Kuzmik, Courtin and colleagues (Merlot group, CNRS Toulouse) posted a bioRxiv preprint on 21 October 2026 — not yet peer-reviewed — that delivers the genomic infrastructure hyperaccumulation genetics has been waiting for: a corrected chromosome-grade assembly of the Cira accession of Noccaea caerulescens, plus high-quality assemblies of five further accessions spanning the species’ edaphic range, from serpentine to calamine to ordinary soil. The species is the closest thing the field has to an Arabidopsis of metal hyperaccumulation — it sits in our database as both a nickel hyperaccumulator (up to 12,000 µg g⁻¹) and a zinc/cadmium one — and this paper is the first to compare its genome structure across edaphic types at contig-level resolution.
Six assemblies, one reference-grade set
The headline numbers are assembly quality, and they are unusually good for a non-model plant. The improved Cira assembly (Ncci v3.1, hybrid PacBio HiFi + Oxford Nanopore) comprises 296 contigs totaling 270 Mb with a contig N50 of 31.1 Mb — individual contigs fully span three of the seven chromosomes (NC1, NC3, NC7) — with 98.72% k-mer completeness, a consensus quality of QV 57.5, and 99.0% of the 1,614 core conserved genes recovered; annotation yielded 30,089 genes, 89% functionally annotated. The five new accessions — Firmiensis (Ncfi), Bergenbach (Ncbe), Monte Prinzera (Ncmp), Col de Marchairuz (Ncmar) and Saint-Laurent-Le-Minier (Ncslm, collected at the Les Avinières calamine tailing pond) — assembled into 249–547 contigs (N50 17.2–29.6 Mb on Nanopore reads; 7.3 Mb for the PacBio-only Ncslm) with k-mer completeness of 92.29–97.26% and BUSCO scores of 98.8–99.2%. Notably, the new long-read assembly also corrects two large rearrangements on NC3 and NC5 that the earlier Hi-C-scaffolded version (v2.3, Belloeil et al. 2026) had introduced — a cautionary detail for anyone who has been doing marker or QTL work against the old reference.
The cadmium dial is copy number
The paper’s central result is that the variation that matters for phytoextraction is largely gene dosage at a small set of loci:
- HMA3, which pumps cadmium into the vacuole, is tandemly amplified to seven copies in the calamine accession Ncslm — matching the count previously reported for the strongly Cd-accumulating Ganges ecotype — and the distal HMA3 duplication is present in every accession except the non-metalliferous Ncmar. The authors argue this amplification is fixed in the calamine populations of the southern Massif Central, the very populations known to carry the species’ highest Cd-accumulation capacity.
- NRAMP1, implicated in Cd uptake and radial transport, reaches eight copies in Ncslm, built from both distant and tandem duplications; NRAMP4 is tandemly duplicated on NC5 in all six accessions but absent as a duplicate in N. praecox and the non-accumulator Microthlaspi erraticum.
- HMA4, the Zn/Cd root-to-shoot loading pump, is tandemly duplicated in every N. caerulescens accession and never in non-accumulating Brassicaceae — with the most duplications (four) in the calamine Ncslm.
For agromining and phytoextraction breeding, this is the useful part: accumulation intensity is partly a copy-number dial at a handful of named loci, which in principle can be selected for with a PCR assay long before anyone grows a plant on contaminated soil.
The two findings that should temper the story
The paper is at its most interesting where it reports what did not happen. First, ZIP-family transporters — long implicated in hyperaccumulation — show no significant copy-number difference between hyperaccumulators and non-accumulators, and metal-homeostasis genes as a class are not over-represented among all genes affected by copy-number variation. The genome did not remodel its metal machinery wholesale; a few loci got turned up. Second, the most evolutionarily suggestive duplication is MGR2, a transporter of the ACDP/CNNM family that exports magnesium from the vacuole: Noccaea genomes carry a distant MGR2 copy absent from all non-accumulating Brassicaceae, expanded to between two and eight tandem copies depending on accession. Because ultramafic soils impose extreme Mg/Ca imbalance, the authors propose this as evidence that adaptation to serpentine soils — not metal toxicity itself — is the ancestral trait in the genus, with metal hyperaccumulation a later exaptation of broadly metal-permeable transporters. That is a elegant hypothesis; the copy-number pattern across the six accessions does not cleanly track substrate (the tandem expansion occurs in the non-metalliferous Ncmar, and single copies occur in ultramafic Ncmp), so it remains a hypothesis.
What it means for the field
Practically, this preprint hands the community a breeding resource: six structurally reliable genomes across the species’ edaphic range, with copy-number states at the HMA3/HMA4/NRAMP1 loci that plausibly explain much of the ecotypic variation in Cd and Zn accumulation that field trials have documented for decades. For conservation the implication cuts the other way: the high-cadmium calamine ecotypes of the southern Massif Central are not just curiosities — their populations carry distinct, functionally consequential genotypes, which makes metallophyte habitat destruction a loss of genetic technology, not just botanical diversity. A genus-level check is also due: the authors note similar duplications in N. praecox, suggesting the copy-number toolkit predates the species.
The honest limits
This is a preprint, and its claims are correlational: copy number is mapped, not functionally tested, and dosage is not expression is not field performance. Several claims rest on a single calamine accession (Ncslm) — the statement that high HMA3 copy number is “fixed in calamine populations” generalizes beyond what six genomes can show. The paper is also internally untidy in ways worth recording rather than smoothing: the Results text reports assembly sizes of 256–278 Mb where Table 1 gives 279–285 Mb; the Methods breakdown of 395 transposable element families sums to roughly 358; MGR2 is described as two-to-eight copies in the Results but “nine tandem copies” in Ncfi in the Discussion; the NC3/NC5 v2.3 errors are called “translocations” in the Results and “inversions” in the Discussion; and Cira’s own soil type is never stated, so the reference accession’s edaphic context is unverifiable from the paper. None of these overturn the main result, but they mark the difference between a genome resource worth building on and conclusions worth waiting for peer review to confirm.
Provenance: all figures verified verbatim against the full CC-BY-NC preprint text (57 pp., Kuzmik, Courtin et al., bioRxiv, posted 21 October 2026, DOI) via independent three-expert cross-check; the paper-internal inconsistencies listed above are reported as found. See /methodology/ for how we source and verify.