The pitch for microbially assisted phytoextraction is usually sold as a general boost: add a plant-growth-promoting organism, get more metal out of the soil. Shahriari, Mosallanejad, Safarzadeh, Zarei and Tahmasebi (Scientific Reports, published 6 October 2026) ran the comparison that complicates it. They grew Camelina sativa for eight weeks in an artificially established soil contaminated simultaneously with Cd, Pb, Zn and Ni at a target of 100 mg kg⁻¹ each, and compared three treatments: an uninoculated control, Bacillus subtilis, and Trichoderma harzianum.
The fungus and the bacterium won different metals
Against the uninoculated control, T. harzianum decreased post-harvest DTPA-extractable Cd, Pb, Ni and Zn by approximately 31%, 32%, 53% and 32% respectively — the Ni response is the standout, a halving of the plant-available pool in eight weeks. But the ranking flipped by metal: the fungal treatment “generally produced greater improvements in Ni and Zn uptake and accumulation indices than Bacillus subtilis, whereas the bacterial treatment resulted in the highest Pb uptake” (the paper’s own words). There is no best inoculant; there is a best inoculant for a target metal, and the choice has to be made metal by metal. For the multi-metal sites that dominate real brownfield contaminations — and for anyone sequencing cleanup on the nickel hub’s priority element versus lead — that metal-specificity is the operational headline.
Read the mass balance before celebrating the 53%
Here is the non-obvious part, and the place where these experiments most often get over-read. The same T. harzianum sentence says the fungus cut DTPA-extractable soil metal while increasing shoot metal uptake and maintaining plant growth. Those two observations are consistent, not redundant: DTPA extraction measures a pool of plant-available metal at one sampling moment, not total soil metal and not cumulative removal. A drop in the extractable pool plus a rise in shoot content is exactly what genuine phytoextraction should look like — but it can also be produced by redistribution into microbial or root biomass, by pH shifts that re-partition metal into non-DTPA fractions, or by sampling timing. Our own earlier analysis of a Sedum alfredii SynCom study found plant uptake accounted for a tiny fraction of the apparent soil-Cd disappearance — a rhizosphere sampling artefact wearing a cleanup result’s clothes. The abstract of this paper reports no absolute tissue concentrations and no total-soil mass balance, so the honest claim is “accelerated transfer into the shoot plus a smaller available pool,” not “half the Ni removed.”
Why this contrasts with the AMF null result
It is tempting to file this under “inoculation works.” Resist. Our coverage of a four-tree Ghana screening just two days ago found arbuscular mycorrhizal inoculation produced no significant effect on Hg accumulation in any species — and the phytoextraction literature is littered with both results. What this experiment adds is not “inoculate and win” but a design principle: Trichoderma and Bacillus operate through different soil–plant acquisition pathways, and under simultaneous multi-metal stress those pathways pay off on different metals. The earlier pieces in this series on hyperaccumulator rhizosphere microbiomes make the same point from the other direction — the community that unlocks one metal can be irrelevant to the next. Inoculant selection is chemistry, not gardening.
The honest limits
Everything above is bounded by the design: one artificially spiked soil at a brutal 100 mg kg⁻¹ per metal (real multi-metal sites rarely deliver four contaminants at once at that level), eight weeks, one crop, greenhouse conditions. The authors themselves scope the conclusion to “early-stage phytoremediation under the experimental conditions” — a hedge worth keeping. C. sativa is a useful accumulator crop for phytomanagement, not a hyperaccumulator, so this is assisted phytoextraction, not agromining; the shoots would still need disposal as contaminated biomass. And because the current version of record is a “shared early” release carrying only the abstract, every number here comes from that abstract, cross-checked verbatim by three independent reviewers against the publisher’s page — no results tables were available to audit.
Provenance: analysis based on the published abstract of Shahriari et al. (2026), verified verbatim against the publisher’s record; methods and limits at /methodology/.