC, a cancer urgently in need of innovative and less toxic therapies. The next step will be to test their efficacy in preclinical models of mCRC using our biobank of patient-derived mCRC organoids (n>50) and patient-derived liver organoids. This screening strategy across different expression levels and tissue types will build the preclinical rationale for selecting ADCs with strong efficacy and reduced toxicity in mCRC.

Pancreatic ductal adenocarcinoma (PDAC) is characterized by an abundant desmoplastic stroma enriched in cancer-associated fibroblasts (CAFs), which profoundly influence tumor growth, extracellular matrix remodeling, immune exclusion, and therapeutic resistance. However, the functional heterogeneity of CAF subtypes and their dynamic interactions with tumor cells remain insufficiently understood due to limitations of conventional 2D culture systems that exclude key components of the tumor microenvironment. Developing in vitro co-culture models for high-throughput ex vivo drug testing is therefore essential to identify actionable therapeutic targets in a physiologically relevant context and to dissect stromal-induced tumor plasticity.

We established a 3D co-culture system combining patient-derived PDAC organoids with primary CAF populations isolated from a matched tumor specimen. Organoids and CAFs were embedded in defined extracellular matrix conditions and maintained in optimized media supporting both cell types. To uncover actionable vulnerabilities, publicly available CRISPR screen datasets were mined, yielding an initial list of 181 druggable targets. A compound library is being assembled and will be tested using acoustic dispensing on fluorescent co-cultures, while drug effects on both cell populations will be quantified through a high-content imaging workflow. For functional validation, organoid models were engineered to express GFP-tagged RNA barcodes, and CAFs were labeled with RFP via lentiviral transgenesis, enabling dynamic monitoring of cell states and plasticity under therapeutic pressure.

Co-culture of PDAC organoids with primary CAFs was successfully established using media compatible with the growth of both components. Initial optimization steps confirmed the feasibility of maintaining fluorescently labeled tumor cells and CAFs in a shared 3D matrix environment. Mining of DepMap CRISPR datasets and the Pharos portal identified 1,048 vulnerabilities, including 181 prioritized candidates, which are currently being integrated into a focused compound library for high-throughput testing. The generation of GFP-barcoded organoids and RFP-labeled CAFs was achieved and will enable forthcoming analyses of cell-state dynamics and therapeutic responses within the co-culture system. Additional functional and transcriptomic characterizations are ongoing.

This advanced co-culture platform provides a powerful tool to identify actionable vulnerabilities in PDAC while integrating the complexity of stromal-tumor interactions. The model enables long-term evaluation of cellular state plasticity driven by CAFs and offers a physiologically relevant framework for testing therapeutic strategies targeting both tumor cells and their microenvironment.

Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive disease with few therapeutic options. Antibody-Drug Conjugates (ADCs) are emerging anticancer agents composed of a monoclonal antibody targeting specific tumor antigens combined with high-potency cytotoxic payload. Trastuzumab Deruxtecan (T-DXd) is an ADC consisting of an anti-HER2 antibody linked to a topoisomerase-I inhibitor (DXd) active in HER2 high and low expressing solid tumors. In PDAC patients, HER2 is expressed in up to 61% of cases, playing a prognostic role. Given the broad expression of HER2 in PDAC samples, we hypothesize that T-DXd, alone and in combination with synergistic compounds, could represent an effective treatment for PDAC patients.

Six PDAC cell lines were analyzed (CAPAN1, PANC1005, ASPC1, PANC1, HS766T, PSN1) according to ERBB2/HER2 expression level by RNA-seq, IF, IHC, WB and FC. Whole exome sequencing was also performed. Each cell line was treated with single agents and combinations of T-DXd, PARP and KRAS inhibitors (PARPi; KRASi) and cell viability was measured by MTT assay. Changes in DNA damage markers and KRAS signalling pathway were investigated by WB (pH2AX/totH2AX, pCHK1/totCHK1, pERK/totERK, pAKT/totAKT, p-p70/tot-p70, p4EBP1/tot4EBP1). Cell cycle distribution was assessed by FC (PI assay).

According to ERBB2 expression level CAPAN1, ASPC1, and PANC1005 were labeled as HER2high cell lines while HS766T, PANC1, and PSN1 as HER2low. KRAS mutations were found in 6/6 lines (KRASmut), BRCA1-2 in 4/6 (BRCA1/2mut). CAPAN1 showed the highest level of total HER2 with highest membrane localization. Accordingly, CAPAN1 (HER2high-KRASmut-BRCA2mut) showed the highest sensitivity to T-DXd single agent (IC50 10 µg/mL), improved by the addition of gemcitabine and cisplatin chemotherapy. Similarly, in HER2high-BRCA1/2mut cells the association of PARPi to T-DXd reduced cell viability increasing DNA damage (pH2AX and pCHK1) and G2-phase cell cycle arrest as compared to monotherapies. In HER2high-KRASmut cells the addition of KRASi to T-DXd promoted cytotoxic activity enhancing the AKT pathway blockage, as showed by the decrease of pERK, pAKT, p-p70, and p4EBP1.

T-DXd based combinations promoted antitumoral effect in HER2high cells. To validate these findings, treatments will be tested in patient-derived organoids (PDO) and primary cells (PDC), and in an orthotopic mouse model. Additionally, transcriptomic analysis of PDO and PDC will be performed to generate response signatures to T-DXd and selected combinations, based on each drug response profiles.

Preliminary findings suggest that T-DXd based combinations are effective in HER2-positive PDAC in vitro models, supporting further validation in preclinical models.

The EPITHOR (“EPIdémiologie en chirurgie THORacique”) database, created in 2003, is a national thoracic surgery registry including more than 250,000 patients from over 100 centers. The LUng CAncer – Prevention and Interception (LUCA-pi) RHU, launched in 2024, aims to develop biological biomarkers and AI-based tools to improve lung cancer prevention and early detection. In this context, the LUCA-pi RETRO database was defined as a subset of EPITHOR, including patients who underwent lung cancer surgery at Hôpital Nord (Marseille) between 2011 and 2024. Despite curative-intent surgery, 20–30% of patients experience cancer recurrence. Predicting recurrence could guide follow-up strategies and (neo-)adjuvant treatment, associated with a lot of adverse effects and a high cost.

As of December 2025, complete clinical, pathological, imaging, and follow-up data are available for 1,938 patients. Only patients with non-small cell lung cancer (NSCLC) were included, with 279 post-preprocess variables. Three cohorts were defined: all patients, stage I patients, and stage II patients. Disease-free survival (DFS), defined as time to recurrence, was the main outcome, with loss to follow-up right-censored. A GitLab-based continuous integration / continuous deployment (CI/CD) data science pipeline was implemented for continuous integration of patient batches. It comprises: preprocessing, statistical analyses (Cox regressions), machine learning predictive analyses and deployment of the reports to a website. An interactive dashboard was also developed and deployed to enable easy exploration of all statistical results.

For all stages, post-operative TNM (Tumor, Nodule, Metastasis) stage was the strongest predictor (C-index = 0.742, HR = 7.08, p < 0.00001). Tumor size and staging variables were consistently significant. Angioinvasion also showed strong predictive value (C-index = 0.667; HR = 3.16; p < 0.00001). In stage I patients, THORACOSCORE was the best predictor (C-index = 0.644; HR = 1.33; p < 0.05), followed by forced expiratory volume in one second (FEV1) (C-index = 0.641; HR = 0.673; p < 0.005). In stage II patients, hospitalization duration was most predictive (C-index = 0.625; HR = 1.49; p < 0.01).

So far, these results confirm the predictive ability of some known variables but the main strength of this dataset lies in enabling deeper investigation. We are currently integrating machine learning methods to identify a predictive signature from these variables.

Future work will incorporate a mechanistic model of tumor growth and metastatic spread to anticipate distant relapse. The long-term goal is to build a practical clinical tool to support decisions on adjuvant therapy and recurrence prevention.

Melanoma is the most aggressive skin cancer, characterized by remarkable cancer cell plasticity, contributing to intra-tumoral heterogeneity and therapeutic resistance. NRAS-mutant melanoma remains a clinical problem, particularly in patients who do not respond to immunotherapies. As a second-line option, MEK inhibitors as single agents fail to provide a significant overall survival benefit. Therefore there is an unmet need for new therapeutic strategies to improve the management of NRAS-mutant melanoma. Here we assessed in vitro and in vivo the response of NRAS-mutant melanoma cells to RMC-6236, a novel non-covalent inhibitor of both oncogenic and wild type RAS isoforms currently undergoing clinical investigation in various cancers.

Loss-of-function approaches using RMC-6236, IAG933 (a YAP-TEAD interaction inhibitor), or siRNAs were employed to evaluate the impact of NRAS inhibition on phenotypic adaptation (RNA-seq, RT-qPCR, western blot analyses) as well as cell proliferation and survival (colony formation assay, flow cytometry) in human and murine NRAS-mutant cell line models. A murine melanoma model using MaNRAS cells grafted into syngeneic C57BL/6 mice was used to assess the effect of RMC-6236 on tumor growth and mouse survival.

Our transcriptomic and proteomic analyses revealed that the anti-proliferative effect of RMC-6236 on NRAS-mutant melanoma cell lines is characterized by a phenotypic transition towards a less differentiated state, with increased expression of mesenchymal and extracellular matrix remodeling markers, along with the activation of a YAP-driven transcriptional signature and focal adhesion kinase (FAK) signaling. In vivo RMC-6236 slowed tumor growth and improved mouse survival. Melanoma cells treated with RMC-6236 in vivo exhibited reduced pigmentation and expressed mesenchymal and neural crest stem cell markers and YAP-target genes. The combination of RMC-6236 and IAG933 synergistically reduced proliferation, prevented phenotypic transition, and induced apoptosis in NRAS-mutant cells. These findings suggest that YAP-TEAD pathway inhibition by IAG933 targets the adaptive response induced by RMC-6236 and enhances treatment efficacy in vitro and in vivo.

RMC-6236 inhibits NRAS-mutant melanoma growth but induces an adaptive phenotypic transition toward a less differentiated, YAP-driven mesenchymal state. YAP–TEAD inhibition with IAG933 blocks this adaptive response, prevents dedifferentiation, and synergistically enhances apoptosis and antitumor efficacy in vitro and in vivo.

NRAS inhibition in melanoma cells induces a mesenchymal phenotypic transition linked to YAP pathway activation. YAP/TEAD inhibition can overcome resistance to NRAS inhibition by preventing adaptive phenotype switching and inducing tumor cell death. This work provides a scientific rationale for treating NRAS-mutant melanomas with a combination of RAS and YAP-TEAD inhibitors.

Triple-negative breast cancer (TNBC) accounts for 10-15% of all breast cancer cases. It is a highly aggressive and heterogeneous cancer characterized by the absence of Hormone Receptors (HR) and Human Epidermal Growth Factor Receptor 2 (HER2). These molecular characteristics limit the treatment options for TNBC patients resulting in a poor prognosis and frequently relapses. Therefore, the identification of new vulnerabilities to overcome chemotherapy resistance are urgently needed. Nuclear Factor Erythroid 2-related factor 2 (NRF2) is a transcription factor that plays a central role in the response to oxidative stress. It is frequently overactivated in cancer, including TNBC, and it is associated with resistance to therapy allowing for metabolic rewiring. However, pharmacological approaches to block NRF2 are still missing. Despite in some tumours the constitutive activation of NRF2 is mainly caused by NRF2 and KEAP1 genes mutations, its hyperactivation can be achieved also independently of these, suggesting that other signalling pathways can sustain NRF2 activity. Protein Tyrosine kinases (PTKs), often overactivated in cancer and influencing several signalling pathways, are promising candidates to explore for their potential impact on NRF2.

Bioinformatic analyses using TCGA and GEO databases were performed to investigate the survival probability of TNBC and non-TNBC patients stratified them based on different expression levels of RTKs and NRF2. Immunoblotting, immunofluorescence, RT-qPCR and RNAseq experiments were used to confirm the interplay between MET and NRF2. Moreover, cell viability and flow cytometry assays were performed to evaluate the efficacy of combinatorial treatments with paclitaxel and specific inhibitors of the MET–NRF2 axis in TNBC cellular models and patient-derived organoids.

Here, we identify a novel MET–SRC signalling axis that regulates NRF2 expression and activity, and demonstrate that its pharmacological targeting sensitizes TNBC cells and patient-derived organoids to the standard Paclitaxel treatment.

Our study shows that RTKs regulate NRF2 expression and activation in TNBC providing a proof of principle for the ability of Tyrosine Kinase Inhibitors (TKIs) to impinge on NRF2 signalling.

Our findings also uncover the value of the MET-SRC-NRF2 interplay as exploitable vulnerability in NRF2-hyperactivated TNBC, paving the way for the repositioning of TKIs as modulators of NRF2 signalling.

Hepatocellular carcinoma (HCC) is a highly aggressive malignancy characterized by pronounced inter- and intra-tumoral heterogeneity, which contributes to poor treatment responses and limited therapeutic options. Traditional two-dimensional cell lines fail to capture the molecular diversity and structural complexity of primary tumours, limiting their translational relevance. Three-dimensional tumoroid models better preserve tissue architecture and cellular interactions, but remain challenging to establish systematically from patient samples. Genetically engineered mouse models, such as the Alb-R26Met mice, develop spontaneous liver tumours that recapitulate key features of aggressive human HCC. These models therefore offer a valuable intermediary system for generating physiologically relevant tumoroids and exploring drug response variability in a controlled background.

Primary tumour cells were isolated from spontaneously arising HCCs in Alb-R26Met mice. We established and characterized a panel of independent primary cell lines and optimized protocols for generating corresponding three-dimensional (3D) tumoroids. Morphological, histological, and molecular profiling were performed using imaging, immunostaining, and transcriptional analyses. Functional assays assessed proliferation rates and sensitivity to targeted therapies, including tyrosine kinase inhibitors (TKIs). The resulting tumoroid panel was then used to test single-agent and combinatorial treatments.

We established eight distinct primary cell lines displaying heterogeneous growth properties, molecular signatures, and pathway activation profiles. All lines efficiently generated robust 3D tumoroids that retained key histopathological features of the tumours of origin. Drug-response assays revealed marked variability in TKI sensitivity, consistent with the biological heterogeneity observed in vivo. Combination treatments further highlighted distinct response patterns across tumoroids, demonstrating the utility of this platform for identifying synergistic effects and resistant phenotypes.

Discussion and conclusion: We present a reproducible workflow for generating a diversified panel of mouse-derived HCC tumoroids that faithfully reflects tumour heterogeneity. This platform provides a tractable and physiologically relevant tool for disease modelling and for accelerating combinatorial drug discovery in HCC.

Pancreatic ductal adenocarcinoma (PDAC) has a 5-year survival rate of 12%, with nearly all patients experiencing recurrence or progression due to pre-existing (primary) or acquired (secondary) chemoresistant cells. This underscores the urgent need to understand PDAC heterogeneity and the effects of treatment on tumor cell populations.

This study aimed to characterize treatment-associated cellular dynamics in PDAC at single-cell resolution. Three patient-derived organoids (PDOs), representing both classical and basal-like transcriptional profiles, were treated with five chemotherapeutic agents : gemcitabine, paclitaxel, 5-fluorouracil (5-FU), oxaliplatin, and SN38. Single-cell RNA sequencing was performed before and after treatment, and clonal tracking was assessed using MitoTrace.

Treatment effects varied depending on the initial phenotype. Gemcitabine and paclitaxel showed the most pronounced transcriptomic changes, inducing a phenotypic shift toward more aggressive, basal-like programs. In contrast, 5-FU, oxaliplatin, and SN38 had comparatively minor effects on gene expression.

Notably, phenotype transitions were clone-specific, suggesting the presence of predetermined phenotypic landscapes that shape treatment-induced plasticity.

These findings highlight chemotherapy-induced clonal plasticity as a mechanism driving resistant phenotypes in PDAC, offering insights for therapeutic stratification and resistance monitoring.

Extracellular vesicles (EVs) are biological nanoparticles that have attracted increasing attention for diverse applications, including diagnostics and therapeutic vectorization [1]. EVs also provide an excellent platform for studying membrane proteins (MPs) in their native lipid environment [2]. Owing to their high stability, EVs allow long-term storage of MPs and enable the investigation of interactions between cell-surface MPs and ligands without the need for detergent-based membrane protein extraction. These properties make EVs particularly suitable for direct analysis of membrane protein–ligand interactions. This work focuses on the kinetic analysis of nanobody binding to integral membrane proteins displayed on extracellular vesicles.

We present a highly sensitive approach to study the kinetics of interactions between extracellular vesicles and nanobodies using the advanced surface-based, label-free biosensing technique known as Grating-Coupled Interferometry (GCI) [3]. The method incorporates novel non-clogging microfluidics that ensure stable baselines and uninterrupted flow, which are essential for accurate kinetic measurements in complex EV samples. The platform employs the waveRAPID assay, an innovative approach that enables full kinetic characterization from a single injection by applying repeated analyte pulses of increasing duration at a constant concentration [4]. This design significantly accelerates data acquisition and enhances experimental throughput.

Using the GCI platform combined with the WaveRAPID assay, we achieved precise kinetic measurements of nanobody interactions with EV-associated membrane proteins. The approach demonstrated high sensitivity and accuracy, even for tight, high-affinity interactions such as nanobody binding to the EV membrane protein SNMP1 [5]. The ability to extract complete kinetic parameters from a single injection substantially reduced assay time while maintaining data robustness.

The combination of non-clogging microfluidics and WaveRAPID-based GCI biosensing addresses key challenges in EV analysis, particularly baseline instability and limited throughput. By preserving membrane proteins in their native EV context, this approach avoids artifacts associated with detergent solubilization and enables physiologically relevant interaction studies. The method is especially advantageous for analyzing high-affinity nanobody–membrane protein interactions, which are often difficult to characterize using conventional techniques.

This study highlights a robust, rapid, and highly sensitive platform for kinetic profiling of nanobody binding to membrane proteins on extracellular vesicles. The integration of GCI technology with WaveRAPID assays enables detailed interaction analysis with minimal sample consumption and high throughput. These advances provide a powerful tool for membrane protein research and support the development of sensitive EV-based diagnostic applications.

Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive solid tumors, characterized by late diagnosis, rapid metastatic spread, and limited response to systemic therapies. FOLFIRINOX (FFX) is currently the standard first-line chemotherapy for eligible patients and improves survival compared to gemcitabine. However, its efficacy is strongly limited by treatment-associated toxicity and, more importantly, by the frequent emergence of chemoresistance. Metabolic reprogramming is a hallmark of PDAC and may play a central role in tumor adaptation to chemotherapy. This study aimed to characterize FOLFIRINOX-induced metabolic alterations associated with resistance in PDAC.

Patient-derived xenograft (PDX) models were generated from tumors of six PDAC patients and treated with escalating doses of FOLFIRINOX to obtain sensitive (FFX-S), resistant (FFX-R), and untreated (NT) tumor phenotypes. Transcriptomic profiling was performed using Agilent microarrays, followed by standardized preprocessing and normalization. Differential expression analysis was conducted using the Limma framework with correction for patient and batch effects. A curated metabolic gene filter integrating Gene Ontology annotations, expert-curated datasets, and literature-derived pathways was applied to focus on metabolism-related genes. Functional enrichment analyses, network visualization, and in silico validation were performed using EnrichR, DAVID, Cytoscape, and public bulk and single-cell RNA sequencing datasets.

Comparative transcriptomic analysis between FFX-resistant and FFX-sensitive tumors identified 141 deregulated metabolic genes, including 47 upregulated and 94 downregulated. Enrichment revealed disruptions in fucosylation, and bilirubin biosynthesis. Applying a stricter threshold narrowed the set to 52 key metabolic genes, with FUOM and BLVRA emerging as prominent candidates. FUOM, involved in fucose metabolism, and BLVRA, a regulator of heme metabolism and oxidative stress, were strongly modulated by FOLFIRINOX. Single-cell RNA-seq showed induction of both genes mainly in tumor cells, with distinct regulation in fibroblasts and macrophages.

These findings demonstrate that FOLFIRINOX resistance in PDAC is associated with profound metabolic reprogramming affecting carbohydrate and heme-related pathways. The identification of FUOM and BLVRA highlights metabolic adaptations that may support tumor survival under chemotherapeutic stress through enhanced glycosylation and oxidative stress regulation. The compartment-specific regulation observed at the single-cell level underscores the importance of tumor–stroma interactions in shaping metabolic resistance mechanisms. Together, these results provide a mechanistic rationale for targeting metabolic pathways in combination with FOLFIRINOX to overcome resistance.

This study identifies novel metabolic signatures associated with FOLFIRINOX resistance in PDAC and proposes FUOM and BLVRA as potential biomarkers and therapeutic targets. Targeting these metabolic adaptations may represent a promising strategy to enhance FOLFIRINOX efficacy and improve patient outcomes.

Long-term survival in cancers with dismal prognosis, such as metastatic pancreatic ductal adenocarcinoma (mPDAC), IDH–wild-type glioblastoma (GBM-IDHwt), and extensive-stage small cell lung cancer (ES-SCLC), is exceptionally rare. Nonetheless, a small subset of patients achieves durable survival without clearly identified clinical or molecular features. These “exceptional survivors” represent a unique opportunity to uncover biological mechanisms underlying effective disease control. The ROSALIND study was designed to systematically dissect the molecular determinants of exceptional survival. The goal is to identify actionable vulnerabilities that could be leveraged to improve survival in patients with standard outcome.

ROSALIND is a retrospective, international, multicenter case-control study encompassing three cohorts: mPDAC, GBM-IDHwt, and ES-SCLC. Long-term survivors are defined as patients surviving more than five years for mPDAC and ES-SCLC, and more than three years for GBM-IDHwt. Controls are matched for tumor stage, age, sample type, and treatment history, with overall survival centered around the median survival reported in pivotal clinical trials for each tumor type. The analytical multi-omics pipeline includes whole-exome sequencing (WES), single-cell RNA sequencing, 10x Genomics Xenium spatial transcriptomics, proteomics, microbiome profiling (16S rRNA), and radiomics.

As of December 2025, 329 patients have been enrolled, including 244 long-term survivors, across the three cohorts: 52 ES-SCLC, 91 GBM-IDHwt, and 101 mPDAC. Patients were recruited through a global network of 87 centers across 34 countries, with an additional 1,466 exceptional survivors identified for potential inclusion. To date, 256 cases have been fully profiled. Preliminary results from the PDAC cohort show pronounced differences in stromal composition distinguishing exceptional survivors from standard survivors, with cancer cells displaying distinct relationships with the immune infiltrate and cancer-associated fibroblasts between the two groups.

ROSALIND extends beyond tumor-intrinsic alterations to incorporate spatial features of the tumor microenvironment as a critical determinant of outcome. Spatially resolved analyses suggest that standard survival in PDAC may be associated with a potentially more permissive immune niche compared with exceptional survivors. These findings point to context-dependent vulnerabilities within the tumor microenvironment. By integrating deeply annotated clinical data with high-resolution molecular and spatial profiling, ROSALIND is uncovering strategies to reprogram tumors toward long term-like states in patients with standard outcomes.

ROSALIND represents the largest systematic efforts to characterize exceptional survival in aggressive cancers using state-of-the-art multi-omic and spatial technologies. Our data provide a rational framework for developing precision oncology strategies aimed at improving survival in patients with otherwise lethal malignancies.