Intratumoral heterogeneity (ITH), encompassing both genetic and non-genetic variation, is a key driver of tumor evolution and therapeutic resistance. It leads to the coexistence of phenotypically and functionally distinct tumor cell populations within a single lesion, each governed by specific transcriptional programs. In parallel, the fate of tumor cells is not solely determined by intrinsic features but is also shaped by interactions with the surrounding tumor microenvironment (TME), which plays a central role in tumor progression and treatment resistance. Deciphering the spatial organization and functional interplay between ITH and the TME is essential to understand mechanisms of immune escape and therapeutic failure. Thus, we aim to characterize how intratumoral heterogeneity is spatially organized within the TME and to determine whether these spatial relationships drive immune evasion and contribute to failure of anti-tumor immune responses.

We employ complementary approaches to dissect the interplay between intratumoral heterogeneity and the immune microenvironment. First, we use breast cancer cell line models recapitulating intratumoral heterogeneity to interrogate tumor–immune interactions through functional co-culture assays. Second, spatial transcriptomic analysis of patient tumor sections is used to map distinct tumor cell states and immune cell populations within the TME, revealing how immune infiltration patterns correlate with specific tumor cell states.

Functional co-culture assays identify differential immune susceptibility across tumor cell populations, enabling the derivation of a resistance-associated transcriptional signature. In parallel, spatial omics analysis reveals organized immune infiltration patterns aligned with distinct tumor cell identities. Moreover, this resistance signature is expressed by a specific tumor cell state that spatially correlates with immune exclusion and altered immune infiltration patterns, validating the biological relevance of the signature and supporting its use in a CRISPRi screen to identify novel mediators of immune escape.

The spatial interplay between tumor heterogeneity and immune cell distribution is a key determinant of immune evasion, yet remains poorly characterized. By revealing which tumor cell states drive immune escape, we can decode mechanisms underlying resistance and identify new therapeutic targets.

By integrating complementary functional and spatial approaches, this work elucidates mechanisms by which intratumoral heterogeneity facilitates immune escape in breast cancer and highlights candidate targets for overcoming resistance.

Advances in cancer neuroscience demonstrate that the peripheral nervous system actively contributes to cancer progression, rather than passively responding to it. Tumors reshape the surrounding neuronal circuits, and sensory neurons promote tumor growth through direct interactions and bioactive signals. This bidirectional crosstalk is particularly evident in pancreatic ductal adenocarcinoma (PDAC). Our team has demonstrated that sensory neurons undergo extensive remodeling within PDAC lesions. Although targeting this neuronal plasticity could limit PDAC progression, the molecular cues driving PDAC innervation remain largely unknown. Given the close interaction between axons and PDAC cells, we hypothesized that PDAC cells actively promote sensory innervation by releasing factors that attract axons within the tumor.

Rodent dorsal root ganglia (DRG)-like nociceptors and mechanoreceptors were cultured in a bicompartmentalized organ-on-a-chip device and exposed to conditioned media (CM) from the human PDAC cell lines Panc1 and MiaPaCa2. Candidate secreted factors were identified via multi-omics meta-analysis, which combined secretomic and transcriptomic data. We validated secretion levels using ELISA and tested axon outgrowth with neutralizing antibodies on rodent and human induced pluripotent stem cell (iPSC)-derived sensory neurons. In parallel, we used patient-derived PDAC lines to evaluate whether our system could capture differential sensory neuron responses to clinically relevant tumor variants.

We found that the CM of Panc1, but not MiaPaCa2, cells promotes the growth of sensory axons in nociceptor-like neuron cell lines, but not mechanoreceptor-like ones. We identified GDF15 (growth differentiation factor 15) and DKK1 (Dickkopf-1) as potential candidates for inducing sensory axon outgrowth. These proteins are overexpressed by Panc1 cells compared to MiaPaCa2 cells. Functional validation of the two candidates revealed that recombinant GDF15 and DKK1 proteins increase the axon length of rodent nociceptors. Furthermore, blocking GDF15 and DKK1 specifically in Panc1 CM but not in MiaPaCa2 CM reduced the axon length of rodent and human iPSC-derived sensory nociceptor neurons. Screening of patient-derived PDAC lines confirmed GDF15 and DKK1 as relevant clinical candidates. GDF15 is broadly overexpressed and strongly associated with classical PurIST/PAMG signatures, while DKK1 correlates with gemcitabine resistance. Next, we will test how these patient-derived lines affect human sensory neurons differently.

Our results demonstrate that PDAC cells stimulate pro-tumor sensory axon remodeling by secreting GDF15 and DKK1. Since both factors promote tumor growth and their expression in patient-derived lines correlates with classical signatures and chemoresistance, targeting them could limit tumor progression and the pro-tumor influence of sensory neurons simultaneously.

Despite the lack of catalytic activity, pseudokinases contribute to cell signalling and have pivotal functions in physiology and diseases. Bulk analyses of many cancer types, including colorectal cancer, have frequently correlated high expression of the pseudokinase PTK7 receptor to poor prognosis and resistance to treatment. However how it intervenes in heterogenous cancer cell populations at the pronostic and molecular levels is yet unknown. Here we uncover a previously unrecognized mechanism by which PTK7 controls signaling and functions of EphA2, an active tyrosine kinase receptor which plays a key role in cell-cell communication and metastasis.

Heterogenous co-expression of PTK7 and EphA2 assessed at the single cell level is observed in malignant colorectal tumors and associated to distinct pathways. Molecularly, we show that PTK7 regulates the oligomeric state of EphA2 through a direct interaction. PTK7 loss stabilizes active EphA2 dimers resisting to K63 ubiquitin and lysosomal-dependent degradation, and perturbs integrin functions. PTK7-deficient cells with loose adhesive properties to extracellular matrices have a more pronounced metastatic potential in mice. Likewise, PTK7low/EphA2high expression in tumors predicts poorer clinical outcome of patients.

Our findings establish the PTK7-EphA2 axis as an essential regulator of integrin-driven adhesion and reveal the duality of PTK7 functions in the prometastatic program and probably in the response to treatment.

Diffuse Midline Glioma (DMG) is a fatal type of pediatric brain tumor. It represents one of the biggest challenges in pediatric oncology with a median overall survival of 9-11 months. After decades of clinical trial failure, the imipridone ONC-201 has just been approved by the FDA for patients with recurrent H3K27M-mutant tumors. While most patients temporarily respond to this molecule, tumor growth inevitably restarts, highlighting the need to identify combination therapies to enhance (or prolong) its efficacy. Approximately 80% of DMG harbour a recurrent somatic mutation on histone H3, the H3K27M mutation. It has been identified as a key driver of DMG, due to the major epigenetic dysregulation it induces. Indeed, recent data support the view that at least two different epigenetic cell states coexist within DMG tumors, thereby creating a specific and high level of intra-tumoral heterogeneity. Moreover, several studies have identified a metabolic reprogramming of DMG tumors and highlighted their dependencies on cholesterol biosynthesis, TCA cycle and de novo pyrimidine biosynthesis.

We hypothesize that the metabolic reprogramming occurring in DMG cells constitutes a targetable vulnerability that could be exploited to develop innovative combination therapies and increase our knowledge of DMG biology. To test this hypothesis, we performed a high-throughput drug screening using a 110-drug focused library in combination with ONC-201 or its derivatives in 8 different models of patient-derived DMG.

Amongst the 330 tested pairwise combinations, the association of ONC-201 (and its derivatives) with NAMPT inhibitors was identified as the most potent. Using a matrix of 6x5 different drug concentrations, we validated the potency of the drug combination in 4 different models, using different NAMPT inhibitors. By calculating the Bliss score, we were able to define the association as highly synergistic. Furthermore, a second high-throughput screening using a dedicated library of 152 metabolic inhibitors confirmed that NAMPT inhibition efficiently impacts DMG neurosphere growth.

Functional validation is currently underway to ascertain ON-target mechanism involved in the synergy between imipridones and NAMPT inhibitors.

Thus, using high-throughput drug screening we identified a metabolic vulnerability in DMG which can be therapeutically exploited by combining ONC compounds with NAMPT inhibitors. To go further, we will study the impact of this combination on the different cell sub-populations using spectral cytometry, to investigate the level of complexity between epigenetic and metabolic reprogramming heterogeneity and identify vulnerabilities as a step towards the development of better tailored treatments for DMG.

Glioblastoma (GBM) is the most aggressive primary brain tumor in adults, with less than 5% survival at 5 years (Stupp 2005). This poor prognosis is due to its multi-layer intra-tumoral heterogeneity. In particular, three main molecular states have been identified, namely the mesenchymal (MES), astrocyte-like (AC) and OPC-like (OPC) state (Neftel 2019). Importantly, GBM cells can shift between molecular states, in particular from OPC to MES state, allowing them to adapt, resist and escape treatments (Schmitt 2021, Hara 2021). In this context, understanding the dynamics of these molecular shifts is crucial to improve GBM outcomes. However, transition monitoring is currently limited by the lack of a specific marker of each state and the limited diversity of molecular states in primary GBM cultures. Here, we designed a complex tumoroid model integrating a defined ratio of molecular states with dynamic dual tools monitoring both OPC and MES states.

Primary GBM cells were transduced with both OPC and MES genetic tracers. Tumoroids were generated with either OPC cells, MES cells or an equal mixture of both. Dynamic molecular transitions were performed using western blot, flow cytometry and videomicroscopy.

We initially validated the specific expression of OPC- and MES- genetic tracer according to the established molecular subtype using OPC and MES cells, respectively. Then, we evaluated the cell state transition upon TNFα. We observed a dynamic transition from an OPC state toward MES state. Finally, since spatial molecular state organization is associated with the metabolic landscape, we investigate whether metabolic restriction impacts GBM molecular dynamics. Both glucose or glutamine restriction triggered the emergence of a hybrid OPC/MES state in OPC tumoroids. Strikingly, similar experiments performed in the presence of MES cells induced a complete transition of OPC cells into MES state.

Since molecular transition has been reported in response to radiotherapy and chemotherapy, this innovative model will allow a better understanding of the temporal dynamics of this process, the molecular mechanisms involved and the identification of potential strategies preventing this dynamic.

This innovative GBM model allows the dynamic monitoring of molecular transition, a crucial step in understanding tumor adaptation to treatment involved in systematic GBM recurrence and poor prognosis.

Colorectal cancer remains a major public health concern in France, ranking among the most common cancers and representing the second leading cause of cancer-related mortality. Extensive research has focused on genetic alterations in epithelial cells, particularly allelic losses of the tumor suppressor gene APC (Adenomatous Polyposis Coli), identified as the initiating event in 80–85% of sporadic CRC cases. Over the past decade, the tumor microenvironment has emerged as a critical determinant of cancer initiation, progression, and therapeutic response. Among the components of the TME, the immune system plays a dual role in both restraining and promoting tumor development. Understanding the mechanisms leading to the establishment of a pro-tumoral immune microenvironment could facilitate the discovery of novel prognostic biomarkers and foster the development of immune-based therapy. We hypothesized that intestinal tuft cells, a specialized population of epithelial cells with immunomodulatory functions, influence tumor initiation by shaping the immune microenvironment during colorectal tumorigenesis.

To test this hypothesis, we used a mouse model of intestinal tumorigenesis (ApcΔ14/+) combined with tuft cell deficiency (Pou2f3⁻/⁻). The immune cell infiltrate in the non-tumoral intestinal mucosa was analyzed to determine how absence of tuft cells affects immune composition. We also employed the DEREG mouse model to specifically deplete regulatory T cells (Tregs) in vivo during tumorigenesis. To identify the mediators through which tuft cells modulate the immune microenvironment, we focused on prostanoids—pro-inflammatory lipid mediators synthesized by COX-2. For this purpose, we generated a conditional Cox-2 knockout in tuft cells (Cox2fl/fl ; Villin-Cre ; ApcΔ14/+) to assess the impact of COX-2 loss on immune infiltration and tumor initiation.

Tuft cell deficiency in ApcΔ14/+ mice led to a significant reduction in tumor numbers, accompanied by a decrease in Treg infiltration in the non-tumoral mucosa compared with controls. Specific depletion of Tregs in DEREG mice similarly resulted in a drastic reduction in tumor initiation, highlighting the essential role of Tregs during early tumorigenesis. We further found that tuft cells are the only epithelial population expressing Cox-2 in the context of Apc heterozygosity. In addition, the specific genetic deletion of Cox2 in tuft cells phenocopied tuft cell deficiency, both in terms of reduced tumor initiation and decreased Treg infiltration.

Overall, our findings demonstrate that tuft cells promote tumor initiation through COX-2-dependent modulation of the immune microenvironment, notably via recruitment of immunosuppressive Tregs. This work highlights tuft cells and COX-2 signaling as promising targets for preventive and therapeutic interventions in colorectal cancer.

Breast cancer screening strategies lead to an increase in the detection of diverse preneoplastic lesions, raising the risk to overdiagnosis and/or overtreatment due to the uncertainty about their evolution to cancer. Understanding the early steps of breast cancer initiation is therefore essential. Tumor initiation theories differ: some base on genetics, with a single somatic mutation driving tumor formation, while others highlight epigenetic changes as a priming event. Together, these findings suggest that epigenetic perturbations may raise cell susceptibility to transformation, promoting additional mutations and tumor development. In the mammary gland, a complex tissue whose epithelium is structured into a cell hierarchy, our lab demonstrated that epigenetic perturbations can disrupt mammary epithelial differentiation and homeostasis, thereby enhancing tumorigenesis. Altogether, this supports our hypothesis that epigenetic priming can represent an initiating event in tumorigenesis, increasing mammary epithelial cell susceptibility to acquire genetic alterations and drive cancer development, while genetic alterations alone are insufficient to initiate tumor.

To decipher these early events, normal human mammary epithelial (HME) cells were used as a model. HME cells reveals a continuum from mammary stem cells (MaSC) to luminal cells (LC), tracing a simplified mammary hierarchy. An epidrug screen was performed to identify epigenetic perturbations disrupting differentiation of MaSC-sorted cells, validated by mammosphere assays. Using colony formation assays and xenotransplantation into humanized fat pads of immunodeficient mice, we tested whether epigenetic priming was sufficient to increase HME cell susceptibility to transformation following an oncogenic stress. Finally, scRNAseq and CUT&RUNseq were performed to study how this epigenetic perturbation impairs differentiation and cell susceptibility.

The epidrug screen revealed an enrichment of bromodomain inhibitors among the drugs with the greatest impact on MaSC differentiation. We selected a BRD4i, a demonstrate that HME treatment with this compound increased MaSC proportion by impairing differentiation, confirmed by an increase in MaSC-module genes in scRNAseq. Then, PIK3CAH1047R was expressed in BRD4i-treated cells. In vitro and in vivo, BRD4i-PIK3CAH1047R cells showed increased growth and formed tumor, whereas PIK3CAH1047R-only cells formed fibrocystic non-tumoral gland. CUT&RUNseq showed perturbation activated enhancers leading to deregulation of gene expression which could explain this impairment of differentiation and increase of cell susceptibility to be transformed.

BRD4i alters MaSC differentiation and primes cells for PIK3CAH1047R-induced transformation. These results support that epigenetic perturbation can act as an early event in breast tumorigenesis and suggest that specific gene expression changes contribute to increased cell susceptibility.

The incidence of pancreatic ductal adenocarcinoma (PDAC) has risen significantly in recent years and, although considered rare, it is projected to become the second deadliest cancer by 2030 due to the lack of early biomarkers and high resistance to current therapies. Immunotherapy has brought new hope for other cancers. However, PDAC is characterized by a highly developed stroma, and crosstalk between tumor cells and the numerous stromal cell types within the tumor microenvironment (TME) coordinates the suppression of anti-tumor immune responses and promotes resistance to current immune therapies. Tumor-associated macrophages dominate the PDAC immune landscape and play a central role in shaping this immunosuppressive barrier.

Tumor- and stromal cell–secreted factors circulate beyond the PDAC TME and reach the bone marrow. To investigate the impact of this systemic signalling on adaptive immunity and hematopoietic niche remodeling, we analysed bone marrow–derived stem and progenitor cells, as well as bone marrow–derived macrophages (BMDMs), from mice with PDAC compared to controls. In parallel, we assessed the functional consequences of this potential immune training on tumor progression in vivo using chimeric mouse models.

We show a series of epigenetic changes in bone marrow–derived populations that correlate with immune reprogramming in tumor-bearing mice. It appears that this reprogramming is heritable, supporting a model of tumor-trained immunity. We show that BMDMs from these tumor-bearing mice displayed a pro-tumor phenotype upon ex vivo stimulation with PDAC-secreted factors. Functionally, our data suggest that this training was associated with increased tumor growth and reduced infiltration of pro-inflammatory tumor-infiltrating lymphocytes (TILs) and tumor-associated macrophages (TAMs). Epigenetic signatures identified in circulating monocytes from PDAC patients support the existence of this model in humans and suggest a potential biomarker for earlier detection.

These findings suggest that systemic signals originating from the tumor microenvironment can remodel the hematopoietic compartment and shape immune responses that favor tumor progression. The identification of heritable epigenetic changes supports a model in which trained immunity contributes to the establishment of an immunosuppressive environment in PDAC.

Taken together, we propose a new model of trained-immunity to explain how the immune system adapts in response to circulating tumor factors. Epigenetic reprogramming associated to this training could provide a therapeutic target to regulate the anti-tumor immune response in PDAC at both primary and metastatic sites, and to prevent relapse in patients.

The limitations of chemotherapy (e.g., toxicities, limited efficacy) have led to the development of nanocarriers for drug delivery to improve pharmacokinetics (PK) and therapeutic outcomes. However, optimizing dosing regimens remains challenging. Moreover, since chemotherapy are mainly administered intravenously (IV), this results in patient discomfort and high treatment cost.

To address these issues, we used PK/pharmacodynamics (PD) modeling and applied it to subcutaneously (SC) injectable polymer prodrug based on paclitaxel (Ptx) and polyacrylamide (PAAm).

PK/PD studies were performed on MCF-7 tumor-bearing mice. The PK model was developed on IV Ptx and SC Ptx-PAAm data. The PD model was developed on control, IV Ptx, and SC Ptx-PAAm groups (15 mg/kg), and validated on an independent group (SC Ptx-PAAm 60 mg/kg). Optimal dosing regimens identified in silico were then validated in vivo with excellent agreement. A dosing regimen combining a loading dose and daily injections achieved a 60% complete response rate without added toxicity, outperforming prior results.

This is the first validated PK/PD model for nanocarriers, offering a framework for more effective, cost-efficient, and ethically refined drug development.

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal cancer with a five-year survival rate under 12%. Chemotherapy, including mFOLFIRINOX or gemcitabine-based regimens, is the main treatment for unresectable cases but achieves tumor regression in only 10%-30% of patients. Nearly all patients experience recurrence or progression, driven by preexisted (primary) or acquired (secondary) chemoresistant cells. The hypothesis of this study is that the drug-response profile and phenotypic evolution dynamics of PDAC tumors are determined by the tumor cell phenotype at the niche level, which in turn drives the tumor’s drug response and the patient’s outcome.

Intratumoral cell phenotypes were identified through the integration of bulk and single-cell transcriptomic data. The identified cell phenotypes were validated in single-cell RNA sequencing (scRNA-seq) cohorts of patients. Additionally, the treatment-associated dynamics of the identified phenotypes was evaluated in three patient-derived organoids (PDO) treated with five chemotherapeutic agents: gemcitabine, paclitaxel, 5-fluorouracil (5-FU), oxaliplatin, and SN38. Lastly the enrichment of the cell phenotypes in bulk RNAseq of patients was assessed across three independent cohorts: 343 resectable cases (PRODIGE-24), 65 primary tumors from metastatic patients, and 29 paired samples collected before and after neoadjuvant treatment.

Five phenotype modules were identified: two associated with pancreatic differentiation, an epithelioid phenotype, a squamous phenotype, and an EMT phenotype. Validation in single-cell cohorts of patients showed significant enrichment of squamous and EMT-related cell clusters in the metastatic stage (P<0.001) and after neoadjuvant treatment (P<0.001). Treatment-associated dynamics showed that the effect varied depending on the initial phenotype. Gemcitabine and paclitaxel showed the most pronounced transcriptomic changes, inducing a phenotypic shift toward more aggressive phenotypes. Analysis of bulk RNA-seq data revealed that patient prognosis was associated with the dominant cell phenotype. In the PRODIGE-24 cohort, patients enriched in the Differentiated 1 phenotype exhibited a better prognosis in both treatment arms: gemcitabine (stratified HR: 0.38; 95% CI, 0.19–0.75; P=0.005) and FOLFIRINOX (stratified HR:0.40; 95% CI, 0.20–0.79; P=0.009), whereas those enriched in squamous had the worst prognosis. A similar pattern was observed in metastatic patients. In the neoadjuvant setting, treatment led to a shift toward more unfavorable phenotypes.

Characterizing intratumoral heterogeneity and treatment-associated dynamics in PDAC reveals tumor cells as key drivers of patient outcome. Patient stratification by tumor cell composition enables improved treatment allocation and identification of resistance pathways for personalized treatment strategies.

This study highlights a link between PDAC cell phenotypes and drug response, suggesting plasticity-driven mechanisms underlying acquired chemoresistance.

Patient Tumor-Derived Organoids (PDTOs) provide an ex vivo platform to test drug sensitivity and guide individualized treatment decisions within functional personalized medicine (FPM). Observational studies show PDTOs can recapitulate clinical responses, but their operational implementation and clinical utility remain largely untested in prospective interventional trials. The primary objective was to assess the feasibility of generating PDTOs and a drug sensitivity profile (chemogram) within 10 weeks in more than 50% of evaluable patients. Secondary objectives were to determine the proportion of patients receiving PDTO-guided treatment and to evaluate treatment efficacy.

In this multicenter phase I/II trial, patients with heavily pretreated metastatic colorectal cancer underwent tumor biopsy for PDTO generation. Each PDTO was tested against a 25-drug panel including off-label CRC treatments. The chemograms generated were reviewed by a dedicated tumor board, which issued personalized treatment recommendations.

A total of 61 patients were enrolled, and 54 biopsied, forming the per-protocol evaluable population. The PDTO take-on rate was 78%, the highest reported in a prospective solid-tumor FPM study. The chemograms were generated within 10 weeks for 39 patients (72%), meeting the primary endpoint. Nineteen patients (35%) received PDTO-guided treatment. Clinical benefit, evaluated in the PDTO-guided population, was observed in four patients (21%), achieving durable disease control lasting between 5.6 and 12.9 months.

ORGANOTREAT-01 demonstrates that PDTO-based drug testing is feasible and can be integrated into routine clinical workflows. The unprecedented take-on rate enabled drug testing in most patients, resulting in durable disease control in a subset. These findings support further optimization of FPM strategies and the expansion of PDTO-guided approaches to other solid tumors.