Mots-Clés

RNA Motifs, translation regulation, translation factors, Motif Finding Algorithms, Probabilistic approach, machine learning, deep learning

Description

Motivation :

Translation of proteins in the sea urchin embryo is stimulated upon fertilization, and is necessary for cell cycle progression and development. The translational activation occurs in an mTOR (mechanistic Target of Rapamycin) dependent manner and involves the assembly of the canonical eIF4 initiation complex on the cap structure of maternally stored mRNAs (review in Pontheaux et al, 20211).
By a polysome profiling pangenomic approach, we have identified a pool of mRNAs that is specific all recruited into polysomes in response to fertilization (Chassé et al, 20182), and we are currently analyzing the translatome in presence of PP242 mTOR inhibitor. The set of translated mRNAs encodes for proteins implicated in different regulatory circuits, such as those controlling cell cycle regulation or RNA binding proteins. Translation of most mRNAs showed sensitivity to PP242 inhibitor, indicative of recruitment via the canonical initiation step as expected, however a fraction of the recruited mRNAs is still addressed to the polysomes in the presence of PP242, suggesting that they could be translated through a non-canonical initiation route.

It is known that mTOR plays a fundamental role in the control of mRNA translation, particularly for the class of transcripts that code for all ribosomal proteins and many translation factors, which bear a 5ʹterminal oligopyrimidine sequence (TOP motif). In sea urchin, despite the activation of the mTOR pathway upon fertilization, the transcripts coding the ribosomal proteins and translation factors were significantly under-represented in the translated set of mRNAs, and were shown to be mainly untranslated. Moreover, many mTOR-regulated mRNAs lack classical TOP sequences (Hsieh et al, 2012; Thoreen et al, 2012; Philippe et al, 2023). While inhibition of the mTOR pathway suppresses the translation of a majority of the mRNAs translated at fertilization, the mechanisms driving this regulation may rely on so far unknown RNA motifs.

Objectifs :

The aim of this stage is to identify new RNA motifs elements within co-translated mRNAs. Translation of a given mRNA can be significantly influenced by mRNA sequence features located in either the 5’UnTranslated Region (secondary structures, upstream open reading frames, internal ribosome entry sites, etc.) or the 3’ UTR (cytoplasmic polyadenylation elements, RNA-binding
motifs, miRNA binding sites, etc.).

Methodology :

We hypothesize that co-translated mRNAs are likely to share elements that allow the co-regulation of these specific mRNA cohorts. Here we will focus on conserved sequence motifs.  We shall use machine learning computational approaches for identifying enriched consensus motifs in the translatome data sets. The consensus motifs identified will then be tested for their ability to drive translation of a reporter mRNA microinjected into unfertilized sea urchin eggs and analyzed following fertilization in normal and
PP242-treated embryos. Furthermore, the consensus motif will be specifically inhibited by co-injection with a targeted morpholino.

We will use the translatomes produced in our group at the egg-to embryo transition (Chassé et al, 2018) and in response to mTOR inhibitor PP242 (unpublished results), to generate pools of mRNA sequences according to their translational status (mRNAs translated in response to fertilization, mRNAs translated dependently and independently of mTOR). In each pool, we will screen for enriched RNA
motifs using machine learning approaches. Several methods could be explored from the traditional ones (enumeration and probabilistic approaches) to the latest development process of deep learning). For this later, several tools with different architectures are available such as deep learning motif mining based on convolutional neural network (CNN), recurrent neural network (RNN), and transformers (He Y et al, 2021).

Scientific environment :

The intern will work within the ECOMAP  and TCCD teams at the Roscoff Biological Station (CNRS and Sorbonne University). He/she will be supervised by J. Bernardes (Assistant professor) and J. Morales (CNRS researcher). 

Skills :

Linux and Bash

Good programming skills in python or C++

Basic Concepts of Machine Learning and/or deep learning.

Master in bioinformatics, computer science, physics, mathematics, etc.

Autonomy, initiative, organizational skills, good analytical and synthesis skills.

References :

-Pontheaux F et al (2021) Chapter 17 - Echinoderms: Focus on the sea urchin model in cellular and developmental biology. In “Established and Emerging Marine organisms in experimental biology” A. Boutet, B. Schierwater (eds) CRC.
-Chassé H et al (2018) Translatome analysis at the egg-to-embryo transition in sea urchin. Nucleic Acids Res. 46(9), 4607–4621.
-Hsieh AC et al (2012) The translational landscape of mTOR signalling steers cancer initiation and metastasis. Nature 485, 55–61.
-Thoreen CC et al (2012) A unifying model for mTORC1-mediated regulation of mRNA translation. Nature 485, 109–113 ;
-Philippe et al (2020) Global analysis of LARP1 translation targets reveals tunable and dynamic features of 5ʹ TOP motifs. Proc. Natl. Acad. Sci. U.S.A. 117, 5319–5328.
-He Y et al (2021) A survey on deep learning in DNA/RNA motif mining. Briefings in Bioinformatics. 22(4):bbaa229.
-Chassé et al (2019) In vivo analysis of protein translation activity in sea urchin eggs and embryos. Methods in Cell Biol Vol 151:335-352.