Jin, Zhou, Deng, Hu, Zhou and Ma (Zhejiang University; Lena Qiying Ma senior author) publish in the Journal of Hazardous Materials (518:143917, 2026) a controlled comparison that every arsenic phytoextraction programme should read before choosing an amendment. They grew the arsenic hyperaccumulator Pteris vittata — the species this database tracks, and the subject of our earlier piece on its phosphate-uptake pathway — for two months in soil spiked to 100 mg kg⁻¹ arsenic, amended with 2.0 g kg⁻¹ calcium supplied three ways: as phosphate rock (PR), as calcium carbonate, or as soluble CaCl₂. Same fern, same soil, same calcium dose, same element.
The paradox the trial exposes
Read the availability numbers first. Relative to phosphate rock, CaCO₃ produced 10–15% higher rhizosphere available arsenic and phosphorus and slightly greater fern biomass. By the standard extraction-based logic that governs most amendment screening — mobilise the metal, then let the plant harvest it — limestone should have won the arsenic race. It lost, and not narrowly: PR increased frond arsenic by 35% to 3,128 mg kg⁻¹ and frond phosphorus by 28% to 3,158 mg kg⁻¹ compared with CaCO₃. On a 100 mg kg⁻¹ soil, that is a shoot-to-soil concentration ratio of roughly 31 (a derived figure: frond concentration divided by the nominal spiked soil concentration, both printed in the paper — a nominal enrichment factor against total soil arsenic, not a transfer coefficient against the measured available pool, which the abstract reports separately and which was higher under CaCO₃). Extractable arsenic predicted the wrong winner because it measured the wrong step of the pathway.
Why the rock wins: a primed transporter, not a bigger pool
The gene-expression data point at the mechanism. Under PR, P. vittata upregulated the high-affinity phosphate/arsenate transporter PvPht1;3 together with the arsenate reductase PvHAC2 and the arsenite antiporters PvACR3 and PvACR3;3 — by 2.0–2.3 fold (the abstract states the comparison baseline only for the CaCO₃ fold-changes below, not for these, so we quote it without one). The logic of that suite is the canonical P. vittata pathway: arsenate enters through the phosphate transporter, HAC2 reduces it to arsenite, and the ACR3 antiporters move arsenite onward — in P. vittata this family is best characterised at the tonoplast, sequestering arsenite in vacuoles. PR is a poorly soluble mineral that dissolves slowly and feeds phosphate at a sustained trickle; the interpretation that fits the transcription pattern is that the fern’s uptake machinery stays primed by steady, low-level phosphate supply instead of being swamped by a pulse. The CaCO₃ treatment ran a visibly different program — PvPht1;3, PvPTB1;2 and PvACR3;1/3;2 induced 1.7–3.3 fold relative to the soil control — a rewiring, not merely a weaker version of the same one. Same species, two different transcriptional routes to two different frond concentrations.
The other half of the story is microbial
PR retained a rhizosphere bacterial community of Shannon diversity 4.20, against 3.76 under CaCO₃, whose elevated pH compressed the community; the PR community carried taxa associated with phosphorus cycling and arsenic tolerance. The paper attributes part of PR’s advantage to this restructuring. That is the non-obvious implication: an amendment’s value to a hyperaccumulator is not only its geochemistry but its ecology. A liming agent that makes the soil solution richer in arsenic can still underperform if it flattens the rhizosphere community that handles phosphorus turnover and arsenic detoxification at the root surface. Screening amendments by soil-extract chemistry alone — the near-universal practice — misses half the mechanism.
The opportunity, honestly bounded
The practical opportunity is real and unglamorous. Phosphate rock is the least-processed phosphorus fertiliser, and here it delivered a 35% larger arsenic crop per harvest cycle in a direct head-to-head. For phytoextraction economics, which are dominated by the number of grow-and-harvest cycles a site needs, a one-third jump in per-cycle removal is the kind of lever that changes project viability. But the bounds must be stated as plainly as the result. This is one soil, one arsenic concentration (100 mg kg⁻¹), one species, one calcium dose, one two-month pot trial, and the full text is paywalled — everything above is verified against the published abstract, which is unusually quantitative but is still the abstract. The paper does not report what else the phosphate rock carried: phosphate rocks are natural ores that vary widely in cadmium, uranium and fluorine content, and an honest field budget must count the contaminants an amendment imports before crediting it with contaminants it removes. Nor is 3,128 mg kg⁻¹ near this species’ demonstrated ceiling — the frond record in this database is 22,630 µg g⁻¹ for P. vittata, so PR moved a mid-range performance upward; it did not unlock new physiology.
What we take from it
The field’s default mental model — mobilise, then harvest — was directly falsified here: the amendment that mobilised more arsenic into solution grew ferns containing less. For a hyperaccumulator whose entire value is running its uptake transporters at high duty cycle, amendment choice should be screened on transporter expression and rhizosphere function, with soil extractions demoted to a supporting role. And the CaCl₂ arm of the trial — soluble calcium with no phosphate or carbonate counter-ion — is the quiet control that says the calcium itself is not the driver; the anion chemistry and its microbial consequences are. More multi-amendment, mechanism-instrumented trials like this one, reported with full metadata, are what our methodology asks for.
Provenance: the article is closed access; all claims are bounded to the peer-reviewed abstract as independently re-retrieved from PubMed (PMID 42887114) and Europe PMC, which agree verbatim. Every load-bearing number above was verified verbatim by three independent reviewers on different vendors on a shared claim template — zero cross-vendor discrepancies on any digit. Derived figures (nominal enrichment factor ≈ 31; comparison of 3,128 to the 22,630 µg g⁻¹ database record) are labelled as derived. Paper-internal limits reported rather than smoothed over: one soil, one dose, one species, pot scale; the abstract states the fold-change baseline only for the CaCO₃ genes, not the PR ones; no impurity analysis of the phosphate rock reported in the abstract. Methods and limits at /methodology/.