CDC45-MCM2-7-GINS (CMG) helicase assembly is the central event in eukaryotic replication initiation. In yeast, a multi-subunit “pre-loading complex” (pre-LC) accompanies GINS to chromatin-bound MCM2-7, leading to CMG formation. Here, we report that DONSON, a metazoan protein mutated in microcephalic primordial dwarfism, is required for CMG assembly in vertebrates. Using AlphaFold to screen for protein-protein interactions followed by experimental validation, we show that DONSON scaffolds a vertebrate pre-LC containing GINS, TOPBP1, and DNA pol ε. Our evidence suggests that DONSON docks the pre-LC onto MCM2-7, delivering GINS to its binding site in CMG. A patient-derived DONSON mutation compromises CMG assembly and recapitulates microcephalic dwarfism in mice. These results unify our understanding of eukaryotic replication initiation, implicate defective CMG assembly in microcephalic dwarfism, and illustrate how in silico protein-protein interaction screening accelerates mechanistic discovery.
Protein-protein interactions (PPIs) are ubiquitous in biology, yet a comprehensive structural characterization of the PPIs underlying cellular processes is lacking. AlphaFold-Multimer (AF-M) has the potential to fill this knowledge gap, but standard AF-M confidence metrics do not reliably separate relevant PPIs from an abundance of false positive predictions. To address this limitation, we used machine learning on curated datasets to train a structure prediction and omics-informed classifier (SPOC) that effectively separates true and false AF-M predictions of PPIs, including in proteome-wide screens. We applied SPOC to an all-by-all matrix of nearly 300 human genome maintenance proteins, generating ∼40,000 predictions that can be viewed at predictomes.org, where users can also score their own predictions with SPOC. High-confidence PPIs discovered using our approach enable hypothesis generation in genome maintenance. Our results provide a framework for interpreting large-scale AF-M screens and help lay the foundation for a proteome-wide structural interactome.
Summary: Protein-protein interactions (PPIs) are ubiquitous in biology, yet a comprehensive structural characterization of the PPIs underlying biochemical processes is lacking. Although AlphaFold-Multimer (AF-M) has the potential to fill this knowledge gap, standard AF-M confidence metrics do not reliably separate relevant PPIs from an abundance of false positive predictions. To address this limitation, we used machine learning on well curated datasets to train a S tructure P rediction and O mics informed C lassifier called SPOC that shows excellent performance in separating true and false PPIs, including in proteome-wide screens. We applied SPOC to an all-by-all matrix of nearly 300 human genome maintenance proteins, generating ∼40,000 predictions that can be viewed at predictomes.org, where users can also score their own predictions with SPOC. High confidence PPIs discovered using our approach suggest novel hypotheses in genome maintenance. Our results provide a framework for interpreting large scale AF-M screens and help lay the foundation for a proteome-wide structural interactome.
Early or sorting endosomes are dynamic organelles that play key roles in proteome control by triaging plasma membrane proteins for either recycling or degradation in the lysosome1,2. These events are coordinated by numerous transiently associated regulatory complexes and integral membrane components that contribute to organelle identity during endosome maturation3. Although a subset of the several hundred protein components and cargoes known to associate with endosomes have been studied at the biochemical and/or structural level, interaction partners and higher-order molecular assemblies for many endosomal components remain unknown. Here, we combine crosslinking and native gel mass spectrometry4–7 of purified early endosomes with AlphaFold8,9 and computational analysis to create a systematic human endosomal structural interactome. We present 229 structural models for endosomal protein pairs and additional higher-order assemblies supported by experimental crosslinks from their native subcellular context, suggesting structural mechanisms for previously reported regulatory processes. Using induced neurons, we validate two candidate complexes whose interactions are supported by crosslinks and structural predictions: TMEM230 as a subunit of ATP8 and ATP11 lipid flippases10 and TMEM9 and TMEM9B as subunits of the chloride–proton antiporters CLCN3, CLCN4 and CLCN5 (ref. 11). This resource and its accompanying structural network viewer provide an experimental framework for understanding organellar structural interactomes and large-scale validation of structural predictions.
Summary: Protein-protein interactions (PPIs) drive virtually all biological processes, yet most PPIs have not been identified and even more remain structurally unresolved. We developed a two-step computational screen for human PPIs. First, a classifier called KIRC (Knowledge-Informed Rapid Classifier), trained on biological features, was used to rank all 200 million possible protein pairs in the human proteome by their interaction likelihood. Second, the ∼1.6 million top-ranked KIRC pairs were subjected to structure prediction by AlphaFold-Multimer and ranked using SPOC (Structure Prediction and Omics Classifier), which identifies functional predictions based on biological and structural features. This pipeline revealed 16,000 high-confidence PPIs (∼90% precision), of which more than 5,000 were not previously recognized and more than 12,000 have not been structurally resolved. We use this “predictome” to formulate new hypotheses in different areas of biology, reinterpret low-resolution cryo-EM maps, and identify and validate novel PPIs that may support replication-coupled chromatin assembly. The predicted PPIs, viewable at predictomes.org , are expected to accelerate characterization of the molecular interactions that underlie vertebrate cell physiology.
A multi-protein system called the divisome promotes bacterial division. This apparatus synthesizes the peptidoglycan (PG) cell wall layer that forms the daughter cell poles and protects them from osmotic lysis. In the model Gram-negative bacterium Escherichia coli , PG synthases called class A penicillin-binding proteins (aPBPs) have been proposed to play crucial roles in division. However, there is limited experimental support for aPBPs playing a specialized role in division that is distinct from their general function in the expansion and fortification of the PG matrix. Here, we present in situ cryogenic electron tomography data indicating that the aPBP-type enzyme PBP1b is required to produce a wedge-like density of PG at the division site. Furthermore, atomic force and live cell microscopy showed that loss of this structure weakens the division site and renders it susceptible to lysis. Surprisingly, we found that the lipoprotein activator LpoB needed to promote the general function of PBP1b was not required for normal division site architecture or its integrity. Additionally, we show that of the two PBP1b isoforms produced in cells, it is the one with an extended cytoplasmic N-terminus that functions in division, likely via recruitment by the FtsA component of the divisome. Altogether, our results demonstrate that PBP1b plays a specialized, LpoB-independent role in E. coli cell division involving the biogenesis of a PG structure that prevents osmotic rupture. The conservation of aPBPs with extended cytoplasmic N-termini suggests that other Gram-negative bacteria may use similar mechanisms to reinforce their division site.