Thèse Rôle du Microbiote Tumoral dans la Polyrésistance aux Traitements du Cancer Pancréatique H/F

Doctorat_Gouv

Établissement : Université Paris-Saclay GS Life Sciences and Health
École doctorale : Cancérologie : Biologie - Médecine - Santé
Laboratoire de recherche : MIcrobiota and MucOsae for cancer immunoSurveillAnce
Direction de la thèse : Laurence ZITVOGEL ORCID 0000000315960998
Début de la thèse : 2026-10-01
Date limite de candidature : 2026-05-03T23:59:59

L'immunothérapie, seule ou associée à d'autres agents cytotoxiques ou antiangiogéniques, constitue actuellement le pilier du traitement clinique des patients atteints de cancer. Malgré un bénéfice clinique significatif attribuable à ces schémas thérapeutiques combinés, une grande partie des patients atteints de cancers avancés ou métastatiques restent résistants au traitement. La plupart des tumeurs malignes sensibles aux inhibiteurs de points de contrôle immunitaire (notamment les cancers du poumon, de la tête et du cou, de la vessie, de l'oesophage et colorectal) se développent dans la peau ou les muqueuses, zones colonisées par un microbiote local. Bien que non directement oncogènes, les microbes intratumoraux (bactéries ou champignons) peuvent induire un microenvironnement prémétastatique et créer un microenvironnement tumoral pro-inflammatoire délétère pour l'hôte. Il a été rapporté que les patients présentent souvent une dysbiose locale (une altération de la composition du microbiote, au moins au niveau intestinal) et des déficits des réponses immunitaires humorales et cellulaires dirigées contre ces pathobiontes associés à la tumeur, comparativement aux sujets sains. De plus, nos données préliminaires soulignent que des titres élevés d'IgG dirigés contre les pathobiontes envahissant les tumeurs des muqueuses sont associés à des réponses durables à l'immunothérapie, tandis que les réponses IgG dirigées contre les commensaux bénéfiques sont délétères lors du blocage des points de contrôle immunitaire. Par conséquent, nous postulons que le transfert passif d'anticorps dotés d'un potentiel neutralisant ou de fonctions cytotoxiques contre les pathobiontes, seuls ou intégrés dans des monocytes, ou encore la vaccination contre les pathobiontes associés au cancer, pourrait révolutionner l'onco-immunologie. Notre priorité sera d'étudier en premier lieu le microbiote tumoral du cancer du pancréas (pDAC) et les bactéries déjà isolées par notre laboratoire à partir de 277 cancers primitifs. En second lieu, nous testerons comment chacun de ces pathobiontes peut induire ou aggraver des modèles tumoraux de pDAC chez la souris. Enfin, nous générerons des anticorps neutralisants ciblant ces bactéries afin de valider ce concept et de développer de nouvelles stratégies thérapeutiques.

Pathobionts or harmful bacteria invade the tumor beds (in colorectal, head and neck, and lung cancer) and reprogram tumor cells towards an inflammatory, immunoresistant and proliferative pattern culminating in oncogenesis exacerbation. The tumor cell-bacterium interaction obeys to a symbiosis rather than an ectoparasitism, suggesting that this is a win-win situation for the tumorigenic process and for the bacterium (at the expense of the host). Metastases contain Fusobacterium nucleatum spp (F.nucleatum) that can be passed on to several generations of tumors implanted in patient derived xenografts in immunocompromised animals [1]. A causality link between presence of intratumoral bacteria and tumor progression has been brought up directly showing carcinogenic transformation of colon enteroids with genotoxic Escherichia coli pks+ strains [2]. Secondly, using a reductionist in vitro co-culture approach and live-cell confocal microscopy, Galeano Niño et al. showed that F.nucleatum-infected (but not uninfected) colorectal cancer spheroids recruited and retained neutrophils concomitantly to the activation of stress kinases [3]. Thirdly, intracytoplasmic bacteria (live facultative anaerobes such as Staphylococcus, Streptococcus and Lactobacillus spp) residing in murine mammary cancers and their metastases caused actin cytoskeleton remodeling and resistance to mechanical stress[4]. Finally, F.nucleatum infection of cancer cells suppressed calreticulin exposure preventing the immunogenic cell death of cancer cells [5] that was rescued by the liposomal formulation of tinidazole [6]. Hence, blocking bacterial entry into the cytosol of cancer cells and stopping bacterial proliferation in the extracellular tumor microenvironment to prevent spreading within myeloid cell and neighboring tumor cells is critical to restrain immunoevasion and prevent metastatic niche formation.
Besides the direct proinflammatory role of the pathobiont, another problem is the lack of immunization of the host against the microbe. Indeed, most patients do not mount neutralizing antibodies against these pathobionts (Fusobacterium nucleatum subsp. animalis or colibactin (pks)+ Escherichia coli). Moreover, our advanced patients rarely exhibited T cell responses against bacterial epitopes presented on tumor cell MHC complexes, in contrast to what was reported in melanoma[7]. Current approaches in oncology rely on antibodies directed against cell surface oncogenic proteins (such as HER2 or EGFR) not expressed in healthy tissues. However, antibodies directed against oncogenic pathways often lead to escape mutations. To our knowledge in oncology, no one targeted protumorigenic bacteria using passive transfer of antibodies or monocyte armed with neutralizing antibodies directed against these pathobionts. Antibiotics will never substitute for such antibodies because of their lack of specificity (destroying gut beneficial bacteria). Our project based on generating antibodies directed against intratumoral bacteria endowed with neutralizing or cytotoxic functions (ADCC, ADCP or complement binding) against these microbes and/or infected tumor cells represents a conceptual breakthrough with considerable impact on cancers located at mucosal sites. In addition, this project may bring solutions to gut dysbiosis, since some of the intratumoral pathobionts may originate from the gut microbiota, and may be also listed in the bacteria associated with disabled ecosystems.
Colon and pancreatic carcinoma represent common and hard to treat malignancies with high metastatic potential. These neoplasias develop within mucosal areas inhabited by a commensalism but also exposed to external pathogens. Although these cancers are amenable to chemoimmunotherapy based on immune checkpoint inhibitors (ICI), they remain a clinical challenge at advanced stages for oncologists. Several laboratories contributed to the demonstration of a clinical relevance of gut microbiota in cancer patients treated with ICI and CAR-T cells [8, 9]. While there is strong molecular evidence of an intratumoral microbiome across >30 cancer types [7, 10-12], in primary and metastatic tumors, the clinical significance of these intratumoral microbes is currently being discussed. Imaging data show the co-localization of pan-bacterial markers with immune and epithelial cell targets, suggesting that the intratumoral microbiota can be intracellular[1, 3, 7, 11]. In vitro and preclinical animal models indicated that bacteria such as Fusobacterium nucleatum subsp. animalis or genotoxic colibactin (pks)+ Escherichia coli (called pathobionts henceforth) in the tumour-associated microbiota have a role in cancer development [13], metastasis [1, 4, 14], immunoevasion [15] and chemoresistance [16, 17], more specifically in colorectal and head and neck cancers developing in nonsterile mucosae.
The current project relies on unpublished work based on culturomics of 143 fresh tumor samples that confirm the presence of pathobionts in lung and head & neck tumors. Moreover, host-microbe interactions are disrupted in cancer patients at three levels. First, shot gun metagenomics analyses of stools highlight a deviated taxonomic composition of the local commensalism (apostrophed dysbiosis) in cancer patients compared with healthy subjects [18]. Secondly, cancer patients exhibit a defect in IgG/IgA antibody titers and Th1 responses directed against pathobionts compared with healthy individuals (our unpublished data). Goubet et al. reported that urothelial cancer patients experiencing iterative urinary infections with uropathogenic E.coli that invade their cancer cells experienced a better clinical outcome in neoadjuvant immunotherapy when mounting E.coli-specific IgG and T helper cells [19]. Similarly, lung and colon cancer patients who have detectable fecal F. nucleatum exhibit a worse prognosis but those who were seroconverted against F. nucleatum exhibited a better progression-free survival during immunotherapy.
Hence, we conclude that humoral and cellular immune responses directed against pathobionts are beneficial to the host when the latter invade their tumors. Therefore, we postulate that passive transfer of antibodies endowed with neutralizing potential or cytotoxic functions against pathobionts, alone or engineered within monocytes, or alternatively vaccination against cancer -associated pathobionts could revolutionize the field of oncoimmunology. The program project will bring this proof-of-concept and generate these novel therapeutic strategies.

The current application relies on preliminary data that are very promising but not yet at a mature stage, with some risks associated with its development. We have already i) identified that 10-15% of primary head and neck, lung or bladder cancers contain tumor infiltrating memory B cells, max. 1% of them binding to F. nucleatum or E. coli pks+, ii) we have already run >800 serologies against 10 gut commensals and identified immunogenic pathobionts (such as F.nucleatum spp. or pks+ E.coli), iii) we have studied memory T cell responses against 6 gut commensals in > 70 patients and 20 healthy volunteers.
In brief, while gut prevalence of Akkermansia muciniphila or F. prausnitzii is associated with clinical benefit to immunotherapy in lung cancers [20, 21], IgG responses directed against this beneficial Gram negative bacterium are of dismal prognosis in first line NSCLC and are associated with the loss of intestinal residency of Akkermansia spp. Mounting memory T cell (Th1 or Th17) responses against beneficial bacteria (Akkermansia and F. prausnitzii) are also deleterious for patients. Consistently, while gut prevalence of E. coli SGB 10068 is associated with stage IV colorectal cancer (CRC) and dismal prognosis during chemo-immunotherapy (our unpublished data), those patients who developed IgG+IgA+ responses against E coli pks+ (about 18% of them) exhibit a favorable clinical outcome. Indeed, CRC patients who have high E.coli-specific IgG+IgA+ have a relative loss of fecal E. coli SGB 10068. In other words, these findings suggest that prevalence of a harmful bug in the stools is not dangerous if a humoral immune response can be developed against it. Finally, we conducted a B cell cloning of memory B cells infiltrating tumors and capable of binding to F. nucleatum, based on the hypothesis that B cells educated in gut-associated lymphoid tissues can migrate to the tumor bed and/or de novo develop within the tertiary structures neoformed in tumors. After cell sorting and expansion, we screened the supernatant of the expanded B cell clones based on their bacterial reactivity. This process allowed us to generate a monoclonal antibody (B6 mAb) targeting F. nucleatum subspp. animalis clade 2 (Figure 2), which strongly binds to a subset of stool bacteria (mostly Rikensenellacae and Prevotellaceae, as shown by sorting and sequencing mAb+ and mAb microbiota fractions). Next, we confirmed that B6 clone recognizes, not only F. nucleatum spp but Alistipes shahii, Alistipes finegoldii, Alistipes jeddahensis (intratumoral bacterium), Prevotella buccae (isolated from a tumor) and Odoribacter splanchnicus strains with a high affinity (but not other Rikensenellaceae ). As previously reported for anti-pneumococcal antibodies [22] or during maternofetal transfer [23], cross-reactivity can be explained by the recognition of glycan structures [22] or murein lipoproteins [23, 24].

1) Develop model system for rapid bacterial-specific intratumoral B cell cloning, plasmablast differentiation and IgG sequencing
2) Generate antibodies directed against intratumoral bacteria endowed with neutralizing or cytotoxic functions against these bacteria and/or infected tumor cells
3) Examine the microbial reactivity of antibodies synthetized from clonal selection of tumor-infiltrating B lymphocytes (Team#2 and coordinator)
4) Proof-of-concept efficacy of Ab-based therapy in mouse models of colon or bladder carcinogenesis colonized with Fusobacterium nucleatum subsp. animalis or pks+ Escherichia coli

Objective 1: Intratumoral bacteria-specific B cell cloning and IgG/M sequencing.

Our laboratory (Dr Zitvogel) has stored fresh tumors from lung, colorectal cancers, head and neck and bladder cancers and tumor-draining lymph nodes (tdLNs) containing high proportions of B cells for subsequent thawing and cloning. Mature B cells capable of binding various bacteria are sorted and expanded using CD40L low-expressing feeder cells (kindly provided by G. Kelsoe) and supplemented culture media containing IL-2, IL-4, IL-21, and BAFF growth factors, according to established methods (Fig. 2A-C) [29]. This effective protocol has been successfully implemented in our lab and will be applied across diverse tumors and tdLNs with various pathobionts (priority will be given to targeting F. nucleatum subspp animalis clade 2, and pks+ E. coli). After 21-25 days, supernatants from expanded B cell clone cultures are screened for their ability to release IgG, IgA, and IgM against F. nucleatum and pks+ E. coli, or for exhibiting cross-reactivity with other bacteria. Each positive well will then have the B cell receptor sequenced. The sequences of heavy and light chains are subsequently forwarded to PROTEOGENIX, a company that specializes in the synthesis of purified monoclonal antibodies. Sequence analysis will not only facilitate the generation of monoclonal antibodies but also enable the identification of clonal expansion within this subset of intratumoral B cells that exhibit reactivity against bacterial components. We already obtained two mAb targeting distinct members of the Prevotellaceae and Veillonellaceae family, renown to be associated with pathogenicity.
Main Deliverable and milestone 1: Validate the methodology by successfully cloning and sequencing a mAb that recognizes one or more pathobionts, with a particular focus on F. nucleatum subsp. animalis clade 2 and pks+ E. coli.



Objective 2: Generate bacteria-specific antibodies with neutralizing or cytotoxic functions.

We will first validate the sensitivity, robustness, and specificity of the recombinant mAb directed towards our target pathobiont and assess its potential cross-reactivity with whole stool samples or various bacterial isolates, including different strains of the same species. We will then test whether our mAb triggers Fc-dependent immune signaling using luciferase reporter cell assays. Pasteurized bacteria will be incubated with IVIG or our clone Ab or isotype Ctl Ab and probed for the activation of FcR-receptor IIa (CD32), a surrogate for ADCP, and IIIa (CD16), a surrogate for ADCC, using luciferase reporter cell assays (collaboration with Johan Malmström). Alternatively, the FcR-expressing THP1 monocyte cell line will be infected with fluorescent pathobionts in the presence or absence of complement and our antibody. Flow cytometry will be used to assess cell death and infectivity.

Main Deliverables
2.1: Determine the affinity and avidity of the mAb clones from Objective 1 to the bacterium of interest
2.2: Demonstrate the ability of our mAb to neutralize or inhibit bacterial infectivity in tumor cells or macrophages/myeloid cells in vitro

Objective 3: Validate E.coli antigenic target OMPc

We contemplate several methods to nail down the target antigens for the clone B6 specific of Fusobacterium nucleatum spp animalis clade 2 and for the antigen relevant for pks+ E.coli.

Firstly, this consortium has implemented two phage-display library of self- and microbial-antigens for PhIP-seq (phage immunoprecipitation and sequencing). This powerful technique will allow us to screen antibodies (notably the B6 clone) against thousands of potential self-antigens simultaneously, providing a more comprehensive view of the autoantigen repertoire targeted by TIL-B cells. Their data suggest that the local B cell responses in tumors may lead to systemic autoantibody production that may cross react with microbial antigens.
Otherwise, we will try several protein extraction protocols for each strain of F. nuc, and will perform 2D electrophoresis, transfer onto nitrocellulose membrane for immunoblot followed by protein extraction and MALDI-TOF -based mass spectrometry. Alternatively, an immunoprecipitation could be performed.

Concerning antibodies targeting E.coli, we have screened more than 100 serum of bladder cancer patients to monitor their seroreactivities against 13 strains of E. coli (from public registries or in house for which we had the entire genome sequence), established a clading and focused on genetic variants associated with antibiotics resistance. A study identified six clades based on omp C and omp F mutations with a strong correlation to PBP3 insertions co-carried with beta-lactamases including bla NDM [30]). One strain of ST12 was recognized by patient serum while other strains from ST10 and ST131 were not, catching our attention (Figure 3). We aligned OMPC C and OMP F sequence differences according to seroreactivity profiles and found two mutated regions in OMPC that deserve a special focus (Figure 3 and 4). Protein variants have been synthetized and we are now testing the serum reactivity against these protein variants using home made ELISA. Two approaches will be conducted after validation of the specific cancer patient seroreactivity against these recombinant variants, to identify B cells binding to the fluorescent protein variant, sequence them and make antibodies directed against this (ese) protein(s) (several methods can be used including that described in Sokal et al.[29] and define the MHC class I and II epitopes that could be recognized by T cells (using IEBD software, designing 15 or 12 mers overlapping peptides binding to the most frequent HLA-DR-DP-DQ). Using patient PBMC and in vitro restimulation assays, we shall nail down peptide sequences that promote recall responses (by monitoring IFNg, IL-17, IL-10 release after peptide stimulation using immunoenzymatic assays) in cancer patients.


Main Deliverables
3.1: Determine the antigen of F. nucleatum or A. shahii recognized by B6 mAb
3.2: Validate the OmpC antigenic variant by tracking peripheral or tumoral B cells binding to it


Objective 4: Establish proof-of-concept for the efficacy of Ab-based therapy in mouse models of colon or PDAC colonized with pathobionts.

We will evaluate the efficacy of the generated mAbs in binding to murine FcR-bearing RAW macrophage cell lines or murine splenocytes, and then assess their effectiveness in i) neutralizing targeted bacteria following local or intravenous administration, ii) inducing tumor regression, and iii) reducing metastasis development in vivo. We will establish orthotopic cancers using the MC38 CRC and MB49 mucosa-invasive bladder cancer cell lines using micro-syringes [31]. We have already established the transplantable KRAS PDAC model where local inoculation of Streptotoccus sanguinis or pks+E.coli significantly inhibited the elicitation of long term protective antitumor immune responses triggered by anti-PD1 Abs (in the point-by-point reply). We will attempt to interfere in this inhibition by coinjecting iv our antibodies directed against these bacteria when we will have them.
The current in vivo experimental setting that we will concentrate on is the following one, already ongoing in our laboratory UMR1015, aimed at testing the capacity of B6 mAb to restrain F. nucleatum sp. animalis clade 2 or Prevotella intermedia (that is a target of B6 Ab) gut colonization. C57BL/6 mice will be implanted with a MB49. We know that when tumors reach a 40 mm2 size, there is a transient gut permeability, a downregulation of MADCAM-1 culminating in gut dysbiosis in a adrenergic-dependent manner (Yonekura et al Cancer Discov. 2022). We will perform an oral enforced gavage with F. nucleatum sp. animalis clade 2 or Prevotella intermedia after a short treatment with broad spectrum antibiotics, to ensure a better intestinal colonization and then inject iv or ip saturating doses of the IgG1 B6 Ab (or IgG1 isotype control). Next, daily qPCR and culturomics of stools will be performed to monitor the reduction of the relative abundance and prevalence of P. intermedia, and compared the relative capacity of B6 versus isotype control to drastically eliminate the bacterium. The system could be modulated using subclinical DSS-mediated colitis that should facilitate the transcytosis of B6 mAb from plasma to stools.

Alternatively, tumor cell lines will be infected in vitro with F. nucleatum subsp. animalis or E. coli ST12 or ST131 at various multiplicities of infection to determine the optimal percentage of cell infection that could mimic bacterial microniches. A bacterial strain with no cross-reactivity with our mAb will be used as a negative control. Alternatively, autologous myeloid cells infected with F. nucleatum subsp. animalis [32] and E. coli ST12 or ST131 can be injected systemically to colonize the TME following established protocols. Tumor colonization can also be accomplished with oral gavage of tumor-challenged mice with fluorescent-labeled bacterial strains.

Main Deliverables
4.1: Evaluate the impact (protumoral) of intratumoral bacteria injection in the proposed orthotopic tumor models on tumor growth and the induction of local immunosuppression.
4.2: Demonstrate the antitumor activity of our anti-pathobiont mAb, injected intratumorally and intravenously, in at least one of the proposed orthotopic tumor models.