How much nickel does a plant contain? For decades the honest answer has been “which plant, which soil?” — scattered across thousands of papers that each measured a handful of elements in a handful of species. Coker, Howe and colleagues at Texas A&M and twelve other institutions have now assembled those papers into a single resource: a global database of plant tissue concentrations spanning 52 elements, compiled from 5,474 samples across 73 countries, 21 climate classes and 26 soil groups, with mixed-effects-modelled global means and a systematic hyperaccumulator screen (Coker et al. 2026). For a field that argues about thresholds as much as about mechanisms, the baselines are as valuable as the headline findings.
The baseline, and a surprise in it
The mixed-effects global means put typical plant tissue at 3.59 mg kg⁻¹ Ni, 33 mg kg⁻¹ Zn, 9.89 mg kg⁻¹ Cu and 0.39 mg kg⁻¹ Cd (supporting information, Table 3). Two nonessential elements top the enrichment league: against published crustal benchmarks, Cd (4.3×) and Se (4.1×) are accumulated more strongly relative to their abundance in Earth’s crust than any other heavy metal or metalloid — the authors note that 20 elements occur in plant tissue at higher concentrations than the micronutrient Mo. Life runs on N, P, K and S, but a measurable share of a plant’s ash is elements it never evolved to use. That is a standing invitation to phytomining — and a warning that the food chain samples the same pool.
The skew is the signal
The raw database tells the same story more bluntly. Across all 1,410 nickel records the mean is 167 mg kg⁻¹ but the median is 1.7; for cadmium the mean is 5.28 against a median of 0.056 mg kg⁻¹ (SI Table 1) — roughly hundred-fold mean-over-median skews, whereas zinc sits at a modest 4.7× (126 vs 26.6). That skew is hyperaccumulation appearing in one summary statistic: a small tail of records, orders of magnitude above background, drags the arithmetic mean away from the typical plant. It is also why the authors needed mixed-effects models rather than simple averages, and it sets the honest scale of phytoextraction ambition — a shoot cadmium threshold of 100 mg kg⁻¹ is about three orders of magnitude above what the median plant record contains.
A systematic screen, and its two blind spots
Using the thresholds of van der Ent et al. (2013) (Ni, Pb and As 1,000; Cu, Co and Cr 300; Cd and Se 100; Zn 3,000; Mn 10,000 mg kg⁻¹ dry mass), the authors report 53 species crossing at least one bar, with high Ni, Cu and Pb uptake the most prevalent indicators. The thirteen-species Ni shoot list reads like a field guide to New Caledonian ultramafic flora: Geissois pruinosa (in our database), Hybanthus austrocaledonicus, Homalium kanaliense — and Sebertia acuminata, the famous nickel tree whose latex runs up to ~25% Ni by dry mass and whose accepted name is Pycnandra acuminata. The Zn shoot list even contains Dichapetalum gelonioides (printed “gelanioides” in the SI), the Shaba Zn accumulator (also in our database).
But the screen’s definition — concentration ≥ threshold in the stated organ, nothing more — has two blind spots a careful reader must hold onto:
- Overcalling. With no paired soil data and roots counting as qualifying organs, the lists absorb contaminated-site and spiked-system records. Lemna minor — duckweed — appears as both a Cd (≥100) and Zn (≥3000) “hyperaccumulator”; Capsicum frutescens as a Se hyperaccumulator; maize and sunflower sit on root lists. These are almost certainly exposure records, not constitutive hyperaccumulation, and none would survive the classic shoot-plus-soil criteria.
- Undercalling. The best-documented hyperaccumulators on Earth — Noccaea caerulescens (our Zn/Cd record), Pteris vittata (our As record), the Odontarrhena alliance, Sedum alfredii — are absent entirely. A 5,474-sample database is a thin skim of a century of trace-element literature, so Table 2 records what this particular compilation captured, not a census of known hyperaccumulators. (The headline “53 species” is the authors’ main-text count; the SI’s element-organ lists alone contain more than 60 distinct names across 92 records.)
What it means for agromining and phytoremediation
The non-obvious result is phylogenetic. Ornstein–Uhlenbeck models fit to major plant lineages indicate stabilising selection toward lineage-specific elemental optima for most elements, and an XGBoost classifier ranked the elements that best predict where a plant sits in the tree of life as S > Ca > N > Zn > Ba > Cu > Pb > Sb > Cd > P — cadmium and lead, which plants never needed, carry more taxonomic signal than phosphorus. Hyperaccumulation capacity is an inherited, lineage-constrained trait, not a soil response any species can dial up. For agromining that cuts both ways: prospecting should be phylogenetically targeted (the ultramafic-flora signal in the Ni list is exactly that pattern), and domestication is unlikely to be a shortcut — the study finds crops and wild plants already diverged in their Zn:Cd, K:Na, N:S and N:P stoichiometries, so high-biomass crop ideotypes and hyperaccumulator physiology sit on different evolutionary trajectories.
For this site the paper is also a methodological mirror. Our own database pairs concentrations with soil context and organ identity for exactly the reason this study’s screen illustrates: numbers without provenance mint duckweed hyperaccumulators while missing Noccaea. Robust baselines like these — the true background for every element in the /metals/nickel/ and /metals/cadmium/ hubs — are the reference frame that makes “unusual” definable at all.
What we could not read
The publisher full text (and the ETH Zürich repository copy) sit behind bot protection this host cannot pass, so this analysis rests on the complete published abstract and the full supporting information (Tables 1–3, figure captions), each verified verbatim by three independent retrievals. Claims bounded accordingly: the composition of the mean-median skew (hyperaccumulator versus contaminated-site records) is not decomposed in the material we could verify, and the precise main-text method behind the 53-species count is the authors’ as reported.
Source: Coker, Denvir, Mokhtari, Lewkowicz, Grove, Mikhailova, Lennox, López-Pozo, Jaumà-Palomeras, Whiteley, Rivero, Barbero Barcenilla & Howe (2026), ACS Environmental Au, DOI 10.1021/acsenvironau.6c00028; supporting information via 10.1021/acsenvironau.6c00028.s001. All load-bearing numbers verified against the abstract and SI by three independent reviewers. See /methodology/ for how we source and check analysis pieces.