A foundational atlas of autism protein interactions reveals molecular convergence

Convergence of autism proteins at the cilium

Hundreds of high confidence autism genes have been identified, yet the relevant etiological mechanisms remain unclear. Gene ontology analyses have repeatedly identified enrichment of proteins with annotated functions in gene expression regulation and neuronal communication. However, proteins are often pleiotropic and these annotations are inherently incomplete. Our recent autism functional genetics work has suggested that these genes may share a common mechanism at the cilium, a membrane-bound organelle critical for neurogenesis, brain patterning, and neuronal activity-all processes strongly implicated in autism. Moreover, autism commonly co-occurs with conditions that are known to involve ciliary-related pathologies, including congenital heart disease, hydrocephalus, and blindness. However, the role of autism genes at the cilium has not been systematically investigated. Here we demonstrate that autism proteins spanning disparate functional annotations converge in expression, localization, and function at cilia, and that patients with pathogenic variants in these genes have cilia-related co-occurring conditions and biomarkers of disrupted ciliary function. This degree of convergence among genes spanning diverse functional annotations strongly suggests that cilia are relevant to autism, as well as to commonly co-occurring conditions, and that this organelle should be explored further for therapeutic potential.

A scalable, high-throughput neural development platform identifies shared impact of ASD genes on cell fate and differentiation

Background

Deleterious mutations in hundreds of genes confer high risk for neurodevelopmental disorders (NDDs), posing significant challenges for therapeutic development. Identifying convergent pathways shared across NDD genes could reveal high-impact therapeutic targets.

Methods

To identity convergent pathways in NDD genes, we optimized Perturb-seq, a method combining CRISPR perturbation with single-cell RNA sequencing (scRNA-seq), and applied structural topic modeling (STM) to simultaneously assess impact on cell fate and developmental stage. We then studied a subset of autism spectrum disorder (ASD) genes implicated in regulation of gene expression using these improved molecular and analytical approaches.

Results

Results from targeting 60 high-confidence ASD risk genes revealed significant effects on neural development. As expected, ASD risk genes impacted both progenitor fate and/or neuronal differentiation. Using STM, we could identify latent topics jointly capturing cell types, cell fate, and differentiation stages. Repression of ASD risk genes led to changes in topic proportions and effects of four genes (DEAF1, KMT2A, MED13L, and MYT1L) were validated in an independent dataset.

Conclusions

Our optimized Perturb-seq method, combined with a novel analytical approach, provides a powerful, cost-effective framework for uncovering convergent mechanisms among genes involved in complex neurodevelopmental processes. Application of these methods advanced understanding of the impact of ASD mutations on multiple dimensions of neural development, and provides a framework for a broader examination of the function of NDD risk genes.

I'm Walking into Spiderwebs: Making Sense of Protein-Protein Interaction Data

Protein-protein interactions (PPIs) are at the heart of the molecular landscape permeating life. Proteomics studies can explore this protein interaction landscape using mass spectrometry (MS). Thanks to their high sensitivity, mass spectrometers can easily identify thousands of proteins within a single sample, but that same sensitivity generates tangled spiderwebs of data that hide biologically relevant findings. So, what does a researcher do when she finds herself walking into spiderwebs? In a field focused on discovery, MS data require rigor in their analysis, experimental validation, or a combination of both. In this Review, we provide a brief primer on MS-based experimental methods to identify PPIs. We discuss approaches to analyze the resulting data and remove the proteomic background. We consider the advantages between comprehensive and targeted studies. We also discuss how scoring might be improved through AI-based protein structure information. Women have been essential to the development of proteomics, so we will specifically highlight work by women that has made this field thrive in recent years.

Predictomes: A classifier-curated database of AlphaFold-modeled protein-protein interactions

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 Structure Prediction and Omics informed Classifier 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.

Autism genes converge on microtubule biology and RNA-binding proteins during excitatory neurogenesis

Recent studies have identified over one hundred high-confidence (hc) autism spectrum disorder (ASD) genes. Systems biological and functional analyses on smaller subsets of these genes have consistently implicated excitatory neurogenesis. However, the extent to which the broader set of hcASD genes are involved in this process has not been explored systematically nor have the biological pathways underlying this convergence been identified. Here, we leveraged CROP-Seq to repress 87 hcASD genes in a human in vitro model of cortical neurogenesis. We identified 17 hcASD genes whose repression significantly alters developmental trajectory and results in a common cellular state characterized by disruptions in proliferation, differentiation, cell cycle, microtubule biology, and RNA-binding proteins (RBPs). We also characterized over 3,000 differentially expressed genes, 286 of which had expression profiles correlated with changes in developmental trajectory. Overall, we uncovered transcriptional disruptions downstream of hcASD gene perturbations, correlated these disruptions with distinct differentiation phenotypes, and reinforced neurogenesis, microtubule biology, and RBPs as convergent points of disruption in ASD.