Comprehensive genome-wide analysis for the safety assessment of microbial biostimulants in agricultural applications

Abstract

Microbial biostimulants (MBs) offer a sustainable approach to agriculture by helping to reduce reliance on synthetic fertilizers. However, as MBs are intentionally released into the environment, their safety should be rigorously assessed. While taxa with qualified presumption of safety (QPS) benefit from established safety indications, non-QPS taxa lack such guidance. To address this gap, we propose a pipeline combining whole genome sequencing (WGS) and extensive literature search (ELS) data to evaluate microbial safety. We analysed public genomes of three QPS species (Rhodopseudomonas palustris, Bacillus velezensis, Priestia megaterium) and four non-QPS genera (Arthrobacter, Azotobacter, Azospirillum, Herbaspirillum), screening them for virulence factors (VFs), antimicrobial resistance (AMR) genes and mobile genetic elements (MGEs). Results confirmed the safety of QPS taxa, revealing no VFs and only a few intrinsic and non-clinically relevant AMRs. Among non-QPS taxa, VF hits were more prevalent in Azotobacter and Azospirillum spp., though they were mostly related to beneficial plant interactions rather than pathogenicity. AMR genes in non-QPS taxa were primarily associated with efflux pumps or were sporadically distributed. Notably, the only genus-wide pattern observed was that most Azospirillum and Herbaspirillum genomes harboured chromosomally encoded β-lactamases sharing similar genetic structures; however, the detected β-lactamase (bla) genes were distantly related to clinically relevant bla variants, and the absence of MGEs suggests a low risk of horizontal gene transfer, indicating the overall safety of these genera. In general, this WGS-ELS framework provides a robust tool for assessing the safety of non-QPS MBs, supporting regulatory decision-making and ensuring their safe use in sustainable agriculture while safeguarding public health.

Keywords: antimicrobial resistance; horizontal gene transfer (HGT); plant growth-promoting rhizobacteria (PGPR); risk analysis; safety evaluation; sustainable agriculture; virulence factors.

Fig. 1.. Framework comparison of the four main steps used to assign the QPS status to microbial strains (left) and the WGS–ELS approach for microbial safety assessment. Two customizable steps (highlighted in red boxes) allow for adjustments based on the genomic analysis findings, such as lowering detection thresholds or expanding database searches, to ensure comprehensive identification of important safety aspects.

Fig. 2.. Phylogenetic classification of the 151 non-QPS bacterial genomes analysed based on their species-level phylogenetic relationships: (a) Azotobacter (27), (b) Azospirillum (78) and (c) Herbaspirillum (46). Colour separation indicates the clustering of genomes according to the species assigned following fastANI results. Numbers reported at the nodes are 1–100 bootstrap values.

Fig. 3.. Bubble plot representing the VF (top) and AMR (bottom) hits in Azotobacter spp. The X-axis indicates the percentage identity, and the Y-axis shows the percentage coverage. Bubbles are coloured along a red-to-blue or yellow-to-green gradient based on the percentage of hits for AMRs or VFs, respectively. Red dashed lines represent the recommended thresholds at 80% identity and 70% coverage.

Fig. 4.. Bubble plot representing the VF (top) and AMR (bottom) hits in Azospirillum spp. The X-axis indicates the percentage identity, and the Y-axis shows the percentage coverage. Bubbles are coloured along a red-to-blue or yellow-to-green gradient based on the percentage of hits for AMRs or VFs, respectively. Red dashed lines represent the recommended thresholds at 80% identity and 70% coverage.

Fig. 5.. Bubble plot representing the VF (top) and AMR (bottom) hits in Herbaspirillum spp. The X-axis indicates the percentage identity, and the Y-axis shows the percentage coverage. Bubbles are coloured along a red-to-blue or yellow-to-green gradient based on the percentage of hits for AMRs or VFs, respectively. Red dashed lines represent the recommended thresholds at 80% identity and 70% coverage.

Fig. 6.. β-lactamase genes phylogenetic tree of sequences identified in (a) Azospirillum and (b) Herbaspirillum genomes. Clinically relevant β-lactamases are included in red for comparison. The branch lengths measure similarity differences in bla sequences at the amino acid level. Numbers reported at the nodes are 1–100 bootstrap values.