A γ-polyglutamic acid-based artificial biofilm for field delivery of Bacillus amyloliquefaciens KNU-28: an uncontrolled multi-orchard demonstration on peach gummosis with exploratory bark fungal community profiling

Peach gummosis, caused primarily by Botryosphaeria dothidea sensu lato , is a destructive canker disease with few environmentally friendly management options. Here we report an uncontrolled pre/post field demonstration of a γ-polyglutamic acid-based artificial biofilm — a biodegradable delivery platform whose agent-agnostic design remains untested — encapsulating Bacillus amyloliquefaciens KNU-28 across 143 trees in 18 commercial peach orchards in South Korea. Severity declined from a pre-treatment mean of 4.12 (April) to 2.32 by August (76.2% of 143 trees improved by ≥1 point) and to 1.78 by October in three regions (88.3% of 128 trees). In an ordinal cumulative-link mixed model, severity fell below the April baseline at every later visit; the average monthly odds ratio was 0.70 (95% CI 0.66–0.75); alternative model specifications agreed. Operator-independent image-based ΔE quantification across 74 trees showed a concordant cohort-level decline (lesion-area declines of 19–30%; significant in one of four regions after FDR correction) but agreed only weakly with the ordinal score at tree level (Spearman ρ = 0.16), and is complementary rather than confirmatory. We sequenced fungal ITS2 from 23 bark samples. Multivariate dispersion differed among bark groups (PERMDISP F = 20.7, p < 0.001), and healthy and gummosis-affected bark did not separate under compositionally aware metrics (Aitchison/robust CLR, p > 0.34); the nominally significant Bray–Curtis PERMANOVA (R 2 = 0.248, p = 0.001) therefore cannot be read as a centroid difference. Early- and later-treatment diseased communities did not differ (p = 0.542; all genera FDR q > 0.5). Gummosis-enrolled bark showed higher relative abundances of Wickerhamomyces , Cytospora and Botryosphaeria (dominant ASVs assigned to B. qingyuanensis ) than healthy bark, though none survived FDR correction. Because the 19 diseased libraries came from only 10 trees, the genus–severity correlations were replaced by a paired analysis, in which no genus differed nominally. Together, this demonstration establishes the operational feasibility of field-scale application and a parameter-transparent lesion-quantification method; the severity decline and mycobiota trends are accompanying observations. Because untreated controls, strain-level KNU-28 tracking, verification of delivery to bark and a powered microbiome cohort were absent, the data do not establish causal efficacy and warrant randomized, untreated-control trials.