Sialylated CD43 forms a glyco-immune barrier that restrains antileukemic immunity

Macrophages can destroy cancer cells through phagocytosis, yet therapies designed to boost this process have failed to improve outcomes in acute myeloid leukemia (AML). To uncover what governs phagocytosis in this context, researchers conducted CRISPR knockout screens in human AML cells grown alongside human macrophages. The well-known "don't eat me" signal CD47 blocked phagocytosis by mouse macrophages but not human ones. Instead, O-linked glycosylation and sialylation emerged as the dominant suppressors, with the mucin-like surface protein CD43 acting as the primary mediator in AML. CD43's inhibitory effect depended on the length of its extracellular domain but not on known sialic acid receptors SIGLEC-1, -7, or -9. Beyond macrophages, CD43 also suppressed natural killer and T cell responses, indicating it functions as a broader glyco-immune barrier that blunts both innate and adaptive antileukemic immunity—making it a potential therapeutic target.

Read the original article →

How artificial intelligence is reengineering protein engineering

Protein engineering has long relied on computational modeling and high-throughput laboratory experiments to identify proteins with useful properties across medicine, agriculture, and industry. This review examines how artificial intelligence is accelerating that search by navigating the vast space of possible protein sequences more efficiently. The authors survey several categories of AI-driven progress: generative models that design sequences, backbone structures, and atomic configurations; methods for fine-tuning those general models toward proteins with specific functions; approaches that extract useful protein representations or score candidate sequences; and library design strategies that account for the practical constraints of synthesis. Throughout, the review frames these diverse techniques within a shared statistical perspective, arguing that this unifying lens clarifies how modern AI methods relate to one another and to the broader goals of protein engineering.

Read the original article →

Luminal surface proteome of the brain vasculature uncovers blood-brain barrier regulators

Researchers developed an in vivo method to profile proteins on the inner surface of blood vessels, applicable across vertebrates. Applying quantitative mass spectrometry to mouse brain vasculature, they mapped how this luminal surface proteome changes from early development through aging. Genetic perturbation experiments revealed that the arginine transporter SLC7A1 and nitric oxide synthase NOS3 are required for blood-brain barrier integrity only in neonates, while the hyaluronan-degrading enzyme HYAL2 maintains barrier function across the entire lifespan. The findings connect nitric oxide and hyaluronan metabolism to blood-brain barrier regulation, offering a richer molecular picture of how vascular barrier properties are established and maintained over time. This proteomic resource and the regulatory proteins it highlights could inform future strategies for targeting the blood-brain barrier in neurological disease.

Read the original article →

Emergent predictability in microbial ecosystems

Microbial ecologists have long suspected that simple, large-scale patterns can arise from complex communities, but this idea has lacked a rigorous framework. Researchers here formally defined "emergent predictability" as the degree to which grouping individual microbial strains into broader functional categories improves predictions of community behavior. Analyzing two published datasets, they found that coarser descriptions actually became *more* predictive as species richness increased — a result that cannot be explained merely by statistical averaging across large communities. Instead, the analysis suggests that physiological or environmental feedback mechanisms suppress noise along certain axes of community variation, making those axes increasingly informative as diversity grows. The work provides a quantitative foundation for understanding why simplified ecological descriptions sometimes work better in complex communities, with implications for modeling and managing microbiomes in health, agriculture, and environmental contexts.

Read the original article →

Nutritional specialization and social evolution in woodroaches and termites

Researchers investigated how the cooperative social behaviors seen in woodroaches and termites emerged from solitary cockroach ancestors that shifted to feeding on nutrient-poor deadwood. By sequencing genomes from eight new Blattodea species, they traced stepwise genomic reductions accompanying this dietary transition. In woodroaches, genes governing oxidative phosphorylation and peroxisome function were inactivated, which aligns with their characteristically slow offspring development. Termites lost genes essential for sperm motility, supporting the idea that monogamous pair-founding was a prerequisite for reproductive division of labor. Termites also repurposed genes from signaling pathways sensitive to nutritional status — including juvenile hormone, insulin, EGFR, and Dpp — to regulate caste development. Most larvae become workers through early high-energy metabolism, while future reproductives ramp up energy metabolism genes later in development. Together these findings explain how obligate worker provisioning and feedback mechanisms enabled the evolution of large, stable termite colonies.

Read the original article →