Mutations in homologous recombination (HR) genes, including BRCA1 , BRCA2 , and the RAD51 paralog RAD51C , predispose to tumorigenesis and sensitize cancers to DNA-damaging agents and poly(ADP ribose) polymerase inhibitors. However, ∼800 missense variants of unknown significance have been identified for RAD51C alone, impairing cancer risk assessment and therapeutic strategies. Here, we interrogated >50 RAD51C missense variants, finding that mutations in residues conserved with RAD51 strongly predicted HR deficiency and disrupted interactions with other RAD51 paralogs. A cluster of mutations was identified in and around the Walker A box that led to impairments in HR, interactions with three other RAD51 paralogs, binding to single-stranded DNA, and ATP hydrolysis. We generated structural models of the two RAD51 paralog complexes containing RAD51C, RAD51B-RAD51C-RAD51D-XRCC2 and RAD51C-XRCC3. Together with our functional and biochemical analyses, the structural models predict ATP binding at the interface of RAD51C interactions with other RAD51 paralogs, similar to interactions between monomers in RAD51 filaments, and explain the failure of RAD51C variants in binding multiple paralogs. Ovarian cancer patients with variants in this cluster showed exceptionally long survival, which may be relevant to the reversion potential of the variants. This comprehensive analysis provides a framework for RAD51C variant classification. Importantly, it also provides insight into the functioning of the RAD51 paralog complexes.

DNA double-strand break (DSB) repair is initiated by DNA end resection. CtIP acts in short-range resection to stimulate MRE11–RAD50–NBS1 (MRN) to endonucleolytically cleave 5′-terminated DNA to bypass protein blocks. CtIP also promotes the DNA2 helicase–nuclease to accelerate long-range resection downstream from MRN. Here, using AlphaFold2, we identified CtIP-F728E-Y736E as a separation-of-function mutant that is still proficient in conjunction with MRN but is not able to stimulate ssDNA degradation by DNA2. Accordingly, CtIP-F728E-Y736E impairs physical interaction with DNA2. Cellular assays revealed that CtIP-F728E-Y736E cells exhibit reduced DSB-dependent chromatin-bound RPA, impaired long-range resection, and increased sensitivity to DSB-inducing drugs. Previously, CtIP was shown to be targeted by PLK1 to inhibit long-range resection, yet the underlying mechanism was unclear. We show that the DNA2-interacting region in CtIP includes the PLK1 target site at S723. The integrity of S723 in CtIP is necessary for the stimulation of DNA2, and phosphorylation of CtIP by PLK1 in vitro is consequently inhibitory, explaining why PLK1 restricts long-range resection. Our data support a model in which CDK-dependent phosphorylation of CtIP activates resection by MRN in S phase, and PLK1-mediated phosphorylation of CtIP disrupts CtIP stimulation of DNA2 to attenuate long-range resection later at G2/M.

The InterEvDock3 protein docking server exploits the constraints of evolution by multiple means to generate structural models of protein assemblies. The server takes as input either several sequences or 3D structures of proteins known to interact. It returns a set of 10 consensus candidate complexes, together with interface predictions to guide further experimental validation interactively. Three key novelties were implemented in InterEvDock3 to help obtain more reliable models: users can (i) generate template-based structural models of assemblies using close and remote homologs of known 3D structure, detected through an automated search protocol, (ii) select the assembly models most consistent with contact maps from external methods that implement covariation-based contact prediction with or without deep learning and (iii) exploit a novel coevolution-based scoring scheme at atomic level, which leads to significantly higher free docking success rates. The performance of the server was validated on two large free docking benchmark databases, containing respectively 230 unbound targets (Weng dataset) and 812 models of unbound targets (PPI4DOCK dataset). Its effectiveness has also been proven on a number of challenging examples. The InterEvDock3 web interface is available at http://bioserv.rpbs.univ-paris-diderot.fr/services/InterEvDock3/.

The revolution brought about by AlphaFold2 and the performance of AlphaFold2-Multimer open promising perspectives to unravel the complexity of protein-protein interaction networks. Nevertheless, the analysis of interaction networks obtained from proteomics experiments does not systematically provide the delimitations of the interaction regions. This is of particular concern in the case of interactions mediated by intrinsically disordered regions, in which the interaction site is generally small. Using a dataset of protein-peptide complexes involving intrinsically disordered protein regions that are non-redundant with the structures used in AlphaFold2 training, we show that when using the full sequences of the proteins involved in the interaction networks, AlphaFold2-Multimer only achieves 40% success rate in identifying the correct site and structure of the interface. By delineating the interaction region into fragments of decreasing size and combining different strategies for integrating evolutionary information, we managed to raise this success rate up to 90%. Beyond the correct identification of the interaction site, our study also explores specificity issues. We show the advantages and limitations of using the AlphaFold2 confidence score to discriminate between alternative binding partners, a task that can be particularly challenging in the case of small interaction motifs.

The biological function of natural non-coding RNAs (ncRNA) is tightly bound to their molecular structure. Sequence analyses such as multiple sequence alignments (MSA) are the bread and butter of bio-molecules functional analysis; however, analyzing sequence and structure simultaneously is a difficult task. In this work, we propose CARNAGE (Clustering/Alignment of RNA with Graph-network Embedding), which leverages a graph neural network encoder to imprint structural information into a sequence-like embedding; therefore, downstream sequence analyses now account implicitly for structural constraints. In contrast to the traditional "supervised" alignment approaches, we trained our network on a masking problem, independent from the alignment or clustering problem. Our method is very versatile and has shown good performances in 1) designing RNAs sequences, 2) clustering sequences, and 3) aligning multiple sequences only using the simplest Needleman and Wunsch's algorithm. Not only can this approach be readily extended to RNA tridimensional structures, but it can also be applied to proteins.