Taxonomic filtering accompanies functional expansion during long-term soil restoration

The restoration of species-rich calcareous grasslands is a critical conservation objective, yet the recovery of the invisible below-ground microbiome remains poorly quantified compared to above-ground vegetation. Using a unique 143-year land-use chronosequence on Salisbury Plain, UK, we investigated the trajectory of ecosystem reassembly across arable, regenerating (23 and 67 years), and ancient grasslands. By integrating vegetation surveys with soil physiochemistry, microbial profiling, and shotgun metagenomics, we identified a decoupling between floral and edaphic recovery. While the diversity of vegetation recovered relatively rapidly, approaching ancient grassland levels within 23–67 years, soil properties exhibited persistent legacy effects and slow convergence. Bacterial richness decreased with restoration age; this taxonomic contraction was conversely matched by an expansion in inferred metagenomic functional potential. This was reflected in increased functional gene richness and shifts in the relative abundance of specific SEED-annotated functions toward metabolic pathways associated with complex carbon cycling and stress tolerance. These shifts were congruent with the emergence of specific, unnamed genera belonging to Pseudomonadota and Actinomycetota, and the Bacillota species Pristimantibacillus. The soil ecosystem remained distinct from the 143-year stage even after 67 years of recovery, characterised by persistent legacy phosphorus and a slow accumulation of soil organic matter. These findings suggest that passive regeneration alone may be insufficient for full soil functional recovery, and that strategies targeting microbial assembly and long-term carbon dynamics warrant further evaluation.