Multi-kingdom gut microbiota analyses define bacterial-fungal interplay in multiple type 2 diabetes cohorts.

The role of the gut microbiome in type 2 diabetes (T2D) remains incompletely defined, particularly across microbial kingdoms and diverse populations. Here, we conducted a meta-analysis of 3,857 fecal metagenomes from six international cohorts, profiling bacteria, fungi, archaea, and viruses. Using supervised machine-learning models trained on harmonized multi-kingdom profiles with cross-cohort validation, we identified conserved alterations in T2D, characterized by reduced bacterial and viral diversity and increased fungal and archaeal diversity. A cross-kingdom panel of 33 microbial markers derived from these models achieved robust diagnostic performance (AUR-OC=0.82), outperforming single-kingdom models. Notably, Saccharomyces cerevisiae was consistently depleted in T2D and inversely correlated with glycemic indices. In mice, oral S. cerevisiae supplementation improved glucose tolerance and insulin sensitivity while reducing the abundance of Eggerthella lenta and Klebsiella pneumoniae, bacterial taxa previously linked to adverse metabolic and inflammatory phenotypes. Together, our findings highlight the diagnostic value and mechanistic relevance of multi-kingdom microbial signatures in T2D and position S. cerevisiae as a potential fungal probiotic candidate for metabolic intervention.