Thèse Utilisation d'Outils Computationnels de Troisième Génération pour Comprendre l'Impact sur l'Agressivité Cellulaire des Isoformes d'Épissage Spécifiques de Cancer H/F - Doctorat.Gouv.Fr

Établissement : Université Paris-Saclay GS Life Sciences and Health École doctorale : Cancérologie : Biologie - Médecine - Santé Laboratoire de recherche : Genome, ARN et cancer Direction de la thèse : Reini FERNANDEZ DE LUCO ORCID 0000000290866507 Début de la thèse : 2026-10-01 Date limite de candidature : 2026-05-03T23:59:59 Des avancées majeures ont permis de mieux comprendre comment les programmes transcriptionnels spécifiques des cellules cancéreuses sont régulés, l'épigénétique jouant un rôle central dans l'identité et la mémoire cellulaires. Cependant, ces dernières années, une nouvelle couche de régulation post-transcriptionnelle s'est imposée : l'épissage alternatif. Ce processus, qui consiste à inclure/exclure des exons (ou retenir des introns) lors de la maturation du pré-ARNm, génère une grande diversité d'ARNm et donc de protéines. Il concerne plus de 90% des gènes humains et peut être si spécifique d'un état cellulaire qu'il s'avère parfois plus informatif que l'expression génique pour identifier les patientes atteintes de cancer du sein au plus mauvais pronostic (1). En effet, les gènes différentiellement exprimés et différentiellement épissés se recouvrent rarement, l'épissage alternatif représente donc une couche fonctionnelle complémentaire pouvant ouvrir de nouvelles pistes thérapeutiques.
Afin d'identifier des variants d'épissage spécifiques du cancer susceptibles de réduire l'agressivité tumorale, nous nous concentrons sur la transition épithélio-mésenchymateuse (EMT), au cours de laquelle une cellule épithéliale acquiert motilité, plasticité et capacités invasives, favorisant la métastase (2). L'EMT dépend de changements coordonnés de la chromatine et de l'épissage de gènes clés (3,4). Nous avons montré que cibler des marques d'histones différentiellement enrichies sur des exons régulés pendant l'EMT est nécessaire et suffisant pour induire des changements d'épissage associés à l'EMT (4,5). Or, l'EMT n'est pas binaire mais progressive, et ce sont souvent les états intermédiaires qui présentent la plus forte agressivité en contexte tumoral, posant la question quels sont les événements d'épissage critiques à cibler, leur fonction, et leur régulation pour limiter l'invasivité.
Dans ce projet, nous aborderons ces questions via des approches multi-omiques. En exploitant des données d'épigénomiques et de transcriptomiques (lectures courtes et lectures longues de troisième génération) au cours de la progression de l'EMT dans des cellules humaines MCF10A, ainsi que de lignées cellulaires et de patients du sein, vessie, prostate issus de CCLE et TCGA, nous viserons à : (i) définir une signature d'épissage pan-cancer caractéristique de tumeurs agressives de type EMT ; (ii) prédire l'impact des isoformes pro-métastatiques sur la stabilité, la structure et la localisation des protéines (AlphaFold, tappAS (6), dynamique moléculaire via GROMACS; (iii) tester l'existence d'une coordination d'exons co-inclus/co-exclus au sein d'une même molécule d'ARN, potentiellement guidée par l'organisation de la chromatine (7), et en évaluer les conséquences fonctionnelles ; (iv) identifier des régulateurs communs, avec un focus sur les marques d'histones dans le contrôle de l'épissage (4,5). Ce travail permettra d'identifier des événements d'épissage clés associés à des phénotypes pro-métastatiques et de proposer de meilleures cibles pour réduire métastases et rechutes (8).
Metastasis and resistance to treatment remain the primary causes of cancer-related mortality. A major challenge is to detect aggressive cells before it is too late and to understand how they acquire such invasive phenotypes to target them for therapeutic purposes. Increasing evidence indicates that many tumours gain aggressiveness through epithelial-to-mesenchymal transition (EMT), a cell reprogramming process that promotes cell plasticity and dissemination. However, EMT is not a simple binary process. It comprises multiple intermediate states which often carry the strongest invasive potential, making difficult to define robust markers of invasiveness that work across patients and that can be targeted for therapy.

Interestingly, in this EMT reprogramming and gain of aggressiveness, it is not only about which genes are expressed, but also how they are processed into distinct protein isoforms with different functions. Through this alternative splicing, the cell can expand its functional toolkit, enabling rapid and stimulus-adapted phenotypic changes without requiring major transcriptional remodelling. Many cancer-specific splicing events have been identified, and splicing signatures have proven more effective than gene-expression patterns at stratifying cancers with the poorest prognosis (1). Even more, our recent analyses across breast and bladder cancer, two high-risk cancers that share a basal-like aggressive state, proved that most aggressive cancer cells can cluster together based on their splicing patterns irrespective of their tissue of origin. This supports the existence of a conserved splicing program underlying aggressiveness and makes alternative splicing a particularly promising entry point to understand and intercept metastasis.

Yet two major questions remain unanswered, what is the functional impact of these key splicing events in cancer cells and how are they regulated? Using third generation long-reads RNA-sequencing combined with state-of-the-art predictors of protein structure (AlphaFold) and impact on the protein function and molecular dynamics (TappAS, GROMACS), we aim at elucidating the functional impact of key splicing events for the acquisition of a more invasive/aggressive phenotype in the cell. We will also address whether there is exon coordination relevant for cancer aggressiveness and how this coordination is regulated by splicing and chromatin regulators, which we suspect to be a major regulatory layer of key splicing events relevant for cancer aggressiveness (3,4,5). Finally, to identify the splicing events essential for the acquisition of a pro-metastatic invasive phenotype, we will mine available patients' data and use machine learning classifiers to identify a pan-cancer splicing signature of cancer aggressiveness for then studying the impact of these splicing isoforms in the protein function and cancer cell biology. This will be one of the first comprehensive analysis taking advantage of clinical data to address the functional impact of key splicing events at the protein level and molecular dynamics to better understand the role of splicing in cancer aggressiveness. Aim 1. Identification of the first pan-cancer splicing signature characteristic of EMT-like aggressive tumours based on clinical data.
Using CCLE RNAseqs from breast, bladder and prostate cancer cells (which are related cancer types with the most aggressive ones corresponding to EMT/basal-like type of cancers), we will define the splicing signature that better cluster together most aggressive/invasive cell lines irrespectively of the tissue of origin. Then this signature will be refined using more clinically-relevant and heterogeneous data from TCGA patients from the same cancer types to reclassify the most aggressive ones with the poorest prognosis/resistance to treatment/apparition of metastasis using random forest classifiers. Finally, this novel splicing signature of pan-cancer aggressiveness will be tested on our own cohorts of 100 muscle-invasive bladder cancers (MIBC) and 124 TNBC RNA-seqs available at Curie to prove its clinical/functional value. The splicing events defining aggressive cancers will be crossed with changes in splicing during EMT (available in the team), favouring the ones changing at early stages of EMT progression as a sign of early onset of a pro-metastatic phenotype. This final splicing list will be used as the pro-metastatic splicing events to be studied in Aim 2.

Aim 2. Address the functional impact of pro-metastatic splicing events in the protein structure/function and thus EMT and acquisition of cancer aggressiveness.
There are two important information layers missing in short-reads RNA-sequencing: if there is exon coordination in the inclusion/exclusion of regulated exons in the mature mRNAs and what is the impact in the protein structure/functional domains/expression of the different splicing isoforms. Both questions can only be answered with third-generation long-reads RNA-sequencing in which splicing changes can be addressed at the entire mRNA molecule. Such improved new data has recently be generated in the team on our EMT model system based on normal human MCF10a cells overexpressing inducible SNAIL. Taking advantage of protein structure predictors such as AlphaFold and the long-reads analyser TappAS (6), developed by our collaborator Ana Conesa, we will be able to address the functional impact at the protein level of the splicing isoforms identified in aggressive cancers also changing in EMT from Aim1. Furthermore, we will be able to address whether there are exons that are jointly spliced in or out in a coordinated way in the same mRNA molecule and what are the regulatory layers behind this coordination, with a special attention into chromatin modifications, which we have proved to play a major role in splicing regulation (3-7). We expect chromatin-marked exons to be co-regulated in the same RNA molecule with a special impact on the protein expression levels via NMD, or a change in a specific protein domain impacting RNA binding, nuclear localization, enzymatic activity, etc. Functional modifications driven by splicing isoforms specific of most aggressive cancer types / invasive EMT will be further investigated through molecular dynamics studies that will help catch the mechanistic switches gained or lost by the new proteins' isoforms. These types of analyses will give new insights into protein interaction potential and changes in enzymatic activity. This latter approach has the potential to help define more suitable therapeutic strategies by targeting key protein isoforms with inhibitory drugs specific of the pro-metastatic protein isoform.

Aim 3. Identification of common regulators of the pro-metastatic EMT signature.
Finally, mining ChIP-ATLAS and POSTAR3 CLIP-seq databases, we will look for common chromatin and splicing regulators of pro-metastatic splicing isoforms with a clear impact in the protein function and thus EMT of key alternatively spliced genes. A special interest will be deposited in the role of histone marks and 3D chromatin organization since we have strong evidence of a role for a higher order chromosome organization in alternative splicing regulation of key genes which splicing regulation needs to be coordinated and highly dynamic for an early cell response (two manuscripts in preparation). Results from this project represent the first of its kind, going from splicing regulation to functional impact in the context of metastasis.

- Master ou équivalent
- Très motivé(e)
- Avec une formation en analyses transcriptomiques et épigénomiques - une connaissance en analyse structurelle des protéines serait un plus.
- Curieux(se) (lit la bibliographie)
- Esprit d'initiative (ingénieux(se))
- Autonome
- Polyvalent(e)