Brassica oleracea — kale — already sits in our database as a facultative thallium accumulator with shoot Tl up to 6,280 µg g⁻¹ dry mass, and that record has just acquired a mechanism. Corzo-Remigio, Harris, van der Ent and colleagues (Universities of Queensland, Adelaide and QUT, with DESY Hamburg and Wageningen) report in Metallomics (18(1):mfag010, 2026; accepted 2026-02-12) what is, to their knowledge, the first simultaneous micro-XRF plus micro-X-ray-diffraction mapping of a living plant, done at the PETRA III P06 beamline on fresh leaves flash-frozen in the cryostream at −140 °C. The question was narrow: what exactly are the microscopic Tl-rich crystals that an earlier study had spotted in kale leaves? The answer reframes how a crop handles one of the most toxic metals in the periodic table.
The crystals are thallium(I) chloride
Hydroponically grown ‘Red Russian’ kale, exposed to a non-toxic 7.5 µM Tl dose, took the element up aggressively: bulk XRF gave 6,000–6,280 µg Tl g⁻¹ in shoots versus 1,720–1,840 µg g⁻¹ in roots (Table 1) — a shoot-to-root ratio of roughly 3.4–3.5, leaf concentrations around 63× the classical 100 µg g⁻¹ hyperaccumulation threshold, and the Tl almost entirely in the monovalent state, Tl(I). Elemental maps show Tl piled along the leaf margins and around the vascular bundles, a distribution quite different from the even Ca and K maps. In those hot spots the diffraction maps identify the crystal phase: cubic thallium chloride, TlCl(s), refined lattice parameter a = 3.8143(9) Å, in crystals about 5 µm across, each diffraction cluster behaving as a single crystal. Weak extra peaks at 13.4° and 20.9° 2θ might indicate a minor KTlCl₄ phase — i.e. some Tl(III) — but the authors flag that identification as uncertain, and prior XANES work on kale found under 5% of total Tl as Tl(III) anyway.
A halophyte trick in a mustard
The proposed mechanism is the paper’s conceptual contribution. Tl(I) has an ionic radius close to K⁺, and is thought to enter roots through K⁺ channels — which is also why raising K in solution can suppress Tl uptake by up to half. From there Tl is translocated to the shoot, co-compartmented with chloride in vacuoles, and, when the cytoplasmic load gets too high, expelled through hydathodes — the guttation pores at leaf margins and vein endings — where the evaporating fluid precipitates TlCl crystals on the leaf surface. That is precisely how halophytes shed excess NaCl and KCl through salt glands, and the authors draw the parallel explicitly, noting a phylogenetic link between salt tolerance and metal hyperaccumulation and citing metal-laden salt crystals in Armeria maritima. A food crop running a salt-excretion pathway with a potassium-mimicking poison is a genuinely non-obvious piece of plant physiology.
Why it cuts both ways for phytomining
Thallium is a USGS-tracked critical-adjacent commodity at roughly US$8,800 kg⁻¹ (2024), and the paper’s worked examples show why agromining keeps circling back to it: at an estimated 4 t ha⁻¹ of Biscutella laevigata biomass at 425 µg Tl g⁻¹ (50 plants m⁻²), a field would yield ~1.7 kg Tl ha⁻¹ yr⁻¹ — about US$15,000 ha⁻¹; a second example using Iberis linifolia at 800 µg g⁻¹ and 10 t ha⁻¹ reaches ~8 kg ha⁻¹ yr⁻¹, ~US$70,400 — in both cases excluding production and processing costs, which the authors concede “will no doubt be substantial.” Here is where the crystal finding bites in both directions:
- The opportunity. Crystallizing Tl as a chloride salt on the leaf surface could simplify downstream metallurgy — Tl is extractable from biomass with carboxylic acids, and a defined crystalline phase is far friendlier to hydrometallurgy than organically complexed metal. Fast-growing, seed-cheap Brassica crops are natural testbeds for that refining chain, on the model of the nickel bio-ore processes already developed.
- The risk. The same hydathode pathway puts the accumulated metal at the surface, exposed. Rain or dew before harvest could redissolve TlCl and wash it straight back to the soil the crop was meant to clean — a silent leakage term in any Tl phytoextraction mass balance that, as far as this paper goes, nobody has measured. And the agronomic levers the mechanism implies — K⁺ competition at uptake, Cl⁻ supply shaping crystal formation — can suppress recovery as easily as tune it.
The food-safety shadow
The other edge is sharper. Germany’s maximum permissible Tl level in food is 0.5 mg kg⁻¹; Canada’s agricultural soil guideline is 1 mg kg⁻¹; Health Canada identifies food as the main human exposure route. Kale in this experiment carried over 12,000× the German food limit in tissue people actually eat, and B. oleracea var. capitata (cabbage) is already flagged in an EU-FORA review for high Tl uptake near sulfide mineralization. The halophyte analogy at least offers a testable mitigation: if Tl exits via hydathodes, surface crystals may be partly removable by washing — but that is a hypothesis, not a result, and the paper does not test it. Regulators, note: the toxicity asymmetry between Tl(I) and Tl(III) (~50,000× in Chlorella assays, per the cited literature) means speciation-aware limits would matter if a cheaper food-screening method ever arrives.
What this study does not show
Keep the bounds in view, because they are real. This is hydroponics, one main cultivar (‘Red Russian’, with ‘Nero di Toscana’ observed secondarily), one non-toxic Tl dose, and no soil — rhizosphere chemistry, mycorrhizae and root-zone chloride supply could all change the picture in field soil. The guttation-expulsion sequence is an inference from spatial patterns and analogy, not a demonstration: no guttation fluid was collected, no mass balance closes the loop on how much Tl actually leaves the leaf. The KTlCl₄ signal is acknowledged to be weak. And the economics are desk calculations on other species — kale itself is a prospect, not a proven phytoextractor. None of that dulls the core result: a synchrotron-resolved mineral phase, TlCl, sitting in a living leaf, with a mechanism that unifies hyperaccumulation and salt excretion. For thallium coverage that is the strongest mechanism paper in years, and it links cleanly to the four Tl species already in the database — B. oleracea, B. laevigata, Iberis intermedia and Silene latifolia.
Source: Corzo-Remigio, Harris, Jones, Wang, Brueckner, Spiers, Garrevoet & van der Ent (2026), Metallomics 18(1):mfag010, DOI 10.1093/mtomcs/mfag010 (open access, CC-BY). This piece is based on the full open-access text, retrieved from PubMed Central (PMC13017095). All load-bearing numbers — shoot 6,000–6,280 and root 1,720–1,840 µg g⁻¹ Tl; ~5 µm cubic TlCl(s) with a = 3.8143(9) Å; the 7.5 µM hydroponic dose; the 100 µg g⁻¹ threshold; B. laevigata 32,700 and I. intermedia 4,000 µg g⁻¹ records; US$8,800 kg⁻¹; 1.7 and 8 kg ha⁻¹ yr⁻¹ worked examples; the 0.5 and 1 mg kg⁻¹ regulatory values; the <5% Tl(III) XANES figure; and the 50,000× Tl(III) toxicity multiplier — were verified verbatim by three independent reviewers retrieving the article separately; no numeric disagreements were found. Reviewer caveats reflected above: the Tl exposure window is ambiguously reported (24 days pre-culture, then Tl, harvest “after 30 days experimental period” — read literally as ~6 days of Tl exposure); the paper prints the Tl(I) salt formula as “Tl(NO₃)₂” (chemically anomalous for a monovalent ion, quoted as printed); Table 1 is benchtop bulk XRF (Z-Spec JP500), not the synchrotron maps; the economic example’s second species is written Iberis linifolia while its cited reference concerns I. intermedia; and KTlCl₄ is a tentative identification. See /methodology/ for how we source and check analysis pieces.