Plant material

Wild populations of both B. oleracea and B. rapa species were collected in France based on information in the National Inventory of Natural Heritage database (INPN, 2024) and Maggioni, Bothmer, Poulsen, and Aloisi (2020). In addition, B. rapa wild populations were gathered in Italy, Algeria, Slovenia and B. oleracea in Spain (Gomez-Campo et al., 2005) based on local flora and long field experiences. Siliques were collected from 30 plants per population (when available), depending on the size and accessibility of populations. Some wild populations of B. oleracea closely related species were identified and added to the analysis: eight B. montana Pourr. populations (six from France and two from Italy), as well as two B. rupestris Raf. (subsp. rupestris), two B. villosa Biv. [subsp. drepanensis (Caruel) Raimondo & Mazzola and subsp. tineoi (Lojac.) Raimondo & Mazzola], one B. macrocarpa Guss., and one B. incana Ten., all from Sicily, Italy. B. oleracea and B. rapa landraces were collected in five different countries either through direct collects on farms in Algeria, Tunisia and Italy or in BRC maintaining old landraces in France (BRC BrACySol) and Slovenia (Slovene Plant Gene Bank in Slovenia, SRGB KIS). In agreement with each country’s policy, the Nagoya Protocol will be applied, pending the introduction of the relevant collected material into the Multilateral System of the FAO’s International Treaty on Plant Genetic Resources for Food and Agriculture. Thus, during this transition period, the material will be available after seed production upon request, either in BRC BrACySol and SRGB for French and Slovenian populations, respectively, or by contacting the partner in each country, as reported in Supplemental Tables 1 and 2.

Each collected population was named following a specific code. It starts with (1) two letters representing the species (BO for B. oleracea, BR for B. rapa, BM for B. montana, BU for B. rupestris, BV for B. villosa, BA for B. macrocarpa, and BI for B. incana), followed (2) by a letter for the country of origin (F for France, I for Italy, S for Slovenia, E for Spain, A for Algeria, or T for Tunisia), (3) then four letters indicating the location of the collecting site, (4) either a W for a wild population or an L for a landrace, (5) and an additional letter (A, B, C, etc.) in case of several collecting sites at the same location (i.e. BR_I_CAST_W_A and BR_I_CAST_W_B). For all these populations, a common sheet was filled for wild populations to describe the environment (Supplemental Table 1) and another one for landrace collects at the farm or when seeds were acquired from Genetic Resource Centres (GRC BrACySol in France, KIS in Slovenia) (Supplemental Table 2).

Thirty plants per population were grown in the greenhouse for taxonomy assessments. For wild populations, we planted one seed of each of the 30 collected mother plants. When seeds were collected from fewer than 30 plants, we sowed several seeds per mother plant, equally represented, to reach a total of 30 seeds. For landraces, 30 seeds were sown.

As controls for the different experiments, we used a known representative of B. oleracea, B. rapa and B. napus species: doubled haploid lines of B. oleracea subsp. italica (HDEM) and B. rapa subsp. trilocularis (Z1) (Belser et al., 2018) and a pure line of B. napus subsp. oleifera, ‘Darmor’.

Cytogenetic control and chromosome counts

Flow cytometry was performed on all plants to assess the chromosome number of each plant using leaves as described by Leflon et al. (2006). Briefly, approximately 0.5cm² of fresh leaves were harvested and transferred to a Petri dish. This material was chopped using a sharp razor blade in 300µl of nuclei extraction staining buffer (from kit Cystain^TM UV Presice P-Sysmex) and incubated at room temperature for 30 to 90sec. 1.2ml of DAPI staining buffer was added per sample and the solution was then filtered through a 50µm nylon mesh. Estimation for each accession was obtained with FlowMax software using a CyFlow space cytometer (Sysmex Inc.). For the screening of B. oleracea and independently of B. rapa populations, the control variety, HDEM for B. oleracea and Z1 for B. rapa, was adjusted to a fluorescence intensity value of 300 for nuclei at G1 stage. Coincidence or deviation was compared with these controls.

For populations for which flow cytometer and chloroplast sequencing data were not congruent, the chromosome number was also determined from mitotic chromosomes observed on metaphasic cells isolated from root tips. Root tips of 0.5–1.5cm in length were treated in the dark with 0.04% 8-hydroxiquinoline for 2h at 4°C followed by 2h at room temperature to accumulate metaphases. They were then fixed in 3:1 ethanol:glacial acetic acid for 48h at 4°C and stored in 70% ethanol at -20°C until use. After being washed in distilled water for 10min, in HCl 0.25 N for 10min, then treated for 15min with a 0.01M citric acid-sodium citrate buffer (pH 4.5), root tips were incubated at 37°C for 30min in an enzymatic mixture (5% Onozuka R-10 cellulase (Sigma), 1% Y23 pectolyase (Sigma)). The enzymatic solution was removed and the digested root tips were then carefully washed with distilled water for 30min. One root tip was transferred to a slide and macerated with a drop of 3:1 fixation solution. Dried slides were then stained by a drop of 4’,6-diamidino-2-phenylindole (DAPI). Cells were viewed with an ORCA-Flash4 (Hamamatsu, Japan) on Axio Imager Z.2 (Zeiss, Oberkochen, Germany) and analyzed using Zen software (Carl Zeiss, Germany).

Fluorescence in situ hybridization (FISH)

The BoB014O06 BAC clone from B. oleracea BAC library (Howell, Kearsey, Jones, King, & Armstrong, 2008) was used as probe for ‘genomic in situ hybridization (GISH)-like’ to distinguish specifically all C-genome chromosomes in B. napus (Suay et al., 2014). The BoB014O06 clone was labelled by random priming with Alexa-594 dUTP (red) (Thermo Fisher Scientific). The ribosomal probe 45S rDNA used in this study was pTa71 (Gerlach & Bedbrook, 1979) which contained a 9-kb EcoRI fragment of rDNA repeat unit (18S-5.8S-26S genes and spacers) isolated from Triticum aestivum L. pTa71 was labelled by random priming with biotin-14-dUTP (Invitrogen, Life Technologies). Biotinylated probes were immunodetected by Fluorescein avidin DN (green) (Vector Laboratories). The chromosomes were mounted and counterstained in Vectashield (Vector Laboratories) containing 2.5µg/mL 4’,6-diamidino-2-phenylindole (DAPI) (grey). Fluorescence images were captured using an ORCA-Flash4 (Hamamatsu, Japon) on an Axio Imager Z.2 (Zeiss, Oberkochen, Germany) and analyzed using Zen software (Carl Zeiss, Germany).

Species identification by sequencing of a chloroplast region

The aim was to amplify a chloroplast genomic region containing diagnostic single nucleotide polymorphisms (SNPs) or indels for B. oleracea, B. rapa or B. napus. To that purpose, we first retrieved and aligned the Brassica chloroplast genome sequences available for the three species fromLi et al. (2017) using Geneious Prime 2022.2.2 (https://www.geneious.com ). We then identified a genomic region and designed consensus primers enabling us to discriminate each species. The consensus primers allowed amplification of 1,118pb for B. oleracea, 1,088pb for B. rapa or 1,084pb for B. napus. DNA of one to three plants per population and of control lines was extracted using 50mg of fresh leaf tissue, which had previously been freeze-dried, and the Nucleospin Plant II kit (Macherey Nagel). The consensus primers used were trnK-rps16_F (5’ CATAAACAGGTAGACTGCTAACTGG 3’) and trnK-rps16_R (5’ GTATTCTTCCTAAAGGTATGAAAACTAAC 3’) with following PCR reagents: 1X buffer, MgCl2 2mM, dNTPs 0.25mM, Primers 0.5µM each, Taq Promega 1.5U and 5ng DNA of the sample analyzed. The PCR conditions were a denaturation 94°C 2min, then 35 cycles 94°C 30sec - 59°C 30sec - 72°C 1min 30sec, with a final elongation 72°C 10min. The amplified region was then sequenced by Sanger (Genoscreen) and analyzed using Geneious software (https://www.geneious.com ). All amplified chloroplast sequence data have been deposited into NCBI/GenBank as PopSet 2716368500.: PP619885 - PP620127).

Taxonomic verification of the collected populations

Among the collected populations (Table 1), the first limiting factor encountered was the germination of the collected seeds, even under favourable controlled conditions applied on automated germination tools for B. rapa, in spite of seed viability confirmed by tetrazolium staining. Specifically, 6.8% of the collected populations showed a poor emergence in the greenhouse with less than 30 plants per population and were not considered for further analyses. This low germination rate may be attributed to two different factors: the high level of seed dormancy (observed here in 18.2% of B. rapa wild populations) and the seed conservation of landraces collected on farms (5.2% and 14.9% of seeds showed very poor germination for B. rapa and B. oleracea landraces, respectively).

To validate the correct species identification of each collected population and to verify the absence of contamination in the collected seeds, we performed flow cytometry on all the plants grown representing a population. As the investigated species have different profiles linked to their differences in DNA content (630Mb for 18 chromosomes in B. oleracea, 529Mb for 20 chromosomes in B. rapa) (Figure 1A), it was possible to determine with +/-2 chromosomes the genomic structure of each plant.

Due to possible contamination with species having a close chromosome number, this analysis was complemented by sequencing a chloroplast genomic region that showed species-specific differences. We chose a genomic region with a sequence specific to each species according to (Li et al., 2017). The size of the amplified regions was 1,118bp, 1,088bp and 1,084bp for B. oleracea, B. rapa and B. napus, respectively (all these chloroplast sequences are available on NCBI/GenBank as PopSet 2716368500.: PP619885 - PP620127). After aligning the sequences, we compared the sequences obtained in the sampled populations with those of the controls for the three species. We observed four SNPs and six Indels specific to B. oleracea, four SNPs and four Indels specific to B. rapa and three SNPs and five Indels specific to B. napus (examples provided in Figure 2A). B. montana (2n=18) differed from B. oleracea at only three SNPs and one Indel whereas B. villosa and B. macrocarpa differed from B. oleracea at 17 SNPs and six Indels. B. rupestris showed exactly the same sequence as the two latter species except for one SNP, indicating that these three species (B. villosa, B. macrocarpa and B. rupestris) are highly related to each other whereas B. montana seems closer to B. oleracea (Figure 2B).

When flow cytometer and sequencing data were not congruent, chromosome counting was performed during mitosis to identify the species. This observation was combined with GISH-like allowing identification of the C chromosomes and of rDNA locus number, specific to each species with four, ten and 12 rDNA loci for B. oleracea, B. rapa and B. napus, respectively (Figure 3).

By FISH, in B. rapa (A genome) (Figure 3B), the 45S rDNA probe (green) marks five different chromosomes. The strongest FISH signal located on the A03 chromosomes reflects a large number of genes. The second gene-rich locus is located on A01 chromosome proximal to the centromere. The remaining sites are located on cytogenetically undistinguishable A05, A06 and A09 chromosomes. B. oleracea (C genome)(Figure 3D) had two pairs of chromosomes (C07, C08) containing 45S rDNA loci. The sites localized on chromosome C08 show extensive decondensation while loci on C07 are fully condensed. In natural B. napus, we observed twelve 45S rDNA signals and the BOB014O06 staining revealed that eight signals were located on A genome and four on the C genome (Książczyk et al., 2011) (Figure 3C).

All the misidentified populations were listed in Supplemental Table 3. The most frequent mistake was confusing B. oleracea or B. rapa with B. napus. Among the 103 B. oleracea populations analyzed, only three were misidentified (one wild and two landraces) and were thereafter confirmed to belong to B. napus using flow cytometry (Figure 1D). This misidentification was also validated by chloroplast sequencing (Figure 2A) and chromosome counting. Among the 146 analyzed B. rapa populations, 15 were misidentified, out of which 12 were identified as B. napus. Nine of these 12 populations were sampled in the wild and are probably volunteers of B. napus, i.e. escaped from the fields. All these data were confirmed by the sequencing of a chloroplast genomic region (Figure 2A) revealing that all carried B. napus chloroplasts except for one wild Tunisian population (BR_T_ARIA_W_A), which had a B. rapa type chloroplast. The B. napus origin of this population was confirmed by cytogenetic analyses, revealing the presence of nine C chromosomes and 12 45S rDNA signals by FISH, eight on A genome and four on C genome (Figure 3E). Among the three remaining misidentified B. rapa populations, one wild population from Tunisia had a cytometry value close to B. rapa but no chloroplast gene amplification was detected; further morphological observations of this population revealed that it probably belongs to the genus Sinapis. The two last cases observed were B. rapa populations (one Slovenian wild population BR_S_LJUB_W_D and one French landrace BR_F_COND_L_A) having a flow cytometry value close to the one of B. napus (Figure 1E and Figure 1F) but a B. rapa chloroplast genomic sequence. Using cytogenetics, we detected no C chromosomes after a GISH-like experiment and 20 45S rDNA were counted, i.e. five rDNA loci per A genome (Figure 3F), which led us to the conclusion that these populations were in fact B. rapa autotetraploids (AAAA, 2n=4x=40).

Most of the populations confirmed as belonging to a specific species had an identical chloroplast sequence. Nevertheless, we observed a few SNPs specific to some populations. In B. oleracea, two SNPs were specific to only seven populations (BO_F_JOUY_L_A, BO_S_LJUB_L_G, BO_S_LJUB_L_H, BO_S_LJUB_L_L, BO_S_LJUB_L_M, BO_S_LJUB_L_N and BO_S_LJUB_L_O) and one allele at a different SNP was specific to BO_F_MERS_W_A. In B. rapa, three variations differentiated a few populations, one SNP in BR_A_DELL_W_A, one base deletion in BR_A_SEBA_W_A and BR_A_BOME_W_A and one SNP in BR_A_BLID_W_A, BR_A_BOUF_W_A, BR_A_CHLE_W_A, BR_A_BARA_W_A. These differences were observed in all the individuals tested per population.

For B. oleracea related species (B. montana, B. rupestris, B. villosa, B. macrocarpa and B. incana), all collected populations per species had the same flow cytometry value and the same chloroplast sequence.

Description of the populations

After discarding the few populations that did not germinate or were misidentified (Supplemental Table 3), we further characterized the remaining populations and their respective data collected during harvest.

Wild B. oleracea populations were collected on cliffs on the Atlantic coast in France and Spain (Figure 4), whereas its related species (B. montana, B. rupestris, B. villosa, B. macrocarpa and B. incana) were growing more in southern regions, on the Mediterranean coast. Their locations and the characteristics of each environment are described in Supplemental Table 1.

B. oleracea landraces were selected by farmers in each country, even in very warm regions such as the south of Algeria (Figure 4; Supplemental Table 2). Selection of different organs for crop production (flowers, leaves, stems or roots) has led to the divergence of highly diverse phenotypes. It is worth mentioning that some morphotypes were difficult to classify in one subspecies as some of them were domesticated at the same time for leaf production (such as subsp. acephala) and for head cabbage (such as subsp. capitata, e.g. BO_A_TAZL_L_A). Additionally, even within the same morphotype, different developmental traits can be observed such as in Mugnuli populations (south of Italy) with several floral heads compared to common broccoli (Laghetti et al., 2005).

Wild B. rapa populations (Figure 5; Supplemental Table 1) were found in locations where competition with other species is lower, such as vineyards, orchards or field margins. Thus, regardless of the country, the populations were generally large.

The majority of the collected B. rapa local landraces (Figure 5; Supplemental Table 2) were turnips (subsp. rapa) with the exception of few broccoletto (subsp. sylvestris var. esculenta) selected by Italian farmers.