The genus Alyssum is famous in this field for a reason that is partly a taxonomic accident: most of the nickel hyperaccumulators formerly placed in it — including Odontarrhena lesbiaca, an obligate hyperaccumulator reaching ~20,000 µg g⁻¹ Ni — have been moved to Odontarrhena, leaving Alyssum proper as a window onto what the other serpentine colonists do. Celestini, Duchoslav, Nezamivand-Chegini, Gerchen, Šrámková, Wijfjes, Krejčová, Kuzmanović, Španiel, Schneeberger, Yant and Kolář now give us that window in an open-access paper in Annals of Botany 138(3):851–867: a reciprocal-transplant experiment plus a de novo genome and four serpentine–non-serpentine population pairs in Alyssum gmelinii and A. spruneri. Their answer is sharp — serpentine-adapted Alyssum survives by avoiding the metals, and it reuses the same genetic playbook as Arabidopsis arenosa across roughly 20 million years of Brassicaceae divergence.
The transplant: a clock and a sieve
In a greenhouse reciprocal transplant of four Central European A. gmelinii populations (720 seeds, 256 seedlings; serpentine soil from Borovsko, Czechia), the serpentine treatment was brutal to the wrong genotype: 95% of plants of non-serpentine origin died in serpentine soil, against 35% of serpentine-origin plants. Surviving serpentine-lineage plants also raced the season — an interaction hazard ratio of 2.844 (95% CI 2–3.9, P < 0.001) for earlier germination, with serpentine soil itself accelerating germination (HR 1.398, 95% CI 1.1–1.8, P = 0.007), consistent with the idea that seedling establishment must precede summer drought on these shallow, dark, metal-rich soils. The soil chemistry matched textbook serpentine: a PCA split serpentine from non-serpentine sites along PC1 (44.8% of variance), driven by Co (20%), Ni (19%) and Mg (14%) on one side and Ca (16%) on the other.
The phenotype: forage calcium, exclude nickel — but tolerate the residue
The ionomics are the mechanistic core. Serpentine-origin plants doubled their tissue Ca:Mg uptake ratio relative to soil availability when grown on serpentine, maintaining near-control intracellular Ca despite a soil that starves them of it, while Ni and Co were held at tissue concentrations below their soil abundance — a pattern the authors read as active exclusion. The nuance that should keep phytoremediation readers honest: exclusion is relative, not absolute. Tissue Ni still averaged 82 mg kg⁻¹, with many individuals above 100 mg kg⁻¹ — Brooks-grade “strong accumulator” territory, not hyperaccumulator (>1,000 mg kg⁻¹) — and Cr averaged 41 mg kg⁻¹ despite its toxicity. The authors’ reading, and it is persuasive, is that these plants run a two-layer defence: restricted uptake plus intracellular tolerance of whatever gets through. That combination is exactly what a phytoextraction engineer would want to transplant into a crop — or what a conservationist should note is being selected for, independently, on every serpentine outcrop.
The genome: selection on transport, not on metallothioneins
The team assembled a 683 Mb reference genome (N50 12.3 Mb, 96.2% BUSCO completeness, 32,073 protein-coding genes) from a diploid A. gmelinii, then resequenced ten plants from each of four serpentine–non-serpentine population pairs (two per species, spanning diploid and autotetraploid cytotypes; pairs 4–26 km apart; 1,082,883 putatively neutral SNPs at ~32× depth). Combining F_ST outlier windows with population-branch-excess statistics left a conservative ~188 candidate genes per population pair, with significantly more gene reuse across pairs than chance (Fisher’s exact P < 0.002). Sixty-one “top candidates” recurred across at least two pairs, and a PicMin order-statistics scan added 122 consensus genes. The functions are unambiguous: transmembrane ion transport and homeostasis — the high-affinity sulphate transporter SULTR1;1, nitrate transporters NRT2;1 and NRT2;3, potassium transporters KUP9 and HAK5, the K⁺ channel GORK, Mg transporter MGT7, ferric-chelate reduction oxidase FRO4, phosphate transporters PHT1;1/PHT1;2 — plus Casparian-strip proteins CASP1, CASP2 and ESB1, which control where root uptake becomes selective, and seed-development loci (SHB1, EDA4, PUB4) matching the germination phenotype.
The convergence: 13 genes, ~17–20 Mya
The headline result is evolutionary. Arabidopsis arenosa — the only other species with a genome-wide serpentine-adaptation candidate list from individually resequenced plants (222 parallel candidates) — shares 13 candidate genes with the Alyssum sets (ten with the top-candidate list, eleven with the PicMin list, 13 unique: NRT2;1, GORK, KUP9, CASP1, MGR2, PLC7, ARPC4, DAYSLEEPER, RHF1A, PDS3, an A20/AN1 zinc-finger, a P-loop NTPase and an ARM-repeat protein; each overlap P < 0.0001). Convergence holds at the network level too: the between-species protein–protein interaction network of the two candidate sets (127 Alyssum + 218 A. arenosa proteins) has 253 nodes and 1,232 edges, denser than 5,000 randomised lists (P = 0.0046), with PHT1;1, GORK and NRT2:1 flagged as possible hubs. These tribes diverged ~17–20 million years ago, so this looks less like shared recent standing variation and more like repeatable recruitment of the same functional module — the same pattern documented for alpine adaptation in Brassicaceae but unlike the gene-level non-convergence found for Arctic adaptation and the only partial parallels reported for post-whole-genome-duplication adaptation.
What it means for phytoremediation, agromining and conservation
Three implications, one genuine risk. First, for phytoextraction of nickel: the repeated targets here (CASP1 as the endodermal gatekeeper; NRT2;1/GORK/KUP9 coupling nutrient and K⁺ fluxes; PHT1;1 as a network hub) are a short, testable list of engineering or rhizosphere targets for controlling uptake selectivity — arguably more valuable for making a crop exclude toxic Ni while taking up Ca than another copy of a known metal pump. Second, for agromining: the study is a reminder that hyperaccumulation is the exception on ultramafic soils. The non-hyperaccumulator strategy — exclude and tolerate, with tissue Ni still at 82 mg kg⁻¹ — explains why wild metallophyte biomass is a poor ore unless the plant is a true hyperaccumulator, and it maps the genetic boundary between the two strategies inside one tribe (Alysseae). Third, for conservation: every serpentine outcrop appears to be independently recruiting overlapping ion-transport gene sets from its own gene pool, which means metallophyte habitat loss is not interchangeable — each outcrop is a replicate experiment in the same genes under different genomic backgrounds, and the functional variants may be locally unique even when the gene names are not.
The risk is epistemic. All of this is statistical genomics: the candidates are genome-scan hits with no functional validation yet, the transplant is one species complex, one serpentine soil and greenhouse conditions, and the ionomics for the non-serpentine-lineage-in-serpentine combination are missing precisely because those plants died (95% mortality). Convergence is also demonstrated only within Brassicaceae, and the authors say as much. Treat the 13 genes as a prioritised, mechanistically coherent candidate list for serpentine adaptation — not as proven causal alleles.
Source: Celestini, Duchoslav, Nezamivand-Chegini, Gerchen, Šrámková, Wijfjes, Krejčová, Kuzmanović, Španiel, Schneeberger, Yant & Kolář (2026), Annals of Botany 138(3):851–867, DOI 10.1093/aob/mcaf141 (CC BY; epub 10 July 2025). Full text read via Europe PMC (PMC13487438); every load-bearing number was verified verbatim by three independent reviewers retrieving the source through separate paths. See /methodology/ for how we source and check analysis pieces.