Abstract
Diabetic foot ulcers (DFU) are a major complication of type 2 diabetes mellitus, but whether the gut microbiome captures systemic microbial features associated with advanced ulcer severity remains unclear. We performed shotgun metagenomic sequencing of stool samples from 43 patients with type 2 diabetes mellitus and active DFU, comparing Wagner grades 1–3 (
n
= 30) with Wagner 4 gangrenous disease (
n
= 13).
De novo
assembly and binning recovered 440 dereplicated metagenome-assembled genomes (MAGs) meeting medium-quality or high-completeness/low-contamination thresholds. Community-level diversity and dominant-taxon composition did not separate Wagner 4 from Wagner 1–3, indicating that advanced disease was not reflected by broad ecological restructuring. Feature-level analysis instead identified a genome-resolved MAG profile. To prioritize robust candidates, we combined two complementary approaches: random forest (RF) stability selection, which identified 24 MAGs with reproducibly high classification importance across resampled folds, and covariate-adjusted MaAsLin2 differential-abundance testing. Intersecting the results of both approaches prioritized three MAGs supported by each method: one
Escherichia coli
MAG enriched in Wagner 4 and two
Collinsella
MAGs depleted in Wagner 4. This three-MAG signature (out-of-bag AUC = 0.703) retained much of the discriminatory information captured by the broader 24-MAG RF classifier, with concordant, opposing abundance directions across classifier interpretation, differential-abundance testing, and per-MAG abundance distributions. Functional annotation further separated the Wagner 4-enriched
Escherichia coli
from the
Collinsella
MAGs. The
Escherichia coli
MAG carried antibiotic-resistance and virulence-factor signals and encoded respiratory metabolic capacity, whereas the two
Collinsella
MAGs lacked detectable resistance and virulence hits and showed metabolically compact profiles. Exploratory clinical association analysis linked the
E. coli
–
Collinsella
abundance score to longer DFU duration, consistent with a gut microbial correlate of chronic or advanced disease burden. These findings support longitudinal gut metagenomic validation to determine whether this signal tracks DFU progression, treatment response, or recovery.