Cvetanoska, Stefova, Bačeva Andonovska and Petreska Stanoeva report in Chemistry & Biodiversity (2026, vol. 23, e71761) the first combined polyphenolic and volatile-chemistry profiles of plants endemic to the Allchar Mining Area on Kožuf Mountain, North Macedonia — the ancient As–Sb–Tl workings that host the world’s only known lorandite (thallium arsenic sulfide) deposit. On 21 June 2024, at the Crven Dol locality in full flower, the team collected aerial parts of six local endemics — Viola arsenica, Viola allchariensis, Viola tricolor subsp. macedonica, Thymus thracicus var. alsarensis, Centaurea leucomalla and Knautia caroli-rechingeri — and profiled them by HPLC/DAD/MSⁿ and headspace GC–MS. It is the first volatile profile for all six taxa and the first polyphenol profile for five of them; only the thyme had been profiled before (Petreska Stanoeva et al. 2017, cited by the authors).
Three metal strategies, three chemistries
The backdrop makes this more than chemotaxonomy. Prior work from the same Macedonian research lineage — much of it cited inside this paper — has established that these co-occurring neighbors handle the site’s metals in three fundamentally different ways. The three Viola are genuine hyperaccumulators: published aerial-tissue values cited by the authors reach up to 58,900 mg kg⁻¹ Tl and 381 mg kg⁻¹ As (Jakovljević et al. 2025, Annals of Botany; Jakovljević et al. 2023, Metallomics, the latter confirming thallium hyperaccumulation by synchrotron μXRF imaging). Our database holds the same trio — V. arsenica at a Tl maximum of 58,900 µg g⁻¹ (our /species/viola-arsenica/ record), V. allchariensis at 23,100 (/species/viola-allchariensis/) and V. tricolor subsp. macedonica at 10,800 µg g⁻¹ (/species/viola-tricolor-macedonica/) — and for the other European Tl-hyperaccumulation system, in southern France, see our /species/iberis-intermedia/ record. Thymus thracicus var. alsarensis splits the difference: it translocates Tl to its above-ground parts but retains As in its roots (Bačeva Andonovska et al. 2015, Plant Biosystems, cited by the authors). And Knautia caroli-rechingeri with Centaurea leucomalla are excluders, holding metals in the roots and keeping shoots comparatively clean (Bačeva Andonovska et al. 2018, cited by the authors).
The new chemistry sorts almost perfectly along that axis. The excluders run enormous hydroxycinnamate economies: total identified polyphenols in K. caroli-rechingeri hit 453.23 mg g⁻¹ dry herb — roughly 45% of leaf mass — of which 405.40 mg g⁻¹ is a single compound, 1,3-dicaffeoylquinic acid, with 5-caffeoylquinic acid next at 28.93 mg g⁻¹. Centaurea leucomalla runs the same programme at lower titre (38.12 mg g⁻¹ total, with 3,4- and 1,5-dicaffeoylquinic acids at 4.81 and 4.50 mg g⁻¹). The thyme’s arsenal is rosmarinic acid — over 50% of its phenolic acids at 5.76 mg g⁻¹ — plus salvianolic acids (two compounds together, almost 40%), and its headspace is dominated by monoterpenes (51.3% of 77 detected volatiles, including D-limonene at 29.7% and the phenolic terpene carvacrol at 16.2%). The three hyperaccumulating violets, by contrast, are comparatively austere: 18–47 mg g⁻¹ total polyphenols dominated by flavone and flavonol glycosides (apigenin-6-C-glucoside-8-C-rhamnoside up to 25.62 mg g⁻¹), with no coumaric or ferulic acid derivatives detected at all, and aldehyde-led volatile profiles rich in furanones, lactones and nitrogen heterocycles (10.2–26.6%).
The negative result is the finding
Here is the non-obvious part, and the authors are commendably blunt about it: none of this chemistry demonstrably tracks the metal. Grouping the taxa by hyperaccumulator-versus-excluder behaviour gave a non-significant PERMANOVA (F = 3.67 for polyphenols, F = 2.06 for volatiles, exact p = 0.100 both, on only six taxa). Exploratory Spearman correlations against published aerial Tl concentrations were uniformly null after Benjamini–Hochberg correction — total polyphenol content versus Tl: ρ = −0.20, p = 0.78 — and the one nominally significant hit (volatile alcohols, ρ = −1.00, p = 0.017) is just the minimum attainable p-value for a perfect rank correlation with n = 5, evaporating after correction (adjusted p = 0.53). What is strongly structured is species identity itself: the polyphenol and volatile datasets agree with each other far better than either agrees with any metal variable (Mantel ρ = 0.72, exact p = 0.0083 between platforms).
For phytoremediation and agromining practice the lesson cuts both ways. The good news: a chemotype — measurable in a single HPLC or headspace run — is a stable, species-specific fingerprint that survives analytical platform choice, so metabolic profiling could serve as a cheap first-pass screen for metal-handling strategy when surveying unfamiliar metallophyte floras, in the same spirit as the XRF triage we have covered before. The caution: the classic stress-folk expectation — that hyperaccumulators must run outsized antioxidant arsenals to survive their own tissue loads — is not visible here, and the biggest phenolic fortress on the hillside belongs to an excluder that keeps metals out of its leaves entirely. A chemotype tells you what the species is; it does not tell you how much metal it is eating.
A conservation dossier written in metabolites
Allchar’s endemics are not hypothetical conservation objects. The flora catalogued by Micevski and Matevski (cited by the authors) — including Odontarrhena kavadarcensis, Centaurea kavadariensis, the two other taxa profiled here and the Viola — is restricted to this one metalliferous hillside, and the site sits on a deposit unique in the world, exploited for centuries and periodically eyed for more. That gives the paper a quiet second value: it is a chemical baseline, taken with voucher specimens deposited at the Macedonian National Herbarium (MKNH 242161–242166), for six plants that exist essentially nowhere else. Baselines of this kind cannot be re-collected after the fact — if a population vanishes under a renewed excavation or even under uncontrolled sampling, so does the reference chemistry for everything from chemotaxonomy to potential antioxidant or essential-oil value. The same coordinates-plus-chemistry format is, as with last week’s Anatolian survey, close to IUCN-ready evidence.
The sober ledger
Bound this tightly, because the paper itself does. One composite aerial sample per taxon, pooled from several individuals at a single locality on a single date; the triplicates were analytical replicates of that one pool, not biological replication. No metal was measured in these samples at all — the Tl correlations lean on published values from other studies, collection periods and sometimes other plant organs (Tl available for five of six taxa, As for three, Sb for one). The mg g⁻¹ figures are equivalent-based quantification (caffeic, luteolin, isorhamnetin and catalpol equivalents), so they are comparable within the paper but not absolute concentrations, and the volatile percentages are uncorrected peak areas. There is also a small internal tension worth flagging rather than smoothing: the text says 1,3-dicaffeoylquinic acid occurs “at lower levels” in the other studied taxa, while the paper’s own Table 1 lists it for Knautia only — we quote the table-consistent reading. And the authors state plainly that these data “do not establish direct mechanisms of metal tolerance”; the antioxidant-fortress reading above is a hypothesis their own design cannot yet test.
Provenance: open-access full text (PMC13616439, Chemistry & Biodiversity 23(9):e71761, CC-BY) retrieved and read in full; all load-bearing numbers verified against the primary source by three independent reviewers (Claude, GPT, Kimi) with zero claim-level discrepancies; paper-internal limitations (single composite sample per taxon, equivalents-based quantification, no concurrent metal measurements, Table-1-versus-text tension on 1,3-diCQA distribution) reported above rather than smoothed over. Methods and limits at /methodology/.