Only Two Soil Cleanup Approaches Actually Work in the Field, Review Finds

A sweeping assessment of rhizosphere engineering has delivered an unusually blunt verdict on the state of soil remediation science: of the many microbiome-based approaches being promoted to clean contaminated farmland, only two have ever been validated in the field, and neither relies on the engineered organisms that dominate headlines. The mini-review, published in Frontiers in Microbiology by Priya Arunkumar of Karunya Institute of Technology and Sciences, examines how deliberately manipulating the microbial communities around plant roots can be used to remediate agricultural soils contaminated with pesticides, heavy metals, petroleum hydrocarbons and microplastics. Its central contribution is not a new technique but a sorting exercise, assigning every approach in the literature to one of three tiers: field-validated, emerging at pot or microcosm scale, or purely conceptual.

The review’s first field-validated domain is metal immobilization and phytoextraction through bioaugmentation and bulk amendment. The supporting numbers are substantial. Combined plant growth-promoting bacterial inoculants raised straw cadmium extraction rates by 156 percent over uninoculated controls in a field application with the hyperaccumulator Sedum alfredii. A plant growth-promoting yeast applied with Sedum plumbizincicola increased hyperaccumulator biomass by 46 percent and reduced extractable soil cadmium by roughly 29 to 39 percent in field experiments. A single wheat-straw biochar amendment reduced extractable cadmium and lead continuously across three consecutive rice paddy seasons, and a combination of sepiolite and biochar cut maize grain cadmium by 24.6 to 47.5 percent on mine-adjacent farmland. Crucially, the review notes, these interventions fix metals in place or move them into harvestable tissue; they do not destroy the contaminant, because metals cannot be degraded at all.

The second field-validated domain is hydrocarbon rhizoremediation by defined bacterial consortia. A five-strain hydrocarbonoclastic consortium, combined with the tree Azadirachta indica and nitrogen-phosphorus-potassium fertilization, degraded up to 67 percent of total petroleum hydrocarbons over 120 days at two contaminated field sites. Separately, an artificial consortium of plant growth-promoting rhizobacteria and polycyclic aromatic hydrocarbon degraders reduced petroleum hydrocarbons in an in situ field trial in a semiarid environment. Both domains, the author stresses, rest on cultured isolates or bulk amendments such as biochar. Neither depends on genome-edited organisms, synthetic pathways or any of the molecular tools that have come to define the field’s public image.

One tier down sit the approaches demonstrated only in pots, mesocosms or microcosms. Inoculation with arbuscular mycorrhizal fungi combined with biochar reduced maize tissue cadmium by 50 to 76 percent in mesocosm experiments, and a defined bacterial consortium called ECO-M removed 94.3 percent of the insecticide chlorpyrifos from contaminated soil in a microcosm, though mineralization was never quantified and no field trial followed. The review illustrates the gap between such figures and field reality with atrazine, where isotope-resolved work showed mineralization varying from 10 percent to above 40 percent between isolates, with a substantial non-extractable residue persisting in every treatment. Field-scale evidence for these approaches simply does not yet exist.

Microplastics receive their own chapter, and the verdict there is the harshest in the review: rhizosphere-mediated microplastic remediation has never been demonstrated in soil at all. The author argues that the literature systematically conflates five distinct processes. Physical interception by roots and adsorption onto root surfaces merely relocate particles without altering polymer chemistry, so a declining bulk-soil particle count can record redistribution rather than loss. Aging and fragmentation, driven by ultraviolet exposure and oxidation, change the physical state of the polymer and can even raise hazard by increasing particle number, specific surface area and the release of sorbed additives. Microbial transformation is polymer-specific: polyesters with hydrolysable ester bonds are attacked far more readily than polyolefins such as polyethylene and polypropylene, whose inert carbon-carbon backbones require oxidative initiation and degrade orders of magnitude more slowly. Only mineralization, the complete conversion of polymer carbon to carbon dioxide and microbial biomass, constitutes genuine removal, and it is the process least often demonstrated. The most frequently cited degradation figures, 17.3 percent and 7.3 percent particle weight loss over 40 days, came from axenic liquid cultures without soil, plant or rhizosphere, while a 240-day soil incubation detected only weak weight loss with surface oxidation, which is evidence of weathering rather than degradation.

The third tier, labeled conceptual, contains the approaches that attract the most attention. Synthetic biology constructs and CRISPR-based community editing have been demonstrated only in cultured, genetically tractable strains, never in soil. In situ editing of resident soil communities by phage-mediated or conjugative CRISPR delivery remains a proposal, as do microbiome-informed digital twins and artificial intelligence-guided consortium design; a systematic review of digital twins applied to soil quality found no published application in a predictive-control context. The review is careful to distinguish genetic potential from demonstrated activity: detecting degradative genes such as opd, atzA or czc in a metagenome establishes only that a related sequence is present, not that it is expressed, that its host is viable, or that the reaction proceeds at a rate relevant to remediation.

Between the conceptual tier and the field lies a set of biosafety obstacles that the review treats as the principal barrier rather than an afterthought. The rhizosphere is among the more permissive environments for horizontal gene transfer, and direct measurement in fabricated rhizosphere ecosystems has shown plasmid transfer between Pseudomonas putida donors and recipients, and onward to a different root-colonizing genus, at frequencies that vary with plant developmental stage, exudate composition and environmental stress. Soil bacteria also carry large integrative and conjugative elements that can exceed 800 kilobases. For in situ editing strategies this is doubly problematic, because conjugative delivery deliberately exploits the same machinery: a delivery system optimized for transfer efficiency is, by construction, optimized for escape. Containment evidence is similarly sobering. In a model engineered organism carrying a CRISPR-Cas9 kill switch, escape rates measured by abundance of the targeted genes were several orders of magnitude higher than escape rates measured by colony-forming units, because the switch arrests growth while leaving target genes intact and detectable inside intact membranes. Culture-based monitoring, the author concludes, systematically understates persistence, and a containment claim supported only by plate counts is not supported.

Ecological risk extends to unmodified inoculants as well. In low-fertility soil, commercial plant growth-promoting rhizobacteria that failed to colonize at all nonetheless narrowed niche breadth and raised niche overlap among resident taxa by 59.2 to 62.4 percent, intensifying competition without delivering any benefit. By contrast, a native synthetic community assembled from the same soils established successfully and raised the maize root-to-shoot ratio by 78 to 121 percent, against 23 to 86 percent for the commercial product. Seed-delivered synthetic communities can dominate seedling microbiota, contributing on average 80 percent of membership, which establishes that deliberate community installation is achievable. Regulation remains fragmented across national frameworks with no coordinated international policy governing release or post-release surveillance, and the review calls for molecular monitoring that tracks construct abundance rather than viable cell counts alone, plus measurement of transfer to non-target taxa wherever mobilizable elements are used.

Practical deployment faces further hurdles that have little to do with molecular biology. Current carrier systems for inoculants, including peat, alginate beads and wettable powders, suffer transport and retention problems under field-relevant stress, and their shelf life generally falls short of the six to twelve months regarded as necessary for commercialization, though the review cautions that shelf life is strain-, carrier- and storage-specific. Contaminant concentrations vary over meters in real fields, so apparent reductions may reflect sampling position rather than treatment, and paired pre- and post-treatment sampling at fixed georeferenced points, unamended control plots and mass balance over a defined soil depth are all needed before any reduction can be attributed to an intervention. Economically, phytoremediation is most competitive in remote regions of low land value, and microbiome-guided intervention should be justified as a reduction in years of land occupancy rather than as a removal percentage.

The review closes with a diagnosis that will discomfort parts of the field: the principal deficit is not technical capability but field-scale evidence reported to a common standard. No microbiome-guided intervention has been evaluated in multi-season, multi-site field trials under a common reporting protocol, inoculant application rates are seldom varied systematically, the trade-off between remediation performance and crop yield is rarely measured in the same experiment, and inoculant survival beyond a single season is almost never reported. Unsuccessful inoculations are unlikely to be published at the rate they occur, so the central tendency of the reported literature probably overstates field performance. For bioaugmentation with cultured isolates and bulk amendment, a practical case can be made now, the author argues. For everything else, the appropriate claim is promise rather than demonstrated performance.

Subject of Research: Evidence tiers of microbiome-guided rhizosphere engineering for remediating contaminated agricultural soils

Article Title: Microbiome-guided rhizosphere engineering for sustainable remediation of agricultural soil contaminants

Article References: Arunkumar, P. (2026). Microbiome-guided rhizosphere engineering for sustainable remediation of agricultural soil contaminants. Frontiers in Microbiology, 17, Article 1994433. https://doi.org/10.3389/fmicb.2026.1994433

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1994433

Keywords: rhizosphere engineering, soil remediation, microbiome, bioaugmentation, heavy metals, phytoextraction, hydrocarbons, microplastics, CRISPR, synthetic biology, biochar, biosafety