Toulouse Onco Week 2025

Complex and Simple AI/ML Models for Harnessing Molecular Data in Oncology

This talk will highlight the application of AI/ML models in predicting the sensitivity and resistance of cancer cell lines to drugs, as well as assessing target safety in early drug discovery programs.

5 February at 14:30 CET  during the session "Artificial Intelligence & Cancer"
Presented by Andrey Zinovyev, Principal Scientist in Evotec’s In Silico R&D

Mice Humanized for the Immune System as Validated Tool for Therapeutic Antibody Development

Lise Pasquet¹, Alice Marchand¹, Marion Mars¹, Sophie Chabot¹, Fréderique Dol-Gleizes¹, Sébastien Tabruyn², Dan Georgess², Pascale Lejeune¹

^{1 EVOTEC France SAS, Campus Curie, 195 route d’Espagne, 31036 Toulouse, France
2 TransCure bioServices, 260 Avenue Marie Curie, 74160 Archamps, France}

Therapeutic antibodies modulating the immune response are frequently solely specific for human proteins and their preclinical development requires suitable mouse models. Therefore, mice carrying a human immune system allow to evaluate immunotherapies in a more relevant immune context. Such “humanized” mice are developed by reconstituting the immune system of immunodeficient animals with human immune cells from 2 possible origins: CD34⁺ hematopoietic stem cells (huCD34⁺ mice) isolated from cord blood, or mature Peripheral Blood Mononuclear Cells prepared from buffy coats (huPBMC mice). According to the compound to test and the goal of the research, one model can be preferred over the other.

HuCD34⁺ mice exhibit multilineage immune populations allowing the evaluation of immune checkpoint inhibitors. We developed the human A375 melanoma model in huCD34⁺ NCG mice to evaluate the combination strategy of relatlimab (anti-LAG3 antibody) and nivolumab (anti-PD1 antibody) approved at the beginning of 2024 as a first-line treatment for specific melanoma indications. Anti-PD1 single treatment tended to, albiet without statistical significance, decrease tumour growth compared to isotype control group while anti-LAG3 alone had no impact on tumour growth. The combination strategy induced a significant decrease in tumour growth compared to single agents in line with the data observed in the RELATIVITY-047 clinical trial (Tawbi H., NEJM, 2022).

We also developed huPBMC mice that exhibit exclusively T-cell lineage and are perfect fit to evaluate T-cell engagers like the CD3xEpCAM bispecific antibody, solitomab. We observed a dose-dependency between the dose of PBMC injected and obtained level of human reconstitution, but the level of reconstitution did not impact engraftment or tumour progression of the A375 melanoma model or the BT474-clone5 breast cancer model. In both models, we confirmed that the tumour growth did not impact the T-cell reconstitution, we also identified the modification of CD8/CD4 ratio in favor of the CD8 as a biomarker of solitomab activity.

Humanized immune system mice presenting either more diverse lineages (CD34+ huNCG) or lineage restricted populations (huPBMC) proved themselves efficient tools to evaluate anti-tumour immunotherapies. Acquiring the expertise to handle such models while recognizing their qualities and limitations contributes to selecting the most useful model for preclinical studies predictive of human therapeutic response.

Presented by Sophie Chabot, Group Leader in Immuno-oncology

Immune Checkpoint Therapy as a Treatment Option for Non-muscle Invasive Bladder Cancer

Sulayman Benmerzoug¹, Katie Louche¹, Thibault Angles¹, Frederique Dol-Gleizes¹, Pascale Lejeune¹ and Sophie Chabot^{1
1 Evotec SE, France, Toulouse}

Bladder cancer (BCa) is one of the most common malignancies with more than 500,000 new cases diagnosed each year. Intravesical instillations of Bacillus Calmette-Guérin (BCG) is the gold-standard treatment for non-muscle invasive BCa (NMIBC) patients despite significant side effects and high rates of refractory disease and/or recurrence. Indeed, treatment failure occurs in ~40% of cases emphasizing the urgent need to evaluate alternative new therapies including immunotherapies. Recently, the use of immune checkpoint inhibitor (ICI) anti-PD-1 therapy has been approved for the treatment of BCG-unresponsive NMIBC patients. Here, we investigated whether targeting PD-1 or CTLA-4 can boost anti-tumor immunity in the orthotopic murine MB49 BCa model. To this purpose, 2.5x10⁵ luciferase-expressing MB49 cells (MB49-Luc) have been instilled into the preconditioned bladder of female C57Bl/6 mice by intra-urethral catheterization. In vivo tumor growth has been monitored by non-invasive bioluminescence (BLI). All the inoculated mice showed a luciferase signal at day 3 post-bladder tumor inoculation, demonstrating 100% of tumor uptake. Based on BLI, a randomization has been performed before ICI treatment consisting of anti-CTLA-4 or anti-PD-1 (10 and 100 μg/mouse, respectively, q3dx4, i.p.). Our data indicate that ICI treatments significantly decreased BLI signal, indicating reduced tumor growth. Furthermore, anti-CTLA-4 and anti-PD-1 administration resulted in a significant long term anti-tumor activity, with ~90% and ~70% overall survival of anti-CTLA-4 and anti-PD-1 treated mice, respectively, while isotype control mice displayed ~10% survival by day 74 post-MB49-Luc cells inoculation. To investigate the potential development of systemic adaptive anti-tumor immunity elicited by ICI treatment, remaining ICI-treated mice and newly onboarded non-treated control mice have been (re)-challenged with MB49 cells subcutaneously (s.c.). While all the control mice showed tumor outgrowth and needed to be sacrificed due to humane endpoint by the day 32 post-s.c. inoculation, rechallenged ICI-treated mice showed potent abscopal effect and displayed no tumor engraftment with 100% overall survival. Altogether, these findings demonstrate that systemic ICI immunotherapies such as PD-1 and CTLA-4 elicit potent anti-tumor activity and induce protective anti-tumor immune memory. ICI might present an alternative or a complementary treatment to BCG therapy for NMIBC patients.

Presented by Sulayman Benmerzoug, Team leader in Immuno-oncology

Time Series Analysis of Cellular Interactions Through Gene Regulatory Network Inference

Hugo CHENEL^1,2, Malvina MARKU¹, Julie BORDENAVE¹, Marcin DOMAGALA¹, Mary POUPOT¹, Loic YSEBAERT⁴, Andrei ZINOVYEV² and Vera PANCALDI^1,3

^{1 CRCT, Université de Toulouse, Inserm, CNRS, Université Toulouse III-Paul Sabatier, Centre de Recherches en Cancérologie de Toulouse, France.
2 Evotec, Toulouse, France.
3 Barcelona Supercomputing Center, Carrer de Jordi Girona, 29-31, Barcelona, Spain.
4 Institut Universitaire du Cancer de Toulouse-Oncopole, Toulouse, France.
Corresponding Author: hugo.chenel@inserm.fr; malvina.marku@inserm.fr; vera.pancaldi@inserm.fr}

Gene regulatory networks (GRNs) are crucial to understand the complex regulatory mechanisms underlying cancer development and progression. They provide insights into molecular mechanisms driving cancer, such as identifying key driver genes, molecular pathways, and the identification of novel therapeutic targets. In the context of the tumor microenvironment (TME), the complex interactions between immune and cancer cells give rise to a cascade of regulatory processes at different levels, defining the cellular behavior and response to external stimuli. This project aims to investigate how regulatory interactions between genes characterize cellular behavior. Using an in vitro model of Chronic Lymphocytic Leukemia (CLL), a blood cancer characterized by the aberrant proliferation of malignant B lymphocytes in the lymph nodes, we studied CLL cell interactions with immune cells. In the lymph node, CLL cells interact with monocytes, promoting their differentiation into macrophages that promote CLL cell survival, similar to tumor associated macrophages in solid cancers. We conducted experiments in three conditions (CLL patient blood, monocytes from healthy individuals with CLL cells, and B-CLL cells alone) and obtained a 14-day gene expression time-course bulk RNAseq of CLL cells. GRNs were inferred for each condition to identify involved genes. Structural analysis and gene set enrichment analysis revealed significant differences in CLL cell responses to macrophages and immune cells, identifying novel transcription factors involved in CLL-immune cell crosstalk.

We applied various analytical tools to the experimental setups including ImpulseDE2, which analyzes continuous changes in gene expression across multiple time points, capturing subtle shifts and trends, and BIODICA, which uses Independent Component Analysis to decompose complex signals/profiles into independent components, representing underlying biological processes. These analyses provide a comprehensive view of CLL cell interactions with immune cells in our in-vitro co-culture system. The integration of multi-omics data using multilayer networks could enhance our understanding of CLL by jointly considering several data types to understand the complex dynamics of cancer-immune cell interactions and identify novel therapeutic strategies. This approach highlights the power of integrating multiple analytical tools to enhance understanding of cancer biology in the context of the tumor microenvironment.

Presented by Hugo Chenel, PhD student in computational biology