The elemental-defence hypothesis says plants hyperaccumulate metals such as selenium to poison herbivores. Dotson, Kurowski and Grillo ask the question that matters for anyone fielding these plants as cleanup crops: what does hyperaccumulation do to everyone else living in and around the root? Their answer, in Plant Biology 28(6):1960–1972 (CC BY; Loyola University Chicago), is a carefully bounded surprise — and a caution for mixed-cropping phytoremediation designs.
One soil, two congeners, four doses
The experiment is clean in its simplicity. Field soil collected beside Astragalus crotalariae plants near Anza-Borrego Desert State Park was genuinely selenium deficient (<2.0 mg kg⁻¹ DW by ICP-AES), so its microbial community had no history of Se exposure. Two co-occurring Astragalus were grown in it: A. crotalariae, one of 25 Astragalus species documented as Se hyperaccumulators (Alford et al. 2012), and A. lentiginosus var. borreganus, a non-accumulator. Sodium selenate was watered in at 0, 20, 80 and 100 µM twice weekly for ten weeks, and bulk soil, rhizosphere and root endosphere were profiled by 16S rRNA amplicon sequencing (59 samples; 1,800,754 reads; 4,523 ASVs after filtering).
The leaves tell one story, tolerance another
The first result overturns a naive expectation. The hyperaccumulator did not out-accumulate its congener: at the high dose both species’ leaves held nearly identical Se — 490 mg kg⁻¹ in A. lentiginosus, 480 mg kg⁻¹ in A. crotalariae — squarely in the secondary accumulator range of 100–1000 mg kg⁻¹, not the >1000 mg kg⁻¹ hyperaccumulator band. Either species accumulated essentially nothing without added selenate (<13 and <3.6 mg kg⁻¹, the ICP reporting limits). What separated the species was tolerance, not uptake: 100 µM selenate was lethal to the non-accumulator, which had to be dropped to 80 µM, while the hyperaccumulator was visibly unaffected.
The paper’s explanation is speciation chemistry — but note carefully: it was not measured here. Drawing on earlier X-ray absorption and biochemical work (Freeman et al. 2006), the authors argue that hyperaccumulating Astragalus store Se as methylated compounds — methyl-selenocysteine and its γ-glutamyl derivative — that cannot be misincorporated into proteins, whereas non-hyperaccumulators assimilate Se as selenomethionine and selenocysteine, which poison proteins nonspecifically. In our database, Astragalus bisulcatus — the species Freeman et al. actually measured — is an obligate hyperaccumulator reaching 13,685 µg g⁻¹ Se in wild leaves, and the wild record for A. crotalariae likewise exceeds the 1,000 µg g⁻¹ threshold on seleniferous soils. Under this greenhouse’s ten-week dosing, neither focal plant got near those numbers, and that dose gap matters when weighing how far the microbiome results travel.
The asymmetry that is the point
The load-bearing result is how little the hyperaccumulator’s microbiome moved. By the paper’s prose, selenium treatment plus its interaction with compartment explained 0.02% of community variance in A. crotalariae versus 6.05% in A. lentiginosus — against 85.98% and 80.76% respectively attributable to compartment alone. A correction a careful reader must know: the paper’s own Table 1 prints different PVCA values (Se main effect 0.03% vs 4.20%; Se × compartment interaction 0.00% vs 33.37%; compartment 96.51% vs 49.71%). Prose and table disagree numerically — an internal inconsistency the authors should reconcile — but they agree completely on the direction and the ordering: in the hyperaccumulator selenium is a rounding error, while in the non-accumulator it is one of the largest structured signals after compartment.
The whole-community statistics tell the same story. A whole-community ANOVA found the Se × species interaction significant for bacterial diversity (F = 7.183, P = 0.0103) and evenness (F = 7.237, P = 0.0101): the two species’ microbiomes respond differently to the same dose. PerMANOVA within compartments found only one community significantly shifted by selenium across all species × compartment combinations — the root endosphere of the non-accumulator (F(2,8) = 4.206, P-adj = 0.0341), though with two honest caveats: no pairwise dose comparison survived multiple-test correction, and the effect rides partly on unequal variances (PermDISP F(2,8) = 8.2494, P-adj = 0.0313). The authors also show the control-only (0 µM) communities of the two species do not differ significantly, so the divergence under dosing is treatment-driven rather than an intrinsic species fingerprint.
The rhizobium is the casualty
The most consequential single number is not a diversity index. ASV1, identified as Allomesorhizobium — a recently validated rhizobial genus (Li et al. 2024) that induces nodules in Astragalus — accounted for 12.5% of all sequencing reads and 67.3% of root endosphere reads, yet was scarce in bulk soil and rhizosphere. In the non-accumulator’s roots this dominant symbiont was significantly more abundant without selenium than with it; in the hyperaccumulator’s roots it was untouched (there, a different rhizobium, Rhizobium ASV6, declined under Se while ASV3 rose — strain-level tolerance splitting even within a genus). Across all comparisons, 84 ASVs shifted significantly with treatment (37 in bulk soil, 32 of them less abundant with Se; 29 in rhizosphere; 18 in roots), and the authors stress that even within a genus, ASVs can respond oppositely — so genus-level “Se-tolerant microbe” lists from the literature should be read at strain resolution or not at all.
Their synthesis is the paper’s deepest point: the rhizobium in the root endosphere appears dependent on its host’s detoxification chemistry for protection — “if the host non-accumulator is unable to protect its symbiont from lethal doses of selenium, the symbiont cannot effectively produce the resources necessary for the host plant to compete.” That is a hedged, mechanism-inferred claim (the evidence is the absence of harm in the hyperaccumulator, not a measured protective pathway), and the paper itself never tests inoculation or nodulation — nodules were sampled only to identify their rhizobia.
What it means for phytoremediation — opportunity and risk
Opportunity: a Se hyperaccumulator appears to be a microbiome-safe cleanup tool. Hyperaccumulation chemistry that methylates Se inside plant tissue does not just protect the plant — on this evidence it shields the entire root-associated community, including the mutualist that matters most for a legume’s growth. Monoculture phytoextraction of seleniferous soils with Astragalus is unlikely to sterilise its own rhizosphere in the way a bare chemical extraction would, and the Se-responsive taxa the study catalogues are a candidate list of partners for deliberately designed rhizospheres — though designing such systems is extrapolation beyond what this paper tests.
Risk: the damage case falls on everything that is not a hyperaccumulator. On a real contaminated site — where agronomy often mixes hyperaccumulators with non-accumulating crops, cover species or legume rotations for biomass and nitrogen — selenate can restructure root communities and knock back N-fixing rhizobia precisely in the plants that have no internal detox chemistry. The same elemental defence that makes a hyperaccumulator a good neighbour makes a non-accumulating crop a susceptible one, and the paper’s closing argument is exactly that ecological partners must be considered when these plants are deployed for phytoremediation.
Bounds, stated plainly: two closely related species (species effects cannot be fully excluded); one low-Se soil whose microbes were not pre-adapted, so pre-adapted seleniferous communities may respond differently; a ten-week greenhouse dosing with selenate only; no Se speciation, no nodulation counts, no biomass and no N-fixation measured; root-tissue Se itself unmeasured (only leaves); and a prose-versus-table PVCA discrepancy flagged above. The qualitative asymmetry, however, is consistent across every analysis the paper runs — and it reframes selenium hyperaccumulation from a herbivore story into a community ecology one.
Source: Dotson, Kurowski & Grillo (2026), Plant Biology 28(6):1960–1972, DOI 10.1111/plb.70263 (CC BY; received 2025-11-06, accepted 2026-04-08). Full text read from the open-access PMC copy (PMC13586643). Every load-bearing number was verified verbatim by three independent reviewers against the retrieved full text (two separate extractions of the JATS XML plus an independent PubMed re-retrieval); the PVCA prose-versus-Table-1 discrepancy and the reporting-limit nature of the low-dose leaf values are flagged in the body. Speciation chemistry is the paper’s literature-based interpretation, not a measurement. See /methodology/ for how we source and check analysis pieces.