Climate-induced range shifts support local plant diversity but don’t reduce extinction risk

Using the BioShifts database of nearly 15,000 observed plant range shift records spanning over 6,500 species, researchers modeled species-specific movement velocities and combined them with habitat suitability projections, climate models, and 6.8 million occurrence records to forecast where roughly 18% of known vascular plant species will survive through 2081–2100. The central finding is that 7–16% of modeled species face losing more than 90% of their range, with climate-driven habitat loss—not dispersal failure—responsible for 70–80% of those losses. While range shifts are projected to boost local species richness across about 28% of land surface and stabilize diversity in tropical and subtropical zones, high-latitude regions above 50°N face sharp richness declines because warming outpaces plant movement. The study concludes that facilitating migration will help local communities reorganize but will do little to curb global extinctions; protecting climate refugia and strengthening seed banks and botanical gardens are likely more effective conservation priorities.

Read the original article →

Human DHX29 detects nonoptimal codon usage to regulate mRNA stability

Cells use synonymous codons—different nucleotide triplets encoding the same amino acid—at varying frequencies, and "nonoptimal" codons are associated with faster mRNA degradation, but how human cells detect and act on this signal has been unclear. Using genome-wide CRISPR screening, researchers identified the RNA-binding protein DHX29 as a key player in codon-dependent gene regulation. Cryo-electron microscopy and selective ribosome profiling showed that DHX29 binds at the A-site entrance of the translating ribosome, precisely where aminoacyl-tRNA complexes are sampled during elongation. Proteomic experiments further revealed that DHX29 recruits the GIGYF2•4EHP repressor complex, which broadly suppresses the stability of mRNAs enriched in nonoptimal codons. Together, these findings reveal a direct molecular pathway connecting codon usage to mRNA decay in human cells, with implications for understanding how codon optimization shapes protein output in health and disease.

Read the original article →

Rapid directed evolution guided by protein language models and epistatic interactions

Improving proteins through directed evolution is hampered by the enormity of possible mutation combinations and the cost of exploring them experimentally. This work introduces MULTI-evolve, a framework that uses protein language models or existing functional data combined with epistatic modeling to predict which groups of mutations will work synergistically. A companion molecular assembly method called MULTI-assembly overcomes a key bottleneck by efficiently constructing multi-mutation variants across sequences spanning multiple kilobases, sidestepping the length limitations of conventional gene synthesis. Testing the approach on three different proteins produced up to tenfold functional improvements within a single round of machine learning–guided evolution. By integrating computational prediction with efficient library construction, MULTI-evolve offers a streamlined end-to-end pipeline for engineering diverse protein types without requiring large pre-existing datasets or iterative experimental cycles.

Read the original article →